mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-01-16 03:52:35 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions
mariadb/storage/maria/ma_locking.c
Monty a6bf4b5807 MDEV-29693 ANALYZE TABLE still flushes table definition cache when engine-independent statistics is used 
This commits enables reloading of engine-independent statistics
without flushing the table from table definition cache.

This is achieved by allowing multiple version of the
TABLE_STATISTICS_CB object and having independent pointers to it in
TABLE and TABLE_SHARE.  The TABLE_STATISTICS_CB object have reference
pointers and are freed when no one is pointing to it anymore.

TABLE's TABLE_STATISTICS_CB pointer is updated to use the
TABLE_SHARE's pointer when read_statistics_for_tables() is called at
the beginning of a query.

Main changes:
- read_statistics_for_table() will allocate an new TABLE_STATISTICS_CB
  object.
- All get_stat_values() functions has a new parameter that tells
  where collected data should be stored. get_stat_values() are not
  using the table_field object anymore to store data.
- All get_stat_values() functions returns 1 if they found any
  data in the statistics tables.

Other things:
- Fixed INSERT DELAYED to not read statistics tables.
- Removed Statistics_state from TABLE_STATISTICS_CB as this is not
  needed anymore as wer are not changing TABLE_SHARE->stats_cb while
  calculating or loading statistics.
- Store values used with store_from_statistical_minmax_field() in
  TABLE_STATISTICS_CB::mem_root. This allowed me to remove the function
  delete_stat_values_for_table_share().
  - Field_blob::store_from_statistical_minmax_field() is implemented
    but is not normally used as we do not yet support EIS statistics
    for blobs. For example Field_blob::update_min() and
    Field_blob::update_max() are not implemented.
    Note that the function can be called if there is an concurrent
    "ALTER TABLE MODIFY field BLOB" running because of a bug in
    ALTER TABLE where it deletes entries from column_stats
    before it has an exclusive lock on the table.
- Use result of field->val_str(&val) as a pointer to the result
  instead of val (safetly fix).
- Allocate memory for collected statistics in THD::mem_root, not in
  in TABLE::mem_root. This could cause the TABLE object to grow if a
  ANALYZE TABLE was run many times on the same table.
  This was done in allocate_statistics_for_table(),
  create_min_max_statistical_fields_for_table() and
  create_min_max_statistical_fields_for_table_share().
- Store in TABLE_STATISTICS_CB::stats_available which statistics was
  found in the statistics tables.
- Removed index_table from class Index_prefix_calc as it was not used.
- Added TABLE_SHARE::LOCK_statistics to ensure we don't load EITS
  in parallel. First thread will load it, others will reuse the
  loaded data.
- Eliminate read_histograms_for_table(). The loading happens within
  read_statistics_for_tables() if histograms are needed.
  One downside is that if we have read statistics without histograms
  before and someone requires histograms, we have to read all statistics
  again (once) from the statistics tables.
  A smaller downside is the need to call alloc_root() for each
  individual histogram. Before we could allocate all the space for
  histograms with a single alloc_root.
- Fixed bug in MyISAM and Aria where they did not properly notice
  that table had changed after analyze table. This was not a problem
  before this patch as then the MyISAM and Aria tables where flushed
  as part of ANALYZE table which did hide this issue.
- Fixed a bug in ANALYZE table where table->records could be seen as 0
  in collect_statistics_for_table(). The effect of this unlikely bug
  was that a full table scan could be done even if
  analyze_sample_percentage was not set to 1.
- Changed multiple mallocs in a row to use multi_alloc_root().
- Added a mutex protection in update_statistics_for_table() to ensure
  that several tables are not updating the statistics at the same time.

Some of the changes in sql_statistics.cc are based on a patch from
Oleg Smirnov <olernov@gmail.com>

Co-authored-by: Oleg Smirnov <olernov@gmail.com>
Co-authored-by: Vicentiu Ciorbaru <cvicentiu@gmail.com>
Reviewer: Sergei Petrunia <sergey@mariadb.com>
2023-08-18 13:28:39 +03:00

607 lines
19 KiB
C
Raw Blame History

/* Copyright (C) 2006 MySQL AB & MySQL Finland AB & TCX DataKonsult AB
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1335 USA */
/*
Locking of Maria-tables.
Must be first request before doing any furter calls to any Maria function.
