diff options
Diffstat (limited to 'plugins/Dbx_mdb/src/mdbx')
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/AUTHORS | 30 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/COPYRIGHT | 22 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/LICENSE | 47 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/README.md | 758 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/bits.h | 1226 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/defs.h | 423 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/lck-windows.c | 452 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/mdbx.c | 11613 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/mdbx.h | 1825 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/osal.c | 1002 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/osal.h | 645 | ||||
| -rw-r--r-- | plugins/Dbx_mdb/src/mdbx/version.c | 34 |
12 files changed, 0 insertions, 18077 deletions
diff --git a/plugins/Dbx_mdb/src/mdbx/AUTHORS b/plugins/Dbx_mdb/src/mdbx/AUTHORS deleted file mode 100644 index 3321baad04..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/AUTHORS +++ /dev/null @@ -1,30 +0,0 @@ -Contributors -============ - -Alexey Naumov <alexey.naumov@gmail.com> -Chris Mikkelson <cmikk@qwest.net> -Claude Brisson <claude.brisson@gmail.com> -David Barbour <dmbarbour@gmail.com> -David Wilson <dw@botanicus.net> -dreamsxin <dreamsxin@126.com> -Hallvard Furuseth <hallvard@openldap.org>, <h.b.furuseth@usit.uio.no> -Heiko Becker <heirecka@exherbo.org> -Howard Chu <hyc@openldap.org>, <hyc@symas.com> -Ignacio Casal Quinteiro <ignacio.casal@nice-software.com> -James Rouzier <rouzier@gmail.com> -Jean-Christophe DUBOIS <jcd@tribudubois.net> -John Hewson <john@jahewson.com> -Klaus Malorny <klaus.malorny@knipp.de> -Kurt Zeilenga <kurt.zeilenga@isode.com> -Leonid Yuriev <leo@yuriev.ru>, <lyuryev@ptsecurity.com> -Lorenz Bauer <lmb@cloudflare.com> -Luke Yeager <lyeager@nvidia.com> -Martin Hedenfalk <martin@bzero.se> -Ondrej Kuznik <ondrej.kuznik@acision.com> -Orivej Desh <orivej@gmx.fr> -Oskari Timperi <oskari.timperi@iki.fi> -Pavel Medvedev <pmedvedev@gmail.com> -Philipp Storz <philipp.storz@bareos.com> -Quanah Gibson-Mount <quanah@openldap.org> -Salvador Ortiz <sog@msg.com.mx> -Sebastien Launay <sebastien@slaunay.fr> diff --git a/plugins/Dbx_mdb/src/mdbx/COPYRIGHT b/plugins/Dbx_mdb/src/mdbx/COPYRIGHT deleted file mode 100644 index 4c5862156f..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/COPYRIGHT +++ /dev/null @@ -1,22 +0,0 @@ -Copyright 2015-2017 Leonid Yuriev <leo@yuriev.ru>. -Copyright 2011-2015 Howard Chu, Symas Corp. -Copyright 2015,2016 Peter-Service R&D LLC. -All rights reserved. - -Redistribution and use in source and binary forms, with or without -modification, are permitted only as authorized by the OpenLDAP -Public License. - -A copy of this license is available in the file LICENSE in the -top-level directory of the distribution or, alternatively, at -<http://www.OpenLDAP.org/license.html>. - -OpenLDAP is a registered trademark of the OpenLDAP Foundation. - -Individual files and/or contributed packages may be copyright by -other parties and/or subject to additional restrictions. - -This work also contains materials derived from public sources. - -Additional information about OpenLDAP can be obtained at -<http://www.openldap.org/>. diff --git a/plugins/Dbx_mdb/src/mdbx/LICENSE b/plugins/Dbx_mdb/src/mdbx/LICENSE deleted file mode 100644 index 05ad7571e4..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/LICENSE +++ /dev/null @@ -1,47 +0,0 @@ -The OpenLDAP Public License - Version 2.8, 17 August 2003 - -Redistribution and use of this software and associated documentation -("Software"), with or without modification, are permitted provided -that the following conditions are met: - -1. Redistributions in source form must retain copyright statements - and notices, - -2. Redistributions in binary form must reproduce applicable copyright - statements and notices, this list of conditions, and the following - disclaimer in the documentation and/or other materials provided - with the distribution, and - -3. Redistributions must contain a verbatim copy of this document. - -The OpenLDAP Foundation may revise this license from time to time. -Each revision is distinguished by a version number. You may use -this Software under terms of this license revision or under the -terms of any subsequent revision of the license. - -THIS SOFTWARE IS PROVIDED BY THE OPENLDAP FOUNDATION AND ITS -CONTRIBUTORS ``AS IS'' AND ANY EXPRESSED OR IMPLIED WARRANTIES, -INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY -AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT -SHALL THE OPENLDAP FOUNDATION, ITS CONTRIBUTORS, OR THE AUTHOR(S) -OR OWNER(S) OF THE SOFTWARE BE LIABLE FOR ANY DIRECT, INDIRECT, -INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, -BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; -LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER -CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT -LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN -ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE -POSSIBILITY OF SUCH DAMAGE. - -The names of the authors and copyright holders must not be used in -advertising or otherwise to promote the sale, use or other dealing -in this Software without specific, written prior permission. Title -to copyright in this Software shall at all times remain with copyright -holders. - -OpenLDAP is a registered trademark of the OpenLDAP Foundation. - -Copyright 1999-2003 The OpenLDAP Foundation, Redwood City, -California, USA. All Rights Reserved. Permission to copy and -distribute verbatim copies of this document is granted. diff --git a/plugins/Dbx_mdb/src/mdbx/README.md b/plugins/Dbx_mdb/src/mdbx/README.md deleted file mode 100644 index c962da2dc3..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/README.md +++ /dev/null @@ -1,758 +0,0 @@ -libmdbx -====================================== -**The revised and extended descendant of [Symas LMDB](https://symas.com/lmdb/).** - -*The Future will Positive. Всё будет хорошо.* -[](https://travis-ci.org/leo-yuriev/libmdbx) -[](https://ci.appveyor.com/project/leo-yuriev/libmdbx/branch/master) -[](https://scan.coverity.com/projects/reopen-libmdbx) - -English version [by Google](https://translate.googleusercontent.com/translate_c?act=url&ie=UTF8&sl=ru&tl=en&u=https://github.com/leo-yuriev/libmdbx/tree/master) -and [by Yandex](https://translate.yandex.ru/translate?url=https%3A%2F%2Fgithub.com%2FReOpen%2Flibmdbx%2Ftree%2Fmaster&lang=ru-en). - -### Project Status - -**Now MDBX is under _active development_** and until 2018 is expected a big -change both of API and database format. Unfortunately those update will lead to -loss of compatibility with previous versions. - -The aim of this revolution in providing a clearer robust API and adding new -features, including the database properties. - - -## Содержание - -- [Обзор](#Обзор) - - [Сравнение с другими СУБД](#Сравнение-с-другими-СУБД) - - [История & Acknowledgements](#История) -- [Основные свойства](#Основные-свойства) -- [Сравнение производительности](#Сравнение-производительности) - - [Интегральная производительность](#Интегральная-производительность) - - [Масштабируемость чтения](#Масштабируемость-чтения) - - [Синхронная фиксация](#Синхронная-фиксация) - - [Отложенная фиксация](#Отложенная-фиксация) - - [Асинхронная фиксация](#Асинхронная-фиксация) - - [Потребление ресурсов](#Потребление-ресурсов) -- [Недостатки и Компромиссы](#Недостатки-и-Компромиссы) - - [Проблема долгих чтений](#Проблема-долгих-чтений) - - [Сохранность данных в режиме асинхронной фиксации](#Сохранность-данных-в-режиме-асинхронной-фиксации) -- [Доработки и усовершенствования относительно LMDB](#Доработки-и-усовершенствования-относительно-lmdb) - - -## Обзор - -_libmdbx_ - это встраиваемый key-value движок хранения со специфическим -набором свойств и возможностей, ориентированный на создание уникальных -легковесных решений с предельной производительностью. - -_libmdbx_ позволяет множеству процессов совместно читать и обновлять -несколько key-value таблиц с соблюдением [ACID](https://ru.wikipedia.org/wiki/ACID), -при минимальных накладных расходах и амортизационной стоимости любых операций Olog(N). - -_libmdbx_ обеспечивает -[serializability](https://en.wikipedia.org/wiki/Serializability) -изменений и согласованность данных после аварий. При этом транзакции -изменяющие данные никак не мешают операциям чтения и выполняются строго -последовательно с использованием единственного -[мьютекса](https://en.wikipedia.org/wiki/Mutual_exclusion). - -_libmdbx_ позволяет выполнять операции чтения с гарантиями -[wait-free](https://en.wikipedia.org/wiki/Non-blocking_algorithm#Wait-freedom), -параллельно на каждом ядре CPU, без использования атомарных операций -и/или примитивов синхронизации. - -_libmdbx_ не использует [LSM](https://en.wikipedia.org/wiki/Log-structured_merge-tree), а основан на [B+Tree](https://en.wikipedia.org/wiki/B%2B_tree) с [отображением](https://en.wikipedia.org/wiki/Memory-mapped_file) всех данных в память, -при этом текущая версия не использует [WAL](https://en.wikipedia.org/wiki/Write-ahead_logging). -Это предопределяет многие свойства, в том числе удачные и противопоказанные сценарии использования. - -### Сравнение с другими СУБД - -Ввиду того, что в _libmdbx_ сейчас происходит революция, я посчитал лучшим решением -ограничится здесь ссылкой на [главу Comparison with other databases](https://github.com/coreos/bbolt#comparison-with-other-databases) в описании _BoltDB_. - - -### История - -_libmdbx_ является результатом переработки и развития "Lightning Memory-Mapped Database", -известной под аббревиатурой -[LMDB](https://en.wikipedia.org/wiki/Lightning_Memory-Mapped_Database). -Изначально доработка производилась в составе проекта -[ReOpenLDAP](https://github.com/leo-yuriev/ReOpenLDAP). Примерно за год -работы внесенные изменения приобрели самостоятельную ценность. Осенью -2015 доработанный движок был выделен в отдельный проект, который был -[представлен на конференции Highload++ -2015](http://www.highload.ru/2015/abstracts/1831.html). - -В начале 2017 года движок _libmdbx_ получил новый импульс развития, -благодаря использованию в [Fast Positive -Tables](https://github.com/leo-yuriev/libfpta), aka ["Позитивные -Таблицы"](https://github.com/leo-yuriev/libfpta) by [Positive -Technologies](https://www.ptsecurity.ru). - - -#### Acknowledgements - -Howard Chu (Symas Corporation) - the author of LMDB, -from which originated the MDBX in 2015. - -Martin Hedenfalk <martin@bzero.se> - the author of `btree.c` code, -which was used for begin development of LMDB. - - -Основные свойства -================= - -_libmdbx_ наследует все ключевые возможности и особенности от -своего прародителя [LMDB](https://en.wikipedia.org/wiki/Lightning_Memory-Mapped_Database), -но с устранением ряда описываемых далее проблем и архитектурных недочетов. - -1. Данные хранятся в упорядоченном отображении (ordered map), ключи всегда - отсортированы, поддерживается выборка диапазонов (range lookups). - -2. Данные отображается в память каждого работающего с БД процесса. - К данным и ключам обеспечивается прямой доступ в памяти без необходимости их - копирования. - -3. Транзакции согласно - [ACID](https://ru.wikipedia.org/wiki/ACID), посредством - [MVCC](https://ru.wikipedia.org/wiki/MVCC) и - [COW](https://ru.wikipedia.org/wiki/%D0%9A%D0%BE%D0%BF%D0%B8%D1%80%D0%BE%D0%B2%D0%B0%D0%BD%D0%B8%D0%B5_%D0%BF%D1%80%D0%B8_%D0%B7%D0%B0%D0%BF%D0%B8%D1%81%D0%B8). - Изменения строго последовательны и не блокируются чтением, - конфликты между транзакциями не возможны. - При этом гарантируется чтение только зафиксированных данных, см [relaxing serializability](https://en.wikipedia.org/wiki/Serializability). - -4. Чтение и поиск [без блокировок](https://ru.wikipedia.org/wiki/%D0%9D%D0%B5%D0%B1%D0%BB%D0%BE%D0%BA%D0%B8%D1%80%D1%83%D1%8E%D1%89%D0%B0%D1%8F_%D1%81%D0%B8%D0%BD%D1%85%D1%80%D0%BE%D0%BD%D0%B8%D0%B7%D0%B0%D1%86%D0%B8%D1%8F), - без [атомарных операций](https://ru.wikipedia.org/wiki/%D0%90%D1%82%D0%BE%D0%BC%D0%B0%D1%80%D0%BD%D0%B0%D1%8F_%D0%BE%D0%BF%D0%B5%D1%80%D0%B0%D1%86%D0%B8%D1%8F). - Читатели не блокируются операциями записи и не конкурируют - между собой, чтение масштабируется линейно по ядрам CPU. - > Для точности следует отметить, что "подключение к БД" (старт первой - > читающей транзакции в потоке) и "отключение от БД" (закрытие БД или - > завершение потока) требуют краткосрочного захвата блокировки для - > регистрации/дерегистрации текущего потока в "таблице читателей". - -5. Эффективное хранение дубликатов (ключей с несколькими - значениями), без дублирования ключей, с сортировкой значений, в - том числе целочисленных (для вторичных индексов). - -6. Эффективная поддержка коротких ключей фиксированной длины, в том числе целочисленных. - -7. Амортизационная стоимость любой операции Olog(N), - [WAF](https://en.wikipedia.org/wiki/Write_amplification) (Write - Amplification Factor) и RAF (Read Amplification Factor) также Olog(N). - -8. Нет [WAL](https://en.wikipedia.org/wiki/Write-ahead_logging) и журнала - транзакций, после сбоев не требуется восстановление. Не требуется компактификация - или какое-либо периодическое обслуживание. Поддерживается резервное копирование - "по горячему", на работающей БД без приостановки изменения данных. - -9. Отсутствует какое-либо внутреннее управление памятью или кэшированием. Всё - необходимое штатно выполняет ядро ОС! - - -Сравнение производительности -============================ - -Все представленные ниже данные получены многократным прогоном тестов на -ноутбуке Lenovo Carbon-2, i7-4600U 2.1 ГГц, 8 Гб ОЗУ, с SSD-диском -SAMSUNG MZNTD512HAGL-000L1 (DXT23L0Q) 512 Гб. - -Исходный код бенчмарка [_IOArena_](https://github.com/pmwkaa/ioarena) и -сценарии тестирования [доступны на -github](https://github.com/pmwkaa/ioarena/tree/HL%2B%2B2015). - --------------------------------------------------------------------------------- - -### Интегральная производительность - -Показана соотнесенная сумма ключевых показателей производительности в трёх -бенчмарках: - - - Чтение/Поиск на машине с 4-мя процессорами; - - - Транзакции с [CRUD](https://ru.wikipedia.org/wiki/CRUD)-операциями - (вставка, чтение, обновление, удаление) в режиме **синхронной фиксации** - данных (fdatasync при завершении каждой транзакции или аналог); - - - Транзакции с [CRUD](https://ru.wikipedia.org/wiki/CRUD)-операциями - (вставка, чтение, обновление, удаление) в режиме **отложенной фиксации** - данных (отложенная запись посредством файловой систем или аналог); - -*Бенчмарк в режиме асинхронной записи не включен по двум причинам:* - - 1. Такое сравнение не совсем правомочно, его следует делать с движками - ориентированными на хранение данных в памяти ([Tarantool](https://tarantool.io/), [Redis](https://redis.io/)). - - 2. Превосходство libmdbx становится еще более подавляющем, что мешает - восприятию информации. - - - --------------------------------------------------------------------------------- - -### Масштабируемость чтения - -Для каждого движка показана суммарная производительность при -одновременном выполнении запросов чтения/поиска в 1-2-4-8 потоков на -машине с 4-мя физическими процессорами. - - - --------------------------------------------------------------------------------- - -### Синхронная фиксация - - - Линейная шкала слева и темные прямоугольники соответствуют количеству - транзакций в секунду, усредненному за всё время теста. - - - Логарифмическая шкала справа и желтые интервальные отрезки - соответствуют времени выполнения транзакций. При этом каждый отрезок - показывает минимальное и максимальное время затраченное на выполнение - транзакций, а крестиком отмечено среднеквадратичное значение. - -Выполняется **10.000 транзакций в режиме синхронной фиксации данных** на -диске. При этом требуется гарантия, что при аварийном выключении питания -(или другом подобном сбое) все данные будут консистентны и полностью -соответствовать последней завершенной транзакции. В _libmdbx_ в этом -режиме при фиксации каждой транзакции выполняется системный вызов -[fdatasync](https://linux.die.net/man/2/fdatasync). - -В каждой транзакции выполняется комбинированная CRUD-операция (две -вставки, одно чтение, одно обновление, одно удаление). Бенчмарк стартует -на пустой базе, а при завершении, в результате выполняемых действий, в -базе насчитывается 10.000 небольших key-value записей. - - - --------------------------------------------------------------------------------- - -### Отложенная фиксация - - - Линейная шкала слева и темные прямоугольники соответствуют количеству - транзакций в секунду, усредненному за всё время теста. - - - Логарифмическая шкала справа и желтые интервальные отрезки - соответствуют времени выполнения транзакций. При этом каждый отрезок - показывает минимальное и максимальное время затраченное на выполнение - транзакций, а крестиком отмечено среднеквадратичное значение. - -Выполняется **100.000 транзакций в режиме отложенной фиксации данных** -на диске. При этом требуется гарантия, что при аварийном выключении -питания (или другом подобном сбое) все данные будут консистентны на -момент завершения одной из транзакций, но допускается потеря изменений -из некоторого количества последних транзакций, что для многих движков -предполагает включение -[WAL](https://en.wikipedia.org/wiki/Write-ahead_logging) (write-ahead -logging) либо журнала транзакций, который в свою очередь опирается на -гарантию упорядоченности данных в журналируемой файловой системе. -_libmdbx_ при этом не ведет WAL, а передает весь контроль файловой -системе и ядру ОС. - -В каждой транзакции выполняется комбинированная CRUD-операция (две -вставки, одно чтение, одно обновление, одно удаление). Бенчмарк стартует -на пустой базе, а при завершении, в результате выполняемых действий, в -базе насчитывается 100.000 небольших key-value записей. - - - --------------------------------------------------------------------------------- - -### Асинхронная фиксация - - - Линейная шкала слева и темные прямоугольники соответствуют количеству - транзакций в секунду, усредненному за всё время теста. - - - Логарифмическая шкала справа и желтые интервальные отрезки - соответствуют времени выполнения транзакций. При этом каждый отрезок - показывает минимальное и максимальное время затраченное на выполнение - транзакций, а крестиком отмечено среднеквадратичное значение. - -Выполняется **1.000.000 транзакций в режиме асинхронной фиксации -данных** на диске. При этом требуется гарантия, что при аварийном -выключении питания (или другом подобном сбое) все данные будут -консистентны на момент завершения одной из транзакций, но допускается -потеря изменений из значительного количества последних транзакций. Во -всех движках при этом включался режим предполагающий минимальную -нагрузку на диск по-записи, и соответственно минимальную гарантию -сохранности данных. В _libmdbx_ при этом используется режим асинхронной -записи измененных страниц на диск посредством ядра ОС и системного -вызова [msync(MS_ASYNC)](https://linux.die.net/man/2/msync). - -В каждой транзакции выполняется комбинированная CRUD-операция (две -вставки, одно чтение, одно обновление, одно удаление). Бенчмарк стартует -на пустой базе, а при завершении, в результате выполняемых действий, в -базе насчитывается 10.000 небольших key-value записей. - - - --------------------------------------------------------------------------------- - -### Потребление ресурсов - -Показана соотнесенная сумма использованных ресурсов в ходе бенчмарка в -режиме отложенной фиксации: - - - суммарное количество операций ввода-вывода (IOPS), как записи, так и - чтения. - - - суммарное затраченное время процессора, как в режиме пользовательских процессов, - так и в режиме ядра ОС. - - - использованное место на диске при завершении теста, после закрытия БД из тестирующего процесса, - но без ожидания всех внутренних операций обслуживания (компактификации LSM и т.п.). - -Движок _ForestDB_ был исключен при оформлении результатов, так как -относительно конкурентов многократно превысил потребление каждого из -ресурсов (потратил процессорное время на генерацию IOPS для заполнения -диска), что не позволяло наглядно сравнить показатели остальных движков -на одной диаграмме. - -Все данные собирались посредством системного вызова -[getrusage()](http://man7.org/linux/man-pages/man2/getrusage.2.html) и -сканированием директорий с данными. - - - --------------------------------------------------------------------------------- - -## Недостатки и Компромиссы - -1. Единовременно может выполняться не более одной транзакция изменения данных - (один писатель). Зато все изменения всегда последовательны, не может быть - конфликтов или логических ошибок при откате транзакций. - -2. Отсутствие [WAL](https://en.wikipedia.org/wiki/Write-ahead_logging) - обуславливает относительно большой - [WAF](https://en.wikipedia.org/wiki/Write_amplification) (Write - Amplification Factor). Поэтому фиксация изменений на диске может быть - достаточно дорогой и являться главным ограничением производительности - при интенсивном изменении данных. - > В качестве компромисса _libmdbx_ предлагает несколько режимов ленивой - > и/или периодической фиксации. В том числе режим `MAPASYNC`, при котором - > изменения происходят только в памяти и асинхронно фиксируются на диске - > ядром ОС. - > - > Однако, следует воспринимать это свойство аккуратно и взвешенно. - > Например, полная фиксация транзакции в БД с журналом потребует минимум 2 - > IOPS (скорее всего 3-4) из-за накладных расходов в файловой системе. В - > _libmdbx_ фиксация транзакции также требует от 2 IOPS. Однако, в БД с - > журналом кол-во IOPS будет меняться в зависимости от файловой системы, - > но не от кол-ва записей или их объема. Тогда как в _libmdbx_ кол-во - > будет расти логарифмически от кол-во записей/строк в БД (по высоте - > b+tree). - -3. [COW](https://ru.wikipedia.org/wiki/%D0%9A%D0%BE%D0%BF%D0%B8%D1%80%D0%BE%D0%B2%D0%B0%D0%BD%D0%B8%D0%B5_%D0%BF%D1%80%D0%B8_%D0%B7%D0%B0%D0%BF%D0%B8%D1%81%D0%B8) - для реализации [MVCC](https://ru.wikipedia.org/wiki/MVCC) выполняется на - уровне страниц в [B+ - дереве](https://ru.wikipedia.org/wiki/B-%D0%B4%D0%B5%D1%80%D0%B5%D0%B2%D0%BE). - Поэтому изменение данных амортизационно требует копирования Olog(N) - страниц, что расходует [пропускную способность оперативной - памяти](https://en.wikipedia.org/wiki/Memory_bandwidth) и является - основным ограничителем производительности в режиме `MAPASYNC`. - > Этот недостаток неустраним, тем не менее следует дать некоторые пояснения. - > Дело в том, что фиксация изменений на диске потребует гораздо более - > значительного копирования данных в памяти и массы других затратных операций. - > Поэтому обусловленное этим недостатком падение производительности становится - > заметным только при отказе от фиксации изменений на диске. - > Соответственно, корректнее сказать что _libmdbx_ позволяет - > получить персистентность ценой минимального падения производительности. - > Если же нет необходимости оперативно сохранять данные, то логичнее - > использовать `std::map`. - -4. В _LMDB_ существует проблема долгих чтений (приостановленных читателей), - которая приводит к деградации производительности и переполнению БД. - > В _libmdbx_ предложены средства для предотвращения, быстрого выхода из - > некомфортной ситуации и устранения её последствий. Подробности ниже. - -5. В _LMDB_ есть вероятность разрушения БД в режиме `WRITEMAP+MAPASYNC`. - В _libmdbx_ для `WRITEMAP+MAPASYNC` гарантируется как сохранность базы, - так и согласованность данных. - > Дополнительно, в качестве альтернативы, предложен режим `UTTERLY_NOSYNC`. - > Подробности ниже. - - -#### Проблема долгих чтений - -*Следует отметить*, что проблема "сборки мусора" так или иначе -существует во всех СУБД (Vacuum в PostgreSQL). Однако в случае _libmdbx_ -и LMDB она проявляется более остро, прежде всего из-за высокой -производительности, а также из-за намеренного упрощения внутренних -механизмов ради производительности. - -Понимание проблемы требует некоторых пояснений, которые -изложены ниже, но могут быть сложны для быстрого восприятия. -Поэтому, тезисно: - -* Изменение данных на фоне долгой операции чтения может - приводить к исчерпанию места в БД. - -* После чего любая попытка обновить данные будет приводить к - ошибке `MAP_FULL` до завершения долгой операции чтения. - -* Характерными примерами долгих чтений являются горячее - резервное копирования и отладка клиентского приложения при - активной транзакции чтения. - -* В оригинальной _LMDB_ после этого будет наблюдаться - устойчивая деградация производительности всех механизмов - обратной записи на диск (в I/O контроллере, в гипервизоре, - в ядре ОС). - -* В _libmdbx_ предусмотрен механизм аварийного прерывания таких - операций, а также режим `LIFO RECLAIM` устраняющий последующую - деградацию производительности. - -Операции чтения выполняются в контексте снимка данных (версии -БД), который был актуальным на момент старта транзакции чтения. Такой -читаемый снимок поддерживается неизменным до завершения операции. В свою -очередь, это не позволяет повторно использовать страницы БД в -последующих версиях (снимках БД). - -Другими словами, если обновление данных выполняется на фоне долгой -операции чтения, то вместо повторного использования "старых" ненужных -страниц будут выделяться новые, так как "старые" страницы составляют -снимок БД, который еще используется долгой операцией чтения. - -В результате, при интенсивном изменении данных и достаточно длительной -операции чтения, в БД могут быть исчерпаны свободные страницы, что не -позволит создавать новые снимки/версии БД. Такая ситуация будет -сохраняться до завершения операции чтения, которая использует старый -снимок данных и препятствует повторному использованию страниц БД. - -Однако, на этом проблемы не заканчиваются. После описанной ситуации, все -дополнительные страницы, которые были выделены пока переработка старых -была невозможна, будут участвовать в цикле выделения/освобождения до -конца жизни экземпляра БД. В оригинальной _LMDB_ этот цикл использования -страниц работает по принципу [FIFO](https://ru.wikipedia.org/wiki/FIFO). -Поэтому увеличение количества циркулирующий страниц, с точки зрения -механизмов кэширования и/или обратной записи, выглядит как увеличение -рабочего набор данных. Проще говоря, однократное попадание в ситуацию -"уснувшего читателя" приводит к устойчивому эффекту вымывания I/O кэша -при всех последующих изменениях данных. - -Для устранения описанных проблемы в _libmdbx_ сделаны существенные -доработки, подробности ниже. Иллюстрации к проблеме "долгих чтений" -можно найти в [слайдах презентации](http://www.slideshare.net/leoyuriev/lmdb). - -Там же приведен пример количественной оценки прироста производительности -за счет эффективной работы [BBWC](https://en.wikipedia.org/wiki/BBWC) -при включении `LIFO RECLAIM` в _libmdbx_. - - -#### Сохранность данных в режиме асинхронной фиксации - -При работе в режиме `WRITEMAP+MAPSYNC` запись измененных страниц -выполняется ядром ОС, что имеет ряд преимуществ. Так например, при крахе -приложения, ядро ОС сохранит все изменения. - -Однако, при аварийном отключении питания или сбое в ядре ОС, на диске -может быть сохранена только часть измененных страниц БД. При этом с большой -вероятностью может оказаться так, что будут сохранены мета-страницы со -ссылками на страницы с новыми версиями данных, но не сами новые данные. -В этом случае БД будет безвозвратна разрушена, даже если до аварии -производилась полная синхронизация данных (посредством -`mdbx_env_sync()`). - -В _libmdbx_ эта проблема устранена путем полной переработки -пути записи данных: - -* В режиме `WRITEMAP+MAPSYNC` _libmdbx_ не обновляет - мета-страницы непосредственно, а поддерживает их теневые копии - с переносом изменений после фиксации данных. - -* При завершении транзакций, в зависимости от состояния - синхронности данных между диском и оперативной память, - _libmdbx_ помечает точки фиксации либо как сильные (strong), - либо как слабые (weak). Так например, в режиме - `WRITEMAP+MAPSYNC` завершаемые транзакции помечаются как - слабые, а при явной синхронизации данных как сильные. - -* В _libmdbx_ поддерживается не две, а три отдельные мета-страницы. - Это позволяет выполнять фиксацию транзакций с формированием как - сильной, так и слабой точки фиксации, без потери двух предыдущих - точек фиксации (из которых одна может быть сильной, а вторая слабой). - В результате, _libmdbx_ позволяет в произвольном порядке чередовать - сильные и слабые точки фиксации без нарушения соответствующих - гарантий в случае неожиданной системной аварии во время фиксации. - -* При открытии БД выполняется автоматический откат к последней - сильной фиксации. Этим обеспечивается гарантия сохранности БД. - -Такая гарантия надежности не дается бесплатно. Для -сохранности данных, страницы формирующие крайний снимок с -сильной фиксацией, не должны повторно использоваться -(перезаписываться) до формирования следующей сильной точки -фиксации. Таким образом, крайняя точка фиксации создает -описанный выше эффект "долгого чтения". Разница же здесь в том, -что при исчерпании свободных страниц ситуация будет -автоматически исправлена, посредством записи изменений на диск -и формированием новой сильной точки фиксации. - -Таким образом, в режиме безопасной асинхронной фиксации _libmdbx_ будет -всегда использовать новые страницы до исчерпания места в БД или до явного -формирования сильной точки фиксации посредством `mdbx_env_sync()`. -При этом суммарный трафик записи на диск будет примерно такой-же, -как если бы отдельно фиксировалась каждая транзакций. - -В текущей версии _libmdbx_ вам предоставляется выбор между безопасным -режимом (по умолчанию) асинхронной фиксации, и режимом `UTTERLY_NOSYNC` когда -при системной аварии есть шанс полного разрушения БД как в LMDB. - -В последующих версиях _libmdbx_ будут предусмотрены средства -для асинхронной записи данных на диск с автоматическим -формированием сильных точек фиксации. - --------------------------------------------------------------------------------- - -Доработки и усовершенствования относительно LMDB -================================================ - -1. Режим `LIFO RECLAIM`. - - Для повторного использования выбираются не самые старые, а - самые новые страницы из доступных. За счет этого цикл - использования страниц всегда имеет минимальную длину и не - зависит от общего числа выделенных страниц. - - В результате механизмы кэширования и обратной записи работают с - максимально возможной эффективностью. В случае использования - контроллера дисков или системы хранения с - [BBWC](https://en.wikipedia.org/wiki/BBWC) возможно - многократное увеличение производительности по записи - (обновлению данных). - -2. Обработчик `OOM-KICK`. - - Посредством `mdbx_env_set_oomfunc()` может быть установлен - внешний обработчик (callback), который будет вызван при - исчерпания свободных страниц из-за долгой операцией чтения. - Обработчику будет передан PID и pthread_id виновника. - В свою очередь обработчик может предпринять одно из действий: - - * нейтрализовать виновника (отправить сигнал kill #9), если - долгое чтение выполняется сторонним процессом; - - * отменить или перезапустить проблемную операцию чтения, если - операция выполняется одним из потоков текущего процесса; - - * подождать некоторое время, в расчете что проблемная операция - чтения будет штатно завершена; - - * прервать текущую операцию изменения данных с возвратом кода - ошибки. - -3. Гарантия сохранности БД в режиме `WRITEMAP+MAPSYNC`. - -В текущей версии _libmdbx_ вам предоставляется выбор между безопасным -режимом (по умолчанию) асинхронной фиксации, и режимом `UTTERLY_NOSYNC` -когда при системной аварии есть шанс полного разрушения БД как в LMDB. -Для подробностей смотрите раздел -[Сохранность данных в режиме асинхронной фиксации](#Сохранность-данных-в-режиме-асинхронной-фиксации). - -4. Возможность автоматического формирования контрольных точек -(сброса данных на диск) при накоплении заданного объёма изменений, -устанавливаемого функцией `mdbx_env_set_syncbytes()`. - -5. Возможность получить отставание текущей транзакции чтения от -последней версии данных в БД посредством `mdbx_txn_straggler()`. - -6. Утилита mdbx_chk для проверки БД и функция `mdbx_env_pgwalk()` для -обхода всех страниц БД. - -7. Управление отладкой и получение отладочных сообщений посредством -`mdbx_setup_debug()`. - -8. Возможность связать с каждой завершаемой транзакцией до 3 -дополнительных маркеров посредством `mdbx_canary_put()`, и прочитать их -в транзакции чтения посредством `mdbx_canary_get()`. - -9. Возможность узнать есть ли за текущей позицией курсора строка данных -посредством `mdbx_cursor_eof()`. - -10. Возможность явно запросить обновление существующей записи, без -создания новой посредством флажка `MDBX_CURRENT` для `mdbx_put()`. - -11. Возможность посредством `mdbx_replace()` обновить или удалить запись -с получением предыдущего значения данных, а также адресно изменить -конкретное multi-значение. - -12. Поддержка ключей и значений нулевой длины, включая сортированные -дубликаты. - -13. Исправленный вариант `mdbx_cursor_count()`, возвращающий корректное -количество дубликатов для всех типов таблиц и любого положения курсора. - -14. Возможность открыть БД в эксклюзивном режиме посредством -`mdbx_env_open_ex()`, например в целях её проверки. - -15. Возможность закрыть БД в "грязном" состоянии (без сброса данных и -формирования сильной точки фиксации) посредством `mdbx_env_close_ex()`. - -16. Возможность получить посредством `mdbx_env_info()` дополнительную -информацию, включая номер самой старой версии БД (снимка данных), -который используется одним из читателей. - -17. Функция `mdbx_del()` не игнорирует дополнительный (уточняющий) -аргумент `data` для таблиц без дубликатов (без флажка `MDBX_DUPSORT`), а -при его ненулевом значении всегда использует его для сверки с удаляемой -записью. - -18. Возможность открыть dbi-таблицу, одновременно с установкой -компараторов для ключей и данных, посредством `mdbx_dbi_open_ex()`. - -19. Возможность посредством `mdbx_is_dirty()` определить находятся ли -некоторый ключ или данные в "грязной" странице БД. Таким образом, -избегая лишнего копирования данных перед выполнением модифицирующих -операций (значения в размещенные "грязных" страницах могут быть -перезаписаны при изменениях, иначе они будут неизменны). - -20. Корректное обновление текущей записи, в том числе сортированного -дубликата, при использовании режима `MDBX_CURRENT` в -`mdbx_cursor_put()`. - -21. Все курсоры, как в транзакциях только для чтения, так и в пишущих, -могут быть переиспользованы посредством `mdbx_cursor_renew()` и ДОЛЖНЫ -ОСВОБОЖДАТЬСЯ ЯВНО. - > - > ## _ВАЖНО_, Обратите внимание! - > - > Это единственное изменение в API, которое значимо меняет - > семантику управления курсорами и может приводить к утечкам - > памяти. Следует отметить, что это изменение вынужденно. - > Так устраняется неоднозначность с массой тяжких последствий: - > - > - обращение к уже освобожденной памяти; - > - попытки повторного освобождения памяти; - > - memory corruption and segfaults. - -22. Дополнительный код ошибки `MDBX_EMULTIVAL`, который возвращается из -`mdbx_put()` и `mdbx_replace()` при попытке выполнить неоднозначное -обновление или удаления одного из нескольких значений с одним ключом. - -23. Возможность посредством `mdbx_get_ex()` получить значение по -заданному ключу, одновременно с количеством дубликатов. - -24. Наличие функций `mdbx_cursor_on_first()` и `mdbx_cursor_on_last()`, -которые позволяют быстро выяснить стоит ли курсор на первой/последней -позиции. - -25. При завершении читающих транзакций, открытые в них DBI-хендлы не -закрываются и не теряются при завершении таких транзакций посредством -`mdbx_txn_abort()` или `mdbx_txn_reset()`. Что позволяет избавится от ряда -сложно обнаруживаемых ошибок. - -26. Генерация последовательностей посредством `mdbx_dbi_sequence()`. - -27. Расширенное динамическое управление размером БД, включая выбор -размера страницы посредством `mdbx_env_set_geometry()`. - -28. Три мета-страницы вместо двух, что позволяет гарантированно -консистентно обновлять слабые контрольные точки фиксации без риска -повредить крайнюю сильную точку фиксации. - -29. В _libmdbx_ реализован автоматический возврат освобождающихся -страниц в область нераспределенного резерва в конце файла данных. При -этом уменьшается количество страниц загруженных в память и участвующих в -цикле обновления данных и записи на диск. Фактически _libmdbx_ выполняет -постоянную компактификацию данных, но не затрачивая на это -дополнительных ресурсов, а только освобождая их. При освобождении места -в БД, в случае наличия поддержки со стороны операционной системы и -установки соответствующих параметров геометрии базы данных, также будет -уменьшаться размер файла на диске. - --------------------------------------------------------------------------------- - -``` -$ objdump -f -h -j .text libmdbx.so - -libmdbx.so: file format elf64-x86-64 -architecture: i386:x86-64, flags 0x00000150: -HAS_SYMS, DYNAMIC, D_PAGED -start address 0x000030e0 - -Sections: -Idx Name Size VMA LMA File off Algn - 11 .text 00014661 000030e0 000030e0 000030e0 2**4 - CONTENTS, ALLOC, LOAD, READONLY, CODE -``` - -``` -$ objdump -C -T libmdbx.so | grep mdbx | sort - -00004057 g DF .text 0000003f Base mdbx_strerror_r -00004096 g DF .text 00000031 Base mdbx_strerror -00004207 g DF .text 00000025 Base mdbx_env_get_maxkeysize -0000422c g DF .text 000000b8 Base mdbx_env_create -000042e4 g DF .text 0000001f Base mdbx_env_set_mapsize -00004f9f g DF .text 00000037 Base mdbx_env_set_maxdbs -00004fd6 g DF .text 00000036 Base mdbx_env_set_maxreaders -0000500c g DF .text 00000027 Base mdbx_env_get_maxreaders -00005033 g DF .text 0000066a Base mdbx_env_open_ex -0000569d g DF .text 00000008 Base mdbx_env_open -000056a5 g DF .text 00000096 Base mdbx_env_close_ex -0000573b g DF .text 00000007 Base mdbx_env_close -00005742 g DF .text 00000047 Base mdbx_env_set_flags -00005789 g DF .text 0000001d Base mdbx_env_get_flags -000057a6 g DF .text 00000014 Base mdbx_env_set_userctx -000057ba g DF .text 0000000f Base mdbx_env_get_userctx -000057c9 g DF .text 0000000d Base mdbx_env_set_assert -000057d6 g DF .text 0000001d Base mdbx_env_get_path -000057f3 g DF .text 00000018 Base mdbx_env_get_fd -0000580b g DF .text 00000056 Base mdbx_env_stat -00005861 g DF .text 00000276 Base mdbx_env_info -00005ad7 g DF .text 00000148 Base mdbx_reader_list -0000656a g DF .text 0000012a Base mdbx_dbi_stat -0000693a g DF .text 00000146 Base mdbx_env_copy2fd -00006a80 g DF .text 0000012e Base mdbx_env_copy -00006bae g DF .text 0000002a Base mdbx_reader_check -00006bd8 g DF .text 000000f9 Base mdbx_setup_debug -00006cd1 g DF .text 00000033 Base mdbx_env_set_syncbytes -00006d04 g DF .text 00000023 Base mdbx_env_set_oomfunc -00006d27 g DF .text 00000019 Base mdbx_env_get_oomfunc -00006d40 g DF .text 00000121 Base mdbx_env_pgwalk -0000ac60 g DF .text 00000163 Base mdbx_dkey -0000add0 g DF .text 00000016 Base mdbx_cmp -0000adf0 g DF .text 00000016 Base mdbx_dcmp -0000ae10 g DF .text 00000271 Base mdbx_env_sync -0000b090 g DF .text 0000001b Base mdbx_txn_env -0000b0b0 g DF .text 0000001c Base mdbx_txn_id -0000b0d0 g DF .text 00000077 Base mdbx_txn_reset -0000b150 g DF .text 00000077 Base mdbx_txn_abort -0000b1d0 g DF .text 00000057 Base mdbx_get_maxkeysize -0000b230 g DF .text 000006b7 Base mdbx_env_set_geometry -0000b8f0 g DF .text 000000ef Base mdbx_cursor_count -0000b9e0 g DF .text 000000ad Base mdbx_cursor_close -0000ba90 g DF .text 0000001b Base mdbx_cursor_txn -0000bab0 g DF .text 00000017 Base mdbx_cursor_dbi -0000bad0 g DF .text 0000007d Base mdbx_dbi_close -0000bb50 g DF .text 000000cc Base mdbx_dbi_flags_ex -0000bc20 g DF .text 00000038 Base mdbx_dbi_flags -0000c250 g DF .text 00000077 Base mdbx_txn_renew -0000c2d0 g DF .text 000004e5 Base mdbx_txn_begin -0000dcb0 g DF .text 00000128 Base mdbx_cursor_open -0000dde0 g DF .text 0000011d Base mdbx_cursor_renew -0000e970 g DF .text 000000fc Base mdbx_get -0000ef00 g DF .text 00000489 Base mdbx_cursor_get -000125e0 g DF .text 00000719 Base mdbx_cursor_del -00012e00 g DF .text 000000e4 Base mdbx_del -00012ef0 g DF .text 000002c3 Base mdbx_drop -000131c0 g DF .text 0000129e Base mdbx_cursor_put -000145d0 g DF .text 000000a7 Base mdbx_put -00014b60 g DF .text 000000bf Base mdbx_dbi_open_ex -00014c20 g DF .text 0000000b Base mdbx_dbi_open -00014c30 g DF .text 00001347 Base mdbx_txn_commit -00015f80 g DF .text 00000105 Base mdbx_txn_straggler -00016090 g DF .text 000000e7 Base mdbx_canary_put -00016180 g DF .text 00000078 Base mdbx_canary_get -00016200 g DF .text 0000006e Base mdbx_cursor_on_first -00016270 g DF .text 00000096 Base mdbx_cursor_on_last -00016310 g DF .text 00000066 Base mdbx_cursor_eof -00016380 g DF .text 00000504 Base mdbx_replace -00016890 g DF .text 0000017d Base mdbx_get_ex -00016a10 g DF .text 000000a4 Base mdbx_is_dirty -00016ac0 g DF .text 00000120 Base mdbx_dbi_sequence -00016be0 g DF .text 00000064 Base mdbx_cursor_get_attr -00016c50 g DF .text 00000064 Base mdbx_get_attr -00016cc0 g DF .text 000000c7 Base mdbx_put_attr -00016d90 g DF .text 000000c7 Base mdbx_cursor_put_attr -00016e60 g DF .text 00000244 Base mdbx_set_attr -``` diff --git a/plugins/Dbx_mdb/src/mdbx/bits.h b/plugins/Dbx_mdb/src/mdbx/bits.h deleted file mode 100644 index 4d3a50cc0b..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/bits.h +++ /dev/null @@ -1,1226 +0,0 @@ -/* - * Copyright 2015-2017 Leonid Yuriev <leo@yuriev.ru> - * and other libmdbx authors: please see AUTHORS file. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted only as authorized by the OpenLDAP - * Public License. - * - * A copy of this license is available in the file LICENSE in the - * top-level directory of the distribution or, alternatively, at - * <http://www.OpenLDAP.org/license.html>. */ - -#pragma once -/* *INDENT-OFF* */ -/* clang-format off */ - -#ifndef MDBX_DEBUG -# define MDBX_DEBUG 0 -#endif - -#if MDBX_DEBUG -# undef NDEBUG -#endif - -/* Features under development */ -#ifndef MDBX_DEVEL -# define MDBX_DEVEL 1 -#endif - -/*----------------------------------------------------------------------------*/ - -/* Should be defined before any includes */ -#ifndef _GNU_SOURCE -# define _GNU_SOURCE 1 -#endif -#ifndef _FILE_OFFSET_BITS -# define _FILE_OFFSET_BITS 64 -#endif - -#ifdef _MSC_VER -# ifndef _CRT_SECURE_NO_WARNINGS -# define _CRT_SECURE_NO_WARNINGS -# endif -#if _MSC_VER > 1800 -# pragma warning(disable : 4464) /* relative include path contains '..' */ -#endif -#pragma warning(disable : 4710) /* 'xyz': function not inlined */ -#pragma warning(disable : 4711) /* function 'xyz' selected for automatic inline expansion */ -#pragma warning(disable : 4201) /* nonstandard extension used : nameless struct / union */ -#pragma warning(disable : 4702) /* unreachable code */ -#pragma warning(disable : 4706) /* assignment within conditional expression */ -#pragma warning(disable : 4127) /* conditional expression is constant */ -#pragma warning(disable : 4324) /* 'xyz': structure was padded due to alignment specifier */ -#pragma warning(disable : 4310) /* cast truncates constant value */ -#pragma warning(disable : 4820) /* bytes padding added after data member for aligment */ -#pragma warning(disable : 4548) /* expression before comma has no effect; expected expression with side - effect */ -#endif /* _MSC_VER (warnings) */ - -#include "./mdbx.h" -#include "./defs.h" - -#if defined(__GNUC__) && !__GNUC_PREREQ(4,2) - /* Actualy libmdbx was not tested with compilers older than GCC from RHEL6. - * But you could remove this #error and try to continue at your own risk. - * In such case please don't rise up an issues related ONLY to old compilers. - */ -# warning "libmdbx required at least GCC 4.2 compatible C/C++ compiler." -#endif - -#if defined(__GLIBC__) && !__GLIBC_PREREQ(2,12) - /* Actualy libmdbx was not tested with something older than glibc 2.12 (from RHEL6). - * But you could remove this #error and try to continue at your own risk. - * In such case please don't rise up an issues related ONLY to old systems. - */ -# warning "libmdbx required at least GLIBC 2.12." -#endif - -#ifdef __SANITIZE_THREAD__ -# warning "libmdbx don't compatible with ThreadSanitizer, you will get a lot of false-positive issues." -#endif /* __SANITIZE_THREAD__ */ - -#include "./osal.h" - -/* *INDENT-ON* */ -/* clang-format on */ - -/*----------------------------------------------------------------------------*/ -/* Basic constants and types */ - -/* The minimum number of keys required in a database page. - * Setting this to a larger value will place a smaller bound on the - * maximum size of a data item. Data items larger than this size will - * be pushed into overflow pages instead of being stored directly in - * the B-tree node. This value used to default to 4. With a page size - * of 4096 bytes that meant that any item larger than 1024 bytes would - * go into an overflow page. That also meant that on average 2-3KB of - * each overflow page was wasted space. The value cannot be lower than - * 2 because then there would no longer be a tree structure. With this - * value, items larger than 2KB will go into overflow pages, and on - * average only 1KB will be wasted. */ -#define MDBX_MINKEYS 2 - -/* A stamp that identifies a file as an MDBX file. - * There's nothing special about this value other than that it is easily - * recognizable, and it will reflect any byte order mismatches. */ -#define MDBX_MAGIC UINT64_C(/* 56-bit prime */ 0x59659DBDEF4C11) - -/* The version number for a database's datafile format. */ -#define MDBX_DATA_VERSION ((MDBX_DEVEL) ? 255 : 2) -/* The version number for a database's lockfile format. */ -#define MDBX_LOCK_VERSION ((MDBX_DEVEL) ? 255 : 2) - -/* handle for the DB used to track free pages. */ -#define FREE_DBI 0 -/* handle for the default DB. */ -#define MAIN_DBI 1 -/* Number of DBs in metapage (free and main) - also hardcoded elsewhere */ -#define CORE_DBS 2 -#define MAX_DBI (INT16_MAX - CORE_DBS) - -/* Number of meta pages - also hardcoded elsewhere */ -#define NUM_METAS 3 - -/* A page number in the database. - * - * MDBX uses 32 bit for page numbers. This limits database - * size up to 2^44 bytes, in case of 4K pages. */ -typedef uint32_t pgno_t; -#define PRIaPGNO PRIu32 -#define MAX_PAGENO ((pgno_t)UINT64_C(0xffffFFFFffff)) -#define MIN_PAGENO NUM_METAS - -/* A transaction ID. */ -typedef uint64_t txnid_t; -#define PRIaTXN PRIi64 -#if MDBX_DEVEL -#define MIN_TXNID (UINT64_MAX - UINT32_MAX) -#elif MDBX_DEBUG -#define MIN_TXNID UINT64_C(0x100000000) -#else -#define MIN_TXNID UINT64_C(1) -#endif /* MIN_TXNID */ - -/* Used for offsets within a single page. - * Since memory pages are typically 4 or 8KB in size, 12-13 bits, - * this is plenty. */ -typedef uint16_t indx_t; - -#define MEGABYTE ((size_t)1 << 20) - -/*----------------------------------------------------------------------------*/ -/* Core structures for database and shared memory (i.e. format definition) */ -#pragma pack(push, 1) - -/* Reader Lock Table - * - * Readers don't acquire any locks for their data access. Instead, they - * simply record their transaction ID in the reader table. The reader - * mutex is needed just to find an empty slot in the reader table. The - * slot's address is saved in thread-specific data so that subsequent - * read transactions started by the same thread need no further locking to - * proceed. - * - * If MDBX_NOTLS is set, the slot address is not saved in thread-specific data. - * No reader table is used if the database is on a read-only filesystem. - * - * Since the database uses multi-version concurrency control, readers don't - * actually need any locking. This table is used to keep track of which - * readers are using data from which old transactions, so that we'll know - * when a particular old transaction is no longer in use. Old transactions - * that have discarded any data pages can then have those pages reclaimed - * for use by a later write transaction. - * - * The lock table is constructed such that reader slots are aligned with the - * processor's cache line size. Any slot is only ever used by one thread. - * This alignment guarantees that there will be no contention or cache - * thrashing as threads update their own slot info, and also eliminates - * any need for locking when accessing a slot. - * - * A writer thread will scan every slot in the table to determine the oldest - * outstanding reader transaction. Any freed pages older than this will be - * reclaimed by the writer. The writer doesn't use any locks when scanning - * this table. This means that there's no guarantee that the writer will - * see the most up-to-date reader info, but that's not required for correct - * operation - all we need is to know the upper bound on the oldest reader, - * we don't care at all about the newest reader. So the only consequence of - * reading stale information here is that old pages might hang around a - * while longer before being reclaimed. That's actually good anyway, because - * the longer we delay reclaiming old pages, the more likely it is that a - * string of contiguous pages can be found after coalescing old pages from - * many old transactions together. */ - -/* The actual reader record, with cacheline padding. */ -typedef struct MDBX_reader { - /* Current Transaction ID when this transaction began, or (txnid_t)-1. - * Multiple readers that start at the same time will probably have the - * same ID here. Again, it's not important to exclude them from - * anything; all we need to know is which version of the DB they - * started from so we can avoid overwriting any data used in that - * particular version. */ - volatile txnid_t mr_txnid; - - /* The information we store in a single slot of the reader table. - * In addition to a transaction ID, we also record the process and - * thread ID that owns a slot, so that we can detect stale information, - * e.g. threads or processes that went away without cleaning up. - * - * NOTE: We currently don't check for stale records. - * We simply re-init the table when we know that we're the only process - * opening the lock file. */ - - /* The process ID of the process owning this reader txn. */ - volatile mdbx_pid_t mr_pid; - /* The thread ID of the thread owning this txn. */ - volatile mdbx_tid_t mr_tid; - - /* cache line alignment */ - uint8_t pad[MDBX_CACHELINE_SIZE - - (sizeof(txnid_t) + sizeof(mdbx_pid_t) + sizeof(mdbx_tid_t)) % - MDBX_CACHELINE_SIZE]; -} __cache_aligned MDBX_reader; - -/* Information about a single database in the environment. */ -typedef struct MDBX_db { - uint16_t md_flags; /* see mdbx_dbi_open */ - uint16_t md_depth; /* depth of this tree */ - uint32_t md_xsize; /* also ksize for LEAF2 pages */ - pgno_t md_root; /* the root page of this tree */ - pgno_t md_branch_pages; /* number of internal pages */ - pgno_t md_leaf_pages; /* number of leaf pages */ - pgno_t md_overflow_pages; /* number of overflow pages */ - uint64_t md_seq; /* table sequence counter */ - uint64_t md_entries; /* number of data items */ - uint64_t md_merkle; /* Merkle tree checksum */ -} MDBX_db; - -/* Meta page content. - * A meta page is the start point for accessing a database snapshot. - * Pages 0-1 are meta pages. Transaction N writes meta page (N % 2). */ -typedef struct MDBX_meta { - /* Stamp identifying this as an MDBX file. - * It must be set to MDBX_MAGIC with MDBX_DATA_VERSION. */ - uint64_t mm_magic_and_version; - - /* txnid that committed this page, the first of a two-phase-update pair */ - volatile txnid_t mm_txnid_a; - - uint16_t mm_extra_flags; /* extra DB flags, zero (nothing) for now */ - uint8_t mm_validator_id; /* ID of checksum and page validation method, - * zero (nothing) for now */ - uint8_t mm_extra_pagehdr; /* extra bytes in the page header, - * zero (nothing) for now */ - - struct { - uint16_t grow; /* datafile growth step in pages */ - uint16_t shrink; /* datafile shrink threshold in pages */ - pgno_t lower; /* minimal size of datafile in pages */ - pgno_t upper; /* maximal size of datafile in pages */ - pgno_t now; /* current size of datafile in pages */ - pgno_t next; /* first unused page in the datafile, - * but actually the file may be shorter. */ - } mm_geo; - - MDBX_db mm_dbs[CORE_DBS]; /* first is free space, 2nd is main db */ - /* The size of pages used in this DB */ -#define mm_psize mm_dbs[FREE_DBI].md_xsize -/* Any persistent environment flags, see mdbx_env */ -#define mm_flags mm_dbs[FREE_DBI].md_flags - mdbx_canary mm_canary; - -#define MDBX_DATASIGN_NONE 0u -#define MDBX_DATASIGN_WEAK 1u -#define SIGN_IS_WEAK(sign) ((sign) == MDBX_DATASIGN_WEAK) -#define SIGN_IS_STEADY(sign) ((sign) > MDBX_DATASIGN_WEAK) -#define META_IS_WEAK(meta) SIGN_IS_WEAK((meta)->mm_datasync_sign) -#define META_IS_STEADY(meta) SIGN_IS_STEADY((meta)->mm_datasync_sign) - volatile uint64_t mm_datasync_sign; - - /* txnid that committed this page, the second of a two-phase-update pair */ - volatile txnid_t mm_txnid_b; -} MDBX_meta; - -/* Common header for all page types. The page type depends on mp_flags. - * - * P_BRANCH and P_LEAF pages have unsorted 'MDBX_node's at the end, with - * sorted mp_ptrs[] entries referring to them. Exception: P_LEAF2 pages - * omit mp_ptrs and pack sorted MDBX_DUPFIXED values after the page header. - * - * P_OVERFLOW records occupy one or more contiguous pages where only the - * first has a page header. They hold the real data of F_BIGDATA nodes. - * - * P_SUBP sub-pages are small leaf "pages" with duplicate data. - * A node with flag F_DUPDATA but not F_SUBDATA contains a sub-page. - * (Duplicate data can also go in sub-databases, which use normal pages.) - * - * P_META pages contain MDBX_meta, the start point of an MDBX snapshot. - * - * Each non-metapage up to MDBX_meta.mm_last_pg is reachable exactly once - * in the snapshot: Either used by a database or listed in a freeDB record. */ -typedef struct MDBX_page { - union { - struct MDBX_page *mp_next; /* for in-memory list of freed pages, - * must be first field, see NEXT_LOOSE_PAGE */ - uint64_t mp_validator; /* checksum of page content or a txnid during - * which the page has been updated */ - }; - uint16_t mp_leaf2_ksize; /* key size if this is a LEAF2 page */ -#define P_BRANCH 0x01 /* branch page */ -#define P_LEAF 0x02 /* leaf page */ -#define P_OVERFLOW 0x04 /* overflow page */ -#define P_META 0x08 /* meta page */ -#define P_DIRTY 0x10 /* dirty page, also set for P_SUBP pages */ -#define P_LEAF2 0x20 /* for MDBX_DUPFIXED records */ -#define P_SUBP 0x40 /* for MDBX_DUPSORT sub-pages */ -#define P_LOOSE 0x4000 /* page was dirtied then freed, can be reused */ -#define P_KEEP 0x8000 /* leave this page alone during spill */ - uint16_t mp_flags; - union { - struct { - indx_t mp_lower; /* lower bound of free space */ - indx_t mp_upper; /* upper bound of free space */ - }; - uint32_t mp_pages; /* number of overflow pages */ - }; - pgno_t mp_pgno; /* page number */ - - /* dynamic size */ - union { - indx_t mp_ptrs[1]; - MDBX_meta mp_meta; - uint8_t mp_data[1]; - }; -} MDBX_page; - -/* Size of the page header, excluding dynamic data at the end */ -#define PAGEHDRSZ ((unsigned)offsetof(MDBX_page, mp_data)) - -/* The maximum size of a database page. -* -* It is 64K, but value-PAGEHDRSZ must fit in MDBX_page.mp_upper. -* -* MDBX will use database pages < OS pages if needed. -* That causes more I/O in write transactions: The OS must -* know (read) the whole page before writing a partial page. -* -* Note that we don't currently support Huge pages. On Linux, -* regular data files cannot use Huge pages, and in general -* Huge pages aren't actually pageable. We rely on the OS -* demand-pager to read our data and page it out when memory -* pressure from other processes is high. So until OSs have -* actual paging support for Huge pages, they're not viable. */ -#define MAX_PAGESIZE 0x10000u -#define MIN_PAGESIZE 512u - -#define MIN_MAPSIZE (MIN_PAGESIZE * MIN_PAGENO) -#if defined(_WIN32) || defined(_WIN64) -#define MAX_MAPSIZE32 UINT32_C(0x38000000) -#else -#define MAX_MAPSIZE32 UINT32_C(0x7ff80000) -#endif -#define MAX_MAPSIZE64 \ - ((sizeof(pgno_t) > 4) ? UINT64_C(0x7fffFFFFfff80000) \ - : MAX_PAGENO * (uint64_t)MAX_PAGESIZE) - -#define MAX_MAPSIZE ((sizeof(size_t) < 8) ? MAX_MAPSIZE32 : MAX_MAPSIZE64) - -/* The header for the reader table (a memory-mapped lock file). */ -typedef struct MDBX_lockinfo { - /* Stamp identifying this as an MDBX file. - * It must be set to MDBX_MAGIC with with MDBX_LOCK_VERSION. */ - uint64_t mti_magic_and_version; - - /* Format of this lock file. Must be set to MDBX_LOCK_FORMAT. */ - uint32_t mti_os_and_format; - - /* Flags which environment was opened. */ - volatile uint32_t mti_envmode; - - union { -#ifdef MDBX_OSAL_LOCK - MDBX_OSAL_LOCK mti_wmutex; -#endif - uint64_t align_wmutex; - }; - - union { - /* The number of slots that have been used in the reader table. - * This always records the maximum count, it is not decremented - * when readers release their slots. */ - volatile unsigned __cache_aligned mti_numreaders; - uint64_t align_numreaders; - }; - - union { -#ifdef MDBX_OSAL_LOCK - /* Mutex protecting access to this table. */ - MDBX_OSAL_LOCK mti_rmutex; -#endif - uint64_t align_rmutex; - }; - - union { - volatile txnid_t mti_oldest; - uint64_t align_oldest; - }; - - union { - volatile uint32_t mti_readers_refresh_flag; - uint64_t align_reader_finished_flag; - }; - - uint8_t pad_align[MDBX_CACHELINE_SIZE - sizeof(uint64_t) * 7]; - - MDBX_reader __cache_aligned mti_readers[1]; -} MDBX_lockinfo; - -#ifdef _MSC_VER -#pragma pack(pop) -#endif /* MSVC: Enable aligment */ - -#define MDBX_LOCKINFO_WHOLE_SIZE \ - ((sizeof(MDBX_lockinfo) + MDBX_CACHELINE_SIZE - 1) & \ - ~((size_t)MDBX_CACHELINE_SIZE - 1)) - -/* Lockfile format signature: version, features and field layout */ -#define MDBX_LOCK_FORMAT \ - ((MDBX_OSAL_LOCK_SIGN << 16) + \ - (uint16_t)(MDBX_LOCKINFO_WHOLE_SIZE + MDBX_CACHELINE_SIZE - 1)) - -#define MDBX_DATA_MAGIC ((MDBX_MAGIC << 8) + MDBX_DATA_VERSION) - -#define MDBX_LOCK_MAGIC ((MDBX_MAGIC << 8) + MDBX_LOCK_VERSION) - -/*----------------------------------------------------------------------------*/ -/* Two kind lists of pages (aka PNL) */ - -/* An PNL is an Page Number List, a sorted array of IDs. The first element of - * the array is a counter for how many actual page-numbers are in the list. - * PNLs are sorted in descending order, this allow cut off a page with lowest - * pgno (at the tail) just truncating the list */ -#define MDBX_PNL_ASCENDING 0 -typedef pgno_t *MDBX_PNL; - -#if MDBX_PNL_ASCENDING -#define MDBX_PNL_ORDERED(first, last) ((first) < (last)) -#define MDBX_PNL_DISORDERED(first, last) ((first) >= (last)) -#else -#define MDBX_PNL_ORDERED(first, last) ((first) > (last)) -#define MDBX_PNL_DISORDERED(first, last) ((first) <= (last)) -#endif - -/* List of txnid, only for MDBX_env.mt_lifo_reclaimed */ -typedef txnid_t *MDBX_TXL; - -/* An ID2 is an ID/pointer pair. */ -typedef struct MDBX_ID2 { - pgno_t mid; /* The ID */ - void *mptr; /* The pointer */ -} MDBX_ID2; - -/* An ID2L is an ID2 List, a sorted array of ID2s. - * The first element's mid member is a count of how many actual - * elements are in the array. The mptr member of the first element is - * unused. The array is sorted in ascending order by mid. */ -typedef MDBX_ID2 *MDBX_ID2L; - -/* PNL sizes - likely should be even bigger - * limiting factors: sizeof(pgno_t), thread stack size */ -#define MDBX_PNL_LOGN 16 /* DB_SIZE is 2^16, UM_SIZE is 2^17 */ -#define MDBX_PNL_DB_SIZE (1 << MDBX_PNL_LOGN) -#define MDBX_PNL_UM_SIZE (1 << (MDBX_PNL_LOGN + 1)) - -#define MDBX_PNL_DB_MAX (MDBX_PNL_DB_SIZE - 1) -#define MDBX_PNL_UM_MAX (MDBX_PNL_UM_SIZE - 1) - -#define MDBX_PNL_SIZEOF(pl) (((pl)[0] + 1) * sizeof(pgno_t)) -#define MDBX_PNL_IS_ZERO(pl) ((pl)[0] == 0) -#define MDBX_PNL_CPY(dst, src) (memcpy(dst, src, MDBX_PNL_SIZEOF(src))) -#define MDBX_PNL_FIRST(pl) ((pl)[1]) -#define MDBX_PNL_LAST(pl) ((pl)[(pl)[0]]) - -/* Current max length of an mdbx_pnl_alloc()ed PNL */ -#define MDBX_PNL_ALLOCLEN(pl) ((pl)[-1]) - -/*----------------------------------------------------------------------------*/ -/* Internal structures */ - -/* Auxiliary DB info. - * The information here is mostly static/read-only. There is - * only a single copy of this record in the environment. */ -typedef struct MDBX_dbx { - MDBX_val md_name; /* name of the database */ - MDBX_cmp_func *md_cmp; /* function for comparing keys */ - MDBX_cmp_func *md_dcmp; /* function for comparing data items */ -} MDBX_dbx; - -/* A database transaction. - * Every operation requires a transaction handle. */ -struct MDBX_txn { -#define MDBX_MT_SIGNATURE UINT32_C(0x93D53A31) - size_t mt_signature; - MDBX_txn *mt_parent; /* parent of a nested txn */ - /* Nested txn under this txn, set together with flag MDBX_TXN_HAS_CHILD */ - MDBX_txn *mt_child; - pgno_t mt_next_pgno; /* next unallocated page */ - pgno_t mt_end_pgno; /* corresponding to the current size of datafile */ - /* The ID of this transaction. IDs are integers incrementing from 1. - * Only committed write transactions increment the ID. If a transaction - * aborts, the ID may be re-used by the next writer. */ - txnid_t mt_txnid; - MDBX_env *mt_env; /* the DB environment */ - /* The list of reclaimed txns from freeDB */ - MDBX_TXL mt_lifo_reclaimed; - /* The list of pages that became unused during this transaction. */ - MDBX_PNL mt_befree_pages; - /* The list of loose pages that became unused and may be reused - * in this transaction, linked through NEXT_LOOSE_PAGE(page). */ - MDBX_page *mt_loose_pages; - /* Number of loose pages (mt_loose_pages) */ - unsigned mt_loose_count; - /* The sorted list of dirty pages we temporarily wrote to disk - * because the dirty list was full. page numbers in here are - * shifted left by 1, deleted slots have the LSB set. */ - MDBX_PNL mt_spill_pages; - union { - /* For write txns: Modified pages. Sorted when not MDBX_WRITEMAP. */ - MDBX_ID2L mt_rw_dirtylist; - /* For read txns: This thread/txn's reader table slot, or NULL. */ - MDBX_reader *mt_ro_reader; - }; - /* Array of records for each DB known in the environment. */ - MDBX_dbx *mt_dbxs; - /* Array of MDBX_db records for each known DB */ - MDBX_db *mt_dbs; - /* Array of sequence numbers for each DB handle */ - unsigned *mt_dbiseqs; - -/* Transaction DB Flags */ -#define DB_DIRTY MDBX_TBL_DIRTY /* DB was written in this txn */ -#define DB_STALE MDBX_TBL_STALE /* Named-DB record is older than txnID */ -#define DB_NEW MDBX_TBL_NEW /* Named-DB handle opened in this txn */ -#define DB_VALID 0x08 /* DB handle is valid, see also MDBX_VALID */ -#define DB_USRVALID 0x10 /* As DB_VALID, but not set for FREE_DBI */ -#define DB_DUPDATA 0x20 /* DB is MDBX_DUPSORT data */ - /* In write txns, array of cursors for each DB */ - MDBX_cursor **mt_cursors; - /* Array of flags for each DB */ - uint8_t *mt_dbflags; - /* Number of DB records in use, or 0 when the txn is finished. - * This number only ever increments until the txn finishes; we - * don't decrement it when individual DB handles are closed. */ - MDBX_dbi mt_numdbs; - -/* Transaction Flags */ -/* mdbx_txn_begin() flags */ -#define MDBX_TXN_BEGIN_FLAGS \ - (MDBX_NOMETASYNC | MDBX_NOSYNC | MDBX_RDONLY | MDBX_TRYTXN) -#define MDBX_TXN_NOMETASYNC \ - MDBX_NOMETASYNC /* don't sync meta for this txn on commit */ -#define MDBX_TXN_NOSYNC MDBX_NOSYNC /* don't sync this txn on commit */ -#define MDBX_TXN_RDONLY MDBX_RDONLY /* read-only transaction */ - /* internal txn flags */ -#define MDBX_TXN_WRITEMAP MDBX_WRITEMAP /* copy of MDBX_env flag in writers */ -#define MDBX_TXN_FINISHED 0x01 /* txn is finished or never began */ -#define MDBX_TXN_ERROR 0x02 /* txn is unusable after an error */ -#define MDBX_TXN_DIRTY 0x04 /* must write, even if dirty list is empty */ -#define MDBX_TXN_SPILLS 0x08 /* txn or a parent has spilled pages */ -#define MDBX_TXN_HAS_CHILD 0x10 /* txn has an MDBX_txn.mt_child */ -/* most operations on the txn are currently illegal */ -#define MDBX_TXN_BLOCKED \ - (MDBX_TXN_FINISHED | MDBX_TXN_ERROR | MDBX_TXN_HAS_CHILD) - unsigned mt_flags; - /* dirtylist room: Array size - dirty pages visible to this txn. - * Includes ancestor txns' dirty pages not hidden by other txns' - * dirty/spilled pages. Thus commit(nested txn) has room to merge - * dirtylist into mt_parent after freeing hidden mt_parent pages. */ - unsigned mt_dirtyroom; - mdbx_tid_t mt_owner; /* thread ID that owns this transaction */ - mdbx_canary mt_canary; -}; - -/* Enough space for 2^32 nodes with minimum of 2 keys per node. I.e., plenty. - * At 4 keys per node, enough for 2^64 nodes, so there's probably no need to - * raise this on a 64 bit machine. */ -#define CURSOR_STACK 32 - -struct MDBX_xcursor; - -/* Cursors are used for all DB operations. - * A cursor holds a path of (page pointer, key index) from the DB - * root to a position in the DB, plus other state. MDBX_DUPSORT - * cursors include an xcursor to the current data item. Write txns - * track their cursors and keep them up to date when data moves. - * Exception: An xcursor's pointer to a P_SUBP page can be stale. - * (A node with F_DUPDATA but no F_SUBDATA contains a subpage). */ -struct MDBX_cursor { -#define MDBX_MC_SIGNATURE UINT32_C(0xFE05D5B1) -#define MDBX_MC_READY4CLOSE UINT32_C(0x2817A047) -#define MDBX_MC_WAIT4EOT UINT32_C(0x90E297A7) - uint32_t mc_signature; - /* The database handle this cursor operates on */ - MDBX_dbi mc_dbi; - /* Next cursor on this DB in this txn */ - MDBX_cursor *mc_next; - /* Backup of the original cursor if this cursor is a shadow */ - MDBX_cursor *mc_backup; - /* Context used for databases with MDBX_DUPSORT, otherwise NULL */ - struct MDBX_xcursor *mc_xcursor; - /* The transaction that owns this cursor */ - MDBX_txn *mc_txn; - /* The database record for this cursor */ - MDBX_db *mc_db; - /* The database auxiliary record for this cursor */ - MDBX_dbx *mc_dbx; - /* The mt_dbflag for this database */ - uint8_t *mc_dbflag; - uint16_t mc_snum; /* number of pushed pages */ - uint16_t mc_top; /* index of top page, normally mc_snum-1 */ - /* Cursor state flags. */ -#define C_INITIALIZED 0x01 /* cursor has been initialized and is valid */ -#define C_EOF 0x02 /* No more data */ -#define C_SUB 0x04 /* Cursor is a sub-cursor */ -#define C_DEL 0x08 /* last op was a cursor_del */ -#define C_UNTRACK 0x40 /* Un-track cursor when closing */ -#define C_RECLAIMING 0x80 /* FreeDB lookup is prohibited */ - unsigned mc_flags; /* see mdbx_cursor */ - MDBX_page *mc_pg[CURSOR_STACK]; /* stack of pushed pages */ - indx_t mc_ki[CURSOR_STACK]; /* stack of page indices */ -}; - -/* Context for sorted-dup records. - * We could have gone to a fully recursive design, with arbitrarily - * deep nesting of sub-databases. But for now we only handle these - * levels - main DB, optional sub-DB, sorted-duplicate DB. */ -typedef struct MDBX_xcursor { - /* A sub-cursor for traversing the Dup DB */ - MDBX_cursor mx_cursor; - /* The database record for this Dup DB */ - MDBX_db mx_db; - /* The auxiliary DB record for this Dup DB */ - MDBX_dbx mx_dbx; - /* The mt_dbflag for this Dup DB */ - uint8_t mx_dbflag; -} MDBX_xcursor; - -/* Check if there is an inited xcursor, so XCURSOR_REFRESH() is proper */ -#define XCURSOR_INITED(mc) \ - ((mc)->mc_xcursor && ((mc)->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED)) - -/* Update sub-page pointer, if any, in mc->mc_xcursor. - * Needed when the node which contains the sub-page may have moved. - * Called with mp = mc->mc_pg[mc->mc_top], ki = mc->mc_ki[mc->mc_top]. */ -#define XCURSOR_REFRESH(mc, mp, ki) \ - do { \ - MDBX_page *xr_pg = (mp); \ - MDBX_node *xr_node = NODEPTR(xr_pg, ki); \ - if ((xr_node->mn_flags & (F_DUPDATA | F_SUBDATA)) == F_DUPDATA) \ - (mc)->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(xr_node); \ - } while (0) - -/* State of FreeDB old pages, stored in the MDBX_env */ -typedef struct MDBX_pgstate { - pgno_t *mf_reclaimed_pglist; /* Reclaimed freeDB pages, or NULL before use */ - txnid_t mf_last_reclaimed; /* ID of last used record, or 0 if - !mf_reclaimed_pglist */ -} MDBX_pgstate; - -/* The database environment. */ -struct MDBX_env { -#define MDBX_ME_SIGNATURE UINT32_C(0x9A899641) - size_t me_signature; - mdbx_mmap_t me_dxb_mmap; /* The main data file */ -#define me_map me_dxb_mmap.dxb -#define me_fd me_dxb_mmap.fd -#define me_mapsize me_dxb_mmap.length - mdbx_mmap_t me_lck_mmap; /* The lock file */ -#define me_lfd me_lck_mmap.fd -#define me_lck me_lck_mmap.lck - -/* Failed to update the meta page. Probably an I/O error. */ -#define MDBX_FATAL_ERROR UINT32_C(0x80000000) -/* Additional flag for mdbx_sync_locked() */ -#define MDBX_SHRINK_ALLOWED UINT32_C(0x40000000) -/* Some fields are initialized. */ -#define MDBX_ENV_ACTIVE UINT32_C(0x20000000) -/* me_txkey is set */ -#define MDBX_ENV_TXKEY UINT32_C(0x10000000) - uint32_t me_flags; /* see mdbx_env */ - unsigned me_psize; /* DB page size, inited from me_os_psize */ - unsigned me_psize2log; /* log2 of DB page size */ - unsigned me_os_psize; /* OS page size, from mdbx_syspagesize() */ - unsigned me_maxreaders; /* size of the reader table */ - /* Max MDBX_lockinfo.mti_numreaders of interest to mdbx_env_close() */ - unsigned me_close_readers; - mdbx_fastmutex_t me_dbi_lock; - MDBX_dbi me_numdbs; /* number of DBs opened */ - MDBX_dbi me_maxdbs; /* size of the DB table */ - mdbx_pid_t me_pid; /* process ID of this env */ - mdbx_thread_key_t me_txkey; /* thread-key for readers */ - char *me_path; /* path to the DB files */ - void *me_pbuf; /* scratch area for DUPSORT put() */ - MDBX_txn *me_txn; /* current write transaction */ - MDBX_txn *me_txn0; /* prealloc'd write transaction */ - MDBX_dbx *me_dbxs; /* array of static DB info */ - uint16_t *me_dbflags; /* array of flags from MDBX_db.md_flags */ - unsigned *me_dbiseqs; /* array of dbi sequence numbers */ - volatile txnid_t *me_oldest; /* ID of oldest reader last time we looked */ - MDBX_pgstate me_pgstate; /* state of old pages from freeDB */ -#define me_last_reclaimed me_pgstate.mf_last_reclaimed -#define me_reclaimed_pglist me_pgstate.mf_reclaimed_pglist - MDBX_page *me_dpages; /* list of malloc'd blocks for re-use */ - /* PNL of pages that became unused in a write txn */ - MDBX_PNL me_free_pgs; - /* ID2L of pages written during a write txn. Length MDBX_PNL_UM_SIZE. */ - MDBX_ID2L me_dirtylist; - /* Max number of freelist items that can fit in a single overflow page */ - unsigned me_maxfree_1pg; - /* Max size of a node on a page */ - unsigned me_nodemax; - unsigned me_maxkey_limit; /* max size of a key */ - mdbx_pid_t me_live_reader; /* have liveness lock in reader table */ - void *me_userctx; /* User-settable context */ - size_t me_sync_pending; /* Total dirty/non-sync'ed bytes - * since the last mdbx_env_sync() */ - size_t me_sync_threshold; /* Treshold of above to force synchronous flush */ - MDBX_oom_func *me_oom_func; /* Callback for kicking laggard readers */ - txnid_t me_oldest_stub; -#if MDBX_DEBUG - MDBX_assert_func *me_assert_func; /* Callback for assertion failures */ -#endif -#ifdef USE_VALGRIND - int me_valgrind_handle; -#endif - - struct { - size_t lower; /* minimal size of datafile */ - size_t upper; /* maximal size of datafile */ - size_t now; /* current size of datafile */ - size_t grow; /* step to grow datafile */ - size_t shrink; /* threshold to shrink datafile */ - } me_dbgeo; /* */ -}; - -/* Nested transaction */ -typedef struct MDBX_ntxn { - MDBX_txn mnt_txn; /* the transaction */ - MDBX_pgstate mnt_pgstate; /* parent transaction's saved freestate */ -} MDBX_ntxn; - -/*----------------------------------------------------------------------------*/ -/* Debug and Logging stuff */ - -extern int mdbx_runtime_flags; -extern MDBX_debug_func *mdbx_debug_logger; -extern txnid_t mdbx_debug_edge; - -void mdbx_debug_log(int type, const char *function, int line, const char *fmt, - ...) -#if defined(__GNUC__) || __has_attribute(format) - __attribute__((format(printf, 4, 5))) -#endif - ; - -void mdbx_panic(const char *fmt, ...) -#if defined(__GNUC__) || __has_attribute(format) - __attribute__((format(printf, 1, 2))) -#endif - ; - -#if MDBX_DEBUG - -#define mdbx_assert_enabled() unlikely(mdbx_runtime_flags &MDBX_DBG_ASSERT) - -#define mdbx_audit_enabled() unlikely(mdbx_runtime_flags &MDBX_DBG_AUDIT) - -#define mdbx_debug_enabled(type) \ - unlikely(mdbx_runtime_flags &(type & (MDBX_DBG_TRACE | MDBX_DBG_EXTRA))) - -#else -#define mdbx_debug_enabled(type) (0) -#define mdbx_audit_enabled() (0) -#ifndef NDEBUG -#define mdbx_assert_enabled() (1) -#else -#define mdbx_assert_enabled() (0) -#endif /* NDEBUG */ -#endif /* MDBX_DEBUG */ - -#define mdbx_print(fmt, ...) \ - mdbx_debug_log(MDBX_DBG_PRINT, NULL, 0, fmt, ##__VA_ARGS__) - -#define mdbx_trace(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_TRACE)) \ - mdbx_debug_log(MDBX_DBG_TRACE, __FUNCTION__, __LINE__, fmt "\n", \ - ##__VA_ARGS__); \ - } while (0) - -#define mdbx_verbose(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_TRACE /* FIXME */)) \ - mdbx_debug_log(MDBX_DBG_TRACE /* FIXME */, __FUNCTION__, __LINE__, \ - fmt "\n", ##__VA_ARGS__); \ - } while (0) - -#define mdbx_info(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_TRACE /* FIXME */)) \ - mdbx_debug_log(MDBX_DBG_TRACE /* FIXME */, __FUNCTION__, __LINE__, \ - fmt "\n", ##__VA_ARGS__); \ - } while (0) - -#define mdbx_notice(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_TRACE /* FIXME */)) \ - mdbx_debug_log(MDBX_DBG_TRACE /* FIXME */, __FUNCTION__, __LINE__, \ - fmt "\n", ##__VA_ARGS__); \ - } while (0) - -#define mdbx_warning(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_TRACE /* FIXME */)) \ - mdbx_debug_log(MDBX_DBG_TRACE /* FIXME */, __FUNCTION__, __LINE__, \ - fmt "\n", ##__VA_ARGS__); \ - } while (0) - -#define mdbx_error(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_TRACE /* FIXME */)) \ - mdbx_debug_log(MDBX_DBG_TRACE /* FIXME */, __FUNCTION__, __LINE__, \ - fmt "\n", ##__VA_ARGS__); \ - } while (0) - -#define mdbx_fatal(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_TRACE /* FIXME */)) \ - mdbx_debug_log(MDBX_DBG_TRACE /* FIXME */, __FUNCTION__, __LINE__, \ - fmt "\n", ##__VA_ARGS__); \ - } while (0) - -#define mdbx_debug(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_TRACE)) \ - mdbx_debug_log(MDBX_DBG_TRACE, __FUNCTION__, __LINE__, fmt "\n", \ - ##__VA_ARGS__); \ - } while (0) - -#define mdbx_debug_print(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_TRACE)) \ - mdbx_debug_log(MDBX_DBG_TRACE, NULL, 0, fmt, ##__VA_ARGS__); \ - } while (0) - -#define mdbx_debug_extra(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_EXTRA)) \ - mdbx_debug_log(MDBX_DBG_EXTRA, __FUNCTION__, __LINE__, fmt, \ - ##__VA_ARGS__); \ - } while (0) - -#define mdbx_debug_extra_print(fmt, ...) \ - do { \ - if (mdbx_debug_enabled(MDBX_DBG_EXTRA)) \ - mdbx_debug_log(MDBX_DBG_EXTRA, NULL, 0, fmt, ##__VA_ARGS__); \ - } while (0) - -#define mdbx_ensure_msg(env, expr, msg) \ - do { \ - if (unlikely(!