mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-01-27 09:14:17 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions
mariadb/sql/sql_reload.cc
Brandon Nesterenko 952ab9a596 MDEV-30260: Slave crashed:reload_acl_and_cache during shutdown 
The signal handler thread can use various different runtime
resources when processing a SIGHUP (e.g. master-info information)
due to calling into reload_acl_and_cache(). Currently, the shutdown
process waits for the termination of the signal thread after
performing cleanup. However, this could cause resources actively
used by the signal handler to be freed while reload_acl_and_cache()
is processing.

The specific resource that caused MDEV-30260 is a race condition for
the hostname_cache, such that mysqld would delete it in
clean_up()::hostname_cache_free(), before the signal handler would
use it in reload_acl_and_cache()::hostname_cache_refresh().

Another similar resource is the active_mi/master_info_index. There
was a race between its deletion by the main thread in end_slave(),
and their usage by the Signal Handler as a part of
Master_info_index::flush_all_relay_logs.read(active_mi) in
reload_acl_and_cache().

This patch fixes these race conditions by relocating where server
shutdown waits for the signal handler to die until after
server-level threads have been killed (i.e., as a last step of
close_connections()). With respect to the hostname_cache, active_mi
and master_info_cache, this ensures that they cannot be destroyed
while the signal handler is still active, and potentially using
them.

Additionally:

 1) This requires that Events memory is still in place for SIGHUP
handling's mysql_print_status(). So event deinitialization is moved
into clean_up(), but the event scheduler still needs to be stopped
in close_connections() at the same spot.

 2) The function kill_server_thread is no longer used, so it is
deleted

 3) The timeout to wait for the death of the signal thread was not
consistent with the comment. The comment mentioned up to 10 seconds,
whereas it was actually 0.01s. The code has been fixed to wait up to
10 seconds.

 4) A warning has been added if the signal handler thread fails to
exit in time.

 5) Added pthread_join() to end of wait_for_signal_thread_to_end()
if it hadn't ended in 10s with a warning. Note this also removes
the pthread_detached attribute from the signal_thread to allow
for the pthread_join().

Reviewed By:
===========
Vladislav Vaintroub <wlad@mariadb.com>
Andrei Elkin <andrei.elkin@mariadb.com>
2024-04-09 14:25:13 -06:00