Is used to allow many process use the same non transactional Maria table
*/
#include "ma_ftdefs.h"
/* lock table by F_UNLCK, F_RDLCK or F_WRLCK */
int maria_lock_database(MARIA_HA *info, int lock_type)
{
int error;
uint count;
MARIA_SHARE *share= info->s;
DBUG_ENTER("maria_lock_database");
DBUG_PRINT("enter",("lock_type: %d old lock %d r_locks: %u w_locks: %u "
"global_changed: %d open_count: %u name: '%s'",
lock_type, info->lock_type, share->r_locks,
share->w_locks,
share->global_changed, share->state.open_count,
share->index_file_name.str));
if (share->options & HA_OPTION_READ_ONLY_DATA ||
info->lock_type == lock_type)
DBUG_RETURN(0);
if (lock_type == F_EXTRA_LCK) /* Used by TMP tables */
{
++share->w_locks;
++share->tot_locks;
info->lock_type= lock_type;
DBUG_RETURN(0);
}
error=0;
if (!info->intern_lock_locked)
mysql_mutex_lock(&share->intern_lock);
if (share->kfile.file >= 0) /* May only be false on windows */
{
switch (lock_type) {
case F_UNLCK:
maria_ftparser_call_deinitializer(info);
if (info->lock_type == F_RDLCK)
{
count= --share->r_locks;
if (share->lock_restore_status)
(*share->lock_restore_status)(info);
}
else
{
count= --share->w_locks;
if (share->lock.update_status)
_ma_update_status_with_lock(info);
}
--share->tot_locks;
if (info->lock_type == F_WRLCK && !share->w_locks)
{
/* pages of transactional tables get flushed at Checkpoint */
if (!share->base.born_transactional && !share->temporary &&
_ma_flush_table_files(info,
share->delay_key_write ? MARIA_FLUSH_DATA :
MARIA_FLUSH_DATA | MARIA_FLUSH_INDEX,
FLUSH_KEEP, FLUSH_KEEP))
error= my_errno;
}
if (info->opt_flag & (READ_CACHE_USED | WRITE_CACHE_USED))
{
if (end_io_cache(&info->rec_cache))
{
error= my_errno;
_ma_set_fatal_error(info, error);
}
}
if (!count)
{
DBUG_PRINT("info",("changed: %u w_locks: %u",
(uint) share->changed, share->w_locks));
if (share->changed && !share->w_locks)
{
#ifdef HAVE_MMAP
if ((share->mmaped_length !=
share->state.state.data_file_length) &&
(share->nonmmaped_inserts > MAX_NONMAPPED_INSERTS))
{
if (share->lock_key_trees)
mysql_rwlock_wrlock(&share->mmap_lock);
_ma_remap_file(info, share->state.state.data_file_length);
share->nonmmaped_inserts= 0;
if (share->lock_key_trees)
mysql_rwlock_unlock(&share->mmap_lock);
}
#endif
#ifdef MARIA_EXTERNAL_LOCKING
share->state.process= share->last_process=share->this_process;
share->state.unique= info->last_unique= info->this_unique;
share->state.update_count= info->last_loop= ++info->this_loop;
#endif
/* transactional tables rather flush their state at Checkpoint */
if (!share->base.born_transactional)
{
if (_ma_state_info_write_sub(share->kfile.file, &share->state,
MA_STATE_INFO_WRITE_DONT_MOVE_OFFSET))
error= my_errno;
else
{
/* A value of 0 means below means "state flushed" */
share->changed= 0;
}
}
if (maria_flush)
{
if (_ma_sync_table_files(info))
error= my_errno;
}
else
share->not_flushed=1;
if (error)
_ma_set_fatal_error(info, error);
}
}
info->opt_flag&= ~(READ_CACHE_USED | WRITE_CACHE_USED);
info->lock_type= F_UNLCK;
break;
case F_RDLCK:
if (info->lock_type == F_WRLCK)
{
/*
Change RW to READONLY
mysqld does not turn write locks to read locks,
so we're never here in mysqld.
*/
share->w_locks--;
share->r_locks++;
info->lock_type=lock_type;
break;
}
#ifdef MARIA_EXTERNAL_LOCKING
if (!share->r_locks && !share->w_locks)
{
/* note that a transactional table should not do this */
if (_ma_state_info_read_dsk(share->kfile.file, &share->state))
{
error=my_errno;
break;
}
}
#endif
_ma_test_if_changed(info);
share->r_locks++;
share->tot_locks++;
info->lock_type=lock_type;
break;
case F_WRLCK:
if (info->lock_type == F_RDLCK)
{ /* Change READONLY to RW */
if (share->r_locks == 1)
{
share->r_locks--;
share->w_locks++;
info->lock_type=lock_type;
break;
}
}
#ifdef MARIA_EXTERNAL_LOCKING
if (!(share->options & HA_OPTION_READ_ONLY_DATA))
{
if (!share->w_locks)
{
if (!share->r_locks)
{
/*
Note that transactional tables should not do this.