(expr))) \ - mdbx_assert_fail(env, msg, __FUNCTION__, __LINE__); \ - } while (0) - -#define mdbx_ensure(env, expr) mdbx_ensure_msg(env, expr, #expr) - -/* assert(3) variant in environment context */ -#define mdbx_assert(env, expr) \ - do { \ - if (mdbx_assert_enabled()) \ - mdbx_ensure(env, expr); \ - } while (0) - -/* assert(3) variant in cursor context */ -#define mdbx_cassert(mc, expr) mdbx_assert((mc)->mc_txn->mt_env, expr) - -/* assert(3) variant in transaction context */ -#define mdbx_tassert(txn, expr) mdbx_assert((txn)->mt_env, expr) - -static __inline void mdbx_jitter4testing(bool tiny) { -#ifndef NDEBUG - if (MDBX_DBG_JITTER & mdbx_runtime_flags) - mdbx_osal_jitter(tiny); -#else - (void)tiny; -#endif -} - -/*----------------------------------------------------------------------------*/ -/* Internal prototypes and inlines */ - -int mdbx_reader_check0(MDBX_env *env, int rlocked, int *dead); -void mdbx_rthc_dtor(void *rthc); -void mdbx_rthc_lock(void); -void mdbx_rthc_unlock(void); -int mdbx_rthc_alloc(mdbx_thread_key_t *key, MDBX_reader *begin, - MDBX_reader *end); -void mdbx_rthc_remove(mdbx_thread_key_t key); -void mdbx_rthc_cleanup(void); - -static __inline bool mdbx_is_power2(size_t x) { return (x & (x - 1)) == 0; } - -static __inline size_t mdbx_roundup2(size_t value, size_t granularity) { - assert(mdbx_is_power2(granularity)); - return (value + granularity - 1) & ~(granularity - 1); -} - -static __inline unsigned mdbx_log2(size_t value) { - assert(mdbx_is_power2(value)); - - unsigned log = 0; - while (value > 1) { - log += 1; - value >>= 1; - } - return log; -} - -#define MDBX_IS_ERROR(rc) \ - ((rc) != MDBX_RESULT_TRUE && (rc) != MDBX_RESULT_FALSE) - -/* Internal error codes, not exposed outside libmdbx */ -#define MDBX_NO_ROOT (MDBX_LAST_ERRCODE + 10) - -/* Debuging output value of a cursor DBI: Negative in a sub-cursor. */ -#define DDBI(mc) \ - (((mc)->mc_flags & C_SUB) ? -(int)(mc)->mc_dbi : (int)(mc)->mc_dbi) - -/* Key size which fits in a DKBUF. */ -#define DKBUF_MAXKEYSIZE 511 /* FIXME */ - -#if MDBX_DEBUG -#define DKBUF char _kbuf[DKBUF_MAXKEYSIZE * 4 + 2] -#define DKEY(x) mdbx_dkey(x, _kbuf, DKBUF_MAXKEYSIZE * 2 + 1) -#define DVAL(x) \ - mdbx_dkey(x, _kbuf + DKBUF_MAXKEYSIZE * 2 + 1, DKBUF_MAXKEYSIZE * 2 + 1) -#else -#define DKBUF ((void)(0)) -#define DKEY(x) ("-") -#define DVAL(x) ("-") -#endif - -/* An invalid page number. - * Mainly used to denote an empty tree. */ -#define P_INVALID (~(pgno_t)0) - -/* Test if the flags f are set in a flag word w. */ -#define F_ISSET(w, f) (((w) & (f)) == (f)) - -/* Round n up to an even number. */ -#define EVEN(n) (((n) + 1U) & -2) /* sign-extending -2 to match n+1U */ - -/* Default size of memory map. - * This is certainly too small for any actual applications. Apps should - * always set the size explicitly using mdbx_env_set_mapsize(). */ -#define DEFAULT_MAPSIZE 1048576 - -/* Number of slots in the reader table. - * This value was chosen somewhat arbitrarily. The 61 is a prime number, - * and such readers plus a couple mutexes fit into single 4KB page. - * Applications should set the table size using mdbx_env_set_maxreaders(). */ -#define DEFAULT_READERS 61 - -/* Address of first usable data byte in a page, after the header */ -#define PAGEDATA(p) ((void *)((char *)(p) + PAGEHDRSZ)) - -/* Number of nodes on a page */ -#define NUMKEYS(p) ((unsigned)(p)->mp_lower >> 1) - -/* The amount of space remaining in the page */ -#define SIZELEFT(p) (indx_t)((p)->mp_upper - (p)->mp_lower) - -/* The percentage of space used in the page, in tenths of a percent. */ -#define PAGEFILL(env, p) \ - (1024UL * ((env)->me_psize - PAGEHDRSZ - SIZELEFT(p)) / \ - ((env)->me_psize - PAGEHDRSZ)) -/* The minimum page fill factor, in tenths of a percent. - * Pages emptier than this are candidates for merging. */ -#define FILL_THRESHOLD 256 - -/* Test if a page is a leaf page */ -#define IS_LEAF(p) F_ISSET((p)->mp_flags, P_LEAF) -/* Test if a page is a LEAF2 page */ -#define IS_LEAF2(p) F_ISSET((p)->mp_flags, P_LEAF2) -/* Test if a page is a branch page */ -#define IS_BRANCH(p) F_ISSET((p)->mp_flags, P_BRANCH) -/* Test if a page is an overflow page */ -#define IS_OVERFLOW(p) unlikely(F_ISSET((p)->mp_flags, P_OVERFLOW)) -/* Test if a page is a sub page */ -#define IS_SUBP(p) F_ISSET((p)->mp_flags, P_SUBP) - -/* The number of overflow pages needed to store the given size. */ -#define OVPAGES(env, size) (bytes2pgno(env, PAGEHDRSZ - 1 + (size)) + 1) - -/* Link in MDBX_txn.mt_loose_pages list. - * Kept outside the page header, which is needed when reusing the page. */ -#define NEXT_LOOSE_PAGE(p) (*(MDBX_page **)((p) + 2)) - -/* Header for a single key/data pair within a page. - * Used in pages of type P_BRANCH and P_LEAF without P_LEAF2. - * We guarantee 2-byte alignment for 'MDBX_node's. - * - * mn_lo and mn_hi are used for data size on leaf nodes, and for child - * pgno on branch nodes. On 64 bit platforms, mn_flags is also used - * for pgno. (Branch nodes have no flags). Lo and hi are in host byte - * order in case some accesses can be optimized to 32-bit word access. - * - * Leaf node flags describe node contents. F_BIGDATA says the node's - * data part is the page number of an overflow page with actual data. - * F_DUPDATA and F_SUBDATA can be combined giving duplicate data in - * a sub-page/sub-database, and named databases (just F_SUBDATA). */ -typedef struct MDBX_node { - union { - struct { -#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ - union { - struct { - uint16_t mn_lo, mn_hi; /* part of data size or pgno */ - }; - uint32_t mn_dsize; - }; - uint16_t mn_flags; /* see mdbx_node */ - uint16_t mn_ksize; /* key size */ -#else - uint16_t mn_ksize; /* key size */ - uint16_t mn_flags; /* see mdbx_node */ - union { - struct { - uint16_t mn_hi, mn_lo; /* part of data size or pgno */ - }; - uint32_t mn_dsize; - }; -#endif - }; - pgno_t mn_ksize_and_pgno; - }; - -/* mdbx_node Flags */ -#define F_BIGDATA 0x01 /* data put on overflow page */ -#define F_SUBDATA 0x02 /* data is a sub-database */ -#define F_DUPDATA 0x04 /* data has duplicates */ - -/* valid flags for mdbx_node_add() */ -#define NODE_ADD_FLAGS (F_DUPDATA | F_SUBDATA | MDBX_RESERVE | MDBX_APPEND) - uint8_t mn_data[1]; /* key and data are appended here */ -} MDBX_node; - -/* Size of the node header, excluding dynamic data at the end */ -#define NODESIZE offsetof(MDBX_node, mn_data) - -/* Bit position of top word in page number, for shifting mn_flags */ -#define PGNO_TOPWORD ((pgno_t)-1 > 0xffffffffu ? 32 : 0) - -/* Size of a node in a branch page with a given key. - * This is just the node header plus the key, there is no data. */ -#define INDXSIZE(k) (NODESIZE + ((k) == NULL ? 0 : (k)->iov_len)) - -/* Size of a node in a leaf page with a given key and data. - * This is node header plus key plus data size. */ -#define LEAFSIZE(k, d) (NODESIZE + (k)->iov_len + (d)->iov_len) - -/* Address of node i in page p */ -static __inline MDBX_node *NODEPTR(MDBX_page *p, unsigned i) { - assert(NUMKEYS(p) > (unsigned)(i)); - return (MDBX_node *)((char *)(p) + (p)->mp_ptrs[i] + PAGEHDRSZ); -} - -/* Address of the key for the node */ -#define NODEKEY(node) (void *)((node)->mn_data) - -/* Address of the data for a node */ -#define NODEDATA(node) (void *)((char *)(node)->mn_data + (node)->mn_ksize) - -/* Get the page number pointed to by a branch node */ -static __inline pgno_t NODEPGNO(const MDBX_node *node) { - pgno_t pgno; - if (UNALIGNED_OK) { - pgno = node->mn_ksize_and_pgno; - if (sizeof(pgno_t) > 4) - pgno &= MAX_PAGENO; - } else { - pgno = node->mn_lo | ((pgno_t)node->mn_hi << 16); - if (sizeof(pgno_t) > 4) - pgno |= ((uint64_t)node->mn_flags) << 32; - } - return pgno; -} - -/* Set the page number in a branch node */ -static __inline void SETPGNO(MDBX_node *node, pgno_t pgno) { - assert(pgno <= MAX_PAGENO); - - if (UNALIGNED_OK) { - if (sizeof(pgno_t) > 4) - pgno |= ((uint64_t)node->mn_ksize) << 48; - node->mn_ksize_and_pgno = pgno; - } else { - node->mn_lo = (uint16_t)pgno; - node->mn_hi = (uint16_t)(pgno >> 16); - if (sizeof(pgno_t) > 4) - node->mn_flags = (uint16_t)((uint64_t)pgno >> 32); - } -} - -/* Get the size of the data in a leaf node */ -static __inline size_t NODEDSZ(const MDBX_node *node) { - size_t size; - if (UNALIGNED_OK) { - size = node->mn_dsize; - } else { - size = node->mn_lo | ((size_t)node->mn_hi << 16); - } - return size; -} - -/* Set the size of the data for a leaf node */ -static __inline void SETDSZ(MDBX_node *node, size_t size) { - assert(size < INT_MAX); - if (UNALIGNED_OK) { - node->mn_dsize = (uint32_t)size; - } else { - node->mn_lo = (uint16_t)size; - node->mn_hi = (uint16_t)(size >> 16); - } -} - -/* The size of a key in a node */ -#define NODEKSZ(node) ((node)->mn_ksize) - -/* The address of a key in a LEAF2 page. - * LEAF2 pages are used for MDBX_DUPFIXED sorted-duplicate sub-DBs. - * There are no node headers, keys are stored contiguously. */ -#define LEAF2KEY(p, i, ks) ((char *)(p) + PAGEHDRSZ + ((i) * (ks))) - -/* Set the node's key into keyptr, if requested. */ -#define MDBX_GET_KEY(node, keyptr) \ - do { \ - if ((keyptr) != NULL) { \ - (keyptr)->iov_len = NODEKSZ(node); \ - (keyptr)->iov_base = NODEKEY(node); \ - } \ - } while (0) - -/* Set the node's key into key. */ -#define MDBX_GET_KEY2(node, key) \ - do { \ - key.iov_len = NODEKSZ(node); \ - key.iov_base = NODEKEY(node); \ - } while (0) - -#define MDBX_VALID 0x8000 /* DB handle is valid, for me_dbflags */ -#define PERSISTENT_FLAGS (0xffff & ~(MDBX_VALID)) -/* mdbx_dbi_open() flags */ -#define VALID_FLAGS \ - (MDBX_REVERSEKEY | MDBX_DUPSORT | MDBX_INTEGERKEY | MDBX_DUPFIXED | \ - MDBX_INTEGERDUP | MDBX_REVERSEDUP | MDBX_CREATE) - -/* max number of pages to commit in one writev() call */ -#define MDBX_COMMIT_PAGES 64 -#if defined(IOV_MAX) && IOV_MAX < MDBX_COMMIT_PAGES /* sysconf(_SC_IOV_MAX) */ -#undef MDBX_COMMIT_PAGES -#define MDBX_COMMIT_PAGES IOV_MAX -#endif - -/* Check txn and dbi arguments to a function */ -#define TXN_DBI_EXIST(txn, dbi, validity) \ - ((dbi) < (txn)->mt_numdbs && ((txn)->mt_dbflags[dbi] & (validity))) - -/* Check for misused dbi handles */ -#define TXN_DBI_CHANGED(txn, dbi) \ - ((txn)->mt_dbiseqs[dbi] != (txn)->mt_env->me_dbiseqs[dbi]) - -/* LY: fast enough on most systems - * - * / - * | -1, a < b - * cmp2int(a,b) = < 0, a == b - * | 1, a > b - * \ - */ -#if 1 -#define mdbx_cmp2int(a, b) (((b) > (a)) ? -1 : (a) > (b)) -#else -#define mdbx_cmp2int(a, b) (((a) > (b)) - ((b) > (a))) -#endif - -static __inline size_t pgno2bytes(const MDBX_env *env, pgno_t pgno) { - mdbx_assert(env, (1u << env->me_psize2log) == env->me_psize); - return ((size_t)pgno) << env->me_psize2log; -} - -static __inline MDBX_page *pgno2page(const MDBX_env *env, pgno_t pgno) { - return (MDBX_page *)(env->me_map + pgno2bytes(env, pgno)); -} - -static __inline pgno_t bytes2pgno(const MDBX_env *env, size_t bytes) { - mdbx_assert(env, (env->me_psize >> env->me_psize2log) == 1); - return (pgno_t)(bytes >> env->me_psize2log); -} - -static __inline pgno_t pgno_add(pgno_t base, pgno_t augend) { - assert(base <= MAX_PAGENO); - return (augend < MAX_PAGENO - base) ? base + augend : MAX_PAGENO; -} - -static __inline pgno_t pgno_sub(pgno_t base, pgno_t subtrahend) { - assert(base >= MIN_PAGENO); - return (subtrahend < base - MIN_PAGENO) ? base - subtrahend : MIN_PAGENO; -} - -static __inline size_t pgno_align2os_bytes(const MDBX_env *env, pgno_t pgno) { - return mdbx_roundup2(pgno2bytes(env, pgno), env->me_os_psize); -} - -static __inline pgno_t pgno_align2os_pgno(const MDBX_env *env, pgno_t pgno) { - return bytes2pgno(env, pgno_align2os_bytes(env, pgno)); -} diff --git a/plugins/Dbx_mdb/src/mdbx/defs.h b/plugins/Dbx_mdb/src/mdbx/defs.h deleted file mode 100644 index 5190fa3a35..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/defs.h +++ /dev/null @@ -1,423 +0,0 @@ -/* - * Copyright 2015-2017 Leonid Yuriev <leo@yuriev.ru> - * and other libmdbx authors: please see AUTHORS file. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted only as authorized by the OpenLDAP - * Public License. - * - * A copy of this license is available in the file LICENSE in the - * top-level directory of the distribution or, alternatively, at - * <http://www.OpenLDAP.org/license.html>. - */ - -#pragma once -/* *INDENT-OFF* */ -/* clang-format off */ - -#ifndef __GNUC_PREREQ -# if defined(__GNUC__) && defined(__GNUC_MINOR__) -# define __GNUC_PREREQ(maj, min) \ - ((__GNUC__ << 16) + __GNUC_MINOR__ >= ((maj) << 16) + (min)) -# else -# define __GNUC_PREREQ(maj, min) (0) -# endif -#endif /* __GNUC_PREREQ */ - -#ifndef __CLANG_PREREQ -# ifdef __clang__ -# define __CLANG_PREREQ(maj,min) \ - ((__clang_major__ << 16) + __clang_minor__ >= ((maj) << 16) + (min)) -# else -# define __CLANG_PREREQ(maj,min) (0) -# endif -#endif /* __CLANG_PREREQ */ - -#ifndef __GLIBC_PREREQ -# if defined(__GLIBC__) && defined(__GLIBC_MINOR__) -# define __GLIBC_PREREQ(maj, min) \ - ((__GLIBC__ << 16) + __GLIBC_MINOR__ >= ((maj) << 16) + (min)) -# else -# define __GLIBC_PREREQ(maj, min) (0) -# endif -#endif /* __GLIBC_PREREQ */ - -#ifndef __has_attribute -# define __has_attribute(x) (0) -#endif - -#ifndef __has_feature -# define __has_feature(x) (0) -#endif - -#ifndef __has_extension -# define __has_extension(x) (0) -#endif - -#ifndef __has_builtin -# define __has_builtin(x) (0) -#endif - -#if __has_feature(thread_sanitizer) -# define __SANITIZE_THREAD__ 1 -#endif - -#if __has_feature(address_sanitizer) -# define __SANITIZE_ADDRESS__ 1 -#endif - -/*----------------------------------------------------------------------------*/ - -#ifndef __extern_C -# ifdef __cplusplus -# define __extern_C extern "C" -# else -# define __extern_C -# endif -#endif /* __extern_C */ - -#ifndef __cplusplus -# ifndef bool -# define bool _Bool -# endif -# ifndef true -# define true (1) -# endif -# ifndef false -# define false (0) -# endif -#endif - -#if !defined(nullptr) && !defined(__cplusplus) || (__cplusplus < 201103L && !defined(_MSC_VER)) -# define nullptr NULL -#endif - -/*----------------------------------------------------------------------------*/ - -#if !defined(__thread) && (defined(_MSC_VER) || defined(__DMC__)) -# define __thread __declspec(thread) -#endif /* __thread */ - -#ifndef __alwaysinline -# if defined(__GNUC__) || __has_attribute(always_inline) -# define __alwaysinline __inline __attribute__((always_inline)) -# elif defined(_MSC_VER) -# define __alwaysinline __forceinline -# else -# define __alwaysinline -# endif -#endif /* __alwaysinline */ - -#ifndef __noinline -# if defined(__GNUC__) || __has_attribute(noinline) -# define __noinline __attribute__((noinline)) -# elif defined(_MSC_VER) -# define __noinline __declspec(noinline) -# elif defined(__SUNPRO_C) || defined(__sun) || defined(sun) -# define __noinline inline -# elif !defined(__INTEL_COMPILER) -# define __noinline /* FIXME ? */ -# endif -#endif /* __noinline */ - -#ifndef __must_check_result -# if defined(__GNUC__) || __has_attribute(warn_unused_result) -# define __must_check_result __attribute__((warn_unused_result)) -# else -# define __must_check_result -# endif -#endif /* __must_check_result */ - -#ifndef __deprecated -# if defined(__GNUC__) || __has_attribute(deprecated) -# define __deprecated __attribute__((deprecated)) -# elif defined(_MSC_VER) -# define __deprecated __declspec(deprecated) -# else -# define __deprecated -# endif -#endif /* __deprecated */ - -#if !defined(__noop) && !defined(_MSC_VER) -# ifdef __cplusplus - static inline void __noop_consume_args() {} - template <typename First, typename... Rest> - static inline void - __noop_consume_args(const First &first, const Rest &... rest) { - (void) first; __noop_consume_args(rest...); - } -# define __noop(...) __noop_consume_args(__VA_ARGS__) -# elif defined(__GNUC__) && (!defined(__STRICT_ANSI__) || !__STRICT_ANSI__) - static __inline void __noop_consume_args(void* anchor, ...) { - (void) anchor; - } -# define __noop(...) __noop_consume_args(0, ##__VA_ARGS__) -# else -# define __noop(...) do {} while(0) -# endif -#endif /* __noop */ - -#ifndef __fallthrough -# if __GNUC_PREREQ(7, 0) || __has_attribute(fallthrough) -# define __fallthrough __attribute__((fallthrough)) -# else -# define __fallthrough __noop() -# endif -#endif /* __fallthrough */ - -#ifndef __unreachable -# if __GNUC_PREREQ(4,5) -# define __unreachable() __builtin_unreachable() -# elif defined(_MSC_VER) -# define __unreachable() __assume(0) -# else -# define __unreachable() __noop() -# endif -#endif /* __unreachable */ - -#ifndef __prefetch -# if defined(__GNUC__) || defined(__clang__) -# define __prefetch(ptr) __builtin_prefetch(ptr) -# else -# define __prefetch(ptr) __noop(ptr) -# endif -#endif /* __prefetch */ - -#ifndef __aligned -# if defined(__GNUC__) || __has_attribute(aligned) -# define __aligned(N) __attribute__((aligned(N))) -# elif defined(_MSC_VER) -# define __aligned(N) __declspec(align(N)) -# else -# define __aligned(N) -# endif -#endif /* __aligned */ - -#ifndef __noreturn -# if defined(__GNUC__) || __has_attribute(noreturn) -# define __noreturn __attribute__((noreturn)) -# elif defined(_MSC_VER) -# define __noreturn __declspec(noreturn) -# else -# define __noreturn -# endif -#endif /* __noreturn */ - -#ifndef __nothrow -# if defined(__GNUC__) || __has_attribute(nothrow) -# define __nothrow __attribute__((nothrow)) -# elif defined(_MSC_VER) && defined(__cplusplus) -# define __nothrow __declspec(nothrow) -# else -# define __nothrow -# endif -#endif /* __nothrow */ - -#ifndef __pure_function - /* Many functions have no effects except the return value and their - * return value depends only on the parameters and/or global variables. - * Such a function can be subject to common subexpression elimination - * and loop optimization just as an arithmetic operator would be. - * These functions should be declared with the attribute pure. */ -# if defined(__GNUC__) || __has_attribute(pure) -# define __pure_function __attribute__((pure)) -# else -# define __pure_function -# endif -#endif /* __pure_function */ - -#ifndef __const_function - /* Many functions do not examine any values except their arguments, - * and have no effects except the return value. Basically this is just - * slightly more strict class than the PURE attribute, since function - * is not allowed to read global memory. - * - * Note that a function that has pointer arguments and examines the - * data pointed to must not be declared const. Likewise, a function - * that calls a non-const function usually must not be const. - * It does not make sense for a const function to return void. */ -# if defined(__GNUC__) || __has_attribute(const) -# define __const_function __attribute__((const)) -# else -# define __const_function -# endif -#endif /* __const_function */ - -#ifndef __dll_hidden -# if defined(__GNUC__) || __has_attribute(visibility) -# define __hidden __attribute__((visibility("hidden"))) -# else -# define __hidden -# endif -#endif /* __dll_hidden */ - -#ifndef __optimize -# if defined(__OPTIMIZE__) -# if defined(__clang__) && !__has_attribute(optimize) -# define __optimize(ops) -# elif defined(__GNUC__) || __has_attribute(optimize) -# define __optimize(ops) __attribute__((optimize(ops))) -# else -# define __optimize(ops) -# endif -# else -# define __optimize(ops) -# endif -#endif /* __optimize */ - -#ifndef __hot -# if defined(__OPTIMIZE__) -# if defined(__clang__) && !__has_attribute(hot) - /* just put frequently used functions in separate section */ -# define __hot __attribute__((section("text.hot"))) __optimize("O3") -# elif defined(__GNUC__) || __has_attribute(hot) -# define __hot __attribute__((hot)) __optimize("O3") -# else -# define __hot __optimize("O3") -# endif -# else -# define __hot -# endif -#endif /* __hot */ - -#ifndef __cold -# if defined(__OPTIMIZE__) -# if defined(__clang__) && !__has_attribute(cold) - /* just put infrequently used functions in separate section */ -# define __cold __attribute__((section("text.unlikely"))) __optimize("Os") -# elif defined(__GNUC__) || __has_attribute(cold) -# define __cold __attribute__((cold)) __optimize("Os") -# else -# define __cold __optimize("Os") -# endif -# else -# define __cold -# endif -#endif /* __cold */ - -#ifndef __flatten -# if defined(__OPTIMIZE__) && (defined(__GNUC__) || __has_attribute(flatten)) -# define __flatten __attribute__((flatten)) -# else -# define __flatten -# endif -#endif /* __flatten */ - -#ifndef likely -# if defined(__GNUC__) || defined(__clang__) -# define likely(cond) __builtin_expect(!!(cond), 1) -# else -# define likely(x) (x) -# endif -#endif /* likely */ - -#ifndef unlikely -# if defined(__GNUC__) || defined(__clang__) -# define unlikely(cond) __builtin_expect(!!(cond), 0) -# else -# define unlikely(x) (x) -# endif -#endif /* unlikely */ - -/* Wrapper around __func__, which is a C99 feature */ -#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L -# define mdbx_func_ __func__ -#elif (defined(__GNUC__) && __GNUC__ >= 2) || defined(__clang__) || defined(_MSC_VER) -# define mdbx_func_ __FUNCTION__ -#else -# define mdbx_func_ "<mdbx_unknown>" -#endif - -/*----------------------------------------------------------------------------*/ - -#if defined(USE_VALGRIND) -# include <valgrind/memcheck.h> -# ifndef VALGRIND_DISABLE_ADDR_ERROR_REPORTING_IN_RANGE - /* LY: available since Valgrind 3.10 */ -# define VALGRIND_DISABLE_ADDR_ERROR_REPORTING_IN_RANGE(a,s) -# define VALGRIND_ENABLE_ADDR_ERROR_REPORTING_IN_RANGE(a,s) -# endif -#else -# define VALGRIND_CREATE_MEMPOOL(h,r,z) -# define VALGRIND_DESTROY_MEMPOOL(h) -# define VALGRIND_MEMPOOL_TRIM(h,a,s) -# define VALGRIND_MEMPOOL_ALLOC(h,a,s) -# define VALGRIND_MEMPOOL_FREE(h,a) -# define VALGRIND_MEMPOOL_CHANGE(h,a,b,s) -# define VALGRIND_MAKE_MEM_NOACCESS(a,s) -# define VALGRIND_MAKE_MEM_DEFINED(a,s) -# define VALGRIND_MAKE_MEM_UNDEFINED(a,s) -# define VALGRIND_DISABLE_ADDR_ERROR_REPORTING_IN_RANGE(a,s) -# define VALGRIND_ENABLE_ADDR_ERROR_REPORTING_IN_RANGE(a,s) -# define VALGRIND_CHECK_MEM_IS_ADDRESSABLE(a,s) (0) -# define VALGRIND_CHECK_MEM_IS_DEFINED(a,s) (0) -# define RUNNING_ON_VALGRIND (0) -#endif /* USE_VALGRIND */ - -#ifdef __SANITIZE_ADDRESS__ -# include <sanitizer/asan_interface.h> -#else -# define ASAN_POISON_MEMORY_REGION(addr, size) \ - ((void)(addr), (void)(size)) -# define ASAN_UNPOISON_MEMORY_REGION(addr, size) \ - ((void)(addr), (void)(size)) -#endif /* __SANITIZE_ADDRESS__ */ - -/*----------------------------------------------------------------------------*/ - -#ifndef ARRAY_LENGTH -# ifdef __cplusplus - template <typename T, size_t N> - char (&__ArraySizeHelper(T (&array)[N]))[N]; -# define ARRAY_LENGTH(array) (sizeof(::__ArraySizeHelper(array))) -# else -# define ARRAY_LENGTH(array) (sizeof(array) / sizeof(array[0])) -# endif -#endif /* ARRAY_LENGTH */ - -#ifndef ARRAY_END -# define ARRAY_END(array) (&array[ARRAY_LENGTH(array)]) -#endif /* ARRAY_END */ - -#ifndef STRINGIFY -# define STRINGIFY_HELPER(x) #x -# define STRINGIFY(x) STRINGIFY_HELPER(x) -#endif /* STRINGIFY */ - -#ifndef offsetof -# define offsetof(type, member) __builtin_offsetof(type, member) -#endif /* offsetof */ - -#ifndef container_of -# define container_of(ptr, type, member) \ - ((type *)((char *)(ptr) - offsetof(type, member))) -#endif /* container_of */ - -#define MDBX_TETRAD(a, b, c, d) \ - ((uint32_t)(a) << 24 | (uint32_t)(b) << 16 | (uint32_t)(c) << 8 | (d)) - -#define MDBX_STRING_TETRAD(str) MDBX_TETRAD(str[0], str[1], str[2], str[3]) - -#define FIXME "FIXME: " __FILE__ ", " STRINGIFY(__LINE__) - -#ifndef STATIC_ASSERT_MSG -# if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 201112L) \ - || __has_feature(c_static_assert) -# define STATIC_ASSERT_MSG(expr, msg) _Static_assert(expr, msg) -# elif defined(static_assert) -# define STATIC_ASSERT_MSG(expr, msg) static_assert(expr, msg) -# elif defined(_MSC_VER) -# include <crtdbg.h> -# define STATIC_ASSERT_MSG(expr, msg) _STATIC_ASSERT(expr) -# else -# define STATIC_ASSERT_MSG(expr, msg) switch (0) {case 0:case (expr):;} -# endif -#endif /* STATIC_ASSERT */ - -#ifndef STATIC_ASSERT -# define STATIC_ASSERT(expr) STATIC_ASSERT_MSG(expr, #expr) -#endif - -/* *INDENT-ON* */ -/* clang-format on */ diff --git a/plugins/Dbx_mdb/src/mdbx/lck-windows.c b/plugins/Dbx_mdb/src/mdbx/lck-windows.c deleted file mode 100644 index 672512a7e1..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/lck-windows.c +++ /dev/null @@ -1,452 +0,0 @@ -/* - * Copyright 2015-2017 Leonid Yuriev <leo@yuriev.ru> - * and other libmdbx authors: please see AUTHORS file. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted only as authorized by the OpenLDAP - * Public License. - * - * A copy of this license is available in the file LICENSE in the - * top-level directory of the distribution or, alternatively, at - * <http://www.OpenLDAP.org/license.html>. - */ - -#include "./bits.h" - -/* PREAMBLE FOR WINDOWS: - * - * We are not concerned for performance here. - * If you are running Windows a performance could NOT be the goal. - * Otherwise please use Linux. - * - * Regards, - * LY - */ - -/*----------------------------------------------------------------------------*/ -/* rthc */ - -static CRITICAL_SECTION rthc_critical_section; - -static void NTAPI tls_callback(PVOID module, DWORD reason, PVOID reserved) { - (void)module; - (void)reserved; - switch (reason) { - case DLL_PROCESS_ATTACH: - InitializeCriticalSection(&rthc_critical_section); - break; - case DLL_PROCESS_DETACH: - DeleteCriticalSection(&rthc_critical_section); - break; - - case DLL_THREAD_ATTACH: - break; - case DLL_THREAD_DETACH: - mdbx_rthc_cleanup(); - break; - } -} - -void mdbx_rthc_lock(void) { EnterCriticalSection(&rthc_critical_section); } - -void mdbx_rthc_unlock(void) { LeaveCriticalSection(&rthc_critical_section); } - -/* *INDENT-OFF* */ -/* clang-format off */ -#if defined(_MSC_VER) -# pragma const_seg(push) -# pragma data_seg(push) - -# ifdef _WIN64 - /* kick a linker to create the TLS directory if not already done */ -# pragma comment(linker, "/INCLUDE:_tls_used") - /* Force some symbol references. */ -# pragma comment(linker, "/INCLUDE:mdbx_tls_callback") - /* specific const-segment for WIN64 */ -# pragma const_seg(".CRT$XLB") - const -# else - /* kick a linker to create the TLS directory if not already done */ -# pragma comment(linker, "/INCLUDE:__tls_used") - /* Force some symbol references. */ -# pragma comment(linker, "/INCLUDE:_mdbx_tls_callback") - /* specific data-segment for WIN32 */ -# pragma data_seg(".CRT$XLB") -# endif - - PIMAGE_TLS_CALLBACK mdbx_tls_callback = tls_callback; -# pragma data_seg(pop) -# pragma const_seg(pop) - -#elif defined(__GNUC__) -# ifdef _WIN64 - const -# endif - PIMAGE_TLS_CALLBACK mdbx_tls_callback __attribute__((section(".CRT$XLB"), used)) - = tls_callback; -#else -# error FIXME -#endif -/* *INDENT-ON* */ -/* clang-format on */ - -/*----------------------------------------------------------------------------*/ - -#define LCK_SHARED 0 -#define LCK_EXCLUSIVE LOCKFILE_EXCLUSIVE_LOCK -#define LCK_WAITFOR 0 -#define LCK_DONTWAIT LOCKFILE_FAIL_IMMEDIATELY - -static __inline BOOL flock(mdbx_filehandle_t fd, DWORD flags, uint64_t offset, - size_t bytes) { - OVERLAPPED ov; - ov.hEvent = 0; - ov.Offset = (DWORD)offset; - ov.OffsetHigh = HIGH_DWORD(offset); - return LockFileEx(fd, flags, 0, (DWORD)bytes, HIGH_DWORD(bytes), &ov); -} - -static __inline BOOL funlock(mdbx_filehandle_t fd, uint64_t offset, - size_t bytes) { - return UnlockFile(fd, (DWORD)offset, HIGH_DWORD(offset), (DWORD)bytes, - HIGH_DWORD(bytes)); -} - -/*----------------------------------------------------------------------------*/ -/* global `write` lock for write-txt processing, - * exclusive locking both meta-pages) */ - -#define LCK_MAXLEN (1u + (size_t)(MAXSSIZE_T)) -#define LCK_META_OFFSET 0 -#define LCK_META_LEN 0x10000u -#define LCK_BODY_OFFSET LCK_META_LEN -#define LCK_BODY_LEN (LCK_MAXLEN - LCK_BODY_OFFSET + 1u) -#define LCK_META LCK_META_OFFSET, LCK_META_LEN -#define LCK_BODY LCK_BODY_OFFSET, LCK_BODY_LEN -#define LCK_WHOLE 0, LCK_MAXLEN - -int mdbx_txn_lock(MDBX_env *env, bool dontwait) { - if (flock(env->me_fd, dontwait ? (LCK_EXCLUSIVE | LCK_DONTWAIT) - : (LCK_EXCLUSIVE | LCK_WAITFOR), - LCK_BODY)) - return MDBX_SUCCESS; - int rc = GetLastError(); - return (!dontwait || rc != ERROR_LOCK_VIOLATION) ? rc : MDBX_BUSY; -} - -void mdbx_txn_unlock(MDBX_env *env) { - if (!funlock(env->me_fd, LCK_BODY)) - mdbx_panic("%s failed: errcode %u", mdbx_func_, GetLastError()); -} - -/*----------------------------------------------------------------------------*/ -/* global `read` lock for readers registration, - * exclusive locking `mti_numreaders` (second) cacheline */ - -#define LCK_LO_OFFSET 0 -#define LCK_LO_LEN offsetof(MDBX_lockinfo, mti_numreaders) -#define LCK_UP_OFFSET LCK_LO_LEN -#define LCK_UP_LEN (MDBX_LOCKINFO_WHOLE_SIZE - LCK_UP_OFFSET) -#define LCK_LOWER LCK_LO_OFFSET, LCK_LO_LEN -#define LCK_UPPER LCK_UP_OFFSET, LCK_UP_LEN - -int mdbx_rdt_lock(MDBX_env *env) { - if (env->me_lfd == INVALID_HANDLE_VALUE) - return MDBX_SUCCESS; /* readonly database in readonly filesystem */ - - /* transite from S-? (used) to S-E (locked), e.g. exclusive lock upper-part */ - if (flock(env->me_lfd, LCK_EXCLUSIVE | LCK_WAITFOR, LCK_UPPER)) - return MDBX_SUCCESS; - return GetLastError(); -} - -void mdbx_rdt_unlock(MDBX_env *env) { - if (env->me_lfd != INVALID_HANDLE_VALUE) { - /* transite from S-E (locked) to S-? (used), e.g. unlock upper-part */ - if (!funlock(env->me_lfd, LCK_UPPER)) - mdbx_panic("%s failed: errcode %u", mdbx_func_, GetLastError()); - } -} - -/*----------------------------------------------------------------------------*/ -/* global `initial` lock for lockfile initialization, - * exclusive/shared locking first cacheline */ - -/* FIXME: locking schema/algo descritpion. - ?-? = free - S-? = used - E-? = exclusive-read - ?-S - ?-E = middle - S-S - S-E = locked - E-S - E-E = exclusive-write -*/ - -int mdbx_lck_init(MDBX_env *env) { - (void)env; - return MDBX_SUCCESS; -} - -/* Seize state as 'exclusive-write' (E-E and returns MDBX_RESULT_TRUE) - * or as 'used' (S-? and returns MDBX_RESULT_FALSE), otherwise returns an error - */ -static int internal_seize_lck(HANDLE lfd) { - int rc; - assert(lfd != INVALID_HANDLE_VALUE); - - /* 1) now on ?-? (free), get ?-E (middle) */ - mdbx_jitter4testing(false); - if (!flock(lfd, LCK_EXCLUSIVE | LCK_WAITFOR, LCK_UPPER)) { - rc = GetLastError() /* 2) something went wrong, give up */; - mdbx_error("%s(%s) failed: errcode %u", mdbx_func_, - "?-?(free) >> ?-E(middle)", rc); - return rc; - } - - /* 3) now on ?-E (middle), try E-E (exclusive-write) */ - mdbx_jitter4testing(false); - if (flock(lfd, LCK_EXCLUSIVE | LCK_DONTWAIT, LCK_LOWER)) - return MDBX_RESULT_TRUE /* 4) got E-E (exclusive-write), done */; - - /* 5) still on ?-E (middle) */ - rc = GetLastError(); - mdbx_jitter4testing(false); - if (rc != ERROR_SHARING_VIOLATION && rc != ERROR_LOCK_VIOLATION) { - /* 6) something went wrong, give up */ - if (!funlock(lfd, LCK_UPPER)) - mdbx_panic("%s(%s) failed: errcode %u", mdbx_func_, - "?-E(middle) >> ?-?(free)", GetLastError()); - return rc; - } - - /* 7) still on ?-E (middle), try S-E (locked) */ - mdbx_jitter4testing(false); - rc = flock(lfd, LCK_SHARED | LCK_DONTWAIT, LCK_LOWER) ? MDBX_RESULT_FALSE - : GetLastError(); - - mdbx_jitter4testing(false); - if (rc != MDBX_RESULT_FALSE) - mdbx_error("%s(%s) failed: errcode %u", mdbx_func_, - "?-E(middle) >> S-E(locked)", rc); - - /* 8) now on S-E (locked) or still on ?-E (middle), - * transite to S-? (used) or ?-? (free) */ - if (!funlock(lfd, LCK_UPPER)) - mdbx_panic("%s(%s) failed: errcode %u", mdbx_func_, - "X-E(locked/middle) >> X-?(used/free)", GetLastError()); - - /* 9) now on S-? (used, DONE) or ?-? (free, FAILURE) */ - return rc; -} - -int mdbx_lck_seize(MDBX_env *env) { - int rc; - - assert(env->me_fd != INVALID_HANDLE_VALUE); - if (env->me_lfd == INVALID_HANDLE_VALUE) { - /* LY: without-lck mode (e.g. on read-only filesystem) */ - mdbx_jitter4testing(false); - if (!flock(env->me_fd, LCK_SHARED | LCK_DONTWAIT, LCK_WHOLE)) { - rc = GetLastError(); - mdbx_error("%s(%s) failed: errcode %u", mdbx_func_, "without-lck", rc); - return rc; - } - return MDBX_RESULT_FALSE; - } - - rc = internal_seize_lck(env->me_lfd); - mdbx_jitter4testing(false); - if (rc == MDBX_RESULT_TRUE && (env->me_flags & MDBX_RDONLY) == 0) { - /* Check that another process don't operates in without-lck mode. - * Doing such check by exclusive locking the body-part of db. Should be - * noted: - * - we need an exclusive lock for do so; - * - we can't lock meta-pages, otherwise other process could get an error - * while opening db in valid (non-conflict) mode. */ - if (!flock(env->me_fd, LCK_EXCLUSIVE | LCK_DONTWAIT, LCK_BODY)) { - rc = GetLastError(); - mdbx_error("%s(%s) failed: errcode %u", mdbx_func_, - "lock-against-without-lck", rc); - mdbx_jitter4testing(false); - mdbx_lck_destroy(env); - } else { - mdbx_jitter4testing(false); - if (!funlock(env->me_fd, LCK_BODY)) - mdbx_panic("%s(%s) failed: errcode %u", mdbx_func_, - "unlock-against-without-lck", GetLastError()); - } - } - - return rc; -} - -int mdbx_lck_downgrade(MDBX_env *env, bool complete) { - /* Transite from exclusive state (E-?) to used (S-?) */ - assert(env->me_fd != INVALID_HANDLE_VALUE); - assert(env->me_lfd != INVALID_HANDLE_VALUE); - - /* 1) must be at E-E (exclusive-write) */ - if (!complete) { - /* transite from E-E to E_? (exclusive-read) */ - if (!funlock(env->me_lfd, LCK_UPPER)) - mdbx_panic("%s(%s) failed: errcode %u", mdbx_func_, - "E-E(exclusive-write) >> E-?(exclusive-read)", GetLastError()); - return MDBX_SUCCESS /* 2) now at E-? (exclusive-read), done */; - } - - /* 3) now at E-E (exclusive-write), transite to ?_E (middle) */ - if (!funlock(env->me_lfd, LCK_LOWER)) - mdbx_panic("%s(%s) failed: errcode %u", mdbx_func_, - "E-E(exclusive-write) >> ?-E(middle)", GetLastError()); - - /* 4) now at ?-E (middle), transite to S-E (locked) */ - if (!flock(env->me_lfd, LCK_SHARED | LCK_DONTWAIT, LCK_LOWER)) { - int rc = GetLastError() /* 5) something went wrong, give up */; - mdbx_error("%s(%s) failed: errcode %u", mdbx_func_, - "?-E(middle) >> S-E(locked)", rc); - return rc; - } - - /* 6) got S-E (locked), continue transition to S-? (used) */ - if (!funlock(env->me_lfd, LCK_UPPER)) - mdbx_panic("%s(%s) failed: errcode %u", mdbx_func_, - "S-E(locked) >> S-?(used)", GetLastError()); - - return MDBX_SUCCESS /* 7) now at S-? (used), done */; -} - -int mdbx_lck_upgrade(MDBX_env *env) { - /* Transite from locked state (S-E) to exclusive-write (E-E) */ - assert(env->me_fd != INVALID_HANDLE_VALUE); - assert(env->me_lfd != INVALID_HANDLE_VALUE); - - /* 1) must be at S-E (locked), transite to ?_E (middle) */ - if (!funlock(env->me_lfd, LCK_LOWER)) - mdbx_panic("%s(%s) failed: errcode %u", mdbx_func_, - "S-E(locked) >> ?-E(middle)", GetLastError()); - - /* 3) now on ?-E (middle), try E-E (exclusive-write) */ - mdbx_jitter4testing(false); - if (flock(env->me_lfd, LCK_EXCLUSIVE | LCK_DONTWAIT, LCK_LOWER)) - return MDBX_RESULT_TRUE; /* 4) got E-E (exclusive-write), done */ - - /* 5) still on ?-E (middle) */ - int rc = GetLastError(); - mdbx_jitter4testing(false); - if (rc != ERROR_SHARING_VIOLATION && rc != ERROR_LOCK_VIOLATION) { - /* 6) something went wrong, report but continue */ - mdbx_error("%s(%s) failed: errcode %u", mdbx_func_, - "?-E(middle) >> E-E(exclusive-write)", rc); - } - - /* 7) still on ?-E (middle), try restore S-E (locked) */ - mdbx_jitter4testing(false); - rc = flock(env->me_lfd, LCK_SHARED | LCK_DONTWAIT, LCK_LOWER) - ? MDBX_RESULT_FALSE - : GetLastError(); - - mdbx_jitter4testing(false); - if (rc != MDBX_RESULT_FALSE) { - mdbx_fatal("%s(%s) failed: errcode %u", mdbx_func_, - "?-E(middle) >> S-E(locked)", rc); - return rc; - } - - /* 8) now on S-E (locked) */ - return MDBX_RESULT_FALSE; -} - -void mdbx_lck_destroy(MDBX_env *env) { - int rc; - - if (env->me_lfd != INVALID_HANDLE_VALUE) { - /* double `unlock` for robustly remove overlapped shared/exclusive locks */ - while (funlock(env->me_lfd, LCK_LOWER)) - ; - rc = GetLastError(); - assert(rc == ERROR_NOT_LOCKED); - (void)rc; - SetLastError(ERROR_SUCCESS); - - while (funlock(env->me_lfd, LCK_UPPER)) - ; - rc = GetLastError(); - assert(rc == ERROR_NOT_LOCKED); - (void)rc; - SetLastError(ERROR_SUCCESS); - } - - if (env->me_fd != INVALID_HANDLE_VALUE) { - /* explicitly unlock to avoid latency for other processes (windows kernel - * releases such locks via deferred queues) */ - while (funlock(env->me_fd, LCK_BODY)) - ; - rc = GetLastError(); - assert(rc == ERROR_NOT_LOCKED); - (void)rc; - SetLastError(ERROR_SUCCESS); - - while (funlock(env->me_fd, LCK_META)) - ; - rc = GetLastError(); - assert(rc == ERROR_NOT_LOCKED); - (void)rc; - SetLastError(ERROR_SUCCESS); - - while (funlock(env->me_fd, LCK_WHOLE)) - ; - rc = GetLastError(); - assert(rc == ERROR_NOT_LOCKED); - (void)rc; - SetLastError(ERROR_SUCCESS); - } -} - -/*----------------------------------------------------------------------------*/ -/* reader checking (by pid) */ - -int mdbx_rpid_set(MDBX_env *env) { - (void)env; - return MDBX_SUCCESS; -} - -int mdbx_rpid_clear(MDBX_env *env) { - (void)env; - return MDBX_SUCCESS; -} - -/* Checks reader by pid. - * - * Returns: - * MDBX_RESULT_TRUE, if pid is live (unable to acquire lock) - * MDBX_RESULT_FALSE, if pid is dead (lock acquired) - * or otherwise the errcode. */ -int mdbx_rpid_check(MDBX_env *env, mdbx_pid_t pid) { - (void)env; - HANDLE hProcess = OpenProcess(PROCESS_QUERY_LIMITED_INFORMATION, FALSE, pid); - int rc; - if (hProcess) { - rc = WaitForSingleObject(hProcess, 0); - CloseHandle(hProcess); - } else { - rc = GetLastError(); - } - - switch (rc) { - case ERROR_INVALID_PARAMETER: - /* pid seem invalid */ - return MDBX_RESULT_FALSE; - case WAIT_OBJECT_0: - /* process just exited */ - return MDBX_RESULT_FALSE; - case WAIT_TIMEOUT: - /* pid running */ - return MDBX_RESULT_TRUE; - default: - /* failure */ - return rc; - } -} diff --git a/plugins/Dbx_mdb/src/mdbx/mdbx.c b/plugins/Dbx_mdb/src/mdbx/mdbx.c deleted file mode 100644 index d55484b802..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/mdbx.c +++ /dev/null @@ -1,11613 +0,0 @@ -/* - * Copyright 2015-2017 Leonid Yuriev <leo@yuriev.ru> - * and other libmdbx authors: please see AUTHORS file. - * All rights reserved. - * - * This code is derived from "LMDB engine" written by - * Howard Chu (Symas Corporation), which itself derived from btree.c - * written by Martin Hedenfalk. - * - * --- - * - * Portions Copyright 2011-2015 Howard Chu, Symas Corp. All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted only as authorized by the OpenLDAP - * Public License. - * - * A copy of this license is available in the file LICENSE in the - * top-level directory of the distribution or, alternatively, at - * <http://www.OpenLDAP.org/license.html>. - * - * --- - * - * Portions Copyright (c) 2009, 2010 Martin Hedenfalk <martin@bzero.se> - * - * Permission to use, copy, modify, and distribute this software for any - * purpose with or without fee is hereby granted, provided that the above - * copyright notice and this permission notice appear in all copies. - * - * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES - * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF - * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR - * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES - * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN - * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF - * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ - -#include "./bits.h" - -/*----------------------------------------------------------------------------*/ -/* rthc (tls keys and destructors) */ - -typedef struct rthc_entry_t { - MDBX_reader *begin; - MDBX_reader *end; - mdbx_thread_key_t key; -} rthc_entry_t; - -#if MDBX_DEBUG -#define RTHC_INITIAL_LIMIT 1 -#else -#define RTHC_INITIAL_LIMIT 16 -#endif - -static unsigned rthc_count; -static unsigned rthc_limit = RTHC_INITIAL_LIMIT; -static rthc_entry_t rthc_table_static[RTHC_INITIAL_LIMIT]; -static rthc_entry_t *rthc_table = rthc_table_static; - -__cold void mdbx_rthc_dtor(void *ptr) { - MDBX_reader *rthc = (MDBX_reader *)ptr; - - mdbx_rthc_lock(); - const mdbx_pid_t self_pid = mdbx_getpid(); - for (unsigned i = 0; i < rthc_count; ++i) { - if (rthc >= rthc_table[i].begin && rthc < rthc_table[i].end) { - if (rthc->mr_pid == self_pid) { - rthc->mr_pid = 0; - mdbx_coherent_barrier(); - } - break; - } - } - mdbx_rthc_unlock(); -} - -__cold void mdbx_rthc_cleanup(void) { - mdbx_rthc_lock(); - const mdbx_pid_t self_pid = mdbx_getpid(); - for (unsigned i = 0; i < rthc_count; ++i) { - mdbx_thread_key_t key = rthc_table[i].key; - MDBX_reader *rthc = mdbx_thread_rthc_get(key); - if (rthc) { - mdbx_thread_rthc_set(key, NULL); - if (rthc->mr_pid == self_pid) { - rthc->mr_pid = 0; - mdbx_coherent_barrier(); - } - } - } - mdbx_rthc_unlock(); -} - -__cold int mdbx_rthc_alloc(mdbx_thread_key_t *key, MDBX_reader *begin, - MDBX_reader *end) { -#ifndef NDEBUG - *key = (mdbx_thread_key_t)0xBADBADBAD; -#endif /* NDEBUG */ - int rc = mdbx_thread_key_create(key); - if (rc != MDBX_SUCCESS) - return rc; - - mdbx_rthc_lock(); - if (rthc_count == rthc_limit) { - rthc_entry_t *new_table = - realloc((rthc_table == rthc_table_static) ? NULL : rthc_table, - sizeof(rthc_entry_t) * rthc_limit * 2); - if (new_table == NULL) { - rc = MDBX_ENOMEM; - goto bailout; - } - if (rthc_table == rthc_table_static) - memcpy(new_table, rthc_table_static, sizeof(rthc_table_static)); - rthc_table = new_table; - rthc_limit *= 2; - } - - rthc_table[rthc_count].key = *key; - rthc_table[rthc_count].begin = begin; - rthc_table[rthc_count].end = end; - ++rthc_count; - mdbx_rthc_unlock(); - return MDBX_SUCCESS; - -bailout: - mdbx_thread_key_delete(*key); - mdbx_rthc_unlock(); - return rc; -} - -__cold void mdbx_rthc_remove(mdbx_thread_key_t key) { - mdbx_rthc_lock(); - mdbx_thread_key_delete(key); - - for (unsigned i = 0; i < rthc_count; ++i) { - if (key == rthc_table[i].key) { - const mdbx_pid_t self_pid = mdbx_getpid(); - for (MDBX_reader *rthc = rthc_table[i].begin; rthc < rthc_table[i].end; - ++rthc) - if (rthc->mr_pid == self_pid) - rthc->mr_pid = 0; - mdbx_coherent_barrier(); - if (--rthc_count > 0) - rthc_table[i] = rthc_table[rthc_count]; - else if (rthc_table != rthc_table_static) { - free(rthc_table); - rthc_table = rthc_table_static; - rthc_limit = RTHC_INITIAL_LIMIT; - } - break; - } - } - - mdbx_rthc_unlock(); -} - -/*----------------------------------------------------------------------------*/ - -/* Allocate an PNL. - * Allocates memory for an PNL of the given size. - * Returns PNL on success, NULL on failure. */ -static MDBX_PNL mdbx_pnl_alloc(size_t size) { - MDBX_PNL pl = malloc((size + 2) * sizeof(pgno_t)); - if (likely(pl)) { - *pl++ = (pgno_t)size; - *pl = 0; - } - return pl; -} - -static MDBX_TXL mdbx_txl_alloc(void) { - const size_t malloc_overhead = sizeof(void *) * 2; - const size_t bytes = mdbx_roundup2(malloc_overhead + sizeof(txnid_t) * 61, - MDBX_CACHELINE_SIZE) - - malloc_overhead; - MDBX_TXL ptr = malloc(bytes); - if (likely(ptr)) { - *ptr++ = bytes / sizeof(txnid_t) - 2; - *ptr = 0; - } - return ptr; -} - -/* Free an PNL. - * [in] pl The PNL to free. */ -static void mdbx_pnl_free(MDBX_PNL pl) { - if (likely(pl)) - free(pl - 1); -} - -static void mdbx_txl_free(MDBX_TXL list) { - if (likely(list)) - free(list - 1); -} - -/* Append ID to PNL. The PNL must be big enough. */ -static __inline void mdbx_pnl_xappend(MDBX_PNL pl, pgno_t id) { - assert(pl[0] + (size_t)1 < MDBX_PNL_ALLOCLEN(pl)); - pl[pl[0] += 1] = id; -} - -static bool mdbx_pnl_check(MDBX_PNL pl) { - if (pl) { - for (const pgno_t *ptr = pl + pl[0]; --ptr > pl;) { - assert(MDBX_PNL_ORDERED(ptr[0], ptr[1])); - assert(ptr[0] >= NUM_METAS); - if (unlikely(MDBX_PNL_DISORDERED(ptr[0], ptr[1]) || ptr[0] < NUM_METAS)) - return false; - } - } - return true; -} - -/* Sort an PNL. - * [in,out] pnl The PNL to sort. */ -static void __hot mdbx_pnl_sort(MDBX_PNL pnl) { - /* Max possible depth of int-indexed tree * 2 items/level */ - int istack[sizeof(int) * CHAR_BIT * 2]; - int i, j, k, l, ir, jstack; - pgno_t a; - -/* Quicksort + Insertion sort for small arrays */ -#define PNL_SMALL 8 -#define PNL_SWAP(a, b) \ - do { \ - pgno_t tmp_pgno = (a); \ - (a) = (b); \ - (b) = tmp_pgno; \ - } while (0) - - ir = (int)pnl[0]; - l = 1; - jstack = 0; - while (1) { - if (ir - l < PNL_SMALL) { /* Insertion sort */ - for (j = l + 1; j <= ir; j++) { - a = pnl[j]; - for (i = j - 1; i >= 1; i--) { - if (MDBX_PNL_DISORDERED(a, pnl[i])) - break; - pnl[i + 1] = pnl[i]; - } - pnl[i + 1] = a; - } - if (jstack == 0) - break; - ir = istack[jstack--]; - l = istack[jstack--]; - } else { - k = (l + ir) >> 1; /* Choose median of left, center, right */ - PNL_SWAP(pnl[k], pnl[l + 1]); - if (MDBX_PNL_ORDERED(pnl[ir], pnl[l])) - PNL_SWAP(pnl[l], pnl[ir]); - - if (MDBX_PNL_ORDERED(pnl[ir], pnl[l + 1])) - PNL_SWAP(pnl[l + 1], pnl[ir]); - - if (MDBX_PNL_ORDERED(pnl[l + 1], pnl[l])) - PNL_SWAP(pnl[l], pnl[l + 1]); - - i = l + 1; - j = ir; - a = pnl[l + 1]; - while (1) { - do - i++; - while (MDBX_PNL_ORDERED(pnl[i], a)); - do - j--; - while (MDBX_PNL_ORDERED(a, pnl[j])); - if (j < i) - break; - PNL_SWAP(pnl[i], pnl[j]); - } - pnl[l + 1] = pnl[j]; - pnl[j] = a; - jstack += 2; - if (ir - i + 1 >= j - l) { - istack[jstack] = ir; - istack[jstack - 1] = i; - ir = j - 1; - } else { - istack[jstack] = j - 1; - istack[jstack - 1] = l; - l = i; - } - } - } -#undef PNL_SMALL -#undef PNL_SWAP - assert(mdbx_pnl_check(pnl)); -} - -/* Search for an ID in an PNL. - * [in] pl The PNL to search. - * [in] id The ID to search for. - * Returns The index of the first ID greater than or equal to id. */ -static unsigned __hot mdbx_pnl_search(MDBX_PNL pnl, pgno_t id) { - assert(mdbx_pnl_check(pnl)); - - /* binary search of id in pl - * if found, returns position of id - * if not found, returns first position greater than id */ - unsigned base = 0; - unsigned cursor = 1; - int val = 0; - unsigned n = pnl[0]; - - while (n > 0) { - unsigned pivot = n >> 1; - cursor = base + pivot + 1; - val = MDBX_PNL_ASCENDING ? mdbx_cmp2int(pnl[cursor], id) - : mdbx_cmp2int(id, pnl[cursor]); - - if (val < 0) { - n = pivot; - } else if (val > 0) { - base = cursor; - n -= pivot + 1; - } else { - return cursor; - } - } - - if (val > 0) - ++cursor; - - return cursor; -} - -/* Shrink an PNL. - * Return the PNL to the default size if it has grown larger. - * [in,out] ppl Address of the PNL to shrink. */ -static void mdbx_pnl_shrink(MDBX_PNL *ppl) { - MDBX_PNL pl = *ppl - 1; - if (unlikely(*pl > MDBX_PNL_UM_MAX)) { - /* shrink to MDBX_PNL_UM_MAX */ - pl = realloc(pl, (MDBX_PNL_UM_MAX + 2) * sizeof(pgno_t)); - if (likely(pl)) { - *pl++ = MDBX_PNL_UM_MAX; - *ppl = pl; - } - } -} - -/* Grow an PNL. - * Return the PNL to the size growed by given number. - * [in,out] ppl Address of the PNL to grow. */ -static int mdbx_pnl_grow(MDBX_PNL *ppl, size_t num) { - MDBX_PNL idn = *ppl - 1; - /* grow it */ - idn = realloc(idn, (*idn + num + 2) * sizeof(pgno_t)); - if (unlikely(!idn)) - return MDBX_ENOMEM; - *idn++ += (pgno_t)num; - *ppl = idn; - return 0; -} - -static int mdbx_txl_grow(MDBX_TXL *ptr, size_t num) { - MDBX_TXL list = *ptr - 1; - /* grow it */ - list = realloc(list, ((size_t)*list + num + 2) * sizeof(txnid_t)); - if (unlikely(!list)) - return MDBX_ENOMEM; - *list++ += num; - *ptr = list; - return 0; -} - -/* Make room for num additional elements in an PNL. - * [in,out] ppl Address of the PNL. - * [in] num Number of elements to make room for. - * Returns 0 on success, MDBX_ENOMEM on failure. */ -static int mdbx_pnl_need(MDBX_PNL *ppl, size_t num) { - MDBX_PNL pl = *ppl; - num += pl[0]; - if (unlikely(num > pl[-1])) { - num = (num + num / 4 + (256 + 2)) & -256; - pl = realloc(pl - 1, num * sizeof(pgno_t)); - if (unlikely(!pl)) - return MDBX_ENOMEM; - *pl++ = (pgno_t)num - 2; - *ppl = pl; - } - return 0; -} - -/* Append an ID onto an PNL. - * [in,out] ppl Address of the PNL to append to. - * [in] id The ID to append. - * Returns 0 on success, MDBX_ENOMEM if the PNL is too large. */ -static int mdbx_pnl_append(MDBX_PNL *ppl, pgno_t id) { - MDBX_PNL pl = *ppl; - /* Too big? */ - if (unlikely(pl[0] >= pl[-1])) { - if (mdbx_pnl_grow(ppl, MDBX_PNL_UM_MAX)) - return MDBX_ENOMEM; - pl = *ppl; - } - pl[0]++; - pl[pl[0]] = id; - return 0; -} - -static int mdbx_txl_append(MDBX_TXL *ptr, txnid_t id) { - MDBX_TXL list = *ptr; - /* Too big? */ - if (unlikely(list[0] >= list[-1])) { - if (mdbx_txl_grow(ptr, (size_t)list[0])) - return MDBX_ENOMEM; - list = *ptr; - } - list[0]++; - list[list[0]] = id; - return 0; -} - -/* Append an PNL onto an PNL. - * [in,out] ppl Address of the PNL to append to. - * [in] app The PNL to append. - * Returns 0 on success, MDBX_ENOMEM if the PNL is too large. */ -static int mdbx_pnl_append_list(MDBX_PNL *ppl, MDBX_PNL app) { - MDBX_PNL pnl = *ppl; - /* Too big? */ - if (unlikely(pnl[0] + app[0] >= pnl[-1])) { - if (mdbx_pnl_grow(ppl, app[0])) - return MDBX_ENOMEM; - pnl = *ppl; - } - memcpy(&pnl[pnl[0] + 1], &app[1], app[0] * sizeof(pgno_t)); - pnl[0] += app[0]; - return 0; -} - -static int mdbx_txl_append_list(MDBX_TXL *ptr, MDBX_TXL append) { - MDBX_TXL list = *ptr; - /* Too big? */ - if (unlikely(list[0] + append[0] >= list[-1])) { - if (mdbx_txl_grow(ptr, (size_t)append[0])) - return MDBX_ENOMEM; - list = *ptr; - } - memcpy(&list[list[0] + 1], &append[1], (size_t)append[0] * sizeof(txnid_t)); - list[0] += append[0]; - return 0; -} - -/* Append an ID range onto an PNL. - * [in,out] ppl Address of the PNL to append to. - * [in] id The lowest ID to append. - * [in] n Number of IDs to append. - * Returns 0 on success, MDBX_ENOMEM if the PNL is too large. */ -static int mdbx_pnl_append_range(MDBX_PNL *ppl, pgno_t id, size_t n) { - pgno_t *pnl = *ppl, len = pnl[0]; - /* Too big? */ - if (unlikely(len + n > pnl[-1])) { - if (mdbx_pnl_grow(ppl, n | MDBX_PNL_UM_MAX)) - return MDBX_ENOMEM; - pnl = *ppl; - } - pnl[0] = len + (pgno_t)n; - pnl += len; - while (n) - pnl[n--] = id++; - return 0; -} - -/* Merge an PNL onto an PNL. The destination PNL must be big enough. - * [in] pl The PNL to merge into. - * [in] merge The PNL to merge. */ -static void __hot mdbx_pnl_xmerge(MDBX_PNL pnl, MDBX_PNL merge) { - assert(mdbx_pnl_check(pnl)); - assert(mdbx_pnl_check(merge)); - pgno_t old_id, merge_id, i = merge[0], j = pnl[0], k = i + j, total = k; - pnl[0] = - MDBX_PNL_ASCENDING ? 0 : ~(pgno_t)0; /* delimiter for pl scan below */ - old_id = pnl[j]; - while (i) { - merge_id = merge[i--]; - for (; MDBX_PNL_ORDERED(merge_id, old_id); old_id = pnl[--j]) - pnl[k--] = old_id; - pnl[k--] = merge_id; - } - pnl[0] = total; - assert(mdbx_pnl_check(pnl)); -} - -/* Search for an ID in an ID2L. - * [in] pnl The ID2L to search. - * [in] id The ID to search for. - * Returns The index of the first ID2 whose mid member is greater than - * or equal to id. */ -static unsigned __hot mdbx_mid2l_search(MDBX_ID2L pnl, pgno_t id) { - /* binary search of id in pnl - * if found, returns position of id - * if not found, returns first position greater than id */ - unsigned base = 0; - unsigned cursor = 1; - int val = 0; - unsigned n = (unsigned)pnl[0].mid; - -#if MDBX_DEBUG - for (const MDBX_ID2 *ptr = pnl + pnl[0].mid; --ptr > pnl;) { - assert(ptr[0].mid < ptr[1].mid); - assert(ptr[0].mid >= NUM_METAS); - } -#endif - - while (n > 0) { - unsigned pivot = n >> 1; - cursor = base + pivot + 1; - val = mdbx_cmp2int(id, pnl[cursor].mid); - - if (val < 0) { - n = pivot; - } else if (val > 0) { - base = cursor; - n -= pivot + 1; - } else { - return cursor; - } - } - - if (val > 0) - ++cursor; - - return cursor; -} - -/* Insert an ID2 into a ID2L. - * [in,out] pnl The ID2L to insert into. - * [in] id The ID2 to insert. - * Returns 0 on success, -1 if the ID was already present in the ID2L. */ -static int mdbx_mid2l_insert(MDBX_ID2L pnl, MDBX_ID2 *id) { - unsigned x = mdbx_mid2l_search(pnl, id->mid); - if (unlikely(x < 1)) - return /* internal error */ -2; - - if (x <= pnl[0].mid && pnl[x].mid == id->mid) - return /* duplicate */ -1; - - if (unlikely(pnl[0].mid >= MDBX_PNL_UM_MAX)) - return /* too big */ -2; - - /* insert id */ - pnl[0].mid++; - for (unsigned i = (unsigned)pnl[0].mid; i > x; i--) - pnl[i] = pnl[i - 1]; - pnl[x] = *id; - return 0; -} - -/* Append an ID2 into a ID2L. - * [in,out] pnl The ID2L to append into. - * [in] id The ID2 to append. - * Returns 0 on success, -2 if the ID2L is too big. */ -static int mdbx_mid2l_append(MDBX_ID2L pnl, MDBX_ID2 *id) { -#if MDBX_DEBUG - for (unsigned i = pnl[0].mid; i > 0; --i) { - assert(pnl[i].mid != id->mid); - if (unlikely(pnl[i].mid == id->mid)) - return -1; - } -#endif - - /* Too big? */ - if (unlikely(pnl[0].mid >= MDBX_PNL_UM_MAX)) - return -2; - - pnl[0].mid++; - pnl[pnl[0].mid] = *id; - return 0; -} - -/*----------------------------------------------------------------------------*/ - -int mdbx_runtime_flags = MDBX_DBG_PRINT -#if MDBX_DEBUG - | MDBX_DBG_ASSERT -#endif -#if MDBX_DEBUG > 1 - | MDBX_DBG_TRACE -#endif -#if MDBX_DEBUG > 2 - | MDBX_DBG_AUDIT -#endif -#if MDBX_DEBUG > 3 - | MDBX_DBG_EXTRA -#endif - ; - -MDBX_debug_func *mdbx_debug_logger; - -static int mdbx_page_alloc(MDBX_cursor *mc, unsigned num, MDBX_page **mp, - int flags); -static int mdbx_page_new(MDBX_cursor *mc, uint32_t flags, unsigned num, - MDBX_page **mp); -static int mdbx_page_touch(MDBX_cursor *mc); -static int mdbx_cursor_touch(MDBX_cursor *mc); - -#define MDBX_END_NAMES \ - { \ - "committed", "empty-commit", "abort", "reset", "reset-tmp", "fail-begin", \ - "fail-beginchild" \ - } -enum { - /* mdbx_txn_end operation number, for logging */ - MDBX_END_COMMITTED, - MDBX_END_EMPTY_COMMIT, - MDBX_END_ABORT, - MDBX_END_RESET, - MDBX_END_RESET_TMP, - MDBX_END_FAIL_BEGIN, - MDBX_END_FAIL_BEGINCHILD -}; -#define MDBX_END_OPMASK 0x0F /* mask for mdbx_txn_end() operation number */ -#define MDBX_END_UPDATE 0x10 /* update env state (DBIs) */ -#define MDBX_END_FREE 0x20 /* free txn unless it is MDBX_env.me_txn0 */ -#define MDBX_END_EOTDONE 0x40 /* txn's cursors already closed */ -#define MDBX_END_SLOT 0x80 /* release any reader slot if MDBX_NOTLS */ -static int mdbx_txn_end(MDBX_txn *txn, unsigned mode); - -static int mdbx_page_get(MDBX_cursor *mc, pgno_t pgno, MDBX_page **mp, - int *lvl); -static int mdbx_page_search_root(MDBX_cursor *mc, MDBX_val *key, int modify); - -#define MDBX_PS_MODIFY 1 -#define MDBX_PS_ROOTONLY 2 -#define MDBX_PS_FIRST 4 -#define MDBX_PS_LAST 8 -static int mdbx_page_search(MDBX_cursor *mc, MDBX_val *key, int flags); -static int mdbx_page_merge(MDBX_cursor *csrc, MDBX_cursor *cdst); - -#define MDBX_SPLIT_REPLACE MDBX_APPENDDUP /* newkey is not new */ -static int mdbx_page_split(MDBX_cursor *mc, MDBX_val *newkey, MDBX_val *newdata, - pgno_t newpgno, unsigned nflags); - -static int mdbx_read_header(MDBX_env *env, MDBX_meta *meta); -static int mdbx_sync_locked(MDBX_env *env, unsigned flags, - MDBX_meta *const pending); -static void mdbx_env_close0(MDBX_env *env); - -static MDBX_node *mdbx_node_search(MDBX_cursor *mc, MDBX_val *key, int *exactp); -static int mdbx_node_add(MDBX_cursor *mc, unsigned indx, MDBX_val *key, - MDBX_val *data, pgno_t pgno, unsigned flags); -static void mdbx_node_del(MDBX_cursor *mc, size_t ksize); -static void mdbx_node_shrink(MDBX_page *mp, unsigned indx); -static int mdbx_node_move(MDBX_cursor *csrc, MDBX_cursor *cdst, int fromleft); -static int mdbx_node_read(MDBX_cursor *mc, MDBX_node *leaf, MDBX_val *data); -static size_t mdbx_leaf_size(MDBX_env *env, MDBX_val *key, MDBX_val *data); -static size_t mdbx_branch_size(MDBX_env *env, MDBX_val *key); - -static int mdbx_rebalance(MDBX_cursor *mc); -static int mdbx_update_key(MDBX_cursor *mc, MDBX_val *key); - -static void mdbx_cursor_pop(MDBX_cursor *mc); -static int mdbx_cursor_push(MDBX_cursor *mc, MDBX_page *mp); - -static int mdbx_cursor_del0(MDBX_cursor *mc); -static int mdbx_del0(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *data, - unsigned flags); -static int mdbx_cursor_sibling(MDBX_cursor *mc, int move_right); -static int mdbx_cursor_next(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data, - MDBX_cursor_op op); -static int mdbx_cursor_prev(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data, - MDBX_cursor_op op); -static int mdbx_cursor_set(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data, - MDBX_cursor_op op, int *exactp); -static int mdbx_cursor_first(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data); -static int mdbx_cursor_last(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data); - -static void mdbx_cursor_init(MDBX_cursor *mc, MDBX_txn *txn, MDBX_dbi dbi, - MDBX_xcursor *mx); -static void mdbx_xcursor_init0(MDBX_cursor *mc); -static void mdbx_xcursor_init1(MDBX_cursor *mc, MDBX_node *node); -static void mdbx_xcursor_init2(MDBX_cursor *mc, MDBX_xcursor *src_mx, - int force); - -static int mdbx_drop0(MDBX_cursor *mc, int subs); - -static MDBX_cmp_func mdbx_cmp_memn, mdbx_cmp_memnr, mdbx_cmp_int_ai, - mdbx_cmp_int_a2, mdbx_cmp_int_ua; - -static const char *__mdbx_strerr(int errnum) { - /* Table of descriptions for MDBX errors */ - static const char *const tbl[] = { - "MDBX_KEYEXIST: Key/data pair already exists", - "MDBX_NOTFOUND: No matching key/data pair found", - "MDBX_PAGE_NOTFOUND: Requested page not found", - "MDBX_CORRUPTED: Database is corrupted", - "MDBX_PANIC: Update of meta page failed or environment had fatal error", - "MDBX_VERSION_MISMATCH: DB version mismatch libmdbx", - "MDBX_INVALID: File is not an MDBX file", - "MDBX_MAP_FULL: Environment mapsize limit reached", - "MDBX_DBS_FULL: Too may DBI (maxdbs reached)", - "MDBX_READERS_FULL: Too many readers (maxreaders reached)", - NULL /* MDBX_TLS_FULL (-30789): unused in MDBX */, - "MDBX_TXN_FULL: Transaction has too many dirty pages - transaction too " - "big", - "MDBX_CURSOR_FULL: Internal error - cursor stack limit reached", - "MDBX_PAGE_FULL: Internal error - page has no more space", - "MDBX_MAP_RESIZED: Database contents grew beyond environment mapsize", - "MDBX_INCOMPATIBLE: Operation and DB incompatible, or DB flags changed", - "MDBX_BAD_RSLOT: Invalid reuse of reader locktable slot", - "MDBX_BAD_TXN: Transaction must abort, has a child, or is invalid", - "MDBX_BAD_VALSIZE: Unsupported size of key/DB name/data, or wrong " - "DUPFIXED size", - "MDBX_BAD_DBI: The specified DBI handle was closed/changed unexpectedly", - "MDBX_PROBLEM: Unexpected problem - txn should abort", - "MDBX_BUSY: Another write transation is started", - }; - - if (errnum >= MDBX_KEYEXIST && errnum <= MDBX_LAST_ERRCODE) { - int i = errnum - MDBX_KEYEXIST; - return tbl[i]; - } - - switch (errnum) { - case MDBX_SUCCESS: - return "MDBX_SUCCESS: Successful"; - case MDBX_EMULTIVAL: - return "MDBX_EMULTIVAL: Unable to update multi-value for the given key"; - case MDBX_EBADSIGN: - return "MDBX_EBADSIGN: Wrong signature of a runtime object(s)"; - case MDBX_WANNA_RECOVERY: - return "MDBX_WANNA_RECOVERY: Database should be recovered, but this could " - "NOT be done in a read-only mode"; - case MDBX_EKEYMISMATCH: - return "MDBX_EKEYMISMATCH: The given key value is mismatched to the " - "current cursor position"; - case MDBX_TOO_LARGE: - return "MDBX_TOO_LARGE: Database is too large for current system, " - "e.g. could NOT be mapped into RAM"; - case MDBX_THREAD_MISMATCH: - return "MDBX_THREAD_MISMATCH: A thread has attempted to use a not " - "owned object, e.g. a transaction that started by another thread"; - default: - return NULL; - } -} - -const char *__cold mdbx_strerror_r(int errnum, char *buf, size_t buflen) { - const char *msg = __mdbx_strerr(errnum); - if (!msg) { - if (!buflen || buflen > INT_MAX) - return NULL; -#ifdef _MSC_VER - size_t size = FormatMessageA( - FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_IGNORE_INSERTS, NULL, - errnum, MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), buf, (DWORD)buflen, - NULL); - return size ? buf : NULL; -#elif defined(_GNU_SOURCE) - /* GNU-specific */ - msg = strerror_r(errnum, buf, buflen); -#elif (_POSIX_C_SOURCE >= 200112L || _XOPEN_SOURCE >= 600) - /* XSI-compliant */ - int rc = strerror_r(errnum, buf, buflen); - if (rc) { - rc = snprintf(buf, buflen, "error %d", errnum); - assert(rc > 0); - } - return buf; -#else - strncpy(buf, strerror(errnum), buflen); - buf[buflen - 1] = '\0'; - return buf; -#endif - } - return msg; -} - -const char *__cold mdbx_strerror(int errnum) { - const char *msg = __mdbx_strerr(errnum); - if (!msg) { -#ifdef _MSC_VER - static __thread char buffer[1024]; - size_t size = FormatMessageA( - FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_IGNORE_INSERTS, NULL, - errnum, MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), buffer, - sizeof(buffer), NULL); - if (size) - msg = buffer; -#else - msg = strerror(errnum); -#endif - } - return msg; -} - -static txnid_t mdbx_oomkick(MDBX_env *env, const txnid_t laggard); - -void __cold mdbx_debug_log(int type, const char *function, int line, - const char *fmt, ...) { - va_list args; - - va_start(args, fmt); - if (mdbx_debug_logger) - mdbx_debug_logger(type, function, line, fmt, args); - else { - if (function && line > 0) - fprintf(stderr, "%s:%d ", function, line); - else if (function) - fprintf(stderr, "%s: ", function); - else if (line > 0) - fprintf(stderr, "%d: ", line); - vfprintf(stderr, fmt, args); - } - va_end(args); -} - -/* Dump a key in ascii or hexadecimal. */ -char *mdbx_dkey(const MDBX_val *key, char *const buf, const size_t bufsize) { - if (!key) - return "<null>"; - if (!buf || bufsize < 4) - return nullptr; - if (!key->iov_len) - return "<empty>"; - - const uint8_t *const data = key->iov_base; - bool is_ascii = true; - unsigned i; - for (i = 0; is_ascii && i < key->iov_len; i++) - if (data[i] < ' ' || data[i] > 127) - is_ascii = false; - - if (is_ascii) { - int len = - snprintf(buf, bufsize, "%.*s", - (key->iov_len > INT_MAX) ? INT_MAX : (int)key->iov_len, data); - assert(len > 0 && (unsigned)len < bufsize); - (void)len; - } else { - char *const detent = buf + bufsize - 2; - char *ptr = buf; - *ptr++ = '<'; - for (i = 0; i < key->iov_len; i++) { - const ptrdiff_t left = detent - ptr; - assert(left > 0); - int len = snprintf(ptr, left, "%02x", data[i]); - if (len < 0 || len >= left) - break; - ptr += len; - } - if (ptr < detent) { - ptr[0] = '>'; - ptr[1] = '\0'; - } - } - return buf; -} - -#if 0 /* LY: debug stuff */ -static const char *mdbx_leafnode_type(MDBX_node *n) { - static char *const tp[2][2] = {{"", ": DB"}, {": sub-page", ": sub-DB"}}; - return F_ISSET(n->mn_flags, F_BIGDATA) ? ": overflow page" - : tp[F_ISSET(n->mn_flags, F_DUPDATA)] - [F_ISSET(n->mn_flags, F_SUBDATA)]; -} - -/* Display all the keys in the page. */ -static void mdbx_page_list(MDBX_page *mp) { - pgno_t pgno = mp->mp_pgno; - const char *type, *state = (mp->mp_flags & P_DIRTY) ? ", dirty" : ""; - MDBX_node *node; - unsigned i, nkeys, nsize, total = 0; - MDBX_val key; - DKBUF; - - switch (mp->mp_flags & - (P_BRANCH | P_LEAF | P_LEAF2 | P_META | P_OVERFLOW | P_SUBP)) { - case P_BRANCH: - type = "Branch page"; - break; - case P_LEAF: - type = "Leaf page"; - break; - case P_LEAF | P_SUBP: - type = "Sub-page"; - break; - case P_LEAF | P_LEAF2: - type = "LEAF2 page"; - break; - case P_LEAF | P_LEAF2 | P_SUBP: - type = "LEAF2 sub-page"; - break; - case P_OVERFLOW: - mdbx_print("Overflow page %" PRIu64 " pages %u%s\n", pgno, mp->mp_pages, - state); - return; - case P_META: - mdbx_print("Meta-page %" PRIu64 " txnid %" PRIu64 "\n", pgno, - ((MDBX_meta *)PAGEDATA(mp))->mm_txnid); - return; - default: - mdbx_print("Bad page %" PRIu64 " flags 0x%X\n", pgno, mp->mp_flags); - return; - } - - nkeys = NUMKEYS(mp); - mdbx_print("%s %" PRIu64 " numkeys %u%s\n", type, pgno, nkeys, state); - - for (i = 0; i < nkeys; i++) { - if (IS_LEAF2(mp)) { /* LEAF2 pages have no mp_ptrs[] or node headers */ - key.iov_len = nsize = mp->mp_leaf2_ksize; - key.iov_base = LEAF2KEY(mp, i, nsize); - total += nsize; - mdbx_print("key %u: nsize %u, %s\n", i, nsize, DKEY(&key)); - continue; - } - node = NODEPTR(mp, i); - key.iov_len = node->mn_ksize; - key.iov_base = node->mn_data; - nsize = NODESIZE + key.iov_len; - if (IS_BRANCH(mp)) { - mdbx_print("key %u: page %" PRIu64 ", %s\n", i, NODEPGNO(node), - DKEY(&key)); - total += nsize; - } else { - if (F_ISSET(node->mn_flags, F_BIGDATA)) - nsize += sizeof(pgno_t); - else - nsize += NODEDSZ(node); - total += nsize; - nsize += sizeof(indx_t); - mdbx_print("key %u: nsize %u, %s%s\n", i, nsize, DKEY(&key), - mdbx_leafnode_type(node)); - } - total = EVEN(total); - } - mdbx_print("Total: header %u + contents %u + unused %u\n", - IS_LEAF2(mp) ? PAGEHDRSZ : PAGEHDRSZ + mp->mp_lower, total, - SIZELEFT(mp)); -} - -static void mdbx_cursor_chk(MDBX_cursor *mc) { - unsigned i; - MDBX_node *node; - MDBX_page *mp; - - if (!mc->mc_snum || !(mc->mc_flags & C_INITIALIZED)) - return; - for (i = 0; i < mc->mc_top; i++) { - mp = mc->mc_pg[i]; - node = NODEPTR(mp, mc->mc_ki[i]); - if (unlikely(NODEPGNO(node) != mc->mc_pg[i + 1]->mp_pgno)) - mdbx_print("oops!\n"); - } - if (unlikely(mc->mc_ki[i] >= NUMKEYS(mc->mc_pg[i]))) - mdbx_print("ack!\n"); - if (XCURSOR_INITED(mc)) { - node = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); - if (((node->mn_flags & (F_DUPDATA | F_SUBDATA)) == F_DUPDATA) && - mc->mc_xcursor->mx_cursor.mc_pg[0] != NODEDATA(node)) { - mdbx_print("blah!\n"); - } - } -} -#endif /* 0 */ - -/* Count all the pages in each DB and in the freelist and make sure - * it matches the actual number of pages being used. - * All named DBs must be open for a correct count. */ -static void mdbx_audit(MDBX_txn *txn) { - MDBX_cursor mc; - MDBX_val key, data; - int rc; - - pgno_t freecount = 0; - mdbx_cursor_init(&mc, txn, FREE_DBI, NULL); - while ((rc = mdbx_cursor_get(&mc, &key, &data, MDBX_NEXT)) == 0) - freecount += *(pgno_t *)data.iov_base; - mdbx_tassert(txn, rc == MDBX_NOTFOUND); - - pgno_t count = 0; - for (MDBX_dbi i = 0; i < txn->mt_numdbs; i++) { - MDBX_xcursor mx; - if (!(txn->mt_dbflags[i] & DB_VALID)) - continue; - mdbx_cursor_init(&mc, txn, i, &mx); - if (txn->mt_dbs[i].md_root == P_INVALID) - continue; - count += txn->mt_dbs[i].md_branch_pages + txn->mt_dbs[i].md_leaf_pages + - txn->mt_dbs[i].md_overflow_pages; - if (txn->mt_dbs[i].md_flags & MDBX_DUPSORT) { - rc = mdbx_page_search(&mc, NULL, MDBX_PS_FIRST); - for (; rc == MDBX_SUCCESS; rc = mdbx_cursor_sibling(&mc, 1)) { - MDBX_page *mp = mc.mc_pg[mc.mc_top]; - for (unsigned j = 0; j < NUMKEYS(mp); j++) { - MDBX_node *leaf = NODEPTR(mp, j); - if (leaf->mn_flags & F_SUBDATA) { - MDBX_db db; - memcpy(&db, NODEDATA(leaf), sizeof(db)); - count += - db.md_branch_pages + db.md_leaf_pages + db.md_overflow_pages; - } - } - } - mdbx_tassert(txn, rc == MDBX_NOTFOUND); - } - } - if (freecount + count + NUM_METAS != txn->mt_next_pgno) { - mdbx_print("audit: %" PRIaTXN " freecount: %" PRIaPGNO " count: %" PRIaPGNO - " total: %" PRIaPGNO " next_pgno: %" PRIaPGNO "\n", - txn->mt_txnid, freecount, count + NUM_METAS, - freecount + count + NUM_METAS, txn->mt_next_pgno); - } -} - -int mdbx_cmp(MDBX_txn *txn, MDBX_dbi dbi, const MDBX_val *a, - const MDBX_val *b) { - mdbx_assert(NULL, txn->mt_signature == MDBX_MT_SIGNATURE); - return txn->mt_dbxs[dbi].md_cmp(a, b); -} - -int mdbx_dcmp(MDBX_txn *txn, MDBX_dbi dbi, const MDBX_val *a, - const MDBX_val *b) { - mdbx_assert(NULL, txn->mt_signature == MDBX_MT_SIGNATURE); - return txn->mt_dbxs[dbi].md_dcmp(a, b); -} - -/* Allocate memory for a page. - * Re-use old malloc'd pages first for singletons, otherwise just malloc. - * Set MDBX_TXN_ERROR on failure. */ -static MDBX_page *mdbx_page_malloc(MDBX_txn *txn, unsigned num) { - MDBX_env *env = txn->mt_env; - MDBX_page *np = env->me_dpages; - size_t size = env->me_psize; - if (likely(num == 1 && np)) { - ASAN_UNPOISON_MEMORY_REGION(np, size); - VALGRIND_MEMPOOL_ALLOC(env, np, size); - VALGRIND_MAKE_MEM_DEFINED(&np->mp_next, sizeof(np->mp_next)); - env->me_dpages = np->mp_next; - } else { - size = pgno2bytes(env, num); - np = malloc(size); - if (unlikely(!np)) { - txn->mt_flags |= MDBX_TXN_ERROR; - return np; - } - VALGRIND_MEMPOOL_ALLOC(env, np, size); - } - - if ((env->me_flags & MDBX_NOMEMINIT) == 0) { - /* For a single page alloc, we init everything after the page header. - * For multi-page, we init the final page; if the caller needed that - * many pages they will be filling in at least up to the last page. */ - size_t skip = PAGEHDRSZ; - if (num > 1) - skip += pgno2bytes(env, num - 1); - memset((char *)np + skip, 0, size - skip); - } -#if MDBX_DEBUG - np->mp_pgno = 0; -#endif - VALGRIND_MAKE_MEM_UNDEFINED(np, size); - np->mp_flags = 0; - np->mp_pages = num; - return np; -} - -/* Free a single page. - * Saves single pages to a list, for future reuse. - * (This is not used for multi-page overflow pages.) */ -static __inline void mdbx_page_free(MDBX_env *env, MDBX_page *mp) { -#if MDBX_DEBUG - mp->mp_pgno = MAX_PAGENO; -#endif - mp->mp_next = env->me_dpages; - VALGRIND_MEMPOOL_FREE(env, mp); - env->me_dpages = mp; -} - -/* Free a dirty page */ -static void mdbx_dpage_free(MDBX_env *env, MDBX_page *dp) { - if (!IS_OVERFLOW(dp) || dp->mp_pages == 1) { - mdbx_page_free(env, dp); - } else { - /* large pages just get freed directly */ - VALGRIND_MEMPOOL_FREE(env, dp); - free(dp); - } -} - -/* Return all dirty pages to dpage list */ -static void mdbx_dlist_free(MDBX_txn *txn) { - MDBX_env *env = txn->mt_env; - MDBX_ID2L dl = txn->mt_rw_dirtylist; - size_t i, n = dl[0].mid; - - for (i = 1; i <= n; i++) - mdbx_dpage_free(env, dl[i].mptr); - - dl[0].mid = 0; -} - -static size_t bytes_align2os_bytes(const MDBX_env *env, size_t bytes) { - return mdbx_roundup2(mdbx_roundup2(bytes, env->me_psize), env->me_os_psize); -} - -static void __cold mdbx_kill_page(MDBX_env *env, pgno_t pgno) { - const size_t offs = pgno2bytes(env, pgno); - const size_t shift = offsetof(MDBX_page, mp_pages); - - if (env->me_flags & MDBX_WRITEMAP) { - MDBX_page *mp = (MDBX_page *)(env->me_map + offs); - memset(&mp->mp_pages, 0x6F /* 'o', 111 */, env->me_psize - shift); - VALGRIND_MAKE_MEM_NOACCESS(&mp->mp_pages, env->me_psize - shift); - ASAN_POISON_MEMORY_REGION(&mp->mp_pages, env->me_psize - shift); - } else { - intptr_t len = env->me_psize - shift; - void *buf = alloca(len); - memset(buf, 0x6F /* 'o', 111 */, len); - (void)mdbx_pwrite(env->me_fd, buf, len, offs + shift); - } -} - -/* Loosen or free a single page. - * - * Saves single pages to a list for future reuse - * in this same txn. It has been pulled from the freeDB - * and already resides on the dirty list, but has been - * deleted. Use these pages first before pulling again - * from the freeDB. - * - * If the page wasn't dirtied in this txn, just add it - * to this txn's free list. */ -static int mdbx_page_loose(MDBX_cursor *mc, MDBX_page *mp) { - int loose = 0; - const pgno_t pgno = mp->mp_pgno; - MDBX_txn *txn = mc->mc_txn; - - if ((mp->mp_flags & P_DIRTY) && mc->mc_dbi != FREE_DBI) { - if (txn->mt_parent) { - mdbx_cassert(mc, (txn->mt_env->me_flags & MDBX_WRITEMAP) == 0); - MDBX_ID2 *dl = txn->mt_rw_dirtylist; - /* If txn has a parent, - * make sure the page is in our dirty list. */ - if (dl[0].mid) { - unsigned x = mdbx_mid2l_search(dl, pgno); - if (x <= dl[0].mid && dl[x].mid == pgno) { - if (unlikely(mp != dl[x].mptr)) { /* bad cursor? */ - mdbx_error("wrong page 0x%p #%" PRIaPGNO - " in the dirtylist[%d], expecting %p", - dl[x].mptr, pgno, x, mp); - mc->mc_flags &= ~(C_INITIALIZED | C_EOF); - txn->mt_flags |= MDBX_TXN_ERROR; - return MDBX_PROBLEM; - } - /* ok, it's ours */ - loose = 1; - } - } - } else { - /* no parent txn, so it's just ours */ - loose = 1; - } - } - if (loose) { - mdbx_debug("loosen db %d page %" PRIaPGNO, DDBI(mc), mp->mp_pgno); - MDBX_page **link = &NEXT_LOOSE_PAGE(mp); - if (unlikely(txn->mt_env->me_flags & MDBX_PAGEPERTURB)) { - mdbx_kill_page(txn->mt_env, pgno); - VALGRIND_MAKE_MEM_UNDEFINED(link, sizeof(MDBX_page *)); - ASAN_UNPOISON_MEMORY_REGION(link, sizeof(MDBX_page *)); - } - *link = txn->mt_loose_pages; - txn->mt_loose_pages = mp; - txn->mt_loose_count++; - mp->mp_flags |= P_LOOSE; - } else { - int rc = mdbx_pnl_append(&txn->mt_befree_pages, pgno); - if (unlikely(rc)) - return rc; - } - - return MDBX_SUCCESS; -} - -/* Set or clear P_KEEP in dirty, non-overflow, non-sub pages watched by txn. - * - * [in] mc A cursor handle for the current operation. - * [in] pflags Flags of the pages to update: - * - P_DIRTY to set P_KEEP, - * - P_DIRTY|P_KEEP to clear it. - * [in] all No shortcuts. Needed except after a full mdbx_page_flush(). - * - * Returns 0 on success, non-zero on failure. */ -static int mdbx_pages_xkeep(MDBX_cursor *mc, unsigned pflags, bool all) { - const unsigned Mask = P_SUBP | P_DIRTY | P_LOOSE | P_KEEP; - MDBX_txn *txn = mc->mc_txn; - MDBX_cursor *m3, *m0 = mc; - MDBX_xcursor *mx; - MDBX_page *dp, *mp; - MDBX_node *leaf; - unsigned i, j; - int rc = MDBX_SUCCESS, level; - - /* Mark pages seen by cursors: First m0, then tracked cursors */ - for (i = txn->mt_numdbs;;) { - if (mc->mc_flags & C_INITIALIZED) { - for (m3 = mc;; m3 = &mx->mx_cursor) { - mp = NULL; - for (j = 0; j < m3->mc_snum; j++) { - mp = m3->mc_pg[j]; - if ((mp->mp_flags & Mask) == pflags) - mp->mp_flags ^= P_KEEP; - } - mx = m3->mc_xcursor; - /* Proceed to mx if it is at a sub-database */ - if (!