675 lines
20 KiB
C++
Raw Blame History

/* Copyright (c) 2010, 2016, Oracle and/or its affiliates.
Copyright (c) 2011, 2016, MariaDB
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 St, Fifth Floor, Boston, MA 02110-1335 USA */
#include "mariadb.h"
#include "sql_reload.h"
#include "sql_priv.h"
#include "mysqld.h" // select_errors
#include "sql_class.h" // THD
#include "sql_acl.h" // acl_reload
#include "sql_servers.h" // servers_reload
#include "sql_connect.h" // reset_mqh
#include "thread_cache.h"
#include "sql_base.h" // close_cached_tables
#include "sql_db.h" // my_dbopt_cleanup
#include "hostname.h" // hostname_cache_refresh
#include "sql_repl.h" // reset_master, reset_slave
#include "rpl_mi.h" // Master_info::data_lock
#include "sql_show.h"
#include "debug_sync.h"
#include "des_key_file.h"
#include "transaction.h"
#ifdef WITH_WSREP
#include "wsrep_mysqld.h"
#endif
static void disable_checkpoints(THD *thd);
/**
Reload/resets privileges and the different caches.
@param thd Thread handler (can be NULL!)
@param options What should be reset/reloaded (tables, privileges, slave...)
@param tables Tables to flush (if any)
@param write_to_binlog < 0 if there was an error while interacting with the binary log inside
reload_acl_and_cache,
0 if we should not write to the binary log,
> 0 if we can write to the binlog.
@note Depending on 'options', it may be very bad to write the
query to the binlog (e.g. FLUSH SLAVE); this is a
pointer where reload_acl_and_cache() will put 0 if
it thinks we really should not write to the binlog.
Otherwise it will put 1.
@return Error status code
@retval 0 Ok
@retval !=0 Error; thd->killed is set or thd->is_error() is true
*/
bool reload_acl_and_cache(THD *thd, unsigned long long options,
TABLE_LIST *tables, int *write_to_binlog)
{
bool result=0;
select_errors=0; /* Write if more errors */
int tmp_write_to_binlog= *write_to_binlog= 1;
#ifndef DBUG_OFF
/*
When invoked for handling a SIGHUP by rpl_shutdown_sighup.test, we need to
force the signal handler to wait after REFRESH_TABLES, as that will check
for a killed server, and we need to call hostname_cache_refresh after
server cleanup has happened to trigger MDEV-30260.
*/
int do_dbug_sleep= 0;
#endif
DBUG_ASSERT(!thd || !thd->in_sub_stmt);
#ifndef NO_EMBEDDED_ACCESS_CHECKS
if (options & REFRESH_GRANT)
{
THD *tmp_thd= 0;
/*
If reload_acl_and_cache() is called from SIGHUP handler we have to
allocate temporary THD for execution of acl_reload()/grant_reload().
*/
if (unlikely(!thd) && (thd= (tmp_thd= new THD(0))))
{
thd->thread_stack= (char*) &tmp_thd;
thd->store_globals();
}
if (likely(thd))
{
bool reload_acl_failed= acl_reload(thd);
bool reload_grants_failed= grant_reload(thd);
bool reload_servers_failed= servers_reload(thd);
if (reload_acl_failed || reload_grants_failed || reload_servers_failed)
{
result= 1;
/*
When an error is returned, my_message may have not been called and
the client will hang waiting for a response.
*/
my_error(ER_UNKNOWN_ERROR, MYF(0));
}
#ifndef DBUG_OFF
DBUG_EXECUTE_IF("hold_sighup_log_refresh", {
DBUG_ASSERT(!debug_sync_set_action(
thd, STRING_WITH_LEN("now SIGNAL in_reload_acl_and_cache "
"WAIT_FOR refresh_logs")));
do_dbug_sleep= 1;
});
#endif
}
opt_noacl= 0;
if (unlikely(tmp_thd))
{
delete tmp_thd;
thd= 0;
}
reset_mqh((LEX_USER *)NULL, TRUE);
}
#endif
if (options & REFRESH_LOG)
{
/*
Flush the normal query log, the update log, the binary log,
the slow query log, the relay log (if it exists) and the log
tables.
*/
options|= REFRESH_BINARY_LOG;
options|= REFRESH_RELAY_LOG;
options|= REFRESH_SLOW_LOG;
options|= REFRESH_GENERAL_LOG;
options|= REFRESH_ENGINE_LOG;
options|= REFRESH_ERROR_LOG;
}
if (options & REFRESH_ERROR_LOG)
if (unlikely(flush_error_log()))
result= 1;
if ((options & REFRESH_SLOW_LOG) && global_system_variables.sql_log_slow)
logger.flush_slow_log();
if ((options & REFRESH_GENERAL_LOG) && opt_log)
logger.flush_general_log();
if (options & REFRESH_ENGINE_LOG)
if (ha_flush_logs())
result= 1;
if (options & REFRESH_BINARY_LOG)
{
/*
Writing this command to the binlog may result in infinite loops
when doing mysqlbinlog|mysql, and anyway it does not really make
sense to log it automatically (would cause more trouble to users