If we enabled this code, we should make sure to skip it if
born_transactional is true. We should not test
now_transactional to decide if we can call
_ma_state_info_read_dsk(), because it can temporarily be 0
(TRUNCATE on a partitioned table) and thus it would make a state
modification below without mutex, confusing a concurrent
checkpoint running.
Even if this code was enabled only for non-transactional tables:
in scenario LOCK TABLE t1 WRITE; INSERT INTO t1; DELETE FROM t1;
state on disk read by DELETE is obsolete as it was not flushed
at the end of INSERT. MyISAM same. It however causes no issue as
maria_delete_all_rows() calls _ma_reset_status() thus is not
influenced by the obsolete read values.
*/
if (_ma_state_info_read_dsk(share->kfile.file, &share->state))
{
error=my_errno;
break;
}
}
}
}
#endif /* defined(MARIA_EXTERNAL_LOCKING) */
_ma_test_if_changed(info);
info->lock_type=lock_type;
info->invalidator=share->invalidator;
share->w_locks++;
share->tot_locks++;
break;
default:
DBUG_ASSERT(0);
break; /* Impossible */
}
}
#ifdef _WIN32
else
{
/*
Check for bad file descriptors if this table is part
of a merge union. Failing to capture this may cause
a crash on windows if the table is renamed and
later on referenced by the merge table.
*/
if( info->owned_by_merge && (info->s)->kfile.file < 0 )
{
error = HA_ERR_NO_SUCH_TABLE;
}
}
#endif
if (!info->intern_lock_locked)
mysql_mutex_unlock(&share->intern_lock);
DBUG_RETURN(error);
} /* maria_lock_database */
/****************************************************************************
** functions to read / write the state
****************************************************************************/
int _ma_readinfo(register MARIA_HA *info __attribute__ ((unused)),
int lock_type __attribute__ ((unused)),
int check_keybuffer __attribute__ ((unused)))
{
#ifdef MARIA_EXTERNAL_LOCKING
DBUG_ENTER("_ma_readinfo");
if (info->lock_type == F_UNLCK)
{
MARIA_SHARE *share= info->s;
if (!share->tot_locks)
{
/* should not be done for transactional tables */
if (_ma_state_info_read_dsk(share->kfile.file, &share->state))
{
if (!my_errno)
my_errno= HA_ERR_FILE_TOO_SHORT;
DBUG_RETURN(1);
}
}
if (check_keybuffer)
VOID(_ma_test_if_changed(info));
info->invalidator=share->invalidator;
}
else if (lock_type == F_WRLCK && info->lock_type == F_RDLCK)
{
my_errno=EACCES; /* Not allowed to change */
DBUG_RETURN(-1); /* when have read_lock() */
}
DBUG_RETURN(0);
#else
return 0;
#endif /* defined(MARIA_EXTERNAL_LOCKING) */
} /* _ma_readinfo */
/*
Every isam-function that updates the isam-database MUST end with this
request
NOTES
my_errno is not changed if this succeeds!