(mx && (mx->mx_cursor.mc_flags & C_INITIALIZED))) - break; - if (!(mp && (mp->mp_flags & P_LEAF))) - break; - leaf = NODEPTR(mp, m3->mc_ki[j - 1]); - if (!(leaf->mn_flags & F_SUBDATA)) - break; - } - } - mc = mc->mc_next; - for (; !mc || mc == m0; mc = txn->mt_cursors[--i]) - if (i == 0) - goto mark_done; - } - -mark_done: - if (all) { - /* Mark dirty root pages */ - for (i = 0; i < txn->mt_numdbs; i++) { - if (txn->mt_dbflags[i] & DB_DIRTY) { - pgno_t pgno = txn->mt_dbs[i].md_root; - if (pgno == P_INVALID) - continue; - if (unlikely((rc = mdbx_page_get(m0, pgno, &dp, &level)) != - MDBX_SUCCESS)) - break; - if ((dp->mp_flags & Mask) == pflags && level <= 1) - dp->mp_flags ^= P_KEEP; - } - } - } - - return rc; -} - -static int mdbx_page_flush(MDBX_txn *txn, pgno_t keep); - -/* Spill pages from the dirty list back to disk. - * This is intended to prevent running into MDBX_TXN_FULL situations, - * but note that they may still occur in a few cases: - * - * 1) our estimate of the txn size could be too small. Currently this - * seems unlikely, except with a large number of MDBX_MULTIPLE items. - * - * 2) child txns may run out of space if their parents dirtied a - * lot of pages and never spilled them. TODO: we probably should do - * a preemptive spill during mdbx_txn_begin() of a child txn, if - * the parent's dirtyroom is below a given threshold. - * - * Otherwise, if not using nested txns, it is expected that apps will - * not run into MDBX_TXN_FULL any more. The pages are flushed to disk - * the same way as for a txn commit, e.g. their P_DIRTY flag is cleared. - * If the txn never references them again, they can be left alone. - * If the txn only reads them, they can be used without any fuss. - * If the txn writes them again, they can be dirtied immediately without - * going thru all of the work of mdbx_page_touch(). Such references are - * handled by mdbx_page_unspill(). - * - * Also note, we never spill DB root pages, nor pages of active cursors, - * because we'll need these back again soon anyway. And in nested txns, - * we can't spill a page in a child txn if it was already spilled in a - * parent txn. That would alter the parent txns' data even though - * the child hasn't committed yet, and we'd have no way to undo it if - * the child aborted. - * - * [in] m0 cursor A cursor handle identifying the transaction and - * database for which we are checking space. - * [in] key For a put operation, the key being stored. - * [in] data For a put operation, the data being stored. - * - * Returns 0 on success, non-zero on failure. */ -static int mdbx_page_spill(MDBX_cursor *m0, MDBX_val *key, MDBX_val *data) { - MDBX_txn *txn = m0->mc_txn; - MDBX_ID2L dl = txn->mt_rw_dirtylist; - - if (m0->mc_flags & C_SUB) - return MDBX_SUCCESS; - - /* Estimate how much space this op will take */ - pgno_t i = m0->mc_db->md_depth; - /* Named DBs also dirty the main DB */ - if (m0->mc_dbi >= CORE_DBS) - i += txn->mt_dbs[MAIN_DBI].md_depth; - /* For puts, roughly factor in the key+data size */ - if (key) - i += bytes2pgno(txn->mt_env, LEAFSIZE(key, data) + txn->mt_env->me_psize); - i += i; /* double it for good measure */ - pgno_t need = i; - - if (txn->mt_dirtyroom > i) - return MDBX_SUCCESS; - - if (!txn->mt_spill_pages) { - txn->mt_spill_pages = mdbx_pnl_alloc(MDBX_PNL_UM_MAX); - if (unlikely(!txn->mt_spill_pages)) - return MDBX_ENOMEM; - } else { - /* purge deleted slots */ - MDBX_PNL sl = txn->mt_spill_pages; - pgno_t num = sl[0], j = 0; - for (i = 1; i <= num; i++) { - if (!(sl[i] & 1)) - sl[++j] = sl[i]; - } - sl[0] = j; - } - - /* Preserve pages which may soon be dirtied again */ - int rc = mdbx_pages_xkeep(m0, P_DIRTY, true); - if (unlikely(rc != MDBX_SUCCESS)) - goto bailout; - - /* Less aggressive spill - we originally spilled the entire dirty list, - * with a few exceptions for cursor pages and DB root pages. But this - * turns out to be a lot of wasted effort because in a large txn many - * of those pages will need to be used again. So now we spill only 1/8th - * of the dirty pages. Testing revealed this to be a good tradeoff, - * better than 1/2, 1/4, or 1/10. */ - if (need < MDBX_PNL_UM_MAX / 8) - need = MDBX_PNL_UM_MAX / 8; - - /* Save the page IDs of all the pages we're flushing */ - /* flush from the tail forward, this saves a lot of shifting later on. */ - for (i = dl[0].mid; i && need; i--) { - pgno_t pn = dl[i].mid << 1; - MDBX_page *dp = dl[i].mptr; - if (dp->mp_flags & (P_LOOSE | P_KEEP)) - continue; - /* Can't spill twice, - * make sure it's not already in a parent's spill list. */ - if (txn->mt_parent) { - MDBX_txn *tx2; - for (tx2 = txn->mt_parent; tx2; tx2 = tx2->mt_parent) { - if (tx2->mt_spill_pages) { - unsigned j = mdbx_pnl_search(tx2->mt_spill_pages, pn); - if (j <= tx2->mt_spill_pages[0] && tx2->mt_spill_pages[j] == pn) { - dp->mp_flags |= P_KEEP; - break; - } - } - } - if (tx2) - continue; - } - rc = mdbx_pnl_append(&txn->mt_spill_pages, pn); - if (unlikely(rc != MDBX_SUCCESS)) - goto bailout; - need--; - } - mdbx_pnl_sort(txn->mt_spill_pages); - - /* Flush the spilled part of dirty list */ - rc = mdbx_page_flush(txn, i); - if (unlikely(rc != MDBX_SUCCESS)) - goto bailout; - - /* Reset any dirty pages we kept that page_flush didn't see */ - rc = mdbx_pages_xkeep(m0, P_DIRTY | P_KEEP, i != 0); - -bailout: - txn->mt_flags |= rc ? MDBX_TXN_ERROR : MDBX_TXN_SPILLS; - return rc; -} - -/*----------------------------------------------------------------------------*/ - -#define METAPAGE(env, n) (&pgno2page(env, n)->mp_meta) - -#define METAPAGE_END(env) METAPAGE(env, NUM_METAS) - -static __inline txnid_t meta_txnid(const MDBX_env *env, const MDBX_meta *meta, - bool allow_volatile) { - mdbx_assert(env, meta >= METAPAGE(env, 0) || meta < METAPAGE_END(env)); - txnid_t a = meta->mm_txnid_a; - txnid_t b = meta->mm_txnid_b; - if (allow_volatile) - return (a == b) ? a : 0; - mdbx_assert(env, a == b); - return a; -} - -static __inline txnid_t mdbx_meta_txnid_stable(const MDBX_env *env, - const MDBX_meta *meta) { - return meta_txnid(env, meta, false); -} - -static __inline txnid_t mdbx_meta_txnid_fluid(const MDBX_env *env, - const MDBX_meta *meta) { - return meta_txnid(env, meta, true); -} - -static __inline void mdbx_meta_update_begin(const MDBX_env *env, - MDBX_meta *meta, txnid_t txnid) { - mdbx_assert(env, meta >= METAPAGE(env, 0) || meta < METAPAGE_END(env)); - mdbx_assert(env, meta->mm_txnid_a < txnid && meta->mm_txnid_b < txnid); - meta->mm_txnid_a = txnid; - (void)env; - mdbx_coherent_barrier(); -} - -static __inline void mdbx_meta_update_end(const MDBX_env *env, MDBX_meta *meta, - txnid_t txnid) { - mdbx_assert(env, meta >= METAPAGE(env, 0) || meta < METAPAGE_END(env)); - mdbx_assert(env, meta->mm_txnid_a == txnid); - mdbx_assert(env, meta->mm_txnid_b < txnid); - - mdbx_jitter4testing(true); - meta->mm_txnid_b = txnid; - mdbx_coherent_barrier(); -} - -static __inline void mdbx_meta_set_txnid(const MDBX_env *env, MDBX_meta *meta, - txnid_t txnid) { - mdbx_assert(env, meta < METAPAGE(env, 0) || meta > METAPAGE_END(env)); - meta->mm_txnid_a = txnid; - meta->mm_txnid_b = txnid; -} - -static __inline uint64_t mdbx_meta_sign(const MDBX_meta *meta) { - uint64_t sign = MDBX_DATASIGN_NONE; -#if 0 /* TODO */ - sign = hippeus_hash64(...); -#else - (void)meta; -#endif - /* LY: newer returns MDBX_DATASIGN_NONE or MDBX_DATASIGN_WEAK */ - return (sign > MDBX_DATASIGN_WEAK) ? sign : ~sign; -} - -enum meta_choise_mode { prefer_last, prefer_noweak, prefer_steady }; - -static __inline bool mdbx_meta_ot(const enum meta_choise_mode mode, - const MDBX_env *env, const MDBX_meta *a, - const MDBX_meta *b) { - mdbx_jitter4testing(true); - txnid_t txnid_a = mdbx_meta_txnid_fluid(env, a); - txnid_t txnid_b = mdbx_meta_txnid_fluid(env, b); - - mdbx_jitter4testing(true); - switch (mode) { - default: - assert(false); - __unreachable(); - /* fall through */ - __fallthrough; - case prefer_steady: - if (META_IS_STEADY(a) != META_IS_STEADY(b)) - return META_IS_STEADY(b); - /* fall through */ - __fallthrough; - case prefer_noweak: - if (META_IS_WEAK(a) != META_IS_WEAK(b)) - return !META_IS_WEAK(b); - /* fall through */ - __fallthrough; - case prefer_last: - mdbx_jitter4testing(true); - if (txnid_a == txnid_b) - return META_IS_STEADY(b) || (META_IS_WEAK(a) && !META_IS_WEAK(b)); - return txnid_a < txnid_b; - } -} - -static __inline bool mdbx_meta_eq(const MDBX_env *env, const MDBX_meta *a, - const MDBX_meta *b) { - mdbx_jitter4testing(true); - const txnid_t txnid = mdbx_meta_txnid_fluid(env, a); - if (!txnid || txnid != mdbx_meta_txnid_fluid(env, b)) - return false; - - mdbx_jitter4testing(true); - if (META_IS_STEADY(a) != META_IS_STEADY(b)) - return false; - - mdbx_jitter4testing(true); - return true; -} - -static int mdbx_meta_eq_mask(const MDBX_env *env) { - MDBX_meta *m0 = METAPAGE(env, 0); - MDBX_meta *m1 = METAPAGE(env, 1); - MDBX_meta *m2 = METAPAGE(env, 2); - - int rc = mdbx_meta_eq(env, m0, m1) ? 1 : 0; - if (mdbx_meta_eq(env, m1, m2)) - rc += 2; - if (mdbx_meta_eq(env, m2, m0)) - rc += 4; - return rc; -} - -static __inline MDBX_meta *mdbx_meta_recent(const enum meta_choise_mode mode, - const MDBX_env *env, MDBX_meta *a, - MDBX_meta *b) { - const bool a_older_that_b = mdbx_meta_ot(mode, env, a, b); - mdbx_assert(env, !mdbx_meta_eq(env, a, b)); - return a_older_that_b ? b : a; -} - -static __inline MDBX_meta *mdbx_meta_ancient(const enum meta_choise_mode mode, - const MDBX_env *env, MDBX_meta *a, - MDBX_meta *b) { - const bool a_older_that_b = mdbx_meta_ot(mode, env, a, b); - mdbx_assert(env, !mdbx_meta_eq(env, a, b)); - return a_older_that_b ? a : b; -} - -static __inline MDBX_meta * -mdbx_meta_mostrecent(const enum meta_choise_mode mode, const MDBX_env *env) { - MDBX_meta *m0 = METAPAGE(env, 0); - MDBX_meta *m1 = METAPAGE(env, 1); - MDBX_meta *m2 = METAPAGE(env, 2); - - MDBX_meta *head = mdbx_meta_recent(mode, env, m0, m1); - head = mdbx_meta_recent(mode, env, head, m2); - return head; -} - -static __hot MDBX_meta *mdbx_meta_steady(const MDBX_env *env) { - return mdbx_meta_mostrecent(prefer_steady, env); -} - -static __hot MDBX_meta *mdbx_meta_head(const MDBX_env *env) { - return mdbx_meta_mostrecent(prefer_last, env); -} - -static __hot txnid_t mdbx_reclaiming_detent(const MDBX_env *env) { - if (F_ISSET(env->me_flags, MDBX_UTTERLY_NOSYNC)) - return env->me_txn->mt_txnid - 1; - - return mdbx_meta_txnid_stable(env, mdbx_meta_steady(env)); -} - -static const char *mdbx_durable_str(const MDBX_meta *const meta) { - if (META_IS_WEAK(meta)) - return "Weak"; - if (META_IS_STEADY(meta)) - return (meta->mm_datasync_sign == mdbx_meta_sign(meta)) ? "Steady" - : "Tainted"; - return "Legacy"; -} - -/*----------------------------------------------------------------------------*/ - -/* Find oldest txnid still referenced. */ -static txnid_t mdbx_find_oldest(MDBX_txn *txn) { - mdbx_tassert(txn, (txn->mt_flags & MDBX_RDONLY) == 0); - const MDBX_env *env = txn->mt_env; - MDBX_lockinfo *const lck = env->me_lck; - - const txnid_t edge = mdbx_reclaiming_detent(env); - mdbx_tassert(txn, edge <= txn->mt_txnid - 1); - const txnid_t last_oldest = lck->mti_oldest; - mdbx_tassert(txn, edge >= last_oldest); - if (last_oldest == edge) - return edge; - - const uint32_t nothing_changed = MDBX_STRING_TETRAD("None"); - const uint32_t snap_readers_refresh_flag = lck->mti_readers_refresh_flag; - mdbx_jitter4testing(false); - if (snap_readers_refresh_flag == nothing_changed) - return last_oldest; - - txnid_t oldest = edge; - lck->mti_readers_refresh_flag = nothing_changed; - mdbx_coherent_barrier(); - const unsigned snap_nreaders = lck->mti_numreaders; - for (unsigned i = 0; i < snap_nreaders; ++i) { - if (lck->mti_readers[i].mr_pid) { - /* mdbx_jitter4testing(true); */ - const txnid_t snap = lck->mti_readers[i].mr_txnid; - if (oldest > snap && last_oldest <= /* ignore pending updates */ snap) { - oldest = snap; - if (oldest == last_oldest) - return oldest; - } - } - } - - if (oldest != last_oldest) { - mdbx_notice("update oldest %" PRIaTXN " -> %" PRIaTXN, last_oldest, oldest); - mdbx_tassert(txn, oldest >= lck->mti_oldest); - lck->mti_oldest = oldest; - } - return oldest; -} - -/* Add a page to the txn's dirty list */ -static void mdbx_page_dirty(MDBX_txn *txn, MDBX_page *mp) { - MDBX_ID2 mid; - int rc, (*insert)(MDBX_ID2L, MDBX_ID2 *); - - if (txn->mt_flags & MDBX_TXN_WRITEMAP) { - insert = mdbx_mid2l_append; - } else { - insert = mdbx_mid2l_insert; - } - mid.mid = mp->mp_pgno; - mid.mptr = mp; - rc = insert(txn->mt_rw_dirtylist, &mid); - mdbx_tassert(txn, rc == 0); - txn->mt_dirtyroom--; -} - -static int mdbx_mapresize(MDBX_env *env, const pgno_t size_pgno, - const pgno_t limit_pgno) { -#ifdef USE_VALGRIND - const size_t prev_mapsize = env->me_mapsize; - void *const prev_mapaddr = env->me_map; -#endif - - const size_t limit_bytes = pgno_align2os_bytes(env, limit_pgno); - const size_t size_bytes = pgno_align2os_bytes(env, size_pgno); - - mdbx_info("resize datafile/mapping: " - "present %" PRIuPTR " -> %" PRIuPTR ", " - "limit %" PRIuPTR " -> %" PRIuPTR, - env->me_dbgeo.now, size_bytes, env->me_dbgeo.upper, limit_bytes); - - mdbx_assert(env, limit_bytes >= size_bytes); - mdbx_assert(env, bytes2pgno(env, size_bytes) == size_pgno); - mdbx_assert(env, bytes2pgno(env, limit_bytes) == limit_pgno); - const int rc = - mdbx_mresize(env->me_flags, &env->me_dxb_mmap, size_bytes, limit_bytes); - - if (rc == MDBX_SUCCESS) { - env->me_dbgeo.now = size_bytes; - env->me_dbgeo.upper = limit_bytes; - } else if (rc != MDBX_RESULT_TRUE) { - mdbx_error("failed resize datafile/mapping: " - "present %" PRIuPTR " -> %" PRIuPTR ", " - "limit %" PRIuPTR " -> %" PRIuPTR ", errcode %d", - env->me_dbgeo.now, size_bytes, env->me_dbgeo.upper, limit_bytes, - rc); - return rc; - } else { - mdbx_notice("unable resize datafile/mapping: " - "present %" PRIuPTR " -> %" PRIuPTR ", " - "limit %" PRIuPTR " -> %" PRIuPTR ", errcode %d", - env->me_dbgeo.now, size_bytes, env->me_dbgeo.upper, limit_bytes, - rc); - } - -#ifdef USE_VALGRIND - if (prev_mapsize != env->me_mapsize || prev_mapaddr != env->me_map) { - VALGRIND_DISCARD(env->me_valgrind_handle); - env->me_valgrind_handle = 0; - if (env->me_mapsize) - env->me_valgrind_handle = - VALGRIND_CREATE_BLOCK(env->me_map, env->me_mapsize, "mdbx"); - } -#endif - - if (env->me_txn) { - mdbx_tassert(env->me_txn, size_pgno >= env->me_txn->mt_next_pgno); - env->me_txn->mt_end_pgno = size_pgno; - } - return MDBX_SUCCESS; -} - -/* Allocate page numbers and memory for writing. Maintain me_last_reclaimed, - * me_reclaimed_pglist and mt_next_pgno. Set MDBX_TXN_ERROR on failure. - * - * If there are free pages available from older transactions, they - * are re-used first. Otherwise allocate a new page at mt_next_pgno. - * Do not modify the freedB, just merge freeDB records into me_reclaimed_pglist - * and move me_last_reclaimed to say which records were consumed. Only this - * function can create me_reclaimed_pglist and move - * me_last_reclaimed/mt_next_pgno. - * - * [in] mc cursor A cursor handle identifying the transaction and - * database for which we are allocating. - * [in] num the number of pages to allocate. - * [out] mp Address of the allocated page(s). Requests for multiple pages - * will always be satisfied by a single contiguous chunk of memory. - * - * Returns 0 on success, non-zero on failure.*/ - -#define MDBX_ALLOC_CACHE 1 -#define MDBX_ALLOC_GC 2 -#define MDBX_ALLOC_NEW 4 -#define MDBX_ALLOC_KICK 8 -#define MDBX_ALLOC_ALL \ - (MDBX_ALLOC_CACHE | MDBX_ALLOC_GC | MDBX_ALLOC_NEW | MDBX_ALLOC_KICK) - -static int mdbx_page_alloc(MDBX_cursor *mc, unsigned num, MDBX_page **mp, - int flags) { - int rc; - MDBX_txn *txn = mc->mc_txn; - MDBX_env *env = txn->mt_env; - MDBX_page *np; - - if (likely(flags & MDBX_ALLOC_GC)) { - flags |= env->me_flags & (MDBX_COALESCE | MDBX_LIFORECLAIM); - if (unlikely(mc->mc_flags & C_RECLAIMING)) { - /* If mc is updating the freeDB, then the freelist cannot play - * catch-up with itself by growing while trying to save it. */ - flags &= - ~(MDBX_ALLOC_GC | MDBX_ALLOC_KICK | MDBX_COALESCE | MDBX_LIFORECLAIM); - } - } - - if (likely(flags & MDBX_ALLOC_CACHE)) { - /* If there are any loose pages, just use them */ - assert(mp && num); - if (likely(num == 1 && txn->mt_loose_pages)) { - np = txn->mt_loose_pages; - txn->mt_loose_pages = NEXT_LOOSE_PAGE(np); - txn->mt_loose_count--; - mdbx_debug("db %d use loose page %" PRIaPGNO, DDBI(mc), np->mp_pgno); - ASAN_UNPOISON_MEMORY_REGION(np, env->me_psize); - mdbx_tassert(txn, np->mp_pgno < txn->mt_next_pgno); - mdbx_ensure(env, np->mp_pgno >= NUM_METAS); - *mp = np; - return MDBX_SUCCESS; - } - } - - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - pgno_t pgno, *repg_list = env->me_reclaimed_pglist; - unsigned repg_pos = 0, repg_len = repg_list ? repg_list[0] : 0; - txnid_t oldest = 0, last = 0; - const unsigned wanna_range = num - 1; - - while (1) { /* oom-kick retry loop */ - /* If our dirty list is already full, we can't do anything */ - if (unlikely(txn->mt_dirtyroom == 0)) { - rc = MDBX_TXN_FULL; - goto fail; - } - - MDBX_cursor recur; - for (MDBX_cursor_op op = MDBX_FIRST;; - op = (flags & MDBX_LIFORECLAIM) ? MDBX_PREV : MDBX_NEXT) { - MDBX_val key, data; - - /* Seek a big enough contiguous page range. - * Prefer pages with lower pgno. */ - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - if (likely(flags & MDBX_ALLOC_CACHE) && repg_len > wanna_range && - (!(flags & MDBX_COALESCE) || op == MDBX_FIRST)) { -#if MDBX_PNL_ASCENDING - for (repg_pos = 1; repg_pos <= repg_len - wanna_range; ++repg_pos) { - pgno = repg_list[repg_pos]; - if (likely(repg_list[repg_pos + wanna_range - 1] == - pgno + wanna_range - 1)) - goto done; - } -#else - repg_pos = repg_len; - do { - pgno = repg_list[repg_pos]; - if (likely(repg_list[repg_pos - wanna_range] == pgno + wanna_range)) - goto done; - } while (--repg_pos > wanna_range); -#endif /* MDBX_PNL sort-order */ - } - - if (op == MDBX_FIRST) { /* 1st iteration, setup cursor, etc */ - if (unlikely(!(flags & MDBX_ALLOC_GC))) - break /* reclaiming is prohibited for now */; - - /* Prepare to fetch more and coalesce */ - oldest = (flags & MDBX_LIFORECLAIM) ? mdbx_find_oldest(txn) - : env->me_oldest[0]; - mdbx_cursor_init(&recur, txn, FREE_DBI, NULL); - if (flags & MDBX_LIFORECLAIM) { - /* Begin from oldest reader if any */ - if (oldest > 2) { - last = oldest - 1; - op = MDBX_SET_RANGE; - } - } else if (env->me_last_reclaimed) { - /* Continue lookup from env->me_last_reclaimed to oldest reader */ - last = env->me_last_reclaimed; - op = MDBX_SET_RANGE; - } - - key.iov_base = &last; - key.iov_len = sizeof(last); - } - - if (!(flags & MDBX_LIFORECLAIM)) { - /* Do not try fetch more if the record will be too recent */ - if (op != MDBX_FIRST && ++last >= oldest) { - oldest = mdbx_find_oldest(txn); - if (oldest <= last) - break; - } - } - - rc = mdbx_cursor_get(&recur, &key, NULL, op); - if (rc == MDBX_NOTFOUND && (flags & MDBX_LIFORECLAIM)) { - if (op == MDBX_SET_RANGE) - continue; - if (oldest < mdbx_find_oldest(txn)) { - oldest = *env->me_oldest; - last = oldest - 1; - key.iov_base = &last; - key.iov_len = sizeof(last); - op = MDBX_SET_RANGE; - rc = mdbx_cursor_get(&recur, &key, NULL, op); - } - } - if (unlikely(rc)) { - if (rc == MDBX_NOTFOUND) - break; - goto fail; - } - - last = *(txnid_t *)key.iov_base; - if (oldest <= last) { - oldest = mdbx_find_oldest(txn); - if (oldest <= last) { - if (flags & MDBX_LIFORECLAIM) - continue; - break; - } - } - - if (flags & MDBX_LIFORECLAIM) { - /* skip IDs of records that already reclaimed */ - if (txn->mt_lifo_reclaimed) { - unsigned i; - for (i = (unsigned)txn->mt_lifo_reclaimed[0]; i > 0; --i) - if (txn->mt_lifo_reclaimed[i] == last) - break; - if (i) - continue; - } - } - - /* Reading next FreeDB record */ - np = recur.mc_pg[recur.mc_top]; - MDBX_node *leaf = NODEPTR(np, recur.mc_ki[recur.mc_top]); - if (unlikely((rc = mdbx_node_read(&recur, leaf, &data)) != MDBX_SUCCESS)) - goto fail; - - if ((flags & MDBX_LIFORECLAIM) && !txn->mt_lifo_reclaimed) { - txn->mt_lifo_reclaimed = mdbx_txl_alloc(); - if (unlikely(!txn->mt_lifo_reclaimed)) { - rc = MDBX_ENOMEM; - goto fail; - } - } - - /* Append PNL from FreeDB record to me_reclaimed_pglist */ - pgno_t *re_pnl = (pgno_t *)data.iov_base; - mdbx_tassert(txn, re_pnl[0] == 0 || - data.iov_len == (re_pnl[0] + 1) * sizeof(pgno_t)); - mdbx_tassert(txn, mdbx_pnl_check(re_pnl)); - repg_pos = re_pnl[0]; - if (!repg_list) { - if (unlikely(!(env->me_reclaimed_pglist = repg_list = - mdbx_pnl_alloc(repg_pos)))) { - rc = MDBX_ENOMEM; - goto fail; - } - } else { - if (unlikely( - (rc = mdbx_pnl_need(&env->me_reclaimed_pglist, repg_pos)) != 0)) - goto fail; - repg_list = env->me_reclaimed_pglist; - } - - /* Remember ID of FreeDB record */ - if (flags & MDBX_LIFORECLAIM) { - if ((rc = mdbx_txl_append(&txn->mt_lifo_reclaimed, last)) != 0) - goto fail; - } - env->me_last_reclaimed = last; - - if (mdbx_debug_enabled(MDBX_DBG_EXTRA)) { - mdbx_debug_extra("PNL read txn %" PRIaTXN " root %" PRIaPGNO - " num %u, PNL", - last, txn->mt_dbs[FREE_DBI].md_root, repg_pos); - unsigned i; - for (i = repg_pos; i; i--) - mdbx_debug_extra_print(" %" PRIaPGNO "", re_pnl[i]); - mdbx_debug_extra_print("\n"); - } - - /* Merge in descending sorted order */ - mdbx_pnl_xmerge(repg_list, re_pnl); - repg_len = repg_list[0]; - if (unlikely((flags & MDBX_ALLOC_CACHE) == 0)) { - /* Done for a kick-reclaim mode, actually no page needed */ - return MDBX_SUCCESS; - } - - mdbx_tassert(txn, - repg_len == 0 || repg_list[repg_len] < txn->mt_next_pgno); - if (repg_len) { - /* Refund suitable pages into "unallocated" space */ - pgno_t tail = txn->mt_next_pgno; - pgno_t *const begin = repg_list + 1; - pgno_t *const end = begin + repg_len; - pgno_t *higest; -#if MDBX_PNL_ASCENDING - for (higest = end; --higest >= begin;) { -#else - for (higest = begin; higest < end; ++higest) { -#endif /* MDBX_PNL sort-order */ - mdbx_tassert(txn, *higest >= NUM_METAS && *higest < tail); - if (*higest != tail - 1) - break; - tail -= 1; - } - if (tail != txn->mt_next_pgno) { -#if MDBX_PNL_ASCENDING - repg_len = (unsigned)(higest + 1 - begin); -#else - repg_len -= (unsigned)(higest - begin); - for (pgno_t *move = begin; higest < end; ++move, ++higest) - *move = *higest; -#endif /* MDBX_PNL sort-order */ - repg_list[0] = repg_len; - mdbx_info("refunded %" PRIaPGNO " pages: %" PRIaPGNO " -> %" PRIaPGNO, - tail - txn->mt_next_pgno, tail, txn->mt_next_pgno); - txn->mt_next_pgno = tail; - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - } - } - - /* Don't try to coalesce too much. */ - if (repg_len > MDBX_PNL_UM_SIZE / 2) - break; - if (flags & MDBX_COALESCE) { - if (repg_len /* current size */ >= env->me_maxfree_1pg / 2 || - repg_pos /* prev size */ >= env->me_maxfree_1pg / 4) - flags &= ~MDBX_COALESCE; - } - } - - if ((flags & (MDBX_COALESCE | MDBX_ALLOC_CACHE)) == - (MDBX_COALESCE | MDBX_ALLOC_CACHE) && - repg_len > wanna_range) { -#if MDBX_PNL_ASCENDING - for (repg_pos = 1; repg_pos <= repg_len - wanna_range; ++repg_pos) { - pgno = repg_list[repg_pos]; - if (likely(repg_list[repg_pos + wanna_range - 1] == - pgno + wanna_range - 1)) - goto done; - } -#else - repg_pos = repg_len; - do { - pgno = repg_list[repg_pos]; - if (likely(repg_list[repg_pos - wanna_range] == pgno + wanna_range)) - goto done; - } while (--repg_pos > wanna_range); -#endif /* MDBX_PNL sort-order */ - } - - /* Use new pages from the map when nothing suitable in the freeDB */ - repg_pos = 0; - pgno = txn->mt_next_pgno; - rc = MDBX_MAP_FULL; - const pgno_t next = pgno_add(pgno, num); - if (likely(next <= txn->mt_end_pgno)) { - rc = MDBX_NOTFOUND; - if (likely(flags & MDBX_ALLOC_NEW)) - goto done; - } - - const MDBX_meta *head = mdbx_meta_head(env); - if ((flags & MDBX_ALLOC_GC) && - ((flags & MDBX_ALLOC_KICK) || rc == MDBX_MAP_FULL)) { - MDBX_meta *steady = mdbx_meta_steady(env); - - if (oldest == mdbx_meta_txnid_stable(env, steady) && - !META_IS_STEADY(head) && META_IS_STEADY(steady)) { - /* LY: Here an oom was happened: - * - all pages had allocated; - * - reclaiming was stopped at the last steady-sync; - * - the head-sync is weak. - * Now we need make a sync to resume reclaiming. If both - * MDBX_NOSYNC and MDBX_MAPASYNC flags are set, then assume that - * utterly no-sync write mode was requested. In such case - * don't make a steady-sync, but only a legacy-mode checkpoint, - * just for resume reclaiming only, not for data consistency. */ - - mdbx_debug("kick-gc: head %" PRIaTXN "-%s, tail %" PRIaTXN - "-%s, oldest %" PRIaTXN "", - mdbx_meta_txnid_stable(env, head), mdbx_durable_str(head), - mdbx_meta_txnid_stable(env, steady), - mdbx_durable_str(steady), oldest); - - const unsigned syncflags = F_ISSET(env->me_flags, MDBX_UTTERLY_NOSYNC) - ? env->me_flags - : env->me_flags & MDBX_WRITEMAP; - MDBX_meta meta = *head; - if (mdbx_sync_locked(env, syncflags, &meta) == MDBX_SUCCESS) { - txnid_t snap = mdbx_find_oldest(txn); - if (snap > oldest) - continue; - } - } - - if (rc == MDBX_MAP_FULL && oldest < txn->mt_txnid - 1) { - if (mdbx_oomkick(env, oldest) > oldest) - continue; - } - } - - if (rc == MDBX_MAP_FULL && next < head->mm_geo.upper) { - mdbx_assert(env, next > txn->mt_end_pgno); - pgno_t aligned = pgno_align2os_pgno( - env, pgno_add(next, head->mm_geo.grow - next % head->mm_geo.grow)); - - if (aligned > head->mm_geo.upper) - aligned = head->mm_geo.upper; - mdbx_assert(env, aligned > txn->mt_end_pgno); - - mdbx_info("try growth datafile to %" PRIaPGNO " pages (+%" PRIaPGNO ")", - aligned, aligned - txn->mt_end_pgno); - rc = mdbx_mapresize(env, aligned, head->mm_geo.upper); - if (rc == MDBX_SUCCESS) { - mdbx_tassert(env->me_txn, txn->mt_end_pgno >= next); - if (!mp) - return rc; - goto done; - } - - mdbx_warning("unable growth datafile to %" PRIaPGNO "pages (+%" PRIaPGNO - "), errcode %d", - aligned, aligned - txn->mt_end_pgno, rc); - } - - fail: - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - if (mp) { - *mp = NULL; - txn->mt_flags |= MDBX_TXN_ERROR; - } - assert(rc); - return rc; - } - -done: - mdbx_tassert(txn, mp && num); - mdbx_ensure(env, pgno >= NUM_METAS); - if (env->me_flags & MDBX_WRITEMAP) { - np = pgno2page(env, pgno); - /* LY: reset no-access flag from mdbx_kill_page() */ - VALGRIND_MAKE_MEM_UNDEFINED(np, pgno2bytes(env, num)); - ASAN_UNPOISON_MEMORY_REGION(np, pgno2bytes(env, num)); - } else { - if (unlikely(!(np = mdbx_page_malloc(txn, num)))) { - rc = MDBX_ENOMEM; - goto fail; - } - } - - if (repg_pos) { - mdbx_tassert(txn, pgno < txn->mt_next_pgno); - mdbx_tassert(txn, pgno == repg_list[repg_pos]); - /* Cutoff allocated pages from me_reclaimed_pglist */ - repg_list[0] = repg_len -= num; - for (unsigned i = repg_pos - num; i < repg_len;) - repg_list[++i] = repg_list[++repg_pos]; - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - } else { - txn->mt_next_pgno = pgno + num; - mdbx_assert(env, txn->mt_next_pgno <= txn->mt_end_pgno); - } - - if (env->me_flags & MDBX_PAGEPERTURB) - memset(np, 0x71 /* 'q', 113 */, pgno2bytes(env, num)); - VALGRIND_MAKE_MEM_UNDEFINED(np, pgno2bytes(env, num)); - - np->mp_pgno = pgno; - np->mp_leaf2_ksize = 0; - np->mp_flags = 0; - np->mp_pages = num; - mdbx_page_dirty(txn, np); - *mp = np; - - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - return MDBX_SUCCESS; -} - -/* Copy the used portions of a non-overflow page. - * [in] dst page to copy into - * [in] src page to copy from - * [in] psize size of a page */ -static void mdbx_page_copy(MDBX_page *dst, MDBX_page *src, unsigned psize) { - STATIC_ASSERT(UINT16_MAX > MAX_PAGESIZE - PAGEHDRSZ); - STATIC_ASSERT(MIN_PAGESIZE > PAGEHDRSZ + NODESIZE * 42); - enum { Align = sizeof(pgno_t) }; - indx_t upper = src->mp_upper, lower = src->mp_lower, unused = upper - lower; - - /* If page isn't full, just copy the used portion. Adjust - * alignment so memcpy may copy words instead of bytes. */ - if ((unused &= -Align) && !IS_LEAF2(src)) { - upper = (upper + PAGEHDRSZ) & -Align; - memcpy(dst, src, (lower + PAGEHDRSZ + (Align - 1)) & -Align); - memcpy((pgno_t *)((char *)dst + upper), (pgno_t *)((char *)src + upper), - psize - upper); - } else { - memcpy(dst, src, psize - unused); - } -} - -/* Pull a page off the txn's spill list, if present. - * - * If a page being referenced was spilled to disk in this txn, bring - * it back and make it dirty/writable again. - * - * [in] txn the transaction handle. - * [in] mp the page being referenced. It must not be dirty. - * [out] ret the writable page, if any. - * ret is unchanged if mp wasn't spilled. */ -static int mdbx_page_unspill(MDBX_txn *txn, MDBX_page *mp, MDBX_page **ret) { - MDBX_env *env = txn->mt_env; - const MDBX_txn *tx2; - unsigned x; - pgno_t pgno = mp->mp_pgno, pn = pgno << 1; - - for (tx2 = txn; tx2; tx2 = tx2->mt_parent) { - if (!tx2->mt_spill_pages) - continue; - x = mdbx_pnl_search(tx2->mt_spill_pages, pn); - if (x <= tx2->mt_spill_pages[0] && tx2->mt_spill_pages[x] == pn) { - MDBX_page *np; - int num; - if (txn->mt_dirtyroom == 0) - return MDBX_TXN_FULL; - num = IS_OVERFLOW(mp) ? mp->mp_pages : 1; - if (env->me_flags & MDBX_WRITEMAP) { - np = mp; - } else { - np = mdbx_page_malloc(txn, num); - if (unlikely(!np)) - return MDBX_ENOMEM; - if (unlikely(num > 1)) - memcpy(np, mp, pgno2bytes(env, num)); - else - mdbx_page_copy(np, mp, env->me_psize); - } - mdbx_debug("unspill page %" PRIaPGNO, mp->mp_pgno); - if (tx2 == txn) { - /* If in current txn, this page is no longer spilled. - * If it happens to be the last page, truncate the spill list. - * Otherwise mark it as deleted by setting the LSB. */ - if (x == txn->mt_spill_pages[0]) - txn->mt_spill_pages[0]--; - else - txn->mt_spill_pages[x] |= 1; - } /* otherwise, if belonging to a parent txn, the - * page remains spilled until child commits */ - - mdbx_page_dirty(txn, np); - np->mp_flags |= P_DIRTY; - *ret = np; - break; - } - } - return MDBX_SUCCESS; -} - -/* Touch a page: make it dirty and re-insert into tree with updated pgno. - * Set MDBX_TXN_ERROR on failure. - * - * [in] mc cursor pointing to the page to be touched - * - * Returns 0 on success, non-zero on failure. */ -static int mdbx_page_touch(MDBX_cursor *mc) { - MDBX_page *mp = mc->mc_pg[mc->mc_top], *np; - MDBX_txn *txn = mc->mc_txn; - MDBX_cursor *m2, *m3; - pgno_t pgno; - int rc; - - mdbx_cassert(mc, !IS_OVERFLOW(mp)); - if (!F_ISSET(mp->mp_flags, P_DIRTY)) { - if (txn->mt_flags & MDBX_TXN_SPILLS) { - np = NULL; - rc = mdbx_page_unspill(txn, mp, &np); - if (unlikely(rc)) - goto fail; - if (likely(np)) - goto done; - } - - if (unlikely((rc = mdbx_pnl_need(&txn->mt_befree_pages, 1)) || - (rc = mdbx_page_alloc(mc, 1, &np, MDBX_ALLOC_ALL)))) - goto fail; - pgno = np->mp_pgno; - mdbx_debug("touched db %d page %" PRIaPGNO " -> %" PRIaPGNO, DDBI(mc), - mp->mp_pgno, pgno); - mdbx_cassert(mc, mp->mp_pgno != pgno); - mdbx_pnl_xappend(txn->mt_befree_pages, mp->mp_pgno); - /* Update the parent page, if any, to point to the new page */ - if (mc->mc_top) { - MDBX_page *parent = mc->mc_pg[mc->mc_top - 1]; - MDBX_node *node = NODEPTR(parent, mc->mc_ki[mc->mc_top - 1]); - SETPGNO(node, pgno); - } else { - mc->mc_db->md_root = pgno; - } - } else if (txn->mt_parent && !IS_SUBP(mp)) { - mdbx_tassert(txn, (txn->mt_env->me_flags & MDBX_WRITEMAP) == 0); - MDBX_ID2 mid, *dl = txn->mt_rw_dirtylist; - pgno = mp->mp_pgno; - /* If txn has a parent, make sure the page is in our dirty list. */ - if (dl[0].mid) { - unsigned x = mdbx_mid2l_search(dl, pgno); - if (x <= dl[0].mid && dl[x].mid == pgno) { - if (unlikely(mp != dl[x].mptr)) { /* bad cursor? */ - mdbx_error("wrong page 0x%p #%" PRIaPGNO - " in the dirtylist[%d], expecting %p", - dl[x].mptr, pgno, x, mp); - mc->mc_flags &= ~(C_INITIALIZED | C_EOF); - txn->mt_flags |= MDBX_TXN_ERROR; - return MDBX_PROBLEM; - } - return MDBX_SUCCESS; - } - } - - mdbx_debug("clone db %d page %" PRIaPGNO, DDBI(mc), mp->mp_pgno); - mdbx_cassert(mc, dl[0].mid < MDBX_PNL_UM_MAX); - /* No - copy it */ - np = mdbx_page_malloc(txn, 1); - if (unlikely(!np)) - return MDBX_ENOMEM; - mid.mid = pgno; - mid.mptr = np; - rc = mdbx_mid2l_insert(dl, &mid); - mdbx_cassert(mc, rc == 0); - } else { - return MDBX_SUCCESS; - } - - mdbx_page_copy(np, mp, txn->mt_env->me_psize); - np->mp_pgno = pgno; - np->mp_flags |= P_DIRTY; - -done: - /* Adjust cursors pointing to mp */ - mc->mc_pg[mc->mc_top] = np; - m2 = txn->mt_cursors[mc->mc_dbi]; - if (mc->mc_flags & C_SUB) { - for (; m2; m2 = m2->mc_next) { - m3 = &m2->mc_xcursor->mx_cursor; - if (m3->mc_snum < mc->mc_snum) - continue; - if (m3->mc_pg[mc->mc_top] == mp) - m3->mc_pg[mc->mc_top] = np; - } - } else { - for (; m2; m2 = m2->mc_next) { - if (m2->mc_snum < mc->mc_snum) - continue; - if (m2 == mc) - continue; - if (m2->mc_pg[mc->mc_top] == mp) { - m2->mc_pg[mc->mc_top] = np; - if (XCURSOR_INITED(m2) && IS_LEAF(np)) - XCURSOR_REFRESH(m2, np, m2->mc_ki[mc->mc_top]); - } - } - } - return MDBX_SUCCESS; - -fail: - txn->mt_flags |= MDBX_TXN_ERROR; - return rc; -} - -int mdbx_env_sync(MDBX_env *env, int force) { - if (unlikely(!env)) - return MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - unsigned flags = env->me_flags & ~MDBX_NOMETASYNC; - if (unlikely(flags & (MDBX_RDONLY | MDBX_FATAL_ERROR))) - return MDBX_EACCESS; - - if (unlikely(!env->me_lck)) - return MDBX_PANIC; - - const bool outside_txn = - (!env->me_txn0 || env->me_txn0->mt_owner != mdbx_thread_self()); - - if (outside_txn) { - int rc = mdbx_txn_lock(env, false); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - } - - MDBX_meta *head = mdbx_meta_head(env); - if (!META_IS_STEADY(head) || env->me_sync_pending) { - - if (force || (env->me_sync_threshold && - env->me_sync_pending >= env->me_sync_threshold)) - flags &= MDBX_WRITEMAP /* clear flags for full steady sync */; - - if (outside_txn && - env->me_sync_pending > - pgno2bytes(env, 16 /* FIXME: define threshold */) && - (flags & MDBX_NOSYNC) == 0) { - assert(((flags ^ env->me_flags) & MDBX_WRITEMAP) == 0); - const size_t usedbytes = pgno_align2os_bytes(env, head->mm_geo.next); - - mdbx_txn_unlock(env); - - /* LY: pre-sync without holding lock to reduce latency for writer(s) */ - int rc = (flags & MDBX_WRITEMAP) - ? mdbx_msync(&env->me_dxb_mmap, 0, usedbytes, - flags & MDBX_MAPASYNC) - : mdbx_filesync(env->me_fd, false); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - rc = mdbx_txn_lock(env, false); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - /* LY: head may be changed. */ - head = mdbx_meta_head(env); - } - - if (!META_IS_STEADY(head) || env->me_sync_pending) { - mdbx_debug("meta-head %" PRIaPGNO ", %s, sync_pending %" PRIuPTR, - container_of(head, MDBX_page, mp_data)->mp_pgno, - mdbx_durable_str(head), env->me_sync_pending); - MDBX_meta meta = *head; - int rc = mdbx_sync_locked(env, flags | MDBX_SHRINK_ALLOWED, &meta); - if (unlikely(rc != MDBX_SUCCESS)) { - if (outside_txn) - mdbx_txn_unlock(env); - return rc; - } - } - } - - if (outside_txn) - mdbx_txn_unlock(env); - return MDBX_SUCCESS; -} - -/* Back up parent txn's cursors, then grab the originals for tracking */ -static int mdbx_cursor_shadow(MDBX_txn *src, MDBX_txn *dst) { - MDBX_cursor *mc, *bk; - MDBX_xcursor *mx; - size_t size; - int i; - - for (i = src->mt_numdbs; --i >= 0;) { - if ((mc = src->mt_cursors[i]) != NULL) { - size = sizeof(MDBX_cursor); - if (mc->mc_xcursor) - size += sizeof(MDBX_xcursor); - for (; mc; mc = bk->mc_next) { - bk = malloc(size); - if (unlikely(!bk)) - return MDBX_ENOMEM; - *bk = *mc; - mc->mc_backup = bk; - mc->mc_db = &dst->mt_dbs[i]; - /* Kill pointers into src to reduce abuse: The - * user may not use mc until dst ends. But we need a valid - * txn pointer here for cursor fixups to keep working. */ - mc->mc_txn = dst; - mc->mc_dbflag = &dst->mt_dbflags[i]; - if ((mx = mc->mc_xcursor) != NULL) { - *(MDBX_xcursor *)(bk + 1) = *mx; - mx->mx_cursor.mc_txn = dst; - } - mc->mc_next = dst->mt_cursors[i]; - dst->mt_cursors[i] = mc; - } - } - } - return MDBX_SUCCESS; -} - -/* Close this write txn's cursors, give parent txn's cursors back to parent. - * - * [in] txn the transaction handle. - * [in] merge true to keep changes to parent cursors, false to revert. - * - * Returns 0 on success, non-zero on failure. */ -static void mdbx_cursors_eot(MDBX_txn *txn, unsigned merge) { - MDBX_cursor **cursors = txn->mt_cursors, *mc, *next, *bk; - MDBX_xcursor *mx; - int i; - - for (i = txn->mt_numdbs; --i >= 0;) { - for (mc = cursors[i]; mc; mc = next) { - unsigned stage = mc->mc_signature; - mdbx_ensure(NULL, - stage == MDBX_MC_SIGNATURE || stage == MDBX_MC_WAIT4EOT); - next = mc->mc_next; - if ((bk = mc->mc_backup) != NULL) { - if (merge) { - /* Commit changes to parent txn */ - mc->mc_next = bk->mc_next; - mc->mc_backup = bk->mc_backup; - mc->mc_txn = bk->mc_txn; - mc->mc_db = bk->mc_db; - mc->mc_dbflag = bk->mc_dbflag; - if ((mx = mc->mc_xcursor) != NULL) - mx->mx_cursor.mc_txn = bk->mc_txn; - } else { - /* Abort nested txn */ - *mc = *bk; - if ((mx = mc->mc_xcursor) != NULL) - *mx = *(MDBX_xcursor *)(bk + 1); - } - bk->mc_signature = 0; - free(bk); - } - if (stage == MDBX_MC_WAIT4EOT) { - mc->mc_signature = 0; - free(mc); - } else { - mc->mc_signature = MDBX_MC_READY4CLOSE; - mc->mc_flags = 0 /* reset C_UNTRACK */; - } - } - cursors[i] = NULL; - } -} - -/* Common code for mdbx_txn_begin() and mdbx_txn_renew(). */ -static int mdbx_txn_renew0(MDBX_txn *txn, unsigned flags) { - MDBX_env *env = txn->mt_env; - int rc; - - if (unlikely(env->me_pid != mdbx_getpid())) { - env->me_flags |= MDBX_FATAL_ERROR; - return MDBX_PANIC; - } - - pgno_t upper_pgno = 0; - if (flags & MDBX_TXN_RDONLY) { - txn->mt_flags = MDBX_TXN_RDONLY; - MDBX_reader *r = txn->mt_ro_reader; - if (likely(env->me_flags & MDBX_ENV_TXKEY)) { - mdbx_assert(env, !(env->me_flags & MDBX_NOTLS)); - r = mdbx_thread_rthc_get(env->me_txkey); - if (likely(r)) { - mdbx_assert(env, r->mr_pid == env->me_pid); - mdbx_assert(env, r->mr_tid == mdbx_thread_self()); - } - } else { - mdbx_assert(env, !env->me_lck || (env->me_flags & MDBX_NOTLS)); - } - - if (likely(r)) { - if (unlikely(r->mr_pid != env->me_pid || r->mr_txnid != ~(txnid_t)0)) - return MDBX_BAD_RSLOT; - } else if (env->me_lck) { - unsigned slot, nreaders; - const mdbx_pid_t pid = env->me_pid; - const mdbx_tid_t tid = mdbx_thread_self(); - mdbx_assert(env, env->me_lck->mti_magic_and_version == MDBX_LOCK_MAGIC); - mdbx_assert(env, env->me_lck->mti_os_and_format == MDBX_LOCK_FORMAT); - - rc = mdbx_rdt_lock(env); - if (unlikely(MDBX_IS_ERROR(rc))) - return rc; - rc = MDBX_SUCCESS; - - if (unlikely(env->me_live_reader != pid)) { - rc = mdbx_rpid_set(env); - if (unlikely(rc != MDBX_SUCCESS)) { - mdbx_rdt_unlock(env); - return rc; - } - env->me_live_reader = pid; - } - - while (1) { - nreaders = env->me_lck->mti_numreaders; - for (slot = 0; slot < nreaders; slot++) - if (env->me_lck->mti_readers[slot].mr_pid == 0) - break; - - if (likely(slot < env->me_maxreaders)) - break; - - rc = mdbx_reader_check0(env, true, NULL); - if (rc != MDBX_RESULT_TRUE) { - mdbx_rdt_unlock(env); - return (rc == MDBX_SUCCESS) ? MDBX_READERS_FULL : rc; - } - } - - STATIC_ASSERT(sizeof(MDBX_reader) == MDBX_CACHELINE_SIZE); - STATIC_ASSERT( - offsetof(MDBX_lockinfo, mti_readers) % MDBX_CACHELINE_SIZE == 0); - r = &env->me_lck->mti_readers[slot]; - /* Claim the reader slot, carefully since other code - * uses the reader table un-mutexed: First reset the - * slot, next publish it in mtb.mti_numreaders. After - * that, it is safe for mdbx_env_close() to touch it. - * When it will be closed, we can finally claim it. */ - r->mr_pid = 0; - r->mr_txnid = ~(txnid_t)0; - r->mr_tid = tid; - mdbx_coherent_barrier(); - if (slot == nreaders) - env->me_lck->mti_numreaders = ++nreaders; - if (env->me_close_readers < nreaders) - env->me_close_readers = nreaders; - r->mr_pid = pid; - mdbx_rdt_unlock(env); - - if (likely(env->me_flags & MDBX_ENV_TXKEY)) - mdbx_thread_rthc_set(env->me_txkey, r); - } - - while (1) { - MDBX_meta *const meta = mdbx_meta_head(env); - mdbx_jitter4testing(false); - const txnid_t snap = mdbx_meta_txnid_fluid(env, meta); - mdbx_jitter4testing(false); - if (r) { - r->mr_txnid = snap; - mdbx_jitter4testing(false); - mdbx_assert(env, r->mr_pid == mdbx_getpid()); - //mdbx_assert(env, r->mr_tid == mdbx_thread_self()); - mdbx_assert(env, r->mr_txnid == snap); - mdbx_coherent_barrier(); - env->me_lck->mti_readers_refresh_flag = true; - } - mdbx_jitter4testing(true); - - /* Snap the state from current meta-head */ - txn->mt_txnid = snap; - txn->mt_next_pgno = meta->mm_geo.next; - txn->mt_end_pgno = meta->mm_geo.now; - upper_pgno = meta->mm_geo.upper; - memcpy(txn->mt_dbs, meta->mm_dbs, CORE_DBS * sizeof(MDBX_db)); - txn->mt_canary = meta->mm_canary; - - /* LY: Retry on a race, ITS#7970. */ - mdbx_compiler_barrier(); - if (likely(meta == mdbx_meta_head(env) && - snap == mdbx_meta_txnid_fluid(env, meta) && - snap >= *env->me_oldest)) { - mdbx_jitter4testing(false); - break; - } - } - - if (unlikely(txn->mt_txnid == 0)) { - mdbx_error("environment corrupted by died writer, must shutdown!"); - rc = MDBX_WANNA_RECOVERY; - goto bailout; - } - mdbx_assert(env, txn->mt_txnid >= *env->me_oldest); - txn->mt_ro_reader = r; - txn->mt_dbxs = env->me_dbxs; /* mostly static anyway */ - } else { - /* Not yet touching txn == env->me_txn0, it may be active */ - mdbx_jitter4testing(false); - rc = mdbx_txn_lock(env, F_ISSET(flags, MDBX_TRYTXN)); - if (unlikely(rc)) - return rc; - - mdbx_jitter4testing(false); - MDBX_meta *meta = mdbx_meta_head(env); - mdbx_jitter4testing(false); - txn->mt_canary = meta->mm_canary; - const txnid_t snap = mdbx_meta_txnid_stable(env, meta); - txn->mt_txnid = snap + 1; - if (unlikely(txn->mt_txnid < snap)) { - mdbx_debug("txnid overflow!"); - rc = MDBX_TXN_FULL; - goto bailout; - } - - txn->mt_flags = flags; - txn->mt_child = NULL; - txn->mt_loose_pages = NULL; - txn->mt_loose_count = 0; - txn->mt_dirtyroom = MDBX_PNL_UM_MAX; - txn->mt_rw_dirtylist = env->me_dirtylist; - txn->mt_rw_dirtylist[0].mid = 0; - txn->mt_befree_pages = env->me_free_pgs; - txn->mt_befree_pages[0] = 0; - txn->mt_spill_pages = NULL; - if (txn->mt_lifo_reclaimed) - txn->mt_lifo_reclaimed[0] = 0; - env->me_txn = txn; - memcpy(txn->mt_dbiseqs, env->me_dbiseqs, env->me_maxdbs * sizeof(unsigned)); - /* Copy the DB info and flags */ - memcpy(txn->mt_dbs, meta->mm_dbs, CORE_DBS * sizeof(MDBX_db)); - /* Moved to here to avoid a data race in read TXNs */ - txn->mt_next_pgno = meta->mm_geo.next; - txn->mt_end_pgno = meta->mm_geo.now; - upper_pgno = meta->mm_geo.upper; - } - - /* Setup db info */ - txn->mt_numdbs = env->me_numdbs; - for (unsigned i = CORE_DBS; i < txn->mt_numdbs; i++) { - unsigned x = env->me_dbflags[i]; - txn->mt_dbs[i].md_flags = x & PERSISTENT_FLAGS; - txn->mt_dbflags[i] = - (x & MDBX_VALID) ? DB_VALID | DB_USRVALID | DB_STALE : 0; - } - txn->mt_dbflags[MAIN_DBI] = DB_VALID | DB_USRVALID; - txn->mt_dbflags[FREE_DBI] = DB_VALID; - - if (unlikely(env->me_flags & MDBX_FATAL_ERROR)) { - mdbx_warning("environment had fatal error, must shutdown!"); - rc = MDBX_PANIC; - } else { - const size_t size = pgno2bytes(env, txn->mt_end_pgno); - if (unlikely(size > env->me_mapsize)) { - if (upper_pgno > MAX_PAGENO || - bytes2pgno(env, pgno2bytes(env, upper_pgno)) != upper_pgno) { - rc = MDBX_MAP_RESIZED; - goto bailout; - } - rc = mdbx_mapresize(env, txn->mt_end_pgno, upper_pgno); - if (rc != MDBX_SUCCESS) - goto bailout; - } - txn->mt_owner = mdbx_thread_self(); - return MDBX_SUCCESS; - } -bailout: - assert(rc != MDBX_SUCCESS); - mdbx_txn_end(txn, MDBX_END_SLOT | MDBX_END_FAIL_BEGIN); - return rc; -} - -int mdbx_txn_renew(MDBX_txn *txn) { - int rc; - - if (unlikely(!txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(!F_ISSET(txn->mt_flags, MDBX_TXN_RDONLY | MDBX_TXN_FINISHED))) - return MDBX_EINVAL; - - if (unlikely(txn->mt_owner != 0)) - return MDBX_THREAD_MISMATCH; - - rc = mdbx_txn_renew0(txn, MDBX_TXN_RDONLY); - if (rc == MDBX_SUCCESS) { - txn->mt_owner = mdbx_thread_self(); - mdbx_debug("renew txn %" PRIaTXN "%c %p on env %p, root page %" PRIaPGNO - "/%" PRIaPGNO, - txn->mt_txnid, (txn->mt_flags & MDBX_TXN_RDONLY) ? 'r' : 'w', - (void *)txn, (void *)txn->mt_env, txn->mt_dbs[MAIN_DBI].md_root, - txn->mt_dbs[FREE_DBI].md_root); - } - return rc; -} - -int mdbx_txn_begin(MDBX_env *env, MDBX_txn *parent, unsigned flags, - MDBX_txn **ret) { - MDBX_txn *txn; - MDBX_ntxn *ntxn; - int rc, size, tsize; - - if (unlikely(!env || !ret)) - return MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(env->me_pid != mdbx_getpid())) { - env->me_flags |= MDBX_FATAL_ERROR; - return MDBX_PANIC; - } - - if (unlikely(!env->me_map)) - return MDBX_EPERM; - - flags &= MDBX_TXN_BEGIN_FLAGS; - flags |= env->me_flags & MDBX_WRITEMAP; - - if (unlikely(env->me_flags & MDBX_RDONLY & - ~flags)) /* write txn in RDONLY env */ - return MDBX_EACCESS; - - if (parent) { - if (unlikely(parent->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EINVAL; - - if (unlikely(parent->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - /* Nested transactions: Max 1 child, write txns only, no writemap */ - flags |= parent->mt_flags; - if (unlikely(flags & (MDBX_RDONLY | MDBX_WRITEMAP | MDBX_TXN_BLOCKED))) { - return (parent->mt_flags & MDBX_TXN_RDONLY) ? MDBX_EINVAL : MDBX_BAD_TXN; - } - /* Child txns save MDBX_pgstate and use own copy of cursors */ - size = env->me_maxdbs * (sizeof(MDBX_db) + sizeof(MDBX_cursor *) + 1); - size += tsize = sizeof(MDBX_ntxn); - } else if (flags & MDBX_RDONLY) { - size = env->me_maxdbs * (sizeof(MDBX_db) + 1); - size += tsize = sizeof(MDBX_txn); - } else { - /* Reuse preallocated write txn. However, do not touch it until - * mdbx_txn_renew0() succeeds, since it currently may be active. */ - txn = env->me_txn0; - goto renew; - } - if (unlikely((txn = calloc(1, size)) == NULL)) { - mdbx_debug("calloc: %s", "failed"); - return MDBX_ENOMEM; - } - txn->mt_dbxs = env->me_dbxs; /* static */ - txn->mt_dbs = (MDBX_db *)((char *)txn + tsize); - txn->mt_dbflags = (uint8_t *)txn + size - env->me_maxdbs; - txn->mt_flags = flags; - txn->mt_env = env; - - if (parent) { - unsigned i; - txn->mt_cursors = (MDBX_cursor **)(txn->mt_dbs + env->me_maxdbs); - txn->mt_dbiseqs = parent->mt_dbiseqs; - txn->mt_rw_dirtylist = malloc(sizeof(MDBX_ID2) * MDBX_PNL_UM_SIZE); - if (!txn->mt_rw_dirtylist || - !(txn->mt_befree_pages = mdbx_pnl_alloc(MDBX_PNL_UM_MAX))) { - free(txn->mt_rw_dirtylist); - free(txn); - return MDBX_ENOMEM; - } - txn->mt_txnid = parent->mt_txnid; - txn->mt_dirtyroom = parent->mt_dirtyroom; - txn->mt_rw_dirtylist[0].mid = 0; - txn->mt_spill_pages = NULL; - txn->mt_next_pgno = parent->mt_next_pgno; - txn->mt_end_pgno = parent->mt_end_pgno; - parent->mt_flags |= MDBX_TXN_HAS_CHILD; - parent->mt_child = txn; - txn->mt_parent = parent; - txn->mt_numdbs = parent->mt_numdbs; - memcpy(txn->mt_dbs, parent->mt_dbs, txn->mt_numdbs * sizeof(MDBX_db)); - /* Copy parent's mt_dbflags, but clear DB_NEW */ - for (i = 0; i < txn->mt_numdbs; i++) - txn->mt_dbflags[i] = parent->mt_dbflags[i] & ~DB_NEW; - rc = 0; - ntxn = (MDBX_ntxn *)txn; - ntxn->mnt_pgstate = - env->me_pgstate; /* save parent me_reclaimed_pglist & co */ - if (env->me_reclaimed_pglist) { - size = MDBX_PNL_SIZEOF(env->me_reclaimed_pglist); - env->me_reclaimed_pglist = mdbx_pnl_alloc(env->me_reclaimed_pglist[0]); - if (likely(env->me_reclaimed_pglist)) - memcpy(env->me_reclaimed_pglist, ntxn->mnt_pgstate.mf_reclaimed_pglist, - size); - else - rc = MDBX_ENOMEM; - } - if (likely(!rc)) - rc = mdbx_cursor_shadow(parent, txn); - if (unlikely(rc)) - mdbx_txn_end(txn, MDBX_END_FAIL_BEGINCHILD); - } else { /* MDBX_RDONLY */ - txn->mt_dbiseqs = env->me_dbiseqs; - renew: - rc = mdbx_txn_renew0(txn, flags); - } - - if (unlikely(rc)) { - if (txn != env->me_txn0) - free(txn); - } else { - txn->mt_signature = MDBX_MT_SIGNATURE; - *ret = txn; - mdbx_debug("begin txn %" PRIaTXN "%c %p on env %p, root page %" PRIaPGNO - "/%" PRIaPGNO, - txn->mt_txnid, (flags & MDBX_RDONLY) ? 'r' : 'w', (void *)txn, - (void *)env, txn->mt_dbs[MAIN_DBI].md_root, - txn->mt_dbs[FREE_DBI].md_root); - } - - return rc; -} - -MDBX_env *mdbx_txn_env(MDBX_txn *txn) { - if (unlikely(!txn || txn->mt_signature != MDBX_MT_SIGNATURE)) - return NULL; - return txn->mt_env; -} - -uint64_t mdbx_txn_id(MDBX_txn *txn) { - if (unlikely(!txn || txn->mt_signature != MDBX_MT_SIGNATURE)) - return ~(txnid_t)0; - return txn->mt_txnid; -} - -/* Export or close DBI handles opened in this txn. */ -static void mdbx_dbis_update(MDBX_txn *txn, int keep) { - MDBX_dbi n = txn->mt_numdbs; - if (n) { - MDBX_env *env = txn->mt_env; - uint8_t *tdbflags = txn->mt_dbflags; - - for (unsigned i = n; --i >= CORE_DBS;) { - if (tdbflags[i] & DB_NEW) { - if (keep) { - env->me_dbflags[i] = txn->mt_dbs[i].md_flags | MDBX_VALID; - } else { - char *ptr = env->me_dbxs[i].md_name.iov_base; - if (ptr) { - env->me_dbxs[i].md_name.iov_base = NULL; - env->me_dbxs[i].md_name.iov_len = 0; - env->me_dbflags[i] = 0; - env->me_dbiseqs[i]++; - free(ptr); - } - } - } - } - if (keep && env->me_numdbs < n) - env->me_numdbs = n; - } -} - -/* End a transaction, except successful commit of a nested transaction. - * May be called twice for readonly txns: First reset it, then abort. - * [in] txn the transaction handle to end - * [in] mode why and how to end the transaction */ -static int mdbx_txn_end(MDBX_txn *txn, unsigned mode) { - MDBX_env *env = txn->mt_env; - static const char *const names[] = MDBX_END_NAMES; - - if (unlikely(txn->mt_env->me_pid != mdbx_getpid())) { - env->me_flags |= MDBX_FATAL_ERROR; - return MDBX_PANIC; - } - - /* Export or close DBI handles opened in this txn */ - mdbx_dbis_update(txn, mode & MDBX_END_UPDATE); - - mdbx_debug("%s txn %" PRIaTXN "%c %p on mdbenv %p, root page %" PRIaPGNO - "/%" PRIaPGNO, - names[mode & MDBX_END_OPMASK], txn->mt_txnid, - (txn->mt_flags & MDBX_TXN_RDONLY) ? 'r' : 'w', (void *)txn, - (void *)env, txn->mt_dbs[MAIN_DBI].md_root, - txn->mt_dbs[FREE_DBI].md_root); - - if (F_ISSET(txn->mt_flags, MDBX_TXN_RDONLY)) { - if (txn->mt_ro_reader) { - txn->mt_ro_reader->mr_txnid = ~(txnid_t)0; - env->me_lck->mti_readers_refresh_flag = true; - if (mode & MDBX_END_SLOT) { - if ((env->me_flags & MDBX_ENV_TXKEY) == 0) - txn->mt_ro_reader->mr_pid = 0; - txn->mt_ro_reader = NULL; - } - } - mdbx_coherent_barrier(); - txn->mt_numdbs = 0; /* prevent further DBI activity */ - txn->mt_flags |= MDBX_TXN_FINISHED; - txn->mt_owner = 0; - } else if (!F_ISSET(txn->mt_flags, MDBX_TXN_FINISHED)) { - pgno_t *pghead = env->me_reclaimed_pglist; - - if (!(mode & MDBX_END_EOTDONE)) /* !(already closed cursors) */ - mdbx_cursors_eot(txn, 0); - if (!(env->me_flags & MDBX_WRITEMAP)) { - mdbx_dlist_free(txn); - } - - if (txn->mt_lifo_reclaimed) { - txn->mt_lifo_reclaimed[0] = 0; - if (txn != env->me_txn0) { - mdbx_txl_free(txn->mt_lifo_reclaimed); - txn->mt_lifo_reclaimed = NULL; - } - } - txn->mt_numdbs = 0; - txn->mt_flags = MDBX_TXN_FINISHED; - - if (!txn->mt_parent) { - mdbx_pnl_shrink(&txn->mt_befree_pages); - env->me_free_pgs = txn->mt_befree_pages; - /* me_pgstate: */ - env->me_reclaimed_pglist = NULL; - env->me_last_reclaimed = 0; - - env->me_txn = NULL; - txn->mt_owner = 0; - txn->mt_signature = 0; - mode = 0; /* txn == env->me_txn0, do not free() it */ - - /* The writer mutex was locked in mdbx_txn_begin. */ - mdbx_txn_unlock(env); - } else { - txn->mt_parent->mt_child = NULL; - txn->mt_parent->mt_flags &= ~MDBX_TXN_HAS_CHILD; - env->me_pgstate = ((MDBX_ntxn *)txn)->mnt_pgstate; - mdbx_pnl_free(txn->mt_befree_pages); - mdbx_pnl_free(txn->mt_spill_pages); - free(txn->mt_rw_dirtylist); - } - - mdbx_pnl_free(pghead); - } - - if (mode & MDBX_END_FREE) { - mdbx_ensure(env, txn != env->me_txn0); - txn->mt_owner = 0; - txn->mt_signature = 0; - free(txn); - } - - return MDBX_SUCCESS; -} - -int mdbx_txn_reset(MDBX_txn *txn) { - if (unlikely(!txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - /* This call is only valid for read-only txns */ - if (unlikely(!(txn->mt_flags & MDBX_TXN_RDONLY))) - return MDBX_EINVAL; - - /* LY: don't close DBI-handles in MDBX mode */ - int rc = mdbx_txn_end(txn, MDBX_END_RESET | MDBX_END_UPDATE); - if (rc == MDBX_SUCCESS) { - assert(txn->mt_signature == MDBX_MT_SIGNATURE); - assert(txn->mt_owner == 0); - } - return rc; -} - -int mdbx_txn_abort(MDBX_txn *txn) { - if (unlikely(!txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != - ((txn->mt_flags & MDBX_TXN_FINISHED) ? 0 : mdbx_thread_self()))) - return MDBX_THREAD_MISMATCH; - - if (F_ISSET(txn->mt_flags, MDBX_TXN_RDONLY)) - /* LY: don't close DBI-handles in MDBX mode */ - return mdbx_txn_end(txn, MDBX_END_ABORT | MDBX_END_UPDATE | MDBX_END_SLOT | - MDBX_END_FREE); - - if (txn->mt_child) - mdbx_txn_abort(txn->mt_child); - - return mdbx_txn_end(txn, MDBX_END_ABORT | MDBX_END_SLOT | MDBX_END_FREE); -} - -static __inline int mdbx_backlog_size(MDBX_txn *txn) { - int reclaimed = txn->mt_env->me_reclaimed_pglist - ? txn->mt_env->me_reclaimed_pglist[0] - : 0; - return reclaimed + txn->mt_loose_count + txn->mt_end_pgno - txn->mt_next_pgno; -} - -/* LY: Prepare a backlog of pages to modify FreeDB itself, - * while reclaiming is prohibited. It should be enough to prevent search - * in mdbx_page_alloc() during a deleting, when freeDB tree is unbalanced. */ -static int mdbx_prep_backlog(MDBX_txn *txn, MDBX_cursor *mc) { - /* LY: extra page(s) for b-tree rebalancing */ - const int extra = (txn->mt_env->me_flags & MDBX_LIFORECLAIM) ? 2 : 1; - - if (mdbx_backlog_size(txn) < mc->mc_db->md_depth + extra) { - int rc = mdbx_cursor_touch(mc); - if (unlikely(rc)) - return rc; - - int backlog; - while (unlikely((backlog = mdbx_backlog_size(txn)) < extra)) { - rc = mdbx_page_alloc(mc, 1, NULL, MDBX_ALLOC_GC); - if (unlikely(rc)) { - if (unlikely(rc != MDBX_NOTFOUND)) - return rc; - break; - } - } - } - - return MDBX_SUCCESS; -} - -/* Save the freelist as of this transaction to the freeDB. - * This changes the freelist. Keep trying until it stabilizes. */ -static int mdbx_freelist_save(MDBX_txn *txn) { - /* env->me_reclaimed_pglist[] can grow and shrink during this call. - * env->me_last_reclaimed and txn->mt_free_pages[] can only grow. - * Page numbers cannot disappear from txn->mt_free_pages[]. */ - MDBX_cursor mc; - MDBX_env *env = txn->mt_env; - int rc, more = 1; - txnid_t cleanup_reclaimed_id = 0, head_id = 0; - pgno_t befree_count = 0; - intptr_t head_room = 0, total_room = 0; - unsigned cleanup_reclaimed_pos = 0, refill_reclaimed_pos = 0; - const bool lifo = (env->me_flags & MDBX_LIFORECLAIM) != 0; - - mdbx_cursor_init(&mc, txn, FREE_DBI, NULL); - - /* MDBX_RESERVE cancels meminit in ovpage malloc (when no WRITEMAP) */ - const intptr_t clean_limit = - (env->me_flags & (MDBX_NOMEMINIT | MDBX_WRITEMAP)) ? SSIZE_MAX - : env->me_maxfree_1pg; - - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); -again_on_freelist_change: - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - while (1) { - /* Come back here after each Put() in case freelist changed */ - MDBX_val key, data; - - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - if (!lifo) { - /* If using records from freeDB which we have not yet deleted, - * now delete them and any we reserved for me_reclaimed_pglist. */ - while (cleanup_reclaimed_id < env->me_last_reclaimed) { - rc = mdbx_cursor_first(&mc, &key, NULL); - if (unlikely(rc)) - goto bailout; - rc = mdbx_prep_backlog(txn, &mc); - if (unlikely(rc)) - goto bailout; - cleanup_reclaimed_id = head_id = *(txnid_t *)key.iov_base; - total_room = head_room = 0; - more = 1; - mdbx_tassert(txn, cleanup_reclaimed_id <= env->me_last_reclaimed); - mc.mc_flags |= C_RECLAIMING; - rc = mdbx_cursor_del(&mc, 0); - mc.mc_flags ^= C_RECLAIMING; - if (unlikely(rc)) - goto bailout; - } - } else if (txn->mt_lifo_reclaimed) { - /* LY: cleanup reclaimed records. */ - while (cleanup_reclaimed_pos < txn->mt_lifo_reclaimed[0]) { - cleanup_reclaimed_id = txn->mt_lifo_reclaimed[++cleanup_reclaimed_pos]; - key.iov_base = &cleanup_reclaimed_id; - key.iov_len = sizeof(cleanup_reclaimed_id); - rc = mdbx_cursor_get(&mc, &key, NULL, MDBX_SET); - if (likely(rc != MDBX_NOTFOUND)) { - if (unlikely(rc)) - goto bailout; - rc = mdbx_prep_backlog(txn, &mc); - if (unlikely(rc)) - goto bailout; - mc.mc_flags |= C_RECLAIMING; - rc = mdbx_cursor_del(&mc, 0); - mc.mc_flags ^= C_RECLAIMING; - if (unlikely(rc)) - goto bailout; - } - } - } - - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - if (txn->mt_loose_pages) { - /* Return loose page numbers to me_reclaimed_pglist, - * though usually none are left at this point. - * The pages themselves remain in dirtylist. */ - if (unlikely(!env->me_reclaimed_pglist) && - !(lifo && env->me_last_reclaimed > 1)) { - /* Put loose page numbers in mt_free_pages, - * since unable to return them to me_reclaimed_pglist. */ - if (unlikely((rc = mdbx_pnl_need(&txn->mt_befree_pages, - txn->mt_loose_count)) != 0)) - return rc; - for (MDBX_page *mp = txn->mt_loose_pages; mp; mp = NEXT_LOOSE_PAGE(mp)) - mdbx_pnl_xappend(txn->mt_befree_pages, mp->mp_pgno); - } else { - /* Room for loose pages + temp PNL with same */ - if ((rc = mdbx_pnl_need(&env->me_reclaimed_pglist, - 2 * txn->mt_loose_count + 1)) != 0) - goto bailout; - MDBX_PNL loose = env->me_reclaimed_pglist + - MDBX_PNL_ALLOCLEN(env->me_reclaimed_pglist) - - txn->mt_loose_count; - unsigned count = 0; - for (MDBX_page *mp = txn->mt_loose_pages; mp; mp = NEXT_LOOSE_PAGE(mp)) - loose[++count] = mp->mp_pgno; - loose[0] = count; - mdbx_pnl_sort(loose); - mdbx_pnl_xmerge(env->me_reclaimed_pglist, loose); - } - - MDBX_ID2L dl = txn->mt_rw_dirtylist; - for (MDBX_page *mp = txn->mt_loose_pages; mp;) { - mdbx_tassert(txn, mp->mp_pgno < txn->mt_next_pgno); - mdbx_ensure(env, mp->mp_pgno >= NUM_METAS); - - unsigned s, d; - for (s = d = 0; ++s <= dl[0].mid;) - if (dl[s].mid != mp->mp_pgno) - dl[++d] = dl[s]; - - dl[0].mid -= 1; - mdbx_tassert(txn, dl[0].mid == d); - - MDBX_page *dp = mp; - mp = NEXT_LOOSE_PAGE(mp); - if ((env->me_flags & MDBX_WRITEMAP) == 0) - mdbx_dpage_free(env, dp); - } - - txn->mt_loose_pages = NULL; - txn->mt_loose_count = 0; - } - - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - if (env->me_reclaimed_pglist) { - /* Refund suitable pages into "unallocated" space */ - pgno_t tail = txn->mt_next_pgno; - pgno_t *const begin = env->me_reclaimed_pglist + 1; - pgno_t *const end = begin + env->me_reclaimed_pglist[0]; - pgno_t *higest; -#if MDBX_PNL_ASCENDING - for (higest = end; --higest >= begin;) { -#else - for (higest = begin; higest < end; ++higest) { -#endif /* MDBX_PNL sort-order */ - mdbx_tassert(txn, *higest >= NUM_METAS && *higest < tail); - if (*higest != tail - 1) - break; - tail -= 1; - } - if (tail != txn->mt_next_pgno) { -#if MDBX_PNL_ASCENDING - env->me_reclaimed_pglist[0] = (unsigned)(higest + 1 - begin); -#else - env->me_reclaimed_pglist[0] -= (unsigned)(higest - begin); - for (pgno_t *move = begin; higest < end; ++move, ++higest) - *move = *higest; -#endif /* MDBX_PNL sort-order */ - mdbx_info("refunded %" PRIaPGNO " pages: %" PRIaPGNO " -> %" PRIaPGNO, - tail - txn->mt_next_pgno, tail, txn->mt_next_pgno); - txn->mt_next_pgno = tail; - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - } - } - - /* Save the PNL of pages freed by this txn, to a single record */ - if (befree_count < txn->mt_befree_pages[0]) { - if (unlikely(!befree_count)) { - /* Make sure last page of freeDB is touched and on freelist */ - rc = mdbx_page_search(&mc, NULL, MDBX_PS_LAST | MDBX_PS_MODIFY); - if (unlikely(rc && rc != MDBX_NOTFOUND)) - goto bailout; - } - pgno_t *befree_pages = txn->mt_befree_pages; - /* Write to last page of freeDB */ - key.iov_len = sizeof(txn->mt_txnid); - key.iov_base = &txn->mt_txnid; - do { - befree_count = befree_pages[0]; - data.iov_len = MDBX_PNL_SIZEOF(befree_pages); - rc = mdbx_cursor_put(&mc, &key, &data, MDBX_RESERVE); - if (unlikely(rc)) - goto bailout; - /* Retry if mt_free_pages[] grew during the Put() */ - befree_pages = txn->mt_befree_pages; - } while (befree_count < befree_pages[0]); - - mdbx_pnl_sort(befree_pages); - memcpy(data.iov_base, befree_pages, data.iov_len); - - if (mdbx_debug_enabled(MDBX_DBG_EXTRA)) { - unsigned i = (unsigned)befree_pages[0]; - mdbx_debug_extra("PNL write txn %" PRIaTXN " root %" PRIaPGNO - " num %u, PNL", - txn->mt_txnid, txn->mt_dbs[FREE_DBI].md_root, i); - for (; i; i--) - mdbx_debug_extra_print(" %" PRIaPGNO "", befree_pages[i]); - mdbx_debug_extra_print("\n"); - } - continue; - } - - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - const intptr_t rpl_len = - (env->me_reclaimed_pglist ? env->me_reclaimed_pglist[0] : 0) + - txn->mt_loose_count; - if (rpl_len && refill_reclaimed_pos == 0) - refill_reclaimed_pos = 1; - - /* Reserve records for me_reclaimed_pglist[]. Split it if multi-page, - * to avoid searching freeDB for a page range. Use keys in - * range [1,me_last_reclaimed]: Smaller than txnid of oldest reader. */ - if (total_room >= rpl_len) { - if (total_room == rpl_len || --more < 0) - break; - } else if (head_room >= (intptr_t)env->me_maxfree_1pg && head_id > 1) { - /* Keep current record (overflow page), add a new one */ - head_id--; - refill_reclaimed_pos++; - head_room = 0; - } - - if (lifo) { - if (refill_reclaimed_pos > - (txn->mt_lifo_reclaimed ? txn->mt_lifo_reclaimed[0] : 0)) { - /* LY: need just a txn-id for save page list. */ - rc = mdbx_page_alloc(&mc, 0, NULL, MDBX_ALLOC_GC | MDBX_ALLOC_KICK); - if (likely(rc == 0)) - /* LY: ok, reclaimed from freedb. */ - continue; - if (unlikely(rc != MDBX_NOTFOUND)) - /* LY: other troubles... */ - goto bailout; - - /* LY: freedb is empty, will look any free txn-id in high2low order. */ - if (unlikely(env->me_last_reclaimed < 1)) { - /* LY: not any txn in the past of freedb. */ - rc = MDBX_MAP_FULL; - goto bailout; - } - - if (unlikely(!txn->mt_lifo_reclaimed)) { - txn->mt_lifo_reclaimed = mdbx_txl_alloc(); - if (unlikely(!txn->mt_lifo_reclaimed)) { - rc = MDBX_ENOMEM; - goto bailout; - } - } - /* LY: append the list. */ - rc = mdbx_txl_append(&txn->mt_lifo_reclaimed, - env->me_last_reclaimed - 1); - if (unlikely(rc)) - goto bailout; - --env->me_last_reclaimed; - /* LY: note that freeDB cleanup is not needed. */ - ++cleanup_reclaimed_pos; - } - mdbx_tassert(txn, txn->mt_lifo_reclaimed != NULL); - head_id = txn->mt_lifo_reclaimed[refill_reclaimed_pos]; - } else { - mdbx_tassert(txn, txn->mt_lifo_reclaimed == NULL); - } - - /* (Re)write {key = head_id, PNL length = head_room} */ - total_room -= head_room; - head_room = rpl_len - total_room; - if (head_room > (intptr_t)env->me_maxfree_1pg && head_id > 1) { - /* Overflow multi-page for part of me_reclaimed_pglist */ - head_room /= (head_id < INT16_MAX) ? (pgno_t)head_id - : INT16_MAX; /* amortize page sizes */ - head_room += env->me_maxfree_1pg - head_room % (env->me_maxfree_1pg + 1); - } else if (head_room < 0) { - /* Rare case, not bothering to delete this record */ - head_room = 0; - continue; - } - key.iov_len = sizeof(head_id); - key.iov_base = &head_id; - data.iov_len = (head_room + 1) * sizeof(pgno_t); - rc = mdbx_cursor_put(&mc, &key, &data, MDBX_RESERVE); - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - if (unlikely(rc)) - goto bailout; - - /* PNL is initially empty, zero out at least the length */ - pgno_t *pgs = (pgno_t *)data.iov_base; - intptr_t i = head_room > clean_limit ? head_room : 0; - do { - pgs[i] = 0; - } while (--i >= 0); - total_room += head_room; - continue; - } - - mdbx_tassert(txn, - cleanup_reclaimed_pos == - (txn->mt_lifo_reclaimed ? txn->mt_lifo_reclaimed[0] : 0)); - - /* Fill in the reserved me_reclaimed_pglist records */ - rc = MDBX_SUCCESS; - mdbx_tassert(txn, mdbx_pnl_check(env->me_reclaimed_pglist)); - if (env->me_reclaimed_pglist && env->me_reclaimed_pglist[0]) { - MDBX_val key, data; - key.iov_len = data.iov_len = 0; /* avoid MSVC warning */ - key.iov_base = data.iov_base = NULL; - - size_t rpl_left = env->me_reclaimed_pglist[0]; - pgno_t *rpl_end = env->me_reclaimed_pglist + rpl_left; - if (txn->mt_lifo_reclaimed == 0) { - mdbx_tassert(txn, lifo == 0); - rc = mdbx_cursor_first(&mc, &key, &data); - if (unlikely(rc)) - goto bailout; - } else { - mdbx_tassert(txn, lifo != 0); - } - - while (1) { - txnid_t id; - if (txn->mt_lifo_reclaimed == 0) { - mdbx_tassert(txn, lifo == 0); - id = *(txnid_t *)key.iov_base; - mdbx_tassert(txn, id <= env->me_last_reclaimed); - } else { - mdbx_tassert(txn, lifo != 0); - mdbx_tassert(txn, - refill_reclaimed_pos > 0 && - refill_reclaimed_pos <= txn->mt_lifo_reclaimed[0]); - id = txn->mt_lifo_reclaimed[refill_reclaimed_pos--]; - key.iov_base = &id; - key.iov_len = sizeof(id); - rc = mdbx_cursor_get(&mc, &key, &data, MDBX_SET); - if (unlikely(rc)) - goto bailout; - } - mdbx_tassert( - txn, cleanup_reclaimed_pos == - (txn->mt_lifo_reclaimed ? txn->mt_lifo_reclaimed[0] : 0)); - - mdbx_tassert(txn, data.iov_len >= sizeof(pgno_t) * 2); - size_t chunk_len = (data.iov_len / sizeof(pgno_t)) - 1; - if (chunk_len > rpl_left) - chunk_len = rpl_left; - data.iov_len = (chunk_len + 1) * sizeof(pgno_t); - key.iov_base = &id; - key.iov_len = sizeof(id); - - rpl_end -= chunk_len; - data.iov_base = rpl_end; - pgno_t save = rpl_end[0]; - rpl_end[0] = (pgno_t)chunk_len; - mdbx_tassert(txn, mdbx_pnl_check(rpl_end)); - mc.mc_flags |= C_RECLAIMING; - rc = mdbx_cursor_put(&mc, &key, &data, MDBX_CURRENT); - mc.mc_flags ^= C_RECLAIMING; - mdbx_tassert(txn, mdbx_pnl_check(rpl_end)); - mdbx_tassert( - txn, cleanup_reclaimed_pos == - (txn->mt_lifo_reclaimed ? txn->mt_lifo_reclaimed[0] : 0)); - rpl_end[0] = save; - if (unlikely(rc)) - goto bailout; - - rpl_left -= chunk_len; - if (rpl_left == 0) - break; - - if (!lifo) { - rc = mdbx_cursor_next(&mc, &key, &data, MDBX_NEXT); - if (unlikely(rc)) - goto bailout; - } - } - } - -bailout: - if (txn->mt_lifo_reclaimed) { - mdbx_tassert(txn, rc || cleanup_reclaimed_pos == txn->mt_lifo_reclaimed[0]); - if (rc == MDBX_SUCCESS && - cleanup_reclaimed_pos != txn->mt_lifo_reclaimed[0]) { - mdbx_tassert(txn, cleanup_reclaimed_pos < txn->mt_lifo_reclaimed[0]); - /* LY: zeroed cleanup_idx to force cleanup - * and refill created freeDB records. */ - cleanup_reclaimed_pos = 0; - /* LY: restart filling */ - total_room = head_room = refill_reclaimed_pos = 0; - more = 1; - goto again_on_freelist_change; - } - txn->mt_lifo_reclaimed[0] = 0; - if (txn != env->me_txn0) { - mdbx_txl_free(txn->mt_lifo_reclaimed); - txn->mt_lifo_reclaimed = NULL; - } - } - - return rc; -} - -/* Flush (some) dirty pages to the map, after clearing their dirty flag. - * [in] txn the transaction that's being committed - * [in] keep number of initial pages in dirtylist to keep dirty. - * Returns 0 on success, non-zero on failure. */ -static int mdbx_page_flush(MDBX_txn *txn, pgno_t keep) { - MDBX_env *env = txn->mt_env; - MDBX_ID2L dl = txn->mt_rw_dirtylist; - unsigned i, j, pagecount = dl[0].mid; - int rc; - size_t size = 0, pos = 0; - pgno_t pgno = 0; - MDBX_page *dp = NULL; - struct iovec iov[MDBX_COMMIT_PAGES]; - intptr_t wpos = 0, wsize = 0; - size_t next_pos = 1; /* impossible pos, so pos != next_pos */ - int n = 0; - - j = i = keep; - - if (env->me_flags & MDBX_WRITEMAP) { - /* Clear dirty flags */ - while (++i <= pagecount) { - dp = dl[i].mptr; - /* Don't flush this page yet */ - if (dp->mp_flags & (P_LOOSE | P_KEEP)) { - dp->mp_flags &= ~P_KEEP; - dl[++j] = dl[i]; - continue; - } - dp->mp_flags &= ~P_DIRTY; - dp->mp_validator = 0 /* TODO */; - env->me_sync_pending += - IS_OVERFLOW(dp) ? pgno2bytes(env, dp->mp_pages) : env->me_psize; - } - goto done; - } - - /* Write the pages */ - for (;;) { - if (++i <= pagecount) { - dp = dl[i].mptr; - /* Don't flush this page yet */ - if (dp->mp_flags & (P_LOOSE | P_KEEP)) { - dp->mp_flags &= ~P_KEEP; - dl[i].mid = 0; - continue; - } - pgno = dl[i].mid; - /* clear dirty flag */ - dp->mp_flags &= ~P_DIRTY; - dp->mp_validator = 0 /* TODO */; - pos = pgno2bytes(env, pgno); - size = IS_OVERFLOW(dp) ? pgno2bytes(env, dp->mp_pages) : env->me_psize; - env->me_sync_pending += size; - } - /* Write up to MDBX_COMMIT_PAGES dirty pages at a time. */ - if (pos != next_pos || n == MDBX_COMMIT_PAGES || wsize + size > MAX_WRITE) { - if (n) { - /* Write previous page(s) */ - rc = mdbx_pwritev(env->me_fd, iov, n, wpos, wsize); - if (unlikely(rc != MDBX_SUCCESS)) { - mdbx_debug("Write error: %s", strerror(rc)); - return rc; - } - n = 0; - } - if (i > pagecount) - break; - wpos = pos; - wsize = 0; - } - mdbx_debug("committing page %" PRIaPGNO "", pgno); - next_pos = pos + size; - iov[n].iov_len = size; - iov[n].iov_base = (char *)dp; - wsize += size; - n++; - } - - mdbx_invalidate_cache(env->me_map, pgno2bytes(env, txn->mt_next_pgno)); - - for (i = keep; ++i <= pagecount;) { - dp = dl[i].mptr; - /* This is a page we skipped above */ - if (!dl[i].mid) { - dl[++j] = dl[i]; - dl[j].mid = dp->mp_pgno; - continue; - } - mdbx_dpage_free(env, dp); - } - -done: - i--; - txn->mt_dirtyroom += i - j; - dl[0].mid = j; - return MDBX_SUCCESS; -} - -int mdbx_txn_commit(MDBX_txn *txn) { - int rc; - - if (unlikely(txn == NULL)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - MDBX_env *env = txn->mt_env; - if (unlikely(env->me_pid != mdbx_getpid())) { - env->me_flags |= MDBX_FATAL_ERROR; - return MDBX_PANIC; - } - - if (txn->mt_child) { - rc = mdbx_txn_commit(txn->mt_child); - txn->mt_child = NULL; - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - } - - /* mdbx_txn_end() mode for a commit which writes nothing */ - unsigned end_mode = - MDBX_END_EMPTY_COMMIT | MDBX_END_UPDATE | MDBX_END_SLOT | MDBX_END_FREE; - if (unlikely(F_ISSET(txn->mt_flags, MDBX_TXN_RDONLY))) - goto done; - - if (unlikely(txn->mt_flags & (MDBX_TXN_FINISHED | MDBX_TXN_ERROR))) { - mdbx_debug("error flag is set, can't commit"); - if (txn->mt_parent) - txn->mt_parent->mt_flags |= MDBX_TXN_ERROR; - rc = MDBX_BAD_TXN; - goto fail; - } - - if (txn->mt_parent) { - MDBX_txn *parent = txn->mt_parent; - MDBX_page **lp; - MDBX_ID2L dst, src; - MDBX_PNL pspill; - unsigned i, x, y, len, ps_len; - - /* Append our reclaim list to parent's */ - if (txn->mt_lifo_reclaimed) { - if (parent->mt_lifo_reclaimed) { - rc = mdbx_txl_append_list(&parent->mt_lifo_reclaimed, - txn->mt_lifo_reclaimed); - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - mdbx_txl_free(txn->mt_lifo_reclaimed); - } else - parent->mt_lifo_reclaimed = txn->mt_lifo_reclaimed; - txn->mt_lifo_reclaimed = NULL; - } - - /* Append our free list to parent's */ - rc = mdbx_pnl_append_list(&parent->mt_befree_pages, txn->mt_befree_pages); - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - mdbx_pnl_free(txn->mt_befree_pages); - /* Failures after this must either undo the changes - * to the parent or set MDBX_TXN_ERROR in the parent. */ - - parent->mt_next_pgno = txn->mt_next_pgno; - parent->mt_end_pgno = txn->mt_end_pgno; - parent->mt_flags = txn->mt_flags; - - /* Merge our cursors into parent's and close them */ - mdbx_cursors_eot(txn, 1); - - /* Update parent's DB table. */ - memcpy(parent->mt_dbs, txn->mt_dbs, txn->mt_numdbs * sizeof(MDBX_db)); - parent->mt_numdbs = txn->mt_numdbs; - parent->mt_dbflags[FREE_DBI] = txn->mt_dbflags[FREE_DBI]; - parent->mt_dbflags[MAIN_DBI] = txn->mt_dbflags[MAIN_DBI]; - for (i = CORE_DBS; i < txn->mt_numdbs; i++) { - /* preserve parent's DB_NEW status */ - parent->mt_dbflags[i] = - txn->mt_dbflags[i] | (parent->mt_dbflags[i] & DB_NEW); - } - - dst = parent->mt_rw_dirtylist; - src = txn->mt_rw_dirtylist; - /* Remove anything in our dirty list from parent's spill list */ - if ((pspill = parent->mt_spill_pages) && (ps_len = pspill[0])) { - x = y = ps_len; - pspill[0] = (pgno_t)-1; - /* Mark our dirty pages as deleted in parent spill list */ - for (i = 0, len = src[0].mid; ++i <= len;) { - pgno_t pn = src[i].mid << 1; - while (pn > pspill[x]) - x--; - if (pn == pspill[x]) { - pspill[x] = 1; - y = --x; - } - } - /* Squash deleted pagenums if we deleted any */ - for (x = y; ++x <= ps_len;) - if (!