than it would help them)
*/
tmp_write_to_binlog= 0;
if (mysql_bin_log.is_open())
{
DYNAMIC_ARRAY *drop_gtid_domain=
(thd && (thd->lex->delete_gtid_domain.elements > 0)) ?
&thd->lex->delete_gtid_domain : NULL;
if (mysql_bin_log.rotate_and_purge(true, drop_gtid_domain))
*write_to_binlog= -1;
/* Note that WSREP(thd) might not be true here e.g. during
SST. */
if (WSREP_ON)
{
/* Wait for last binlog checkpoint event to be logged. */
mysql_bin_log.wait_for_last_checkpoint_event();
}
}
}
if (options & REFRESH_RELAY_LOG)
{
#ifdef HAVE_REPLICATION
LEX_CSTRING connection_name;
Master_info *mi;
if (thd)
connection_name= thd->lex->relay_log_connection_name;
else
{
connection_name.str= (char*) "";
connection_name.length= 0;
}
/*
Writing this command to the binlog may cause problems as the
slave is not likely to have the same connection names.
*/
tmp_write_to_binlog= 0;
if (connection_name.length == 0)
{
if (master_info_index->flush_all_relay_logs())
*write_to_binlog= -1;
}
else if (!(mi= (get_master_info(&connection_name,
Sql_condition::WARN_LEVEL_ERROR))))
{
result= 1;
}
else
{
mysql_mutex_lock(&mi->data_lock);
if (rotate_relay_log(mi))
*write_to_binlog= -1;
mysql_mutex_unlock(&mi->data_lock);
mi->release();
}
#endif
}
#ifdef HAVE_QUERY_CACHE
if (options & REFRESH_QUERY_CACHE_FREE)
{
query_cache.pack(thd); // FLUSH QUERY CACHE
options &= ~REFRESH_QUERY_CACHE; // Don't flush cache, just free memory
}
if (options & (REFRESH_TABLES | REFRESH_QUERY_CACHE))
{
query_cache.flush(); // RESET QUERY CACHE
}
#endif /*HAVE_QUERY_CACHE*/
DBUG_ASSERT(!thd || thd->locked_tables_mode ||
!thd->mdl_context.has_locks() ||
thd->handler_tables_hash.records ||
thd->ull_hash.records ||
thd->global_read_lock.is_acquired() ||
thd->mdl_backup_lock ||
thd->current_backup_stage != BACKUP_FINISHED
);
/*
Note that if REFRESH_READ_LOCK bit is set then REFRESH_TABLES is set too
(see sql_yacc.yy)
*/
if (options & (REFRESH_TABLES | REFRESH_READ_LOCK))
{
if ((options & REFRESH_READ_LOCK) && thd)
{
DBUG_ASSERT(!(options & REFRESH_FAST) && !tables);
/*
On the first hand we need write lock on the tables to be flushed,
on the other hand we must not try to aspire a global read lock
if we have a write locked table as this would lead to a deadlock
when trying to reopen (and re-lock) the table after the flush.
*/
if (thd->locked_tables_mode)
{
my_error(ER_LOCK_OR_ACTIVE_TRANSACTION, MYF(0));
return 1;
}
/*
Writing to the binlog could cause deadlocks, as we don't log
UNLOCK TABLES
*/
tmp_write_to_binlog= 0;
if (thd->global_read_lock.lock_global_read_lock(thd))
return 1; // Killed
if (flush_tables(thd, FLUSH_ALL))
{
/*
NOTE: my_error() has been already called by reopen_tables() within
close_cached_tables().
*/
thd->global_read_lock.unlock_global_read_lock(thd);
return 1;
}
if (thd->global_read_lock.make_global_read_lock_block_commit(thd)) // Killed
{
/* Don't leave things in a half-locked state */
thd->global_read_lock.unlock_global_read_lock(thd);
return 1;
}
if (options & REFRESH_CHECKPOINT)
disable_checkpoints(thd);
/*
We need to do it second time after wsrep appliers were blocked in
make_global_read_lock_block_commit(thd) above since they could have
modified the tables too.
*/
if (WSREP(thd) && flush_tables(thd, FLUSH_ALL))
result= 1;
}
else
{
if (thd && thd->locked_tables_mode)
{
/*
If we are under LOCK TABLES we should have a write
lock on tables which we are going to flush.
*/
if (tables)
{
int err;
for (TABLE_LIST *t= tables; t; t= t->next_local)
if (!find_table_for_mdl_upgrade(thd, t->db.str, t->table_name.str, &err))
{
if (is_locked_view(thd, t))
t->next_local= t->next_global;
else
{
my_error(err, MYF(0), t->table_name.str);
return 1;
}
}
}
else
{
/*
It is not safe to upgrade the metadata lock without GLOBAL IX lock.
This can happen with FLUSH TABLES <list> WITH READ LOCK as we in
these cases don't take a GLOBAL IX lock in order to be compatible
with global read lock.
*/
if (thd->open_tables &&
!thd->mdl_context.is_lock_owner(MDL_key::BACKUP, "", "",
MDL_BACKUP_DDL))
{
my_error(ER_TABLE_NOT_LOCKED_FOR_WRITE, MYF(0),
thd->open_tables->s->table_name.str);
return true;
}
for (TABLE *tab= thd->open_tables; tab; tab= tab->next)
{
if (! tab->mdl_ticket->is_upgradable_or_exclusive())
{