*/
int _ma_writeinfo(register MARIA_HA *info, uint operation)
{
int error,olderror;
MARIA_SHARE *share= info->s;
DBUG_ENTER("_ma_writeinfo");
DBUG_PRINT("info",("operation: %u tot_locks: %u", operation,
share->tot_locks));
error=0;
if (share->tot_locks == 0 && !share->base.born_transactional)
{
/* transactional tables flush their state at Checkpoint */
if (operation)
{ /* Two threads can't be here */
CRASH_IF_S3_TABLE(info->s); /* S3 readonly doesn't come here */
olderror= my_errno; /* Remember last error */
#ifdef MARIA_EXTERNAL_LOCKING
/*
The following only makes sense if we want to be allow two different
processes access the same table at the same time
*/
share->state.process= share->last_process= share->this_process;
share->state.unique= info->last_unique= info->this_unique;
share->state.update_count= info->last_loop= ++info->this_loop;
#endif
if ((error=
_ma_state_info_write_sub(share->kfile.file,
&share->state,
MA_STATE_INFO_WRITE_DONT_MOVE_OFFSET)))
olderror=my_errno;
#ifdef _WIN32
if (maria_flush)
{
_commit(share->kfile.file);
_commit(info->dfile.file);
}
#endif
my_errno=olderror;
}
}
else if (operation)
share->changed= 1; /* Mark keyfile changed */
DBUG_RETURN(error);
} /* _ma_writeinfo */
/*
Test if an external process has changed the database
(Should be called after readinfo)
*/
int _ma_test_if_changed(register MARIA_HA *info)
{
#ifdef MARIA_EXTERNAL_LOCKING
MARIA_SHARE *share= info->s;
if (share->state.process != share->last_process ||
share->state.unique != info->last_unique ||
share->state.update_count != info->last_loop)
{ /* Keyfile has changed */
DBUG_PRINT("info",("index file changed"));
if (share->state.process != share->this_process)
VOID(flush_pagecache_blocks(share->pagecache, &share->kfile,
FLUSH_RELEASE));
share->last_process=share->state.process;
info->last_unique= share->state.unique;
info->last_loop= share->state.update_count;
info->update|= HA_STATE_WRITTEN; /* Must use file on next */
info->data_changed= 1; /* For maria_is_changed */
return 1;
}
#endif
return (!(info->update & HA_STATE_AKTIV) ||
(info->update & (HA_STATE_WRITTEN | HA_STATE_DELETED |
HA_STATE_KEY_CHANGED)));
} /* _ma_test_if_changed */
/*
Put a mark in the .MAI file that someone is updating the table
DOCUMENTATION
state.open_count in the .MAI file is used the following way:
- For the first change of the .MYI file in this process open_count is
incremented by _ma_mark_file_changed(). (We have a write lock on the file
when this happens)
- In maria_close() it's decremented by _ma_decrement_open_count() if it
was incremented in the same process.
This mean that if we are the only process using the file, the open_count
tells us if the MARIA file wasn't properly closed. (This is true if
my_disable_locking is set).
open_count is not maintained on disk for temporary tables.
*/
#define _MA_ALREADY_MARKED_FILE_CHANGED \
((share->state.changed & STATE_CHANGED) && share->global_changed)
int _ma_mark_file_changed(register MARIA_SHARE *share)
{
if (!share->base.born_transactional)
{
if (!_MA_ALREADY_MARKED_FILE_CHANGED)
{
int res= _ma_mark_file_changed_now(share);
/*
Ensure that STATE_NOT_ANALYZED is reset on table changes
*/
share->state.changed|= (STATE_CHANGED | STATE_NOT_ANALYZED |
STATE_NOT_OPTIMIZED_KEYS);
return res;
}
}
else
{
/*
For transactional tables, the table is marked changed when the first page
is written. Here we just mark the state to be updated so that caller
can do 'analyze table' and find that is has changed before any pages
are written.
*/
if (! test_all_bits(share->state.changed,
(STATE_CHANGED | STATE_NOT_ANALYZED |
STATE_NOT_OPTIMIZED_KEYS)))
{
mysql_mutex_lock(&share->intern_lock);
share->state.changed|=(STATE_CHANGED | STATE_NOT_ANALYZED |
STATE_NOT_OPTIMIZED_KEYS);
mysql_mutex_unlock(&share->intern_lock);
}
}
return 0;
}
int _ma_mark_file_changed_now(register MARIA_SHARE *share)
{
uchar buff[3];
int error= 1;
DBUG_ENTER("_ma_mark_file_changed_now");
if (_MA_ALREADY_MARKED_FILE_CHANGED)
DBUG_RETURN(0);
mysql_mutex_lock(&share->intern_lock); /* recheck under mutex */
if (! _MA_ALREADY_MARKED_FILE_CHANGED)
{
share->state.changed|=(STATE_CHANGED | STATE_NOT_ANALYZED |
STATE_NOT_OPTIMIZED_KEYS);
if (!share->global_changed)
{
share->changed= share->global_changed= 1;
share->state.open_count++;
}
/*
Temp tables don't need an open_count as they are removed on crash.
In theory transactional tables are fixed by log-based recovery, so don't
need an open_count either, but if recovery has failed and logs have been
removed (by maria-force-start-after-recovery-failures), we still need to
detect dubious tables.
If we didn't maintain open_count on disk for a table, after a crash
we wouldn't know if it was closed at crash time (thus does not need a
check) or not. So we would have to check all tables: overkill.