(pspill[x] & 1)) - pspill[++y] = pspill[x]; - pspill[0] = y; - } - - /* Remove anything in our spill list from parent's dirty list */ - if (txn->mt_spill_pages && txn->mt_spill_pages[0]) { - for (i = 1; i <= txn->mt_spill_pages[0]; i++) { - pgno_t pn = txn->mt_spill_pages[i]; - if (pn & 1) - continue; /* deleted spillpg */ - pn >>= 1; - y = mdbx_mid2l_search(dst, pn); - if (y <= dst[0].mid && dst[y].mid == pn) { - free(dst[y].mptr); - while (y < dst[0].mid) { - dst[y] = dst[y + 1]; - y++; - } - dst[0].mid--; - } - } - } - - /* Find len = length of merging our dirty list with parent's */ - x = dst[0].mid; - dst[0].mid = 0; /* simplify loops */ - if (parent->mt_parent) { - len = x + src[0].mid; - y = mdbx_mid2l_search(src, dst[x].mid + 1) - 1; - for (i = x; y && i; y--) { - pgno_t yp = src[y].mid; - while (yp < dst[i].mid) - i--; - if (yp == dst[i].mid) { - i--; - len--; - } - } - } else { /* Simplify the above for single-ancestor case */ - len = MDBX_PNL_UM_MAX - txn->mt_dirtyroom; - } - /* Merge our dirty list with parent's */ - y = src[0].mid; - for (i = len; y; dst[i--] = src[y--]) { - pgno_t yp = src[y].mid; - while (yp < dst[x].mid) - dst[i--] = dst[x--]; - if (yp == dst[x].mid) - free(dst[x--].mptr); - } - mdbx_tassert(txn, i == x); - dst[0].mid = len; - free(txn->mt_rw_dirtylist); - parent->mt_dirtyroom = txn->mt_dirtyroom; - if (txn->mt_spill_pages) { - if (parent->mt_spill_pages) { - /* TODO: Prevent failure here, so parent does not fail */ - rc = mdbx_pnl_append_list(&parent->mt_spill_pages, txn->mt_spill_pages); - if (unlikely(rc != MDBX_SUCCESS)) - parent->mt_flags |= MDBX_TXN_ERROR; - mdbx_pnl_free(txn->mt_spill_pages); - mdbx_pnl_sort(parent->mt_spill_pages); - } else { - parent->mt_spill_pages = txn->mt_spill_pages; - } - } - - /* Append our loose page list to parent's */ - for (lp = &parent->mt_loose_pages; *lp; lp = &NEXT_LOOSE_PAGE(*lp)) - ; - *lp = txn->mt_loose_pages; - parent->mt_loose_count += txn->mt_loose_count; - - parent->mt_child = NULL; - mdbx_pnl_free(((MDBX_ntxn *)txn)->mnt_pgstate.mf_reclaimed_pglist); - txn->mt_signature = 0; - free(txn); - return rc; - } - - if (unlikely(txn != env->me_txn)) { - mdbx_debug("attempt to commit unknown transaction"); - rc = MDBX_EINVAL; - goto fail; - } - - mdbx_cursors_eot(txn, 0); - end_mode |= MDBX_END_EOTDONE; - - if (!txn->mt_rw_dirtylist[0].mid && - !(txn->mt_flags & (MDBX_TXN_DIRTY | MDBX_TXN_SPILLS))) - goto done; - - mdbx_debug("committing txn %" PRIaTXN " %p on mdbenv %p, root page %" PRIaPGNO - "/%" PRIaPGNO, - txn->mt_txnid, (void *)txn, (void *)env, - txn->mt_dbs[MAIN_DBI].md_root, txn->mt_dbs[FREE_DBI].md_root); - - /* Update DB root pointers */ - if (txn->mt_numdbs > CORE_DBS) { - MDBX_cursor mc; - MDBX_dbi i; - MDBX_val data; - data.iov_len = sizeof(MDBX_db); - - mdbx_cursor_init(&mc, txn, MAIN_DBI, NULL); - for (i = CORE_DBS; i < txn->mt_numdbs; i++) { - if (txn->mt_dbflags[i] & DB_DIRTY) { - if (unlikely(TXN_DBI_CHANGED(txn, i))) { - rc = MDBX_BAD_DBI; - goto fail; - } - data.iov_base = &txn->mt_dbs[i]; - rc = mdbx_cursor_put(&mc, &txn->mt_dbxs[i].md_name, &data, F_SUBDATA); - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - } - } - } - - rc = mdbx_freelist_save(txn); - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - - mdbx_pnl_free(env->me_reclaimed_pglist); - env->me_reclaimed_pglist = NULL; - mdbx_pnl_shrink(&txn->mt_befree_pages); - - if (mdbx_audit_enabled()) - mdbx_audit(txn); - - rc = mdbx_page_flush(txn, 0); - if (likely(rc == MDBX_SUCCESS)) { - MDBX_meta meta, *head = mdbx_meta_head(env); - - meta.mm_magic_and_version = head->mm_magic_and_version; - meta.mm_extra_flags = head->mm_extra_flags; - meta.mm_validator_id = head->mm_validator_id; - meta.mm_extra_pagehdr = head->mm_extra_pagehdr; - - meta.mm_geo = head->mm_geo; - meta.mm_geo.next = txn->mt_next_pgno; - meta.mm_geo.now = txn->mt_end_pgno; - meta.mm_dbs[FREE_DBI] = txn->mt_dbs[FREE_DBI]; - meta.mm_dbs[MAIN_DBI] = txn->mt_dbs[MAIN_DBI]; - meta.mm_canary = txn->mt_canary; - mdbx_meta_set_txnid(env, &meta, txn->mt_txnid); - - rc = mdbx_sync_locked( - env, env->me_flags | txn->mt_flags | MDBX_SHRINK_ALLOWED, &meta); - } - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - env->me_lck->mti_readers_refresh_flag = false; - end_mode = MDBX_END_COMMITTED | MDBX_END_UPDATE | MDBX_END_EOTDONE; - -done: - return mdbx_txn_end(txn, end_mode); - -fail: - mdbx_txn_abort(txn); - return rc; -} - -/* Read the environment parameters of a DB environment - * before mapping it into memory. */ -static int __cold mdbx_read_header(MDBX_env *env, MDBX_meta *meta) { - assert(offsetof(MDBX_page, mp_meta) == PAGEHDRSZ); - memset(meta, 0, sizeof(MDBX_meta)); - meta->mm_datasync_sign = MDBX_DATASIGN_WEAK; - int rc = MDBX_CORRUPTED; - - /* Read twice all meta pages so we can find the latest one. */ - unsigned loop_limit = NUM_METAS * 2; - for (unsigned loop_count = 0; loop_count < loop_limit; ++loop_count) { - MDBX_page page; - - /* We don't know the page size on first time. - * So, just guess it. */ - unsigned guess_pagesize = meta->mm_psize; - if (guess_pagesize == 0) - guess_pagesize = - (loop_count > NUM_METAS) ? env->me_psize : env->me_os_psize; - - const unsigned meta_number = loop_count % NUM_METAS; - const unsigned offset = guess_pagesize * meta_number; - - unsigned retryleft = 42; - while (1) { - int err = mdbx_pread(env->me_fd, &page, sizeof(page), offset); - if (err != MDBX_SUCCESS) { - mdbx_error("read meta[%u,%u]: %i, %s", offset, (unsigned)sizeof(page), - err, mdbx_strerror(err)); - return err; - } - - MDBX_page again; - err = mdbx_pread(env->me_fd, &again, sizeof(again), offset); - if (err != MDBX_SUCCESS) { - mdbx_error("read meta[%u,%u]: %i, %s", offset, (unsigned)sizeof(again), - err, mdbx_strerror(err)); - return err; - } - - if (memcmp(&page, &again, sizeof(page)) == 0 || --retryleft == 0) - break; - - mdbx_info("meta[%u] was updated, re-read it", meta_number); - } - - if (!retryleft) { - mdbx_error("meta[%u] is too volatile, skip it", meta_number); - continue; - } - - if (page.mp_pgno != meta_number) { - mdbx_error("meta[%u] has invalid pageno %" PRIaPGNO, meta_number, - page.mp_pgno); - return MDBX_INVALID; - } - - if (!F_ISSET(page.mp_flags, P_META)) { - mdbx_error("page #%u not a meta-page", meta_number); - return MDBX_INVALID; - } - - if (page.mp_meta.mm_magic_and_version != MDBX_DATA_MAGIC) { - mdbx_error("meta[%u] has invalid magic/version MDBX_DEVEL=%d", - meta_number, MDBX_DEVEL); - return ((page.mp_meta.mm_magic_and_version >> 8) != MDBX_MAGIC) - ? MDBX_INVALID - : MDBX_VERSION_MISMATCH; - } - - if (page.mp_meta.mm_txnid_a != page.mp_meta.mm_txnid_b) { - mdbx_warning("meta[%u] not completely updated, skip it", meta_number); - continue; - } - - /* LY: check signature as a checksum */ - if (META_IS_STEADY(&page.mp_meta) && - page.mp_meta.mm_datasync_sign != mdbx_meta_sign(&page.mp_meta)) { - mdbx_notice("meta[%u] has invalid steady-checksum (0x%" PRIx64 - " != 0x%" PRIx64 "), skip it", - meta_number, page.mp_meta.mm_datasync_sign, - mdbx_meta_sign(&page.mp_meta)); - continue; - } - - /* LY: check pagesize */ - if (!mdbx_is_power2(page.mp_meta.mm_psize) || - page.mp_meta.mm_psize < MIN_PAGESIZE || - page.mp_meta.mm_psize > MAX_PAGESIZE) { - mdbx_notice("meta[%u] has invalid pagesize (%u), skip it", meta_number, - page.mp_meta.mm_psize); - rc = MDBX_VERSION_MISMATCH; - continue; - } - - mdbx_debug("read meta%" PRIaPGNO " = root %" PRIaPGNO "/%" PRIaPGNO - ", geo %" PRIaPGNO "/%" PRIaPGNO "-%" PRIaPGNO "/%" PRIaPGNO - " +%u -%u, txn_id %" PRIaTXN ", %s", - page.mp_pgno, page.mp_meta.mm_dbs[MAIN_DBI].md_root, - page.mp_meta.mm_dbs[FREE_DBI].md_root, page.mp_meta.mm_geo.lower, - page.mp_meta.mm_geo.next, page.mp_meta.mm_geo.now, - page.mp_meta.mm_geo.upper, page.mp_meta.mm_geo.grow, - page.mp_meta.mm_geo.shrink, page.mp_meta.mm_txnid_a, - mdbx_durable_str(&page.mp_meta)); - - /* LY: check min-pages value */ - if (page.mp_meta.mm_geo.lower < MIN_PAGENO || - page.mp_meta.mm_geo.lower > MAX_PAGENO) { - mdbx_notice("meta[%u] has invalid min-pages (%" PRIaPGNO "), skip it", - meta_number, page.mp_meta.mm_geo.lower); - rc = MDBX_INVALID; - continue; - } - - /* LY: check max-pages value */ - if (page.mp_meta.mm_geo.upper < MIN_PAGENO || - page.mp_meta.mm_geo.upper > MAX_PAGENO || - page.mp_meta.mm_geo.upper < page.mp_meta.mm_geo.lower) { - mdbx_notice("meta[%u] has invalid max-pages (%" PRIaPGNO "), skip it", - meta_number, page.mp_meta.mm_geo.upper); - rc = MDBX_INVALID; - continue; - } - - /* LY: check mapsize limits */ - const uint64_t mapsize_min = - page.mp_meta.mm_geo.lower * (uint64_t)page.mp_meta.mm_psize; - const uint64_t mapsize_max = - page.mp_meta.mm_geo.upper * (uint64_t)page.mp_meta.mm_psize; - STATIC_ASSERT(MAX_MAPSIZE < SSIZE_MAX - MAX_PAGESIZE); - STATIC_ASSERT(MIN_MAPSIZE < MAX_MAPSIZE); - if (mapsize_min < MIN_MAPSIZE || mapsize_max > MAX_MAPSIZE) { - mdbx_notice("meta[%u] has invalid min-mapsize (%" PRIu64 "), skip it", - meta_number, mapsize_min); - rc = MDBX_VERSION_MISMATCH; - continue; - } - - STATIC_ASSERT(MIN_MAPSIZE < MAX_MAPSIZE); - if (mapsize_max > MAX_MAPSIZE || - MAX_PAGENO < mdbx_roundup2((size_t)mapsize_max, env->me_os_psize) / - (uint64_t)page.mp_meta.mm_psize) { - mdbx_notice("meta[%u] has too large max-mapsize (%" PRIu64 "), skip it", - meta_number, mapsize_max); - rc = MDBX_TOO_LARGE; - continue; - } - - /* LY: check end_pgno */ - if (page.mp_meta.mm_geo.now < page.mp_meta.mm_geo.lower || - page.mp_meta.mm_geo.now > page.mp_meta.mm_geo.upper) { - mdbx_notice("meta[%u] has invalid end-pageno (%" PRIaPGNO "), skip it", - meta_number, page.mp_meta.mm_geo.now); - rc = MDBX_CORRUPTED; - continue; - } - - /* LY: check last_pgno */ - if (page.mp_meta.mm_geo.next < MIN_PAGENO || - page.mp_meta.mm_geo.next - 1 > MAX_PAGENO) { - mdbx_notice("meta[%u] has invalid next-pageno (%" PRIaPGNO "), skip it", - meta_number, page.mp_meta.mm_geo.next); - rc = MDBX_CORRUPTED; - continue; - } - - if (page.mp_meta.mm_geo.next > page.mp_meta.mm_geo.now) { - mdbx_notice("meta[%u] next-pageno (%" PRIaPGNO - ") is beyond end-pgno (%" PRIaPGNO "), skip it", - meta_number, page.mp_meta.mm_geo.next, - page.mp_meta.mm_geo.now); - rc = MDBX_CORRUPTED; - continue; - } - - /* LY: FreeDB root */ - if (page.mp_meta.mm_dbs[FREE_DBI].md_root == P_INVALID) { - if (page.mp_meta.mm_dbs[FREE_DBI].md_branch_pages || - page.mp_meta.mm_dbs[FREE_DBI].md_depth || - page.mp_meta.mm_dbs[FREE_DBI].md_entries || - page.mp_meta.mm_dbs[FREE_DBI].md_leaf_pages || - page.mp_meta.mm_dbs[FREE_DBI].md_overflow_pages) { - mdbx_notice("meta[%u] has false-empty freedb, skip it", meta_number); - rc = MDBX_CORRUPTED; - continue; - } - } else if (page.mp_meta.mm_dbs[FREE_DBI].md_root >= - page.mp_meta.mm_geo.next) { - mdbx_notice("meta[%u] has invalid freedb-root %" PRIaPGNO ", skip it", - meta_number, page.mp_meta.mm_dbs[FREE_DBI].md_root); - rc = MDBX_CORRUPTED; - continue; - } - - /* LY: MainDB root */ - if (page.mp_meta.mm_dbs[MAIN_DBI].md_root == P_INVALID) { - if (page.mp_meta.mm_dbs[MAIN_DBI].md_branch_pages || - page.mp_meta.mm_dbs[MAIN_DBI].md_depth || - page.mp_meta.mm_dbs[MAIN_DBI].md_entries || - page.mp_meta.mm_dbs[MAIN_DBI].md_leaf_pages || - page.mp_meta.mm_dbs[MAIN_DBI].md_overflow_pages) { - mdbx_notice("meta[%u] has false-empty maindb", meta_number); - rc = MDBX_CORRUPTED; - continue; - } - } else if (page.mp_meta.mm_dbs[MAIN_DBI].md_root >= - page.mp_meta.mm_geo.next) { - mdbx_notice("meta[%u] has invalid maindb-root %" PRIaPGNO ", skip it", - meta_number, page.mp_meta.mm_dbs[MAIN_DBI].md_root); - rc = MDBX_CORRUPTED; - continue; - } - - if (page.mp_meta.mm_txnid_a == 0) { - mdbx_warning("meta[%u] has zero txnid, skip it", meta_number); - continue; - } - - if (mdbx_meta_ot(prefer_noweak, env, meta, &page.mp_meta)) { - *meta = page.mp_meta; - if (META_IS_WEAK(meta)) - loop_limit += 1; /* LY: should re-read to hush race with update */ - mdbx_info("latch meta[%u]", meta_number); - } - } - - if (META_IS_WEAK(meta)) { - mdbx_error("no usable meta-pages, database is corrupted"); - return rc; - } - - return MDBX_SUCCESS; -} - -static MDBX_page *__cold mdbx_meta_model(const MDBX_env *env, MDBX_page *model, - unsigned num) { - - mdbx_ensure(env, mdbx_is_power2(env->me_psize)); - mdbx_ensure(env, env->me_psize >= MIN_PAGESIZE); - mdbx_ensure(env, env->me_psize <= MAX_PAGESIZE); - mdbx_ensure(env, env->me_dbgeo.lower >= MIN_MAPSIZE); - mdbx_ensure(env, env->me_dbgeo.upper <= MAX_MAPSIZE); - mdbx_ensure(env, env->me_dbgeo.now >= env->me_dbgeo.lower); - mdbx_ensure(env, env->me_dbgeo.now <= env->me_dbgeo.upper); - - memset(model, 0, sizeof(*model)); - model->mp_pgno = num; - model->mp_flags = P_META; - model->mp_meta.mm_magic_and_version = MDBX_DATA_MAGIC; - - model->mp_meta.mm_geo.lower = bytes2pgno(env, env->me_dbgeo.lower); - model->mp_meta.mm_geo.upper = bytes2pgno(env, env->me_dbgeo.upper); - model->mp_meta.mm_geo.grow = (uint16_t)bytes2pgno(env, env->me_dbgeo.grow); - model->mp_meta.mm_geo.shrink = - (uint16_t)bytes2pgno(env, env->me_dbgeo.shrink); - model->mp_meta.mm_geo.now = bytes2pgno(env, env->me_dbgeo.now); - model->mp_meta.mm_geo.next = NUM_METAS; - - mdbx_ensure(env, model->mp_meta.mm_geo.lower >= MIN_PAGENO); - mdbx_ensure(env, model->mp_meta.mm_geo.upper <= MAX_PAGENO); - mdbx_ensure(env, model->mp_meta.mm_geo.now >= model->mp_meta.mm_geo.lower); - mdbx_ensure(env, model->mp_meta.mm_geo.now <= model->mp_meta.mm_geo.upper); - mdbx_ensure(env, model->mp_meta.mm_geo.next >= MIN_PAGENO); - mdbx_ensure(env, model->mp_meta.mm_geo.next <= model->mp_meta.mm_geo.now); - mdbx_ensure(env, model->mp_meta.mm_geo.grow == - bytes2pgno(env, env->me_dbgeo.grow)); - mdbx_ensure(env, model->mp_meta.mm_geo.shrink == - bytes2pgno(env, env->me_dbgeo.shrink)); - - model->mp_meta.mm_psize = env->me_psize; - model->mp_meta.mm_flags = (uint16_t)env->me_flags; - model->mp_meta.mm_flags |= - MDBX_INTEGERKEY; /* this is mm_dbs[FREE_DBI].md_flags */ - model->mp_meta.mm_dbs[FREE_DBI].md_root = P_INVALID; - model->mp_meta.mm_dbs[MAIN_DBI].md_root = P_INVALID; - mdbx_meta_set_txnid(env, &model->mp_meta, MIN_TXNID + num); - model->mp_meta.mm_datasync_sign = mdbx_meta_sign(&model->mp_meta); - return (MDBX_page *)((uint8_t *)model + env->me_psize); -} - -/* Fill in most of the zeroed meta-pages for an empty database environment. - * Return pointer to recenly (head) meta-page. */ -static MDBX_page *__cold mdbx_init_metas(const MDBX_env *env, void *buffer) { - MDBX_page *page0 = (MDBX_page *)buffer; - MDBX_page *page1 = mdbx_meta_model(env, page0, 0); - MDBX_page *page2 = mdbx_meta_model(env, page1, 1); - mdbx_meta_model(env, page2, 2); - page2->mp_meta.mm_datasync_sign = MDBX_DATASIGN_WEAK; - mdbx_assert(env, !mdbx_meta_eq(env, &page0->mp_meta, &page1->mp_meta)); - mdbx_assert(env, !mdbx_meta_eq(env, &page1->mp_meta, &page2->mp_meta)); - mdbx_assert(env, !mdbx_meta_eq(env, &page2->mp_meta, &page0->mp_meta)); - return page1; -} - -static int mdbx_sync_locked(MDBX_env *env, unsigned flags, - MDBX_meta *const pending) { - mdbx_assert(env, ((env->me_flags ^ flags) & MDBX_WRITEMAP) == 0); - MDBX_meta *const meta0 = METAPAGE(env, 0); - MDBX_meta *const meta1 = METAPAGE(env, 1); - MDBX_meta *const meta2 = METAPAGE(env, 2); - MDBX_meta *const head = mdbx_meta_head(env); - - mdbx_assert(env, mdbx_meta_eq_mask(env) == 0); - mdbx_assert(env, - pending < METAPAGE(env, 0) || pending > METAPAGE(env, NUM_METAS)); - mdbx_assert(env, (env->me_flags & (MDBX_RDONLY | MDBX_FATAL_ERROR)) == 0); - mdbx_assert(env, !META_IS_STEADY(head) || env->me_sync_pending != 0); - mdbx_assert(env, pending->mm_geo.next <= pending->mm_geo.now); - - const size_t usedbytes = pgno_align2os_bytes(env, pending->mm_geo.next); - if (env->me_sync_threshold && env->me_sync_pending >= env->me_sync_threshold) - flags &= MDBX_WRITEMAP | MDBX_SHRINK_ALLOWED; - - /* LY: step#1 - sync previously written/updated data-pages */ - int rc = MDBX_RESULT_TRUE; - if (env->me_sync_pending && (flags & MDBX_NOSYNC) == 0) { - mdbx_assert(env, ((flags ^ env->me_flags) & MDBX_WRITEMAP) == 0); - MDBX_meta *const steady = mdbx_meta_steady(env); - if (flags & MDBX_WRITEMAP) { - rc = mdbx_msync(&env->me_dxb_mmap, 0, usedbytes, flags & MDBX_MAPASYNC); - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - if ((flags & MDBX_MAPASYNC) == 0) { - if (unlikely(pending->mm_geo.next > steady->mm_geo.now)) { - rc = mdbx_filesize_sync(env->me_fd); - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - } - env->me_sync_pending = 0; - } - } else { - rc = mdbx_filesync(env->me_fd, pending->mm_geo.next > steady->mm_geo.now); - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - env->me_sync_pending = 0; - } - } - -#if defined(_WIN32) || defined(_WIN64) -/* Windows is unable shrinking a mapped file */ -#else - /* LY: check conditions to shrink datafile */ - pgno_t shrink = 0; - if ((flags & MDBX_SHRINK_ALLOWED) && pending->mm_geo.shrink && - pending->mm_geo.now - pending->mm_geo.next > pending->mm_geo.shrink) { - const pgno_t aligner = - pending->mm_geo.grow ? pending->mm_geo.grow : pending->mm_geo.shrink; - const pgno_t aligned = pgno_align2os_pgno( - env, pending->mm_geo.next + aligner - pending->mm_geo.next % aligner); - const pgno_t bottom = - (aligned > pending->mm_geo.lower) ? aligned : pending->mm_geo.lower; - if (pending->mm_geo.now > bottom) { - shrink = pending->mm_geo.now - bottom; - pending->mm_geo.now = bottom; - if (mdbx_meta_txnid_stable(env, head) == pending->mm_txnid_a) - mdbx_meta_set_txnid(env, pending, pending->mm_txnid_a + 1); - } - } -#endif /* not a Windows */ - - /* Steady or Weak */ - if (env->me_sync_pending == 0) { - pending->mm_datasync_sign = mdbx_meta_sign(pending); - } else { - pending->mm_datasync_sign = - (flags & MDBX_UTTERLY_NOSYNC) == MDBX_UTTERLY_NOSYNC - ? MDBX_DATASIGN_NONE - : MDBX_DATASIGN_WEAK; - } - - MDBX_meta *target = nullptr; - if (mdbx_meta_txnid_stable(env, head) == pending->mm_txnid_a) { - mdbx_assert(env, memcmp(&head->mm_dbs, &pending->mm_dbs, - sizeof(head->mm_dbs)) == 0); - mdbx_assert(env, memcmp(&head->mm_canary, &pending->mm_canary, - sizeof(head->mm_canary)) == 0); - mdbx_assert(env, memcmp(&head->mm_geo, &pending->mm_geo, - sizeof(pending->mm_geo)) == 0); - if (!META_IS_STEADY(head) && META_IS_STEADY(pending)) - target = head; - else { - mdbx_ensure(env, mdbx_meta_eq(env, head, pending)); - mdbx_debug("skip update meta"); - return MDBX_SUCCESS; - } - } else if (head == meta0) - target = mdbx_meta_ancient(prefer_steady, env, meta1, meta2); - else if (head == meta1) - target = mdbx_meta_ancient(prefer_steady, env, meta0, meta2); - else { - mdbx_assert(env, head == meta2); - target = mdbx_meta_ancient(prefer_steady, env, meta0, meta1); - } - - /* LY: step#2 - update meta-page. */ - mdbx_debug("writing meta%" PRIaPGNO " = root %" PRIaPGNO "/%" PRIaPGNO - ", geo %" PRIaPGNO "/%" PRIaPGNO "-%" PRIaPGNO "/%" PRIaPGNO - " +%u -%u, txn_id %" PRIaTXN ", %s", - container_of(target, MDBX_page, mp_data)->mp_pgno, - pending->mm_dbs[MAIN_DBI].md_root, - pending->mm_dbs[FREE_DBI].md_root, pending->mm_geo.lower, - pending->mm_geo.next, pending->mm_geo.now, pending->mm_geo.upper, - pending->mm_geo.grow, pending->mm_geo.shrink, pending->mm_txnid_a, - mdbx_durable_str(pending)); - - mdbx_debug("meta0: %s, %s, txn_id %" PRIaTXN ", root %" PRIaPGNO - "/%" PRIaPGNO, - (meta0 == head) ? "head" : (meta0 == target) ? "tail" : "stay", - mdbx_durable_str(meta0), mdbx_meta_txnid_fluid(env, meta0), - meta0->mm_dbs[MAIN_DBI].md_root, meta0->mm_dbs[FREE_DBI].md_root); - mdbx_debug("meta1: %s, %s, txn_id %" PRIaTXN ", root %" PRIaPGNO - "/%" PRIaPGNO, - (meta1 == head) ? "head" : (meta1 == target) ? "tail" : "stay", - mdbx_durable_str(meta1), mdbx_meta_txnid_fluid(env, meta1), - meta1->mm_dbs[MAIN_DBI].md_root, meta1->mm_dbs[FREE_DBI].md_root); - mdbx_debug("meta2: %s, %s, txn_id %" PRIaTXN ", root %" PRIaPGNO - "/%" PRIaPGNO, - (meta2 == head) ? "head" : (meta2 == target) ? "tail" : "stay", - mdbx_durable_str(meta2), mdbx_meta_txnid_fluid(env, meta2), - meta2->mm_dbs[MAIN_DBI].md_root, meta2->mm_dbs[FREE_DBI].md_root); - - mdbx_assert(env, !mdbx_meta_eq(env, pending, meta0)); - mdbx_assert(env, !mdbx_meta_eq(env, pending, meta1)); - mdbx_assert(env, !mdbx_meta_eq(env, pending, meta2)); - - mdbx_assert(env, ((env->me_flags ^ flags) & MDBX_WRITEMAP) == 0); - mdbx_ensure(env, - target == head || - mdbx_meta_txnid_stable(env, target) < pending->mm_txnid_a); - if (env->me_flags & MDBX_WRITEMAP) { - mdbx_jitter4testing(true); - if (likely(target != head)) { - /* LY: 'invalidate' the meta. */ - target->mm_datasync_sign = MDBX_DATASIGN_WEAK; - mdbx_meta_update_begin(env, target, pending->mm_txnid_a); -#ifndef NDEBUG - /* debug: provoke failure to catch a violators */ - memset(target->mm_dbs, 0xCC, - sizeof(target->mm_dbs) + sizeof(target->mm_canary)); - mdbx_jitter4testing(false); -#endif - - /* LY: update info */ - target->mm_geo = pending->mm_geo; - target->mm_dbs[FREE_DBI] = pending->mm_dbs[FREE_DBI]; - target->mm_dbs[MAIN_DBI] = pending->mm_dbs[MAIN_DBI]; - target->mm_canary = pending->mm_canary; - mdbx_jitter4testing(true); - mdbx_coherent_barrier(); - - /* LY: 'commit' the meta */ - mdbx_meta_update_end(env, target, pending->mm_txnid_b); - mdbx_jitter4testing(true); - } else { - /* dangerous case (target == head), only mm_datasync_sign could - * me updated, check assertions once again */ - mdbx_ensure(env, - mdbx_meta_txnid_stable(env, head) == pending->mm_txnid_a && - !META_IS_STEADY(head) && META_IS_STEADY(pending)); - mdbx_ensure(env, memcmp(&head->mm_geo, &pending->mm_geo, - sizeof(head->mm_geo)) == 0); - mdbx_ensure(env, memcmp(&head->mm_dbs, &pending->mm_dbs, - sizeof(head->mm_dbs)) == 0); - mdbx_ensure(env, memcmp(&head->mm_canary, &pending->mm_canary, - sizeof(head->mm_canary)) == 0); - } - target->mm_datasync_sign = pending->mm_datasync_sign; - mdbx_coherent_barrier(); - mdbx_jitter4testing(true); - } else { - rc = mdbx_pwrite(env->me_fd, pending, sizeof(MDBX_meta), - (uint8_t *)target - env->me_map); - if (unlikely(rc != MDBX_SUCCESS)) { - undo: - mdbx_debug("write failed, disk error?"); - /* On a failure, the pagecache still contains the new data. - * Try write some old data back, to prevent it from being used. */ - mdbx_pwrite(env->me_fd, (void *)target, sizeof(MDBX_meta), - (uint8_t *)target - env->me_map); - goto fail; - } - mdbx_invalidate_cache(target, sizeof(MDBX_meta)); - } - - /* LY: step#3 - sync meta-pages. */ - mdbx_assert(env, ((env->me_flags ^ flags) & MDBX_WRITEMAP) == 0); - if ((flags & (MDBX_NOSYNC | MDBX_NOMETASYNC)) == 0) { - mdbx_assert(env, ((flags ^ env->me_flags) & MDBX_WRITEMAP) == 0); - if (flags & MDBX_WRITEMAP) { - const size_t offset = - ((uint8_t *)container_of(head, MDBX_page, mp_meta)) - - env->me_dxb_mmap.dxb; - const size_t paged_offset = offset & ~(env->me_os_psize - 1); - const size_t paged_length = mdbx_roundup2( - env->me_psize + offset - paged_offset, env->me_os_psize); - rc = mdbx_msync(&env->me_dxb_mmap, paged_offset, paged_length, - flags & MDBX_MAPASYNC); - if (unlikely(rc != MDBX_SUCCESS)) - goto fail; - } else { - rc = mdbx_filesync(env->me_fd, false); - if (rc != MDBX_SUCCESS) - goto undo; - } - } - -#if defined(_WIN32) || defined(_WIN64) -/* Windows is unable shrinking a mapped file */ -#else - /* LY: shrink datafile if needed */ - if (unlikely(shrink)) { - mdbx_info("shrink to %" PRIaPGNO " pages (-%" PRIaPGNO ")", - pending->mm_geo.now, shrink); - rc = mdbx_mapresize(env, pending->mm_geo.now, pending->mm_geo.upper); - if (MDBX_IS_ERROR(rc)) - goto fail; - } -#endif /* not a Windows */ - - return MDBX_SUCCESS; - -fail: - env->me_flags |= MDBX_FATAL_ERROR; - return rc; -} - -int __cold mdbx_env_get_maxkeysize(MDBX_env *env) { - if (!env || env->me_signature != MDBX_ME_SIGNATURE || !env->me_maxkey_limit) - return -MDBX_EINVAL; - return env->me_maxkey_limit; -} - -#define mdbx_nodemax(pagesize) \ - (((((pagesize)-PAGEHDRSZ) / MDBX_MINKEYS) & -(intptr_t)2) - sizeof(indx_t)) - -#define mdbx_maxkey(nodemax) ((nodemax) - (NODESIZE + sizeof(MDBX_db))) - -#define mdbx_maxfree1pg(pagesize) (((pagesize)-PAGEHDRSZ) / sizeof(pgno_t) - 1) - -int mdbx_get_maxkeysize(size_t pagesize) { - if (pagesize == 0) - pagesize = mdbx_syspagesize(); - - intptr_t nodemax = mdbx_nodemax(pagesize); - if (nodemax < 0) - return -MDBX_EINVAL; - - intptr_t maxkey = mdbx_maxkey(nodemax); - return (maxkey > 0 && maxkey < INT_MAX) ? (int)maxkey : -MDBX_EINVAL; -} - -static void __cold mdbx_setup_pagesize(MDBX_env *env, const size_t pagesize) { - STATIC_ASSERT(SSIZE_MAX > MAX_MAPSIZE); - STATIC_ASSERT(MIN_PAGESIZE > sizeof(MDBX_page)); - mdbx_ensure(env, mdbx_is_power2(pagesize)); - mdbx_ensure(env, pagesize >= MIN_PAGESIZE); - mdbx_ensure(env, pagesize <= MAX_PAGESIZE); - env->me_psize = (unsigned)pagesize; - - STATIC_ASSERT(mdbx_maxfree1pg(MIN_PAGESIZE) > 42); - STATIC_ASSERT(mdbx_maxfree1pg(MAX_PAGESIZE) < MDBX_PNL_DB_MAX); - const intptr_t maxfree_1pg = (pagesize - PAGEHDRSZ) / sizeof(pgno_t) - 1; - mdbx_ensure(env, maxfree_1pg > 42 && maxfree_1pg < MDBX_PNL_DB_MAX); - env->me_maxfree_1pg = (unsigned)maxfree_1pg; - - STATIC_ASSERT(mdbx_nodemax(MIN_PAGESIZE) > 42); - STATIC_ASSERT(mdbx_nodemax(MAX_PAGESIZE) < UINT16_MAX); - const intptr_t nodemax = mdbx_nodemax(pagesize); - mdbx_ensure(env, nodemax > 42 && nodemax < UINT16_MAX); - env->me_nodemax = (unsigned)nodemax; - - STATIC_ASSERT(mdbx_maxkey(MIN_PAGESIZE) > 42); - STATIC_ASSERT(mdbx_maxkey(MIN_PAGESIZE) < MIN_PAGESIZE); - STATIC_ASSERT(mdbx_maxkey(MAX_PAGESIZE) > 42); - STATIC_ASSERT(mdbx_maxkey(MAX_PAGESIZE) < MAX_PAGESIZE); - const intptr_t maxkey_limit = mdbx_maxkey(env->me_nodemax); - mdbx_ensure(env, maxkey_limit > 42 && (size_t)maxkey_limit < pagesize); - env->me_maxkey_limit = (unsigned)maxkey_limit; - - env->me_psize2log = mdbx_log2(pagesize); - mdbx_assert(env, pgno2bytes(env, 1) == pagesize); - mdbx_assert(env, bytes2pgno(env, pagesize + pagesize) == 2); -} - -int __cold mdbx_env_create(MDBX_env **penv) { - MDBX_env *env = calloc(1, sizeof(MDBX_env)); - if (!env) - return MDBX_ENOMEM; - - env->me_maxreaders = DEFAULT_READERS; - env->me_maxdbs = env->me_numdbs = CORE_DBS; - env->me_fd = INVALID_HANDLE_VALUE; - env->me_lfd = INVALID_HANDLE_VALUE; - env->me_pid = mdbx_getpid(); - - int rc; - const size_t os_psize = mdbx_syspagesize(); - if (!mdbx_is_power2(os_psize) || os_psize < MIN_PAGESIZE) { - mdbx_error("unsuitable system pagesize %" PRIuPTR, os_psize); - rc = MDBX_INCOMPATIBLE; - goto bailout; - } - env->me_os_psize = (unsigned)os_psize; - mdbx_setup_pagesize(env, env->me_os_psize); - - rc = mdbx_fastmutex_init(&env->me_dbi_lock); - if (unlikely(rc != MDBX_SUCCESS)) - goto bailout; - - VALGRIND_CREATE_MEMPOOL(env, 0, 0); - env->me_signature = MDBX_ME_SIGNATURE; - *penv = env; - return MDBX_SUCCESS; - -bailout: - free(env); - *penv = nullptr; - return rc; -} - -static int __cold mdbx_env_map(MDBX_env *env, size_t usedsize) { - int rc = mdbx_mmap(env->me_flags, &env->me_dxb_mmap, env->me_dbgeo.now, - env->me_dbgeo.upper); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - -#ifdef MADV_DONTFORK - if (madvise(env->me_map, env->me_mapsize, MADV_DONTFORK)) - return errno; -#endif - -#ifdef MADV_NOHUGEPAGE - (void)madvise(env->me_map, env->me_mapsize, MADV_NOHUGEPAGE); -#endif - -#if defined(MADV_DODUMP) && defined(MADV_DONTDUMP) - const size_t meta_length = pgno2bytes(env, NUM_METAS); - (void)madvise(env->me_map, meta_length, MADV_DODUMP); - if (!(env->me_flags & MDBX_PAGEPERTURB)) - (void)madvise(env->me_map + meta_length, env->me_mapsize - meta_length, - MADV_DONTDUMP); -#endif - -#ifdef MADV_REMOVE - if (usedsize && (env->me_flags & MDBX_WRITEMAP)) { - (void)madvise(env->me_map + usedsize, env->me_mapsize - usedsize, - MADV_REMOVE); - } -#else - (void)usedsize; -#endif - -#if defined(MADV_RANDOM) && defined(MADV_WILLNEED) - /* Turn on/off readahead. It's harmful when the DB is larger than RAM. */ - if (madvise(env->me_map, env->me_mapsize, - (env->me_flags & MDBX_NORDAHEAD) ? MADV_RANDOM : MADV_WILLNEED)) - return errno; -#endif - -#ifdef USE_VALGRIND - env->me_valgrind_handle = - VALGRIND_CREATE_BLOCK(env->me_map, env->me_mapsize, "mdbx"); -#endif - - return MDBX_SUCCESS; -} - -LIBMDBX_API int mdbx_env_set_geometry(MDBX_env *env, intptr_t size_lower, - intptr_t size_now, intptr_t size_upper, - intptr_t growth_step, - intptr_t shrink_threshold, - intptr_t pagesize) { - if (unlikely(!env)) - return MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - const bool outside_txn = - (!env->me_txn0 || env->me_txn0->mt_owner != mdbx_thread_self()); - -#if MDBX_DEBUG - if (growth_step < 0) - growth_step = 1; - if (shrink_threshold < 0) - shrink_threshold = 1; -#endif - - int rc = MDBX_PROBLEM; - if (env->me_map) { - /* env already mapped */ - if (!env->me_lck || (env->me_flags & MDBX_RDONLY)) - return MDBX_EACCESS; - - if (outside_txn) { - int err = mdbx_txn_lock(env, false); - if (unlikely(err != MDBX_SUCCESS)) - return err; - } - MDBX_meta *head = mdbx_meta_head(env); - - if (pagesize < 0) - pagesize = env->me_psize; - if (pagesize != (intptr_t)env->me_psize) { - rc = MDBX_EINVAL; - goto bailout; - } - - if (size_lower < 0) - size_lower = pgno2bytes(env, head->mm_geo.lower); - if (size_now < 0) - size_now = pgno2bytes(env, head->mm_geo.now); - if (size_upper < 0) - size_upper = pgno2bytes(env, head->mm_geo.upper); - if (growth_step < 0) - growth_step = pgno2bytes(env, head->mm_geo.grow); - if (shrink_threshold < 0) - shrink_threshold = pgno2bytes(env, head->mm_geo.shrink); - - const size_t usedbytes = pgno2bytes(env, head->mm_geo.next); - if ((size_t)size_upper < usedbytes) { - rc = MDBX_MAP_FULL; - goto bailout; - } - if ((size_t)size_now < usedbytes) - size_now = usedbytes; -#if defined(_WIN32) || defined(_WIN64) - if ((size_t)size_now < env->me_dbgeo.now || - (size_t)size_upper < env->me_dbgeo.upper) { - /* Windows is unable shrinking a mapped file */ - return ERROR_USER_MAPPED_FILE; - } -#endif /* Windows */ - } else { - /* env NOT yet mapped */ - if (!outside_txn) { - rc = MDBX_PANIC; - goto bailout; - } - - if (pagesize < 0) { - pagesize = env->me_os_psize; - if (pagesize > MAX_PAGESIZE) - pagesize = MAX_PAGESIZE; - mdbx_assert(env, pagesize >= MIN_PAGESIZE); - } - } - - if (pagesize < MIN_PAGESIZE || pagesize > MAX_PAGESIZE || - !mdbx_is_power2(pagesize)) { - rc = MDBX_EINVAL; - goto bailout; - } - - if (size_lower < 0) { - size_lower = MIN_MAPSIZE; - if (MIN_MAPSIZE / pagesize < MIN_PAGENO) - size_lower = MIN_PAGENO * pagesize; - } - - if (size_now < 0) { - size_now = DEFAULT_MAPSIZE; - if (size_now < size_lower) - size_now = size_lower; - } - - if (size_upper < 0) { - if ((size_t)size_now >= MAX_MAPSIZE / 2) - size_upper = MAX_MAPSIZE; - else if (MAX_MAPSIZE != MAX_MAPSIZE32 && - (size_t)size_now >= MAX_MAPSIZE32 / 2 && - (size_t)size_now <= MAX_MAPSIZE32 / 4 * 3) - size_upper = MAX_MAPSIZE32; - else { - size_upper = size_now + size_now; - if ((size_t)size_upper < DEFAULT_MAPSIZE * 2) - size_upper = DEFAULT_MAPSIZE * 2; - } - if ((size_t)size_upper / pagesize > MAX_PAGENO) - size_upper = pagesize * MAX_PAGENO; - } - - if (unlikely(size_lower < MIN_MAPSIZE || size_lower > size_upper)) { - rc = MDBX_EINVAL; - goto bailout; - } - - if ((uint64_t)size_lower / pagesize < MIN_PAGENO) { - rc = MDBX_EINVAL; - goto bailout; - } - - if (unlikely((size_t)size_upper > MAX_MAPSIZE || - (uint64_t)size_upper / pagesize > MAX_PAGENO)) { - rc = MDBX_TOO_LARGE; - goto bailout; - } - - size_lower = mdbx_roundup2(size_lower, env->me_os_psize); - size_upper = mdbx_roundup2(size_upper, env->me_os_psize); - size_now = mdbx_roundup2(size_now, env->me_os_psize); - - /* LY: подбираем значение size_upper: - * - кратное размеру системной страницы - * - без нарушения MAX_MAPSIZE или MAX_PAGENO */ - while (unlikely((size_t)size_upper > MAX_MAPSIZE || - (uint64_t)size_upper / pagesize > MAX_PAGENO)) { - if ((size_t)size_upper < env->me_os_psize + MIN_MAPSIZE || - (size_t)size_upper < env->me_os_psize * (MIN_PAGENO + 1)) { - /* паранойа на случай переполнения при невероятных значениях */ - rc = MDBX_EINVAL; - goto bailout; - } - size_upper -= env->me_os_psize; - if ((size_t)size_upper > (size_t)size_lower) - size_lower = size_upper; - } - mdbx_assert(env, (size_upper - size_lower) % env->me_os_psize == 0); - - if (size_now < size_lower) - size_now = size_lower; - if (size_now > size_upper) - size_now = size_upper; - - if (growth_step < 0) { - growth_step = ((size_t)(size_upper - size_lower)) / 42; - if (growth_step > size_lower) - growth_step = size_lower; - if (growth_step < 65536) - growth_step = 65536; - if ((size_t)growth_step > MEGABYTE * 16) - growth_step = MEGABYTE * 16; - } - growth_step = mdbx_roundup2(growth_step, env->me_os_psize); - if (bytes2pgno(env, growth_step) > UINT16_MAX) - growth_step = pgno2bytes(env, UINT16_MAX); - - if (shrink_threshold < 0) { - shrink_threshold = growth_step + growth_step; - if (shrink_threshold < growth_step) - shrink_threshold = growth_step; - } - shrink_threshold = mdbx_roundup2(shrink_threshold, env->me_os_psize); - if (bytes2pgno(env, shrink_threshold) > UINT16_MAX) - shrink_threshold = pgno2bytes(env, UINT16_MAX); - - /* save user's geo-params for future open/create */ - env->me_dbgeo.lower = size_lower; - env->me_dbgeo.now = size_now; - env->me_dbgeo.upper = size_upper; - env->me_dbgeo.grow = growth_step; - env->me_dbgeo.shrink = shrink_threshold; - rc = MDBX_SUCCESS; - - if (env->me_map) { - /* apply new params */ - mdbx_assert(env, pagesize == (intptr_t)env->me_psize); - - MDBX_meta *head = mdbx_meta_head(env); - MDBX_meta meta = *head; - meta.mm_geo.lower = bytes2pgno(env, env->me_dbgeo.lower); - meta.mm_geo.now = bytes2pgno(env, env->me_dbgeo.now); - meta.mm_geo.upper = bytes2pgno(env, env->me_dbgeo.upper); - meta.mm_geo.grow = (uint16_t)bytes2pgno(env, env->me_dbgeo.grow); - meta.mm_geo.shrink = (uint16_t)bytes2pgno(env, env->me_dbgeo.shrink); - - mdbx_assert(env, env->me_dbgeo.lower >= MIN_MAPSIZE); - mdbx_assert(env, meta.mm_geo.lower >= MIN_PAGENO); - mdbx_assert(env, env->me_dbgeo.upper <= MAX_MAPSIZE); - mdbx_assert(env, meta.mm_geo.upper <= MAX_PAGENO); - mdbx_assert(env, meta.mm_geo.now >= meta.mm_geo.next); - mdbx_assert(env, env->me_dbgeo.upper >= env->me_dbgeo.lower); - mdbx_assert(env, meta.mm_geo.upper >= meta.mm_geo.now); - mdbx_assert(env, meta.mm_geo.now >= meta.mm_geo.lower); - mdbx_assert(env, meta.mm_geo.grow == bytes2pgno(env, env->me_dbgeo.grow)); - mdbx_assert(env, - meta.mm_geo.shrink == bytes2pgno(env, env->me_dbgeo.shrink)); - - if (memcmp(&meta.mm_geo, &head->mm_geo, sizeof(meta.mm_geo))) { - if (meta.mm_geo.now != head->mm_geo.now || - meta.mm_geo.upper != head->mm_geo.upper) { - rc = mdbx_mapresize(env, meta.mm_geo.now, meta.mm_geo.upper); - if (unlikely(rc != MDBX_SUCCESS)) - goto bailout; - } - env->me_sync_pending += env->me_psize; - mdbx_meta_set_txnid(env, &meta, mdbx_meta_txnid_stable(env, head) + 1); - rc = mdbx_sync_locked(env, env->me_flags, &meta); - } - } else if (pagesize != (intptr_t)env->me_psize) { - mdbx_setup_pagesize(env, pagesize); - } - -bailout: - if (env->me_map && outside_txn) - mdbx_txn_unlock(env); - return rc; -} - -int __cold mdbx_env_set_mapsize(MDBX_env *env, size_t size) { - return mdbx_env_set_geometry(env, -1, size, -1, -1, -1, -1); -} - -int __cold mdbx_env_set_maxdbs(MDBX_env *env, MDBX_dbi dbs) { - if (unlikely(dbs > MAX_DBI)) - return MDBX_EINVAL; - - if (unlikely(!env)) - return MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(env->me_map)) - return MDBX_EPERM; - - env->me_maxdbs = dbs + CORE_DBS; - return MDBX_SUCCESS; -} - -int __cold mdbx_env_set_maxreaders(MDBX_env *env, unsigned readers) { - if (unlikely(readers < 1 || readers > INT16_MAX)) - return MDBX_EINVAL; - - if (unlikely(!env)) - return MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(env->me_map)) - return MDBX_EPERM; - - env->me_maxreaders = readers; - return MDBX_SUCCESS; -} - -int __cold mdbx_env_get_maxreaders(MDBX_env *env, unsigned *readers) { - if (!env || !readers) - return MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - *readers = env->me_maxreaders; - return MDBX_SUCCESS; -} - -/* Further setup required for opening an MDBX environment */ -static int __cold mdbx_setup_dxb(MDBX_env *env, int lck_rc) { - MDBX_meta meta; - int rc = MDBX_RESULT_FALSE; - int err = mdbx_read_header(env, &meta); - if (unlikely(err != MDBX_SUCCESS)) { - if (lck_rc != /* lck exclusive */ MDBX_RESULT_TRUE || err != MDBX_ENODATA || - (env->me_flags & MDBX_RDONLY) != 0) - return err; - - mdbx_debug("create new database"); - rc = /* new database */ MDBX_RESULT_TRUE; - - if (!env->me_dbgeo.now) { - /* set defaults if not configured */ - err = mdbx_env_set_mapsize(env, DEFAULT_MAPSIZE); - if (unlikely(err != MDBX_SUCCESS)) - return err; - } - - void *buffer = calloc(NUM_METAS, env->me_psize); - if (!buffer) - return MDBX_ENOMEM; - - meta = mdbx_init_metas(env, buffer)->mp_meta; - err = mdbx_pwrite(env->me_fd, buffer, env->me_psize * NUM_METAS, 0); - free(buffer); - if (unlikely(err != MDBX_SUCCESS)) - return err; - - err = mdbx_ftruncate(env->me_fd, env->me_dbgeo.now); - if (unlikely(err != MDBX_SUCCESS)) - return err; - -#ifndef NDEBUG /* just for checking */ - err = mdbx_read_header(env, &meta); - if (unlikely(err != MDBX_SUCCESS)) - return err; -#endif - } - - mdbx_info("header: root %" PRIaPGNO "/%" PRIaPGNO ", geo %" PRIaPGNO - "/%" PRIaPGNO "-%" PRIaPGNO "/%" PRIaPGNO - " +%u -%u, txn_id %" PRIaTXN ", %s", - meta.mm_dbs[MAIN_DBI].md_root, meta.mm_dbs[FREE_DBI].md_root, - meta.mm_geo.lower, meta.mm_geo.next, meta.mm_geo.now, - meta.mm_geo.upper, meta.mm_geo.grow, meta.mm_geo.shrink, - meta.mm_txnid_a, mdbx_durable_str(&meta)); - - mdbx_setup_pagesize(env, meta.mm_psize); - if ((env->me_flags & MDBX_RDONLY) /* readonly */ - || lck_rc != MDBX_RESULT_TRUE /* not exclusive */) { - /* use present params from db */ - err = mdbx_env_set_geometry( - env, meta.mm_geo.lower * (uint64_t)meta.mm_psize, - meta.mm_geo.now * (uint64_t)meta.mm_psize, - meta.mm_geo.upper * (uint64_t)meta.mm_psize, - meta.mm_geo.grow * (uint64_t)meta.mm_psize, - meta.mm_geo.shrink * (uint64_t)meta.mm_psize, meta.mm_psize); - if (unlikely(err != MDBX_SUCCESS)) { - mdbx_error("could not use present dbsize-params from db"); - return MDBX_INCOMPATIBLE; - } - } else if (env->me_dbgeo.now) { - /* silently growth to last used page */ - const size_t used_bytes = pgno2bytes(env, meta.mm_geo.next); - if (env->me_dbgeo.lower < used_bytes) - env->me_dbgeo.lower = used_bytes; - if (env->me_dbgeo.now < used_bytes) - env->me_dbgeo.now = used_bytes; - if (env->me_dbgeo.upper < used_bytes) - env->me_dbgeo.upper = used_bytes; - - /* apply preconfigured params, but only if substantial changes: - * - upper or lower limit changes - * - shrink theshold or growth step - * But ignore just chagne just a 'now/current' size. */ - if (bytes_align2os_bytes(env, env->me_dbgeo.upper) != - pgno_align2os_bytes(env, meta.mm_geo.upper) || - bytes_align2os_bytes(env, env->me_dbgeo.lower) != - pgno_align2os_bytes(env, meta.mm_geo.lower) || - bytes_align2os_bytes(env, env->me_dbgeo.shrink) != - pgno_align2os_bytes(env, meta.mm_geo.shrink) || - bytes_align2os_bytes(env, env->me_dbgeo.grow) != - pgno_align2os_bytes(env, meta.mm_geo.grow)) { - - if (env->me_dbgeo.shrink && env->me_dbgeo.now > used_bytes) - /* pre-shrink if enabled */ - env->me_dbgeo.now = used_bytes + env->me_dbgeo.shrink - - used_bytes % env->me_dbgeo.shrink; - - err = mdbx_env_set_geometry(env, env->me_dbgeo.lower, env->me_dbgeo.now, - env->me_dbgeo.upper, env->me_dbgeo.grow, - env->me_dbgeo.shrink, meta.mm_psize); - if (unlikely(err != MDBX_SUCCESS)) { - mdbx_error("could not apply preconfigured dbsize-params to db"); - return MDBX_INCOMPATIBLE; - } - - /* update meta fields */ - meta.mm_geo.now = bytes2pgno(env, env->me_dbgeo.now); - meta.mm_geo.lower = bytes2pgno(env, env->me_dbgeo.lower); - meta.mm_geo.upper = bytes2pgno(env, env->me_dbgeo.upper); - meta.mm_geo.grow = (uint16_t)bytes2pgno(env, env->me_dbgeo.grow); - meta.mm_geo.shrink = (uint16_t)bytes2pgno(env, env->me_dbgeo.shrink); - - mdbx_info("amended: root %" PRIaPGNO "/%" PRIaPGNO ", geo %" PRIaPGNO - "/%" PRIaPGNO "-%" PRIaPGNO "/%" PRIaPGNO - " +%u -%u, txn_id %" PRIaTXN ", %s", - meta.mm_dbs[MAIN_DBI].md_root, meta.mm_dbs[FREE_DBI].md_root, - meta.mm_geo.lower, meta.mm_geo.next, meta.mm_geo.now, - meta.mm_geo.upper, meta.mm_geo.grow, meta.mm_geo.shrink, - meta.mm_txnid_a, mdbx_durable_str(&meta)); - } - mdbx_ensure(env, meta.mm_geo.now >= meta.mm_geo.next); - } else { - /* geo-params not pre-configured by user, - * get current values from a meta. */ - env->me_dbgeo.now = pgno2bytes(env, meta.mm_geo.now); - env->me_dbgeo.lower = pgno2bytes(env, meta.mm_geo.lower); - env->me_dbgeo.upper = pgno2bytes(env, meta.mm_geo.upper); - env->me_dbgeo.grow = pgno2bytes(env, meta.mm_geo.grow); - env->me_dbgeo.shrink = pgno2bytes(env, meta.mm_geo.shrink); - } - - uint64_t filesize_before_mmap; - err = mdbx_filesize(env->me_fd, &filesize_before_mmap); - if (unlikely(err != MDBX_SUCCESS)) - return err; - - const size_t expected_bytes = - mdbx_roundup2(pgno2bytes(env, meta.mm_geo.now), env->me_os_psize); - const size_t used_bytes = pgno2bytes(env, meta.mm_geo.next); - mdbx_ensure(env, expected_bytes >= used_bytes); - if (filesize_before_mmap != expected_bytes) { - if (lck_rc != /* lck exclusive */ MDBX_RESULT_TRUE) { - mdbx_info("filesize mismatch (expect %" PRIuPTR "/%" PRIaPGNO - ", have %" PRIu64 "/%" PRIaPGNO "), " - "assume collision in non-exclusive mode", - expected_bytes, bytes2pgno(env, expected_bytes), - filesize_before_mmap, - bytes2pgno(env, (size_t)filesize_before_mmap)); - } else { - mdbx_notice("filesize mismatch (expect %" PRIuPTR "/%" PRIaPGNO - ", have %" PRIu64 "/%" PRIaPGNO ")", - expected_bytes, bytes2pgno(env, expected_bytes), - filesize_before_mmap, - bytes2pgno(env, (size_t)filesize_before_mmap)); - if (filesize_before_mmap < used_bytes) { - mdbx_error("last-page beyond end-of-file (last %" PRIaPGNO - ", have %" PRIaPGNO ")", - meta.mm_geo.next, - bytes2pgno(env, (size_t)filesize_before_mmap)); - return MDBX_CORRUPTED; - } - - if (env->me_flags & MDBX_RDONLY) { - mdbx_notice("ignore filesize mismatch in readonly-mode"); - } else { - mdbx_info("resize datafile to %" PRIu64 " bytes, %" PRIaPGNO " pages", - expected_bytes, bytes2pgno(env, expected_bytes)); - err = mdbx_ftruncate(env->me_fd, expected_bytes); - if (unlikely(err != MDBX_SUCCESS)) { - mdbx_error("error %d, while resize datafile to %" PRIu64 - " bytes, %" PRIaPGNO " pages", - rc, expected_bytes, bytes2pgno(env, expected_bytes)); - return err; - } - filesize_before_mmap = expected_bytes; - } - } - } - - err = mdbx_env_map(env, (lck_rc != /* lck exclusive */ MDBX_RESULT_TRUE) - ? 0 - : expected_bytes); - if (err != MDBX_SUCCESS) - return err; - - const unsigned meta_clash_mask = mdbx_meta_eq_mask(env); - if (meta_clash_mask) { - mdbx_error("meta-pages are clashed: mask 0x%d", meta_clash_mask); - return MDBX_WANNA_RECOVERY; - } - - while (1) { - MDBX_meta *head = mdbx_meta_head(env); - const txnid_t head_txnid = mdbx_meta_txnid_fluid(env, head); - if (head_txnid == meta.mm_txnid_a) - break; - - if (lck_rc == /* lck exclusive */ MDBX_RESULT_TRUE) { - assert(META_IS_STEADY(&meta) && !META_IS_STEADY(head)); - if (env->me_flags & MDBX_RDONLY) { - mdbx_error("rollback needed: (from head %" PRIaTXN - " to steady %" PRIaTXN "), but unable in read-only mode", - head_txnid, meta.mm_txnid_a); - return MDBX_WANNA_RECOVERY /* LY: could not recovery/rollback */; - } - - /* LY: rollback weak checkpoint */ - mdbx_trace("rollback: from %" PRIaTXN ", to %" PRIaTXN, head_txnid, - meta.mm_txnid_a); - mdbx_ensure(env, head_txnid == mdbx_meta_txnid_stable(env, head)); - - if (env->me_flags & MDBX_WRITEMAP) { - head->mm_txnid_a = 0; - head->mm_datasync_sign = MDBX_DATASIGN_WEAK; - head->mm_txnid_b = 0; - const size_t offset = - ((uint8_t *)container_of(head, MDBX_page, mp_meta)) - - env->me_dxb_mmap.dxb; - const size_t paged_offset = offset & ~(env->me_os_psize - 1); - const size_t paged_length = mdbx_roundup2( - env->me_psize + offset - paged_offset, env->me_os_psize); - err = mdbx_msync(&env->me_dxb_mmap, paged_offset, paged_length, false); - } else { - MDBX_meta rollback = *head; - mdbx_meta_set_txnid(env, &rollback, 0); - rollback.mm_datasync_sign = MDBX_DATASIGN_WEAK; - err = mdbx_pwrite(env->me_fd, &rollback, sizeof(MDBX_meta), - (uint8_t *)head - (uint8_t *)env->me_map); - } - if (err) - return err; - - mdbx_invalidate_cache(env->me_map, pgno2bytes(env, NUM_METAS)); - mdbx_ensure(env, 0 == mdbx_meta_txnid_fluid(env, head)); - mdbx_ensure(env, 0 == mdbx_meta_eq_mask(env)); - continue; - } - - if (!env->me_lck) { - /* LY: without-lck (read-only) mode, so it is imposible that other - * process made weak checkpoint. */ - mdbx_error("without-lck, unable recovery/rollback"); - return MDBX_WANNA_RECOVERY; - } - - /* LY: assume just have a collision with other running process, - * or someone make a weak checkpoint */ - mdbx_info("assume collision or online weak checkpoint"); - break; - } - - const MDBX_meta *head = mdbx_meta_head(env); - if (lck_rc == /* lck exclusive */ MDBX_RESULT_TRUE) { - /* re-check file size after mmap */ - uint64_t filesize_after_mmap; - err = mdbx_filesize(env->me_fd, &filesize_after_mmap); - if (unlikely(err != MDBX_SUCCESS)) - return err; - if (filesize_after_mmap != expected_bytes) { - if (filesize_after_mmap != filesize_before_mmap) - mdbx_info("datafile resized by system to %" PRIu64 " bytes", - filesize_after_mmap); - if (filesize_after_mmap % env->me_os_psize || - filesize_after_mmap > env->me_dbgeo.upper || - filesize_after_mmap < used_bytes) { - mdbx_info("unacceptable/unexpected datafile size %" PRIu64, - filesize_after_mmap); - return MDBX_PROBLEM; - } - if ((env->me_flags & MDBX_RDONLY) == 0) { - meta.mm_geo.now = - bytes2pgno(env, env->me_dbgeo.now = (size_t)filesize_after_mmap); - mdbx_info("update meta-geo to filesize %" PRIuPTR " bytes, %" PRIaPGNO - " pages", - env->me_dbgeo.now, meta.mm_geo.now); - } - } - - if (memcmp(&meta.mm_geo, &head->mm_geo, sizeof(meta.mm_geo))) { - const txnid_t txnid = mdbx_meta_txnid_stable(env, head); - mdbx_info("updating meta.geo: " - "from l%" PRIaPGNO "-n%" PRIaPGNO "-u%" PRIaPGNO - "/s%u-g%u (txn#%" PRIaTXN "), " - "to l%" PRIaPGNO "-n%" PRIaPGNO "-u%" PRIaPGNO - "/s%u-g%u (txn#%" PRIaTXN ")", - head->mm_geo.lower, head->mm_geo.now, head->mm_geo.upper, - head->mm_geo.shrink, head->mm_geo.grow, txnid, - meta.mm_geo.lower, meta.mm_geo.now, meta.mm_geo.upper, - meta.mm_geo.shrink, meta.mm_geo.grow, txnid + 1); - - mdbx_ensure(env, mdbx_meta_eq(env, &meta, head)); - mdbx_meta_set_txnid(env, &meta, txnid + 1); - env->me_sync_pending += env->me_psize; - err = mdbx_sync_locked(env, env->me_flags | MDBX_SHRINK_ALLOWED, &meta); - if (err) { - mdbx_info("error %d, while updating meta.geo: " - "from l%" PRIaPGNO "-n%" PRIaPGNO "-u%" PRIaPGNO - "/s%u-g%u (txn#%" PRIaTXN "), " - "to l%" PRIaPGNO "-n%" PRIaPGNO "-u%" PRIaPGNO - "/s%u-g%u (txn#%" PRIaTXN ")", - err, head->mm_geo.lower, head->mm_geo.now, head->mm_geo.upper, - head->mm_geo.shrink, head->mm_geo.grow, txnid, - meta.mm_geo.lower, meta.mm_geo.now, meta.mm_geo.upper, - meta.mm_geo.shrink, meta.mm_geo.grow, txnid + 1); - return err; - } - } - } - - return rc; -} - -/****************************************************************************/ - -/* Open and/or initialize the lock region for the environment. */ -static int __cold mdbx_setup_lck(MDBX_env *env, char *lck_pathname, - mode_t mode) { - assert(env->me_fd != INVALID_HANDLE_VALUE); - assert(env->me_lfd == INVALID_HANDLE_VALUE); - - int err = mdbx_openfile(lck_pathname, O_RDWR | O_CREAT, mode, &env->me_lfd); - if (err != MDBX_SUCCESS) { - if (err != MDBX_EROFS || (env->me_flags & MDBX_RDONLY) == 0) - return err; - /* LY: without-lck mode (e.g. on read-only filesystem) */ - env->me_lfd = INVALID_HANDLE_VALUE; - env->me_oldest = &env->me_oldest_stub; - env->me_maxreaders = UINT_MAX; - mdbx_debug("lck-setup: %s ", "lockless mode (readonly)"); - return MDBX_SUCCESS; - } - - /* Try to get exclusive lock. If we succeed, then - * nobody is using the lock region and we should initialize it. */ - const int rc = mdbx_lck_seize(env); - if (MDBX_IS_ERROR(rc)) - return rc; - - mdbx_debug("lck-setup: %s ", - (rc == MDBX_RESULT_TRUE) ? "exclusive" : "shared"); - - uint64_t size; - err = mdbx_filesize(env->me_lfd, &size); - if (unlikely(err != MDBX_SUCCESS)) - return err; - - if (rc == MDBX_RESULT_TRUE) { - uint64_t wanna = mdbx_roundup2( - (env->me_maxreaders - 1) * sizeof(MDBX_reader) + sizeof(MDBX_lockinfo), - env->me_os_psize); -#ifndef NDEBUG - err = mdbx_ftruncate(env->me_lfd, size = 0); - if (unlikely(err != MDBX_SUCCESS)) - return err; -#endif - mdbx_jitter4testing(false); - - if (size != wanna) { - err = mdbx_ftruncate(env->me_lfd, wanna); - if (unlikely(err != MDBX_SUCCESS)) - return err; - size = wanna; - } - } else if (size > SSIZE_MAX || (size & (env->me_os_psize - 1)) || - size < env->me_os_psize) { - mdbx_notice("lck-file has invalid size %" PRIu64 " bytes", size); - return MDBX_PROBLEM; - } - - const size_t maxreaders = - ((size_t)size - sizeof(MDBX_lockinfo)) / sizeof(MDBX_reader) + 1; - if (maxreaders > UINT16_MAX) { - mdbx_error("lck-size too big (up to %" PRIuPTR " readers)", maxreaders); - return MDBX_PROBLEM; - } - env->me_maxreaders = (unsigned)maxreaders; - - err = mdbx_mmap(MDBX_WRITEMAP, &env->me_lck_mmap, (size_t)size, (size_t)size); - if (unlikely(err != MDBX_SUCCESS)) - return err; - -#ifdef MADV_DODUMP - (void)madvise(env->me_lck, size, MADV_DODUMP); -#endif - -#ifdef MADV_DONTFORK - if (madvise(env->me_lck, size, MADV_DONTFORK) < 0) - return errno; -#endif - -#ifdef MADV_WILLNEED - if (madvise(env->me_lck, size, MADV_WILLNEED) < 0) - return errno; -#endif - -#ifdef MADV_RANDOM - if (madvise(env->me_lck, size, MADV_RANDOM) < 0) - return errno; -#endif - - if (rc == MDBX_RESULT_TRUE) { - /* LY: exlcusive mode, init lck */ - memset(env->me_lck, 0, (size_t)size); - err = mdbx_lck_init(env); - if (err) - return err; - - env->me_lck->mti_magic_and_version = MDBX_LOCK_MAGIC; - env->me_lck->mti_os_and_format = MDBX_LOCK_FORMAT; - } else { - if (env->me_lck->mti_magic_and_version != MDBX_LOCK_MAGIC) { - mdbx_error("lock region has invalid magic/version"); - return ((env->me_lck->mti_magic_and_version >> 8) != MDBX_MAGIC) - ? MDBX_INVALID - : MDBX_VERSION_MISMATCH; - } - if (env->me_lck->mti_os_and_format != MDBX_LOCK_FORMAT) { - mdbx_error("lock region has os/format 0x%" PRIx32 ", expected 0x%" PRIx32, - env->me_lck->mti_os_and_format, MDBX_LOCK_FORMAT); - return MDBX_VERSION_MISMATCH; - } - } - - mdbx_assert(env, !MDBX_IS_ERROR(rc)); - env->me_oldest = &env->me_lck->mti_oldest; - return rc; -} - -/* Only a subset of the mdbx_env flags can be changed - * at runtime. Changing other flags requires closing the - * environment and re-opening it with the new flags. */ -#define CHANGEABLE \ - (MDBX_NOSYNC | MDBX_NOMETASYNC | MDBX_MAPASYNC | MDBX_NOMEMINIT | \ - MDBX_COALESCE | MDBX_PAGEPERTURB) -#define CHANGELESS \ - (MDBX_NOSUBDIR | MDBX_RDONLY | MDBX_WRITEMAP | MDBX_NOTLS | MDBX_NORDAHEAD | \ - MDBX_LIFORECLAIM) - -#if VALID_FLAGS & PERSISTENT_FLAGS & (CHANGEABLE | CHANGELESS) -#error "Persistent DB flags & env flags overlap, but both go in mm_flags" -#endif - -int __cold mdbx_env_open_ex(MDBX_env *env, const char *path, unsigned flags, - mode_t mode, int *exclusive) { - if (unlikely(!env || !path)) - return MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - if (env->me_fd != INVALID_HANDLE_VALUE || - (flags & ~(CHANGEABLE | CHANGELESS))) - return MDBX_EINVAL; - - size_t len_full, len = strlen(path); - if (flags & MDBX_NOSUBDIR) { - len_full = len + sizeof(MDBX_LOCK_SUFFIX) + len + 1; - } else { - len_full = len + sizeof(MDBX_LOCKNAME) + len + sizeof(MDBX_DATANAME); - } - char *lck_pathname = malloc(len_full); - if (!lck_pathname) - return MDBX_ENOMEM; - - char *dxb_pathname; - if (flags & MDBX_NOSUBDIR) { - dxb_pathname = lck_pathname + len + sizeof(MDBX_LOCK_SUFFIX); - sprintf(lck_pathname, "%s" MDBX_LOCK_SUFFIX, path); - strcpy(dxb_pathname, path); - } else { - dxb_pathname = lck_pathname + len + sizeof(MDBX_LOCKNAME); - sprintf(lck_pathname, "%s" MDBX_LOCKNAME, path); - sprintf(dxb_pathname, "%s" MDBX_DATANAME, path); - } - - int rc = MDBX_SUCCESS; - flags |= env->me_flags; - if (flags & MDBX_RDONLY) { - /* LY: silently ignore irrelevant flags when - * we're only getting read access */ - flags &= ~(MDBX_WRITEMAP | MDBX_MAPASYNC | MDBX_NOSYNC | MDBX_NOMETASYNC | - MDBX_COALESCE | MDBX_LIFORECLAIM | MDBX_NOMEMINIT); - } else { - if (!((env->me_free_pgs = mdbx_pnl_alloc(MDBX_PNL_UM_MAX)) && - (env->me_dirtylist = calloc(MDBX_PNL_UM_SIZE, sizeof(MDBX_ID2))))) - rc = MDBX_ENOMEM; - } - env->me_flags = flags |= MDBX_ENV_ACTIVE; - if (rc) - goto bailout; - - env->me_path = mdbx_strdup(path); - env->me_dbxs = calloc(env->me_maxdbs, sizeof(MDBX_dbx)); - env->me_dbflags = calloc(env->me_maxdbs, sizeof(uint16_t)); - env->me_dbiseqs = calloc(env->me_maxdbs, sizeof(unsigned)); - if (!(env->me_dbxs && env->me_path && env->me_dbflags && env->me_dbiseqs)) { - rc = MDBX_ENOMEM; - goto bailout; - } - env->me_dbxs[FREE_DBI].md_cmp = mdbx_cmp_int_ai; /* aligned MDBX_INTEGERKEY */ - - int oflags; - if (F_ISSET(flags, MDBX_RDONLY)) - oflags = O_RDONLY; - else - oflags = O_RDWR | O_CREAT; - - rc = mdbx_openfile(dxb_pathname, oflags, mode, &env->me_fd); - if (rc != MDBX_SUCCESS) - goto bailout; - - const int lck_rc = mdbx_setup_lck(env, lck_pathname, mode); - if (MDBX_IS_ERROR(lck_rc)) { - rc = lck_rc; - goto bailout; - } - - const int dxb_rc = mdbx_setup_dxb(env, lck_rc); - if (MDBX_IS_ERROR(dxb_rc)) { - rc = dxb_rc; - goto bailout; - } - - mdbx_debug("opened dbenv %p", (void *)env); - const unsigned mode_flags = - MDBX_WRITEMAP | MDBX_NOSYNC | MDBX_NOMETASYNC | MDBX_MAPASYNC; - if (lck_rc == MDBX_RESULT_TRUE) { - env->me_lck->mti_envmode = env->me_flags & (mode_flags | MDBX_RDONLY); - if (exclusive == NULL || *exclusive < 2) { - /* LY: downgrade lock only if exclusive access not requested. - * in case exclusive==1, just leave value as is. */ - rc = mdbx_lck_downgrade(env, true); - mdbx_debug("lck-downgrade-full: rc %i ", rc); - } else { - rc = mdbx_lck_downgrade(env, false); - mdbx_debug("lck-downgrade-partial: rc %i ", rc); - } - if (rc != MDBX_SUCCESS) - goto bailout; - } else { - if (exclusive) { - /* LY: just indicate that is not an exclusive access. */ - *exclusive = 0; - } - if ((env->me_flags & MDBX_RDONLY) == 0) { - while (env->me_lck->mti_envmode == MDBX_RDONLY) { - if (mdbx_atomic_compare_and_swap32(&env->me_lck->mti_envmode, - MDBX_RDONLY, - env->me_flags & mode_flags)) - break; - /* TODO: yield/relax cpu */ - } - if ((env->me_lck->mti_envmode ^ env->me_flags) & mode_flags) { - mdbx_error("current mode/flags incompatible with requested"); - rc = MDBX_INCOMPATIBLE; - goto bailout; - } - } - } - - if (env->me_lck && (env->me_flags & MDBX_NOTLS) == 0) { - rc = mdbx_rthc_alloc(&env->me_txkey, &env->me_lck->mti_readers[0], - &env->me_lck->mti_readers[env->me_maxreaders]); - if (unlikely(rc != MDBX_SUCCESS)) - goto bailout; - env->me_flags |= MDBX_ENV_TXKEY; - } - - if ((flags & MDBX_RDONLY) == 0) { - MDBX_txn *txn; - int tsize = sizeof(MDBX_txn), - size = tsize + - env->me_maxdbs * (sizeof(MDBX_db) + sizeof(MDBX_cursor *) + - sizeof(unsigned) + 1); - if ((env->me_pbuf = calloc(1, env->me_psize)) && (txn = calloc(1, size))) { - txn->mt_dbs = (MDBX_db *)((char *)txn + tsize); - txn->mt_cursors = (MDBX_cursor **)(txn->mt_dbs + env->me_maxdbs); - txn->mt_dbiseqs = (unsigned *)(txn->mt_cursors + env->me_maxdbs); - txn->mt_dbflags = (uint8_t *)(txn->mt_dbiseqs + env->me_maxdbs); - txn->mt_env = env; - txn->mt_dbxs = env->me_dbxs; - txn->mt_flags = MDBX_TXN_FINISHED; - env->me_txn0 = txn; - } else { - rc = MDBX_ENOMEM; - } - } - -#if MDBX_DEBUG - if (rc == MDBX_SUCCESS) { - MDBX_meta *meta = mdbx_meta_head(env); - MDBX_db *db = &meta->mm_dbs[MAIN_DBI]; - - mdbx_debug("opened database version %u, pagesize %u", - (uint8_t)meta->mm_magic_and_version, env->me_psize); - mdbx_debug("using meta page %" PRIaPGNO ", txn %" PRIaTXN "", - container_of(meta, MDBX_page, mp_data)->mp_pgno, - mdbx_meta_txnid_fluid(env, meta)); - mdbx_debug("depth: %u", db->md_depth); - mdbx_debug("entries: %" PRIu64 "", db->md_entries); - mdbx_debug("branch pages: %" PRIaPGNO "", db->md_branch_pages); - mdbx_debug("leaf pages: %" PRIaPGNO "", db->md_leaf_pages); - mdbx_debug("overflow pages: %" PRIaPGNO "", db->md_overflow_pages); - mdbx_debug("root: %" PRIaPGNO "", db->md_root); - } -#endif - -bailout: - if (rc) - mdbx_env_close0(env); - free(lck_pathname); - return rc; -} - -int __cold mdbx_env_open(MDBX_env *env, const char *path, unsigned flags, - mode_t mode) { - return mdbx_env_open_ex(env, path, flags, mode, NULL); -} - -/* Destroy resources from mdbx_env_open(), clear our readers & DBIs */ -static void __cold mdbx_env_close0(MDBX_env *env) { - if (!(env->me_flags & MDBX_ENV_ACTIVE)) - return; - env->me_flags &= ~MDBX_ENV_ACTIVE; - - /* Doing this here since me_dbxs may not exist during mdbx_env_close */ - if (env->me_dbxs) { - for (unsigned i = env->me_maxdbs; --i >= CORE_DBS;) - free(env->me_dbxs[i].md_name.iov_base); - free(env->me_dbxs); - } - - free(env->me_pbuf); - free(env->me_dbiseqs); - free(env->me_dbflags); - free(env->me_path); - free(env->me_dirtylist); - if (env->me_txn0) { - mdbx_txl_free(env->me_txn0->mt_lifo_reclaimed); - free(env->me_txn0); - } - mdbx_pnl_free(env->me_free_pgs); - - if (env->me_flags & MDBX_ENV_TXKEY) { - mdbx_rthc_remove(env->me_txkey); - env->me_flags &= ~MDBX_ENV_TXKEY; - } - - if (env->me_map) { - mdbx_munmap(&env->me_dxb_mmap); -#ifdef USE_VALGRIND - VALGRIND_DISCARD(env->me_valgrind_handle); - env->me_valgrind_handle = -1; -#endif - } - if (env->me_fd != INVALID_HANDLE_VALUE) { - (void)mdbx_closefile(env->me_fd); - env->me_fd = INVALID_HANDLE_VALUE; - } - - if (env->me_lck) - mdbx_munmap(&env->me_lck_mmap); - env->me_pid = 0; - env->me_oldest = nullptr; - - mdbx_lck_destroy(env); - if (env->me_lfd != INVALID_HANDLE_VALUE) { - (void)mdbx_closefile(env->me_lfd); - env->me_lfd = INVALID_HANDLE_VALUE; - } -} - -int __cold mdbx_env_close_ex(MDBX_env *env, int dont_sync) { - MDBX_page *dp; - int rc = MDBX_SUCCESS; - - if (unlikely(!env)) - return MDBX_EINVAL; - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - if (!dont_sync && !(env->me_flags & MDBX_RDONLY)) - rc = mdbx_env_sync(env, true); - - VALGRIND_DESTROY_MEMPOOL(env); - while ((dp = env->me_dpages) != NULL) { - ASAN_UNPOISON_MEMORY_REGION(&dp->mp_next, sizeof(dp->mp_next)); - VALGRIND_MAKE_MEM_DEFINED(&dp->mp_next, sizeof(dp->mp_next)); - env->me_dpages = dp->mp_next; - free(dp); - } - - mdbx_env_close0(env); - mdbx_ensure(env, mdbx_fastmutex_destroy(&env->me_dbi_lock) == MDBX_SUCCESS); - env->me_signature = 0; - free(env); - - return rc; -} - -void __cold mdbx_env_close(MDBX_env *env) { mdbx_env_close_ex(env, 0); } - -/* Compare two items pointing at aligned unsigned int's. */ -static int __hot mdbx_cmp_int_ai(const MDBX_val *a, const MDBX_val *b) { - mdbx_assert(NULL, a->iov_len == b->iov_len); - mdbx_assert(NULL, 0 == (uintptr_t)a->iov_base % sizeof(int) && - 0 == (uintptr_t)b->iov_base % sizeof(int)); - switch (a->iov_len) { - case 4: - return mdbx_cmp2int(*(uint32_t *)a->iov_base, *(uint32_t *)b->iov_base); - case 8: - return mdbx_cmp2int(*(uint64_t *)a->iov_base, *(uint64_t *)b->iov_base); - default: - mdbx_assert_fail(NULL, "invalid size for INTEGERKEY/INTEGERDUP", mdbx_func_, - __LINE__); - return 0; - } -} - -/* Compare two items pointing at 2-byte aligned unsigned int's. */ -static int __hot mdbx_cmp_int_a2(const MDBX_val *a, const MDBX_val *b) { - mdbx_assert(NULL, a->iov_len == b->iov_len); - mdbx_assert(NULL, 0 == (uintptr_t)a->iov_base % sizeof(uint16_t) && - 0 == (uintptr_t)b->iov_base % sizeof(uint16_t)); -#if UNALIGNED_OK - switch (a->iov_len) { - case 4: - return mdbx_cmp2int(*(uint32_t *)a->iov_base, *(uint32_t *)b->iov_base); - case 8: - return mdbx_cmp2int(*(uint64_t *)a->iov_base, *(uint64_t *)b->iov_base); - default: - mdbx_assert_fail(NULL, "invalid size for INTEGERKEY/INTEGERDUP", mdbx_func_, - __LINE__); - return 0; - } -#else - mdbx_assert(NULL, 0 == a->iov_len % sizeof(uint16_t)); - { - int diff; - const uint16_t *pa, *pb, *end; - -#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ - end = (const uint16_t *)a->iov_base; - pa = (const uint16_t *)((char *)a->iov_base + a->iov_len); - pb = (const uint16_t *)((char *)b->iov_base + a->iov_len); - do { - diff = *--pa - *--pb; -#else /* __BYTE_ORDER__ */ - end = (const uint16_t *)((char *)a->iov_base + a->iov_len); - pa = (const uint16_t *)a->iov_base; - pb = (const uint16_t *)b->iov_base; - do { - diff = *pa++ - *pb++; -#endif /* __BYTE_ORDER__ */ - if (likely(diff != 0)) - break; - } while (pa != end); - return diff; - } -#endif /* UNALIGNED_OK */ -} - -/* Compare two items pointing at unsigneds of unknown alignment. - * - * This is also set as MDBX_INTEGERDUP|MDBX_DUPFIXED's MDBX_dbx.md_dcmp. */ -static int __hot mdbx_cmp_int_ua(const MDBX_val *a, const MDBX_val *b) { - mdbx_assert(NULL, a->iov_len == b->iov_len); -#if UNALIGNED_OK - switch (a->iov_len) { - case 4: - return mdbx_cmp2int(*(uint32_t *)a->iov_base, *(uint32_t *)b->iov_base); - case 8: - return mdbx_cmp2int(*(uint64_t *)a->iov_base, *(uint64_t *)b->iov_base); - default: - mdbx_assert_fail(NULL, "invalid size for INTEGERKEY/INTEGERDUP", mdbx_func_, - __LINE__); - return 0; - } -#else - mdbx_assert(NULL, a->iov_len == sizeof(int) || a->iov_len == sizeof(size_t)); -#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ - { - int diff; - const uint8_t *pa, *pb; - - pa = (const uint8_t *)a->iov_base + a->iov_len; - pb = (const uint8_t *)b->iov_base + a->iov_len; - - do { - diff = *--pa - *--pb; - if (likely(diff != 0)) - break; - } while (pa != a->iov_base); - return diff; - } -#else /* __BYTE_ORDER__ */ - return memcmp(a->iov_base, b->iov_base, a->iov_len); -#endif /* __BYTE_ORDER__ */ -#endif /* UNALIGNED_OK */ -} - -/* Compare two items lexically */ -static int __hot mdbx_cmp_memn(const MDBX_val *a, const MDBX_val *b) { -/* LY: assumes that length of keys are NOT equal for most cases, - * if no then branch-prediction should mitigate the problem */ -#if 0 - /* LY: without branch instructions on x86, - * but isn't best for equal length of keys */ - int diff_len = mdbx_cmp2int(a->iov_len, b->iov_len); -#else - /* LY: best when length of keys are equal, - * but got a branch-penalty otherwise */ - if (likely(a->iov_len == b->iov_len)) - return memcmp(a->iov_base, b->iov_base, a->iov_len); - int diff_len = (a->iov_len < b->iov_len) ? -1 : 1; -#endif - size_t shortest = (a->iov_len < b->iov_len) ? a->iov_len : b->iov_len; - int diff_data = memcmp(a->iov_base, b->iov_base, shortest); - return likely(diff_data) ? diff_data : diff_len; -} - -/* Compare two items in reverse byte order */ -static int __hot mdbx_cmp_memnr(const MDBX_val *a, const MDBX_val *b) { - const uint8_t *pa, *pb, *end; - - pa = (const uint8_t *)a->iov_base + a->iov_len; - pb = (const uint8_t *)b->iov_base + b->iov_len; - size_t minlen = (a->iov_len < b->iov_len) ? a->iov_len : b->iov_len; - end = pa - minlen; - - while (pa != end) { - int diff = *--pa - *--pb; - if (likely(diff)) - return diff; - } - return mdbx_cmp2int(a->iov_len, b->iov_len); -} - -/* Search for key within a page, using binary search. - * Returns the smallest entry larger or equal to the key. - * If exactp is non-null, stores whether the found entry was an exact match - * in *exactp (1 or 0). - * Updates the cursor index with the index of the found entry. - * If no entry larger or equal to the key is found, returns NULL. */ -static MDBX_node *__hot mdbx_node_search(MDBX_cursor *mc, MDBX_val *key, - int *exactp) { - unsigned i = 0, nkeys; - int low, high; - int rc = 0; - MDBX_page *mp = mc->mc_pg[mc->mc_top]; - MDBX_node *node = NULL; - MDBX_val nodekey; - MDBX_cmp_func *cmp; - DKBUF; - - nkeys = NUMKEYS(mp); - - mdbx_debug("searching %u keys in %s %spage %" PRIaPGNO "", nkeys, - IS_LEAF(mp) ? "leaf" : "branch", IS_SUBP(mp) ? "sub-" : "", - mp->mp_pgno); - - low = IS_LEAF(mp) ? 0 : 1; - high = nkeys - 1; - cmp = mc->mc_dbx->md_cmp; - - /* Branch pages have no data, so if using integer keys, - * alignment is guaranteed. Use faster mdbx_cmp_int_ai. - */ - if (cmp == mdbx_cmp_int_a2 && IS_BRANCH(mp)) - cmp = mdbx_cmp_int_ai; - - if (IS_LEAF2(mp)) { - nodekey.iov_len = mc->mc_db->md_xsize; - node = NODEPTR(mp, 0); /* fake */ - while (low <= high) { - i = (low + high) >> 1; - nodekey.iov_base = LEAF2KEY(mp, i, nodekey.iov_len); - rc = cmp(key, &nodekey); - mdbx_debug("found leaf index %u [%s], rc = %i", i, DKEY(&nodekey), rc); - if (rc == 0) - break; - if (rc > 0) - low = i + 1; - else - high = i - 1; - } - } else { - while (low <= high) { - i = (low + high) >> 1; - - node = NODEPTR(mp, i); - nodekey.iov_len = NODEKSZ(node); - nodekey.iov_base = NODEKEY(node); - - rc = cmp(key, &nodekey); - if (IS_LEAF(mp)) - mdbx_debug("found leaf index %u [%s], rc = %i", i, DKEY(&nodekey), rc); - else - mdbx_debug("found branch index %u [%s -> %" PRIaPGNO "], rc = %i", i, - DKEY(&nodekey), NODEPGNO(node), rc); - if (rc == 0) - break; - if (rc > 0) - low = i + 1; - else - high = i - 1; - } - } - - if (rc > 0) /* Found entry is less than the key. */ - i++; /* Skip to get the smallest entry larger than key. */ - - if (exactp) - *exactp = (rc == 0 && nkeys > 0); - /* store the key index */ - mdbx_cassert(mc, i <= UINT16_MAX); - mc->mc_ki[mc->mc_top] = (indx_t)i; - if (i >= nkeys) - /* There is no entry larger or equal to the key. */ - return NULL; - - /* nodeptr is fake for LEAF2 */ - return IS_LEAF2(mp) ? node : NODEPTR(mp, i); -} - -#if 0 /* unused for now */ -static void mdbx_cursor_adjust(MDBX_cursor *mc, func) { - MDBX_cursor *m2; - - for (m2 = mc->mc_txn->mt_cursors[mc->mc_dbi]; m2; m2 = m2->mc_next) { - if (m2->mc_pg[m2->mc_top] == mc->mc_pg[mc->mc_top]) { - func(mc, m2); - } - } -} -#endif - -/* Pop a page off the top of the cursor's stack. */ -static void mdbx_cursor_pop(MDBX_cursor *mc) { - if (mc->mc_snum) { - mdbx_debug("popped page %" PRIaPGNO " off db %d cursor %p", - mc->mc_pg[mc->mc_top]->mp_pgno, DDBI(mc), (void *)mc); - - mc->mc_snum--; - if (mc->mc_snum) { - mc->mc_top--; - } else { - mc->mc_flags &= ~C_INITIALIZED; - } - } -} - -/* Push a page onto the top of the cursor's stack. - * Set MDBX_TXN_ERROR on failure. */ -static int mdbx_cursor_push(MDBX_cursor *mc, MDBX_page *mp) { - mdbx_debug("pushing page %" PRIaPGNO " on db %d cursor %p", mp->mp_pgno, - DDBI(mc), (void *)mc); - - if (unlikely(mc->mc_snum >= CURSOR_STACK)) { - mc->mc_txn->mt_flags |= MDBX_TXN_ERROR; - return MDBX_CURSOR_FULL; - } - - mdbx_cassert(mc, mc->mc_snum < UINT16_MAX); - mc->mc_top = mc->mc_snum++; - mc->mc_pg[mc->mc_top] = mp; - mc->mc_ki[mc->mc_top] = 0; - - return MDBX_SUCCESS; -} - -/* Find the address of the page corresponding to a given page number. - * Set MDBX_TXN_ERROR on failure. - * - * [in] mc the cursor accessing the page. - * [in] pgno the page number for the page to retrieve. - * [out] ret address of a pointer where the page's address will be - * stored. - * [out] lvl dirtylist inheritance level of found page. 1=current txn, - * 0=mapped page. - * - * Returns 0 on success, non-zero on failure. */ -static int mdbx_page_get(MDBX_cursor *mc, pgno_t pgno, MDBX_page **ret, - int *lvl) { - MDBX_txn *txn = mc->mc_txn; - MDBX_env *env = txn->mt_env; - MDBX_page *p = NULL; - int level; - - if (!