my_error(ER_TABLE_NOT_LOCKED_FOR_WRITE, MYF(0),
tab->s->table_name.str);
return 1;
}
}
}
}
#ifdef WITH_WSREP
/* In case of applier thread, do not call flush tables */
if (!thd || !thd->wsrep_applier)
#endif /* WITH_WSREP */
{
if (close_cached_tables(thd, tables,
((options & REFRESH_FAST) ? FALSE : TRUE),
(thd ? thd->variables.lock_wait_timeout :
LONG_TIMEOUT)))
{
/*
NOTE: my_error() has been already called by reopen_tables() within
close_cached_tables().
*/
result= 1;
}
}
}
my_dbopt_cleanup();
}
#ifndef DBUG_OFF
if (do_dbug_sleep)
my_sleep(3000000); // 3s
#endif
if (options & REFRESH_HOSTS)
hostname_cache_refresh();
if (thd && (options & REFRESH_STATUS))
refresh_status(thd);
if (options & REFRESH_THREADS)
thread_cache.flush();
#ifdef HAVE_REPLICATION
if (options & REFRESH_MASTER)
{
DBUG_ASSERT(thd);
tmp_write_to_binlog= 0;
if (reset_master(thd, NULL, 0, thd->lex->next_binlog_file_number))
{
/* NOTE: my_error() has been already called by reset_master(). */
result= 1;
}
}
#endif
#ifdef HAVE_OPENSSL
if (options & REFRESH_DES_KEY_FILE)
{
if (des_key_file && load_des_key_file(des_key_file))
{
/* NOTE: my_error() has been already called by load_des_key_file(). */
result= 1;
}
}
#endif
#ifdef HAVE_REPLICATION
if (options & REFRESH_SLAVE)
{
LEX_MASTER_INFO* lex_mi= &thd->lex->mi;
Master_info *mi;
tmp_write_to_binlog= 0;
if (!(mi= get_master_info(&lex_mi->connection_name,
Sql_condition::WARN_LEVEL_ERROR)))
{
result= 1;
}
else
{
/* The following will fail if slave is running */
if (reset_slave(thd, mi))
{
mi->release();
/* NOTE: my_error() has been already called by reset_slave(). */
result= 1;
}
else if (mi->connection_name.length && thd->lex->reset_slave_info.all)
{
/* If not default connection and 'all' is used */
mi->release();
mysql_mutex_lock(&LOCK_active_mi);
if (master_info_index->remove_master_info(mi))
result= 1;
mysql_mutex_unlock(&LOCK_active_mi);
}
else
mi->release();
}
}
#endif
if (options & REFRESH_USER_RESOURCES)
reset_mqh((LEX_USER *) NULL, 0); /* purecov: inspected */
if (options & REFRESH_SSL)
{
if (reinit_ssl())
result= 1;
#ifdef WITH_WSREP
if (!result &&
WSREP_ON && wsrep_reload_ssl())
{
my_message(ER_UNKNOWN_ERROR, "Failed to refresh WSREP SSL.", MYF(0));
result= 1;
}
#endif
}
if (options & REFRESH_GENERIC)
{
List_iterator_fast<LEX_CSTRING> li(thd->lex->view_list);
LEX_CSTRING *ls;
while ((ls= li++))
{
ST_SCHEMA_TABLE *table= find_schema_table(thd, ls);
if (table->reset_table())
result= 1;
}
}
if (*write_to_binlog != -1)
*write_to_binlog= tmp_write_to_binlog;
/*
If the query was killed then this function must fail.
*/
return result || (thd ? thd->killed : 0);
}
/**
Implementation of FLUSH TABLES <table_list> WITH READ LOCK
and FLUSH TABLES <table_list> FOR EXPORT
In brief: take exclusive locks, expel tables from the table
cache, reopen the tables, enter the 'LOCKED TABLES' mode,
downgrade the locks.
Note: the function is written to be called from
mysql_execute_command(), it is not reusable in arbitrary
execution context.
Required privileges
-------------------
Since the statement implicitly enters LOCK TABLES mode,
it requires LOCK TABLES privilege on every table.
But since the rest of FLUSH commands require
the global RELOAD_ACL, it also requires RELOAD_ACL.
Compatibility with the global read lock
---------------------------------------
We don't wait for the GRL, since neither the
5.1 combination that this new statement is intended to
replace (LOCK TABLE <list> WRITE; FLUSH TABLES;),
nor FLUSH TABLES WITH READ LOCK do.
@todo: this is not implemented, Dmitry disagrees.
Currently we wait for GRL in another connection,
but are compatible with a GRL in our own connection.
Behaviour under LOCK TABLES
---------------------------
Bail out: i.e. don't perform an implicit UNLOCK TABLES.
This is not consistent with LOCK TABLES statement, but is
in line with behaviour of FLUSH TABLES WITH READ LOCK, and we
try to not introduce any new statements with implicit
semantics.
Compatibility with parallel updates
-----------------------------------
As a result, we will wait for all open transactions
against the tables to complete. After the lock downgrade,
new transactions will be able to read the tables, but not
write to them.
Differences from FLUSH TABLES <list>