*/
if (!share->temporary)
{
CRASH_IF_S3_TABLE(share);
mi_int2store(buff,share->state.open_count);
buff[2]=1; /* Mark that it's changed */
if (my_pwrite(share->kfile.file, buff, sizeof(buff),
sizeof(share->state.header) +
MARIA_FILE_OPEN_COUNT_OFFSET,
MYF(MY_NABP)))
goto err;
}
/* Set uuid of file if not yet set (zerofilled file) */
if (share->base.born_transactional &&
!(share->state.org_changed & STATE_NOT_MOVABLE))
{
CRASH_IF_S3_TABLE(share);
/* Lock table to current installation */
if (_ma_set_uuid(share, 0) ||
(share->state.create_rename_lsn == LSN_NEEDS_NEW_STATE_LSNS &&
_ma_update_state_lsns_sub(share, LSN_IMPOSSIBLE,
trnman_get_min_trid(),
TRUE, TRUE)))
goto err;
share->state.changed|= STATE_NOT_MOVABLE;
share->state.org_changed|= STATE_NOT_MOVABLE;
}
}
error= 0;
err:
mysql_mutex_unlock(&share->intern_lock);
DBUG_RETURN(error);
#undef _MA_ALREADY_MARKED_FILE_CHANGED
}
/*
Check that a region is all zero
SYNOPSIS
check_if_zero()
pos Start of memory to check
length length of memory region
NOTES
Used mainly to detect rows with wrong extent information
*/
my_bool _ma_check_if_zero(uchar *pos, size_t length)
{
uchar *end;
for (end= pos+ length; pos != end ; pos++)
if (pos[0] != 0)
return 1;
return 0;
}
/*
This is only called by close or by extra(HA_FLUSH) if the OS has the pwrite()
call. In these context the following code should be safe!
*/
int _ma_decrement_open_count(MARIA_HA *info, my_bool lock_tables)
{
uchar buff[2];
register MARIA_SHARE *share= info->s;
int lock_error=0,write_error=0;
DBUG_ENTER("_ma_decrement_open_count");
if (share->global_changed)
{
uint old_lock=info->lock_type;
share->global_changed=0;
lock_error= (my_disable_locking || ! lock_tables ? 0 :
maria_lock_database(info, F_WRLCK));
/* Its not fatal even if we couldn't get the lock ! */
if (share->state.open_count > 0)
{
CRASH_IF_S3_TABLE(share);
share->state.open_count--;
share->changed= 1; /* We have to update state */
/*
For temporary tables that will just be deleted, we don't have
to decrement state. For transactional tables the state will be
updated in maria_close().
*/
if (!share->temporary && !share->now_transactional)
{
mi_int2store(buff,share->state.open_count);
write_error= (int) my_pwrite(share->kfile.file, buff, sizeof(buff),
sizeof(share->state.header) +
MARIA_FILE_OPEN_COUNT_OFFSET,
MYF(MY_NABP));
}
}
if (!lock_error && !my_disable_locking && lock_tables)
lock_error=maria_lock_database(info,old_lock);
}
DBUG_RETURN(MY_TEST(lock_error || write_error));
}
/** @brief mark file as crashed */
void _ma_mark_file_crashed(MARIA_SHARE *share)
{
uchar buff[2];
DBUG_ENTER("_ma_mark_file_crashed");
share->state.changed|= STATE_CRASHED;
if (share->no_status_updates)
DBUG_VOID_RETURN; /* Safety */
mi_int2store(buff, share->state.changed);
/*
We can ignore the errors, as if the mark failed, there isn't anything
else we can do; The user should already have got an error that the
table was crashed.
*/
(void) my_pwrite(share->kfile.file, buff, sizeof(buff),
sizeof(share->state.header) +
MARIA_FILE_CHANGED_OFFSET,
MYF(MY_NABP));
DBUG_VOID_RETURN;
}
/**
@brief Set uuid of for a Maria file
@fn _ma_set_uuid()
@param share Maria share
@param reset_uuid Instead of setting file to maria_uuid, set it to
0 to mark it as movable
*/
my_bool _ma_set_uuid(MARIA_SHARE *share, my_bool reset_uuid)
{
uchar buff[MY_UUID_SIZE], *uuid;
uuid= maria_uuid;
if (reset_uuid)
{
bzero(buff, sizeof(buff));
uuid= buff;
}
CRASH_IF_S3_TABLE(share);
return (my_bool) my_pwrite(share->kfile.file, uuid, MY_UUID_SIZE,
mi_uint2korr(share->state.header.base_pos),
MYF(MY_NABP));
}
Reference in a new issue View git blame Copy permalink