(txn->mt_flags & (MDBX_TXN_RDONLY | MDBX_TXN_WRITEMAP))) { - MDBX_txn *tx2 = txn; - level = 1; - do { - MDBX_ID2L dl = tx2->mt_rw_dirtylist; - unsigned x; - /* Spilled pages were dirtied in this txn and flushed - * because the dirty list got full. Bring this page - * back in from the map (but don't unspill it here, - * leave that unless page_touch happens again). */ - if (tx2->mt_spill_pages) { - pgno_t pn = pgno << 1; - x = mdbx_pnl_search(tx2->mt_spill_pages, pn); - if (x <= tx2->mt_spill_pages[0] && tx2->mt_spill_pages[x] == pn) - goto mapped; - } - if (dl[0].mid) { - unsigned y = mdbx_mid2l_search(dl, pgno); - if (y <= dl[0].mid && dl[y].mid == pgno) { - p = dl[y].mptr; - goto done; - } - } - level++; - } while ((tx2 = tx2->mt_parent) != NULL); - } - - if (unlikely(pgno >= txn->mt_next_pgno)) { - mdbx_debug("page %" PRIaPGNO " not found", pgno); - txn->mt_flags |= MDBX_TXN_ERROR; - return MDBX_PAGE_NOTFOUND; - } - level = 0; - -mapped: - p = pgno2page(env, pgno); -/* TODO: check p->mp_validator here */ - -done: - *ret = p; - if (lvl) - *lvl = level; - return MDBX_SUCCESS; -} - -/* Finish mdbx_page_search() / mdbx_page_search_lowest(). - * The cursor is at the root page, set up the rest of it. */ -static int mdbx_page_search_root(MDBX_cursor *mc, MDBX_val *key, int flags) { - MDBX_page *mp = mc->mc_pg[mc->mc_top]; - int rc; - DKBUF; - - while (IS_BRANCH(mp)) { - MDBX_node *node; - indx_t i; - - mdbx_debug("branch page %" PRIaPGNO " has %u keys", mp->mp_pgno, - NUMKEYS(mp)); - /* Don't assert on branch pages in the FreeDB. We can get here - * while in the process of rebalancing a FreeDB branch page; we must - * let that proceed. ITS#8336 */ - mdbx_cassert(mc, !mc->mc_dbi || NUMKEYS(mp) > 1); - mdbx_debug("found index 0 to page %" PRIaPGNO "", NODEPGNO(NODEPTR(mp, 0))); - - if (flags & (MDBX_PS_FIRST | MDBX_PS_LAST)) { - i = 0; - if (flags & MDBX_PS_LAST) { - i = NUMKEYS(mp) - 1; - /* if already init'd, see if we're already in right place */ - if (mc->mc_flags & C_INITIALIZED) { - if (mc->mc_ki[mc->mc_top] == i) { - mc->mc_top = mc->mc_snum++; - mp = mc->mc_pg[mc->mc_top]; - goto ready; - } - } - } - } else { - int exact; - node = mdbx_node_search(mc, key, &exact); - if (node == NULL) - i = NUMKEYS(mp) - 1; - else { - i = mc->mc_ki[mc->mc_top]; - if (!exact) { - mdbx_cassert(mc, i > 0); - i--; - } - } - mdbx_debug("following index %u for key [%s]", i, DKEY(key)); - } - - mdbx_cassert(mc, i < NUMKEYS(mp)); - node = NODEPTR(mp, i); - - if (unlikely((rc = mdbx_page_get(mc, NODEPGNO(node), &mp, NULL)) != 0)) - return rc; - - mc->mc_ki[mc->mc_top] = i; - if (unlikely(rc = mdbx_cursor_push(mc, mp))) - return rc; - - ready: - if (flags & MDBX_PS_MODIFY) { - if (unlikely((rc = mdbx_page_touch(mc)) != 0)) - return rc; - mp = mc->mc_pg[mc->mc_top]; - } - } - - if (unlikely(!IS_LEAF(mp))) { - mdbx_debug("internal error, index points to a page with 0x%02x flags!?", - mp->mp_flags); - mc->mc_txn->mt_flags |= MDBX_TXN_ERROR; - return MDBX_CORRUPTED; - } - - mdbx_debug("found leaf page %" PRIaPGNO " for key [%s]", mp->mp_pgno, - DKEY(key)); - mc->mc_flags |= C_INITIALIZED; - mc->mc_flags &= ~C_EOF; - - return MDBX_SUCCESS; -} - -/* Search for the lowest key under the current branch page. - * This just bypasses a NUMKEYS check in the current page - * before calling mdbx_page_search_root(), because the callers - * are all in situations where the current page is known to - * be underfilled. */ -static int mdbx_page_search_lowest(MDBX_cursor *mc) { - MDBX_page *mp = mc->mc_pg[mc->mc_top]; - MDBX_node *node = NODEPTR(mp, 0); - int rc; - - if (unlikely((rc = mdbx_page_get(mc, NODEPGNO(node), &mp, NULL)) != 0)) - return rc; - - mc->mc_ki[mc->mc_top] = 0; - if (unlikely(rc = mdbx_cursor_push(mc, mp))) - return rc; - return mdbx_page_search_root(mc, NULL, MDBX_PS_FIRST); -} - -/* Search for the page a given key should be in. - * Push it and its parent pages on the cursor stack. - * - * [in,out] mc the cursor for this operation. - * [in] key the key to search for, or NULL for first/last page. - * [in] flags If MDBX_PS_MODIFY is set, visited pages in the DB - * are touched (updated with new page numbers). - * If MDBX_PS_FIRST or MDBX_PS_LAST is set, find first or last - * leaf. - * This is used by mdbx_cursor_first() and mdbx_cursor_last(). - * If MDBX_PS_ROOTONLY set, just fetch root node, no further - * lookups. - * - * Returns 0 on success, non-zero on failure. */ -static int mdbx_page_search(MDBX_cursor *mc, MDBX_val *key, int flags) { - int rc; - pgno_t root; - - /* Make sure the txn is still viable, then find the root from - * the txn's db table and set it as the root of the cursor's stack. */ - if (unlikely(mc->mc_txn->mt_flags & MDBX_TXN_BLOCKED)) { - mdbx_debug("transaction has failed, must abort"); - return MDBX_BAD_TXN; - } - - mdbx_cassert(mc, mc->mc_txn->mt_txnid >= mc->mc_txn->mt_env->me_oldest[0]); - /* Make sure we're using an up-to-date root */ - if (unlikely(*mc->mc_dbflag & DB_STALE)) { - MDBX_cursor mc2; - if (unlikely(TXN_DBI_CHANGED(mc->mc_txn, mc->mc_dbi))) - return MDBX_BAD_DBI; - mdbx_cursor_init(&mc2, mc->mc_txn, MAIN_DBI, NULL); - rc = mdbx_page_search(&mc2, &mc->mc_dbx->md_name, 0); - if (rc) - return rc; - { - MDBX_val data; - int exact = 0; - MDBX_node *leaf = mdbx_node_search(&mc2, &mc->mc_dbx->md_name, &exact); - if (!exact) - return MDBX_NOTFOUND; - if (unlikely((leaf->mn_flags & (F_DUPDATA | F_SUBDATA)) != F_SUBDATA)) - return MDBX_INCOMPATIBLE; /* not a named DB */ - rc = mdbx_node_read(&mc2, leaf, &data); - if (rc) - return rc; - - uint16_t md_flags; - memcpy(&md_flags, ((char *)data.iov_base + offsetof(MDBX_db, md_flags)), - sizeof(uint16_t)); - /* The txn may not know this DBI, or another process may - * have dropped and recreated the DB with other flags. */ - if (unlikely((mc->mc_db->md_flags & PERSISTENT_FLAGS) != md_flags)) - return MDBX_INCOMPATIBLE; - memcpy(mc->mc_db, data.iov_base, sizeof(MDBX_db)); - } - *mc->mc_dbflag &= ~DB_STALE; - } - root = mc->mc_db->md_root; - - if (unlikely(root == P_INVALID)) { /* Tree is empty. */ - mdbx_debug("tree is empty"); - return MDBX_NOTFOUND; - } - - mdbx_cassert(mc, mc->mc_txn->mt_txnid >= mc->mc_txn->mt_env->me_oldest[0]); - mdbx_cassert(mc, root >= NUM_METAS); - if (!mc->mc_pg[0] || mc->mc_pg[0]->mp_pgno != root) - if (unlikely((rc = mdbx_page_get(mc, root, &mc->mc_pg[0], NULL)) != 0)) - return rc; - - mc->mc_snum = 1; - mc->mc_top = 0; - - mdbx_debug("db %d root page %" PRIaPGNO " has flags 0x%X", DDBI(mc), root, - mc->mc_pg[0]->mp_flags); - - if (flags & MDBX_PS_MODIFY) { - if (unlikely(rc = mdbx_page_touch(mc))) - return rc; - } - - if (flags & MDBX_PS_ROOTONLY) - return MDBX_SUCCESS; - - return mdbx_page_search_root(mc, key, flags); -} - -static int mdbx_ovpage_free(MDBX_cursor *mc, MDBX_page *mp) { - MDBX_txn *txn = mc->mc_txn; - pgno_t pg = mp->mp_pgno; - unsigned x = 0, ovpages = mp->mp_pages; - MDBX_env *env = txn->mt_env; - MDBX_PNL sl = txn->mt_spill_pages; - pgno_t pn = pg << 1; - int rc; - - mdbx_debug("free ov page %" PRIaPGNO " (%u)", pg, ovpages); - /* If the page is dirty or on the spill list we just acquired it, - * so we should give it back to our current free list, if any. - * Otherwise put it onto the list of pages we freed in this txn. - * - * Won't create me_reclaimed_pglist: me_last_reclaimed must be inited along - * with it. - * Unsupported in nested txns: They would need to hide the page - * range in ancestor txns' dirty and spilled lists. */ - if (env->me_reclaimed_pglist && !txn->mt_parent && - ((mp->mp_flags & P_DIRTY) || - (sl && (x = mdbx_pnl_search(sl, pn)) <= sl[0] && sl[x] == pn))) { - unsigned i, j; - pgno_t *mop; - MDBX_ID2 *dl, ix, iy; - rc = mdbx_pnl_need(&env->me_reclaimed_pglist, ovpages); - if (unlikely(rc)) - return rc; - if (!(mp->mp_flags & P_DIRTY)) { - /* This page is no longer spilled */ - if (x == sl[0]) - sl[0]--; - else - sl[x] |= 1; - goto release; - } - /* Remove from dirty list */ - dl = txn->mt_rw_dirtylist; - x = dl[0].mid--; - for (ix = dl[x]; ix.mptr != mp; ix = iy) { - if (likely(x > 1)) { - x--; - iy = dl[x]; - dl[x] = ix; - } else { - mdbx_cassert(mc, x > 1); - mdbx_error("not found page 0x%p #%" PRIaPGNO " in the dirtylist", mp, - mp->mp_pgno); - j = ++(dl[0].mid); - dl[j] = ix; /* Unsorted. OK when MDBX_TXN_ERROR. */ - txn->mt_flags |= MDBX_TXN_ERROR; - return MDBX_PROBLEM; - } - } - txn->mt_dirtyroom++; - if (!(env->me_flags & MDBX_WRITEMAP)) - mdbx_dpage_free(env, mp); - release: - /* Insert in me_reclaimed_pglist */ - mop = env->me_reclaimed_pglist; - j = mop[0] + ovpages; - for (i = mop[0]; i && mop[i] < pg; i--) - mop[j--] = mop[i]; - while (j > i) - mop[j--] = pg++; - mop[0] += ovpages; - } else { - rc = mdbx_pnl_append_range(&txn->mt_befree_pages, pg, ovpages); - if (unlikely(rc)) - return rc; - } - mc->mc_db->md_overflow_pages -= ovpages; - return 0; -} - -/* Return the data associated with a given node. - * - * [in] mc The cursor for this operation. - * [in] leaf The node being read. - * [out] data Updated to point to the node's data. - * - * Returns 0 on success, non-zero on failure. */ -static __inline int mdbx_node_read(MDBX_cursor *mc, MDBX_node *leaf, - MDBX_val *data) { - MDBX_page *omp; /* overflow page */ - pgno_t pgno; - int rc; - - mdbx_cassert(mc, mc->mc_txn->mt_txnid >= mc->mc_txn->mt_env->me_oldest[0]); - if (!F_ISSET(leaf->mn_flags, F_BIGDATA)) { - data->iov_len = NODEDSZ(leaf); - data->iov_base = NODEDATA(leaf); - return MDBX_SUCCESS; - } - - /* Read overflow data. */ - data->iov_len = NODEDSZ(leaf); - memcpy(&pgno, NODEDATA(leaf), sizeof(pgno)); - if (unlikely((rc = mdbx_page_get(mc, pgno, &omp, NULL)) != 0)) { - mdbx_debug("read overflow page %" PRIaPGNO " failed", pgno); - return rc; - } - data->iov_base = PAGEDATA(omp); - - return MDBX_SUCCESS; -} - -int mdbx_get(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *data) { - MDBX_cursor mc; - MDBX_xcursor mx; - int exact = 0; - DKBUF; - - mdbx_debug("===> get db %u key [%s]", dbi, DKEY(key)); - - if (unlikely(!key || !data || !txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - if (unlikely(txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - mdbx_cursor_init(&mc, txn, dbi, &mx); - return mdbx_cursor_set(&mc, key, data, MDBX_SET, &exact); -} - -/* Find a sibling for a page. - * Replaces the page at the top of the cursor's stack with the specified - * sibling, if one exists. - * - * [in] mc The cursor for this operation. - * [in] move_right Non-zero if the right sibling is requested, - * otherwise the left sibling. - * - * Returns 0 on success, non-zero on failure. */ -static int mdbx_cursor_sibling(MDBX_cursor *mc, int move_right) { - int rc; - MDBX_node *indx; - MDBX_page *mp; - - mdbx_cassert(mc, mc->mc_txn->mt_txnid >= mc->mc_txn->mt_env->me_oldest[0]); - if (unlikely(mc->mc_snum < 2)) { - return MDBX_NOTFOUND; /* root has no siblings */ - } - - mdbx_cursor_pop(mc); - mdbx_debug("parent page is page %" PRIaPGNO ", index %u", - mc->mc_pg[mc->mc_top]->mp_pgno, mc->mc_ki[mc->mc_top]); - - if (move_right - ? (mc->mc_ki[mc->mc_top] + 1u >= NUMKEYS(mc->mc_pg[mc->mc_top])) - : (mc->mc_ki[mc->mc_top] == 0)) { - mdbx_debug("no more keys left, moving to %s sibling", - move_right ? "right" : "left"); - if (unlikely((rc = mdbx_cursor_sibling(mc, move_right)) != MDBX_SUCCESS)) { - /* undo cursor_pop before returning */ - mc->mc_top++; - mc->mc_snum++; - return rc; - } - } else { - if (move_right) - mc->mc_ki[mc->mc_top]++; - else - mc->mc_ki[mc->mc_top]--; - mdbx_debug("just moving to %s index key %u", move_right ? "right" : "left", - mc->mc_ki[mc->mc_top]); - } - mdbx_cassert(mc, IS_BRANCH(mc->mc_pg[mc->mc_top])); - - indx = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); - if (unlikely((rc = mdbx_page_get(mc, NODEPGNO(indx), &mp, NULL)) != 0)) { - /* mc will be inconsistent if caller does mc_snum++ as above */ - mc->mc_flags &= ~(C_INITIALIZED | C_EOF); - return rc; - } - - mdbx_cursor_push(mc, mp); - if (!move_right) - mc->mc_ki[mc->mc_top] = NUMKEYS(mp) - 1; - - return MDBX_SUCCESS; -} - -/* Move the cursor to the next data item. */ -static int mdbx_cursor_next(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data, - MDBX_cursor_op op) { - MDBX_page *mp; - MDBX_node *leaf; - int rc; - - if ((mc->mc_flags & C_DEL) && op == MDBX_NEXT_DUP) - return MDBX_NOTFOUND; - - if (!(mc->mc_flags & C_INITIALIZED)) - return mdbx_cursor_first(mc, key, data); - - mp = mc->mc_pg[mc->mc_top]; - if (mc->mc_flags & C_EOF) { - if (mc->mc_ki[mc->mc_top] + 1u >= NUMKEYS(mp)) - return MDBX_NOTFOUND; - mc->mc_flags ^= C_EOF; - } - - if (mc->mc_db->md_flags & MDBX_DUPSORT) { - leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - if (op == MDBX_NEXT || op == MDBX_NEXT_DUP) { - rc = - mdbx_cursor_next(&mc->mc_xcursor->mx_cursor, data, NULL, MDBX_NEXT); - if (op != MDBX_NEXT || rc != MDBX_NOTFOUND) { - if (likely(rc == MDBX_SUCCESS)) - MDBX_GET_KEY(leaf, key); - return rc; - } - } - } else { - mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED | C_EOF); - if (op == MDBX_NEXT_DUP) - return MDBX_NOTFOUND; - } - } - - mdbx_debug("cursor_next: top page is %" PRIaPGNO " in cursor %p", mp->mp_pgno, - (void *)mc); - if (mc->mc_flags & C_DEL) { - mc->mc_flags ^= C_DEL; - goto skip; - } - - if (mc->mc_ki[mc->mc_top] + 1u >= NUMKEYS(mp)) { - mdbx_debug("=====> move to next sibling page"); - if (unlikely((rc = mdbx_cursor_sibling(mc, 1)) != MDBX_SUCCESS)) { - mc->mc_flags |= C_EOF; - return rc; - } - mp = mc->mc_pg[mc->mc_top]; - mdbx_debug("next page is %" PRIaPGNO ", key index %u", mp->mp_pgno, - mc->mc_ki[mc->mc_top]); - } else - mc->mc_ki[mc->mc_top]++; - -skip: - mdbx_debug("==> cursor points to page %" PRIaPGNO - " with %u keys, key index %u", - mp->mp_pgno, NUMKEYS(mp), mc->mc_ki[mc->mc_top]); - - if (IS_LEAF2(mp)) { - key->iov_len = mc->mc_db->md_xsize; - key->iov_base = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], key->iov_len); - return MDBX_SUCCESS; - } - - mdbx_cassert(mc, IS_LEAF(mp)); - leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); - - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_xcursor_init1(mc, leaf); - } - if (data) { - if (unlikely((rc = mdbx_node_read(mc, leaf, data)) != MDBX_SUCCESS)) - return rc; - - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - rc = mdbx_cursor_first(&mc->mc_xcursor->mx_cursor, data, NULL); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - } - } - - MDBX_GET_KEY(leaf, key); - return MDBX_SUCCESS; -} - -/* Move the cursor to the previous data item. */ -static int mdbx_cursor_prev(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data, - MDBX_cursor_op op) { - MDBX_page *mp; - MDBX_node *leaf; - int rc; - - if ((mc->mc_flags & C_DEL) && op == MDBX_PREV_DUP) - return MDBX_NOTFOUND; - - if (!(mc->mc_flags & C_INITIALIZED)) { - rc = mdbx_cursor_last(mc, key, data); - if (unlikely(rc)) - return rc; - mc->mc_ki[mc->mc_top]++; - } - - mp = mc->mc_pg[mc->mc_top]; - if ((mc->mc_db->md_flags & MDBX_DUPSORT) && - mc->mc_ki[mc->mc_top] < NUMKEYS(mp)) { - leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - if (op == MDBX_PREV || op == MDBX_PREV_DUP) { - rc = - mdbx_cursor_prev(&mc->mc_xcursor->mx_cursor, data, NULL, MDBX_PREV); - if (op != MDBX_PREV || rc != MDBX_NOTFOUND) { - if (likely(rc == MDBX_SUCCESS)) { - MDBX_GET_KEY(leaf, key); - mc->mc_flags &= ~C_EOF; - } - return rc; - } - } - } else { - mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED | C_EOF); - if (op == MDBX_PREV_DUP) - return MDBX_NOTFOUND; - } - } - - mdbx_debug("cursor_prev: top page is %" PRIaPGNO " in cursor %p", mp->mp_pgno, - (void *)mc); - - mc->mc_flags &= ~(C_EOF | C_DEL); - - if (mc->mc_ki[mc->mc_top] == 0) { - mdbx_debug("=====> move to prev sibling page"); - if ((rc = mdbx_cursor_sibling(mc, 0)) != MDBX_SUCCESS) { - return rc; - } - mp = mc->mc_pg[mc->mc_top]; - mc->mc_ki[mc->mc_top] = NUMKEYS(mp) - 1; - mdbx_debug("prev page is %" PRIaPGNO ", key index %u", mp->mp_pgno, - mc->mc_ki[mc->mc_top]); - } else - mc->mc_ki[mc->mc_top]--; - - mdbx_debug("==> cursor points to page %" PRIaPGNO - " with %u keys, key index %u", - mp->mp_pgno, NUMKEYS(mp), mc->mc_ki[mc->mc_top]); - - if (IS_LEAF2(mp)) { - key->iov_len = mc->mc_db->md_xsize; - key->iov_base = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], key->iov_len); - return MDBX_SUCCESS; - } - - mdbx_cassert(mc, IS_LEAF(mp)); - leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); - - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_xcursor_init1(mc, leaf); - } - if (data) { - if (unlikely((rc = mdbx_node_read(mc, leaf, data)) != MDBX_SUCCESS)) - return rc; - - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - rc = mdbx_cursor_last(&mc->mc_xcursor->mx_cursor, data, NULL); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - } - } - - MDBX_GET_KEY(leaf, key); - return MDBX_SUCCESS; -} - -/* Set the cursor on a specific data item. */ -static int mdbx_cursor_set(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data, - MDBX_cursor_op op, int *exactp) { - int rc; - MDBX_page *mp; - MDBX_node *leaf = NULL; - DKBUF; - - mdbx_cassert(mc, mc->mc_txn->mt_txnid >= mc->mc_txn->mt_env->me_oldest[0]); - if ((mc->mc_db->md_flags & MDBX_INTEGERKEY) && - unlikely(key->iov_len != sizeof(uint32_t) && - key->iov_len != sizeof(uint64_t))) { - mdbx_cassert(mc, !"key-size is invalid for MDBX_INTEGERKEY"); - return MDBX_BAD_VALSIZE; - } - - if (mc->mc_xcursor) - mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED | C_EOF); - - /* See if we're already on the right page */ - if (mc->mc_flags & C_INITIALIZED) { - MDBX_val nodekey; - - mp = mc->mc_pg[mc->mc_top]; - if (!NUMKEYS(mp)) { - mc->mc_ki[mc->mc_top] = 0; - return MDBX_NOTFOUND; - } - if (mp->mp_flags & P_LEAF2) { - nodekey.iov_len = mc->mc_db->md_xsize; - nodekey.iov_base = LEAF2KEY(mp, 0, nodekey.iov_len); - } else { - leaf = NODEPTR(mp, 0); - MDBX_GET_KEY2(leaf, nodekey); - } - rc = mc->mc_dbx->md_cmp(key, &nodekey); - if (rc == 0) { - /* Probably happens rarely, but first node on the page - * was the one we wanted. - */ - mc->mc_ki[mc->mc_top] = 0; - if (exactp) - *exactp = 1; - goto set1; - } - if (rc > 0) { - unsigned i; - unsigned nkeys = NUMKEYS(mp); - if (nkeys > 1) { - if (mp->mp_flags & P_LEAF2) { - nodekey.iov_base = LEAF2KEY(mp, nkeys - 1, nodekey.iov_len); - } else { - leaf = NODEPTR(mp, nkeys - 1); - MDBX_GET_KEY2(leaf, nodekey); - } - rc = mc->mc_dbx->md_cmp(key, &nodekey); - if (rc == 0) { - /* last node was the one we wanted */ - mdbx_cassert(mc, nkeys >= 1 && nkeys <= UINT16_MAX + 1); - mc->mc_ki[mc->mc_top] = (indx_t)(nkeys - 1); - if (exactp) - *exactp = 1; - goto set1; - } - if (rc < 0) { - if (mc->mc_ki[mc->mc_top] < NUMKEYS(mp)) { - /* This is definitely the right page, skip search_page */ - if (mp->mp_flags & P_LEAF2) { - nodekey.iov_base = - LEAF2KEY(mp, mc->mc_ki[mc->mc_top], nodekey.iov_len); - } else { - leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); - MDBX_GET_KEY2(leaf, nodekey); - } - rc = mc->mc_dbx->md_cmp(key, &nodekey); - if (rc == 0) { - /* current node was the one we wanted */ - if (exactp) - *exactp = 1; - goto set1; - } - } - rc = 0; - mc->mc_flags &= ~C_EOF; - goto set2; - } - } - /* If any parents have right-sibs, search. - * Otherwise, there's nothing further. */ - for (i = 0; i < mc->mc_top; i++) - if (mc->mc_ki[i] < NUMKEYS(mc->mc_pg[i]) - 1) - break; - if (i == mc->mc_top) { - /* There are no other pages */ - mdbx_cassert(mc, nkeys <= UINT16_MAX); - mc->mc_ki[mc->mc_top] = (uint16_t)nkeys; - return MDBX_NOTFOUND; - } - } - if (!mc->mc_top) { - /* There are no other pages */ - mc->mc_ki[mc->mc_top] = 0; - if (op == MDBX_SET_RANGE && !exactp) { - rc = 0; - goto set1; - } else - return MDBX_NOTFOUND; - } - } else { - mc->mc_pg[0] = 0; - } - - rc = mdbx_page_search(mc, key, 0); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - mp = mc->mc_pg[mc->mc_top]; - mdbx_cassert(mc, IS_LEAF(mp)); - -set2: - leaf = mdbx_node_search(mc, key, exactp); - if (exactp != NULL && !*exactp) { - /* MDBX_SET specified and not an exact match. */ - return MDBX_NOTFOUND; - } - - if (leaf == NULL) { - mdbx_debug("===> inexact leaf not found, goto sibling"); - if (unlikely((rc = mdbx_cursor_sibling(mc, 1)) != MDBX_SUCCESS)) { - mc->mc_flags |= C_EOF; - return rc; /* no entries matched */ - } - mp = mc->mc_pg[mc->mc_top]; - mdbx_cassert(mc, IS_LEAF(mp)); - leaf = NODEPTR(mp, 0); - } - -set1: - mc->mc_flags |= C_INITIALIZED; - mc->mc_flags &= ~C_EOF; - - if (IS_LEAF2(mp)) { - if (op == MDBX_SET_RANGE || op == MDBX_SET_KEY) { - key->iov_len = mc->mc_db->md_xsize; - key->iov_base = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], key->iov_len); - } - return MDBX_SUCCESS; - } - - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_xcursor_init1(mc, leaf); - } - if (likely(data)) { - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - if (op == MDBX_SET || op == MDBX_SET_KEY || op == MDBX_SET_RANGE) { - rc = mdbx_cursor_first(&mc->mc_xcursor->mx_cursor, data, NULL); - } else { - int ex2, *ex2p; - if (op == MDBX_GET_BOTH) { - ex2p = &ex2; - ex2 = 0; - } else { - ex2p = NULL; - } - rc = mdbx_cursor_set(&mc->mc_xcursor->mx_cursor, data, NULL, - MDBX_SET_RANGE, ex2p); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - } - } else if (op == MDBX_GET_BOTH || op == MDBX_GET_BOTH_RANGE) { - MDBX_val olddata; - if (unlikely((rc = mdbx_node_read(mc, leaf, &olddata)) != MDBX_SUCCESS)) - return rc; - rc = mc->mc_dbx->md_dcmp(data, &olddata); - if (rc) { - if (op == MDBX_GET_BOTH || rc > 0) - return MDBX_NOTFOUND; - rc = 0; - } - *data = olddata; - } else { - if (mc->mc_xcursor) - mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED | C_EOF); - if (unlikely((rc = mdbx_node_read(mc, leaf, data)) != MDBX_SUCCESS)) - return rc; - } - } - - /* The key already matches in all other cases */ - if (op == MDBX_SET_RANGE || op == MDBX_SET_KEY) - MDBX_GET_KEY(leaf, key); - - mdbx_debug("==> cursor placed on key [%s], data [%s]", DKEY(key), DVAL(data)); - return rc; -} - -/* Move the cursor to the first item in the database. */ -static int mdbx_cursor_first(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data) { - int rc; - MDBX_node *leaf; - - if (mc->mc_xcursor) - mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED | C_EOF); - - if (!(mc->mc_flags & C_INITIALIZED) || mc->mc_top) { - rc = mdbx_page_search(mc, NULL, MDBX_PS_FIRST); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - } - mdbx_cassert(mc, IS_LEAF(mc->mc_pg[mc->mc_top])); - - leaf = NODEPTR(mc->mc_pg[mc->mc_top], 0); - mc->mc_flags |= C_INITIALIZED; - mc->mc_flags &= ~C_EOF; - - mc->mc_ki[mc->mc_top] = 0; - - if (IS_LEAF2(mc->mc_pg[mc->mc_top])) { - key->iov_len = mc->mc_db->md_xsize; - key->iov_base = LEAF2KEY(mc->mc_pg[mc->mc_top], 0, key->iov_len); - return MDBX_SUCCESS; - } - - if (likely(data)) { - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_cassert(mc, mc->mc_xcursor != nullptr); - mdbx_xcursor_init1(mc, leaf); - rc = mdbx_cursor_first(&mc->mc_xcursor->mx_cursor, data, NULL); - if (unlikely(rc)) - return rc; - } else { - if (unlikely((rc = mdbx_node_read(mc, leaf, data)) != MDBX_SUCCESS)) - return rc; - } - } - MDBX_GET_KEY(leaf, key); - return MDBX_SUCCESS; -} - -/* Move the cursor to the last item in the database. */ -static int mdbx_cursor_last(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data) { - int rc; - MDBX_node *leaf; - - if (mc->mc_xcursor) - mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED | C_EOF); - - if (likely((mc->mc_flags & (C_EOF | C_DEL)) != C_EOF)) { - if (!(mc->mc_flags & C_INITIALIZED) || mc->mc_top) { - rc = mdbx_page_search(mc, NULL, MDBX_PS_LAST); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - } - mdbx_cassert(mc, IS_LEAF(mc->mc_pg[mc->mc_top])); - } - - mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top]) - 1; - mc->mc_flags |= C_INITIALIZED | C_EOF; - leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); - - if (IS_LEAF2(mc->mc_pg[mc->mc_top])) { - key->iov_len = mc->mc_db->md_xsize; - key->iov_base = - LEAF2KEY(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], key->iov_len); - return MDBX_SUCCESS; - } - - if (likely(data)) { - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_cassert(mc, mc->mc_xcursor != nullptr); - mdbx_xcursor_init1(mc, leaf); - rc = mdbx_cursor_last(&mc->mc_xcursor->mx_cursor, data, NULL); - if (unlikely(rc)) - return rc; - } else { - if (unlikely((rc = mdbx_node_read(mc, leaf, data)) != MDBX_SUCCESS)) - return rc; - } - } - - MDBX_GET_KEY(leaf, key); - return MDBX_SUCCESS; -} - -int mdbx_cursor_get(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data, - MDBX_cursor_op op) { - int rc; - int exact = 0; - int (*mfunc)(MDBX_cursor * mc, MDBX_val * key, MDBX_val * data); - - if (unlikely(mc == NULL)) - return MDBX_EINVAL; - - if (unlikely(mc->mc_signature != MDBX_MC_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(mc->mc_txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - mdbx_cassert(mc, mc->mc_txn->mt_txnid >= mc->mc_txn->mt_env->me_oldest[0]); - switch (op) { - case MDBX_GET_CURRENT: { - if (unlikely(!(mc->mc_flags & C_INITIALIZED))) - return MDBX_EINVAL; - MDBX_page *mp = mc->mc_pg[mc->mc_top]; - unsigned nkeys = NUMKEYS(mp); - if (mc->mc_ki[mc->mc_top] >= nkeys) { - mdbx_cassert(mc, nkeys <= UINT16_MAX); - mc->mc_ki[mc->mc_top] = (uint16_t)nkeys; - return MDBX_NOTFOUND; - } - assert(nkeys > 0); - - rc = MDBX_SUCCESS; - if (IS_LEAF2(mp)) { - key->iov_len = mc->mc_db->md_xsize; - key->iov_base = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], key->iov_len); - } else { - MDBX_node *leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); - MDBX_GET_KEY(leaf, key); - if (data) { - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - if (unlikely(!(mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED))) { - mdbx_xcursor_init1(mc, leaf); - rc = mdbx_cursor_first(&mc->mc_xcursor->mx_cursor, data, NULL); - if (unlikely(rc)) - return rc; - } - rc = mdbx_cursor_get(&mc->mc_xcursor->mx_cursor, data, NULL, - MDBX_GET_CURRENT); - } else { - rc = mdbx_node_read(mc, leaf, data); - } - if (unlikely(rc)) - return rc; - } - } - break; - } - case MDBX_GET_BOTH: - case MDBX_GET_BOTH_RANGE: - if (unlikely(data == NULL)) - return MDBX_EINVAL; - if (unlikely(mc->mc_xcursor == NULL)) - return MDBX_INCOMPATIBLE; - /* FALLTHRU */ - case MDBX_SET: - case MDBX_SET_KEY: - case MDBX_SET_RANGE: - if (unlikely(key == NULL)) - return MDBX_EINVAL; - rc = mdbx_cursor_set(mc, key, data, op, - op == MDBX_SET_RANGE ? NULL : &exact); - break; - case MDBX_GET_MULTIPLE: - if (unlikely(data == NULL || !(mc->mc_flags & C_INITIALIZED))) - return MDBX_EINVAL; - if (unlikely(!(mc->mc_db->md_flags & MDBX_DUPFIXED))) - return MDBX_INCOMPATIBLE; - rc = MDBX_SUCCESS; - if (!(mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) || - (mc->mc_xcursor->mx_cursor.mc_flags & C_EOF)) - break; - goto fetchm; - case MDBX_NEXT_MULTIPLE: - if (unlikely(data == NULL)) - return MDBX_EINVAL; - if (unlikely(!(mc->mc_db->md_flags & MDBX_DUPFIXED))) - return MDBX_INCOMPATIBLE; - rc = mdbx_cursor_next(mc, key, data, MDBX_NEXT_DUP); - if (rc == MDBX_SUCCESS) { - if (mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) { - MDBX_cursor *mx; - fetchm: - mx = &mc->mc_xcursor->mx_cursor; - data->iov_len = NUMKEYS(mx->mc_pg[mx->mc_top]) * mx->mc_db->md_xsize; - data->iov_base = PAGEDATA(mx->mc_pg[mx->mc_top]); - mx->mc_ki[mx->mc_top] = NUMKEYS(mx->mc_pg[mx->mc_top]) - 1; - } else { - rc = MDBX_NOTFOUND; - } - } - break; - case MDBX_PREV_MULTIPLE: - if (data == NULL) - return MDBX_EINVAL; - if (!(mc->mc_db->md_flags & MDBX_DUPFIXED)) - return MDBX_INCOMPATIBLE; - rc = MDBX_SUCCESS; - if (!(mc->mc_flags & C_INITIALIZED)) - rc = mdbx_cursor_last(mc, key, data); - if (rc == MDBX_SUCCESS) { - MDBX_cursor *mx = &mc->mc_xcursor->mx_cursor; - if (mx->mc_flags & C_INITIALIZED) { - rc = mdbx_cursor_sibling(mx, 0); - if (rc == MDBX_SUCCESS) - goto fetchm; - } else { - rc = MDBX_NOTFOUND; - } - } - break; - case MDBX_NEXT: - case MDBX_NEXT_DUP: - case MDBX_NEXT_NODUP: - rc = mdbx_cursor_next(mc, key, data, op); - break; - case MDBX_PREV: - case MDBX_PREV_DUP: - case MDBX_PREV_NODUP: - rc = mdbx_cursor_prev(mc, key, data, op); - break; - case MDBX_FIRST: - rc = mdbx_cursor_first(mc, key, data); - break; - case MDBX_FIRST_DUP: - mfunc = mdbx_cursor_first; - mmove: - if (unlikely(data == NULL || !(mc->mc_flags & C_INITIALIZED))) - return MDBX_EINVAL; - if (unlikely(mc->mc_xcursor == NULL)) - return MDBX_INCOMPATIBLE; - if (mc->mc_ki[mc->mc_top] >= NUMKEYS(mc->mc_pg[mc->mc_top])) { - mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top]); - return MDBX_NOTFOUND; - } - { - MDBX_node *leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); - if (!F_ISSET(leaf->mn_flags, F_DUPDATA)) { - MDBX_GET_KEY(leaf, key); - rc = mdbx_node_read(mc, leaf, data); - break; - } - } - if (unlikely(!(mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED))) - return MDBX_EINVAL; - rc = mfunc(&mc->mc_xcursor->mx_cursor, data, NULL); - break; - case MDBX_LAST: - rc = mdbx_cursor_last(mc, key, data); - break; - case MDBX_LAST_DUP: - mfunc = mdbx_cursor_last; - goto mmove; - default: - mdbx_debug("unhandled/unimplemented cursor operation %u", op); - return MDBX_EINVAL; - } - - mc->mc_flags &= ~C_DEL; - return rc; -} - -/* Touch all the pages in the cursor stack. Set mc_top. - * Makes sure all the pages are writable, before attempting a write operation. - * [in] mc The cursor to operate on. */ -static int mdbx_cursor_touch(MDBX_cursor *mc) { - int rc = MDBX_SUCCESS; - - if (mc->mc_dbi >= CORE_DBS && !(*mc->mc_dbflag & (DB_DIRTY | DB_DUPDATA))) { - /* Touch DB record of named DB */ - MDBX_cursor mc2; - MDBX_xcursor mcx; - if (TXN_DBI_CHANGED(mc->mc_txn, mc->mc_dbi)) - return MDBX_BAD_DBI; - mdbx_cursor_init(&mc2, mc->mc_txn, MAIN_DBI, &mcx); - rc = mdbx_page_search(&mc2, &mc->mc_dbx->md_name, MDBX_PS_MODIFY); - if (unlikely(rc)) - return rc; - *mc->mc_dbflag |= DB_DIRTY; - } - mc->mc_top = 0; - if (mc->mc_snum) { - do { - rc = mdbx_page_touch(mc); - } while (!rc && ++(mc->mc_top) < mc->mc_snum); - mc->mc_top = mc->mc_snum - 1; - } - return rc; -} - -/* Do not spill pages to disk if txn is getting full, may fail instead */ -#define MDBX_NOSPILL 0x8000 - -int mdbx_cursor_put(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data, - unsigned flags) { - MDBX_env *env; - MDBX_page *fp, *sub_root = NULL; - uint16_t fp_flags; - MDBX_val xdata, *rdata, dkey, olddata; - MDBX_db dummy; - unsigned mcount = 0, dcount = 0, nospill; - size_t nsize; - int rc = MDBX_SUCCESS, rc2; - unsigned nflags; - DKBUF; - - if (unlikely(mc == NULL || key == NULL)) - return MDBX_EINVAL; - - if (unlikely(mc->mc_signature != MDBX_MC_SIGNATURE)) - return MDBX_EBADSIGN; - - env = mc->mc_txn->mt_env; - - /* Check this first so counter will always be zero on any early failures. */ - if (flags & MDBX_MULTIPLE) { - if (unlikely(!F_ISSET(mc->mc_db->md_flags, MDBX_DUPFIXED))) - return MDBX_INCOMPATIBLE; - if (unlikely(data[1].iov_len >= INT_MAX)) - return MDBX_EINVAL; - dcount = (unsigned)data[1].iov_len; - data[1].iov_len = 0; - } - - if (flags & MDBX_RESERVE) { - if (unlikely(mc->mc_db->md_flags & (MDBX_DUPSORT | MDBX_REVERSEDUP))) - return MDBX_INCOMPATIBLE; - data->iov_base = nullptr; - } - - nospill = flags & MDBX_NOSPILL; - flags &= ~MDBX_NOSPILL; - - if (unlikely(mc->mc_txn->mt_flags & (MDBX_TXN_RDONLY | MDBX_TXN_BLOCKED))) - return (mc->mc_txn->mt_flags & MDBX_TXN_RDONLY) ? MDBX_EACCESS - : MDBX_BAD_TXN; - - if (unlikely(key->iov_len > env->me_maxkey_limit)) - return MDBX_BAD_VALSIZE; - - if (unlikely(data->iov_len > ((mc->mc_db->md_flags & MDBX_DUPSORT) - ? env->me_maxkey_limit - : MDBX_MAXDATASIZE))) - return MDBX_BAD_VALSIZE; - - if ((mc->mc_db->md_flags & MDBX_INTEGERKEY) && - unlikely(key->iov_len != sizeof(uint32_t) && - key->iov_len != sizeof(uint64_t))) { - mdbx_cassert(mc, !"key-size is invalid for MDBX_INTEGERKEY"); - return MDBX_BAD_VALSIZE; - } - - if ((mc->mc_db->md_flags & MDBX_INTEGERDUP) && - unlikely(data->iov_len != sizeof(uint32_t) && - data->iov_len != sizeof(uint64_t))) { - mdbx_cassert(mc, !"data-size is invalid MDBX_INTEGERDUP"); - return MDBX_BAD_VALSIZE; - } - - mdbx_debug("==> put db %d key [%s], size %" PRIuPTR - ", data [%s] size %" PRIuPTR, - DDBI(mc), DKEY(key), key ? key->iov_len : 0, - DVAL((flags & MDBX_RESERVE) ? nullptr : data), data->iov_len); - - int dupdata_flag = 0; - if (flags & MDBX_CURRENT) { - /* Опция MDBX_CURRENT означает, что запрошено обновление текущей записи, - * на которой сейчас стоит курсор. Проверяем что переданный ключ совпадает - * со значением в текущей позиции курсора. - * Здесь проще вызвать mdbx_cursor_get(), так как для обслуживания таблиц - * с MDBX_DUPSORT также требуется текущий размер данных. */ - MDBX_val current_key, current_data; - rc = mdbx_cursor_get(mc, ¤t_key, ¤t_data, MDBX_GET_CURRENT); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - if (mc->mc_dbx->md_cmp(key, ¤t_key) != 0) - return MDBX_EKEYMISMATCH; - - if (F_ISSET(mc->mc_db->md_flags, MDBX_DUPSORT)) { - MDBX_node *leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_cassert(mc, - mc->mc_xcursor != NULL && - (mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED)); - /* Если за ключом более одного значения, либо если размер данных - * отличается, то вместо inplace обновления требуется удаление и - * последующая вставка. */ - if (mc->mc_xcursor->mx_db.md_entries > 1 || - current_data.iov_len != data->iov_len) { - rc = mdbx_cursor_del(mc, 0); - if (rc != MDBX_SUCCESS) - return rc; - flags -= MDBX_CURRENT; - } - } - } - } - - if (mc->mc_db->md_root == P_INVALID) { - /* new database, cursor has nothing to point to */ - mc->mc_snum = 0; - mc->mc_top = 0; - mc->mc_flags &= ~C_INITIALIZED; - rc = MDBX_NO_ROOT; - } else if ((flags & MDBX_CURRENT) == 0) { - int exact = 0; - MDBX_val d2; - if (flags & MDBX_APPEND) { - MDBX_val k2; - rc = mdbx_cursor_last(mc, &k2, &d2); - if (rc == 0) { - rc = mc->mc_dbx->md_cmp(key, &k2); - if (rc > 0) { - rc = MDBX_NOTFOUND; - mc->mc_ki[mc->mc_top]++; - } else { - /* new key is <= last key */ - rc = MDBX_EKEYMISMATCH; - } - } - } else { - rc = mdbx_cursor_set(mc, key, &d2, MDBX_SET, &exact); - } - if ((flags & MDBX_NOOVERWRITE) && rc == 0) { - mdbx_debug("duplicate key [%s]", DKEY(key)); - *data = d2; - return MDBX_KEYEXIST; - } - if (rc && unlikely(rc != MDBX_NOTFOUND)) - return rc; - } - - mc->mc_flags &= ~C_DEL; - - /* Cursor is positioned, check for room in the dirty list */ - if (!nospill) { - if (flags & MDBX_MULTIPLE) { - rdata = &xdata; - xdata.iov_len = data->iov_len * dcount; - } else { - rdata = data; - } - if (unlikely(rc2 = mdbx_page_spill(mc, key, rdata))) - return rc2; - } - - if (rc == MDBX_NO_ROOT) { - MDBX_page *np; - /* new database, write a root leaf page */ - mdbx_debug("allocating new root leaf page"); - if (unlikely(rc2 = mdbx_page_new(mc, P_LEAF, 1, &np))) { - return rc2; - } - assert(np->mp_flags & P_LEAF); - mdbx_cursor_push(mc, np); - mc->mc_db->md_root = np->mp_pgno; - mc->mc_db->md_depth++; - *mc->mc_dbflag |= DB_DIRTY; - if ((mc->mc_db->md_flags & (MDBX_DUPSORT | MDBX_DUPFIXED)) == MDBX_DUPFIXED) - np->mp_flags |= P_LEAF2; - mc->mc_flags |= C_INITIALIZED; - } else { - /* make sure all cursor pages are writable */ - rc2 = mdbx_cursor_touch(mc); - if (unlikely(rc2)) - return rc2; - } - - bool insert_key, insert_data, do_sub = false; - insert_key = insert_data = (rc != MDBX_SUCCESS); - if (insert_key) { - /* The key does not exist */ - mdbx_debug("inserting key at index %i", mc->mc_ki[mc->mc_top]); - if ((mc->mc_db->md_flags & MDBX_DUPSORT) && - LEAFSIZE(key, data) > env->me_nodemax) { - /* Too big for a node, insert in sub-DB. Set up an empty - * "old sub-page" for prep_subDB to expand to a full page. */ - fp_flags = P_LEAF | P_DIRTY; - fp = env->me_pbuf; - fp->mp_leaf2_ksize = (uint16_t)data->iov_len; /* used if MDBX_DUPFIXED */ - fp->mp_lower = fp->mp_upper = 0; - olddata.iov_len = PAGEHDRSZ; - goto prep_subDB; - } - } else { - /* there's only a key anyway, so this is a no-op */ - if (IS_LEAF2(mc->mc_pg[mc->mc_top])) { - char *ptr; - unsigned ksize = mc->mc_db->md_xsize; - if (key->iov_len != ksize) - return MDBX_BAD_VALSIZE; - ptr = LEAF2KEY(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], ksize); - memcpy(ptr, key->iov_base, ksize); - fix_parent: - /* if overwriting slot 0 of leaf, need to - * update branch key if there is a parent page */ - if (mc->mc_top && !mc->mc_ki[mc->mc_top]) { - unsigned dtop = 1; - mc->mc_top--; - /* slot 0 is always an empty key, find real slot */ - while (mc->mc_top && !mc->mc_ki[mc->mc_top]) { - mc->mc_top--; - dtop++; - } - if (mc->mc_ki[mc->mc_top]) - rc2 = mdbx_update_key(mc, key); - else - rc2 = MDBX_SUCCESS; - mdbx_cassert(mc, mc->mc_top + dtop < UINT16_MAX); - mc->mc_top += (uint16_t)dtop; - if (rc2) - return rc2; - } - return MDBX_SUCCESS; - } - - more:; - MDBX_node *leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); - olddata.iov_len = NODEDSZ(leaf); - olddata.iov_base = NODEDATA(leaf); - - /* DB has dups? */ - if (F_ISSET(mc->mc_db->md_flags, MDBX_DUPSORT)) { - /* Prepare (sub-)page/sub-DB to accept the new item, if needed. - * fp: old sub-page or a header faking it. - * mp: new (sub-)page. offset: growth in page size. - * xdata: node data with new page or DB. */ - unsigned i, offset = 0; - MDBX_page *mp = fp = xdata.iov_base = env->me_pbuf; - mp->mp_pgno = mc->mc_pg[mc->mc_top]->mp_pgno; - - /* Was a single item before, must convert now */ - if (!F_ISSET(leaf->mn_flags, F_DUPDATA)) { - - /* does data match? */ - if (!mc->mc_dbx->md_dcmp(data, &olddata)) { - if (unlikely(flags & (MDBX_NODUPDATA | MDBX_APPENDDUP))) - return MDBX_KEYEXIST; - /* overwrite it */ - goto current; - } - - /* Just overwrite the current item */ - if (flags & MDBX_CURRENT) - goto current; - - /* Back up original data item */ - dupdata_flag = 1; - dkey.iov_len = olddata.iov_len; - dkey.iov_base = memcpy(fp + 1, olddata.iov_base, olddata.iov_len); - - /* Make sub-page header for the dup items, with dummy body */ - fp->mp_flags = P_LEAF | P_DIRTY | P_SUBP; - fp->mp_lower = 0; - xdata.iov_len = PAGEHDRSZ + dkey.iov_len + data->iov_len; - if (mc->mc_db->md_flags & MDBX_DUPFIXED) { - fp->mp_flags |= P_LEAF2; - fp->mp_leaf2_ksize = (uint16_t)data->iov_len; - xdata.iov_len += 2 * data->iov_len; /* leave space for 2 more */ - } else { - xdata.iov_len += 2 * (sizeof(indx_t) + NODESIZE) + - (dkey.iov_len & 1) + (data->iov_len & 1); - } - fp->mp_upper = (uint16_t)(xdata.iov_len - PAGEHDRSZ); - olddata.iov_len = xdata.iov_len; /* pretend olddata is fp */ - } else if (leaf->mn_flags & F_SUBDATA) { - /* Data is on sub-DB, just store it */ - flags |= F_DUPDATA | F_SUBDATA; - goto put_sub; - } else { - /* Data is on sub-page */ - fp = olddata.iov_base; - switch (flags) { - default: - if (!(mc->mc_db->md_flags & MDBX_DUPFIXED)) { - offset = EVEN(NODESIZE + sizeof(indx_t) + data->iov_len); - break; - } - offset = fp->mp_leaf2_ksize; - if (SIZELEFT(fp) < offset) { - offset *= 4; /* space for 4 more */ - break; - } - /* FALLTHRU: Big enough MDBX_DUPFIXaED sub-page */ - __fallthrough; - case MDBX_CURRENT | MDBX_NODUPDATA: - case MDBX_CURRENT: - fp->mp_flags |= P_DIRTY; - fp->mp_pgno = mp->mp_pgno; - mc->mc_xcursor->mx_cursor.mc_pg[0] = fp; - flags |= F_DUPDATA; - goto put_sub; - } - xdata.iov_len = olddata.iov_len + offset; - } - - fp_flags = fp->mp_flags; - if (NODESIZE + NODEKSZ(leaf) + xdata.iov_len > env->me_nodemax) { - /* Too big for a sub-page, convert to sub-DB */ - fp_flags &= ~P_SUBP; - prep_subDB: - dummy.md_xsize = 0; - dummy.md_flags = 0; - if (mc->mc_db->md_flags & MDBX_DUPFIXED) { - fp_flags |= P_LEAF2; - dummy.md_xsize = fp->mp_leaf2_ksize; - dummy.md_flags = MDBX_DUPFIXED; - if (mc->mc_db->md_flags & MDBX_INTEGERDUP) - dummy.md_flags |= MDBX_INTEGERKEY; - } - dummy.md_depth = 1; - dummy.md_branch_pages = 0; - dummy.md_leaf_pages = 1; - dummy.md_overflow_pages = 0; - dummy.md_entries = NUMKEYS(fp); - xdata.iov_len = sizeof(MDBX_db); - xdata.iov_base = &dummy; - if ((rc = mdbx_page_alloc(mc, 1, &mp, MDBX_ALLOC_ALL))) - return rc; - mdbx_cassert(mc, env->me_psize > olddata.iov_len); - offset = env->me_psize - (unsigned)olddata.iov_len; - flags |= F_DUPDATA | F_SUBDATA; - dummy.md_root = mp->mp_pgno; - sub_root = mp; - } - if (mp != fp) { - mp->mp_flags = fp_flags | P_DIRTY; - mp->mp_leaf2_ksize = fp->mp_leaf2_ksize; - mp->mp_lower = fp->mp_lower; - mdbx_cassert(mc, fp->mp_upper + offset <= UINT16_MAX); - mp->mp_upper = (indx_t)(fp->mp_upper + offset); - if (fp_flags & P_LEAF2) { - memcpy(PAGEDATA(mp), PAGEDATA(fp), NUMKEYS(fp) * fp->mp_leaf2_ksize); - } else { - memcpy((char *)mp + mp->mp_upper + PAGEHDRSZ, - (char *)fp + fp->mp_upper + PAGEHDRSZ, - olddata.iov_len - fp->mp_upper - PAGEHDRSZ); - for (i = 0; i < NUMKEYS(fp); i++) { - mdbx_cassert(mc, fp->mp_ptrs[i] + offset <= UINT16_MAX); - mp->mp_ptrs[i] = (indx_t)(fp->mp_ptrs[i] + offset); - } - } - } - - rdata = &xdata; - flags |= F_DUPDATA; - do_sub = true; - if (!insert_key) - mdbx_node_del(mc, 0); - goto new_sub; - } - current: - /* MDBX passes F_SUBDATA in 'flags' to write a DB record */ - if (unlikely((leaf->mn_flags ^ flags) & F_SUBDATA)) - return MDBX_INCOMPATIBLE; - /* overflow page overwrites need special handling */ - if (F_ISSET(leaf->mn_flags, F_BIGDATA)) { - MDBX_page *omp; - pgno_t pg; - int level, ovpages, dpages = OVPAGES(env, data->iov_len); - - memcpy(&pg, olddata.iov_base, sizeof(pg)); - if (unlikely((rc2 = mdbx_page_get(mc, pg, &omp, &level)) != 0)) - return rc2; - ovpages = omp->mp_pages; - - /* Is the ov page large enough? */ - if (ovpages >= dpages) { - if (!(omp->mp_flags & P_DIRTY) && - (level || (env->me_flags & MDBX_WRITEMAP))) { - rc = mdbx_page_unspill(mc->mc_txn, omp, &omp); - if (unlikely(rc)) - return rc; - level = 0; /* dirty in this txn or clean */ - } - /* Is it dirty? */ - if (omp->mp_flags & P_DIRTY) { - /* yes, overwrite it. Note in this case we don't - * bother to try shrinking the page if the new data - * is smaller than the overflow threshold. */ - if (unlikely(level > 1)) { - /* It is writable only in a parent txn */ - MDBX_page *np = mdbx_page_malloc(mc->mc_txn, ovpages); - MDBX_ID2 id2; - if (unlikely(!np)) - return MDBX_ENOMEM; - id2.mid = pg; - id2.mptr = np; - /* Note - this page is already counted in parent's dirtyroom */ - rc2 = mdbx_mid2l_insert(mc->mc_txn->mt_rw_dirtylist, &id2); - mdbx_cassert(mc, rc2 == 0); - - /* Currently we make the page look as with put() in the - * parent txn, in case the user peeks at MDBX_RESERVEd - * or unused parts. Some users treat ovpages specially. */ - const size_t whole = pgno2bytes(env, ovpages); - /* Skip the part where MDBX will put *data. - * Copy end of page, adjusting alignment so - * compiler may copy words instead of bytes. */ - const size_t off = - (PAGEHDRSZ + data->iov_len) & -(intptr_t)sizeof(size_t); - memcpy((size_t *)((char *)np + off), (size_t *)((char *)omp + off), - whole - off); - memcpy(np, omp, PAGEHDRSZ); /* Copy header of page */ - omp = np; - } - SETDSZ(leaf, data->iov_len); - if (F_ISSET(flags, MDBX_RESERVE)) - data->iov_base = PAGEDATA(omp); - else - memcpy(PAGEDATA(omp), data->iov_base, data->iov_len); - return MDBX_SUCCESS; - } - } - if ((rc2 = mdbx_ovpage_free(mc, omp)) != MDBX_SUCCESS) - return rc2; - } else if (data->iov_len == olddata.iov_len) { - assert(EVEN(key->iov_len) == EVEN(leaf->mn_ksize)); - /* same size, just replace it. Note that we could - * also reuse this node if the new data is smaller, - * but instead we opt to shrink the node in that case. */ - if (F_ISSET(flags, MDBX_RESERVE)) - data->iov_base = olddata.iov_base; - else if (!(mc->mc_flags & C_SUB)) - memcpy(olddata.iov_base, data->iov_base, data->iov_len); - else { - mdbx_cassert(mc, NUMKEYS(mc->mc_pg[mc->mc_top]) == 1); - mdbx_cassert(mc, mc->mc_pg[mc->mc_top]->mp_upper == - mc->mc_pg[mc->mc_top]->mp_lower); - mdbx_cassert(mc, IS_LEAF(mc->mc_pg[mc->mc_top]) && - !IS_LEAF2(mc->mc_pg[mc->mc_top])); - mdbx_cassert(mc, NODEDSZ(leaf) == 0); - mdbx_cassert(mc, leaf->mn_flags == 0); - mdbx_cassert(mc, key->iov_len < UINT16_MAX); - leaf->mn_ksize = (uint16_t)key->iov_len; - memcpy(NODEKEY(leaf), key->iov_base, key->iov_len); - assert((char *)NODEDATA(leaf) + NODEDSZ(leaf) < - (char *)(mc->mc_pg[mc->mc_top]) + env->me_psize); - goto fix_parent; - } - return MDBX_SUCCESS; - } - mdbx_node_del(mc, 0); - } - - rdata = data; - -new_sub: - nflags = flags & NODE_ADD_FLAGS; - nsize = IS_LEAF2(mc->mc_pg[mc->mc_top]) ? key->iov_len - : mdbx_leaf_size(env, key, rdata); - if (SIZELEFT(mc->mc_pg[mc->mc_top]) < nsize) { - if ((flags & (F_DUPDATA | F_SUBDATA)) == F_DUPDATA) - nflags &= ~MDBX_APPEND; /* sub-page may need room to grow */ - if (!insert_key) - nflags |= MDBX_SPLIT_REPLACE; - rc = mdbx_page_split(mc, key, rdata, P_INVALID, nflags); - } else { - /* There is room already in this leaf page. */ - rc = mdbx_node_add(mc, mc->mc_ki[mc->mc_top], key, rdata, 0, nflags); - if (likely(rc == 0)) { - /* Adjust other cursors pointing to mp */ - MDBX_cursor *m2, *m3; - MDBX_dbi dbi = mc->mc_dbi; - unsigned i = mc->mc_top; - MDBX_page *mp = mc->mc_pg[i]; - - for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) { - if (mc->mc_flags & C_SUB) - m3 = &m2->mc_xcursor->mx_cursor; - else - m3 = m2; - if (m3 == mc || m3->mc_snum < mc->mc_snum || m3->mc_pg[i] != mp) - continue; - if (m3->mc_ki[i] >= mc->mc_ki[i] && insert_key) { - m3->mc_ki[i]++; - } - if (XCURSOR_INITED(m3)) - XCURSOR_REFRESH(m3, mp, m3->mc_ki[i]); - } - } - } - - if (likely(rc == MDBX_SUCCESS)) { - /* Now store the actual data in the child DB. Note that we're - * storing the user data in the keys field, so there are strict - * size limits on dupdata. The actual data fields of the child - * DB are all zero size. */ - if (do_sub) { - int xflags; - size_t ecount; - put_sub: - xdata.iov_len = 0; - xdata.iov_base = ""; - MDBX_node *leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); - if (flags & MDBX_CURRENT) { - xflags = (flags & MDBX_NODUPDATA) - ? MDBX_CURRENT | MDBX_NOOVERWRITE | MDBX_NOSPILL - : MDBX_CURRENT | MDBX_NOSPILL; - } else { - mdbx_xcursor_init1(mc, leaf); - xflags = (flags & MDBX_NODUPDATA) ? MDBX_NOOVERWRITE | MDBX_NOSPILL - : MDBX_NOSPILL; - } - if (sub_root) - mc->mc_xcursor->mx_cursor.mc_pg[0] = sub_root; - /* converted, write the original data first */ - if (dupdata_flag) { - rc = mdbx_cursor_put(&mc->mc_xcursor->mx_cursor, &dkey, &xdata, xflags); - if (unlikely(rc)) - goto bad_sub; - /* we've done our job */ - dkey.iov_len = 0; - } - if (!(leaf->mn_flags & F_SUBDATA) || sub_root) { - /* Adjust other cursors pointing to mp */ - MDBX_cursor *m2; - MDBX_xcursor *mx = mc->mc_xcursor; - unsigned i = mc->mc_top; - MDBX_page *mp = mc->mc_pg[i]; - int nkeys = NUMKEYS(mp); - - for (m2 = mc->mc_txn->mt_cursors[mc->mc_dbi]; m2; m2 = m2->mc_next) { - if (m2 == mc || m2->mc_snum < mc->mc_snum) - continue; - if (!(m2->mc_flags & C_INITIALIZED)) - continue; - if (m2->mc_pg[i] == mp) { - if (m2->mc_ki[i] == mc->mc_ki[i]) { - mdbx_xcursor_init2(m2, mx, dupdata_flag); - } else if (!insert_key && m2->mc_ki[i] < nkeys) { - XCURSOR_REFRESH(m2, mp, m2->mc_ki[i]); - } - } - } - } - mdbx_cassert(mc, mc->mc_xcursor->mx_db.md_entries < SIZE_MAX); - ecount = (size_t)mc->mc_xcursor->mx_db.md_entries; - if (flags & MDBX_APPENDDUP) - xflags |= MDBX_APPEND; - rc = mdbx_cursor_put(&mc->mc_xcursor->mx_cursor, data, &xdata, xflags); - if (flags & F_SUBDATA) { - void *db = NODEDATA(leaf); - memcpy(db, &mc->mc_xcursor->mx_db, sizeof(MDBX_db)); - } - insert_data = (ecount != (size_t)mc->mc_xcursor->mx_db.md_entries); - } - /* Increment count unless we just replaced an existing item. */ - if (insert_data) - mc->mc_db->md_entries++; - if (insert_key) { - /* Invalidate txn if we created an empty sub-DB */ - if (unlikely(rc)) - goto bad_sub; - /* If we succeeded and the key didn't exist before, - * make sure the cursor is marked valid. */ - mc->mc_flags |= C_INITIALIZED; - } - if (flags & MDBX_MULTIPLE) { - if (!rc) { - mcount++; - /* let caller know how many succeeded, if any */ - data[1].iov_len = mcount; - if (mcount < dcount) { - data[0].iov_base = (char *)data[0].iov_base + data[0].iov_len; - insert_key = insert_data = false; - goto more; - } - } - } - return rc; - bad_sub: - if (unlikely(rc == MDBX_KEYEXIST)) - mdbx_error("unexpected %s", "MDBX_KEYEXIST"); - /* should not happen, we deleted that item */ - rc = MDBX_PROBLEM; - } - mc->mc_txn->mt_flags |= MDBX_TXN_ERROR; - return rc; -} - -int mdbx_cursor_del(MDBX_cursor *mc, unsigned flags) { - MDBX_node *leaf; - MDBX_page *mp; - int rc; - - if (unlikely(!mc)) - return MDBX_EINVAL; - - if (unlikely(mc->mc_signature != MDBX_MC_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(mc->mc_txn->mt_flags & (MDBX_TXN_RDONLY | MDBX_TXN_BLOCKED))) - return (mc->mc_txn->mt_flags & MDBX_TXN_RDONLY) ? MDBX_EACCESS - : MDBX_BAD_TXN; - - if (unlikely(!(mc->mc_flags & C_INITIALIZED))) - return MDBX_EINVAL; - - if (unlikely(mc->mc_ki[mc->mc_top] >= NUMKEYS(mc->mc_pg[mc->mc_top]))) - return MDBX_NOTFOUND; - - if (unlikely(!(flags & MDBX_NOSPILL) && - (rc = mdbx_page_spill(mc, NULL, NULL)))) - return rc; - - rc = mdbx_cursor_touch(mc); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - mp = mc->mc_pg[mc->mc_top]; - if (IS_LEAF2(mp)) - goto del_key; - leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); - - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - if (flags & MDBX_NODUPDATA) { - /* mdbx_cursor_del0() will subtract the final entry */ - mc->mc_db->md_entries -= mc->mc_xcursor->mx_db.md_entries - 1; - mc->mc_xcursor->mx_cursor.mc_flags &= ~C_INITIALIZED; - } else { - if (!F_ISSET(leaf->mn_flags, F_SUBDATA)) { - mc->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(leaf); - } - rc = mdbx_cursor_del(&mc->mc_xcursor->mx_cursor, MDBX_NOSPILL); - if (unlikely(rc)) - return rc; - /* If sub-DB still has entries, we're done */ - if (mc->mc_xcursor->mx_db.md_entries) { - if (leaf->mn_flags & F_SUBDATA) { - /* update subDB info */ - void *db = NODEDATA(leaf); - memcpy(db, &mc->mc_xcursor->mx_db, sizeof(MDBX_db)); - } else { - MDBX_cursor *m2; - /* shrink fake page */ - mdbx_node_shrink(mp, mc->mc_ki[mc->mc_top]); - leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); - mc->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(leaf); - /* fix other sub-DB cursors pointed at fake pages on this page */ - for (m2 = mc->mc_txn->mt_cursors[mc->mc_dbi]; m2; m2 = m2->mc_next) { - if (m2 == mc || m2->mc_snum < mc->mc_snum) - continue; - if (!(m2->mc_flags & C_INITIALIZED)) - continue; - if (m2->mc_pg[mc->mc_top] == mp) { - MDBX_node *n2 = leaf; - if (m2->mc_ki[mc->mc_top] >= NUMKEYS(mp)) - continue; - if (m2->mc_ki[mc->mc_top] != mc->mc_ki[mc->mc_top]) { - n2 = NODEPTR(mp, m2->mc_ki[mc->mc_top]); - if (n2->mn_flags & F_SUBDATA) - continue; - } - m2->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(n2); - } - } - } - mc->mc_db->md_entries--; - return rc; - } else { - mc->mc_xcursor->mx_cursor.mc_flags &= ~C_INITIALIZED; - } - /* otherwise fall thru and delete the sub-DB */ - } - - if (leaf->mn_flags & F_SUBDATA) { - /* add all the child DB's pages to the free list */ - rc = mdbx_drop0(&mc->mc_xcursor->mx_cursor, 0); - if (unlikely(rc)) - goto fail; - } - } - /* MDBX passes F_SUBDATA in 'flags' to delete a DB record */ - else if (unlikely((leaf->mn_flags ^ flags) & F_SUBDATA)) { - rc = MDBX_INCOMPATIBLE; - goto fail; - } - - /* add overflow pages to free list */ - if (F_ISSET(leaf->mn_flags, F_BIGDATA)) { - MDBX_page *omp; - pgno_t pg; - - memcpy(&pg, NODEDATA(leaf), sizeof(pg)); - if (unlikely((rc = mdbx_page_get(mc, pg, &omp, NULL)) || - (rc = mdbx_ovpage_free(mc, omp)))) - goto fail; - } - -del_key: - return mdbx_cursor_del0(mc); - -fail: - mc->mc_txn->mt_flags |= MDBX_TXN_ERROR; - return rc; -} - -/* Allocate and initialize new pages for a database. - * Set MDBX_TXN_ERROR on failure. - * - * [in] mc a cursor on the database being added to. - * [in] flags flags defining what type of page is being allocated. - * [in] num the number of pages to allocate. This is usually 1, - * unless allocating overflow pages for a large record. - * [out] mp Address of a page, or NULL on failure. - * - * Returns 0 on success, non-zero on failure. */ -static int mdbx_page_new(MDBX_cursor *mc, unsigned flags, unsigned num, - MDBX_page **mp) { - MDBX_page *np; - int rc; - - if (unlikely((rc = mdbx_page_alloc(mc, num, &np, MDBX_ALLOC_ALL)))) - return rc; - mdbx_debug("allocated new page #%" PRIaPGNO ", size %u", np->mp_pgno, - mc->mc_txn->mt_env->me_psize); - np->mp_flags = (uint16_t)(flags | P_DIRTY); - np->mp_lower = 0; - np->mp_upper = (indx_t)(mc->mc_txn->mt_env->me_psize - PAGEHDRSZ); - - if (IS_BRANCH(np)) - mc->mc_db->md_branch_pages++; - else if (IS_LEAF(np)) - mc->mc_db->md_leaf_pages++; - else if (IS_OVERFLOW(np)) { - mc->mc_db->md_overflow_pages += num; - np->mp_pages = num; - } - *mp = np; - - return MDBX_SUCCESS; -} - -/* Calculate the size of a leaf node. - * - * The size depends on the environment's page size; if a data item - * is too large it will be put onto an overflow page and the node - * size will only include the key and not the data. Sizes are always - * rounded up to an even number of bytes, to guarantee 2-byte alignment - * of the MDBX_node headers. - * - * [in] env The environment handle. - * [in] key The key for the node. - * [in] data The data for the node. - * - * Returns The number of bytes needed to store the node. */ -static __inline size_t mdbx_leaf_size(MDBX_env *env, MDBX_val *key, - MDBX_val *data) { - size_t sz; - - sz = LEAFSIZE(key, data); - if (sz > env->me_nodemax) { - /* put on overflow page */ - sz -= data->iov_len - sizeof(pgno_t); - } - - return EVEN(sz + sizeof(indx_t)); -} - -/* Calculate the size of a branch node. - * - * The size should depend on the environment's page size but since - * we currently don't support spilling large keys onto overflow - * pages, it's simply the size of the MDBX_node header plus the - * size of the key. Sizes are always rounded up to an even number - * of bytes, to guarantee 2-byte alignment of the MDBX_node headers. - * - * [in] env The environment handle. - * [in] key The key for the node. - * - * Returns The number of bytes needed to store the node. */ -static __inline size_t mdbx_branch_size(MDBX_env *env, MDBX_val *key) { - size_t sz; - - sz = INDXSIZE(key); - if (unlikely(sz > env->me_nodemax)) { - /* put on overflow page */ - /* not implemented */ - mdbx_assert_fail(env, "INDXSIZE(key) <= env->me_nodemax", __FUNCTION__, - __LINE__); - sz -= key->iov_len - sizeof(pgno_t); - } - - return sz + sizeof(indx_t); -} - -/* Add a node to the page pointed to by the cursor. - * Set MDBX_TXN_ERROR on failure. - * - * [in] mc The cursor for this operation. - * [in] indx The index on the page where the new node should be added. - * [in] key The key for the new node. - * [in] data The data for the new node, if any. - * [in] pgno The page number, if adding a branch node. - * [in] flags Flags for the node. - * - * Returns 0 on success, non-zero on failure. Possible errors are: - * - * MDBX_ENOMEM - failed to allocate overflow pages for the node. - * MDBX_PAGE_FULL - there is insufficient room in the page. This error - * should never happen since all callers already calculate - * the page's free space before calling this function. */ -static int mdbx_node_add(MDBX_cursor *mc, unsigned indx, MDBX_val *key, - MDBX_val *data, pgno_t pgno, unsigned flags) { - unsigned i; - size_t node_size = NODESIZE; - intptr_t room; - MDBX_node *node; - MDBX_page *mp = mc->mc_pg[mc->mc_top]; - MDBX_page *ofp = NULL; /* overflow page */ - void *ndata; - DKBUF; - - mdbx_cassert(mc, mp->mp_upper >= mp->mp_lower); - - mdbx_debug("add to %s %spage %" PRIaPGNO " index %i, data size %" PRIuPTR - " key size %" PRIuPTR " [%s]", - IS_LEAF(mp) ? "leaf" : "branch", IS_SUBP(mp) ? "sub-" : "", - mp->mp_pgno, indx, data ? data->iov_len : 0, - key ? key->iov_len : 0, DKEY(key)); - - if (IS_LEAF2(mp)) { - mdbx_cassert(mc, key); - /* Move higher keys up one slot. */ - const int ksize = mc->mc_db->md_xsize; - char *const ptr = LEAF2KEY(mp, indx, ksize); - const int diff = NUMKEYS(mp) - indx; - if (diff > 0) - memmove(ptr + ksize, ptr, diff * ksize); - /* insert new key */ - memcpy(ptr, key->iov_base, ksize); - - /* Just using these for counting */ - mdbx_cassert(mc, UINT16_MAX - mp->mp_lower >= (int)sizeof(indx_t)); - mp->mp_lower += sizeof(indx_t); - mdbx_cassert(mc, mp->mp_upper >= ksize - sizeof(indx_t)); - mp->mp_upper -= (indx_t)(ksize - sizeof(indx_t)); - return MDBX_SUCCESS; - } - - room = (intptr_t)SIZELEFT(mp) - (intptr_t)sizeof(indx_t); - if (key != NULL) - node_size += key->iov_len; - if (IS_LEAF(mp)) { - mdbx_cassert(mc, key && data); - if (unlikely(F_ISSET(flags, F_BIGDATA))) { - /* Data already on overflow page. */ - node_size += sizeof(pgno_t); - } else if (unlikely(node_size + data->iov_len > - mc->mc_txn->mt_env->me_nodemax)) { - pgno_t ovpages = OVPAGES(mc->mc_txn->mt_env, data->iov_len); - int rc; - /* Put data on overflow page. */ - mdbx_debug("data size is %" PRIuPTR ", node would be %" PRIuPTR - ", put data on overflow page", - data->iov_len, node_size + data->iov_len); - node_size = EVEN(node_size + sizeof(pgno_t)); - if ((intptr_t)node_size > room) - goto full; - if ((rc = mdbx_page_new(mc, P_OVERFLOW, ovpages, &ofp))) - return rc; - mdbx_debug("allocated overflow page %" PRIaPGNO "", ofp->mp_pgno); - flags |= F_BIGDATA; - goto update; - } else { - node_size += data->iov_len; - } - } - node_size = EVEN(node_size); - if (unlikely((intptr_t)node_size > room)) - goto full; - -update: - /* Move higher pointers up one slot. */ - for (i = NUMKEYS(mp); i > indx; i--) - mp->mp_ptrs[i] = mp->mp_ptrs[i - 1]; - - /* Adjust free space offsets. */ - size_t ofs = mp->mp_upper - node_size; - mdbx_cassert(mc, ofs >= mp->mp_lower + sizeof(indx_t)); - mdbx_cassert(mc, ofs <= UINT16_MAX); - mp->mp_ptrs[indx] = (uint16_t)ofs; - mp->mp_upper = (uint16_t)ofs; - mp->mp_lower += sizeof(indx_t); - - /* Write the node data. */ - node = NODEPTR(mp, indx); - node->mn_ksize = (key == NULL) ? 0 : (uint16_t)key->iov_len; - node->mn_flags = (uint16_t)flags; - if (IS_LEAF(mp)) - SETDSZ(node, data->iov_len); - else - SETPGNO(node, pgno); - - if (key) - memcpy(NODEKEY(node), key->iov_base, key->iov_len); - - if (IS_LEAF(mp)) { - ndata = NODEDATA(node); - if (unlikely(ofp == NULL)) { - if (unlikely(F_ISSET(flags, F_BIGDATA))) - memcpy(ndata, data->iov_base, sizeof(pgno_t)); - else if (F_ISSET(flags, MDBX_RESERVE)) - data->iov_base = ndata; - else if (likely(ndata != data->iov_base)) - memcpy(ndata, data->iov_base, data->iov_len); - } else { - memcpy(ndata, &ofp->mp_pgno, sizeof(pgno_t)); - ndata = PAGEDATA(ofp); - if (F_ISSET(flags, MDBX_RESERVE)) - data->iov_base = ndata; - else if (likely(ndata != data->iov_base)) - memcpy(ndata, data->iov_base, data->iov_len); - } - } - - return MDBX_SUCCESS; - -full: - mdbx_debug("not enough room in page %" PRIaPGNO ", got %u ptrs", mp->mp_pgno, - NUMKEYS(mp)); - mdbx_debug("upper-lower = %u - %u = %" PRIiPTR "", mp->mp_upper, mp->mp_lower, - room); - mdbx_debug("node size = %" PRIuPTR "", node_size); - mc->mc_txn->mt_flags |= MDBX_TXN_ERROR; - return MDBX_PAGE_FULL; -} - -/* Delete the specified node from a page. - * [in] mc Cursor pointing to the node to delete. - * [in] ksize The size of a node. Only used if the page is - * part of a MDBX_DUPFIXED database. */ -static void mdbx_node_del(MDBX_cursor *mc, size_t ksize) { - MDBX_page *mp = mc->mc_pg[mc->mc_top]; - indx_t indx = mc->mc_ki[mc->mc_top]; - indx_t i, j, numkeys, ptr; - MDBX_node *node; - char *base; - - mdbx_debug("delete node %u on %s page %" PRIaPGNO "", indx, - IS_LEAF(mp) ? "leaf" : "branch", mp->mp_pgno); - numkeys = NUMKEYS(mp); - mdbx_cassert(mc, indx < numkeys); - - if (IS_LEAF2(mp)) { - mdbx_cassert(mc, ksize >= sizeof(indx_t)); - unsigned diff = numkeys - 1 - indx; - base = LEAF2KEY(mp, indx, ksize); - if (diff) - memmove(base, base + ksize, diff * ksize); - mdbx_cassert(mc, mp->mp_lower >= sizeof(indx_t)); - mp->mp_lower -= sizeof(indx_t); - mdbx_cassert(mc, - (size_t)UINT16_MAX - mp->mp_upper >= ksize - sizeof(indx_t)); - mp->mp_upper += (indx_t)(ksize - sizeof(indx_t)); - return; - } - - node = NODEPTR(mp, indx); - size_t sz = NODESIZE + node->mn_ksize; - if (IS_LEAF(mp)) { - if (F_ISSET(node->mn_flags, F_BIGDATA)) - sz += sizeof(pgno_t); - else - sz += NODEDSZ(node); - } - sz = EVEN(sz); - - ptr = mp->mp_ptrs[indx]; - for (i = j = 0; i < numkeys; i++) { - if (i != indx) { - mp->mp_ptrs[j] = mp->mp_ptrs[i]; - if (mp->mp_ptrs[i] < ptr) { - mdbx_cassert(mc, (size_t)UINT16_MAX - mp->mp_ptrs[j] >= sz); - mp->mp_ptrs[j] += (indx_t)sz; - } - j++; - } - } - - base = (char *)mp + mp->mp_upper + PAGEHDRSZ; - memmove(base + sz, base, ptr - mp->mp_upper); - - mdbx_cassert(mc, mp->mp_lower >= sizeof(indx_t)); - mp->mp_lower -= sizeof(indx_t); - mdbx_cassert(mc, (size_t)UINT16_MAX - mp->mp_upper >= sz); - mp->mp_upper += (indx_t)sz; -} - -/* Compact the main page after deleting a node on a subpage. - * [in] mp The main page to operate on. - * [in] indx The index of the subpage on the main page. */ -static void mdbx_node_shrink(MDBX_page *mp, unsigned indx) { - MDBX_node *node; - MDBX_page *sp, *xp; - char *base; - size_t nsize, delta, len, ptr; - int i; - - node = NODEPTR(mp, indx); - sp = (MDBX_page *)NODEDATA(node); - delta = SIZELEFT(sp); - nsize = NODEDSZ(node) - delta; - - /* Prepare to shift upward, set len = length(subpage part to shift) */ - if (IS_LEAF2(sp)) { - len = nsize; - if (nsize & 1) - return; /* do not make the node uneven-sized */ - } else { - xp = (MDBX_page *)((char *)sp + delta); /* destination subpage */ - for (i = NUMKEYS(sp); --i >= 0;) { - assert(sp->mp_ptrs[i] >= delta); - xp->mp_ptrs[i] = (indx_t)(sp->mp_ptrs[i] - delta); - } - len = PAGEHDRSZ; - } - sp->mp_upper = sp->mp_lower; - sp->mp_pgno = mp->mp_pgno; - SETDSZ(node, nsize); - - /* Shift <lower nodes...initial part of subpage> upward */ - base = (char *)mp + mp->mp_upper + PAGEHDRSZ; - memmove(base + delta, base, (char *)sp + len - base); - - ptr = mp->mp_ptrs[indx]; - for (i = NUMKEYS(mp); --i >= 0;) { - if (mp->mp_ptrs[i] <= ptr) { - assert((size_t)UINT16_MAX - mp->mp_ptrs[i] >= delta); - mp->mp_ptrs[i] += (indx_t)delta; - } - } - assert((size_t)UINT16_MAX - mp->mp_upper >= delta); - mp->mp_upper += (indx_t)delta; -} - -/* Initial setup of a sorted-dups cursor. - * - * Sorted duplicates are implemented as a sub-database for the given key. - * The duplicate data items are actually keys of the sub-database. - * Operations on the duplicate data items are performed using a sub-cursor - * initialized when the sub-database is first accessed. This function does - * the preliminary setup of the sub-cursor, filling in the fields that - * depend only on the parent DB. - * - * [in] mc The main cursor whose sorted-dups cursor is to be initialized. */ -static void mdbx_xcursor_init0(MDBX_cursor *mc) { - MDBX_xcursor *mx = mc->mc_xcursor; - - mx->mx_cursor.mc_xcursor = NULL; - mx->mx_cursor.mc_txn = mc->mc_txn; - mx->mx_cursor.mc_db = &mx->mx_db; - mx->mx_cursor.mc_dbx = &mx->mx_dbx; - mx->mx_cursor.mc_dbi = mc->mc_dbi; - mx->mx_cursor.mc_dbflag = &mx->mx_dbflag; - mx->mx_cursor.mc_snum = 0; - mx->mx_cursor.mc_top = 0; - mx->mx_cursor.mc_flags = C_SUB; - mx->mx_dbx.md_name.iov_len = 0; - mx->mx_dbx.md_name.iov_base = NULL; - mx->mx_dbx.md_cmp = mc->mc_dbx->md_dcmp; - mx->mx_dbx.md_dcmp = NULL; -} - -/* Final setup of a sorted-dups cursor. - * Sets up the fields that depend on the data from the main cursor. - * [in] mc The main cursor whose sorted-dups cursor is to be initialized. - * [in] node The data containing the MDBX_db record for the sorted-dup database. - */ -static void mdbx_xcursor_init1(MDBX_cursor *mc, MDBX_node *node) { - MDBX_xcursor *mx = mc->mc_xcursor; - - mdbx_cassert(mc, mc->mc_txn->mt_txnid >= mc->mc_txn->mt_env->me_oldest[0]); - if (node->mn_flags & F_SUBDATA) { - memcpy(&mx->mx_db, NODEDATA(node), sizeof(MDBX_db)); - mx->mx_cursor.mc_pg[0] = 0; - mx->mx_cursor.mc_snum = 0; - mx->mx_cursor.mc_top = 0; - mx->mx_cursor.mc_flags = C_SUB; - } else { - MDBX_page *fp = NODEDATA(node); - mx->mx_db.md_xsize = 0; - mx->mx_db.md_flags = 0; - mx->mx_db.md_depth = 1; - mx->mx_db.md_branch_pages = 0; - mx->mx_db.md_leaf_pages = 1; - mx->mx_db.md_overflow_pages = 0; - mx->mx_db.md_entries = NUMKEYS(fp); - mx->mx_db.md_root = fp->mp_pgno; - mx->mx_cursor.mc_snum = 1; - mx->mx_cursor.mc_top = 0; - mx->mx_cursor.mc_flags = C_INITIALIZED | C_SUB; - mx->mx_cursor.mc_pg[0] = fp; - mx->mx_cursor.mc_ki[0] = 0; - if (mc->mc_db->md_flags & MDBX_DUPFIXED) { - mx->mx_db.md_flags = MDBX_DUPFIXED; - mx->mx_db.md_xsize = fp->mp_leaf2_ksize; - if (mc->mc_db->md_flags & MDBX_INTEGERDUP) - mx->mx_db.md_flags |= MDBX_INTEGERKEY; - } - } - mdbx_debug("Sub-db -%u root page %" PRIaPGNO "", mx->mx_cursor.mc_dbi, - mx->mx_db.md_root); - mx->mx_dbflag = DB_VALID | DB_USRVALID | DB_DUPDATA; - /* FIXME: #if UINT_MAX < SIZE_MAX - if (mx->mx_dbx.md_cmp == mdbx_cmp_int && mx->mx_db.md_pad == - sizeof(size_t)) - mx->mx_dbx.md_cmp = mdbx_cmp_clong; - #endif */ -} - -/* Fixup a sorted-dups cursor due to underlying update. - * Sets up some fields that depend on the data from the main cursor. - * Almost the same as init1, but skips initialization steps if the - * xcursor had already been used. - * [in] mc The main cursor whose sorted-dups cursor is to be fixed up. - * [in] src_mx The xcursor of an up-to-date cursor. - * [in] new_dupdata True if converting from a non-F_DUPDATA item. */ -static void mdbx_xcursor_init2(MDBX_cursor *mc, MDBX_xcursor *src_mx, - int new_dupdata) { - MDBX_xcursor *mx = mc->mc_xcursor; - - mdbx_cassert(mc, mc->mc_txn->mt_txnid >= mc->mc_txn->mt_env->me_oldest[0]); - if (new_dupdata) { - mx->mx_cursor.mc_snum = 1; - mx->mx_cursor.mc_top = 0; - mx->mx_cursor.mc_flags |= C_INITIALIZED; - mx->mx_cursor.mc_ki[0] = 0; - mx->mx_dbflag = DB_VALID | DB_USRVALID | DB_DUPDATA; - mx->mx_dbx.md_cmp = src_mx->mx_dbx.md_cmp; - } else if (!(mx->mx_cursor.mc_flags & C_INITIALIZED)) { - return; - } - mx->mx_db = src_mx->mx_db; - mx->mx_cursor.mc_pg[0] = src_mx->mx_cursor.mc_pg[0]; - mdbx_debug("Sub-db -%u root page %" PRIaPGNO "", mx->mx_cursor.mc_dbi, - mx->mx_db.md_root); -} - -/* Initialize a cursor for a given transaction and database. */ -static void mdbx_cursor_init(MDBX_cursor *mc, MDBX_txn *txn, MDBX_dbi dbi, - MDBX_xcursor *mx) { - mc->mc_signature = MDBX_MC_SIGNATURE; - mc->mc_next = NULL; - mc->mc_backup = NULL; - mc->mc_dbi = dbi; - mc->mc_txn = txn; - mc->mc_db = &txn->mt_dbs[dbi]; - mc->mc_dbx = &txn->mt_dbxs[dbi]; - mc->mc_dbflag = &txn->mt_dbflags[dbi]; - mc->mc_snum = 0; - mc->mc_top = 0; - mc->mc_pg[0] = 0; - mc->mc_flags = 0; - mc->mc_ki[0] = 0; - mc->mc_xcursor = NULL; - if (txn->mt_dbs[dbi].md_flags & MDBX_DUPSORT) { - mdbx_tassert(txn, mx != NULL); - mx->mx_cursor.mc_signature = MDBX_MC_SIGNATURE; - mc->mc_xcursor = mx; - mdbx_xcursor_init0(mc); - } - mdbx_cassert(mc, mc->mc_txn->mt_txnid >= mc->mc_txn->mt_env->me_oldest[0]); - if (unlikely(*mc->mc_dbflag & DB_STALE)) { - mdbx_page_search(mc, NULL, MDBX_PS_ROOTONLY); - } -} - -int mdbx_cursor_open(MDBX_txn *txn, MDBX_dbi dbi, MDBX_cursor **ret) { - MDBX_cursor *mc; - size_t size = sizeof(MDBX_cursor); - - if (unlikely(!ret || !txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_VALID))) - return MDBX_EINVAL; - - if (unlikely(txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - if (unlikely(dbi == FREE_DBI && !F_ISSET(txn->mt_flags, MDBX_TXN_RDONLY))) - return MDBX_EINVAL; - - if (txn->mt_dbs[dbi].md_flags & MDBX_DUPSORT) - size += sizeof(MDBX_xcursor); - - if (likely((mc = malloc(size)) != NULL)) { - mdbx_cursor_init(mc, txn, dbi, (MDBX_xcursor *)(mc + 1)); - if (txn->mt_cursors) { - mc->mc_next = txn->mt_cursors[dbi]; - txn->mt_cursors[dbi] = mc; - mc->mc_flags |= C_UNTRACK; - } - } else { - return MDBX_ENOMEM; - } - - *ret = mc; - - return MDBX_SUCCESS; -} - -int mdbx_cursor_renew(MDBX_txn *txn, MDBX_cursor *mc) { - if (unlikely(!mc || !txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(mc->mc_signature != MDBX_MC_SIGNATURE && - mc->mc_signature != MDBX_MC_READY4CLOSE)) - return MDBX_EINVAL; - - if (unlikely(!TXN_DBI_EXIST(txn, mc->mc_dbi, DB_VALID))) - return MDBX_EINVAL; - - if (unlikely(mc->mc_backup)) - return MDBX_EINVAL; - - if (unlikely((mc->mc_flags & C_UNTRACK) || txn->mt_cursors)) { - MDBX_cursor **prev = &mc->mc_txn->mt_cursors[mc->mc_dbi]; - while (*prev && *prev != mc) - prev = &(*prev)->mc_next; - if (*prev == mc) - *prev = mc->mc_next; - mc->mc_signature = MDBX_MC_READY4CLOSE; - } - - if (unlikely(txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - mdbx_cursor_init(mc, txn, mc->mc_dbi, mc->mc_xcursor); - return MDBX_SUCCESS; -} - -/* Return the count of duplicate data items for the current key */ -int mdbx_cursor_count(MDBX_cursor *mc, size_t *countp) { - if (unlikely(mc == NULL || countp == NULL)) - return MDBX_EINVAL; - - if (unlikely(mc->mc_signature != MDBX_MC_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(mc->mc_txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(mc->mc_txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - if (unlikely(!(mc->mc_flags & C_INITIALIZED))) - return MDBX_EINVAL; - - if (!mc->mc_snum) { - *countp = 0; - return MDBX_NOTFOUND; - } - - MDBX_page *mp = mc->mc_pg[mc->mc_top]; - if ((mc->mc_flags & C_EOF) && mc->mc_ki[mc->mc_top] >= NUMKEYS(mp)) { - *countp = 0; - return MDBX_NOTFOUND; - } - - *countp = 1; - if (mc->mc_xcursor != NULL) { - MDBX_node *leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_cassert(mc, mc->mc_xcursor && (mc->mc_xcursor->mx_cursor.mc_flags & - C_INITIALIZED)); - *countp = unlikely(mc->mc_xcursor->mx_db.md_entries > SIZE_MAX) - ? SIZE_MAX - : (size_t)mc->mc_xcursor->mx_db.md_entries; - } - } - return MDBX_SUCCESS; -} - -void mdbx_cursor_close(MDBX_cursor *mc) { - if (mc) { - mdbx_ensure(NULL, mc->mc_signature == MDBX_MC_SIGNATURE || - mc->mc_signature == MDBX_MC_READY4CLOSE); - if (!mc->mc_backup) { - /* Remove from txn, if tracked. - * A read-only txn (!C_UNTRACK) may have been freed already, - * so do not peek inside it. Only write txns track cursors. */ - if ((mc->mc_flags & C_UNTRACK) && mc->mc_txn->mt_cursors) { - MDBX_cursor **prev = &mc->mc_txn->mt_cursors[mc->mc_dbi]; - while (*prev && *prev != mc) - prev = &(*prev)->mc_next; - if (*prev == mc) - *prev = mc->mc_next; - } - mc->mc_signature = 0; - free(mc); - } else { - /* cursor closed before nested txn ends */ - mdbx_cassert(mc, mc->mc_signature == MDBX_MC_SIGNATURE); - mc->mc_signature = MDBX_MC_WAIT4EOT; - } - } -} - -MDBX_txn *mdbx_cursor_txn(MDBX_cursor *mc) { - if (unlikely(!mc || mc->mc_signature != MDBX_MC_SIGNATURE)) - return NULL; - return mc->mc_txn; -} - -MDBX_dbi mdbx_cursor_dbi(MDBX_cursor *mc) { - if (unlikely(!mc || mc->mc_signature != MDBX_MC_SIGNATURE)) - return UINT_MAX; - return mc->mc_dbi; -} - -/* Replace the key for a branch node with a new key. - * Set MDBX_TXN_ERROR on failure. - * [in] mc Cursor pointing to the node to operate on. - * [in] key The new key to use. - * Returns 0 on success, non-zero on failure. */ -static int mdbx_update_key(MDBX_cursor *mc, MDBX_val *key) { - MDBX_page *mp; - MDBX_node *node; - char *base; - size_t len; - int delta, ksize, oksize; - indx_t ptr, i, numkeys, indx; - DKBUF; - - indx = mc->mc_ki[mc->mc_top]; - mp = mc->mc_pg[mc->mc_top]; - node = NODEPTR(mp, indx); - ptr = mp->mp_ptrs[indx]; - if (MDBX_DEBUG) { - MDBX_val k2; - char kbuf2[DKBUF_MAXKEYSIZE * 2 + 1]; - k2.iov_base = NODEKEY(node); - k2.iov_len = node->mn_ksize; - mdbx_debug("update key %u (ofs %u) [%s] to [%s] on page %" PRIaPGNO "", - indx, ptr, mdbx_dkey(&k2, kbuf2, sizeof(kbuf2)), DKEY(key), - mp->mp_pgno); - } - - /* Sizes must be 2-byte aligned. */ - ksize = EVEN(key->iov_len); - oksize = EVEN(node->mn_ksize); - delta = ksize - oksize; - - /* Shift node contents if EVEN(key length) changed. */ - if (delta) { - if (delta > 0 && SIZELEFT(mp) < delta) { - pgno_t pgno; - /* not enough space left, do a delete and split */ - mdbx_debug("Not enough room, delta = %d, splitting...", delta); - pgno = NODEPGNO(node); - mdbx_node_del(mc, 0); - return mdbx_page_split(mc, key, NULL, pgno, MDBX_SPLIT_REPLACE); - } - - numkeys = NUMKEYS(mp); - for (i = 0; i < numkeys; i++) { - if (mp->mp_ptrs[i] <= ptr) { - mdbx_cassert(mc, mp->mp_ptrs[i] >= delta); - mp->mp_ptrs[i] -= (indx_t)delta; - } - } - - base = (char *)mp + mp->mp_upper + PAGEHDRSZ; - len = ptr - mp->mp_upper + NODESIZE; - memmove(base - delta, base, len); - mdbx_cassert(mc, mp->mp_upper >= delta); - mp->mp_upper -= (indx_t)delta; - - node = NODEPTR(mp, indx); - } - - /* But even if no shift was needed, update ksize */ - if (node->mn_ksize != key->iov_len) - node->mn_ksize = (uint16_t)key->iov_len; - - if (key->iov_len) - memcpy(NODEKEY(node), key->iov_base, key->iov_len); - - return MDBX_SUCCESS; -} - -static void mdbx_cursor_copy(const MDBX_cursor *csrc, MDBX_cursor *cdst); - -/* Perform act while tracking temporary cursor mn */ -#define WITH_CURSOR_TRACKING(mn, act) \ - do { \ - mdbx_cassert(&(mn), \ - mn.mc_txn->mt_cursors != NULL /* must be not rdonly txt */); \ - MDBX_cursor mc_dummy, *tracked, \ - **tp = &(mn).mc_txn->mt_cursors[mn.mc_dbi]; \ - if ((mn).mc_flags & C_SUB) { \ - mc_dummy.mc_flags = C_INITIALIZED; \ - mc_dummy.mc_xcursor = (MDBX_xcursor *)&(mn); \ - tracked = &mc_dummy; \ - } else { \ - tracked = &(mn); \ - } \ - tracked->mc_next = *tp; \ - *tp = tracked; \ - { act; } \ - *tp = tracked->mc_next; \ - } while (0) - -/* Move a node from csrc to cdst. */ -static int mdbx_node_move(MDBX_cursor *csrc, MDBX_cursor *cdst, int fromleft) { - MDBX_node *srcnode; - MDBX_val key, data; - pgno_t srcpg; - MDBX_cursor mn; - int rc; - unsigned flags; - - DKBUF; - - /* Mark src and dst as dirty. */ - if (unlikely((rc = mdbx_page_touch(csrc)) || (rc = mdbx_page_touch(cdst)))) - return rc; - - if (IS_LEAF2(csrc->mc_pg[csrc->mc_top])) { - key.iov_len = csrc->mc_db->md_xsize; - key.iov_base = LEAF2KEY(csrc->mc_pg[csrc->mc_top], - csrc->mc_ki[csrc->mc_top], key.iov_len); - data.iov_len = 0; - data.iov_base = NULL; - srcpg = 0; - flags = 0; - } else { - srcnode = NODEPTR(csrc->mc_pg[csrc->mc_top], csrc->mc_ki[csrc->mc_top]); - mdbx_cassert(csrc, !((size_t)srcnode & 1)); - srcpg = NODEPGNO(srcnode); - flags = srcnode->mn_flags; - if (csrc->mc_ki[csrc->mc_top] == 0 && - IS_BRANCH(csrc->mc_pg[csrc->mc_top])) { - unsigned snum = csrc->mc_snum; - MDBX_node *s2; - /* must find the lowest key below src */ - rc = mdbx_page_search_lowest(csrc); - if (unlikely(rc)) - return rc; - if (IS_LEAF2(csrc->mc_pg[csrc->mc_top])) { - key.iov_len = csrc->mc_db->md_xsize; - key.iov_base = LEAF2KEY(csrc->mc_pg[csrc->mc_top], 0, key.iov_len); - } else { - s2 = NODEPTR(csrc->mc_pg[csrc->mc_top], 0); - key.iov_len = NODEKSZ(s2); - key.iov_base = NODEKEY(s2); - } - mdbx_cassert(csrc, snum >= 1 && snum <= UINT16_MAX); - csrc->mc_snum = (uint16_t)snum--; - csrc->mc_top = (uint16_t)snum; - } else { - key.iov_len = NODEKSZ(srcnode); - key.iov_base = NODEKEY(srcnode); - } - data.iov_len = NODEDSZ(srcnode); - data.iov_base = NODEDATA(srcnode); - } - mn.mc_xcursor = NULL; - if (IS_BRANCH(cdst->mc_pg[cdst->mc_top]) && cdst->mc_ki[cdst->mc_top] == 0) { - unsigned snum = cdst->mc_snum; - MDBX_node *s2; - MDBX_val bkey; - /* must find the lowest key below dst */ - mdbx_cursor_copy(cdst, &mn); - rc = mdbx_page_search_lowest(&mn); - if (unlikely(rc)) - return rc; - if (IS_LEAF2(mn.mc_pg[mn.mc_top])) { - bkey.iov_len = mn.mc_db->md_xsize; - bkey.iov_base = LEAF2KEY(mn.mc_pg[mn.mc_top], 0, bkey.iov_len); - } else { - s2 = NODEPTR(mn.mc_pg[mn.mc_top], 0); - bkey.iov_len = NODEKSZ(s2); - bkey.iov_base = NODEKEY(s2); - } - mdbx_cassert(csrc, snum >= 1 && snum <= UINT16_MAX); - mn.mc_snum = (uint16_t)snum--; - mn.mc_top = (uint16_t)snum; - mn.mc_ki[snum] = 0; - rc = mdbx_update_key(&mn, &bkey); - if (unlikely(rc)) - return rc; - } - - mdbx_debug("moving %s node %u [%s] on page %" PRIaPGNO - " to node %u on page %" PRIaPGNO "", - IS_LEAF(csrc->mc_pg[csrc->mc_top]) ? "leaf" : "branch", - csrc->mc_ki[csrc->mc_top], DKEY(&key), - csrc->mc_pg[csrc->mc_top]->mp_pgno, cdst->mc_ki[cdst->mc_top], - cdst->mc_pg[cdst->mc_top]->mp_pgno); - - /* Add the node to the destination page. */ - rc = - mdbx_node_add(cdst, cdst->mc_ki[cdst->mc_top], &key, &data, srcpg, flags); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - /* Delete the node from the source page. */ - mdbx_node_del(csrc, key.iov_len); - - { - /* Adjust other cursors pointing to mp */ - MDBX_cursor *m2, *m3; - MDBX_dbi dbi = csrc->mc_dbi; - MDBX_page *mpd, *mps; - - mps = csrc->mc_pg[csrc->mc_top]; - /* If we're adding on the left, bump others up */ - if (fromleft) { - mpd = cdst->mc_pg[csrc->mc_top]; - for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) { - if (csrc->mc_flags & C_SUB) - m3 = &m2->mc_xcursor->mx_cursor; - else - m3 = m2; - if (!