-------------------------------------
- you can't flush WITH READ LOCK a non-existent table
- you can't flush WITH READ LOCK under LOCK TABLES
Effect on views and temporary tables.
------------------------------------
You can only apply this command to existing base tables.
If a view with such name exists, ER_WRONG_OBJECT is returned.
If a temporary table with such name exists, it's ignored:
if there is a base table, it's used, otherwise ER_NO_SUCH_TABLE
is returned.
Handling of MERGE tables
------------------------
For MERGE table this statement will open and lock child tables
for read (it is impossible to lock parent table without it).
Child tables won't be flushed unless they are explicitly present
in the statement's table list.
Implicit commit
---------------
This statement causes an implicit commit before and
after it.
HANDLER SQL
-----------
If this connection has HANDLERs open against
some of the tables being FLUSHed, these handlers
are implicitly flushed (lose their position).
*/
bool flush_tables_with_read_lock(THD *thd, TABLE_LIST *all_tables)
{
Lock_tables_prelocking_strategy lock_tables_prelocking_strategy;
/*
This is called from SQLCOM_FLUSH, the transaction has
been committed implicitly.
*/
if (thd->locked_tables_mode)
{
my_error(ER_LOCK_OR_ACTIVE_TRANSACTION, MYF(0));
goto error;
}
if (thd->current_backup_stage != BACKUP_FINISHED)
{
my_error(ER_BACKUP_LOCK_IS_ACTIVE, MYF(0));
goto error;
}
/* Should not flush tables while BACKUP LOCK is active */
if (thd->mdl_backup_lock)
{
my_error(ER_LOCK_OR_ACTIVE_TRANSACTION, MYF(0));
goto error;
}
if (thd->lex->type & REFRESH_READ_LOCK)
{
/*
Acquire SNW locks on tables to be flushed. Don't acquire global
IX and database-scope IX locks on the tables as this will make
this statement incompatible with FLUSH TABLES WITH READ LOCK.
*/
if (lock_table_names(thd, all_tables, NULL,
thd->variables.lock_wait_timeout,
MYSQL_OPEN_SKIP_SCOPED_MDL_LOCK))
goto error;
DEBUG_SYNC(thd,"flush_tables_with_read_lock_after_acquire_locks");
/* Reset ticket to satisfy asserts in open_tables(). */
for (auto table_list= all_tables; table_list;
table_list= table_list->next_global)
table_list->mdl_request.ticket= NULL;
}
thd->variables.option_bits|= OPTION_TABLE_LOCK;
/*
Before opening and locking tables the below call also waits
for old shares to go away, so the fact that we don't pass
MYSQL_OPEN_IGNORE_FLUSH flag to it is important.
Also we don't pass MYSQL_OPEN_HAS_MDL_LOCK flag as we want
to open underlying tables if merge table is flushed.
For underlying tables of the merge the below call has to
acquire SNW locks to ensure that they can be locked for
read without further waiting.
*/
if (open_and_lock_tables(thd, all_tables, FALSE,
MYSQL_OPEN_SKIP_SCOPED_MDL_LOCK,
&lock_tables_prelocking_strategy))
goto error_reset_bits;
if (thd->lex->type & REFRESH_FOR_EXPORT)
{
// Check if all storage engines support FOR EXPORT.
for (TABLE_LIST *table_list= all_tables; table_list;
table_list= table_list->next_global)
{
if (!(table_list->table->file->ha_table_flags() & HA_CAN_EXPORT))
{
my_error(ER_ILLEGAL_HA, MYF(0),table_list->table->file->table_type(),
table_list->db.str, table_list->table_name.str);
goto error_reset_bits;
}
}
}
if (thd->lex->type & REFRESH_READ_LOCK)
{
for (auto table_list= all_tables; table_list;
table_list= table_list->next_global)
{
if (table_list->table->file->extra(HA_EXTRA_FLUSH))
goto error_reset_bits;
}
}
if (thd->locked_tables_list.init_locked_tables(thd))
goto error_reset_bits;
/*
We don't downgrade MDL_SHARED_NO_WRITE here as the intended
post effect of this call is identical to LOCK TABLES <...> READ,
and we didn't use thd->in_lock_talbes and
thd->sql_command= SQLCOM_LOCK_TABLES hacks to enter the LTM.
*/
return FALSE;
error_reset_bits:
trans_rollback_stmt(thd);
close_thread_tables(thd);
thd->variables.option_bits&= ~OPTION_TABLE_LOCK;
error:
return TRUE;
}
/**
Disable checkpoints for all handlers
This is released in unlock_global_read_lock()
*/
static void disable_checkpoints(THD *thd)
{
if (!thd->global_disable_checkpoint)
{
thd->global_disable_checkpoint= 1;
if (!global_disable_checkpoint++)
ha_checkpoint_state(1); // Disable checkpoints
}
}
Reference in a new issue View git blame Copy permalink