(m3->mc_flags & C_INITIALIZED) || m3->mc_top < csrc->mc_top) - continue; - if (m3 != cdst && m3->mc_pg[csrc->mc_top] == mpd && - m3->mc_ki[csrc->mc_top] >= cdst->mc_ki[csrc->mc_top]) { - m3->mc_ki[csrc->mc_top]++; - } - if (m3 != csrc && m3->mc_pg[csrc->mc_top] == mps && - m3->mc_ki[csrc->mc_top] == csrc->mc_ki[csrc->mc_top]) { - m3->mc_pg[csrc->mc_top] = cdst->mc_pg[cdst->mc_top]; - m3->mc_ki[csrc->mc_top] = cdst->mc_ki[cdst->mc_top]; - m3->mc_ki[csrc->mc_top - 1]++; - } - if (XCURSOR_INITED(m3) && IS_LEAF(mps)) - XCURSOR_REFRESH(m3, m3->mc_pg[csrc->mc_top], m3->mc_ki[csrc->mc_top]); - } - } else - /* Adding on the right, bump others down */ - { - for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) { - if (csrc->mc_flags & C_SUB) - m3 = &m2->mc_xcursor->mx_cursor; - else - m3 = m2; - if (m3 == csrc) - continue; - if (!(m3->mc_flags & C_INITIALIZED) || m3->mc_top < csrc->mc_top) - continue; - if (m3->mc_pg[csrc->mc_top] == mps) { - if (!m3->mc_ki[csrc->mc_top]) { - m3->mc_pg[csrc->mc_top] = cdst->mc_pg[cdst->mc_top]; - m3->mc_ki[csrc->mc_top] = cdst->mc_ki[cdst->mc_top]; - m3->mc_ki[csrc->mc_top - 1]--; - } else { - m3->mc_ki[csrc->mc_top]--; - } - if (XCURSOR_INITED(m3) && IS_LEAF(mps)) - XCURSOR_REFRESH(m3, m3->mc_pg[csrc->mc_top], - m3->mc_ki[csrc->mc_top]); - } - } - } - } - - /* Update the parent separators. */ - if (csrc->mc_ki[csrc->mc_top] == 0) { - if (csrc->mc_ki[csrc->mc_top - 1] != 0) { - if (IS_LEAF2(csrc->mc_pg[csrc->mc_top])) { - key.iov_base = LEAF2KEY(csrc->mc_pg[csrc->mc_top], 0, key.iov_len); - } else { - srcnode = NODEPTR(csrc->mc_pg[csrc->mc_top], 0); - key.iov_len = NODEKSZ(srcnode); - key.iov_base = NODEKEY(srcnode); - } - mdbx_debug("update separator for source page %" PRIaPGNO " to [%s]", - csrc->mc_pg[csrc->mc_top]->mp_pgno, DKEY(&key)); - mdbx_cursor_copy(csrc, &mn); - mn.mc_snum--; - mn.mc_top--; - /* We want mdbx_rebalance to find mn when doing fixups */ - WITH_CURSOR_TRACKING(mn, rc = mdbx_update_key(&mn, &key)); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - } - if (IS_BRANCH(csrc->mc_pg[csrc->mc_top])) { - MDBX_val nullkey; - indx_t ix = csrc->mc_ki[csrc->mc_top]; - nullkey.iov_len = 0; - csrc->mc_ki[csrc->mc_top] = 0; - rc = mdbx_update_key(csrc, &nullkey); - csrc->mc_ki[csrc->mc_top] = ix; - mdbx_cassert(csrc, rc == MDBX_SUCCESS); - } - } - - if (cdst->mc_ki[cdst->mc_top] == 0) { - if (cdst->mc_ki[cdst->mc_top - 1] != 0) { - if (IS_LEAF2(csrc->mc_pg[csrc->mc_top])) { - key.iov_base = LEAF2KEY(cdst->mc_pg[cdst->mc_top], 0, key.iov_len); - } else { - srcnode = NODEPTR(cdst->mc_pg[cdst->mc_top], 0); - key.iov_len = NODEKSZ(srcnode); - key.iov_base = NODEKEY(srcnode); - } - mdbx_debug("update separator for destination page %" PRIaPGNO " to [%s]", - cdst->mc_pg[cdst->mc_top]->mp_pgno, DKEY(&key)); - mdbx_cursor_copy(cdst, &mn); - mn.mc_snum--; - mn.mc_top--; - /* We want mdbx_rebalance to find mn when doing fixups */ - WITH_CURSOR_TRACKING(mn, rc = mdbx_update_key(&mn, &key)); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - } - if (IS_BRANCH(cdst->mc_pg[cdst->mc_top])) { - MDBX_val nullkey; - indx_t ix = cdst->mc_ki[cdst->mc_top]; - nullkey.iov_len = 0; - cdst->mc_ki[cdst->mc_top] = 0; - rc = mdbx_update_key(cdst, &nullkey); - cdst->mc_ki[cdst->mc_top] = ix; - mdbx_cassert(cdst, rc == MDBX_SUCCESS); - } - } - - return MDBX_SUCCESS; -} - -/* Merge one page into another. - * - * The nodes from the page pointed to by csrc will be copied to the page - * pointed to by cdst and then the csrc page will be freed. - * - * [in] csrc Cursor pointing to the source page. - * [in] cdst Cursor pointing to the destination page. - * - * Returns 0 on success, non-zero on failure. */ -static int mdbx_page_merge(MDBX_cursor *csrc, MDBX_cursor *cdst) { - MDBX_page *psrc, *pdst; - MDBX_node *srcnode; - MDBX_val key, data; - unsigned nkeys; - int rc; - unsigned i, j; - - psrc = csrc->mc_pg[csrc->mc_top]; - pdst = cdst->mc_pg[cdst->mc_top]; - - mdbx_debug("merging page %" PRIaPGNO " into %" PRIaPGNO "", psrc->mp_pgno, - pdst->mp_pgno); - - mdbx_cassert(csrc, csrc->mc_snum > 1); /* can't merge root page */ - mdbx_cassert(csrc, cdst->mc_snum > 1); - - /* Mark dst as dirty. */ - if (unlikely(rc = mdbx_page_touch(cdst))) - return rc; - - /* get dst page again now that we've touched it. */ - pdst = cdst->mc_pg[cdst->mc_top]; - - /* Move all nodes from src to dst. */ - j = nkeys = NUMKEYS(pdst); - if (IS_LEAF2(psrc)) { - key.iov_len = csrc->mc_db->md_xsize; - key.iov_base = PAGEDATA(psrc); - for (i = 0; i < NUMKEYS(psrc); i++, j++) { - rc = mdbx_node_add(cdst, j, &key, NULL, 0, 0); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - key.iov_base = (char *)key.iov_base + key.iov_len; - } - } else { - for (i = 0; i < NUMKEYS(psrc); i++, j++) { - srcnode = NODEPTR(psrc, i); - if (i == 0 && IS_BRANCH(psrc)) { - MDBX_cursor mn; - MDBX_node *s2; - mdbx_cursor_copy(csrc, &mn); - mn.mc_xcursor = NULL; - /* must find the lowest key below src */ - rc = mdbx_page_search_lowest(&mn); - if (unlikely(rc)) - return rc; - if (IS_LEAF2(mn.mc_pg[mn.mc_top])) { - key.iov_len = mn.mc_db->md_xsize; - key.iov_base = LEAF2KEY(mn.mc_pg[mn.mc_top], 0, key.iov_len); - } else { - s2 = NODEPTR(mn.mc_pg[mn.mc_top], 0); - key.iov_len = NODEKSZ(s2); - key.iov_base = NODEKEY(s2); - } - } else { - key.iov_len = srcnode->mn_ksize; - key.iov_base = NODEKEY(srcnode); - } - - data.iov_len = NODEDSZ(srcnode); - data.iov_base = NODEDATA(srcnode); - rc = mdbx_node_add(cdst, j, &key, &data, NODEPGNO(srcnode), - srcnode->mn_flags); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - } - } - - mdbx_debug("dst page %" PRIaPGNO " now has %u keys (%.1f%% filled)", - pdst->mp_pgno, NUMKEYS(pdst), - (float)PAGEFILL(cdst->mc_txn->mt_env, pdst) / 10); - - /* Unlink the src page from parent and add to free list. */ - csrc->mc_top--; - mdbx_node_del(csrc, 0); - if (csrc->mc_ki[csrc->mc_top] == 0) { - key.iov_len = 0; - rc = mdbx_update_key(csrc, &key); - if (unlikely(rc)) { - csrc->mc_top++; - return rc; - } - } - csrc->mc_top++; - - psrc = csrc->mc_pg[csrc->mc_top]; - /* If not operating on FreeDB, allow this page to be reused - * in this txn. Otherwise just add to free list. */ - rc = mdbx_page_loose(csrc, psrc); - if (unlikely(rc)) - return rc; - if (IS_LEAF(psrc)) - csrc->mc_db->md_leaf_pages--; - else - csrc->mc_db->md_branch_pages--; - { - /* Adjust other cursors pointing to mp */ - MDBX_cursor *m2, *m3; - MDBX_dbi dbi = csrc->mc_dbi; - unsigned top = csrc->mc_top; - - for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) { - if (csrc->mc_flags & C_SUB) - m3 = &m2->mc_xcursor->mx_cursor; - else - m3 = m2; - if (m3 == csrc) - continue; - if (m3->mc_snum < csrc->mc_snum) - continue; - if (m3->mc_pg[top] == psrc) { - m3->mc_pg[top] = pdst; - mdbx_cassert(m3, nkeys + m3->mc_ki[top] <= UINT16_MAX); - m3->mc_ki[top] += (indx_t)nkeys; - m3->mc_ki[top - 1] = cdst->mc_ki[top - 1]; - } else if (m3->mc_pg[top - 1] == csrc->mc_pg[top - 1] && - m3->mc_ki[top - 1] > csrc->mc_ki[top - 1]) { - m3->mc_ki[top - 1]--; - } - if (XCURSOR_INITED(m3) && IS_LEAF(psrc)) - XCURSOR_REFRESH(m3, m3->mc_pg[top], m3->mc_ki[top]); - } - } - { - unsigned snum = cdst->mc_snum; - uint16_t depth = cdst->mc_db->md_depth; - mdbx_cursor_pop(cdst); - rc = mdbx_rebalance(cdst); - /* Did the tree height change? */ - if (depth != cdst->mc_db->md_depth) - snum += cdst->mc_db->md_depth - depth; - mdbx_cassert(cdst, snum >= 1 && snum <= UINT16_MAX); - cdst->mc_snum = (uint16_t)snum; - cdst->mc_top = (uint16_t)(snum - 1); - } - return rc; -} - -/* Copy the contents of a cursor. - * [in] csrc The cursor to copy from. - * [out] cdst The cursor to copy to. */ -static void mdbx_cursor_copy(const MDBX_cursor *csrc, MDBX_cursor *cdst) { - unsigned i; - - mdbx_cassert(csrc, - csrc->mc_txn->mt_txnid >= csrc->mc_txn->mt_env->me_oldest[0]); - cdst->mc_txn = csrc->mc_txn; - cdst->mc_dbi = csrc->mc_dbi; - cdst->mc_db = csrc->mc_db; - cdst->mc_dbx = csrc->mc_dbx; - cdst->mc_snum = csrc->mc_snum; - cdst->mc_top = csrc->mc_top; - cdst->mc_flags = csrc->mc_flags; - - for (i = 0; i < csrc->mc_snum; i++) { - cdst->mc_pg[i] = csrc->mc_pg[i]; - cdst->mc_ki[i] = csrc->mc_ki[i]; - } -} - -/* Rebalance the tree after a delete operation. - * [in] mc Cursor pointing to the page where rebalancing should begin. - * Returns 0 on success, non-zero on failure. */ -static int mdbx_rebalance(MDBX_cursor *mc) { - MDBX_node *node; - int rc, fromleft; - unsigned ptop, minkeys, thresh; - MDBX_cursor mn; - indx_t oldki; - - if (IS_BRANCH(mc->mc_pg[mc->mc_top])) { - minkeys = 2; - thresh = 1; - } else { - minkeys = 1; - thresh = FILL_THRESHOLD; - } - mdbx_debug("rebalancing %s page %" PRIaPGNO " (has %u keys, %.1f%% full)", - IS_LEAF(mc->mc_pg[mc->mc_top]) ? "leaf" : "branch", - mc->mc_pg[mc->mc_top]->mp_pgno, NUMKEYS(mc->mc_pg[mc->mc_top]), - (float)PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) / 10); - - if (PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) >= thresh && - NUMKEYS(mc->mc_pg[mc->mc_top]) >= minkeys) { - mdbx_debug("no need to rebalance page %" PRIaPGNO ", above fill threshold", - mc->mc_pg[mc->mc_top]->mp_pgno); - return MDBX_SUCCESS; - } - - if (mc->mc_snum < 2) { - MDBX_page *mp = mc->mc_pg[0]; - unsigned nkeys = NUMKEYS(mp); - if (IS_SUBP(mp)) { - mdbx_debug("Can't rebalance a subpage, ignoring"); - return MDBX_SUCCESS; - } - if (nkeys == 0) { - mdbx_debug("tree is completely empty"); - mc->mc_db->md_root = P_INVALID; - mc->mc_db->md_depth = 0; - mc->mc_db->md_leaf_pages = 0; - rc = mdbx_pnl_append(&mc->mc_txn->mt_befree_pages, mp->mp_pgno); - if (unlikely(rc)) - return rc; - /* Adjust cursors pointing to mp */ - mc->mc_snum = 0; - mc->mc_top = 0; - mc->mc_flags &= ~C_INITIALIZED; - { - MDBX_cursor *m2, *m3; - MDBX_dbi dbi = mc->mc_dbi; - - for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) { - if (mc->mc_flags & C_SUB) - m3 = &m2->mc_xcursor->mx_cursor; - else - m3 = m2; - if (!(m3->mc_flags & C_INITIALIZED) || (m3->mc_snum < mc->mc_snum)) - continue; - if (m3->mc_pg[0] == mp) { - m3->mc_snum = 0; - m3->mc_top = 0; - m3->mc_flags &= ~C_INITIALIZED; - } - } - } - } else if (IS_BRANCH(mp) && NUMKEYS(mp) == 1) { - int i; - mdbx_debug("collapsing root page!"); - rc = mdbx_pnl_append(&mc->mc_txn->mt_befree_pages, mp->mp_pgno); - if (unlikely(rc)) - return rc; - mc->mc_db->md_root = NODEPGNO(NODEPTR(mp, 0)); - rc = mdbx_page_get(mc, mc->mc_db->md_root, &mc->mc_pg[0], NULL); - if (unlikely(rc)) - return rc; - mc->mc_db->md_depth--; - mc->mc_db->md_branch_pages--; - mc->mc_ki[0] = mc->mc_ki[1]; - for (i = 1; i < mc->mc_db->md_depth; i++) { - mc->mc_pg[i] = mc->mc_pg[i + 1]; - mc->mc_ki[i] = mc->mc_ki[i + 1]; - } - { - /* Adjust other cursors pointing to mp */ - MDBX_cursor *m2, *m3; - MDBX_dbi dbi = mc->mc_dbi; - - for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) { - if (mc->mc_flags & C_SUB) - m3 = &m2->mc_xcursor->mx_cursor; - else - m3 = m2; - if (m3 == mc) - continue; - if (!(m3->mc_flags & C_INITIALIZED)) - continue; - if (m3->mc_pg[0] == mp) { - for (i = 0; i < mc->mc_db->md_depth; i++) { - m3->mc_pg[i] = m3->mc_pg[i + 1]; - m3->mc_ki[i] = m3->mc_ki[i + 1]; - } - m3->mc_snum--; - m3->mc_top--; - } - } - } - } else { - mdbx_debug("root page %" PRIaPGNO - " doesn't need rebalancing (flags 0x%x)", - mp->mp_pgno, mp->mp_flags); - } - return MDBX_SUCCESS; - } - - /* The parent (branch page) must have at least 2 pointers, - * otherwise the tree is invalid. */ - ptop = mc->mc_top - 1; - mdbx_cassert(mc, NUMKEYS(mc->mc_pg[ptop]) > 1); - - /* Leaf page fill factor is below the threshold. - * Try to move keys from left or right neighbor, or - * merge with a neighbor page. */ - - /* Find neighbors. */ - mdbx_cursor_copy(mc, &mn); - mn.mc_xcursor = NULL; - - oldki = mc->mc_ki[mc->mc_top]; - if (mc->mc_ki[ptop] == 0) { - /* We're the leftmost leaf in our parent. */ - mdbx_debug("reading right neighbor"); - mn.mc_ki[ptop]++; - node = NODEPTR(mc->mc_pg[ptop], mn.mc_ki[ptop]); - rc = mdbx_page_get(mc, NODEPGNO(node), &mn.mc_pg[mn.mc_top], NULL); - if (unlikely(rc)) - return rc; - mn.mc_ki[mn.mc_top] = 0; - mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top]); - fromleft = 0; - } else { - /* There is at least one neighbor to the left. */ - mdbx_debug("reading left neighbor"); - mn.mc_ki[ptop]--; - node = NODEPTR(mc->mc_pg[ptop], mn.mc_ki[ptop]); - rc = mdbx_page_get(mc, NODEPGNO(node), &mn.mc_pg[mn.mc_top], NULL); - if (unlikely(rc)) - return rc; - mn.mc_ki[mn.mc_top] = NUMKEYS(mn.mc_pg[mn.mc_top]) - 1; - mc->mc_ki[mc->mc_top] = 0; - fromleft = 1; - } - - mdbx_debug("found neighbor page %" PRIaPGNO " (%u keys, %.1f%% full)", - mn.mc_pg[mn.mc_top]->mp_pgno, NUMKEYS(mn.mc_pg[mn.mc_top]), - (float)PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) / 10); - - /* If the neighbor page is above threshold and has enough keys, - * move one key from it. Otherwise we should try to merge them. - * (A branch page must never have less than 2 keys.) */ - if (PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) >= thresh && - NUMKEYS(mn.mc_pg[mn.mc_top]) > minkeys) { - rc = mdbx_node_move(&mn, mc, fromleft); - if (fromleft) { - /* if we inserted on left, bump position up */ - oldki++; - } - } else { - if (!fromleft) { - rc = mdbx_page_merge(&mn, mc); - } else { - oldki += NUMKEYS(mn.mc_pg[mn.mc_top]); - mn.mc_ki[mn.mc_top] += mc->mc_ki[mn.mc_top] + 1; - /* We want mdbx_rebalance to find mn when doing fixups */ - WITH_CURSOR_TRACKING(mn, rc = mdbx_page_merge(mc, &mn)); - mdbx_cursor_copy(&mn, mc); - } - mc->mc_flags &= ~C_EOF; - } - mc->mc_ki[mc->mc_top] = oldki; - return rc; -} - -/* Complete a delete operation started by mdbx_cursor_del(). */ -static int mdbx_cursor_del0(MDBX_cursor *mc) { - int rc; - MDBX_page *mp; - indx_t ki; - unsigned nkeys; - MDBX_cursor *m2, *m3; - MDBX_dbi dbi = mc->mc_dbi; - - ki = mc->mc_ki[mc->mc_top]; - mp = mc->mc_pg[mc->mc_top]; - mdbx_node_del(mc, mc->mc_db->md_xsize); - mc->mc_db->md_entries--; - { - /* Adjust other cursors pointing to mp */ - for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) { - m3 = (mc->mc_flags & C_SUB) ? &m2->mc_xcursor->mx_cursor : m2; - if (!(m2->mc_flags & m3->mc_flags & C_INITIALIZED)) - continue; - if (m3 == mc || m3->mc_snum < mc->mc_snum) - continue; - if (m3->mc_pg[mc->mc_top] == mp) { - if (m3->mc_ki[mc->mc_top] == ki) { - m3->mc_flags |= C_DEL; - if (mc->mc_db->md_flags & MDBX_DUPSORT) { - /* Sub-cursor referred into dataset which is gone */ - m3->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED | C_EOF); - } - continue; - } else if (m3->mc_ki[mc->mc_top] > ki) { - m3->mc_ki[mc->mc_top]--; - } - if (XCURSOR_INITED(m3)) - XCURSOR_REFRESH(m3, m3->mc_pg[mc->mc_top], m3->mc_ki[mc->mc_top]); - } - } - } - rc = mdbx_rebalance(mc); - - if (likely(rc == MDBX_SUCCESS)) { - /* DB is totally empty now, just bail out. - * Other cursors adjustments were already done - * by mdbx_rebalance and aren't needed here. */ - if (!mc->mc_snum) { - mc->mc_flags |= C_DEL | C_EOF; - return rc; - } - - mp = mc->mc_pg[mc->mc_top]; - nkeys = NUMKEYS(mp); - - /* Adjust other cursors pointing to mp */ - for (m2 = mc->mc_txn->mt_cursors[dbi]; !rc && m2; m2 = m2->mc_next) { - m3 = (mc->mc_flags & C_SUB) ? &m2->mc_xcursor->mx_cursor : m2; - if (!(m2->mc_flags & m3->mc_flags & C_INITIALIZED)) - continue; - if (m3->mc_snum < mc->mc_snum) - continue; - if (m3->mc_pg[mc->mc_top] == mp) { - /* if m3 points past last node in page, find next sibling */ - if (m3->mc_ki[mc->mc_top] >= mc->mc_ki[mc->mc_top]) { - if (m3->mc_ki[mc->mc_top] >= nkeys) { - rc = mdbx_cursor_sibling(m3, 1); - if (rc == MDBX_NOTFOUND) { - m3->mc_flags |= C_EOF; - rc = MDBX_SUCCESS; - continue; - } - } - if (mc->mc_db->md_flags & MDBX_DUPSORT) { - MDBX_node *node = - NODEPTR(m3->mc_pg[m3->mc_top], m3->mc_ki[m3->mc_top]); - /* If this node has dupdata, it may need to be reinited - * because its data has moved. - * If the xcursor was not initd it must be reinited. - * Else if node points to a subDB, nothing is needed. */ - if (node->mn_flags & F_DUPDATA) { - if (m3->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) { - if (!(node->mn_flags & F_SUBDATA)) - m3->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(node); - } else { - mdbx_xcursor_init1(m3, node); - m3->mc_xcursor->mx_cursor.mc_flags |= C_DEL; - } - } - } - } - } - } - mc->mc_flags |= C_DEL; - } - - if (unlikely(rc)) - mc->mc_txn->mt_flags |= MDBX_TXN_ERROR; - return rc; -} - -int mdbx_del(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *data) { - if (unlikely(!key || !txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - if (unlikely(txn->mt_flags & (MDBX_TXN_RDONLY | MDBX_TXN_BLOCKED))) - return (txn->mt_flags & MDBX_TXN_RDONLY) ? MDBX_EACCESS : MDBX_BAD_TXN; - - return mdbx_del0(txn, dbi, key, data, 0); -} - -static int mdbx_del0(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *data, - unsigned flags) { - MDBX_cursor mc; - MDBX_xcursor mx; - MDBX_cursor_op op; - MDBX_val rdata; - int rc, exact = 0; - DKBUF; - - mdbx_debug("====> delete db %u key [%s], data [%s]", dbi, DKEY(key), - DVAL(data)); - - mdbx_cursor_init(&mc, txn, dbi, &mx); - - if (data) { - op = MDBX_GET_BOTH; - rdata = *data; - data = &rdata; - } else { - op = MDBX_SET; - flags |= MDBX_NODUPDATA; - } - rc = mdbx_cursor_set(&mc, key, data, op, &exact); - if (likely(rc == MDBX_SUCCESS)) { - /* let mdbx_page_split know about this cursor if needed: - * delete will trigger a rebalance; if it needs to move - * a node from one page to another, it will have to - * update the parent's separator key(s). If the new sepkey - * is larger than the current one, the parent page may - * run out of space, triggering a split. We need this - * cursor to be consistent until the end of the rebalance. */ - mc.mc_next = txn->mt_cursors[dbi]; - txn->mt_cursors[dbi] = &mc; - rc = mdbx_cursor_del(&mc, flags); - txn->mt_cursors[dbi] = mc.mc_next; - } - return rc; -} - -/* Split a page and insert a new node. - * Set MDBX_TXN_ERROR on failure. - * [in,out] mc Cursor pointing to the page and desired insertion index. - * The cursor will be updated to point to the actual page and index where - * the node got inserted after the split. - * [in] newkey The key for the newly inserted node. - * [in] newdata The data for the newly inserted node. - * [in] newpgno The page number, if the new node is a branch node. - * [in] nflags The NODE_ADD_FLAGS for the new node. - * Returns 0 on success, non-zero on failure. */ -static int mdbx_page_split(MDBX_cursor *mc, MDBX_val *newkey, MDBX_val *newdata, - pgno_t newpgno, unsigned nflags) { - unsigned flags; - int rc = MDBX_SUCCESS, new_root = 0, did_split = 0; - pgno_t pgno = 0; - unsigned i, ptop; - MDBX_env *env = mc->mc_txn->mt_env; - MDBX_node *node; - MDBX_val sepkey, rkey, xdata, *rdata = &xdata; - MDBX_page *copy = NULL; - MDBX_page *rp, *pp; - MDBX_cursor mn; - DKBUF; - - MDBX_page *mp = mc->mc_pg[mc->mc_top]; - unsigned newindx = mc->mc_ki[mc->mc_top]; - unsigned nkeys = NUMKEYS(mp); - - mdbx_debug("-----> splitting %s page %" PRIaPGNO - " and adding [%s] at index %i/%i", - IS_LEAF(mp) ? "leaf" : "branch", mp->mp_pgno, DKEY(newkey), - mc->mc_ki[mc->mc_top], nkeys); - - /* Create a right sibling. */ - if ((rc = mdbx_page_new(mc, mp->mp_flags, 1, &rp))) - return rc; - rp->mp_leaf2_ksize = mp->mp_leaf2_ksize; - mdbx_debug("new right sibling: page %" PRIaPGNO "", rp->mp_pgno); - - /* Usually when splitting the root page, the cursor - * height is 1. But when called from mdbx_update_key, - * the cursor height may be greater because it walks - * up the stack while finding the branch slot to update. */ - if (mc->mc_top < 1) { - if ((rc = mdbx_page_new(mc, P_BRANCH, 1, &pp))) - goto done; - /* shift current top to make room for new parent */ - for (i = mc->mc_snum; i > 0; i--) { - mc->mc_pg[i] = mc->mc_pg[i - 1]; - mc->mc_ki[i] = mc->mc_ki[i - 1]; - } - mc->mc_pg[0] = pp; - mc->mc_ki[0] = 0; - mc->mc_db->md_root = pp->mp_pgno; - mdbx_debug("root split! new root = %" PRIaPGNO "", pp->mp_pgno); - new_root = mc->mc_db->md_depth++; - - /* Add left (implicit) pointer. */ - if (unlikely((rc = mdbx_node_add(mc, 0, NULL, NULL, mp->mp_pgno, 0)) != - MDBX_SUCCESS)) { - /* undo the pre-push */ - mc->mc_pg[0] = mc->mc_pg[1]; - mc->mc_ki[0] = mc->mc_ki[1]; - mc->mc_db->md_root = mp->mp_pgno; - mc->mc_db->md_depth--; - goto done; - } - mc->mc_snum++; - mc->mc_top++; - ptop = 0; - } else { - ptop = mc->mc_top - 1; - mdbx_debug("parent branch page is %" PRIaPGNO "", mc->mc_pg[ptop]->mp_pgno); - } - - mdbx_cursor_copy(mc, &mn); - mn.mc_xcursor = NULL; - mn.mc_pg[mn.mc_top] = rp; - mn.mc_ki[ptop] = mc->mc_ki[ptop] + 1; - - unsigned split_indx; - if (nflags & MDBX_APPEND) { - mn.mc_ki[mn.mc_top] = 0; - sepkey = *newkey; - split_indx = newindx; - nkeys = 0; - } else { - split_indx = (nkeys + 1) / 2; - - if (IS_LEAF2(rp)) { - char *split, *ins; - int x; - unsigned lsize, rsize, ksize; - /* Move half of the keys to the right sibling */ - x = mc->mc_ki[mc->mc_top] - split_indx; - ksize = mc->mc_db->md_xsize; - split = LEAF2KEY(mp, split_indx, ksize); - rsize = (nkeys - split_indx) * ksize; - lsize = (nkeys - split_indx) * sizeof(indx_t); - mdbx_cassert(mc, mp->mp_lower >= lsize); - mp->mp_lower -= (indx_t)lsize; - mdbx_cassert(mc, rp->mp_lower + lsize <= UINT16_MAX); - rp->mp_lower += (indx_t)lsize; - mdbx_cassert(mc, mp->mp_upper + rsize - lsize <= UINT16_MAX); - mp->mp_upper += (indx_t)(rsize - lsize); - mdbx_cassert(mc, rp->mp_upper >= rsize - lsize); - rp->mp_upper -= (indx_t)(rsize - lsize); - sepkey.iov_len = ksize; - if (newindx == split_indx) { - sepkey.iov_base = newkey->iov_base; - } else { - sepkey.iov_base = split; - } - if (x < 0) { - mdbx_cassert(mc, ksize >= sizeof(indx_t)); - ins = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], ksize); - memcpy(rp->mp_ptrs, split, rsize); - sepkey.iov_base = rp->mp_ptrs; - memmove(ins + ksize, ins, (split_indx - mc->mc_ki[mc->mc_top]) * ksize); - memcpy(ins, newkey->iov_base, ksize); - mdbx_cassert(mc, UINT16_MAX - mp->mp_lower >= (int)sizeof(indx_t)); - mp->mp_lower += sizeof(indx_t); - mdbx_cassert(mc, mp->mp_upper >= ksize - sizeof(indx_t)); - mp->mp_upper -= (indx_t)(ksize - sizeof(indx_t)); - } else { - if (x) - memcpy(rp->mp_ptrs, split, x * ksize); - ins = LEAF2KEY(rp, x, ksize); - memcpy(ins, newkey->iov_base, ksize); - memcpy(ins + ksize, split + x * ksize, rsize - x * ksize); - mdbx_cassert(mc, UINT16_MAX - rp->mp_lower >= (int)sizeof(indx_t)); - rp->mp_lower += sizeof(indx_t); - mdbx_cassert(mc, rp->mp_upper >= ksize - sizeof(indx_t)); - rp->mp_upper -= (indx_t)(ksize - sizeof(indx_t)); - mdbx_cassert(mc, x <= UINT16_MAX); - mc->mc_ki[mc->mc_top] = (indx_t)x; - } - } else { - size_t psize, nsize, k; - /* Maximum free space in an empty page */ - unsigned pmax = env->me_psize - PAGEHDRSZ; - if (IS_LEAF(mp)) - nsize = mdbx_leaf_size(env, newkey, newdata); - else - nsize = mdbx_branch_size(env, newkey); - nsize = EVEN(nsize); - - /* grab a page to hold a temporary copy */ - copy = mdbx_page_malloc(mc->mc_txn, 1); - if (unlikely(copy == NULL)) { - rc = MDBX_ENOMEM; - goto done; - } - copy->mp_pgno = mp->mp_pgno; - copy->mp_flags = mp->mp_flags; - copy->mp_lower = 0; - mdbx_cassert(mc, env->me_psize - PAGEHDRSZ <= UINT16_MAX); - copy->mp_upper = (indx_t)(env->me_psize - PAGEHDRSZ); - - /* prepare to insert */ - for (unsigned j = i = 0; i < nkeys; i++) { - if (i == newindx) - copy->mp_ptrs[j++] = 0; - copy->mp_ptrs[j++] = mp->mp_ptrs[i]; - } - - /* When items are relatively large the split point needs - * to be checked, because being off-by-one will make the - * difference between success or failure in mdbx_node_add. - * - * It's also relevant if a page happens to be laid out - * such that one half of its nodes are all "small" and - * the other half of its nodes are "large." If the new - * item is also "large" and falls on the half with - * "large" nodes, it also may not fit. - * - * As a final tweak, if the new item goes on the last - * spot on the page (and thus, onto the new page), bias - * the split so the new page is emptier than the old page. - * This yields better packing during sequential inserts. - */ - int dir; - if (nkeys < 20 || nsize > pmax / 16 || newindx >= nkeys) { - /* Find split point */ - psize = 0; - if (newindx <= split_indx || newindx >= nkeys) { - i = 0; - dir = 1; - k = (newindx >= nkeys) ? nkeys : split_indx + 1 + IS_LEAF(mp); - } else { - i = nkeys; - dir = -1; - k = split_indx - 1; - } - for (; i != k; i += dir) { - if (i == newindx) { - psize += nsize; - node = NULL; - } else { - node = (MDBX_node *)((char *)mp + copy->mp_ptrs[i] + PAGEHDRSZ); - psize += NODESIZE + NODEKSZ(node) + sizeof(indx_t); - if (IS_LEAF(mp)) { - if (F_ISSET(node->mn_flags, F_BIGDATA)) - psize += sizeof(pgno_t); - else - psize += NODEDSZ(node); - } - psize = EVEN(psize); - } - if (psize > pmax || i == k - dir) { - split_indx = i + (dir < 0); - break; - } - } - } - if (split_indx == newindx) { - sepkey.iov_len = newkey->iov_len; - sepkey.iov_base = newkey->iov_base; - } else { - node = - (MDBX_node *)((char *)mp + copy->mp_ptrs[split_indx] + PAGEHDRSZ); - sepkey.iov_len = node->mn_ksize; - sepkey.iov_base = NODEKEY(node); - } - } - } - - mdbx_debug("separator is %d [%s]", split_indx, DKEY(&sepkey)); - - /* Copy separator key to the parent. */ - if (SIZELEFT(mn.mc_pg[ptop]) < mdbx_branch_size(env, &sepkey)) { - int snum = mc->mc_snum; - mn.mc_snum--; - mn.mc_top--; - did_split = 1; - /* We want other splits to find mn when doing fixups */ - WITH_CURSOR_TRACKING( - mn, rc = mdbx_page_split(&mn, &sepkey, NULL, rp->mp_pgno, 0)); - if (unlikely(rc != MDBX_SUCCESS)) - goto done; - - /* root split? */ - if (mc->mc_snum > snum) - ptop++; - - /* Right page might now have changed parent. - * Check if left page also changed parent. */ - if (mn.mc_pg[ptop] != mc->mc_pg[ptop] && - mc->mc_ki[ptop] >= NUMKEYS(mc->mc_pg[ptop])) { - for (i = 0; i < ptop; i++) { - mc->mc_pg[i] = mn.mc_pg[i]; - mc->mc_ki[i] = mn.mc_ki[i]; - } - mc->mc_pg[ptop] = mn.mc_pg[ptop]; - if (mn.mc_ki[ptop]) { - mc->mc_ki[ptop] = mn.mc_ki[ptop] - 1; - } else { - /* find right page's left sibling */ - mc->mc_ki[ptop] = mn.mc_ki[ptop]; - rc = mdbx_cursor_sibling(mc, 0); - } - } - } else { - mn.mc_top--; - rc = mdbx_node_add(&mn, mn.mc_ki[ptop], &sepkey, NULL, rp->mp_pgno, 0); - mn.mc_top++; - } - if (unlikely(rc != MDBX_SUCCESS)) { - if (rc == MDBX_NOTFOUND) /* improper mdbx_cursor_sibling() result */ { - mdbx_error("unexpected %s", "MDBX_NOTFOUND"); - rc = MDBX_PROBLEM; - } - goto done; - } - if (nflags & MDBX_APPEND) { - mc->mc_pg[mc->mc_top] = rp; - mc->mc_ki[mc->mc_top] = 0; - rc = mdbx_node_add(mc, 0, newkey, newdata, newpgno, nflags); - if (rc) - goto done; - for (i = 0; i < mc->mc_top; i++) - mc->mc_ki[i] = mn.mc_ki[i]; - } else if (!IS_LEAF2(mp)) { - /* Move nodes */ - mc->mc_pg[mc->mc_top] = rp; - i = split_indx; - indx_t n = 0; - do { - if (i == newindx) { - rkey.iov_base = newkey->iov_base; - rkey.iov_len = newkey->iov_len; - if (IS_LEAF(mp)) { - rdata = newdata; - } else - pgno = newpgno; - flags = nflags; - /* Update index for the new key. */ - mc->mc_ki[mc->mc_top] = n; - } else { - node = (MDBX_node *)((char *)mp + copy->mp_ptrs[i] + PAGEHDRSZ); - rkey.iov_base = NODEKEY(node); - rkey.iov_len = node->mn_ksize; - if (IS_LEAF(mp)) { - xdata.iov_base = NODEDATA(node); - xdata.iov_len = NODEDSZ(node); - rdata = &xdata; - } else - pgno = NODEPGNO(node); - flags = node->mn_flags; - } - - if (!IS_LEAF(mp) && n == 0) { - /* First branch index doesn't need key data. */ - rkey.iov_len = 0; - } - - rc = mdbx_node_add(mc, n, &rkey, rdata, pgno, flags); - if (rc) - goto done; - if (i == nkeys) { - i = 0; - n = 0; - mc->mc_pg[mc->mc_top] = copy; - } else { - i++; - n++; - } - } while (i != split_indx); - - nkeys = NUMKEYS(copy); - for (i = 0; i < nkeys; i++) - mp->mp_ptrs[i] = copy->mp_ptrs[i]; - mp->mp_lower = copy->mp_lower; - mp->mp_upper = copy->mp_upper; - memcpy(NODEPTR(mp, nkeys - 1), NODEPTR(copy, nkeys - 1), - env->me_psize - copy->mp_upper - PAGEHDRSZ); - - /* reset back to original page */ - if (newindx < split_indx) { - mc->mc_pg[mc->mc_top] = mp; - } else { - mc->mc_pg[mc->mc_top] = rp; - mc->mc_ki[ptop]++; - /* Make sure mc_ki is still valid. */ - if (mn.mc_pg[ptop] != mc->mc_pg[ptop] && - mc->mc_ki[ptop] >= NUMKEYS(mc->mc_pg[ptop])) { - for (i = 0; i <= ptop; i++) { - mc->mc_pg[i] = mn.mc_pg[i]; - mc->mc_ki[i] = mn.mc_ki[i]; - } - } - } - if (nflags & MDBX_RESERVE) { - node = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); - if (!(node->mn_flags & F_BIGDATA)) - newdata->iov_base = NODEDATA(node); - } - } else { - if (newindx >= split_indx) { - mc->mc_pg[mc->mc_top] = rp; - mc->mc_ki[ptop]++; - /* Make sure mc_ki is still valid. */ - if (mn.mc_pg[ptop] != mc->mc_pg[ptop] && - mc->mc_ki[ptop] >= NUMKEYS(mc->mc_pg[ptop])) { - for (i = 0; i <= ptop; i++) { - mc->mc_pg[i] = mn.mc_pg[i]; - mc->mc_ki[i] = mn.mc_ki[i]; - } - } - } - } - - { - /* Adjust other cursors pointing to mp */ - MDBX_cursor *m2, *m3; - MDBX_dbi dbi = mc->mc_dbi; - nkeys = NUMKEYS(mp); - - for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) { - if (mc->mc_flags & C_SUB) - m3 = &m2->mc_xcursor->mx_cursor; - else - m3 = m2; - if (m3 == mc) - continue; - if (!(m2->mc_flags & m3->mc_flags & C_INITIALIZED)) - continue; - if (new_root) { - int k; - /* sub cursors may be on different DB */ - if (m3->mc_pg[0] != mp) - continue; - /* root split */ - for (k = new_root; k >= 0; k--) { - m3->mc_ki[k + 1] = m3->mc_ki[k]; - m3->mc_pg[k + 1] = m3->mc_pg[k]; - } - if (m3->mc_ki[0] >= nkeys) { - m3->mc_ki[0] = 1; - } else { - m3->mc_ki[0] = 0; - } - m3->mc_pg[0] = mc->mc_pg[0]; - m3->mc_snum++; - m3->mc_top++; - } - if (m3->mc_top >= mc->mc_top && m3->mc_pg[mc->mc_top] == mp) { - if (m3->mc_ki[mc->mc_top] >= newindx && !(nflags & MDBX_SPLIT_REPLACE)) - m3->mc_ki[mc->mc_top]++; - if (m3->mc_ki[mc->mc_top] >= nkeys) { - m3->mc_pg[mc->mc_top] = rp; - mdbx_cassert(mc, m3->mc_ki[mc->mc_top] >= nkeys); - m3->mc_ki[mc->mc_top] -= (indx_t)nkeys; - for (i = 0; i < mc->mc_top; i++) { - m3->mc_ki[i] = mn.mc_ki[i]; - m3->mc_pg[i] = mn.mc_pg[i]; - } - } - } else if (!did_split && m3->mc_top >= ptop && - m3->mc_pg[ptop] == mc->mc_pg[ptop] && - m3->mc_ki[ptop] >= mc->mc_ki[ptop]) { - m3->mc_ki[ptop]++; - } - if (XCURSOR_INITED(m3) && IS_LEAF(mp)) - XCURSOR_REFRESH(m3, m3->mc_pg[mc->mc_top], m3->mc_ki[mc->mc_top]); - } - } - mdbx_debug("mp left: %d, rp left: %d", SIZELEFT(mp), SIZELEFT(rp)); - -done: - if (copy) /* tmp page */ - mdbx_page_free(env, copy); - if (unlikely(rc)) - mc->mc_txn->mt_flags |= MDBX_TXN_ERROR; - return rc; -} - -int mdbx_put(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *data, - unsigned flags) { - MDBX_cursor mc; - MDBX_xcursor mx; - - if (unlikely(!key || !data || !txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - if (unlikely(flags & - ~(MDBX_NOOVERWRITE | MDBX_NODUPDATA | MDBX_RESERVE | - MDBX_APPEND | MDBX_APPENDDUP | MDBX_CURRENT))) - return MDBX_EINVAL; - - if (unlikely(txn->mt_flags & (MDBX_TXN_RDONLY | MDBX_TXN_BLOCKED))) - return (txn->mt_flags & MDBX_TXN_RDONLY) ? MDBX_EACCESS : MDBX_BAD_TXN; - - mdbx_cursor_init(&mc, txn, dbi, &mx); - mc.mc_next = txn->mt_cursors[dbi]; - txn->mt_cursors[dbi] = &mc; - - int rc = MDBX_SUCCESS; - /* LY: support for update (explicit overwrite) */ - if (flags & MDBX_CURRENT) { - rc = mdbx_cursor_get(&mc, key, NULL, MDBX_SET); - if (likely(rc == MDBX_SUCCESS) && - (txn->mt_dbs[dbi].md_flags & MDBX_DUPSORT)) { - /* LY: allows update (explicit overwrite) only for unique keys */ - MDBX_node *leaf = NODEPTR(mc.mc_pg[mc.mc_top], mc.mc_ki[mc.mc_top]); - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_tassert(txn, XCURSOR_INITED(&mc) && - mc.mc_xcursor->mx_db.md_entries > 1); - rc = MDBX_EMULTIVAL; - } - } - } - - if (likely(rc == MDBX_SUCCESS)) - rc = mdbx_cursor_put(&mc, key, data, flags); - txn->mt_cursors[dbi] = mc.mc_next; - - return rc; -} - -#ifndef MDBX_WBUF -#define MDBX_WBUF (1024 * 1024) -#endif -#define MDBX_EOF 0x10 /* mdbx_env_copyfd1() is done reading */ - -/* State needed for a double-buffering compacting copy. */ -typedef struct mdbx_copy { - MDBX_env *mc_env; - MDBX_txn *mc_txn; - mdbx_condmutex_t mc_condmutex; - char *mc_wbuf[2]; - char *mc_over[2]; - size_t mc_wlen[2]; - size_t mc_olen[2]; - mdbx_filehandle_t mc_fd; - volatile int mc_error; - pgno_t mc_next_pgno; - short mc_toggle; /* Buffer number in provider */ - short mc_new; /* (0-2 buffers to write) | (MDBX_EOF at end) */ - /* Error code. Never cleared if set. Both threads can set nonzero - * to fail the copy. Not mutex-protected, MDBX expects atomic int. */ -} mdbx_copy; - -/* Dedicated writer thread for compacting copy. */ -static THREAD_RESULT __cold THREAD_CALL mdbx_env_copythr(void *arg) { - mdbx_copy *my = arg; - char *ptr; - int toggle = 0; - int rc; - -#if defined(F_SETNOSIGPIPE) - /* OS X delivers SIGPIPE to the whole process, not the thread that caused it. - * Disable SIGPIPE using platform specific fcntl. */ - int enabled = 1; - if (fcntl(my->mc_fd, F_SETNOSIGPIPE, &enabled)) - my->mc_error = errno; -#endif - -#if defined(SIGPIPE) && !defined(_WIN32) && !defined(_WIN64) - sigset_t set; - sigemptyset(&set); - sigaddset(&set, SIGPIPE); - rc = pthread_sigmask(SIG_BLOCK, &set, NULL); - if (rc != 0) - my->mc_error = rc; -#endif - - mdbx_condmutex_lock(&my->mc_condmutex); - while (!my->mc_error) { - while (!my->mc_new) - mdbx_condmutex_wait(&my->mc_condmutex); - if (my->mc_new == 0 + MDBX_EOF) /* 0 buffers, just EOF */ - break; - size_t wsize = my->mc_wlen[toggle]; - ptr = my->mc_wbuf[toggle]; - again: - if (wsize > 0 && !my->mc_error) { - rc = mdbx_write(my->mc_fd, ptr, wsize); - if (rc != MDBX_SUCCESS) { -#if defined(SIGPIPE) && !defined(_WIN32) && !defined(_WIN64) - if (rc == EPIPE) { - /* Collect the pending SIGPIPE, otherwise (at least OS X) - * gives it to the process on thread-exit (ITS#8504). */ - int tmp; - sigwait(&set, &tmp); - } -#endif - my->mc_error = rc; - } - } - - /* If there's an overflow page tail, write it too */ - if (my->mc_olen[toggle]) { - wsize = my->mc_olen[toggle]; - ptr = my->mc_over[toggle]; - my->mc_olen[toggle] = 0; - goto again; - } - my->mc_wlen[toggle] = 0; - toggle ^= 1; - /* Return the empty buffer to provider */ - my->mc_new--; - mdbx_condmutex_signal(&my->mc_condmutex); - } - mdbx_condmutex_unlock(&my->mc_condmutex); - return (THREAD_RESULT)0; -} - -/* Give buffer and/or MDBX_EOF to writer thread, await unused buffer. - * - * [in] my control structure. - * [in] adjust (1 to hand off 1 buffer) | (MDBX_EOF when ending). */ -static int __cold mdbx_env_cthr_toggle(mdbx_copy *my, int adjust) { - mdbx_condmutex_lock(&my->mc_condmutex); - my->mc_new += (short)adjust; - mdbx_condmutex_signal(&my->mc_condmutex); - while (my->mc_new & 2) /* both buffers in use */ - mdbx_condmutex_wait(&my->mc_condmutex); - mdbx_condmutex_unlock(&my->mc_condmutex); - - my->mc_toggle ^= (adjust & 1); - /* Both threads reset mc_wlen, to be safe from threading errors */ - my->mc_wlen[my->mc_toggle] = 0; - return my->mc_error; -} - -/* Depth-first tree traversal for compacting copy. - * [in] my control structure. - * [in,out] pg database root. - * [in] flags includes F_DUPDATA if it is a sorted-duplicate sub-DB. */ -static int __cold mdbx_env_cwalk(mdbx_copy *my, pgno_t *pg, int flags) { - MDBX_cursor mc; - MDBX_node *ni; - MDBX_page *mo, *mp, *leaf; - char *buf, *ptr; - int rc, toggle; - unsigned i; - - /* Empty DB, nothing to do */ - if (*pg == P_INVALID) - return MDBX_SUCCESS; - - memset(&mc, 0, sizeof(mc)); - mc.mc_snum = 1; - mc.mc_txn = my->mc_txn; - - rc = mdbx_page_get(&mc, *pg, &mc.mc_pg[0], NULL); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - rc = mdbx_page_search_root(&mc, NULL, MDBX_PS_FIRST); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - /* Make cursor pages writable */ - buf = ptr = malloc(pgno2bytes(my->mc_env, mc.mc_snum)); - if (buf == NULL) - return MDBX_ENOMEM; - - for (i = 0; i < mc.mc_top; i++) { - mdbx_page_copy((MDBX_page *)ptr, mc.mc_pg[i], my->mc_env->me_psize); - mc.mc_pg[i] = (MDBX_page *)ptr; - ptr += my->mc_env->me_psize; - } - - /* This is writable space for a leaf page. Usually not needed. */ - leaf = (MDBX_page *)ptr; - - toggle = my->mc_toggle; - while (mc.mc_snum > 0) { - unsigned n; - mp = mc.mc_pg[mc.mc_top]; - n = NUMKEYS(mp); - - if (IS_LEAF(mp)) { - if (!IS_LEAF2(mp) && !(flags & F_DUPDATA)) { - for (i = 0; i < n; i++) { - ni = NODEPTR(mp, i); - if (ni->mn_flags & F_BIGDATA) { - MDBX_page *omp; - - /* Need writable leaf */ - if (mp != leaf) { - mc.mc_pg[mc.mc_top] = leaf; - mdbx_page_copy(leaf, mp, my->mc_env->me_psize); - mp = leaf; - ni = NODEPTR(mp, i); - } - - pgno_t pgno; - memcpy(&pgno, NODEDATA(ni), sizeof(pgno)); - memcpy(NODEDATA(ni), &my->mc_next_pgno, sizeof(pgno_t)); - rc = mdbx_page_get(&mc, pgno, &omp, NULL); - if (unlikely(rc != MDBX_SUCCESS)) - goto done; - if (my->mc_wlen[toggle] >= MDBX_WBUF) { - rc = mdbx_env_cthr_toggle(my, 1); - if (unlikely(rc != MDBX_SUCCESS)) - goto done; - toggle = my->mc_toggle; - } - mo = (MDBX_page *)(my->mc_wbuf[toggle] + my->mc_wlen[toggle]); - memcpy(mo, omp, my->mc_env->me_psize); - mo->mp_pgno = my->mc_next_pgno; - my->mc_next_pgno += omp->mp_pages; - my->mc_wlen[toggle] += my->mc_env->me_psize; - if (omp->mp_pages > 1) { - my->mc_olen[toggle] = pgno2bytes(my->mc_env, omp->mp_pages - 1); - my->mc_over[toggle] = (char *)omp + my->mc_env->me_psize; - rc = mdbx_env_cthr_toggle(my, 1); - if (unlikely(rc != MDBX_SUCCESS)) - goto done; - toggle = my->mc_toggle; - } - } else if (ni->mn_flags & F_SUBDATA) { - MDBX_db db; - - /* Need writable leaf */ - if (mp != leaf) { - mc.mc_pg[mc.mc_top] = leaf; - mdbx_page_copy(leaf, mp, my->mc_env->me_psize); - mp = leaf; - ni = NODEPTR(mp, i); - } - - memcpy(&db, NODEDATA(ni), sizeof(db)); - my->mc_toggle = (short)toggle; - rc = mdbx_env_cwalk(my, &db.md_root, ni->mn_flags & F_DUPDATA); - if (rc) - goto done; - toggle = my->mc_toggle; - memcpy(NODEDATA(ni), &db, sizeof(db)); - } - } - } - } else { - mc.mc_ki[mc.mc_top]++; - if (mc.mc_ki[mc.mc_top] < n) { - pgno_t pgno; - again: - ni = NODEPTR(mp, mc.mc_ki[mc.mc_top]); - pgno = NODEPGNO(ni); - rc = mdbx_page_get(&mc, pgno, &mp, NULL); - if (unlikely(rc != MDBX_SUCCESS)) - goto done; - mc.mc_top++; - mc.mc_snum++; - mc.mc_ki[mc.mc_top] = 0; - if (IS_BRANCH(mp)) { - /* Whenever we advance to a sibling branch page, - * we must proceed all the way down to its first leaf. */ - mdbx_page_copy(mc.mc_pg[mc.mc_top], mp, my->mc_env->me_psize); - goto again; - } else - mc.mc_pg[mc.mc_top] = mp; - continue; - } - } - if (my->mc_wlen[toggle] >= MDBX_WBUF) { - rc = mdbx_env_cthr_toggle(my, 1); - if (unlikely(rc != MDBX_SUCCESS)) - goto done; - toggle = my->mc_toggle; - } - mo = (MDBX_page *)(my->mc_wbuf[toggle] + my->mc_wlen[toggle]); - mdbx_page_copy(mo, mp, my->mc_env->me_psize); - mo->mp_pgno = my->mc_next_pgno++; - my->mc_wlen[toggle] += my->mc_env->me_psize; - if (mc.mc_top) { - /* Update parent if there is one */ - ni = NODEPTR(mc.mc_pg[mc.mc_top - 1], mc.mc_ki[mc.mc_top - 1]); - SETPGNO(ni, mo->mp_pgno); - mdbx_cursor_pop(&mc); - } else { - /* Otherwise we're done */ - *pg = mo->mp_pgno; - break; - } - } -done: - free(buf); - return rc; -} - -/* Copy environment with compaction. */ -static int __cold mdbx_env_compact(MDBX_env *env, mdbx_filehandle_t fd) { - MDBX_txn *txn = NULL; - mdbx_thread_t thr; - mdbx_copy my; - memset(&my, 0, sizeof(my)); - - int rc = mdbx_condmutex_init(&my.mc_condmutex); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - rc = mdbx_memalign_alloc(env->me_os_psize, MDBX_WBUF * 2, - (void **)&my.mc_wbuf[0]); - if (unlikely(rc != MDBX_SUCCESS)) - goto done; - - memset(my.mc_wbuf[0], 0, MDBX_WBUF * 2); - my.mc_wbuf[1] = my.mc_wbuf[0] + MDBX_WBUF; - my.mc_next_pgno = NUM_METAS; - my.mc_env = env; - my.mc_fd = fd; - rc = mdbx_thread_create(&thr, mdbx_env_copythr, &my); - if (unlikely(rc != MDBX_SUCCESS)) - goto done; - - rc = mdbx_txn_begin(env, NULL, MDBX_RDONLY, &txn); - if (unlikely(rc != MDBX_SUCCESS)) - goto finish; - - MDBX_page *meta = mdbx_init_metas(env, my.mc_wbuf[0]); - - /* Set metapage 1 with current main DB */ - pgno_t new_root, root = txn->mt_dbs[MAIN_DBI].md_root; - if ((new_root = root) != P_INVALID) { - /* Count free pages + freeDB pages. Subtract from last_pg - * to find the new last_pg, which also becomes the new root. */ - pgno_t freecount = 0; - MDBX_cursor mc; - MDBX_val key, data; - - mdbx_cursor_init(&mc, txn, FREE_DBI, NULL); - while ((rc = mdbx_cursor_get(&mc, &key, &data, MDBX_NEXT)) == 0) - freecount += *(pgno_t *)data.iov_base; - if (unlikely(rc != MDBX_NOTFOUND)) - goto finish; - - freecount += txn->mt_dbs[FREE_DBI].md_branch_pages + - txn->mt_dbs[FREE_DBI].md_leaf_pages + - txn->mt_dbs[FREE_DBI].md_overflow_pages; - - new_root = txn->mt_next_pgno - 1 - freecount; - meta->mp_meta.mm_geo.next = meta->mp_meta.mm_geo.now = new_root + 1; - meta->mp_meta.mm_dbs[MAIN_DBI] = txn->mt_dbs[MAIN_DBI]; - meta->mp_meta.mm_dbs[MAIN_DBI].md_root = new_root; - } else { - /* When the DB is empty, handle it specially to - * fix any breakage like page leaks from ITS#8174. */ - meta->mp_meta.mm_dbs[MAIN_DBI].md_flags = txn->mt_dbs[MAIN_DBI].md_flags; - } - - /* copy canary sequenses if present */ - if (txn->mt_canary.v) { - meta->mp_meta.mm_canary = txn->mt_canary; - meta->mp_meta.mm_canary.v = mdbx_meta_txnid_stable(env, &meta->mp_meta); - } - - /* update signature */ - meta->mp_meta.mm_datasync_sign = mdbx_meta_sign(&meta->mp_meta); - - my.mc_wlen[0] = pgno2bytes(env, NUM_METAS); - my.mc_txn = txn; - rc = mdbx_env_cwalk(&my, &root, 0); - if (rc == MDBX_SUCCESS && root != new_root) { - mdbx_error("unexpected root %" PRIaPGNO " (%" PRIaPGNO ")", root, new_root); - rc = MDBX_PROBLEM; /* page leak or corrupt DB */ - } - -finish: - if (rc != MDBX_SUCCESS) - my.mc_error = rc; - mdbx_env_cthr_toggle(&my, 1 | MDBX_EOF); - rc = mdbx_thread_join(thr); - mdbx_txn_abort(txn); - -done: - mdbx_memalign_free(my.mc_wbuf[0]); - mdbx_condmutex_destroy(&my.mc_condmutex); - return rc ? rc : my.mc_error; -} - -/* Copy environment as-is. */ -static int __cold mdbx_env_copy_asis(MDBX_env *env, mdbx_filehandle_t fd) { - MDBX_txn *txn = NULL; - - /* Do the lock/unlock of the reader mutex before starting the - * write txn. Otherwise other read txns could block writers. */ - int rc = mdbx_txn_begin(env, NULL, MDBX_RDONLY, &txn); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - /* We must start the actual read txn after blocking writers */ - rc = mdbx_txn_end(txn, MDBX_END_RESET_TMP); - if (unlikely(rc != MDBX_SUCCESS)) - goto bailout; /* FIXME: or just return? */ - - /* Temporarily block writers until we snapshot the meta pages */ - rc = mdbx_txn_lock(env, false); - if (unlikely(rc != MDBX_SUCCESS)) - goto bailout; - - rc = mdbx_txn_renew0(txn, MDBX_RDONLY); - if (unlikely(rc != MDBX_SUCCESS)) { - mdbx_txn_unlock(env); - goto bailout; - } - - rc = mdbx_write(fd, env->me_map, pgno2bytes(env, NUM_METAS)); - MDBX_meta *const head = mdbx_meta_head(env); - const uint64_t size = - mdbx_roundup2(pgno2bytes(env, head->mm_geo.now), env->me_os_psize); - mdbx_txn_unlock(env); - - if (likely(rc == MDBX_SUCCESS)) - rc = mdbx_write(fd, env->me_map + pgno2bytes(env, NUM_METAS), - pgno2bytes(env, txn->mt_next_pgno - NUM_METAS)); - - if (likely(rc == MDBX_SUCCESS)) - rc = mdbx_ftruncate(fd, size); - -bailout: - mdbx_txn_abort(txn); - return rc; -} - -int __cold mdbx_env_copy2fd(MDBX_env *env, mdbx_filehandle_t fd, - unsigned flags) { - if (flags & MDBX_CP_COMPACT) - return mdbx_env_compact(env, fd); - - return mdbx_env_copy_asis(env, fd); -} - -int __cold mdbx_env_copy(MDBX_env *env, const char *path, unsigned flags) { - char *lck_pathname; - mdbx_filehandle_t newfd = INVALID_HANDLE_VALUE; - - if (env->me_flags & MDBX_NOSUBDIR) { - lck_pathname = (char *)path; - } else { - size_t len = strlen(path); - len += sizeof(MDBX_DATANAME); - lck_pathname = malloc(len); - if (!lck_pathname) - return MDBX_ENOMEM; - sprintf(lck_pathname, "%s" MDBX_DATANAME, path); - } - - /* The destination path must exist, but the destination file must not. - * We don't want the OS to cache the writes, since the source data is - * already in the OS cache. */ - int rc = - mdbx_openfile(lck_pathname, O_WRONLY | O_CREAT | O_EXCL, 0666, &newfd); - if (rc == MDBX_SUCCESS) { - if (env->me_psize >= env->me_os_psize) { -#ifdef F_NOCACHE /* __APPLE__ */ - (void)fcntl(newfd, F_NOCACHE, 1); -#elif defined(O_DIRECT) && defined(F_GETFL) - /* Set O_DIRECT if the file system supports it */ - if ((rc = fcntl(newfd, F_GETFL)) != -1) - (void)fcntl(newfd, F_SETFL, rc | O_DIRECT); -#endif - } - rc = mdbx_env_copy2fd(env, newfd, flags); - } - - if (!(env->me_flags & MDBX_NOSUBDIR)) - free(lck_pathname); - - if (newfd != INVALID_HANDLE_VALUE) { - int err = mdbx_closefile(newfd); - if (rc == MDBX_SUCCESS && err != rc) - rc = err; - } - - return rc; -} - -int __cold mdbx_env_set_flags(MDBX_env *env, unsigned flags, int onoff) { - if (unlikely(flags & ~CHANGEABLE)) - return MDBX_EINVAL; - - int rc = mdbx_txn_lock(env, false); - if (unlikely(rc)) - return rc; - - if (onoff) - env->me_flags |= flags; - else - env->me_flags &= ~flags; - - mdbx_txn_unlock(env); - return MDBX_SUCCESS; -} - -int __cold mdbx_env_get_flags(MDBX_env *env, unsigned *arg) { - if (unlikely(!env || !arg)) - return MDBX_EINVAL; - - *arg = env->me_flags & (CHANGEABLE | CHANGELESS); - return MDBX_SUCCESS; -} - -int __cold mdbx_env_set_userctx(MDBX_env *env, void *ctx) { - if (unlikely(!env)) - return MDBX_EINVAL; - env->me_userctx = ctx; - return MDBX_SUCCESS; -} - -void *__cold mdbx_env_get_userctx(MDBX_env *env) { - return env ? env->me_userctx : NULL; -} - -int __cold mdbx_env_set_assert(MDBX_env *env, MDBX_assert_func *func) { - if (unlikely(!env)) - return MDBX_EINVAL; -#if MDBX_DEBUG - env->me_assert_func = func; - return MDBX_SUCCESS; -#else - (void)func; - return MDBX_ENOSYS; -#endif -} - -int __cold mdbx_env_get_path(MDBX_env *env, const char **arg) { - if (unlikely(!env || !arg)) - return MDBX_EINVAL; - - *arg = env->me_path; - return MDBX_SUCCESS; -} - -int __cold mdbx_env_get_fd(MDBX_env *env, mdbx_filehandle_t *arg) { - if (unlikely(!env || !arg)) - return MDBX_EINVAL; - - *arg = env->me_fd; - return MDBX_SUCCESS; -} - -/* Common code for mdbx_dbi_stat() and mdbx_env_stat(). - * [in] env the environment to operate in. - * [in] db the MDBX_db record containing the stats to return. - * [out] arg the address of an MDBX_stat structure to receive the stats. - * Returns 0, this function always succeeds. */ -static int __cold mdbx_stat0(MDBX_env *env, MDBX_db *db, MDBX_stat *arg) { - arg->ms_psize = env->me_psize; - arg->ms_depth = db->md_depth; - arg->ms_branch_pages = db->md_branch_pages; - arg->ms_leaf_pages = db->md_leaf_pages; - arg->ms_overflow_pages = db->md_overflow_pages; - arg->ms_entries = db->md_entries; - return MDBX_SUCCESS; -} - -int __cold mdbx_env_stat(MDBX_env *env, MDBX_stat *arg, size_t bytes) { - MDBX_meta *meta; - - if (unlikely(env == NULL || arg == NULL)) - return MDBX_EINVAL; - if (unlikely(bytes != sizeof(MDBX_stat))) - return MDBX_EINVAL; - - meta = mdbx_meta_head(env); - return mdbx_stat0(env, &meta->mm_dbs[MAIN_DBI], arg); -} - -int __cold mdbx_env_info(MDBX_env *env, MDBX_envinfo *arg, size_t bytes) { - - if (unlikely(env == NULL || arg == NULL)) - return MDBX_EINVAL; - - if (bytes != sizeof(MDBX_envinfo)) - return MDBX_EINVAL; - - const MDBX_meta *const meta0 = METAPAGE(env, 0); - const MDBX_meta *const meta1 = METAPAGE(env, 1); - const MDBX_meta *const meta2 = METAPAGE(env, 2); - const MDBX_meta *meta; - do { - meta = mdbx_meta_head(env); - arg->mi_recent_txnid = mdbx_meta_txnid_fluid(env, meta); - arg->mi_meta0_txnid = mdbx_meta_txnid_fluid(env, meta0); - arg->mi_meta0_sign = meta0->mm_datasync_sign; - arg->mi_meta1_txnid = mdbx_meta_txnid_fluid(env, meta1); - arg->mi_meta1_sign = meta1->mm_datasync_sign; - arg->mi_meta2_txnid = mdbx_meta_txnid_fluid(env, meta2); - arg->mi_meta2_sign = meta2->mm_datasync_sign; - arg->mi_last_pgno = meta->mm_geo.next - 1; - arg->mi_geo.lower = pgno2bytes(env, meta->mm_geo.lower); - arg->mi_geo.upper = pgno2bytes(env, meta->mm_geo.upper); - arg->mi_geo.current = pgno2bytes(env, meta->mm_geo.now); - arg->mi_geo.shrink = pgno2bytes(env, meta->mm_geo.shrink); - arg->mi_geo.grow = pgno2bytes(env, meta->mm_geo.grow); - arg->mi_mapsize = env->me_mapsize; - mdbx_compiler_barrier(); - } while (unlikely(arg->mi_meta0_txnid != mdbx_meta_txnid_fluid(env, meta0) || - arg->mi_meta0_sign != meta0->mm_datasync_sign || - arg->mi_meta1_txnid != mdbx_meta_txnid_fluid(env, meta1) || - arg->mi_meta1_sign != meta1->mm_datasync_sign || - arg->mi_meta2_txnid != mdbx_meta_txnid_fluid(env, meta2) || - arg->mi_meta2_sign != meta2->mm_datasync_sign || - meta != mdbx_meta_head(env) || - arg->mi_recent_txnid != mdbx_meta_txnid_fluid(env, meta))); - - arg->mi_maxreaders = env->me_maxreaders; - arg->mi_numreaders = env->me_lck ? env->me_lck->mti_numreaders : INT32_MAX; - arg->mi_dxb_pagesize = env->me_psize; - arg->mi_sys_pagesize = env->me_os_psize; - - arg->mi_latter_reader_txnid = 0; - if (env->me_lck) { - MDBX_reader *r = env->me_lck->mti_readers; - arg->mi_latter_reader_txnid = arg->mi_recent_txnid; - for (unsigned i = 0; i < arg->mi_numreaders; ++i) { - if (r[i].mr_pid) { - const txnid_t txnid = r[i].mr_txnid; - if (arg->mi_latter_reader_txnid > txnid) - arg->mi_latter_reader_txnid = txnid; - } - } - } - - return MDBX_SUCCESS; -} - -static MDBX_cmp_func *mdbx_default_keycmp(unsigned flags) { - return (flags & MDBX_REVERSEKEY) ? mdbx_cmp_memnr : (flags & MDBX_INTEGERKEY) - ? mdbx_cmp_int_a2 - : mdbx_cmp_memn; -} - -static MDBX_cmp_func *mdbx_default_datacmp(unsigned flags) { - return !(flags & MDBX_DUPSORT) - ? 0 - : ((flags & MDBX_INTEGERDUP) - ? mdbx_cmp_int_ua - : ((flags & MDBX_REVERSEDUP) ? mdbx_cmp_memnr - : mdbx_cmp_memn)); -} - -static int mdbx_dbi_bind(MDBX_txn *txn, const MDBX_dbi dbi, unsigned user_flags, - MDBX_cmp_func *keycmp, MDBX_cmp_func *datacmp) { - /* LY: so, accepting only three cases for the table's flags: - * 1) user_flags and both comparators are zero - * = assume that a by-default mode/flags is requested for reading; - * 2) user_flags exactly the same - * = assume that the target mode/flags are requested properly; - * 3) user_flags differs, but table is empty and MDBX_CREATE is provided - * = assume that a properly create request with custom flags; - */ - if ((user_flags ^ txn->mt_dbs[dbi].md_flags) & PERSISTENT_FLAGS) { - /* flags ara differs, check other conditions */ - if (!user_flags && (!keycmp || keycmp == txn->mt_dbxs[dbi].md_cmp) && - (!datacmp || datacmp == txn->mt_dbxs[dbi].md_dcmp)) { - /* no comparators were provided and flags are zero, - * seems that is case #1 above */ - user_flags = txn->mt_dbs[dbi].md_flags; - } else if ((user_flags & MDBX_CREATE) && txn->mt_dbs[dbi].md_entries == 0) { - if (txn->mt_flags & MDBX_TXN_RDONLY) - return /* FIXME: return extended info */ MDBX_EACCESS; - /* make sure flags changes get committed */ - txn->mt_dbs[dbi].md_flags = user_flags & PERSISTENT_FLAGS; - txn->mt_flags |= MDBX_TXN_DIRTY; - } else { - return /* FIXME: return extended info */ MDBX_INCOMPATIBLE; - } - } - - if (!txn->mt_dbxs[dbi].md_cmp || MDBX_DEBUG) { - if (!keycmp) - keycmp = mdbx_default_keycmp(user_flags); - assert(!txn->mt_dbxs[dbi].md_cmp || txn->mt_dbxs[dbi].md_cmp == keycmp); - txn->mt_dbxs[dbi].md_cmp = keycmp; - } - - if (!txn->mt_dbxs[dbi].md_dcmp || MDBX_DEBUG) { - if (!datacmp) - datacmp = mdbx_default_datacmp(user_flags); - assert(!txn->mt_dbxs[dbi].md_dcmp || txn->mt_dbxs[dbi].md_dcmp == datacmp); - txn->mt_dbxs[dbi].md_dcmp = datacmp; - } - - return MDBX_SUCCESS; -} - -int mdbx_dbi_open_ex(MDBX_txn *txn, const char *table_name, unsigned user_flags, - MDBX_dbi *dbi, MDBX_cmp_func *keycmp, - MDBX_cmp_func *datacmp) { - if (unlikely(!txn || !dbi || (user_flags & ~VALID_FLAGS) != 0)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - /* main table? */ - if (!table_name) { - *dbi = MAIN_DBI; - return mdbx_dbi_bind(txn, MAIN_DBI, user_flags, keycmp, datacmp); - } - - if (txn->mt_dbxs[MAIN_DBI].md_cmp == NULL) { - txn->mt_dbxs[MAIN_DBI].md_cmp = - mdbx_default_keycmp(txn->mt_dbs[MAIN_DBI].md_flags); - txn->mt_dbxs[MAIN_DBI].md_dcmp = - mdbx_default_datacmp(txn->mt_dbs[MAIN_DBI].md_flags); - } - - /* Is the DB already open? */ - size_t len = strlen(table_name); - MDBX_dbi scan, slot = txn->mt_numdbs; - for (scan = txn->mt_numdbs; --scan >= CORE_DBS;) { - if (!txn->mt_dbxs[scan].md_name.iov_len) { - /* Remember this free slot */ - slot = scan; - continue; - } - if (len == txn->mt_dbxs[scan].md_name.iov_len && - !strncmp(table_name, txn->mt_dbxs[scan].md_name.iov_base, len)) { - *dbi = scan; - return mdbx_dbi_bind(txn, scan, user_flags, keycmp, datacmp); - } - } - - /* Fail, if no free slot and max hit */ - MDBX_env *env = txn->mt_env; - if (unlikely(slot >= env->me_maxdbs)) - return MDBX_DBS_FULL; - - /* Cannot mix named table with some main-table flags */ - if (unlikely(txn->mt_dbs[MAIN_DBI].md_flags & - (MDBX_DUPSORT | MDBX_INTEGERKEY))) - return (user_flags & MDBX_CREATE) ? MDBX_INCOMPATIBLE : MDBX_NOTFOUND; - - /* Find the DB info */ - int exact = 0; - MDBX_val key, data; - key.iov_len = len; - key.iov_base = (void *)table_name; - MDBX_cursor mc; - mdbx_cursor_init(&mc, txn, MAIN_DBI, NULL); - int rc = mdbx_cursor_set(&mc, &key, &data, MDBX_SET, &exact); - if (unlikely(rc != MDBX_SUCCESS)) { - if (rc != MDBX_NOTFOUND || !(user_flags & MDBX_CREATE)) - return rc; - } else { - /* make sure this is actually a table */ - MDBX_node *node = NODEPTR(mc.mc_pg[mc.mc_top], mc.mc_ki[mc.mc_top]); - if (unlikely((node->mn_flags & (F_DUPDATA | F_SUBDATA)) != F_SUBDATA)) - return MDBX_INCOMPATIBLE; - } - - if (rc != MDBX_SUCCESS && unlikely(txn->mt_flags & MDBX_TXN_RDONLY)) - return MDBX_EACCESS; - - /* Done here so we cannot fail after creating a new DB */ - char *namedup = mdbx_strdup(table_name); - if (unlikely(!namedup)) - return MDBX_ENOMEM; - - int err = mdbx_fastmutex_acquire(&env->me_dbi_lock); - if (unlikely(err != MDBX_SUCCESS)) { - free(namedup); - return err; - } - - unsigned dbflag = DB_NEW | DB_VALID | DB_USRVALID; - if (unlikely(rc)) { - /* MDBX_NOTFOUND and MDBX_CREATE: Create new DB */ - assert(rc == MDBX_NOTFOUND); - MDBX_db db_dummy; - memset(&db_dummy, 0, sizeof(db_dummy)); - db_dummy.md_root = P_INVALID; - db_dummy.md_flags = user_flags & PERSISTENT_FLAGS; - data.iov_len = sizeof(db_dummy); - data.iov_base = &db_dummy; - WITH_CURSOR_TRACKING( - mc, - rc = mdbx_cursor_put(&mc, &key, &data, F_SUBDATA | MDBX_NOOVERWRITE)); - - if (unlikely(rc != MDBX_SUCCESS)) - goto bailout; - - dbflag |= DB_DIRTY; - } - - /* Got info, register DBI in this txn */ - txn->mt_dbxs[slot].md_name.iov_base = namedup; - txn->mt_dbxs[slot].md_name.iov_len = len; - txn->mt_dbxs[slot].md_cmp = nullptr; - txn->mt_dbxs[slot].md_dcmp = nullptr; - txn->mt_dbflags[slot] = (uint8_t)dbflag; - txn->mt_dbiseqs[slot] = (env->me_dbiseqs[slot] += 1); - - txn->mt_dbs[slot] = *(MDBX_db *)data.iov_base; - rc = mdbx_dbi_bind(txn, slot, user_flags, keycmp, datacmp); - if (unlikely(rc != MDBX_SUCCESS)) { - assert((dbflag & DB_DIRTY) == 0); - /* cleanup slot */ - txn->mt_dbxs[slot].md_name.iov_base = NULL; - txn->mt_dbxs[slot].md_name.iov_len = 0; - txn->mt_dbflags[slot] = 0; - bailout: - free(namedup); - } else { - *dbi = slot; - if (slot == txn->mt_numdbs) - txn->mt_numdbs++; - } - - mdbx_ensure(env, mdbx_fastmutex_release(&env->me_dbi_lock) == MDBX_SUCCESS); - return rc; -} - -int mdbx_dbi_open(MDBX_txn *txn, const char *table_name, unsigned table_flags, - MDBX_dbi *dbi) { - return mdbx_dbi_open_ex(txn, table_name, table_flags, dbi, nullptr, nullptr); -} - -int __cold mdbx_dbi_stat(MDBX_txn *txn, MDBX_dbi dbi, MDBX_stat *arg, - size_t bytes) { - if (unlikely(!arg || !txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_VALID))) - return MDBX_EINVAL; - - if (unlikely(bytes != sizeof(MDBX_stat))) - return MDBX_EINVAL; - - if (unlikely(txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - if (unlikely(txn->mt_dbflags[dbi] & DB_STALE)) { - MDBX_cursor mc; - MDBX_xcursor mx; - /* Stale, must read the DB's root. cursor_init does it for us. */ - mdbx_cursor_init(&mc, txn, dbi, &mx); - } - return mdbx_stat0(txn->mt_env, &txn->mt_dbs[dbi], arg); -} - -static int mdbx_dbi_close_locked(MDBX_env *env, MDBX_dbi dbi) { - if (unlikely(dbi < CORE_DBS || dbi >= env->me_maxdbs)) - return MDBX_EINVAL; - - char *ptr = env->me_dbxs[dbi].md_name.iov_base; - /* If there was no name, this was already closed */ - if (unlikely(!ptr)) - return MDBX_BAD_DBI; - - env->me_dbxs[dbi].md_name.iov_base = NULL; - env->me_dbxs[dbi].md_name.iov_len = 0; - env->me_dbflags[dbi] = 0; - env->me_dbiseqs[dbi]++; - free(ptr); - return MDBX_SUCCESS; -} - -int mdbx_dbi_close(MDBX_env *env, MDBX_dbi dbi) { - if (unlikely(dbi < CORE_DBS || dbi >= env->me_maxdbs)) - return MDBX_EINVAL; - - int rc = mdbx_fastmutex_acquire(&env->me_dbi_lock); - if (likely(rc == MDBX_SUCCESS)) { - rc = mdbx_dbi_close_locked(env, dbi); - mdbx_ensure(env, mdbx_fastmutex_release(&env->me_dbi_lock) == MDBX_SUCCESS); - } - return rc; -} - -int mdbx_dbi_flags_ex(MDBX_txn *txn, MDBX_dbi dbi, unsigned *flags, - unsigned *state) { - if (unlikely(!txn || !flags || !state)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_VALID))) - return MDBX_EINVAL; - - *flags = txn->mt_dbs[dbi].md_flags & PERSISTENT_FLAGS; - *state = txn->mt_dbflags[dbi] & (DB_NEW | DB_DIRTY | DB_STALE); - - return MDBX_SUCCESS; -} - -int mdbx_dbi_flags(MDBX_txn *txn, MDBX_dbi dbi, unsigned *flags) { - unsigned state; - return mdbx_dbi_flags_ex(txn, dbi, flags, &state); -} - -/* Add all the DB's pages to the free list. - * [in] mc Cursor on the DB to free. - * [in] subs non-Zero to check for sub-DBs in this DB. - * Returns 0 on success, non-zero on failure. */ -static int mdbx_drop0(MDBX_cursor *mc, int subs) { - int rc; - - rc = mdbx_page_search(mc, NULL, MDBX_PS_FIRST); - if (likely(rc == MDBX_SUCCESS)) { - MDBX_txn *txn = mc->mc_txn; - MDBX_node *ni; - MDBX_cursor mx; - unsigned i; - - /* DUPSORT sub-DBs have no ovpages/DBs. Omit scanning leaves. - * This also avoids any P_LEAF2 pages, which have no nodes. - * Also if the DB doesn't have sub-DBs and has no overflow - * pages, omit scanning leaves. */ - if ((mc->mc_flags & C_SUB) || (!subs && !mc->mc_db->md_overflow_pages)) - mdbx_cursor_pop(mc); - - mdbx_cursor_copy(mc, &mx); - while (mc->mc_snum > 0) { - MDBX_page *mp = mc->mc_pg[mc->mc_top]; - unsigned n = NUMKEYS(mp); - if (IS_LEAF(mp)) { - for (i = 0; i < n; i++) { - ni = NODEPTR(mp, i); - if (ni->mn_flags & F_BIGDATA) { - MDBX_page *omp; - pgno_t pg; - memcpy(&pg, NODEDATA(ni), sizeof(pg)); - rc = mdbx_page_get(mc, pg, &omp, NULL); - if (unlikely(rc)) - goto done; - mdbx_cassert(mc, IS_OVERFLOW(omp)); - rc = - mdbx_pnl_append_range(&txn->mt_befree_pages, pg, omp->mp_pages); - if (unlikely(rc)) - goto done; - mc->mc_db->md_overflow_pages -= omp->mp_pages; - if (!mc->mc_db->md_overflow_pages && !subs) - break; - } else if (subs && (ni->mn_flags & F_SUBDATA)) { - mdbx_xcursor_init1(mc, ni); - rc = mdbx_drop0(&mc->mc_xcursor->mx_cursor, 0); - if (unlikely(rc)) - goto done; - } - } - if (!subs && !mc->mc_db->md_overflow_pages) - goto pop; - } else { - if (unlikely((rc = mdbx_pnl_need(&txn->mt_befree_pages, n)) != 0)) - goto done; - for (i = 0; i < n; i++) { - pgno_t pg; - ni = NODEPTR(mp, i); - pg = NODEPGNO(ni); - /* free it */ - mdbx_pnl_xappend(txn->mt_befree_pages, pg); - } - } - if (!mc->mc_top) - break; - mdbx_cassert(mc, i <= UINT16_MAX); - mc->mc_ki[mc->mc_top] = (indx_t)i; - rc = mdbx_cursor_sibling(mc, 1); - if (rc) { - if (unlikely(rc != MDBX_NOTFOUND)) - goto done; - /* no more siblings, go back to beginning - * of previous level. */ - pop: - mdbx_cursor_pop(mc); - mc->mc_ki[0] = 0; - for (i = 1; i < mc->mc_snum; i++) { - mc->mc_ki[i] = 0; - mc->mc_pg[i] = mx.mc_pg[i]; - } - } - } - /* free it */ - rc = mdbx_pnl_append(&txn->mt_befree_pages, mc->mc_db->md_root); - done: - if (unlikely(rc)) - txn->mt_flags |= MDBX_TXN_ERROR; - } else if (rc == MDBX_NOTFOUND) { - rc = MDBX_SUCCESS; - } - mc->mc_flags &= ~C_INITIALIZED; - return rc; -} - -int mdbx_drop(MDBX_txn *txn, MDBX_dbi dbi, int del) { - if (unlikely(1 < (unsigned)del || !txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - if (unlikely(TXN_DBI_CHANGED(txn, dbi))) - return MDBX_BAD_DBI; - - if (unlikely(F_ISSET(txn->mt_flags, MDBX_TXN_RDONLY))) - return MDBX_EACCESS; - - MDBX_cursor *mc; - int rc = mdbx_cursor_open(txn, dbi, &mc); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - MDBX_env *env = txn->mt_env; - rc = mdbx_fastmutex_acquire(&env->me_dbi_lock); - if (unlikely(rc != MDBX_SUCCESS)) { - mdbx_cursor_close(mc); - return rc; - } - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) { - rc = MDBX_EINVAL; - goto bailout; - } - - if (unlikely(TXN_DBI_CHANGED(txn, dbi))) { - rc = MDBX_BAD_DBI; - goto bailout; - } - - rc = mdbx_drop0(mc, mc->mc_db->md_flags & MDBX_DUPSORT); - /* Invalidate the dropped DB's cursors */ - for (MDBX_cursor *m2 = txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) - m2->mc_flags &= ~(C_INITIALIZED | C_EOF); - if (unlikely(rc)) - goto bailout; - - /* Can't delete the main DB */ - if (del && dbi >= CORE_DBS) { - rc = mdbx_del0(txn, MAIN_DBI, &mc->mc_dbx->md_name, NULL, F_SUBDATA); - if (likely(!rc)) { - txn->mt_dbflags[dbi] = DB_STALE; - mdbx_dbi_close_locked(env, dbi); - } else { - txn->mt_flags |= MDBX_TXN_ERROR; - } - } else { - /* reset the DB record, mark it dirty */ - txn->mt_dbflags[dbi] |= DB_DIRTY; - txn->mt_dbs[dbi].md_depth = 0; - txn->mt_dbs[dbi].md_branch_pages = 0; - txn->mt_dbs[dbi].md_leaf_pages = 0; - txn->mt_dbs[dbi].md_overflow_pages = 0; - txn->mt_dbs[dbi].md_entries = 0; - txn->mt_dbs[dbi].md_root = P_INVALID; - txn->mt_dbs[dbi].md_seq = 0; - - txn->mt_flags |= MDBX_TXN_DIRTY; - } - -bailout: - mdbx_cursor_close(mc); - mdbx_ensure(env, mdbx_fastmutex_release(&env->me_dbi_lock) == MDBX_SUCCESS); - return rc; -} - -int mdbx_set_compare(MDBX_txn *txn, MDBX_dbi dbi, MDBX_cmp_func *cmp) { - if (unlikely(!txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - txn->mt_dbxs[dbi].md_cmp = cmp; - return MDBX_SUCCESS; -} - -int mdbx_set_dupsort(MDBX_txn *txn, MDBX_dbi dbi, MDBX_cmp_func *cmp) { - if (unlikely(!txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - txn->mt_dbxs[dbi].md_dcmp = cmp; - return MDBX_SUCCESS; -} - -int __cold mdbx_reader_list(MDBX_env *env, MDBX_msg_func *func, void *ctx) { - char buf[64]; - int rc = 0, first = 1; - - if (unlikely(!env || !func)) - return -MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - const MDBX_lockinfo *const lck = env->me_lck; - const unsigned snap_nreaders = lck->mti_numreaders; - for (unsigned i = 0; i < snap_nreaders; i++) { - if (lck->mti_readers[i].mr_pid) { - const txnid_t txnid = lck->mti_readers[i].mr_txnid; - if (txnid == ~(txnid_t)0) - snprintf(buf, sizeof(buf), "%10" PRIuPTR " %" PRIxPTR " -\n", - (uintptr_t)lck->mti_readers[i].mr_pid, - (uintptr_t)lck->mti_readers[i].mr_tid); - else - snprintf(buf, sizeof(buf), "%10" PRIuPTR " %" PRIxPTR " %" PRIaTXN "\n", - (uintptr_t)lck->mti_readers[i].mr_pid, - (uintptr_t)lck->mti_readers[i].mr_tid, txnid); - - if (first) { - first = 0; - rc = func(" pid thread txnid\n", ctx); - if (rc < 0) - break; - } - rc = func(buf, ctx); - if (rc < 0) - break; - } - } - if (first) - rc = func("(no active readers)\n", ctx); - - return rc; -} - -/* Insert pid into list if not already present. - * return -1 if already present. */ -static int __cold mdbx_pid_insert(mdbx_pid_t *ids, mdbx_pid_t pid) { - /* binary search of pid in list */ - unsigned base = 0; - unsigned cursor = 1; - int val = 0; - unsigned n = ids[0]; - - while (n > 0) { - unsigned pivot = n >> 1; - cursor = base + pivot + 1; - val = pid - ids[cursor]; - - if (val < 0) { - n = pivot; - } else if (val > 0) { - base = cursor; - n -= pivot + 1; - } else { - /* found, so it's a duplicate */ - return -1; - } - } - - if (val > 0) - ++cursor; - - ids[0]++; - for (n = ids[0]; n > cursor; n--) - ids[n] = ids[n - 1]; - ids[n] = pid; - return 0; -} - -int __cold mdbx_reader_check(MDBX_env *env, int *dead) { - if (unlikely(!env || env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EINVAL; - if (dead) - *dead = 0; - return mdbx_reader_check0(env, false, dead); -} - -/* Return: - * MDBX_RESULT_TRUE - done and mutex recovered - * MDBX_SUCCESS - done - * Otherwise errcode. */ -int __cold mdbx_reader_check0(MDBX_env *env, int rdt_locked, int *dead) { - assert(rdt_locked >= 0); - - if (unlikely(env->me_pid != mdbx_getpid())) { - env->me_flags |= MDBX_FATAL_ERROR; - return MDBX_PANIC; - } - - MDBX_lockinfo *const lck = env->me_lck; - const unsigned snap_nreaders = lck->mti_numreaders; - mdbx_pid_t *pids = alloca((snap_nreaders + 1) * sizeof(mdbx_pid_t)); - pids[0] = 0; - - int rc = MDBX_SUCCESS, count = 0; - for (unsigned i = 0; i < snap_nreaders; i++) { - const mdbx_pid_t pid = lck->mti_readers[i].mr_pid; - if (pid == 0) - continue /* skip empty */; - if (pid == env->me_pid) - continue /* skip self */; - if (mdbx_pid_insert(pids, pid) != 0) - continue /* such pid already processed */; - - int err = mdbx_rpid_check(env, pid); - if (err == MDBX_RESULT_TRUE) - continue /* reader is live */; - - if (err != MDBX_SUCCESS) { - rc = err; - break /* mdbx_rpid_check() failed */; - } - - /* stale reader found */ - if (!rdt_locked) { - err = mdbx_rdt_lock(env); - if (MDBX_IS_ERROR(err)) { - rc = err; - break; - } - - rdt_locked = -1; - if (err == MDBX_RESULT_TRUE) { - /* mutex recovered, the mdbx_mutex_failed() checked all readers */ - rc = MDBX_RESULT_TRUE; - break; - } - - /* a other process may have clean and reused slot, recheck */ - if (lck->mti_readers[i].mr_pid != pid) - continue; - - err = mdbx_rpid_check(env, pid); - if (MDBX_IS_ERROR(err)) { - rc = err; - break; - } - - if (err != MDBX_SUCCESS) - continue /* the race with other process, slot reused */; - } - - /* clean it */ - for (unsigned j = i; j < snap_nreaders; j++) { - if (lck->mti_readers[j].mr_pid == pid) { - mdbx_debug("clear stale reader pid %" PRIuPTR " txn %" PRIaTXN "", - (size_t)pid, lck->mti_readers[j].mr_txnid); - lck->mti_readers[j].mr_pid = 0; - lck->mti_readers_refresh_flag = true; - count++; - } - } - } - - if (rdt_locked < 0) - mdbx_rdt_unlock(env); - - if (dead) - *dead = count; - return rc; -} - -int __cold mdbx_setup_debug(int flags, MDBX_debug_func *logger) { - unsigned ret = mdbx_runtime_flags; - mdbx_runtime_flags = flags; - -#ifdef __linux__ - if (flags & MDBX_DBG_DUMP) { - int core_filter_fd = open("/proc/self/coredump_filter", O_TRUNC | O_RDWR); - if (core_filter_fd >= 0) { - char buf[32]; - const unsigned r = pread(core_filter_fd, buf, sizeof(buf), 0); - if (r > 0 && r < sizeof(buf)) { - buf[r] = 0; - unsigned long mask = strtoul(buf, NULL, 16); - if (mask != ULONG_MAX) { - mask |= 1 << 3 /* Dump file-backed shared mappings */; - mask |= 1 << 6 /* Dump shared huge pages */; - mask |= 1 << 8 /* Dump shared DAX pages */; - unsigned w = snprintf(buf, sizeof(buf), "0x%lx\n", mask); - if (w > 0 && w < sizeof(buf)) { - w = pwrite(core_filter_fd, buf, w, 0); - (void)w; - } - } - } - close(core_filter_fd); - } - } -#endif /* __linux__ */ - - mdbx_debug_logger = logger; - return ret; -} - -static txnid_t __cold mdbx_oomkick(MDBX_env *env, const txnid_t laggard) { - mdbx_debug("DB size maxed out"); - - int retry; - for (retry = 0; retry < INT_MAX; ++retry) { - txnid_t oldest = mdbx_reclaiming_detent(env); - mdbx_assert(env, oldest < env->me_txn0->mt_txnid); - mdbx_assert(env, oldest >= laggard); - mdbx_assert(env, oldest >= env->me_oldest[0]); - if (oldest == laggard) - return oldest; - - if (MDBX_IS_ERROR(mdbx_reader_check0(env, false, NULL))) - break; - - MDBX_reader *const rtbl = env->me_lck->mti_readers; - MDBX_reader *asleep = nullptr; - for (int i = env->me_lck->mti_numreaders; --i >= 0;) { - if (rtbl[i].mr_pid) { - mdbx_jitter4testing(true); - const txnid_t snap = rtbl[i].mr_txnid; - if (oldest > snap && laggard <= /* ignore pending updates */ snap) { - oldest = snap; - asleep = &rtbl[i]; - } - } - } - - if (laggard < oldest || !asleep) { - if (retry && env->me_oom_func) { - /* LY: notify end of oom-loop */ - const txnid_t gap = oldest - laggard; - env->me_oom_func(env, 0, 0, laggard, - (gap < UINT_MAX) ? (unsigned)gap : UINT_MAX, -retry); - } - mdbx_notice("oom-kick: update oldest %" PRIaTXN " -> %" PRIaTXN, - env->me_oldest[0], oldest); - mdbx_assert(env, env->me_oldest[0] <= oldest); - return env->me_oldest[0] = oldest; - } - - mdbx_tid_t tid; - mdbx_pid_t pid; - int rc; - - if (!env->me_oom_func) - break; - - pid = asleep->mr_pid; - tid = asleep->mr_tid; - if (asleep->mr_txnid != laggard || pid <= 0) - continue; - - const txnid_t gap = - mdbx_meta_txnid_stable(env, mdbx_meta_head(env)) - laggard; - rc = env->me_oom_func(env, pid, tid, laggard, - (gap < UINT_MAX) ? (unsigned)gap : UINT_MAX, retry); - if (rc < 0) - break; - - if (rc) { - asleep->mr_txnid = ~(txnid_t)0; - env->me_lck->mti_readers_refresh_flag = true; - if (rc > 1) { - asleep->mr_tid = 0; - asleep->mr_pid = 0; - mdbx_coherent_barrier(); - } - } - } - - if (retry && env->me_oom_func) { - /* LY: notify end of oom-loop */ - env->me_oom_func(env, 0, 0, laggard, 0, -retry); - } - return mdbx_find_oldest(env->me_txn); -} - -int __cold mdbx_env_set_syncbytes(MDBX_env *env, size_t bytes) { - if (unlikely(!env)) - return MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - env->me_sync_threshold = bytes; - return env->me_map ? mdbx_env_sync(env, 0) : MDBX_SUCCESS; -} - -int __cold mdbx_env_set_oomfunc(MDBX_env *env, MDBX_oom_func *oomfunc) { - if (unlikely(!env)) - return MDBX_EINVAL; - - if (unlikely(env->me_signature != MDBX_ME_SIGNATURE)) - return MDBX_EBADSIGN; - - env->me_oom_func = oomfunc; - return MDBX_SUCCESS; -} - -MDBX_oom_func *__cold mdbx_env_get_oomfunc(MDBX_env *env) { - return likely(env && env->me_signature == MDBX_ME_SIGNATURE) - ? env->me_oom_func - : NULL; -} - -#ifdef __SANITIZE_THREAD__ -/* LY: avoid tsan-trap by me_txn, mm_last_pg and mt_next_pgno */ -__attribute__((no_sanitize_thread, noinline)) -#endif -int mdbx_txn_straggler(MDBX_txn *txn, int *percent) -{ - if (unlikely(!txn)) - return -MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - MDBX_env *env = txn->mt_env; - if (unlikely((txn->mt_flags & MDBX_RDONLY) == 0)) { - if (percent) - *percent = - (int)((txn->mt_next_pgno * UINT64_C(100) + txn->mt_end_pgno / 2) / - txn->mt_end_pgno); - return -1; - } - - txnid_t recent; - MDBX_meta *meta; - do { - meta = mdbx_meta_head(env); - recent = mdbx_meta_txnid_fluid(env, meta); - if (percent) { - const pgno_t maxpg = meta->mm_geo.now; - *percent = (int)((meta->mm_geo.next * UINT64_C(100) + maxpg / 2) / maxpg); - } - } while (unlikely(recent != mdbx_meta_txnid_fluid(env, meta))); - - txnid_t lag = recent - txn->mt_ro_reader->mr_txnid; - return (lag > INT_MAX) ? INT_MAX : (int)lag; -} - -typedef struct mdbx_walk_ctx { - MDBX_txn *mw_txn; - void *mw_user; - MDBX_pgvisitor_func *mw_visitor; -} mdbx_walk_ctx_t; - -/* Depth-first tree traversal. */ -static int __cold mdbx_env_walk(mdbx_walk_ctx_t *ctx, const char *dbi, - pgno_t pg, int deep) { - MDBX_page *mp; - int rc, i, nkeys; - size_t header_size, unused_size, payload_size, align_bytes; - const char *type; - - if (pg == P_INVALID) - return MDBX_SUCCESS; /* empty db */ - - MDBX_cursor mc; - memset(&mc, 0, sizeof(mc)); - mc.mc_snum = 1; - mc.mc_txn = ctx->mw_txn; - - rc = mdbx_page_get(&mc, pg, &mp, NULL); - if (rc) - return rc; - if (pg != mp->mp_pgno) - return MDBX_CORRUPTED; - - nkeys = NUMKEYS(mp); - header_size = IS_LEAF2(mp) ? PAGEHDRSZ : PAGEHDRSZ + mp->mp_lower; - unused_size = SIZELEFT(mp); - payload_size = 0; - - /* LY: Don't use mask here, e.g bitwise - * (P_BRANCH|P_LEAF|P_LEAF2|P_META|P_OVERFLOW|P_SUBP). - * Pages should not me marked dirty/loose or otherwise. */ - switch (mp->mp_flags) { - case P_BRANCH: - type = "branch"; - if (nkeys < 1) - return MDBX_CORRUPTED; - break; - case P_LEAF: - type = "leaf"; - break; - case P_LEAF | P_SUBP: - type = "dupsort-subleaf"; - break; - case P_LEAF | P_LEAF2: - type = "dupfixed-leaf"; - break; - case P_LEAF | P_LEAF2 | P_SUBP: - type = "dupsort-dupfixed-subleaf"; - break; - case P_META: - case P_OVERFLOW: - __fallthrough; - default: - return MDBX_CORRUPTED; - } - - for (align_bytes = i = 0; i < nkeys; - align_bytes += ((payload_size + align_bytes) & 1), i++) { - MDBX_node *node; - - if (IS_LEAF2(mp)) { - /* LEAF2 pages have no mp_ptrs[] or node headers */ - payload_size += mp->mp_leaf2_ksize; - continue; - } - - node = NODEPTR(mp, i); - payload_size += NODESIZE + node->mn_ksize; - - if (IS_BRANCH(mp)) { - rc = mdbx_env_walk(ctx, dbi, NODEPGNO(node), deep); - if (rc) - return rc; - continue; - } - - assert(IS_LEAF(mp)); - if (node->mn_flags & F_BIGDATA) { - MDBX_page *omp; - pgno_t *opg; - size_t over_header, over_payload, over_unused; - - payload_size += sizeof(pgno_t); - opg = NODEDATA(node); - rc = mdbx_page_get(&mc, *opg, &omp, NULL); - if (rc) - return rc; - if (*opg != omp->mp_pgno) - return MDBX_CORRUPTED; - /* LY: Don't use mask here, e.g bitwise - * (P_BRANCH|P_LEAF|P_LEAF2|P_META|P_OVERFLOW|P_SUBP). - * Pages should not me marked dirty/loose or otherwise. */ - if (P_OVERFLOW != omp->mp_flags) - return MDBX_CORRUPTED; - - over_header = PAGEHDRSZ; - over_payload = NODEDSZ(node); - over_unused = pgno2bytes(ctx->mw_txn->mt_env, omp->mp_pages) - - over_payload - over_header; - - rc = ctx->mw_visitor(*opg, omp->mp_pages, ctx->mw_user, dbi, - "overflow-data", 1, over_payload, over_header, - over_unused); - if (rc) - return rc; - continue; - } - - payload_size += NODEDSZ(node); - if (node->mn_flags & F_SUBDATA) { - MDBX_db *db = NODEDATA(node); - char *name = NULL; - - if (!(node->mn_flags & F_DUPDATA)) { - name = NODEKEY(node); - ptrdiff_t namelen = (char *)db - name; - name = memcpy(alloca(namelen + 1), name, namelen); - name[namelen] = 0; - } - rc = mdbx_env_walk(ctx, (name && name[0]) ? name : dbi, db->md_root, - deep + 1); - if (rc) - return rc; - } - } - - return ctx->mw_visitor(mp->mp_pgno, 1, ctx->mw_user, dbi, type, nkeys, - payload_size, header_size, unused_size + align_bytes); -} - -int __cold mdbx_env_pgwalk(MDBX_txn *txn, MDBX_pgvisitor_func *visitor, - void *user) { - if (unlikely(!txn)) - return MDBX_BAD_TXN; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - mdbx_walk_ctx_t ctx; - ctx.mw_txn = txn; - ctx.mw_user = user; - ctx.mw_visitor = visitor; - - int rc = visitor(0, NUM_METAS, user, "meta", "meta", NUM_METAS, - sizeof(MDBX_meta) * NUM_METAS, PAGEHDRSZ * NUM_METAS, - (txn->mt_env->me_psize - sizeof(MDBX_meta) - PAGEHDRSZ) * - NUM_METAS); - if (!rc) - rc = mdbx_env_walk(&ctx, "free", txn->mt_dbs[FREE_DBI].md_root, 0); - if (!rc) - rc = mdbx_env_walk(&ctx, "main", txn->mt_dbs[MAIN_DBI].md_root, 0); - if (!rc) - rc = visitor(P_INVALID, 0, user, NULL, NULL, 0, 0, 0, 0); - return rc; -} - -int mdbx_canary_put(MDBX_txn *txn, const mdbx_canary *canary) { - if (unlikely(!txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - if (unlikely(F_ISSET(txn->mt_flags, MDBX_TXN_RDONLY))) - return MDBX_EACCESS; - - if (likely(canary)) { - if (txn->mt_canary.x == canary->x && txn->mt_canary.y == canary->y && - txn->mt_canary.z == canary->z) - return MDBX_SUCCESS; - txn->mt_canary.x = canary->x; - txn->mt_canary.y = canary->y; - txn->mt_canary.z = canary->z; - } - txn->mt_canary.v = txn->mt_txnid; - - if ((txn->mt_flags & MDBX_TXN_DIRTY) == 0) { - MDBX_env *env = txn->mt_env; - txn->mt_flags |= MDBX_TXN_DIRTY; - env->me_sync_pending += env->me_psize; - } - - return MDBX_SUCCESS; -} - -int mdbx_canary_get(MDBX_txn *txn, mdbx_canary *canary) { - if (unlikely(txn == NULL || canary == NULL)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - *canary = txn->mt_canary; - return MDBX_SUCCESS; -} - -int mdbx_cursor_on_first(MDBX_cursor *mc) { - if (unlikely(mc == NULL)) - return MDBX_EINVAL; - - if (unlikely(mc->mc_signature != MDBX_MC_SIGNATURE)) - return MDBX_EBADSIGN; - - if (!(mc->mc_flags & C_INITIALIZED)) - return MDBX_RESULT_FALSE; - - for (unsigned i = 0; i < mc->mc_snum; ++i) { - if (mc->mc_ki[i]) - return MDBX_RESULT_FALSE; - } - - return MDBX_RESULT_TRUE; -} - -int mdbx_cursor_on_last(MDBX_cursor *mc) { - if (unlikely(mc == NULL)) - return MDBX_EINVAL; - - if (unlikely(mc->mc_signature != MDBX_MC_SIGNATURE)) - return MDBX_EBADSIGN; - - if (!(mc->mc_flags & C_INITIALIZED)) - return MDBX_RESULT_FALSE; - - for (unsigned i = 0; i < mc->mc_snum; ++i) { - unsigned nkeys = NUMKEYS(mc->mc_pg[i]); - if (mc->mc_ki[i] < nkeys - 1) - return MDBX_RESULT_FALSE; - } - - return MDBX_RESULT_TRUE; -} - -int mdbx_cursor_eof(MDBX_cursor *mc) { - if (unlikely(mc == NULL)) - return MDBX_EINVAL; - - if (unlikely(mc->mc_signature != MDBX_MC_SIGNATURE)) - return MDBX_EBADSIGN; - - if ((mc->mc_flags & C_INITIALIZED) == 0) - return MDBX_RESULT_TRUE; - - if (mc->mc_snum == 0) - return MDBX_RESULT_TRUE; - - if ((mc->mc_flags & C_EOF) && - mc->mc_ki[mc->mc_top] >= NUMKEYS(mc->mc_pg[mc->mc_top])) - return MDBX_RESULT_TRUE; - - return MDBX_RESULT_FALSE; -} - -static int mdbx_is_samedata(const MDBX_val *a, const MDBX_val *b) { - return a->iov_len == b->iov_len && - memcmp(a->iov_base, b->iov_base, a->iov_len) == 0; -} - -/* Позволяет обновить или удалить существующую запись с получением - * в old_data предыдущего значения данных. При этом если new_data равен - * нулю, то выполняется удаление, иначе обновление/вставка. - * - * Текущее значение может находиться в уже измененной (грязной) странице. - * В этом случае страница будет перезаписана при обновлении, а само старое - * значение утрачено. Поэтому исходно в old_data должен быть передан - * дополнительный буфер для копирования старого значения. - * Если переданный буфер слишком мал, то функция вернет -1, установив - * old_data->iov_len в соответствующее значение. - * - * Для не-уникальных ключей также возможен второй сценарий использования, - * когда посредством old_data из записей с одинаковым ключом для - * удаления/обновления выбирается конкретная. Для выбора этого сценария - * во flags следует одновременно указать MDBX_CURRENT и MDBX_NOOVERWRITE. - * Именно эта комбинация выбрана, так как она лишена смысла, и этим позволяет - * идентифицировать запрос такого сценария. - * - * Функция может быть замещена соответствующими операциями с курсорами - * после двух доработок (TODO): - * - внешняя аллокация курсоров, в том числе на стеке (без malloc). - * - получения статуса страницы по адресу (знать о P_DIRTY). - */ -int mdbx_replace(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *new_data, - MDBX_val *old_data, unsigned flags) { - if (unlikely(!key || !old_data || !txn || old_data == new_data)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(old_data->iov_base == NULL && old_data->iov_len)) - return MDBX_EINVAL; - - if (unlikely(new_data == NULL && !(flags & MDBX_CURRENT))) - return MDBX_EINVAL; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - if (unlikely(flags & - ~(MDBX_NOOVERWRITE | MDBX_NODUPDATA | MDBX_RESERVE | - MDBX_APPEND | MDBX_APPENDDUP | MDBX_CURRENT))) - return MDBX_EINVAL; - - if (unlikely(txn->mt_flags & (MDBX_TXN_RDONLY | MDBX_TXN_BLOCKED))) - return (txn->mt_flags & MDBX_TXN_RDONLY) ? MDBX_EACCESS : MDBX_BAD_TXN; - - MDBX_cursor mc; - MDBX_xcursor mx; - mdbx_cursor_init(&mc, txn, dbi, &mx); - mc.mc_next = txn->mt_cursors[dbi]; - txn->mt_cursors[dbi] = &mc; - - int rc; - MDBX_val present_key = *key; - if (F_ISSET(flags, MDBX_CURRENT | MDBX_NOOVERWRITE)) { - /* в old_data значение для выбора конкретного дубликата */ - if (unlikely(!(txn->mt_dbs[dbi].md_flags & MDBX_DUPSORT))) { - rc = MDBX_EINVAL; - goto bailout; - } - - /* убираем лишний бит, он был признаком запрошенного режима */ - flags -= MDBX_NOOVERWRITE; - - rc = mdbx_cursor_get(&mc, &present_key, old_data, MDBX_GET_BOTH); - if (rc != MDBX_SUCCESS) - goto bailout; - - if (new_data) { - /* обновление конкретного дубликата */ - if (mdbx_is_samedata(old_data, new_data)) - /* если данные совпадают, то ничего делать не надо */ - goto bailout; - } - } else { - /* в old_data буфер для сохранения предыдущего значения */ - if (unlikely(new_data && old_data->iov_base == new_data->iov_base)) - return MDBX_EINVAL; - MDBX_val present_data; - rc = mdbx_cursor_get(&mc, &present_key, &present_data, MDBX_SET_KEY); - if (unlikely(rc != MDBX_SUCCESS)) { - old_data->iov_base = NULL; - old_data->iov_len = rc; - if (rc != MDBX_NOTFOUND || (flags & MDBX_CURRENT)) - goto bailout; - } else if (flags & MDBX_NOOVERWRITE) { - rc = MDBX_KEYEXIST; - *old_data = present_data; - goto bailout; - } else { - MDBX_page *page = mc.mc_pg[mc.mc_top]; - if (txn->mt_dbs[dbi].md_flags & MDBX_DUPSORT) { - if (flags & MDBX_CURRENT) { - /* для не-уникальных ключей позволяем update/delete только если ключ - * один */ - MDBX_node *leaf = NODEPTR(page, mc.mc_ki[mc.mc_top]); - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_tassert(txn, XCURSOR_INITED(&mc) && - mc.mc_xcursor->mx_db.md_entries > 1); - if (mc.mc_xcursor->mx_db.md_entries > 1) { - rc = MDBX_EMULTIVAL; - goto bailout; - } - } - /* если данные совпадают, то ничего делать не надо */ - if (new_data && mdbx_is_samedata(&present_data, new_data)) { - *old_data = *new_data; - goto bailout; - } - /* В оригинальной LMDB фладок MDBX_CURRENT здесь приведет - * к замене данных без учета MDBX_DUPSORT сортировки, - * но здесь это в любом случае допустимо, так как мы - * проверили что для ключа есть только одно значение. */ - } else if ((flags & MDBX_NODUPDATA) && - mdbx_is_samedata(&present_data, new_data)) { - /* если данные совпадают и установлен MDBX_NODUPDATA */ - rc = MDBX_KEYEXIST; - goto bailout; - } - } else { - /* если данные совпадают, то ничего делать не надо */ - if (new_data && mdbx_is_samedata(&present_data, new_data)) { - *old_data = *new_data; - goto bailout; - } - flags |= MDBX_CURRENT; - } - - if (page->mp_flags & P_DIRTY) { - if (unlikely(old_data->iov_len < present_data.iov_len)) { - old_data->iov_base = NULL; - old_data->iov_len = present_data.iov_len; - rc = MDBX_RESULT_TRUE; - goto bailout; - } - memcpy(old_data->iov_base, present_data.iov_base, present_data.iov_len); - old_data->iov_len = present_data.iov_len; - } else { - *old_data = present_data; - } - } - } - - if (likely(new_data)) - rc = mdbx_cursor_put(&mc, key, new_data, flags); - else - rc = mdbx_cursor_del(&mc, 0); - -bailout: - txn->mt_cursors[dbi] = mc.mc_next; - return rc; -} - -int mdbx_get_ex(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *data, - size_t *values_count) { - DKBUF; - mdbx_debug("===> get db %u key [%s]", dbi, DKEY(key)); - - if (unlikely(!key || !data || !txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - if (unlikely(txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - MDBX_cursor mc; - MDBX_xcursor mx; - mdbx_cursor_init(&mc, txn, dbi, &mx); - - int exact = 0; - int rc = mdbx_cursor_set(&mc, key, data, MDBX_SET_KEY, &exact); - if (unlikely(rc != MDBX_SUCCESS)) { - if (rc == MDBX_NOTFOUND && values_count) - *values_count = 0; - return rc; - } - - if (values_count) { - *values_count = 1; - if (mc.mc_xcursor != NULL) { - MDBX_node *leaf = NODEPTR(mc.mc_pg[mc.mc_top], mc.mc_ki[mc.mc_top]); - if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { - mdbx_tassert(txn, mc.mc_xcursor == &mx && - (mx.mx_cursor.mc_flags & C_INITIALIZED)); - *values_count = (sizeof(*values_count) >= sizeof(mx.mx_db.md_entries) || - mx.mx_db.md_entries <= SIZE_MAX) - ? (size_t)mx.mx_db.md_entries - : SIZE_MAX; - } - } - } - return MDBX_SUCCESS; -} - -/* Функция сообщает находится ли указанный адрес в "грязной" странице у - * заданной пишущей транзакции. В конечном счете это позволяет избавиться от - * лишнего копирования данных из НЕ-грязных страниц. - * - * "Грязные" страницы - это те, которые уже были изменены в ходе пишущей - * транзакции. Соответственно, какие-либо дальнейшие изменения могут привести - * к перезаписи таких страниц. Поэтому все функции, выполняющие изменения, в - * качестве аргументов НЕ должны получать указатели на данные в таких - * страницах. В свою очередь "НЕ грязные" страницы перед модификацией будут - * скопированы. - * - * Другими словами, данные из "грязных" страниц должны быть либо скопированы - * перед передачей в качестве аргументов для дальнейших модификаций, либо - * отвергнуты на стадии проверки корректности аргументов. - * - * Таким образом, функция позволяет как избавится от лишнего копирования, - * так и выполнить более полную проверку аргументов. - * - * ВАЖНО: Передаваемый указатель должен указывать на начало данных. Только - * так гарантируется что актуальный заголовок страницы будет физически - * расположен в той-же странице памяти, в том числе для многостраничных - * P_OVERFLOW страниц с длинными данными. */ -int mdbx_is_dirty(const MDBX_txn *txn, const void *ptr) { - if (unlikely(!txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(txn->mt_flags & MDBX_TXN_RDONLY)) - return MDBX_RESULT_FALSE; - - const MDBX_env *env = txn->mt_env; - const uintptr_t mask = ~(uintptr_t)(env->me_psize - 1); - const MDBX_page *page = (const MDBX_page *)((uintptr_t)ptr & mask); - - /* LY: Тут не всё хорошо с абсолютной достоверностью результата, - * так как флажок P_DIRTY в LMDB может означать не совсем то, - * что было исходно задумано, детали см в логике кода mdbx_page_touch(). - * - * Более того, в режиме БЕЗ WRITEMAP грязные страницы выделяются через - * malloc(), т.е. находятся вне mmap-диапазона и тогда чтобы отличить - * действительно грязную страницу от указателя на данные пользователя - * следует сканировать dirtylist, что накладно. - * - * Тем не менее, однозначно страница "не грязная" (не будет переписана - * во время транзакции) если адрес находится внутри mmap-диапазона - * и в заголовке страницы нет флажка P_DIRTY. */ - if (env->me_map < (uint8_t *)page) { - const size_t usedbytes = pgno2bytes(env, txn->mt_next_pgno); - if ((uint8_t *)page < env->me_map + usedbytes) { - /* страница внутри диапазона, смотрим на флажки */ - return (page->mp_flags & (P_DIRTY | P_LOOSE | P_KEEP)) - ? MDBX_RESULT_TRUE - : MDBX_RESULT_FALSE; - } - /* Гипотетически здесь возможна ситуация, когда указатель адресует что-то - * в пределах mmap, но за границей распределенных страниц. Это тяжелая - * ошибка, к которой не возможно прийти без каких-то больших нарушений. - * Поэтому не проверяем этот случай кроме как assert-ом, на то что - * страница вне mmap-диаппазона. */ - mdbx_tassert(txn, (uint8_t *)page >= env->me_map + env->me_mapsize); - } - - /* Страница вне используемого mmap-диапазона, т.е. либо в функцию был - * передан некорректный адрес, либо адрес в теневой странице, которая была - * выделена посредством malloc(). - * - * Для WRITE_MAP режима такая страница однозначно "не грязная", - * а для режимов без WRITE_MAP следует просматривать списки dirty - * и spilled страниц у каких-либо транзакций (в том числе дочерних). - * - * Поэтому для WRITE_MAP возвращаем false, а для остальных режимов - * всегда true. Такая логика имеет ряд преимуществ: - * - не тратим время на просмотр списков; - * - результат всегда безопасен (может быть ложно-положительным, - * но не ложно-отрицательным); - * - результат не зависит от вложенности транзакций и от относительного - * положения переданной транзакции в этой рекурсии. */ - return (env->me_flags & MDBX_WRITEMAP) ? MDBX_RESULT_FALSE : MDBX_RESULT_TRUE; -} - -int mdbx_dbi_sequence(MDBX_txn *txn, MDBX_dbi dbi, uint64_t *result, - uint64_t increment) { - if (unlikely(!txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_EBADSIGN; - - if (unlikely(txn->mt_owner != mdbx_thread_self())) - return MDBX_THREAD_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - if (unlikely(TXN_DBI_CHANGED(txn, dbi))) - return MDBX_BAD_DBI; - - MDBX_db *dbs = &txn->mt_dbs[dbi]; - if (likely(result)) - *result = dbs->md_seq; - - if (likely(increment > 0)) { - if (unlikely(txn->mt_flags & MDBX_TXN_BLOCKED)) - return MDBX_BAD_TXN; - - if (unlikely(F_ISSET(txn->mt_flags, MDBX_TXN_RDONLY))) - return MDBX_EACCESS; - - uint64_t new = dbs->md_seq + increment; - if (unlikely(new < increment)) - return MDBX_RESULT_TRUE; - - assert(new > dbs->md_seq); - dbs->md_seq = new; - txn->mt_flags |= MDBX_TXN_DIRTY; - txn->mt_dbflags[dbi] |= DB_DIRTY; - } - - return MDBX_SUCCESS; -} - -/*----------------------------------------------------------------------------*/ -/* attribute support functions for Nexenta */ - -static __inline int mdbx_attr_peek(MDBX_val *data, mdbx_attr_t *attrptr) { - if (unlikely(data->iov_len < sizeof(mdbx_attr_t))) - return MDBX_INCOMPATIBLE; - - if (likely(attrptr != NULL)) - *attrptr = *(mdbx_attr_t *)data->iov_base; - data->iov_len -= sizeof(mdbx_attr_t); - data->iov_base = - likely(data->iov_len > 0) ? ((mdbx_attr_t *)data->iov_base) + 1 : NULL; - - return MDBX_SUCCESS; -} - -static __inline int mdbx_attr_poke(MDBX_val *reserved, MDBX_val *data, - mdbx_attr_t attr, unsigned flags) { - mdbx_attr_t *space = reserved->iov_base; - if (flags & MDBX_RESERVE) { - if (likely(data != NULL)) { - data->iov_base = data->iov_len ? space + 1 : NULL; - } - } else { - *space = attr; - if (likely(data != NULL)) { - memcpy(space + 1, data->iov_base, data->iov_len); - } - } - - return MDBX_SUCCESS; -} - -int mdbx_cursor_get_attr(MDBX_cursor *mc, MDBX_val *key, MDBX_val *data, - mdbx_attr_t *attrptr, MDBX_cursor_op op) { - int rc = mdbx_cursor_get(mc, key, data, op); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - return mdbx_attr_peek(data, attrptr); -} - -int mdbx_get_attr(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *data, - uint64_t *attrptr) { - int rc = mdbx_get(txn, dbi, key, data); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - return mdbx_attr_peek(data, attrptr); -} - -int mdbx_put_attr(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *data, - mdbx_attr_t attr, unsigned flags) { - MDBX_val reserve; - reserve.iov_base = NULL; - reserve.iov_len = (data ? data->iov_len : 0) + sizeof(mdbx_attr_t); - - int rc = mdbx_put(txn, dbi, key, &reserve, flags | MDBX_RESERVE); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - return mdbx_attr_poke(&reserve, data, attr, flags); -} - -int mdbx_cursor_put_attr(MDBX_cursor *cursor, MDBX_val *key, MDBX_val *data, - mdbx_attr_t attr, unsigned flags) { - MDBX_val reserve; - reserve.iov_base = NULL; - reserve.iov_len = (data ? data->iov_len : 0) + sizeof(mdbx_attr_t); - - int rc = mdbx_cursor_put(cursor, key, &reserve, flags | MDBX_RESERVE); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - return mdbx_attr_poke(&reserve, data, attr, flags); -} - -int mdbx_set_attr(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, MDBX_val *data, - mdbx_attr_t attr) { - if (unlikely(!key || !txn)) - return MDBX_EINVAL; - - if (unlikely(txn->mt_signature != MDBX_MT_SIGNATURE)) - return MDBX_VERSION_MISMATCH; - - if (unlikely(!TXN_DBI_EXIST(txn, dbi, DB_USRVALID))) - return MDBX_EINVAL; - - if (unlikely(txn->mt_flags & (MDBX_TXN_RDONLY | MDBX_TXN_BLOCKED))) - return (txn->mt_flags & MDBX_TXN_RDONLY) ? MDBX_EACCESS : MDBX_BAD_TXN; - - MDBX_cursor mc; - MDBX_xcursor mx; - MDBX_val old_data; - mdbx_cursor_init(&mc, txn, dbi, &mx); - int rc = mdbx_cursor_set(&mc, key, &old_data, MDBX_SET, NULL); - if (unlikely(rc != MDBX_SUCCESS)) { - if (rc == MDBX_NOTFOUND && data) { - mc.mc_next = txn->mt_cursors[dbi]; - txn->mt_cursors[dbi] = &mc; - rc = mdbx_cursor_put_attr(&mc, key, data, attr, 0); - txn->mt_cursors[dbi] = mc.mc_next; - } - return rc; - } - - mdbx_attr_t old_attr = 0; - rc = mdbx_attr_peek(&old_data, &old_attr); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - - if (old_attr == attr && (!data || (data->iov_len == old_data.iov_len && - memcmp(data->iov_base, old_data.iov_base, - old_data.iov_len) == 0))) - return MDBX_SUCCESS; - - mc.mc_next = txn->mt_cursors[dbi]; - txn->mt_cursors[dbi] = &mc; - rc = mdbx_cursor_put_attr(&mc, key, data ? data : &old_data, attr, - MDBX_CURRENT); - txn->mt_cursors[dbi] = mc.mc_next; - return rc; -} - -//---------------------------------------------------------------------------- - -#ifdef __SANITIZE_ADDRESS__ -LIBMDBX_API __attribute__((weak)) const char *__asan_default_options() { - return "symbolize=1:allow_addr2line=1:" -#ifdef _DEBUG - "debug=1:" -#endif /* _DEBUG */ - "report_globals=1:" - "replace_str=1:replace_intrin=1:" - "malloc_context_size=9:" - "detect_leaks=1:" - "check_printf=1:" - "detect_deadlocks=1:" -#ifndef LTO_ENABLED - "check_initialization_order=1:" -#endif - "detect_stack_use_after_return=1:" - "intercept_tls_get_addr=1:" - "decorate_proc_maps=1:" - "abort_on_error=1"; -} -#endif /* __SANITIZE_ADDRESS__ */ diff --git a/plugins/Dbx_mdb/src/mdbx/mdbx.h b/plugins/Dbx_mdb/src/mdbx/mdbx.h deleted file mode 100644 index 421d833b70..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/mdbx.h +++ /dev/null @@ -1,1825 +0,0 @@ -/* LICENSE AND COPYRUSTING ***************************************************** - * - * Copyright 2015-2017 Leonid Yuriev <leo@yuriev.ru> - * and other libmdbx authors: please see AUTHORS file. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted only as authorized by the OpenLDAP - * Public License. - * - * A copy of this license is available in the file LICENSE in the - * top-level directory of the distribution or, alternatively, at - * <http://www.OpenLDAP.org/license.html>. - * - * --- - * - * This code is derived from "LMDB engine" written by - * Howard Chu (Symas Corporation), which itself derived from btree.c - * written by Martin Hedenfalk. - * - * --- - * - * Portions Copyright 2011-2015 Howard Chu, Symas Corp. All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted only as authorized by the OpenLDAP - * Public License. - * - * A copy of this license is available in the file LICENSE in the - * top-level directory of the distribution or, alternatively, at - * <http://www.OpenLDAP.org/license.html>. - * - * --- - * - * Portions Copyright (c) 2009, 2010 Martin Hedenfalk <martin@bzero.se> - * - * Permission to use, copy, modify, and distribute this software for any - * purpose with or without fee is hereby granted, provided that the above - * copyright notice and this permission notice appear in all copies. - * - * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES - * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF - * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR - * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES - * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN - * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF - * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ - -/* ACKNOWLEDGEMENTS ************************************************************ - * - * Howard Chu (Symas Corporation) - the author of LMDB, - * from which originated the MDBX in 2015. - * - * Martin Hedenfalk <martin@bzero.se> - the author of `btree.c` code, - * which was used for begin development of LMDB. */ - -#pragma once -#ifndef LIBMDBX_H -#define LIBMDBX_H - -/* IMPENDING CHANGES WARNING *************************************************** - * - * Now MDBX is under active development and until November 2017 is expected a - * big change both of API and database format. Unfortunately those update will - * lead to loss of compatibility with previous versions. - * - * The aim of this revolution in providing a clearer robust API and adding new - * features, including the database properties. */ - -/*--------------------------------------------------------------------------*/ - -#ifdef _MSC_VER -#pragma warning(push, 1) -#pragma warning(disable : 4548) /* expression before comma has no effect; \ - expected expression with side - effect */ -#pragma warning(disable : 4530) /* C++ exception handler used, but unwind \ - * semantics are not enabled. Specify /EHsc */ -#pragma warning(disable : 4577) /* 'noexcept' used with no exception handling \ - * mode specified; termination on exception is \ - * not guaranteed. Specify /EHsc */ -#endif /* _MSC_VER (warnings) */ - -#include <stdarg.h> -#include <stddef.h> -#include <stdint.h> - -#if defined(_WIN32) || defined(_WIN64) - -#include <windows.h> -#include <winnt.h> -#ifndef __mode_t_defined -typedef unsigned short mode_t; -#endif -typedef HANDLE mdbx_filehandle_t; -typedef DWORD mdbx_pid_t; -typedef DWORD mdbx_tid_t; -#define MDBX_ENODATA ERROR_HANDLE_EOF -#define MDBX_EINVAL ERROR_INVALID_PARAMETER -#define MDBX_EACCESS ERROR_ACCESS_DENIED -#define MDBX_ENOMEM ERROR_OUTOFMEMORY -#define MDBX_EROFS ERROR_FILE_READ_ONLY -#define MDBX_ENOSYS ERROR_NOT_SUPPORTED -#define MDBX_EIO ERROR_WRITE_FAULT -#define MDBX_EPERM ERROR_INVALID_FUNCTION -#define MDBX_EINTR ERROR_CANCELLED - -#else - -#include <errno.h> /* for error codes */ -#include <pthread.h> /* for pthread_t */ -#include <sys/types.h> /* for pid_t */ -#include <sys/uio.h> /* for truct iovec */ -#define HAVE_STRUCT_IOVEC 1 -typedef int mdbx_filehandle_t; -typedef pid_t mdbx_pid_t; -typedef pthread_t mdbx_tid_t; -#define MDBX_ENODATA ENODATA -#define MDBX_EINVAL EINVAL -#define MDBX_EACCESS EACCES -#define MDBX_ENOMEM ENOMEM -#define MDBX_EROFS EROFS -#define MDBX_ENOSYS ENOSYS -#define MDBX_EIO EIO -#define MDBX_EPERM EPERM -#define MDBX_EINTR EINTR -#endif - -#ifdef _MSC_VER -#pragma warning(pop) -#endif - -/*--------------------------------------------------------------------------*/ - -#ifndef __has_attribute -#define __has_attribute(x) (0) -#endif - -#ifndef __dll_export -#if defined(_WIN32) || defined(__CYGWIN__) -#if defined(__GNUC__) || __has_attribute(dllexport) -#define __dll_export __attribute__((dllexport)) -#elif defined(_MSC_VER) -#define __dll_export __declspec(dllexport) -#else -#define __dll_export -#endif -#elif defined(__GNUC__) || __has_attribute(visibility) -#define __dll_export __attribute__((visibility("default"))) -#else -#define __dll_export -#endif -#endif /* __dll_export */ - -#ifndef __dll_import -#if defined(_WIN32) || defined(__CYGWIN__) -#if defined(__GNUC__) || __has_attribute(dllimport) -#define __dll_import __attribute__((dllimport)) -#elif defined(_MSC_VER) -#define __dll_import __declspec(dllimport) -#else -#define __dll_import -#endif -#else -#define __dll_import -#endif -#endif /* __dll_import */ - -/*--------------------------------------------------------------------------*/ - -#define MDBX_VERSION_MAJOR 0 -#define MDBX_VERSION_MINOR 0 - -#if defined(LIBMDBX_EXPORTS) -#define LIBMDBX_API __dll_export -#elif defined(LIBMDBX_IMPORTS) -#define LIBMDBX_API __dll_import -#else -#define LIBMDBX_API -#endif /* LIBMDBX_API */ - -#ifdef __cplusplus -extern "C" { -#endif - -typedef struct mdbx_version_info { - uint8_t major; - uint8_t minor; - uint16_t release; - uint32_t revision; - struct { - const char *datetime; - const char *tree; - const char *commit; - const char *describe; - } git; -} mdbx_version_info; - -typedef struct mdbx_build_info { - const char *datetime; - const char *target; - const char *options; - const char *compiler; - const char *flags; -} mdbx_build_info; - -extern LIBMDBX_API const mdbx_version_info mdbx_version; -extern LIBMDBX_API const mdbx_build_info mdbx_build; - -/* The name of the lock file in the DB environment */ -#define MDBX_LOCKNAME "/mdbx.lck" -/* The name of the data file in the DB environment */ -#define MDBX_DATANAME "/mdbx.dat" -/* The suffix of the lock file when no subdir is used */ -#define MDBX_LOCK_SUFFIX "-lck" - -/* Opaque structure for a database environment. - * - * A DB environment supports multiple databases, all residing in the same - * shared-memory map. */ -typedef struct MDBX_env MDBX_env; - -/* Opaque structure for a transaction handle. - * - * All database operations require a transaction handle. Transactions may be - * read-only or read-write. */ -typedef struct MDBX_txn MDBX_txn; - -/* A handle for an individual database in the DB environment. */ -typedef uint32_t MDBX_dbi; - -/* Opaque structure for navigating through a database */ -typedef struct MDBX_cursor MDBX_cursor; - -/* Generic structure used for passing keys and data in and out - * of the database. - * - * Values returned from the database are valid only until a subsequent - * update operation, or the end of the transaction. Do not modify or - * free them, they commonly point into the database itself. - * - * Key sizes must be between 1 and mdbx_env_get_maxkeysize() inclusive. - * The same applies to data sizes in databases with the MDBX_DUPSORT flag. - * Other data items can in theory be from 0 to 0xffffffff bytes long. */ -#ifndef HAVE_STRUCT_IOVEC -struct iovec { - void *iov_base; - size_t iov_len; -}; -#define HAVE_STRUCT_IOVEC -#endif /* HAVE_STRUCT_IOVEC */ - -typedef struct iovec MDBX_val; - -/* The maximum size of a data item. - * MDBX only store a 32 bit value for node sizes. */ -#define MDBX_MAXDATASIZE INT32_MAX - -/* A callback function used to compare two keys in a database */ -typedef int(MDBX_cmp_func)(const MDBX_val *a, const MDBX_val *b); - -/* Environment Flags */ -/* no environment directory */ -#define MDBX_NOSUBDIR 0x4000u -/* don't fsync after commit */ -#define MDBX_NOSYNC 0x10000u -/* read only */ -#define MDBX_RDONLY 0x20000u -/* don't fsync metapage after commit */ -#define MDBX_NOMETASYNC 0x40000u -/* use writable mmap */ -#define MDBX_WRITEMAP 0x80000u -/* use asynchronous msync when MDBX_WRITEMAP is used */ -#define MDBX_MAPASYNC 0x100000u -/* tie reader locktable slots to MDBX_txn objects instead of to threads */ -#define MDBX_NOTLS 0x200000u -/* don't do any locking, caller must manage their own locks - * WARNING: libmdbx don't support this mode. */ -#define MDBX_NOLOCK__UNSUPPORTED 0x400000u -/* don't do readahead */ -#define MDBX_NORDAHEAD 0x800000u -/* don't initialize malloc'd memory before writing to datafile */ -#define MDBX_NOMEMINIT 0x1000000u -/* aim to coalesce FreeDB records */ -#define MDBX_COALESCE 0x2000000u -/* LIFO policy for reclaiming FreeDB records */ -#define MDBX_LIFORECLAIM 0x4000000u -/* make a steady-sync only on close and explicit env-sync */ -#define MDBX_UTTERLY_NOSYNC (MDBX_NOSYNC | MDBX_MAPASYNC) -/* debuging option, fill/perturb released pages */ -#define MDBX_PAGEPERTURB 0x8000000u - -/* Database Flags */ -/* use reverse string keys */ -#define MDBX_REVERSEKEY 0x02u -/* use sorted duplicates */ -#define MDBX_DUPSORT 0x04u -/* numeric keys in native byte order, either uint32_t or uint64_t. - * The keys must all be of the same size. */ -#define MDBX_INTEGERKEY 0x08u -/* with MDBX_DUPSORT, sorted dup items have fixed size */ -#define MDBX_DUPFIXED 0x10u -/* with MDBX_DUPSORT, dups are MDBX_INTEGERKEY-style integers */ -#define MDBX_INTEGERDUP 0x20u -/* with MDBX_DUPSORT, use reverse string dups */ -#define MDBX_REVERSEDUP 0x40u -/* create DB if not already existing */ -#define MDBX_CREATE 0x40000u - -/* Write Flags */ -/* For put: Don't write if the key already exists. */ -#define MDBX_NOOVERWRITE 0x10u -/* Only for MDBX_DUPSORT - * For put: don't write if the key and data pair already exist. - * For mdbx_cursor_del: remove all duplicate data items. */ -#define MDBX_NODUPDATA 0x20u -/* For mdbx_cursor_put: overwrite the current key/data pair - * MDBX allows this flag for mdbx_put() for explicit overwrite/update without - * insertion. */ -#define MDBX_CURRENT 0x40u -/* For put: Just reserve space for data, don't copy it. Return a - * pointer to the reserved space. */ -#define MDBX_RESERVE 0x10000u -/* Data is being appended, don't split full pages. */ -#define MDBX_APPEND 0x20000u -/* Duplicate data is being appended, don't split full pages. */ -#define MDBX_APPENDDUP 0x40000u -/* Store multiple data items in one call. Only for MDBX_DUPFIXED. */ -#define MDBX_MULTIPLE 0x80000u - -/* Transaction Flags */ -/* Do not block when starting a write transaction */ -#define MDBX_TRYTXN 0x10000000u - -/* Copy Flags */ -/* Compacting copy: Omit free space from copy, and renumber all - * pages sequentially. */ -#define MDBX_CP_COMPACT 1u - -/* Cursor Get operations. - * - * This is the set of all operations for retrieving data - * using a cursor. */ -typedef enum MDBX_cursor_op { - MDBX_FIRST, /* Position at first key/data item */ - MDBX_FIRST_DUP, /* MDBX_DUPSORT-only: Position at first data item - * of current key. */ - MDBX_GET_BOTH, /* MDBX_DUPSORT-only: Position at key/data pair. */ - MDBX_GET_BOTH_RANGE, /* MDBX_DUPSORT-only: position at key, nearest data. */ - MDBX_GET_CURRENT, /* Return key/data at current cursor position */ - MDBX_GET_MULTIPLE, /* MDBX_DUPFIXED-only: Return key and up to a page of - * duplicate data items from current cursor position. - * Move cursor to prepare for MDBX_NEXT_MULTIPLE.*/ - MDBX_LAST, /* Position at last key/data item */ - MDBX_LAST_DUP, /* MDBX_DUPSORT-only: Position at last data item - * of current key. */ - MDBX_NEXT, /* Position at next data item */ - MDBX_NEXT_DUP, /* MDBX_DUPSORT-only: Position at next data item - * of current key. */ - MDBX_NEXT_MULTIPLE, /* MDBX_DUPFIXED-only: Return key and up to a page of - * duplicate data items from next cursor position. - * Move cursor to prepare for MDBX_NEXT_MULTIPLE. */ - MDBX_NEXT_NODUP, /* Position at first data item of next key */ - MDBX_PREV, /* Position at previous data item */ - MDBX_PREV_DUP, /* MDBX_DUPSORT-only: Position at previous data item - * of current key. */ - MDBX_PREV_NODUP, /* Position at last data item of previous key */ - MDBX_SET, /* Position at specified key */ - MDBX_SET_KEY, /* Position at specified key, return both key and data */ - MDBX_SET_RANGE, /* Position at first key greater than or equal to - * specified key. */ - MDBX_PREV_MULTIPLE /* MDBX_DUPFIXED-only: Position at previous page and - * return key and up to a page of duplicate data items. */ -} MDBX_cursor_op; - -/* Return Codes - * BerkeleyDB uses -30800 to -30999, we'll go under them */ - -/* Successful result */ -#define MDBX_SUCCESS 0 -#define MDBX_RESULT_FALSE MDBX_SUCCESS -#define MDBX_RESULT_TRUE (-1) - -/* key/data pair already exists */ -#define MDBX_KEYEXIST (-30799) -/* key/data pair not found (EOF) */ -#define MDBX_NOTFOUND (-30798) -/* Requested page not found - this usually indicates corruption */ -#define MDBX_PAGE_NOTFOUND (-30797) -/* Located page was wrong type */ -#define MDBX_CORRUPTED (-30796) -/* Update of meta page failed or environment had fatal error */ -#define MDBX_PANIC (-30795) -/* DB file version mismatch with libmdbx */ -#define MDBX_VERSION_MISMATCH (-30794) -/* File is not a valid MDBX file */ -#define MDBX_INVALID (-30793) -/* Environment mapsize reached */ -#define MDBX_MAP_FULL (-30792) -/* Environment maxdbs reached */ -#define MDBX_DBS_FULL (-30791) -/* Environment maxreaders reached */ -#define MDBX_READERS_FULL (-30790) -/* Txn has too many dirty pages */ -#define MDBX_TXN_FULL (-30788) -/* Cursor stack too deep - internal error */ -#define MDBX_CURSOR_FULL (-30787) -/* Page has not enough space - internal error */ -#define MDBX_PAGE_FULL (-30786) -/* Database contents grew beyond environment mapsize */ -#define MDBX_MAP_RESIZED (-30785) -/* Operation and DB incompatible, or DB type changed. This can mean: - * - The operation expects an MDBX_DUPSORT / MDBX_DUPFIXED database. - * - Opening a named DB when the unnamed DB has MDBX_DUPSORT/MDBX_INTEGERKEY. - * - Accessing a data record as a database, or vice versa. - * - The database was dropped and recreated with different flags. */ -#define MDBX_INCOMPATIBLE (-30784) -/* Invalid reuse of reader locktable slot */ -#define MDBX_BAD_RSLOT (-30783) -/* Transaction must abort, has a child, or is invalid */ -#define MDBX_BAD_TXN (-30782) -/* Unsupported size of key/DB name/data, or wrong DUPFIXED size */ -#define MDBX_BAD_VALSIZE (-30781) -/* The specified DBI was changed unexpectedly */ -#define MDBX_BAD_DBI (-30780) -/* Unexpected problem - txn should abort */ -#define MDBX_PROBLEM (-30779) -/* Unexpected problem - txn should abort */ -#define MDBX_BUSY (-30778) -/* The last defined error code */ -#define MDBX_LAST_ERRCODE MDBX_BUSY - -/* The mdbx_put() or mdbx_replace() was called for key, - that has more that one associated value. */ -#define MDBX_EMULTIVAL (-30421) - -/* Bad signature of a runtime object(s), this can mean: - * - memory corruption or double-free; - * - ABI version mismatch (rare case); */ -#define MDBX_EBADSIGN (-30420) - -/* Database should be recovered, but this could NOT be done automatically - * right now (e.g. in readonly mode and so forth). */ -#define MDBX_WANNA_RECOVERY (-30419) - -/* The given key value is mismatched to the current cursor position, - * when mdbx_cursor_put() called with MDBX_CURRENT option. */ -#define MDBX_EKEYMISMATCH (-30418) - -/* Database is too large for current system, - * e.g. could NOT be mapped into RAM. */ -#define MDBX_TOO_LARGE (-30417) - -/* A thread has attempted to use a not owned object, - * e.g. a transaction that started by another thread. */ -#define MDBX_THREAD_MISMATCH (-30416) - -/* Statistics for a database in the environment */ -typedef struct MDBX_stat { - uint32_t ms_psize; /* Size of a database page. - * This is currently the same for all databases. */ - uint32_t ms_depth; /* Depth (height) of the B-tree */ - uint64_t ms_branch_pages; /* Number of internal (non-leaf) pages */ - uint64_t ms_leaf_pages; /* Number of leaf pages */ - uint64_t ms_overflow_pages; /* Number of overflow pages */ - uint64_t ms_entries; /* Number of data items */ -} MDBX_stat; - -/* Information about the environment */ -typedef struct MDBX_envinfo { - struct { - uint64_t lower; /* lower limit for datafile size */ - uint64_t upper; /* upper limit for datafile size */ - uint64_t current; /* current datafile size */ - uint64_t shrink; /* shrink theshold for datafile */ - uint64_t grow; /* growth step for datafile */ - } mi_geo; - uint64_t mi_mapsize; /* Size of the data memory map */ - uint64_t mi_last_pgno; /* ID of the last used page */ - uint64_t mi_recent_txnid; /* ID of the last committed transaction */ - uint64_t mi_latter_reader_txnid; /* ID of the last reader transaction */ - uint64_t mi_meta0_txnid, mi_meta0_sign; - uint64_t mi_meta1_txnid, mi_meta1_sign; - uint64_t mi_meta2_txnid, mi_meta2_sign; - uint32_t mi_maxreaders; /* max reader slots in the environment */ - uint32_t mi_numreaders; /* max reader slots used in the environment */ - uint32_t mi_dxb_pagesize; /* database pagesize */ - uint32_t mi_sys_pagesize; /* system pagesize */ -} MDBX_envinfo; - -/* Return a string describing a given error code. - * - * This function is a superset of the ANSI C X3.159-1989 (ANSI C) strerror(3) - * function. If the error code is greater than or equal to 0, then the string - * returned by the system function strerror(3) is returned. If the error code - * is less than 0, an error string corresponding to the MDBX library error is - * returned. See errors for a list of MDBX-specific error codes. - * - * [in] err The error code - * - * Returns "error message" The description of the error */ -LIBMDBX_API const char *mdbx_strerror(int errnum); -LIBMDBX_API const char *mdbx_strerror_r(int errnum, char *buf, size_t buflen); - -/* Create an MDBX environment handle. - * - * This function allocates memory for a MDBX_env structure. To release - * the allocated memory and discard the handle, call mdbx_env_close(). - * Before the handle may be used, it must be opened using mdbx_env_open(). - * Various other options may also need to be set before opening the handle, - * e.g. mdbx_env_set_mapsize(), mdbx_env_set_maxreaders(), - * mdbx_env_set_maxdbs(), depending on usage requirements. - * - * [out] env The address where the new handle will be stored - * - * Returns A non-zero error value on failure and 0 on success. */ -LIBMDBX_API int mdbx_env_create(MDBX_env **penv); - -/* Open an environment handle. - * - * If this function fails, mdbx_env_close() must be called to discard - * the MDBX_env handle. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] path The directory in which the database files reside. - * This directory must already exist and be writable. - * [in] flags Special options for this environment. This parameter - * must be set to 0 or by bitwise OR'ing together one - * or more of the values described here. - * - * Flags set by mdbx_env_set_flags() are also used: - * - MDBX_NOSUBDIR - * By default, MDBX creates its environment in a directory whose - * pathname is given in path, and creates its data and lock files - * under that directory. With this option, path is used as-is for - * the database main data file. The database lock file is the path - * with "-lock" appended. - * - * - MDBX_RDONLY - * Open the environment in read-only mode. No write operations will - * be allowed. MDBX will still modify the lock file - except on - * read-only filesystems, where MDBX does not use locks. - * - * - MDBX_WRITEMAP - * Use a writeable memory map unless MDBX_RDONLY is set. This uses fewer - * mallocs but loses protection from application bugs like wild pointer - * writes and other bad updates into the database. - * This may be slightly faster for DBs that fit entirely in RAM, - * but is slower for DBs larger than RAM. - * Incompatible with nested transactions. - * Do not mix processes with and without MDBX_WRITEMAP on the same - * environment. This can defeat durability (mdbx_env_sync etc). - * - * - MDBX_NOMETASYNC - * Flush system buffers to disk only once per transaction, omit the - * metadata flush. Defer that until the system flushes files to disk, - * or next non-MDBX_RDONLY commit or mdbx_env_sync(). This optimization - * maintains database integrity, but a system crash may undo the last - * committed transaction. I.e. it preserves the ACI (atomicity, - * consistency, isolation) but not D (durability) database property. - * This flag may be changed at any time using mdbx_env_set_flags(). - * - * - MDBX_NOSYNC - * Don't flush system buffers to disk when committing a transaction. - * This optimization means a system crash can corrupt the database or - * lose the last transactions if buffers are not yet flushed to disk. - * The risk is governed by how often the system flushes dirty buffers - * to disk and how often mdbx_env_sync() is called. However, if the - * filesystem preserves write order and the MDBX_WRITEMAP and/or - * MDBX_LIFORECLAIM flags are not used, transactions exhibit ACI - * (atomicity, consistency, isolation) properties and only lose D - * (durability). I.e. database integrity is maintained, but a system - * crash may undo the final transactions. - * - * Note that (MDBX_NOSYNC | MDBX_WRITEMAP) leaves the system with no - * hint for when to write transactions to disk. - * Therefore the (MDBX_MAPASYNC | MDBX_WRITEMAP) may be preferable. - * This flag may be changed at any time using mdbx_env_set_flags(). - * - * - MDBX_UTTERLY_NOSYNC (internally MDBX_NOSYNC | MDBX_MAPASYNC) - * FIXME: TODO - * - * - MDBX_MAPASYNC - * When using MDBX_WRITEMAP, use asynchronous flushes to disk. As with - * MDBX_NOSYNC, a system crash can then corrupt the database or lose - * the last transactions. Calling mdbx_env_sync() ensures on-disk - * database integrity until next commit. This flag may be changed at - * any time using mdbx_env_set_flags(). - * - * - MDBX_NOTLS - * Don't use Thread-Local Storage. Tie reader locktable slots to - * MDBX_txn objects instead of to threads. I.e. mdbx_txn_reset() keeps - * the slot reseved for the MDBX_txn object. A thread may use parallel - * read-only transactions. A read-only transaction may span threads if - * the user synchronizes its use. Applications that multiplex many - * user threads over individual OS threads need this option. Such an - * application must also serialize the write transactions in an OS - * thread, since MDBX's write locking is unaware of the user threads. - * - * - MDBX_NOLOCK (don't supported by MDBX) - * Don't do any locking. If concurrent access is anticipated, the - * caller must manage all concurrency itself. For proper operation - * the caller must enforce single-writer semantics, and must ensure - * that no readers are using old transactions while a writer is - * active. The simplest approach is to use an exclusive lock so that - * no readers may be active at all when a writer begins. - * - * - MDBX_NORDAHEAD - * Turn off readahead. Most operating systems perform readahead on - * read requests by default. This option turns it off if the OS - * supports it. Turning it off may help random read performance - * when the DB is larger than RAM and system RAM is full. - * - * - MDBX_NOMEMINIT - * Don't initialize malloc'd memory before writing to unused spaces - * in the data file. By default, memory for pages written to the data - * file is obtained using malloc. While these pages may be reused in - * subsequent transactions, freshly malloc'd pages will be initialized - * to zeroes before use. This avoids persisting leftover data from other - * code (that used the heap and subsequently freed the memory) into the - * data file. Note that many other system libraries may allocate and free - * memory from the heap for arbitrary uses. E.g., stdio may use the heap - * for file I/O buffers. This initialization step has a modest performance - * cost so some applications may want to disable it using this flag. This - * option can be a problem for applications which handle sensitive data - * like passwords, and it makes memory checkers like Valgrind noisy. This - * flag is not needed with MDBX_WRITEMAP, which writes directly to the - * mmap instead of using malloc for pages. The initialization is also - * skipped if MDBX_RESERVE is used; the caller is expected to overwrite - * all of the memory that was reserved in that case. This flag may be - * changed at any time using mdbx_env_set_flags(). - * - * - MDBX_COALESCE - * Aim to coalesce records while reclaiming FreeDB. This flag may be - * changed at any time using mdbx_env_set_flags(). - * FIXME: TODO - * - * - MDBX_LIFORECLAIM - * LIFO policy for reclaiming FreeDB records. This significantly reduce - * write IPOs in case MDBX_NOSYNC with periodically checkpoints. - * FIXME: TODO - * - * [in] mode The UNIX permissions to set on created files. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_VERSION_MISMATCH - the version of the MDBX library doesn't match the - * version that created the database environment. - * - MDBX_INVALID - the environment file headers are corrupted. - * - MDBX_ENOENT - the directory specified by the path parameter - * doesn't exist. - * - MDBX_EACCES - the user didn't have permission to access - * the environment files. - * - MDBX_EAGAIN - the environment was locked by another process. */ -LIBMDBX_API int mdbx_env_open(MDBX_env *env, const char *path, unsigned flags, - mode_t mode); -LIBMDBX_API int mdbx_env_open_ex(MDBX_env *env, const char *path, - unsigned flags, mode_t mode, int *exclusive); - -/* Copy an MDBX environment to the specified path, with options. - * - * This function may be used to make a backup of an existing environment. - * No lockfile is created, since it gets recreated at need. - * NOTE: This call can trigger significant file size growth if run in - * parallel with write transactions, because it employs a read-only - * transaction. See long-lived transactions under "Caveats" section. - * - * [in] env An environment handle returned by mdbx_env_create(). It must - * have already been opened successfully. - * [in] path The directory in which the copy will reside. This directory - * must already exist and be writable but must otherwise be empty. - * [in] flags Special options for this operation. This parameter must be set - * to 0 or by bitwise OR'ing together one or more of the values - * described here: - * - * - MDBX_CP_COMPACT - * Perform compaction while copying: omit free pages and sequentially - * renumber all pages in output. This option consumes little bit more - * CPU for processing, but may running quickly than the default, on - * account skipping free pages. - * - * NOTE: Currently it fails if the environment has suffered a page leak. - * - * Returns A non-zero error value on failure and 0 on success. */ -LIBMDBX_API int mdbx_env_copy(MDBX_env *env, const char *path, unsigned flags); - -/* Copy an MDBX environment to the specified file descriptor, - * with options. - * - * This function may be used to make a backup of an existing environment. - * No lockfile is created, since it gets recreated at need. See - * mdbx_env_copy() for further details. - * - * NOTE: This call can trigger significant file size growth if run in - * parallel with write transactions, because it employs a read-only - * transaction. See long-lived transactions under "Caveats" section. - * - * [in] env An environment handle returned by mdbx_env_create(). It must - * have already been opened successfully. - * [in] fd The filedescriptor to write the copy to. It must have already - * been opened for Write access. - * [in] flags Special options for this operation. See mdbx_env_copy() for - * options. - * - * Returns A non-zero error value on failure and 0 on success. */ -LIBMDBX_API int mdbx_env_copy2fd(MDBX_env *env, mdbx_filehandle_t fd, - unsigned flags); - -/* Return statistics about the MDBX environment. - * - * [in] env An environment handle returned by mdbx_env_create() - * [out] stat The address of an MDBX_stat structure where the statistics - * will be copied */ -LIBMDBX_API int mdbx_env_stat(MDBX_env *env, MDBX_stat *stat, size_t bytes); - -/* Return information about the MDBX environment. - * - * [in] env An environment handle returned by mdbx_env_create() - * [out] stat The address of an MDBX_envinfo structure - * where the information will be copied */ -LIBMDBX_API int mdbx_env_info(MDBX_env *env, MDBX_envinfo *info, size_t bytes); - -/* Flush the data buffers to disk. - * - * Data is always written to disk when mdbx_txn_commit() is called, - * but the operating system may keep it buffered. MDBX always flushes - * the OS buffers upon commit as well, unless the environment was - * opened with MDBX_NOSYNC or in part MDBX_NOMETASYNC. This call is - * not valid if the environment was opened with MDBX_RDONLY. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] force If non-zero, force a synchronous flush. Otherwise if the - * environment has the MDBX_NOSYNC flag set the flushes will be - * omitted, and with MDBX_MAPASYNC they will be asynchronous. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EACCES - the environment is read-only. - * - MDBX_EINVAL - an invalid parameter was specified. - * - MDBX_EIO - an error occurred during synchronization. */ -LIBMDBX_API int mdbx_env_sync(MDBX_env *env, int force); - -/* Close the environment and release the memory map. - * - * Only a single thread may call this function. All transactions, databases, - * and cursors must already be closed before calling this function. Attempts - * to use any such handles after calling this function will cause a SIGSEGV. - * The environment handle will be freed and must not be used again after this - * call. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] dont_sync A dont'sync flag, if non-zero the last checkpoint (meta-page - * update) will be kept "as is" and may be still "weak" in the - * NOSYNC/MAPASYNC modes. Such "weak" checkpoint will be - * ignored on opening next time, and transactions since the - * last non-weak checkpoint (meta-page update) will rolledback - * for consistency guarantee. */ -LIBMDBX_API void mdbx_env_close(MDBX_env *env); - -/* Set environment flags. - * - * This may be used to set some flags in addition to those from - * mdbx_env_open(), or to unset these flags. If several threads - * change the flags at the same time, the result is undefined. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] flags The flags to change, bitwise OR'ed together - * [in] onoff A non-zero value sets the flags, zero clears them. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_env_set_flags(MDBX_env *env, unsigned flags, int onoff); - -/* Get environment flags. - * - * [in] env An environment handle returned by mdbx_env_create() - * [out] flags The address of an integer to store the flags - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_env_get_flags(MDBX_env *env, unsigned *flags); - -/* Return the path that was used in mdbx_env_open(). - * - * [in] env An environment handle returned by mdbx_env_create() - * [out] path Address of a string pointer to contain the path. - * This is the actual string in the environment, not a copy. - * It should not be altered in any way. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_env_get_path(MDBX_env *env, const char **path); - -/* Return the file descriptor for the given environment. - * - * NOTE: All MDBX file descriptors have FD_CLOEXEC and - * could't be used after exec() and or fork(). - * - * [in] env An environment handle returned by mdbx_env_create() - * [out] fd Address of a int to contain the descriptor. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_env_get_fd(MDBX_env *env, mdbx_filehandle_t *fd); - -/* Set the size of the memory map to use for this environment. - * - * The size should be a multiple of the OS page size. The default is - * 10485760 bytes. The size of the memory map is also the maximum size - * of the database. The value should be chosen as large as possible, - * to accommodate future growth of the database. - * This function should be called after mdbx_env_create() and before - * mdbx_env_open(). It may be called at later times if no transactions - * are active in this process. Note that the library does not check for - * this condition, the caller must ensure it explicitly. - * - * The new size takes effect immediately for the current process but - * will not be persisted to any others until a write transaction has been - * committed by the current process. Also, only mapsize increases are - * persisted into the environment. - * - * If the mapsize is increased by another process, and data has grown - * beyond the range of the current mapsize, mdbx_txn_begin() will - * return MDBX_MAP_RESIZED. This function may be called with a size - * of zero to adopt the new size. - * - * Any attempt to set a size smaller than the space already consumed by the - * environment will be silently changed to the current size of the used space. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] size The size in bytes - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified, - * or the environment has an active write transaction. */ -LIBMDBX_API int mdbx_env_set_mapsize(MDBX_env *env, size_t size); -LIBMDBX_API int mdbx_env_set_geometry(MDBX_env *env, intptr_t size_lower, - intptr_t size_now, intptr_t size_upper, - intptr_t growth_step, - intptr_t shrink_threshold, - intptr_t pagesize); - -/* Set the maximum number of threads/reader slots for the environment. - * - * This defines the number of slots in the lock table that is used to track - * readers in the the environment. The default is 61. - * Starting a read-only transaction normally ties a lock table slot to the - * current thread until the environment closes or the thread exits. If - * MDBX_NOTLS is in use, mdbx_txn_begin() instead ties the slot to the - * MDBX_txn object until it or the MDBX_env object is destroyed. - * This function may only be called after mdbx_env_create() and before - * mdbx_env_open(). - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] readers The maximum number of reader lock table slots - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified, - * or the environment is already open. */ -LIBMDBX_API int mdbx_env_set_maxreaders(MDBX_env *env, unsigned readers); - -/* Get the maximum number of threads/reader slots for the environment. - * - * [in] env An environment handle returned by mdbx_env_create() - * [out] readers Address of an integer to store the number of readers - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_env_get_maxreaders(MDBX_env *env, unsigned *readers); - -/* Set the maximum number of named databases for the environment. - * - * This function is only needed if multiple databases will be used in the - * environment. Simpler applications that use the environment as a single - * unnamed database can ignore this option. - * This function may only be called after mdbx_env_create() and before - * mdbx_env_open(). - * - * Currently a moderate number of slots are cheap but a huge number gets - * expensive: 7-120 words per transaction, and every mdbx_dbi_open() - * does a linear search of the opened slots. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] dbs The maximum number of databases - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified, - * or the environment is already open. */ -LIBMDBX_API int mdbx_env_set_maxdbs(MDBX_env *env, MDBX_dbi dbs); - -/* Get the maximum size of keys and MDBX_DUPSORT data we can write. - * - * [in] env An environment handle returned by mdbx_env_create() - * - * Returns The maximum size of a key we can write. */ -LIBMDBX_API int mdbx_env_get_maxkeysize(MDBX_env *env); -LIBMDBX_API int mdbx_get_maxkeysize(size_t pagesize); - -/* Set application information associated with the MDBX_env. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] ctx An arbitrary pointer for whatever the application needs. - * - * Returns A non-zero error value on failure and 0 on success. */ -LIBMDBX_API int mdbx_env_set_userctx(MDBX_env *env, void *ctx); - -/* Get the application information associated with the MDBX_env. - * - * [in] env An environment handle returned by mdbx_env_create() - * Returns The pointer set by mdbx_env_set_userctx(). */ -LIBMDBX_API void *mdbx_env_get_userctx(MDBX_env *env); - -/* A callback function for most MDBX assert() failures, - * called before printing the message and aborting. - * - * [in] env An environment handle returned by mdbx_env_create(). - * [in] msg The assertion message, not including newline. */ -typedef void MDBX_assert_func(const MDBX_env *env, const char *msg, - const char *function, unsigned line); - -/* Set or reset the assert() callback of the environment. - * - * Disabled if libmdbx is buillt with MDBX_DEBUG=0. - * NOTE: This hack should become obsolete as mdbx's error handling matures. - * - * [in] env An environment handle returned by mdbx_env_create(). - * [in] func An MDBX_assert_func function, or 0. - * - * Returns A non-zero error value on failure and 0 on success. */ -LIBMDBX_API int mdbx_env_set_assert(MDBX_env *env, MDBX_assert_func *func); - -/* Create a transaction for use with the environment. - * - * The transaction handle may be discarded using mdbx_txn_abort() - * or mdbx_txn_commit(). - * NOTE: A transaction and its cursors must only be used by a single - * thread, and a thread may only have a single transaction at a time. - * If MDBX_NOTLS is in use, this does not apply to read-only transactions. - * NOTE: Cursors may not span transactions. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] parent If this parameter is non-NULL, the new transaction will be - * a nested transaction, with the transaction indicated by parent - * as its parent. Transactions may be nested to any level. - * A parent transaction and its cursors may not issue any other - * operations than mdbx_txn_commit and mdbx_txn_abort while it - * has active child transactions. - * [in] flags Special options for this transaction. This parameter - * must be set to 0 or by bitwise OR'ing together one or more - * of the values described here. - * - * - MDBX_RDONLY - * This transaction will not perform any write operations. - * - * - MDBX_TRYTXN - * Do not block when starting a write transaction - * - * [out] txn Address where the new MDBX_txn handle will be stored - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_PANIC - a fatal error occurred earlier and the environment - * must be shut down. - * - MDBX_MAP_RESIZED - another process wrote data beyond this MDBX_env's - * mapsize and this environment's map must be resized - * as well. See mdbx_env_set_mapsize(). - * - MDBX_READERS_FULL - a read-only transaction was requested and the reader - * lock table is full. See mdbx_env_set_maxreaders(). - * - MDBX_ENOMEM - out of memory. - * - MDBX_BUSY - a write transaction is already started. */ -LIBMDBX_API int mdbx_txn_begin(MDBX_env *env, MDBX_txn *parent, unsigned flags, - MDBX_txn **txn); - -/* Returns the transaction's MDBX_env - * - * [in] txn A transaction handle returned by mdbx_txn_begin() */ -LIBMDBX_API MDBX_env *mdbx_txn_env(MDBX_txn *txn); - -/* Return the transaction's ID. - * - * This returns the identifier associated with this transaction. For a - * read-only transaction, this corresponds to the snapshot being read; - * concurrent readers will frequently have the same transaction ID. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * - * Returns A transaction ID, valid if input is an active transaction. */ -LIBMDBX_API uint64_t mdbx_txn_id(MDBX_txn *txn); - -/* Commit all the operations of a transaction into the database. - * - * The transaction handle is freed. It and its cursors must not be used - * again after this call, except with mdbx_cursor_renew(). - * - * A cursor must be closed explicitly always, before - * or after its transaction ends. It can be reused with - * mdbx_cursor_renew() before finally closing it. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified. - * - MDBX_ENOSPC - no more disk space. - * - MDBX_EIO - a low-level I/O error occurred while writing. - * - MDBX_ENOMEM - out of memory. */ -LIBMDBX_API int mdbx_txn_commit(MDBX_txn *txn); - -/* Abandon all the operations of the transaction instead of saving them. - * - * The transaction handle is freed. It and its cursors must not be used - * again after this call, except with mdbx_cursor_renew(). - * - * A cursor must be closed explicitly always, before or after its transaction - * ends. It can be reused with mdbx_cursor_renew() before finally closing it. - * - * [in] txn A transaction handle returned by mdbx_txn_begin(). */ -LIBMDBX_API int mdbx_txn_abort(MDBX_txn *txn); - -/* Reset a read-only transaction. - * - * Abort the transaction like mdbx_txn_abort(), but keep the transaction - * handle. Therefore mdbx_txn_renew() may reuse the handle. This saves - * allocation overhead if the process will start a new read-only transaction - * soon, and also locking overhead if MDBX_NOTLS is in use. The reader table - * lock is released, but the table slot stays tied to its thread or - * MDBX_txn. Use mdbx_txn_abort() to discard a reset handle, and to free - * its lock table slot if MDBX_NOTLS is in use. - * - * Cursors opened within the transaction must not be used - * again after this call, except with mdbx_cursor_renew(). - * - * Reader locks generally don't interfere with writers, but they keep old - * versions of database pages allocated. Thus they prevent the old pages - * from being reused when writers commit new data, and so under heavy load - * the database size may grow much more rapidly than otherwise. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() */ -LIBMDBX_API int mdbx_txn_reset(MDBX_txn *txn); - -/* Renew a read-only transaction. - * - * This acquires a new reader lock for a transaction handle that had been - * released by mdbx_txn_reset(). It must be called before a reset transaction - * may be used again. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_PANIC - a fatal error occurred earlier and the environment - * must be shut down. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_txn_renew(MDBX_txn *txn); - -/* Open a table in the environment. - * - * A table handle denotes the name and parameters of a table, independently - * of whether such a table exists. The table handle may be discarded by - * calling mdbx_dbi_close(). The old table handle is returned if the table - * was already open. The handle may only be closed once. - * - * The table handle will be private to the current transaction until - * the transaction is successfully committed. If the transaction is - * aborted the handle will be closed automatically. - * After a successful commit the handle will reside in the shared - * environment, and may be used by other transactions. - * - * This function must not be called from multiple concurrent - * transactions in the same process. A transaction that uses - * this function must finish (either commit or abort) before - * any other transaction in the process may use this function. - * - * To use named table (with name != NULL), mdbx_env_set_maxdbs() - * must be called before opening the environment. Table names are - * keys in the internal unnamed table, and may be read but not written. - * - * [in] txn transaction handle returned by mdbx_txn_begin() - * [in] name The name of the table to open. If only a single - * table is needed in the environment, this value may be NULL. - * [in] flags Special options for this table. This parameter must be set - * to 0 or by bitwise OR'ing together one or more of the values - * described here: - * - MDBX_REVERSEKEY - * Keys are strings to be compared in reverse order, from the end - * of the strings to the beginning. By default, Keys are treated as - * strings and compared from beginning to end. - * - MDBX_DUPSORT - * Duplicate keys may be used in the table. Or, from another point of - * view, keys may have multiple data items, stored in sorted order. By - * default keys must be unique and may have only a single data item. - * - MDBX_INTEGERKEY - * Keys are binary integers in native byte order, either uin32_t or - * uint64_t, and will be sorted as such. The keys must all be of the - * same size. - * - MDBX_DUPFIXED - * This flag may only be used in combination with MDBX_DUPSORT. This - * option tells the library that the data items for this database are - * all the same size, which allows further optimizations in storage and - * retrieval. When all data items are the same size, the MDBX_GET_MULTIPLE, - * MDBX_NEXT_MULTIPLE and MDBX_PREV_MULTIPLE cursor operations may be used - * to retrieve multiple items at once. - * - MDBX_INTEGERDUP - * This option specifies that duplicate data items are binary integers, - * similar to MDBX_INTEGERKEY keys. - * - MDBX_REVERSEDUP - * This option specifies that duplicate data items should be compared as - * strings in reverse order (the comparison is performed in the direction - * from the last byte to the first). - * - MDBX_CREATE - * Create the named database if it doesn't exist. This option is not - * allowed in a read-only transaction or a read-only environment. - * - * [out] dbi Address where the new MDBX_dbi handle will be stored - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_NOTFOUND - the specified database doesn't exist in the - * environment and MDBX_CREATE was not specified. - * - MDBX_DBS_FULL - too many databases have been opened. - * See mdbx_env_set_maxdbs(). */ -LIBMDBX_API int mdbx_dbi_open_ex(MDBX_txn *txn, const char *name, - unsigned flags, MDBX_dbi *dbi, - MDBX_cmp_func *keycmp, MDBX_cmp_func *datacmp); -LIBMDBX_API int mdbx_dbi_open(MDBX_txn *txn, const char *name, unsigned flags, - MDBX_dbi *dbi); - -/* Retrieve statistics for a database. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [out] stat The address of an MDBX_stat structure where the statistics - * will be copied - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_dbi_stat(MDBX_txn *txn, MDBX_dbi dbi, MDBX_stat *stat, - size_t bytes); - -/* Retrieve the DB flags for a database handle. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [out] flags Address where the flags will be returned. - * [out] state Address where the state will be returned. - * - * Returns A non-zero error value on failure and 0 on success. */ -#define MDBX_TBL_DIRTY 0x01 /* DB was written in this txn */ -#define MDBX_TBL_STALE 0x02 /* Named-DB record is older than txnID */ -#define MDBX_TBL_NEW 0x04 /* Named-DB handle opened in this txn */ -LIBMDBX_API int mdbx_dbi_flags_ex(MDBX_txn *txn, MDBX_dbi dbi, unsigned *flags, - unsigned *state); -LIBMDBX_API int mdbx_dbi_flags(MDBX_txn *txn, MDBX_dbi dbi, unsigned *flags); - -/* Close a database handle. Normally unnecessary. - * - * Use with care: - * FIXME: This call is not mutex protected. Handles should only be closed by - * a single thread, and only if no other threads are going to reference - * the database handle or one of its cursors any further. Do not close - * a handle if an existing transaction has modified its database. - * Doing so can cause misbehavior from database corruption to errors - * like MDBX_BAD_VALSIZE (since the DB name is gone). - * - * Closing a database handle is not necessary, but lets mdbx_dbi_open() - * reuse the handle value. Usually it's better to set a bigger - * mdbx_env_set_maxdbs(), unless that value would be large. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] dbi A database handle returned by mdbx_dbi_open() - */ -LIBMDBX_API int mdbx_dbi_close(MDBX_env *env, MDBX_dbi dbi); - -/* Empty or delete+close a database. - * - * See mdbx_dbi_close() for restrictions about closing the DB handle. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [in] del 0 to empty the DB, 1 to delete it from the environment - * and close the DB handle. - * - * Returns A non-zero error value on failure and 0 on success. */ -LIBMDBX_API int mdbx_drop(MDBX_txn *txn, MDBX_dbi dbi, int del); - -/* Get items from a database. - * - * This function retrieves key/data pairs from the database. The address - * and length of the data associated with the specified key are returned - * in the structure to which data refers. - * If the database supports duplicate keys (MDBX_DUPSORT) then the - * first data item for the key will be returned. Retrieval of other - * items requires the use of mdbx_cursor_get(). - * - * NOTE: The memory pointed to by the returned values is owned by the - * database. The caller need not dispose of the memory, and may not - * modify it in any way. For values returned in a read-only transaction - * any modification attempts will cause a SIGSEGV. - * - * NOTE: Values returned from the database are valid only until a - * subsequent update operation, or the end of the transaction. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [in] key The key to search for in the database - * [in,out] data The data corresponding to the key - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_NOTFOUND - the key was not in the database. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_get(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, - MDBX_val *data); - -/* Store items into a database. - * - * This function stores key/data pairs in the database. The default behavior - * is to enter the new key/data pair, replacing any previously existing key - * if duplicates are disallowed, or adding a duplicate data item if - * duplicates are allowed (MDBX_DUPSORT). - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [in] key The key to store in the database - * [in,out] data The data to store - * [in] flags Special options for this operation. This parameter must be - * set to 0 or by bitwise OR'ing together one or more of the - * values described here. - * - * - MDBX_NODUPDATA - * Enter the new key/data pair only if it does not already appear - * in the database. This flag may only be specified if the database - * was opened with MDBX_DUPSORT. The function will return MDBX_KEYEXIST - * if the key/data pair already appears in the database. - * - * - MDBX_NOOVERWRITE - * Enter the new key/data pair only if the key does not already appear - * in the database. The function will return MDBX_KEYEXIST if the key - * already appears in the database, even if the database supports - * duplicates (MDBX_DUPSORT). The data parameter will be set to point - * to the existing item. - * - * - MDBX_CURRENT - * Update an single existing entry, but not add new ones. The function - * will return MDBX_NOTFOUND if the given key not exist in the database. - * Or the MDBX_EMULTIVAL in case duplicates for the given key. - * - * - MDBX_RESERVE - * Reserve space for data of the given size, but don't copy the given - * data. Instead, return a pointer to the reserved space, which the - * caller can fill in later - before the next update operation or the - * transaction ends. This saves an extra memcpy if the data is being - * generated later. MDBX does nothing else with this memory, the caller - * is expected to modify all of the space requested. This flag must not - * be specified if the database was opened with MDBX_DUPSORT. - * - * - MDBX_APPEND - * Append the given key/data pair to the end of the database. This option - * allows fast bulk loading when keys are already known to be in the - * correct order. Loading unsorted keys with this flag will cause - * a MDBX_EKEYMISMATCH error. - * - * - MDBX_APPENDDUP - * As above, but for sorted dup data. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_KEYEXIST - * - MDBX_MAP_FULL - the database is full, see mdbx_env_set_mapsize(). - * - MDBX_TXN_FULL - the transaction has too many dirty pages. - * - MDBX_EACCES - an attempt was made to write in a read-only transaction. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_put(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, - MDBX_val *data, unsigned flags); - -/* Delete items from a database. - * - * This function removes key/data pairs from the database. - * - * The data parameter is NOT ignored regardless the database does - * support sorted duplicate data items or not. If the data parameter - * is non-NULL only the matching data item will be deleted. - * - * This function will return MDBX_NOTFOUND if the specified key/data - * pair is not in the database. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [in] key The key to delete from the database - * [in] data The data to delete - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EACCES - an attempt was made to write in a read-only transaction. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_del(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, - MDBX_val *data); - -/* Create a cursor handle. - * - * A cursor is associated with a specific transaction and database. - * A cursor cannot be used when its database handle is closed. Nor - * when its transaction has ended, except with mdbx_cursor_renew(). - * It can be discarded with mdbx_cursor_close(). - * - * A cursor must be closed explicitly always, before - * or after its transaction ends. It can be reused with - * mdbx_cursor_renew() before finally closing it. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [out] cursor Address where the new MDBX_cursor handle will be stored - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_cursor_open(MDBX_txn *txn, MDBX_dbi dbi, - MDBX_cursor **cursor); - -/* Close a cursor handle. - * - * The cursor handle will be freed and must not be used again after this call. - * Its transaction must still be live if it is a write-transaction. - * - * [in] cursor A cursor handle returned by mdbx_cursor_open() */ -LIBMDBX_API void mdbx_cursor_close(MDBX_cursor *cursor); - -/* Renew a cursor handle. - * - * A cursor is associated with a specific transaction and database. - * Cursors that are only used in read-only transactions may be re-used, - * to avoid unnecessary malloc/free overhead. The cursor may be associated - * with a new read-only transaction, and referencing the same database handle - * as it was created with. - * - * This may be done whether the previous transaction is live or dead. - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] cursor A cursor handle returned by mdbx_cursor_open() - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_cursor_renew(MDBX_txn *txn, MDBX_cursor *cursor); - -/* Return the cursor's transaction handle. - * - * [in] cursor A cursor handle returned by mdbx_cursor_open() */ -LIBMDBX_API MDBX_txn *mdbx_cursor_txn(MDBX_cursor *cursor); - -/* Return the cursor's database handle. - * - * [in] cursor A cursor handle returned by mdbx_cursor_open() */ -LIBMDBX_API MDBX_dbi mdbx_cursor_dbi(MDBX_cursor *cursor); - -/* Retrieve by cursor. - * - * This function retrieves key/data pairs from the database. The address and - * length of the key are returned in the object to which key refers (except - * for the case of the MDBX_SET option, in which the key object is unchanged), - * and the address and length of the data are returned in the object to which - * data refers. See mdbx_get() for restrictions on using the output values. - * - * [in] cursor A cursor handle returned by mdbx_cursor_open() - * [in,out] key The key for a retrieved item - * [in,out] data The data of a retrieved item - * [in] op A cursor operation MDBX_cursor_op - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_NOTFOUND - no matching key found. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_cursor_get(MDBX_cursor *cursor, MDBX_val *key, - MDBX_val *data, MDBX_cursor_op op); - -/* Store by cursor. - * - * This function stores key/data pairs into the database. The cursor is - * positioned at the new item, or on failure usually near it. - * - * [in] cursor A cursor handle returned by mdbx_cursor_open() - * [in] key The key operated on. - * [in] data The data operated on. - * [in] flags Options for this operation. This parameter - * must be set to 0 or one of the values described here: - * - * - MDBX_CURRENT - * Replace the item at the current cursor position. The key parameter - * must still be provided, and must match it, otherwise the function - * return MDBX_EKEYMISMATCH. - * - * NOTE: MDBX unlike LMDB allows you to change the size of the data and - * automatically handles reordering for sorted duplicates (MDBX_DUPSORT). - * - * - MDBX_NODUPDATA - * Enter the new key/data pair only if it does not already appear in the - * database. This flag may only be specified if the database was opened - * with MDBX_DUPSORT. The function will return MDBX_KEYEXIST if the - * key/data pair already appears in the database. - * - * - MDBX_NOOVERWRITE - * Enter the new key/data pair only if the key does not already appear - * in the database. The function will return MDBX_KEYEXIST if the key - * already appears in the database, even if the database supports - * duplicates (MDBX_DUPSORT). - * - * - MDBX_RESERVE - * Reserve space for data of the given size, but don't copy the given - * data. Instead, return a pointer to the reserved space, which the - * caller can fill in later - before the next update operation or the - * transaction ends. This saves an extra memcpy if the data is being - * generated later. This flag must not be specified if the database - * was opened with MDBX_DUPSORT. - * - * - MDBX_APPEND - * Append the given key/data pair to the end of the database. No key - * comparisons are performed. This option allows fast bulk loading when - * keys are already known to be in the correct order. Loading unsorted - * keys with this flag will cause a MDBX_KEYEXIST error. - * - * - MDBX_APPENDDUP - * As above, but for sorted dup data. - * - * - MDBX_MULTIPLE - * Store multiple contiguous data elements in a single request. This flag - * may only be specified if the database was opened with MDBX_DUPFIXED. - * The data argument must be an array of two MDBX_vals. The iov_len of the - * first MDBX_val must be the size of a single data element. The iov_base - * of the first MDBX_val must point to the beginning of the array of - * contiguous data elements. The iov_len of the second MDBX_val must be - * the count of the number of data elements to store. On return this - * field will be set to the count of the number of elements actually - * written. The iov_base of the second MDBX_val is unused. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EKEYMISMATCH - * - MDBX_MAP_FULL - the database is full, see mdbx_env_set_mapsize(). - * - MDBX_TXN_FULL - the transaction has too many dirty pages. - * - MDBX_EACCES - an attempt was made to write in a read-only transaction. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_cursor_put(MDBX_cursor *cursor, MDBX_val *key, - MDBX_val *data, unsigned flags); - -/* Delete current key/data pair - * - * This function deletes the key/data pair to which the cursor refers. - * - * [in] cursor A cursor handle returned by mdbx_cursor_open() - * [in] flags Options for this operation. This parameter must be set to 0 - * or one of the values described here. - * - * - MDBX_NODUPDATA - * Delete all of the data items for the current key. This flag may only - * be specified if the database was opened with MDBX_DUPSORT. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EACCES - an attempt was made to write in a read-only transaction. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_cursor_del(MDBX_cursor *cursor, unsigned flags); - -/* Return count of duplicates for current key. - * - * This call is only valid on databases that support sorted duplicate data - * items MDBX_DUPSORT. - * - * [in] cursor A cursor handle returned by mdbx_cursor_open() - * [out] countp Address where the count will be stored - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EINVAL - cursor is not initialized, or an invalid parameter - * was specified. */ -LIBMDBX_API int mdbx_cursor_count(MDBX_cursor *cursor, size_t *countp); - -/* Compare two data items according to a particular database. - * - * This returns a comparison as if the two data items were keys in the - * specified database. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [in] a The first item to compare - * [in] b The second item to compare - * - * Returns < 0 if a < b, 0 if a == b, > 0 if a > b */ -LIBMDBX_API int mdbx_cmp(MDBX_txn *txn, MDBX_dbi dbi, const MDBX_val *a, - const MDBX_val *b); - -/* Compare two data items according to a particular database. - * - * This returns a comparison as if the two items were data items of - * the specified database. The database must have the MDBX_DUPSORT flag. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [in] a The first item to compare - * [in] b The second item to compare - * - * Returns < 0 if a < b, 0 if a == b, > 0 if a > b */ -LIBMDBX_API int mdbx_dcmp(MDBX_txn *txn, MDBX_dbi dbi, const MDBX_val *a, - const MDBX_val *b); - -/* A callback function used to print a message from the library. - * - * [in] msg The string to be printed. - * [in] ctx An arbitrary context pointer for the callback. - * - * Returns < 0 on failure, >= 0 on success. */ -typedef int(MDBX_msg_func)(const char *msg, void *ctx); - -/* Dump the entries in the reader lock table. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] func A MDBX_msg_func function - * [in] ctx Anything the message function needs - * - * Returns < 0 on failure, >= 0 on success. */ -LIBMDBX_API int mdbx_reader_list(MDBX_env *env, MDBX_msg_func *func, void *ctx); - -/* Check for stale entries in the reader lock table. - * - * [in] env An environment handle returned by mdbx_env_create() - * [out] dead Number of stale slots that were cleared - * - * Returns 0 on success, non-zero on failure. */ -LIBMDBX_API int mdbx_reader_check(MDBX_env *env, int *dead); - -LIBMDBX_API char *mdbx_dkey(const MDBX_val *key, char *const buf, - const size_t bufsize); - -LIBMDBX_API int mdbx_env_close_ex(MDBX_env *env, int dont_sync); - -/* Set threshold to force flush the data buffers to disk, - * even of MDBX_NOSYNC, MDBX_NOMETASYNC and MDBX_MAPASYNC flags - * in the environment. - * - * Data is always written to disk when mdbx_txn_commit() is called, - * but the operating system may keep it buffered. MDBX always flushes - * the OS buffers upon commit as well, unless the environment was - * opened with MDBX_NOSYNC or in part MDBX_NOMETASYNC. - * - * The default is 0, than mean no any threshold checked, and no additional - * flush will be made. - * - * [in] env An environment handle returned by mdbx_env_create() - * [in] bytes The size in bytes of summary changes when a synchronous - * flush would be made. - * - * Returns A non-zero error value on failure and 0 on success. */ -LIBMDBX_API int mdbx_env_set_syncbytes(MDBX_env *env, size_t bytes); - -/* Returns a lag of the reading for the given transaction. - * - * Returns an information for estimate how much given read-only - * transaction is lagging relative the to actual head. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [out] percent Percentage of page allocation in the database. - * - * Returns Number of transactions committed after the given was started for - * read, or -1 on failure. */ -LIBMDBX_API int mdbx_txn_straggler(MDBX_txn *txn, int *percent); - -/* A callback function for killing a laggard readers, - * but also could waiting ones. Called in case of MDBX_MAP_FULL error. - * - * [in] env An environment handle returned by mdbx_env_create(). - * [in] pid pid of the reader process. - * [in] tid thread_id of the reader thread. - * [in] txn Transaction number on which stalled. - * [in] gap A lag from the last commited txn. - * [in] retry A retry number, less that zero for notify end of OOM-loop. - * - * Returns -1 on failure (reader is not killed), - * 0 should wait or retry, - * 1 drop reader txn-lock (reading-txn was aborted), - * >1 drop reader registration (reader process was killed). */ -typedef int(MDBX_oom_func)(MDBX_env *env, int pid, mdbx_tid_t tid, uint64_t txn, - unsigned gap, int retry); - -/* Set the OOM callback. - * - * Callback will be called only on out-of-pages case for killing - * a laggard readers to allowing reclaiming of freeDB. - * - * [in] env An environment handle returned by mdbx_env_create(). - * [in] oomfunc A MDBX_oom_func function or NULL to disable. - * - * Returns A non-zero error value on failure and 0 on success. */ -LIBMDBX_API int mdbx_env_set_oomfunc(MDBX_env *env, MDBX_oom_func *oom_func); - -/* Get the current oom_func callback. - * - * Callback will be called only on out-of-pages case for killing - * a laggard readers to allowing reclaiming of freeDB. - * - * [in] env An environment handle returned by mdbx_env_create(). - * - * Returns A MDBX_oom_func function or NULL if disabled. */ -LIBMDBX_API MDBX_oom_func *mdbx_env_get_oomfunc(MDBX_env *env); - -#define MDBX_DBG_ASSERT 1 -#define MDBX_DBG_PRINT 2 -#define MDBX_DBG_TRACE 4 -#define MDBX_DBG_EXTRA 8 -#define MDBX_DBG_AUDIT 16 -#define MDBX_DBG_JITTER 32 -#define MDBX_DBG_DUMP 64 - -typedef void MDBX_debug_func(int type, const char *function, int line, - const char *msg, va_list args); - -LIBMDBX_API int mdbx_setup_debug(int flags, MDBX_debug_func *logger); - -typedef int MDBX_pgvisitor_func(uint64_t pgno, unsigned pgnumber, void *ctx, - const char *dbi, const char *type, - size_t nentries, size_t payload_bytes, - size_t header_bytes, size_t unused_bytes); -LIBMDBX_API int mdbx_env_pgwalk(MDBX_txn *txn, MDBX_pgvisitor_func *visitor, - void *ctx); - -typedef struct mdbx_canary { uint64_t x, y, z, v; } mdbx_canary; - -LIBMDBX_API int mdbx_canary_put(MDBX_txn *txn, const mdbx_canary *canary); -LIBMDBX_API int mdbx_canary_get(MDBX_txn *txn, mdbx_canary *canary); - -/* Returns: - * - MDBX_RESULT_TRUE - * when no more data available or cursor not positioned; - * - MDBX_RESULT_FALSE - * when data available; - * - Otherwise the error code. */ -LIBMDBX_API int mdbx_cursor_eof(MDBX_cursor *mc); - -/* Returns: MDBX_RESULT_TRUE, MDBX_RESULT_FALSE or Error code. */ -LIBMDBX_API int mdbx_cursor_on_first(MDBX_cursor *mc); - -/* Returns: MDBX_RESULT_TRUE, MDBX_RESULT_FALSE or Error code. */ -LIBMDBX_API int mdbx_cursor_on_last(MDBX_cursor *mc); - -LIBMDBX_API int mdbx_replace(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, - MDBX_val *new_data, MDBX_val *old_data, - unsigned flags); -/* Same as mdbx_get(), but: - * 1) if values_count is not NULL, then returns the count - * of multi-values/duplicates for a given key. - * 2) updates the key for pointing to the actual key's data inside DB. */ -LIBMDBX_API int mdbx_get_ex(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, - MDBX_val *data, size_t *values_count); - -LIBMDBX_API int mdbx_is_dirty(const MDBX_txn *txn, const void *ptr); - -LIBMDBX_API int mdbx_dbi_sequence(MDBX_txn *txn, MDBX_dbi dbi, uint64_t *result, - uint64_t increment); - -/*----------------------------------------------------------------------------*/ -/* attribute support functions for Nexenta */ -typedef uint_fast64_t mdbx_attr_t; - -/* Store by cursor with attribute. - * - * This function stores key/data pairs into the database. The cursor is - * positioned at the new item, or on failure usually near it. - * - * NOTE: Internally based on MDBX_RESERVE feature, - * therefore doesn't support MDBX_DUPSORT. - * - * [in] cursor A cursor handle returned by mdbx_cursor_open() - * [in] key The key operated on. - * [in] data The data operated on. - * [in] attr The attribute. - * [in] flags Options for this operation. This parameter must be set to 0 - * or one of the values described here: - * - * - MDBX_CURRENT - * Replace the item at the current cursor position. The key parameter - * must still be provided, and must match it, otherwise the function - * return MDBX_EKEYMISMATCH. - * - * - MDBX_APPEND - * Append the given key/data pair to the end of the database. No key - * comparisons are performed. This option allows fast bulk loading when - * keys are already known to be in the correct order. Loading unsorted - * keys with this flag will cause a MDBX_KEYEXIST error. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_EKEYMISMATCH - * - MDBX_MAP_FULL - the database is full, see mdbx_env_set_mapsize(). - * - MDBX_TXN_FULL - the transaction has too many dirty pages. - * - MDBX_EACCES - an attempt was made to write in a read-only transaction. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_cursor_put_attr(MDBX_cursor *cursor, MDBX_val *key, - MDBX_val *data, mdbx_attr_t attr, - unsigned flags); - -/* Store items and attributes into a database. - * - * This function stores key/data pairs in the database. The default behavior - * is to enter the new key/data pair, replacing any previously existing key - * if duplicates are disallowed. - * - * NOTE: Internally based on MDBX_RESERVE feature, - * therefore doesn't support MDBX_DUPSORT. - * - * [in] txn A transaction handle returned by mdbx_txn_begin(). - * [in] dbi A database handle returned by mdbx_dbi_open(). - * [in] key The key to store in the database. - * [in] attr The attribute to store in the database. - * [in,out] data The data to store. - * [in] flags Special options for this operation. This parameter must be - * set to 0 or by bitwise OR'ing together one or more of the - * values described here: - * - * - MDBX_NOOVERWRITE - * Enter the new key/data pair only if the key does not already appear - * in the database. The function will return MDBX_KEYEXIST if the key - * already appears in the database. The data parameter will be set to - * point to the existing item. - * - * - MDBX_CURRENT - * Update an single existing entry, but not add new ones. The function - * will return MDBX_NOTFOUND if the given key not exist in the database. - * Or the MDBX_EMULTIVAL in case duplicates for the given key. - * - * - MDBX_APPEND - * Append the given key/data pair to the end of the database. This option - * allows fast bulk loading when keys are already known to be in the - * correct order. Loading unsorted keys with this flag will cause - * a MDBX_EKEYMISMATCH error. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_KEYEXIST - * - MDBX_MAP_FULL - the database is full, see mdbx_env_set_mapsize(). - * - MDBX_TXN_FULL - the transaction has too many dirty pages. - * - MDBX_EACCES - an attempt was made to write in a read-only transaction. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_put_attr(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, - MDBX_val *data, mdbx_attr_t attr, unsigned flags); - -/* Set items attribute from a database. - * - * This function stores key/data pairs attribute to the database. - * - * NOTE: Internally based on MDBX_RESERVE feature, - * therefore doesn't support MDBX_DUPSORT. - * - * [in] txn A transaction handle returned by mdbx_txn_begin(). - * [in] dbi A database handle returned by mdbx_dbi_open(). - * [in] key The key to search for in the database. - * [in] data The data to be stored or NULL to save previous value. - * [in] attr The attribute to be stored. - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_NOTFOUND - the key-value pair was not in the database. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_set_attr(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, - MDBX_val *data, mdbx_attr_t attr); - -/* Get items attribute from a database cursor. - * - * This function retrieves key/data pairs from the database. The address and - * length of the key are returned in the object to which key refers (except - * for the case of the MDBX_SET option, in which the key object is unchanged), - * and the address and length of the data are returned in the object to which - * data refers. See mdbx_get() for restrictions on using the output values. - * - * [in] cursor A cursor handle returned by mdbx_cursor_open() - * [in,out] key The key for a retrieved item - * [in,out] data The data of a retrieved item - * [in] op A cursor operation MDBX_cursor_op - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_NOTFOUND - no matching key found. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_cursor_get_attr(MDBX_cursor *mc, MDBX_val *key, - MDBX_val *data, mdbx_attr_t *attrptr, - MDBX_cursor_op op); - -/* Get items attribute from a database. - * - * This function retrieves key/data pairs from the database. The address - * and length of the data associated with the specified key are returned - * in the structure to which data refers. - * If the database supports duplicate keys (MDBX_DUPSORT) then the - * first data item for the key will be returned. Retrieval of other - * items requires the use of mdbx_cursor_get(). - * - * NOTE: The memory pointed to by the returned values is owned by the - * database. The caller need not dispose of the memory, and may not - * modify it in any way. For values returned in a read-only transaction - * any modification attempts will cause a SIGSEGV. - * - * NOTE: Values returned from the database are valid only until a - * subsequent update operation, or the end of the transaction. - * - * [in] txn A transaction handle returned by mdbx_txn_begin() - * [in] dbi A database handle returned by mdbx_dbi_open() - * [in] key The key to search for in the database - * [in,out] data The data corresponding to the key - * - * Returns A non-zero error value on failure and 0 on success, some - * possible errors are: - * - MDBX_NOTFOUND - the key was not in the database. - * - MDBX_EINVAL - an invalid parameter was specified. */ -LIBMDBX_API int mdbx_get_attr(MDBX_txn *txn, MDBX_dbi dbi, MDBX_val *key, - MDBX_val *data, mdbx_attr_t *attrptr); - -#ifdef __cplusplus -} -#endif - -#endif /* LIBMDBX_H */ diff --git a/plugins/Dbx_mdb/src/mdbx/osal.c b/plugins/Dbx_mdb/src/mdbx/osal.c deleted file mode 100644 index 352929f98a..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/osal.c +++ /dev/null @@ -1,1002 +0,0 @@ -/* https://en.wikipedia.org/wiki/Operating_system_abstraction_layer */ - -/* - * Copyright 2015-2017 Leonid Yuriev <leo@yuriev.ru> - * and other libmdbx authors: please see AUTHORS file. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted only as authorized by the OpenLDAP - * Public License. - * - * A copy of this license is available in the file LICENSE in the - * top-level directory of the distribution or, alternatively, at - * <http://www.OpenLDAP.org/license.html>. - */ - -#include "./bits.h" - -#if defined(_WIN32) || defined(_WIN64) -#include <winternl.h> - -static int waitstatus2errcode(DWORD result) { - switch (result) { - case WAIT_OBJECT_0: - return MDBX_SUCCESS; - case WAIT_FAILED: - return GetLastError(); - case WAIT_ABANDONED: - return ERROR_ABANDONED_WAIT_0; - case WAIT_IO_COMPLETION: - return ERROR_USER_APC; - case WAIT_TIMEOUT: - return ERROR_TIMEOUT; - default: - return ERROR_UNHANDLED_ERROR; - } -} - -/* Map a result from an NTAPI call to WIN32 error code. */ -static int ntstatus2errcode(NTSTATUS status) { - DWORD dummy; - OVERLAPPED ov; - memset(&ov, 0, sizeof(ov)); - ov.Internal = status; - return GetOverlappedResult(NULL, &ov, &dummy, FALSE) ? MDBX_SUCCESS - : GetLastError(); -} - -/* We use native NT APIs to setup the memory map, so that we can - * let the DB file grow incrementally instead of always preallocating - * the full size. These APIs are defined in <wdm.h> and <ntifs.h> - * but those headers are meant for driver-level development and - * conflict with the regular user-level headers, so we explicitly - * declare them here. Using these APIs also means we must link to - * ntdll.dll, which is not linked by default in user code. */ -#pragma comment(lib, "ntdll.lib") - -extern NTSTATUS NTAPI NtCreateSection( - OUT PHANDLE SectionHandle, IN ACCESS_MASK DesiredAccess, - IN OPTIONAL POBJECT_ATTRIBUTES ObjectAttributes, - IN OPTIONAL PLARGE_INTEGER MaximumSize, IN ULONG SectionPageProtection, - IN ULONG AllocationAttributes, IN OPTIONAL HANDLE FileHandle); - -extern NTSTATUS NTAPI NtExtendSection(IN HANDLE SectionHandle, - IN PLARGE_INTEGER NewSectionSize); - -typedef enum _SECTION_INHERIT { ViewShare = 1, ViewUnmap = 2 } SECTION_INHERIT; - -extern NTSTATUS NTAPI NtMapViewOfSection( - IN HANDLE SectionHandle, IN HANDLE ProcessHandle, IN OUT PVOID *BaseAddress, - IN ULONG_PTR ZeroBits, IN SIZE_T CommitSize, - IN OUT OPTIONAL PLARGE_INTEGER SectionOffset, IN OUT PSIZE_T ViewSize, - IN SECTION_INHERIT InheritDisposition, IN ULONG AllocationType, - IN ULONG Win32Protect); - -extern NTSTATUS NTAPI NtUnmapViewOfSection(IN HANDLE ProcessHandle, - IN OPTIONAL PVOID BaseAddress); - -extern NTSTATUS NTAPI NtClose(HANDLE Handle); - -extern NTSTATUS NTAPI NtAllocateVirtualMemory( - IN HANDLE ProcessHandle, IN OUT PVOID *BaseAddress, IN ULONG ZeroBits, - IN OUT PULONG RegionSize, IN ULONG AllocationType, IN ULONG Protect); - -extern NTSTATUS NTAPI NtFreeVirtualMemory(IN HANDLE ProcessHandle, - IN PVOID *BaseAddress, - IN OUT PULONG RegionSize, - IN ULONG FreeType); - -#ifndef FILE_PROVIDER_CURRENT_VERSION -typedef struct _FILE_PROVIDER_EXTERNAL_INFO_V1 { - ULONG Version; - ULONG Algorithm; - ULONG Flags; -} FILE_PROVIDER_EXTERNAL_INFO_V1, *PFILE_PROVIDER_EXTERNAL_INFO_V1; -#endif - -#ifndef STATUS_OBJECT_NOT_EXTERNALLY_BACKED -#define STATUS_OBJECT_NOT_EXTERNALLY_BACKED ((NTSTATUS)0xC000046DL) -#endif -#ifndef STATUS_INVALID_DEVICE_REQUEST -#define STATUS_INVALID_DEVICE_REQUEST ((NTSTATUS)0xC0000010L) -#endif - -extern NTSTATUS -NtFsControlFile(IN HANDLE FileHandle, IN OUT HANDLE Event, - IN OUT PVOID /* PIO_APC_ROUTINE */ ApcRoutine, - IN OUT PVOID ApcContext, OUT PIO_STATUS_BLOCK IoStatusBlock, - IN ULONG FsControlCode, IN OUT PVOID InputBuffer, - IN ULONG InputBufferLength, OUT OPTIONAL PVOID OutputBuffer, - IN ULONG OutputBufferLength); - -#endif /* _WIN32 || _WIN64 */ - -/*----------------------------------------------------------------------------*/ - -#ifndef _MSC_VER -/* Prototype should match libc runtime. ISO POSIX (2003) & LSB 3.1 */ -__nothrow __noreturn void __assert_fail(const char *assertion, const char *file, - unsigned line, const char *function); -#else -__extern_C __declspec(dllimport) void __cdecl _assert(char const *message, - char const *filename, - unsigned line); -#endif /* _MSC_VER */ - -#ifndef mdbx_assert_fail -void __cold mdbx_assert_fail(const MDBX_env *env, const char *msg, - const char *func, int line) { -#if MDBX_DEBUG - if (env && env->me_assert_func) { - env->me_assert_func(env, msg, func, line); - return; - } -#else - (void)env; -#endif /* MDBX_DEBUG */ - - if (mdbx_debug_logger) - mdbx_debug_log(MDBX_DBG_ASSERT, func, line, "assert: %s\n", msg); -#ifndef _MSC_VER - __assert_fail(msg, "mdbx", line, func); -#else - _assert(msg, func, line); -#endif /* _MSC_VER */ -} -#endif /* mdbx_assert_fail */ - -__cold void mdbx_panic(const char *fmt, ...) { - va_list ap; - va_start(ap, fmt); -#ifdef _MSC_VER - if (IsDebuggerPresent()) { - OutputDebugStringA("\r\n" FIXME "\r\n"); - FatalExit(ERROR_UNHANDLED_ERROR); - } -#elif _XOPEN_SOURCE >= 700 || _POSIX_C_SOURCE >= 200809L || \ - (__GLIBC_PREREQ(1, 0) && !__GLIBC_PREREQ(2, 10) && defined(_GNU_SOURCE)) - vdprintf(STDERR_FILENO, fmt, ap); -#else -#error FIXME -#endif - va_end(ap); - abort(); -} - -/*----------------------------------------------------------------------------*/ - -#ifndef mdbx_asprintf -int mdbx_asprintf(char **strp, const char *fmt, ...) { - va_list ap, ones; - - va_start(ap, fmt); - va_copy(ones, ap); -#ifdef _MSC_VER - int needed = _vscprintf(fmt, ap); -#elif defined(vsnprintf) || defined(_BSD_SOURCE) || _XOPEN_SOURCE >= 500 || \ - defined(_ISOC99_SOURCE) || _POSIX_C_SOURCE >= 200112L - int needed = vsnprintf(nullptr, 0, fmt, ap); -#else -#error FIXME -#endif - va_end(ap); - - if (unlikely(needed < 0 || needed >= INT_MAX)) { - *strp = nullptr; - va_end(ones); - return needed; - } - - *strp = malloc(needed + 1); - if (unlikely(*strp == nullptr)) { - va_end(ones); - SetLastError(MDBX_ENOMEM); - return -1; - } - -#if defined(vsnprintf) || defined(_BSD_SOURCE) || _XOPEN_SOURCE >= 500 || \ - defined(_ISOC99_SOURCE) || _POSIX_C_SOURCE >= 200112L - int actual = vsnprintf(*strp, needed + 1, fmt, ones); -#else -#error FIXME -#endif - va_end(ones); - - assert(actual == needed); - if (unlikely(actual < 0)) { - free(*strp); - *strp = nullptr; - } - return actual; -} -#endif /* mdbx_asprintf */ - -#ifndef mdbx_memalign_alloc -int mdbx_memalign_alloc(size_t alignment, size_t bytes, void **result) { -#if _MSC_VER - *result = _aligned_malloc(bytes, alignment); - return *result ? MDBX_SUCCESS : MDBX_ENOMEM /* ERROR_OUTOFMEMORY */; -#elif __GLIBC_PREREQ(2, 16) || __STDC_VERSION__ >= 201112L - *result = memalign(alignment, bytes); - return *result ? MDBX_SUCCESS : errno; -#elif _POSIX_VERSION >= 200112L - *result = nullptr; - return posix_memalign(result, alignment, bytes); -#else -#error FIXME -#endif -} -#endif /* mdbx_memalign_alloc */ - -#ifndef mdbx_memalign_free -void mdbx_memalign_free(void *ptr) { -#if _MSC_VER - _aligned_free(ptr); -#else - free(ptr); -#endif -} -#endif /* mdbx_memalign_free */ - -/*----------------------------------------------------------------------------*/ - -int mdbx_condmutex_init(mdbx_condmutex_t *condmutex) { -#if defined(_WIN32) || defined(_WIN64) - int rc = MDBX_SUCCESS; - condmutex->event = NULL; - condmutex->mutex = CreateMutex(NULL, FALSE, NULL); - if (!condmutex->mutex) - return GetLastError(); - - condmutex->event = CreateEvent(NULL, FALSE, FALSE, NULL); - if (!condmutex->event) { - rc = GetLastError(); - (void)CloseHandle(condmutex->mutex); - condmutex->mutex = NULL; - } - return rc; -#else - memset(condmutex, 0, sizeof(mdbx_condmutex_t)); - int rc = pthread_mutex_init(&condmutex->mutex, NULL); - if (rc == 0) { - rc = pthread_cond_init(&condmutex->cond, NULL); - if (rc != 0) - (void)pthread_mutex_destroy(&condmutex->mutex); - } - return rc; -#endif -} - -static bool is_allzeros(const void *ptr, size_t bytes) { - const uint8_t *u8 = ptr; - for (size_t i = 0; i < bytes; ++i) - if (u8[i] != 0) - return false; - return true; -} - -int mdbx_condmutex_destroy(mdbx_condmutex_t *condmutex) { - int rc = MDBX_EINVAL; -#if defined(_WIN32) || defined(_WIN64) - if (condmutex->event) { - rc = CloseHandle(condmutex->event) ? MDBX_SUCCESS : GetLastError(); - if (rc == MDBX_SUCCESS) - condmutex->event = NULL; - } - if (condmutex->mutex) { - rc = CloseHandle(condmutex->mutex) ? MDBX_SUCCESS : GetLastError(); - if (rc == MDBX_SUCCESS) - condmutex->mutex = NULL; - } -#else - if (!is_allzeros(&condmutex->cond, sizeof(condmutex->cond))) { - rc = pthread_cond_destroy(&condmutex->cond); - if (rc == 0) - memset(&condmutex->cond, 0, sizeof(condmutex->cond)); - } - if (!is_allzeros(&condmutex->mutex, sizeof(condmutex->mutex))) { - rc = pthread_mutex_destroy(&condmutex->mutex); - if (rc == 0) - memset(&condmutex->mutex, 0, sizeof(condmutex->mutex)); - } -#endif - return rc; -} - -int mdbx_condmutex_lock(mdbx_condmutex_t *condmutex) { -#if defined(_WIN32) || defined(_WIN64) - DWORD code = WaitForSingleObject(condmutex->mutex, INFINITE); - return waitstatus2errcode(code); -#else - return pthread_mutex_lock(&condmutex->mutex); -#endif -} - -int mdbx_condmutex_unlock(mdbx_condmutex_t *condmutex) { -#if defined(_WIN32) || defined(_WIN64) - return ReleaseMutex(condmutex->mutex) ? MDBX_SUCCESS : GetLastError(); -#else - return pthread_mutex_unlock(&condmutex->mutex); -#endif -} - -int mdbx_condmutex_signal(mdbx_condmutex_t *condmutex) { -#if defined(_WIN32) || defined(_WIN64) - return SetEvent(condmutex->event) ? MDBX_SUCCESS : GetLastError(); -#else - return pthread_cond_signal(&condmutex->cond); -#endif -} - -int mdbx_condmutex_wait(mdbx_condmutex_t *condmutex) { -#if defined(_WIN32) || defined(_WIN64) - DWORD code = - SignalObjectAndWait(condmutex->mutex, condmutex->event, INFINITE, FALSE); - if (code == WAIT_OBJECT_0) - code = WaitForSingleObject(condmutex->mutex, INFINITE); - return waitstatus2errcode(code); -#else - return pthread_cond_wait(&condmutex->cond, &condmutex->mutex); -#endif -} - -/*----------------------------------------------------------------------------*/ - -int mdbx_fastmutex_init(mdbx_fastmutex_t *fastmutex) { -#if defined(_WIN32) || defined(_WIN64) - InitializeCriticalSection(fastmutex); - return MDBX_SUCCESS; -#else - return pthread_mutex_init(fastmutex, NULL); -#endif -} - -int mdbx_fastmutex_destroy(mdbx_fastmutex_t *fastmutex) { -#if defined(_WIN32) || defined(_WIN64) - DeleteCriticalSection(fastmutex); - return MDBX_SUCCESS; -#else - return pthread_mutex_destroy(fastmutex); -#endif -} - -int mdbx_fastmutex_acquire(mdbx_fastmutex_t *fastmutex) { -#if defined(_WIN32) || defined(_WIN64) - EnterCriticalSection(fastmutex); - return MDBX_SUCCESS; -#else - return pthread_mutex_lock(fastmutex); -#endif -} - -int mdbx_fastmutex_release(mdbx_fastmutex_t *fastmutex) { -#if defined(_WIN32) || defined(_WIN64) - LeaveCriticalSection(fastmutex); - return MDBX_SUCCESS; -#else - return pthread_mutex_unlock(fastmutex); -#endif -} - -/*----------------------------------------------------------------------------*/ - -int mdbx_openfile(const char *pathname, int flags, mode_t mode, - mdbx_filehandle_t *fd) { - *fd = INVALID_HANDLE_VALUE; -#if defined(_WIN32) || defined(_WIN64) - (void)mode; - - DWORD DesiredAccess; - DWORD ShareMode = FILE_SHARE_READ | FILE_SHARE_WRITE; - DWORD FlagsAndAttributes = FILE_ATTRIBUTE_NORMAL; - switch (flags & (O_RDONLY | O_WRONLY | O_RDWR)) { - default: - return ERROR_INVALID_PARAMETER; - case O_RDONLY: - DesiredAccess = GENERIC_READ; - break; - case O_WRONLY: /* assume for MDBX_env_copy() and friends output */ - DesiredAccess = GENERIC_WRITE; - ShareMode = 0; - FlagsAndAttributes |= FILE_FLAG_NO_BUFFERING | FILE_FLAG_WRITE_THROUGH; - break; - case O_RDWR: - DesiredAccess = GENERIC_READ | GENERIC_WRITE; - break; - } - - DWORD CreationDisposition; - switch (flags & (O_EXCL | O_CREAT)) { - default: - return ERROR_INVALID_PARAMETER; - case 0: - CreationDisposition = OPEN_EXISTING; - break; - case O_EXCL | O_CREAT: - CreationDisposition = CREATE_NEW; - FlagsAndAttributes |= FILE_ATTRIBUTE_NOT_CONTENT_INDEXED; - break; - case O_CREAT: - CreationDisposition = OPEN_ALWAYS; - FlagsAndAttributes |= FILE_ATTRIBUTE_NOT_CONTENT_INDEXED; - break; - } - - *fd = CreateFileA(pathname, DesiredAccess, ShareMode, NULL, - CreationDisposition, FlagsAndAttributes, NULL); - - if (*fd == INVALID_HANDLE_VALUE) - return GetLastError(); - if ((flags & O_CREAT) && GetLastError() != ERROR_ALREADY_EXISTS) { - /* set FILE_ATTRIBUTE_NOT_CONTENT_INDEXED for new file */ - DWORD FileAttributes = GetFileAttributesA(pathname); - if (FileAttributes == INVALID_FILE_ATTRIBUTES || - !SetFileAttributesA(pathname, FileAttributes | - FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)) { - int rc = GetLastError(); - CloseHandle(*fd); - *fd = INVALID_HANDLE_VALUE; - return rc; - } - } -#else - -#ifdef O_CLOEXEC - flags |= O_CLOEXEC; -#endif - *fd = open(pathname, flags, mode); - if (*fd < 0) - return errno; -#if defined(FD_CLOEXEC) && defined(F_GETFD) - flags = fcntl(*fd, F_GETFD); - if (flags >= 0) - (void)fcntl(*fd, F_SETFD, flags | FD_CLOEXEC); -#endif -#endif - return MDBX_SUCCESS; -} - -int mdbx_closefile(mdbx_filehandle_t fd) { -#if defined(_WIN32) || defined(_WIN64) - return CloseHandle(fd) ? MDBX_SUCCESS : GetLastError(); -#else - return (close(fd) == 0) ? MDBX_SUCCESS : errno; -#endif -} - -int mdbx_pread(mdbx_filehandle_t fd, void *buf, size_t bytes, uint64_t offset) { - if (bytes > MAX_WRITE) - return MDBX_EINVAL; -#if defined(_WIN32) || defined(_WIN64) - - OVERLAPPED ov; - ov.hEvent = 0; - ov.Offset = (DWORD)offset; - ov.OffsetHigh = HIGH_DWORD(offset); - - DWORD read = 0; - if (unlikely(!ReadFile(fd, buf, (DWORD)bytes, &read, &ov))) { - int rc = GetLastError(); - return (rc == MDBX_SUCCESS) ? /* paranoia */ ERROR_READ_FAULT : rc; - } -#else - STATIC_ASSERT_MSG(sizeof(off_t) >= sizeof(size_t), - "libmdbx requires 64-bit file I/O on 64-bit systems"); - intptr_t read = pread(fd, buf, bytes, offset); - if (read < 0) { - int rc = errno; - return (rc == MDBX_SUCCESS) ? /* paranoia */ MDBX_EIO : rc; - } -#endif - return (bytes == (size_t)read) ? MDBX_SUCCESS : MDBX_ENODATA; -} - -int mdbx_pwrite(mdbx_filehandle_t fd, const void *buf, size_t bytes, - uint64_t offset) { -#if defined(_WIN32) || defined(_WIN64) - if (bytes > MAX_WRITE) - return ERROR_INVALID_PARAMETER; - - OVERLAPPED ov; - ov.hEvent = 0; - ov.Offset = (DWORD)offset; - ov.OffsetHigh = HIGH_DWORD(offset); - - DWORD written; - if (likely(WriteFile(fd, buf, (DWORD)bytes, &written, &ov))) - return (bytes == written) ? MDBX_SUCCESS : MDBX_EIO /* ERROR_WRITE_FAULT */; - return GetLastError(); -#else - int rc; - intptr_t written; - do { - STATIC_ASSERT_MSG(sizeof(off_t) >= sizeof(size_t), - "libmdbx requires 64-bit file I/O on 64-bit systems"); - written = pwrite(fd, buf, bytes, offset); - if (likely(bytes == (size_t)written)) - return MDBX_SUCCESS; - rc = errno; - } while (rc == EINTR); - return (written < 0) ? rc : MDBX_EIO /* Use which error code (ENOSPC)? */; -#endif -} - -int mdbx_pwritev(mdbx_filehandle_t fd, struct iovec *iov, int iovcnt, - uint64_t offset, size_t expected_written) { -#if defined(_WIN32) || defined(_WIN64) - size_t written = 0; - for (int i = 0; i < iovcnt; ++i) { - int rc = mdbx_pwrite(fd, iov[i].iov_base, iov[i].iov_len, offset); - if (unlikely(rc != MDBX_SUCCESS)) - return rc; - written += iov[i].iov_len; - offset += iov[i].iov_len; - } - return (expected_written == written) ? MDBX_SUCCESS - : MDBX_EIO /* ERROR_WRITE_FAULT */; -#else - int rc; - intptr_t written; - do { - STATIC_ASSERT_MSG(sizeof(off_t) >= sizeof(size_t), - "libmdbx requires 64-bit file I/O on 64-bit systems"); - written = pwritev(fd, iov, iovcnt, offset); - if (likely(expected_written == (size_t)written)) - return MDBX_SUCCESS; - rc = errno; - } while (rc == EINTR); - return (written < 0) ? rc : MDBX_EIO /* Use which error code? */; -#endif -} - -int mdbx_write(mdbx_filehandle_t fd, const void *buf, size_t bytes) { -#ifdef SIGPIPE - sigset_t set, old; - sigemptyset(&set); - sigaddset(&set, SIGPIPE); - int rc = pthread_sigmask(SIG_BLOCK, &set, &old); - if (rc != 0) - return rc; -#endif - - const char *ptr = buf; - for (;;) { - size_t chunk = (MAX_WRITE < bytes) ? MAX_WRITE : bytes; -#if defined(_WIN32) || defined(_WIN64) - DWORD written; - if (unlikely(!WriteFile(fd, ptr, (DWORD)chunk, &written, NULL))) - return GetLastError(); -#else - intptr_t written = write(fd, ptr, chunk); - if (written < 0) { - int rc = errno; -#ifdef SIGPIPE - if (rc == EPIPE) { - /* Collect the pending SIGPIPE, otherwise at least OS X - * gives it to the process on thread-exit (ITS#8504). */ - int tmp; - sigwait(&set, &tmp); - written = 0; - continue; - } - pthread_sigmask(SIG_SETMASK, &old, NULL); -#endif - return rc; - } -#endif - if (likely(bytes == (size_t)written)) { -#ifdef SIGPIPE - pthread_sigmask(SIG_SETMASK, &old, NULL); -#endif - return MDBX_SUCCESS; - } - ptr += written; - bytes -= written; - } -} - -int mdbx_filesync(mdbx_filehandle_t fd, bool filesize_changed) { -#if defined(_WIN32) || defined(_WIN64) - (void)filesize_changed; - return FlushFileBuffers(fd) ? MDBX_SUCCESS : GetLastError(); -#elif __GLIBC_PREREQ(2, 16) || _BSD_SOURCE || _XOPEN_SOURCE || \ - (__GLIBC_PREREQ(2, 8) && _POSIX_C_SOURCE >= 200112L) - for (;;) { -/* LY: It is no reason to use fdatasync() here, even in case - * no such bug in a kernel. Because "no-bug" mean that a kernel - * internally do nearly the same, e.g. fdatasync() == fsync() - * when no-kernel-bug and file size was changed. - * - * So, this code is always safe and without appreciable - * performance degradation. - * - * For more info about of a corresponding fdatasync() bug - * see http://www.spinics.net/lists/linux-ext4/msg33714.html */ -#if _POSIX_C_SOURCE >= 199309L || _XOPEN_SOURCE >= 500 || \ - defined(_POSIX_SYNCHRONIZED_IO) - if (!filesize_changed && fdatasync(fd) == 0) - return MDBX_SUCCESS; -#else - (void)filesize_changed; -#endif - if (fsync(fd) == 0) - return MDBX_SUCCESS; - int rc = errno; - if (rc != EINTR) - return rc; - } -#else -#error FIXME -#endif -} - -int mdbx_filesize_sync(mdbx_filehandle_t fd) { -#if defined(_WIN32) || defined(_WIN64) - (void)fd; - /* Nothing on Windows (i.e. newer 100% steady) */ - return MDBX_SUCCESS; -#else - for (;;) { - if (fsync(fd) == 0) - return MDBX_SUCCESS; - int rc = errno; - if (rc != EINTR) - return rc; - } -#endif -} - -int mdbx_filesize(mdbx_filehandle_t fd, uint64_t *length) { -#if defined(_WIN32) || defined(_WIN64) - BY_HANDLE_FILE_INFORMATION info; - if (!GetFileInformationByHandle(fd, &info)) - return GetLastError(); - *length = info.nFileSizeLow | (uint64_t)info.nFileSizeHigh << 32; -#else - struct stat st; - - STATIC_ASSERT_MSG(sizeof(off_t) <= sizeof(uint64_t), - "libmdbx requires 64-bit file I/O on 64-bit systems"); - if (fstat(fd, &st)) - return errno; - - *length = st.st_size; -#endif - return MDBX_SUCCESS; -} - -int mdbx_ftruncate(mdbx_filehandle_t fd, uint64_t length) { -#if defined(_WIN32) || defined(_WIN64) - LARGE_INTEGER li; - li.QuadPart = length; - return (SetFilePointerEx(fd, li, NULL, FILE_BEGIN) && SetEndOfFile(fd)) - ? MDBX_SUCCESS - : GetLastError(); -#else - STATIC_ASSERT_MSG(sizeof(off_t) >= sizeof(size_t), - "libmdbx requires 64-bit file I/O on 64-bit systems"); - return ftruncate(fd, length) == 0 ? MDBX_SUCCESS : errno; -#endif -} - -/*----------------------------------------------------------------------------*/ - -int mdbx_thread_key_create(mdbx_thread_key_t *key) { -#if defined(_WIN32) || defined(_WIN64) - *key = TlsAlloc(); - return (*key != TLS_OUT_OF_INDEXES) ? MDBX_SUCCESS : GetLastError(); -#else - return pthread_key_create(key, mdbx_rthc_dtor); -#endif -} - -void mdbx_thread_key_delete(mdbx_thread_key_t key) { -#if defined(_WIN32) || defined(_WIN64) - mdbx_ensure(NULL, TlsFree(key)); -#else - mdbx_ensure(NULL, pthread_key_delete(key) == 0); -#endif -} - -void *mdbx_thread_rthc_get(mdbx_thread_key_t key) { -#if defined(_WIN32) || defined(_WIN64) - return TlsGetValue(key); -#else - return pthread_getspecific(key); -#endif -} - -void mdbx_thread_rthc_set(mdbx_thread_key_t key, const void *value) { -#if defined(_WIN32) || defined(_WIN64) - mdbx_ensure(NULL, TlsSetValue(key, (void *)value)); -#else - mdbx_ensure(NULL, pthread_setspecific(key, value) == 0); -#endif -} - -int mdbx_thread_create(mdbx_thread_t *thread, - THREAD_RESULT(THREAD_CALL *start_routine)(void *), - void *arg) { -#if defined(_WIN32) || defined(_WIN64) - *thread = CreateThread(NULL, 0, start_routine, arg, 0, NULL); - return *thread ? MDBX_SUCCESS : GetLastError(); -#else - return pthread_create(thread, NULL, start_routine, arg); -#endif -} - -int mdbx_thread_join(mdbx_thread_t thread) { -#if defined(_WIN32) || defined(_WIN64) - DWORD code = WaitForSingleObject(thread, INFINITE); - return waitstatus2errcode(code); -#else - void *unused_retval = &unused_retval; - return pthread_join(thread, &unused_retval); -#endif -} - -/*----------------------------------------------------------------------------*/ - -int mdbx_msync(mdbx_mmap_t *map, size_t offset, size_t length, int async) { - uint8_t *ptr = (uint8_t *)map->address + offset; -#if defined(_WIN32) || defined(_WIN64) - if (FlushViewOfFile(ptr, length) && (async || FlushFileBuffers(map->fd))) - return MDBX_SUCCESS; - return GetLastError(); -#else - const int mode = async ? MS_ASYNC : MS_SYNC; - return (msync(ptr, length, mode) == 0) ? MDBX_SUCCESS : errno; -#endif -} - -int mdbx_mmap(int flags, mdbx_mmap_t *map, size_t must, size_t limit) { - assert(must <= limit); -#if defined(_WIN32) || defined(_WIN64) - map->length = 0; - map->current = 0; - map->section = NULL; - map->address = nullptr; - - if (GetFileType(map->fd) != FILE_TYPE_DISK) - return ERROR_FILE_OFFLINE; - - FILE_REMOTE_PROTOCOL_INFO RemoteProtocolInfo; - if (GetFileInformationByHandleEx(map->fd, FileRemoteProtocolInfo, - &RemoteProtocolInfo, - sizeof(RemoteProtocolInfo))) { - if ((RemoteProtocolInfo.Flags & (REMOTE_PROTOCOL_INFO_FLAG_LOOPBACK | - REMOTE_PROTOCOL_INFO_FLAG_OFFLINE)) != - REMOTE_PROTOCOL_INFO_FLAG_LOOPBACK) - return ERROR_FILE_OFFLINE; - } - - NTSTATUS rc; -#if defined(_WIN64) && defined(WOF_CURRENT_VERSION) - struct { - WOF_EXTERNAL_INFO wof_info; - union { - WIM_PROVIDER_EXTERNAL_INFO wim_info; - FILE_PROVIDER_EXTERNAL_INFO_V1 file_info; - }; - size_t reserved_for_microsoft_madness[42]; - } GetExternalBacking_OutputBuffer; - IO_STATUS_BLOCK StatusBlock; - rc = NtFsControlFile(map->fd, NULL, NULL, NULL, &StatusBlock, - FSCTL_GET_EXTERNAL_BACKING, NULL, 0, - &GetExternalBacking_OutputBuffer, - sizeof(GetExternalBacking_OutputBuffer)); - if (rc != STATUS_OBJECT_NOT_EXTERNALLY_BACKED && - rc != STATUS_INVALID_DEVICE_REQUEST) - return NT_SUCCESS(rc) ? ERROR_FILE_OFFLINE : ntstatus2errcode(rc); -#endif - - WCHAR PathBuffer[INT16_MAX]; - DWORD VolumeSerialNumber, FileSystemFlags; - if (!GetVolumeInformationByHandleW(map->fd, PathBuffer, INT16_MAX, - &VolumeSerialNumber, NULL, - &FileSystemFlags, NULL, 0)) - return GetLastError(); - - if ((flags & MDBX_RDONLY) == 0) { - if (FileSystemFlags & (FILE_SEQUENTIAL_WRITE_ONCE | FILE_READ_ONLY_VOLUME | - FILE_VOLUME_IS_COMPRESSED)) - return ERROR_FILE_OFFLINE; - } - - if (!GetFinalPathNameByHandleW(map->fd, PathBuffer, INT16_MAX, - FILE_NAME_NORMALIZED | VOLUME_NAME_NT)) - return GetLastError(); - - if (_wcsnicmp(PathBuffer, L"\\Device\\Mup\\", 12) == 0) - return ERROR_FILE_OFFLINE; - - if (GetFinalPathNameByHandleW(map->fd, PathBuffer, INT16_MAX, - FILE_NAME_NORMALIZED | VOLUME_NAME_DOS)) { - UINT DriveType = GetDriveTypeW(PathBuffer); - if (DriveType == DRIVE_NO_ROOT_DIR && - wcsncmp(PathBuffer, L"\\\\?\\", 4) == 0 && - wcsncmp(PathBuffer + 5, L":\\", 2) == 0) { - PathBuffer[7] = 0; - DriveType = GetDriveTypeW(PathBuffer + 4); - } - switch (DriveType) { - case DRIVE_CDROM: - if (flags & MDBX_RDONLY) - break; - // fall through - case DRIVE_UNKNOWN: - case DRIVE_NO_ROOT_DIR: - case DRIVE_REMOTE: - default: - return ERROR_FILE_OFFLINE; - case DRIVE_REMOVABLE: - case DRIVE_FIXED: - case DRIVE_RAMDISK: - break; - } - } - - rc = NtCreateSection( - &map->section, - /* DesiredAccess */ SECTION_MAP_READ | SECTION_EXTEND_SIZE | - ((flags & MDBX_WRITEMAP) ? SECTION_MAP_WRITE : 0), - /* ObjectAttributes */ NULL, /* MaximumSize */ NULL, - /* SectionPageProtection */ (flags & MDBX_RDONLY) ? PAGE_READONLY - : PAGE_READWRITE, - /* AllocationAttributes */ SEC_RESERVE, map->fd); - - if (!NT_SUCCESS(rc)) - return ntstatus2errcode(rc); - - map->address = nullptr; - SIZE_T ViewSize = (flags & MDBX_RDONLY) ? must : limit; - rc = NtMapViewOfSection( - map->section, GetCurrentProcess(), &map->address, - /* ZeroBits */ 0, - /* CommitSize */ must, - /* SectionOffset */ NULL, &ViewSize, - /* InheritDisposition */ ViewUnmap, - /* AllocationType */ (flags & MDBX_RDONLY) ? 0 : MEM_RESERVE, - /* Win32Protect */ (flags & MDBX_WRITEMAP) ? PAGE_READWRITE - : PAGE_READONLY); - - if (!NT_SUCCESS(rc)) { - NtClose(map->section); - map->section = 0; - map->address = nullptr; - return ntstatus2errcode(rc); - } - assert(map->address != MAP_FAILED); - - uint64_t filesize; - rc = mdbx_filesize(map->fd, &filesize); - if (rc != MDBX_SUCCESS) { - NtClose(map->section); - NtUnmapViewOfSection(GetCurrentProcess(), map->address); - map->section = 0; - map->address = nullptr; - return rc; - } - - map->current = (must < filesize) ? must : (size_t)filesize; - map->length = ViewSize; - return MDBX_SUCCESS; -#else - (void)must; - map->address = mmap( - NULL, limit, (flags & MDBX_WRITEMAP) ? PROT_READ | PROT_WRITE : PROT_READ, - MAP_SHARED, map->fd, 0); - if (likely(map->address != MAP_FAILED)) { - map->length = limit; - return MDBX_SUCCESS; - } - map->length = 0; - map->address = nullptr; - return errno; -#endif -} - -int mdbx_munmap(mdbx_mmap_t *map) { -#if defined(_WIN32) || defined(_WIN64) - if (map->section) - NtClose(map->section); - NTSTATUS rc = NtUnmapViewOfSection(GetCurrentProcess(), map->address); - if (!NT_SUCCESS(rc)) - ntstatus2errcode(rc); - map->length = 0; - map->current = 0; - map->address = nullptr; -#else - if (unlikely(munmap(map->address, map->length))) - return errno; - map->length = 0; - map->address = nullptr; -#endif - return MDBX_SUCCESS; -} - -int mdbx_mresize(int flags, mdbx_mmap_t *map, size_t atleast, size_t limit) { - assert(atleast <= limit); -#if defined(_WIN32) || defined(_WIN64) - if (limit < map->length) { - /* Windows is unable shrinking a mapped section */ - return ERROR_USER_MAPPED_FILE; - } - if (limit > map->length) { - /* extend */ - LARGE_INTEGER new_size; - new_size.QuadPart = limit; - NTSTATUS rc = NtExtendSection(map->section, &new_size); - if (!NT_SUCCESS(rc)) - return ntstatus2errcode(rc); - map->length = limit; - } - if (atleast < map->current) { - /* Windows is unable shrinking a mapped file */ - uint8_t *ptr = (uint8_t *)map->address + atleast; - if (!VirtualFree(ptr, map->current - atleast, MEM_DECOMMIT)) - return MDBX_RESULT_TRUE; - - map->current = atleast; - int rc = mdbx_ftruncate(map->fd, atleast); - return (rc != MDBX_SUCCESS) ? MDBX_RESULT_TRUE : rc; - } - if (atleast > map->current) { - /* growth */ - uint8_t *ptr = (uint8_t *)map->address + map->current; - if (ptr != - VirtualAlloc(ptr, atleast - map->current, MEM_COMMIT, - (flags & MDBX_WRITEMAP) ? PAGE_READWRITE : PAGE_READONLY)) - return GetLastError(); - map->current = atleast; - } - - uint64_t filesize; - int rc = mdbx_filesize(map->fd, &filesize); - if (rc != MDBX_SUCCESS) - return rc; - if (filesize < atleast) { - rc = mdbx_ftruncate(map->fd, atleast); - if (rc != MDBX_SUCCESS) - return rc; - } - return MDBX_SUCCESS; -#else - (void)flags; - if (limit != map->length) { - void *ptr = mremap(map->address, map->length, limit, MREMAP_MAYMOVE); - if (ptr == MAP_FAILED) - return errno; - map->address = ptr; - map->length = limit; - } - return mdbx_ftruncate(map->fd, atleast); -#endif -} - -/*----------------------------------------------------------------------------*/ - -__cold void mdbx_osal_jitter(bool tiny) { - for (;;) { -#if defined(_M_IX86) || defined(_M_X64) || defined(__i386__) || \ - defined(__x86_64__) - const unsigned salt = 277u * (unsigned)__rdtsc(); -#else - const unsigned salt = rand(); -#endif - - const unsigned coin = salt % (tiny ? 29u : 43u); - if (coin < 43 / 3) - break; -#if defined(_WIN32) || defined(_WIN64) - SwitchToThread(); - if (coin > 43 * 2 / 3) - Sleep(1); -#else - sched_yield(); - if (coin > 43 * 2 / 3) - usleep(coin); -#endif - } -} diff --git a/plugins/Dbx_mdb/src/mdbx/osal.h b/plugins/Dbx_mdb/src/mdbx/osal.h deleted file mode 100644 index 871e52b0dc..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/osal.h +++ /dev/null @@ -1,645 +0,0 @@ -/* https://en.wikipedia.org/wiki/Operating_system_abstraction_layer */ - -/* - * Copyright 2015-2017 Leonid Yuriev <leo@yuriev.ru> - * and other libmdbx authors: please see AUTHORS file. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted only as authorized by the OpenLDAP - * Public License. - * - * A copy of this license is available in the file LICENSE in the - * top-level directory of the distribution or, alternatively, at - * <http://www.OpenLDAP.org/license.html>. - */ - -#pragma once - -/*----------------------------------------------------------------------------*/ -/* Microsoft compiler generates a lot of warning for self includes... */ - -#ifdef _MSC_VER -#pragma warning(push, 1) -#pragma warning(disable : 4548) /* expression before comma has no effect; \ - expected expression with side - effect */ -#pragma warning(disable : 4530) /* C++ exception handler used, but unwind \ - * semantics are not enabled. Specify /EHsc */ -#pragma warning(disable : 4577) /* 'noexcept' used with no exception handling \ - * mode specified; termination on exception is \ - * not guaranteed. Specify /EHsc */ -#if !defined(_CRT_SECURE_NO_WARNINGS) -#define _CRT_SECURE_NO_WARNINGS -#endif -#endif /* _MSC_VER (warnings) */ - -/*----------------------------------------------------------------------------*/ -/* C99 includes */ - -#include <inttypes.h> -#include <stddef.h> -#include <stdint.h> -#include <stdlib.h> - -#include <assert.h> -#include <fcntl.h> -#include <limits.h> -#include <malloc.h> -#include <stdio.h> -#include <string.h> -#include <time.h> - -#ifndef _POSIX_C_SOURCE -#ifdef _POSIX_SOURCE -#define _POSIX_C_SOURCE 1 -#else -#define _POSIX_C_SOURCE 0 -#endif -#endif - -#ifndef _XOPEN_SOURCE -#define _XOPEN_SOURCE 0 -#endif - -/*----------------------------------------------------------------------------*/ -/* Systems includes */ - -#if defined(_WIN32) || defined(_WIN64) -#include <windows.h> -#include <winnt.h> -#define HAVE_SYS_STAT_H -#define HAVE_SYS_TYPES_H -typedef HANDLE mdbx_thread_t; -typedef unsigned mdbx_thread_key_t; -#define MDBX_OSAL_SECTION HANDLE -#define MAP_FAILED NULL -#define HIGH_DWORD(v) ((DWORD)((sizeof(v) > 4) ? ((uint64_t)(v) >> 32) : 0)) -#define THREAD_CALL WINAPI -#define THREAD_RESULT DWORD -typedef struct { - HANDLE mutex; - HANDLE event; -} mdbx_condmutex_t; -typedef CRITICAL_SECTION mdbx_fastmutex_t; -#else -#include <pthread.h> -#include <signal.h> -#include <sys/file.h> -#include <sys/mman.h> -#include <sys/param.h> -#include <sys/stat.h> -#include <sys/uio.h> -#include <unistd.h> -typedef pthread_t mdbx_thread_t; -typedef pthread_key_t mdbx_thread_key_t; -#define INVALID_HANDLE_VALUE (-1) -#define THREAD_CALL -#define THREAD_RESULT void * -typedef struct { - pthread_mutex_t mutex; - pthread_cond_t cond; -} mdbx_condmutex_t; -typedef pthread_mutex_t mdbx_fastmutex_t; -#endif /* Platform */ - -#ifndef SSIZE_MAX -#define SSIZE_MAX INTPTR_MAX -#endif - -#ifdef HAVE_SYS_STAT_H -#include <sys/stat.h> -#endif -#ifdef HAVE_SYS_TYPES_H -#include <sys/types.h> -#endif -#ifdef HAVE_SYS_FILE_H -#include <sys/file.h> -#endif - -#if !defined(UNALIGNED_OK) -#if defined(__i386) || defined(__x86_64__) || defined(_M_IX86) || \ - defined(_M_X64) || defined(i386) || defined(_X86_) || defined(__i386__) || \ - defined(_X86_64_) -#define UNALIGNED_OK 1 -#else -#define UNALIGNED_OK 0 -#endif -#endif /* UNALIGNED_OK */ - -#if (-6 & 5) || CHAR_BIT != 8 || UINT_MAX < 0xffffffff || ULONG_MAX % 0xFFFF -#error \ - "Sanity checking failed: Two's complement, reasonably sized integer types" -#endif - -/*----------------------------------------------------------------------------*/ -/* Compiler's includes for builtins/intrinsics */ - -#ifdef _MSC_VER - -#if _MSC_FULL_VER < 190024215 -#if _MSC_FULL_VER < 180040629 && defined(_M_IX86) -#error Please use Visual Studio 2015 (MSC 19.0) or newer for 32-bit target. -#else -#pragma message( \ - "It is recommended to use Visual Studio 2015 (MSC 19.0) or newer.") -#endif -#endif - -#include <intrin.h> - -#elif __GNUC_PREREQ(4, 4) || defined(__clang__) -#if defined(__i386__) || defined(__x86_64__) -#include <cpuid.h> -#include <x86intrin.h> -#endif -#elif defined(__INTEL_COMPILER) -#include <intrin.h> -#elif defined(__SUNPRO_C) || defined(__sun) || defined(sun) -#include <mbarrier.h> -#elif (defined(_HPUX_SOURCE) || defined(__hpux) || defined(__HP_aCC)) && \ - (defined(HP_IA64) || defined(__ia64)) -#include <machine/sys/inline.h> -#elif defined(__IBMC__) && defined(__powerpc) -#include <atomic.h> -#elif defined(_AIX) -#include <builtins.h> -#include <sys/atomic_op.h> -#elif (defined(__osf__) && defined(__DECC)) || defined(__alpha) -#include <c_asm.h> -#include <machine/builtins.h> -#elif defined(__MWERKS__) -/* CodeWarrior - troubles ? */ -#pragma gcc_extensions -#elif defined(__SNC__) -/* Sony PS3 - troubles ? */ -#else -#error Unknown C compiler, please use GNU C 5.x or newer -#endif /* Compiler */ - -/*----------------------------------------------------------------------------*/ -/* Byteorder */ - -#if !defined(__BYTE_ORDER__) || !defined(__ORDER_LITTLE_ENDIAN__) || \ - !defined(__ORDER_BIG_ENDIAN__) - -#if defined(HAVE_ENDIAN_H) -#include <endian.h> -#elif defined(HAVE_SYS_PARAM_H) -#include <sys/param.h> /* for endianness */ -#elif defined(HAVE_NETINET_IN_H) && defined(HAVE_RESOLV_H) -#include <netinet/in.h> -#include <resolv.h> /* defines BYTE_ORDER on HPUX and Solaris */ -#endif - -#if defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && defined(__BIG_ENDIAN) -#define __ORDER_LITTLE_ENDIAN__ __LITTLE_ENDIAN -#define __ORDER_BIG_ENDIAN__ __BIG_ENDIAN -#define __BYTE_ORDER__ __BYTE_ORDER -#else -#define __ORDER_LITTLE_ENDIAN__ 1234 -#define __ORDER_BIG_ENDIAN__ 4321 -#if defined(__LITTLE_ENDIAN__) || defined(_LITTLE_ENDIAN) || \ - defined(__ARMEL__) || defined(__THUMBEL__) || defined(__AARCH64EL__) || \ - defined(__MIPSEL__) || defined(_MIPSEL) || defined(__MIPSEL) || \ - defined(__i386) || defined(__x86_64__) || defined(_M_IX86) || \ - defined(_M_X64) || defined(i386) || defined(_X86_) || defined(__i386__) || \ - defined(_X86_64_) || defined(_M_ARM) || defined(_M_ARM64) || \ - defined(__e2k__) -#define __BYTE_ORDER__ __ORDER_LITTLE_ENDIAN__ -#elif defined(__BIG_ENDIAN__) || defined(_BIG_ENDIAN) || defined(__ARMEB__) || \ - defined(__THUMBEB__) || defined(__AARCH64EB__) || defined(__MIPSEB__) || \ - defined(_MIPSEB) || defined(__MIPSEB) || defined(_M_IA64) -#define __BYTE_ORDER__ __ORDER_BIG_ENDIAN__ -#else -#error __BYTE_ORDER__ should be defined. -#endif -#endif -#endif /* __BYTE_ORDER__ || __ORDER_LITTLE_ENDIAN__ || __ORDER_BIG_ENDIAN__ */ - -#if __BYTE_ORDER__ != __ORDER_LITTLE_ENDIAN__ && \ - __BYTE_ORDER__ != __ORDER_BIG_ENDIAN__ -#error Unsupported byte order. -#endif - -/*----------------------------------------------------------------------------*/ -/* Memory/Compiler barriers, cache coherence */ - -static __inline void mdbx_compiler_barrier(void) { -#if defined(__clang__) || defined(__GNUC__) - __asm__ __volatile__("" ::: "memory"); -#elif defined(_MSC_VER) - _ReadWriteBarrier(); -#elif defined(__INTEL_COMPILER) /* LY: Intel Compiler may mimic GCC and MSC */ - __memory_barrier(); - if (type > MDBX_BARRIER_COMPILER) -#if defined(__ia64__) || defined(__ia64) || defined(_M_IA64) - __mf(); -#elif defined(__i386__) || defined(__x86_64__) - _mm_mfence(); -#else -#error "Unknown target for Intel Compiler, please report to us." -#endif -#elif defined(__SUNPRO_C) || defined(__sun) || defined(sun) - __compiler_barrier(); -#elif (defined(_HPUX_SOURCE) || defined(__hpux) || defined(__HP_aCC)) && \ - (defined(HP_IA64) || defined(__ia64)) - _Asm_sched_fence(/* LY: no-arg meaning 'all expect ALU', e.g. 0x3D3D */); -#elif defined(_AIX) || defined(__ppc__) || defined(__powerpc__) || \ - defined(__ppc64__) || defined(__powerpc64__) - __fence(); -#else -#error "Could not guess the kind of compiler, please report to us." -#endif -} - -static __inline void mdbx_memory_barrier(void) { -#if __has_extension(c_atomic) || __has_extension(cxx_atomic) - __c11_atomic_thread_fence(__ATOMIC_SEQ_CST); -#elif defined(__ATOMIC_SEQ_CST) - __atomic_thread_fence(__ATOMIC_SEQ_CST); -#elif defined(__clang__) || defined(__GNUC__) - __sync_synchronize(); -#elif defined(_MSC_VER) - MemoryBarrier(); -#elif defined(__INTEL_COMPILER) /* LY: Intel Compiler may mimic GCC and MSC */ -#if defined(__ia64__) || defined(__ia64) || defined(_M_IA64) - __mf(); -#elif defined(__i386__) || defined(__x86_64__) - _mm_mfence(); -#else -#error "Unknown target for Intel Compiler, please report to us." -#endif -#elif defined(__SUNPRO_C) || defined(__sun) || defined(sun) - __machine_rw_barrier(); -#elif (defined(_HPUX_SOURCE) || defined(__hpux) || defined(__HP_aCC)) && \ - (defined(HP_IA64) || defined(__ia64)) - _Asm_mf(); -#elif defined(_AIX) || defined(__ppc__) || defined(__powerpc__) || \ - defined(__ppc64__) || defined(__powerpc64__) - __lwsync(); -#else -#error "Could not guess the kind of compiler, please report to us." -#endif -} - -/*----------------------------------------------------------------------------*/ -/* Cache coherence and invalidation */ - -#if defined(__i386__) || defined(__x86_64__) || defined(_M_AMD64) || \ - defined(_M_IX86) || defined(__i386) || defined(__amd64) || \ - defined(i386) || defined(__x86_64) || defined(_AMD64_) || defined(_M_X64) -#define MDBX_CACHE_IS_COHERENT 1 -#elif defined(__hppa) || defined(__hppa__) -#define MDBX_CACHE_IS_COHERENT 1 -#endif - -#ifndef MDBX_CACHE_IS_COHERENT -#define MDBX_CACHE_IS_COHERENT 0 -#endif - -#ifndef MDBX_CACHELINE_SIZE -#if defined(SYSTEM_CACHE_ALIGNMENT_SIZE) -#define MDBX_CACHELINE_SIZE SYSTEM_CACHE_ALIGNMENT_SIZE -#elif defined(__ia64__) || defined(__ia64) || defined(_M_IA64) -#define MDBX_CACHELINE_SIZE 128 -#else -#define MDBX_CACHELINE_SIZE 64 -#endif -#endif /* MDBX_CACHELINE_SIZE */ - -#ifndef __cache_aligned -#define __cache_aligned __aligned(MDBX_CACHELINE_SIZE) -#endif - -#if MDBX_CACHE_IS_COHERENT -#define mdbx_coherent_barrier() mdbx_compiler_barrier() -#else -#define mdbx_coherent_barrier() mdbx_memory_barrier() -#endif - -#if defined(__mips) && defined(__linux) -/* Only MIPS has explicit cache control */ -#include <asm/cachectl.h> -#endif - -static __inline void mdbx_invalidate_cache(void *addr, size_t nbytes) { - mdbx_coherent_barrier(); -#if defined(__mips) && defined(__linux) - /* MIPS has cache coherency issues. - * Note: for any nbytes >= on-chip cache size, entire is flushed. */ - cacheflush(addr, nbytes, DCACHE); -#elif defined(_M_MRX000) || defined(_MIPS_) -#error "Sorry, cacheflush() for MIPS not implemented" -#else - /* LY: assume no relevant mmap/dcache issues. */ - (void)addr; - (void)nbytes; -#endif -} - -/*----------------------------------------------------------------------------*/ -/* libc compatibility stuff */ - -#ifndef mdbx_assert_fail -void mdbx_assert_fail(const MDBX_env *env, const char *msg, const char *func, - int line); -#endif /* mdbx_assert_fail */ - -#if __GLIBC_PREREQ(2, 1) -#define mdbx_asprintf asprintf -#else -int mdbx_asprintf(char **strp, const char *fmt, ...); -#endif - -#ifdef _MSC_VER - -#ifndef snprintf -#define snprintf(buffer, buffer_size, format, ...) \ - _snprintf_s(buffer, buffer_size, _TRUNCATE, format, __VA_ARGS__) -#endif /* snprintf */ - -#ifndef vsnprintf -#define vsnprintf(buffer, buffer_size, format, args) \ - _vsnprintf_s(buffer, buffer_size, _TRUNCATE, format, args) -#endif /* vsnprintf */ - -#ifdef _ASSERTE -#undef assert -#define assert _ASSERTE -#endif - -#endif /* _MSC_VER */ - -/*----------------------------------------------------------------------------*/ -/* OS abstraction layer stuff */ - -/* max bytes to write in one call */ -#define MAX_WRITE UINT32_C(0x3fff0000) - -/* Get the size of a memory page for the system. - * This is the basic size that the platform's memory manager uses, and is - * fundamental to the use of memory-mapped files. */ -static __inline size_t mdbx_syspagesize(void) { -#if defined(_WIN32) || defined(_WIN64) - SYSTEM_INFO si; - GetSystemInfo(&si); - return si.dwPageSize; -#else - return sysconf(_SC_PAGE_SIZE); -#endif -} - -static __inline char *mdbx_strdup(const char *str) { -#ifdef _MSC_VER - return _strdup(str); -#else - return strdup(str); -#endif -} - -static __inline int mdbx_get_errno(void) { -#if defined(_WIN32) || defined(_WIN64) - DWORD rc = GetLastError(); -#else - int rc = errno; -#endif - return rc; -} - -int mdbx_memalign_alloc(size_t alignment, size_t bytes, void **result); -void mdbx_memalign_free(void *ptr); - -int mdbx_condmutex_init(mdbx_condmutex_t *condmutex); -int mdbx_condmutex_lock(mdbx_condmutex_t *condmutex); -int mdbx_condmutex_unlock(mdbx_condmutex_t *condmutex); -int mdbx_condmutex_signal(mdbx_condmutex_t *condmutex); -int mdbx_condmutex_wait(mdbx_condmutex_t *condmutex); -int mdbx_condmutex_destroy(mdbx_condmutex_t *condmutex); - -int mdbx_fastmutex_init(mdbx_fastmutex_t *fastmutex); -int mdbx_fastmutex_acquire(mdbx_fastmutex_t *fastmutex); -int mdbx_fastmutex_release(mdbx_fastmutex_t *fastmutex); -int mdbx_fastmutex_destroy(mdbx_fastmutex_t *fastmutex); - -int mdbx_pwritev(mdbx_filehandle_t fd, struct iovec *iov, int iovcnt, - uint64_t offset, size_t expected_written); -int mdbx_pread(mdbx_filehandle_t fd, void *buf, size_t count, uint64_t offset); -int mdbx_pwrite(mdbx_filehandle_t fd, const void *buf, size_t count, - uint64_t offset); -int mdbx_write(mdbx_filehandle_t fd, const void *buf, size_t count); - -int mdbx_thread_create(mdbx_thread_t *thread, - THREAD_RESULT(THREAD_CALL *start_routine)(void *), - void *arg); -int mdbx_thread_join(mdbx_thread_t thread); -int mdbx_thread_key_create(mdbx_thread_key_t *key); -void mdbx_thread_key_delete(mdbx_thread_key_t key); -void *mdbx_thread_rthc_get(mdbx_thread_key_t key); -void mdbx_thread_rthc_set(mdbx_thread_key_t key, const void *value); - -int mdbx_filesync(mdbx_filehandle_t fd, bool fullsync); -int mdbx_filesize_sync(mdbx_filehandle_t fd); -int mdbx_ftruncate(mdbx_filehandle_t fd, uint64_t length); -int mdbx_filesize(mdbx_filehandle_t fd, uint64_t *length); -int mdbx_openfile(const char *pathname, int flags, mode_t mode, - mdbx_filehandle_t *fd); -int mdbx_closefile(mdbx_filehandle_t fd); - -typedef struct mdbx_mmap_param { - union { - void *address; - uint8_t *dxb; - struct MDBX_lockinfo *lck; - }; - mdbx_filehandle_t fd; - size_t length; /* mapping length, but NOT a size of file or DB */ -#if defined(_WIN32) || defined(_WIN64) - size_t current; /* mapped region size, e.g. file and DB */ -#endif -#ifdef MDBX_OSAL_SECTION - MDBX_OSAL_SECTION section; -#endif -} mdbx_mmap_t; - -int mdbx_mmap(int flags, mdbx_mmap_t *map, size_t must, size_t limit); -int mdbx_munmap(mdbx_mmap_t *map); -int mdbx_mresize(int flags, mdbx_mmap_t *map, size_t current, size_t wanna); -int mdbx_msync(mdbx_mmap_t *map, size_t offset, size_t length, int async); - -static __inline mdbx_pid_t mdbx_getpid(void) { -#if defined(_WIN32) || defined(_WIN64) - return GetCurrentProcessId(); -#else - return getpid(); -#endif -} - -static __inline mdbx_tid_t mdbx_thread_self(void) { -#if defined(_WIN32) || defined(_WIN64) - return GetCurrentThreadId(); -#else - return pthread_self(); -#endif -} - -void mdbx_osal_jitter(bool tiny); - -/*----------------------------------------------------------------------------*/ -/* lck stuff */ - -#if defined(_WIN32) || defined(_WIN64) -#undef MDBX_OSAL_LOCK -#define MDBX_OSAL_LOCK_SIGN UINT32_C(0xF10C) -#else -#define MDBX_OSAL_LOCK pthread_mutex_t -#define MDBX_OSAL_LOCK_SIGN UINT32_C(0x8017) -#endif - -int mdbx_lck_init(MDBX_env *env); - -int mdbx_lck_seize(MDBX_env *env); -int mdbx_lck_downgrade(MDBX_env *env, bool complete); -int mdbx_lck_upgrade(MDBX_env *env); -void mdbx_lck_destroy(MDBX_env *env); - -int mdbx_rdt_lock(MDBX_env *env); -void mdbx_rdt_unlock(MDBX_env *env); - -int mdbx_txn_lock(MDBX_env *env, bool dontwait); -void mdbx_txn_unlock(MDBX_env *env); - -int mdbx_rpid_set(MDBX_env *env); -int mdbx_rpid_clear(MDBX_env *env); - -/* Checks reader by pid. - * - * Returns: - * MDBX_RESULT_TRUE, if pid is live (unable to acquire lock) - * MDBX_RESULT_FALSE, if pid is dead (lock acquired) - * or otherwise the errcode. */ -int mdbx_rpid_check(MDBX_env *env, mdbx_pid_t pid); - -/*----------------------------------------------------------------------------*/ -/* Atomics */ - -#if !defined(__cplusplus) && (__STDC_VERSION__ >= 201112L) && \ - !defined(__STDC_NO_ATOMICS__) && \ - (__GNUC_PREREQ(4, 9) || __CLANG_PREREQ(3, 8) || \ - !(defined(__GNUC__) || defined(__clang__))) -#include <stdatomic.h> -#elif defined(__GNUC__) || defined(__clang__) -/* LY: nothing required */ -#elif defined(_MSC_VER) -#pragma warning(disable : 4163) /* 'xyz': not available as an intrinsic */ -#pragma warning(disable : 4133) /* 'function': incompatible types - from \ - 'size_t' to 'LONGLONG' */ -#pragma warning(disable : 4244) /* 'return': conversion from 'LONGLONG' to \ - 'std::size_t', possible loss of data */ -#pragma warning(disable : 4267) /* 'function': conversion from 'size_t' to \ - 'long', possible loss of data */ -#pragma intrinsic(_InterlockedExchangeAdd, _InterlockedCompareExchange) -#pragma intrinsic(_InterlockedExchangeAdd64, _InterlockedCompareExchange64) -#elif defined(__APPLE__) -#include <libkern/OSAtomic.h> -#else -#error FIXME atomic-ops -#endif - -static __inline uint32_t mdbx_atomic_add32(volatile uint32_t *p, uint32_t v) { -#if !defined(__cplusplus) && defined(ATOMIC_VAR_INIT) - assert(atomic_is_lock_free(p)); - return atomic_fetch_add((_Atomic uint32_t *)p, v); -#elif defined(__GNUC__) || defined(__clang__) - return __sync_fetch_and_add(p, v); -#else -#ifdef _MSC_VER - return _InterlockedExchangeAdd(p, v); -#endif -#ifdef __APPLE__ - return OSAtomicAdd32(v, (volatile int32_t *)p); -#endif -#endif -} - -static __inline uint64_t mdbx_atomic_add64(volatile uint64_t *p, uint64_t v) { -#if !defined(__cplusplus) && defined(ATOMIC_VAR_INIT) - assert(atomic_is_lock_free(p)); - return atomic_fetch_add((_Atomic uint64_t *)p, v); -#elif defined(__GNUC__) || defined(__clang__) - return __sync_fetch_and_add(p, v); -#else -#ifdef _MSC_VER - return _InterlockedExchangeAdd64((volatile int64_t *)p, v); -#endif -#ifdef __APPLE__ - return OSAtomicAdd64(v, (volatile int64_t *)p); -#endif -#endif -} - -#define mdbx_atomic_sub32(p, v) mdbx_atomic_add32(p, 0 - (v)) -#define mdbx_atomic_sub64(p, v) mdbx_atomic_add64(p, 0 - (v)) - -static __inline bool mdbx_atomic_compare_and_swap32(volatile uint32_t *p, - uint32_t c, uint32_t v) { -#if !defined(__cplusplus) && defined(ATOMIC_VAR_INIT) - assert(atomic_is_lock_free(p)); - return atomic_compare_exchange_strong((_Atomic uint32_t *)p, &c, v); -#elif defined(__GNUC__) || defined(__clang__) - return __sync_bool_compare_and_swap(p, c, v); -#else -#ifdef _MSC_VER - return c == _InterlockedCompareExchange(p, v, c); -#endif -#ifdef __APPLE__ - return c == OSAtomicCompareAndSwap32Barrier(c, v, (volatile int32_t *)p); -#endif -#endif -} - -static __inline bool mdbx_atomic_compare_and_swap64(volatile uint64_t *p, - uint64_t c, uint64_t v) { -#if !defined(__cplusplus) && defined(ATOMIC_VAR_INIT) - assert(atomic_is_lock_free(p)); - return atomic_compare_exchange_strong((_Atomic uint64_t *)p, &c, v); -#elif defined(__GNUC__) || defined(__clang__) - return __sync_bool_compare_and_swap(p, c, v); -#else -#ifdef _MSC_VER - return c == _InterlockedCompareExchange64((volatile int64_t *)p, v, c); -#endif -#ifdef __APPLE__ - return c == OSAtomicCompareAndSwap64Barrier(c, v, (volatile uint64_t *)p); -#endif -#endif -} - -/*----------------------------------------------------------------------------*/ - -#if defined(_MSC_VER) && _MSC_VER >= 1900 && _MSC_VER < 1920 -/* LY: MSVC 2015/2017 has buggy/inconsistent PRIuPTR/PRIxPTR macros - * for internal format-args checker. */ -#undef PRIuPTR -#undef PRIiPTR -#undef PRIdPTR -#undef PRIxPTR -#define PRIuPTR "Iu" -#define PRIiPTR "Ii" -#define PRIdPTR "Id" -#define PRIxPTR "Ix" -#define PRIuSIZE "zu" -#define PRIiSIZE "zi" -#define PRIdSIZE "zd" -#define PRIxSIZE "zx" -#endif /* fix PRI*PTR for _MSC_VER */ - -#ifndef PRIuSIZE -#define PRIuSIZE PRIuPTR -#define PRIiSIZE PRIiPTR -#define PRIdSIZE PRIdPTR -#define PRIxSIZE PRIxPTR -#endif /* PRI*SIZE macros for MSVC */ - -#ifdef _MSC_VER -#pragma warning(pop) -#endif diff --git a/plugins/Dbx_mdb/src/mdbx/version.c b/plugins/Dbx_mdb/src/mdbx/version.c deleted file mode 100644 index ef4f99088d..0000000000 --- a/plugins/Dbx_mdb/src/mdbx/version.c +++ /dev/null @@ -1,34 +0,0 @@ -/* - * Copyright 2015-2017 Leonid Yuriev <leo@yuriev.ru> - * and other libmdbx authors: please see AUTHORS file. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted only as authorized by the OpenLDAP - * Public License. - * - * A copy of this license is available in the file LICENSE in the - * top-level directory of the distribution or, alternatively, at - * <http://www.OpenLDAP.org/license.html>. - */ - -#include "./bits.h" - -#if MDBX_VERSION_MAJOR != 0 || MDBX_VERSION_MINOR != 0 -#error "API version mismatch!" -#endif - -#define MDBX_VERSION_RELEASE 0 -#define MDBX_VERSION_REVISION 0 - -/*LIBMDBX_EXPORTS*/ const mdbx_version_info mdbx_version = { - MDBX_VERSION_MAJOR, - MDBX_VERSION_MINOR, - MDBX_VERSION_RELEASE, - MDBX_VERSION_REVISION, - {"@MDBX_GIT_TIMESTAMP@", "@MDBX_GIT_TREE@", "@MDBX_GIT_COMMIT@", - "@MDBX_GIT_DESCRIBE@"}}; - -/*LIBMDBX_EXPORTS*/ const mdbx_build_info mdbx_build = { - "@MDBX_BUILD_TIMESTAMP@", "@MDBX_BUILD_TAGRET@", "@MDBX_BUILD_OPTIONS@", - "@MDBX_BUILD_COMPILER@", "@MDBX_BUILD_FLAGS@"}; |