[Commits] Rev 3954: MDEV-5363: Make parallel replication waits killable in http://bazaar.launchpad.net/~maria-captains/maria/10.0

knielsen at knielsen-hq.org knielsen at knielsen-hq.org
Fri Dec 13 15:26:52 EET 2013


At http://bazaar.launchpad.net/~maria-captains/maria/10.0

------------------------------------------------------------
revno: 3954
revision-id: knielsen at knielsen-hq.org-20131213132651-b90mxo3xqsq89r6u
parent: knielsen at knielsen-hq.org-20131206122823-chprmuvs2u7xkf75
committer: knielsen at knielsen-hq.org
branch nick: tmp-10.0-base
timestamp: Fri 2013-12-13 14:26:51 +0100
message:
  MDEV-5363: Make parallel replication waits killable
  
  Add a test case for killing a waiting query in parallel replication.
  
  Fix several bugs found:
  
   - We should not wakeup_subsequent_commits() in ha_rollback_trans(), since we
     do not know the right wakeup_error() to give.
  
   - When a wait_for_prior_commit() is killed, we must unregister from the
     waitee so we do not race and get an extra (non-kill) wakeup.
  
   - We need to deal with error propagation correctly in queue_for_group_commit
     when one thread is killed.
  
   - Fix one locking issue in queue_for_group_commit(), we could unlock the
     waitee lock too early and this end up processing wakeup() with insufficient
     locking.
  
   - Fix Xid_log_event::do_apply_event; if commit fails it must not update the
     in-memory @@gtid_slave_pos state.
  
   - Fix and cleanup some things in the rpl_parallel.cc error handling.
  
   - Add a missing check for killed in the slave sql driver thread, to avoid a
     race.
=== modified file 'mysql-test/suite/rpl/r/rpl_parallel.result'
--- a/mysql-test/suite/rpl/r/rpl_parallel.result	2013-10-08 12:36:06 +0000
+++ b/mysql-test/suite/rpl/r/rpl_parallel.result	2013-12-13 13:26:51 +0000
@@ -7,6 +7,7 @@ SET GLOBAL slave_parallel_threads=10;
 CHANGE MASTER TO master_use_gtid=slave_pos;
 include/start_slave.inc
 *** Test long-running query in domain 1 can run in parallel with short queries in domain 0 ***
+ALTER TABLE mysql.gtid_slave_pos ENGINE=InnoDB;
 CREATE TABLE t1 (a int PRIMARY KEY) ENGINE=MyISAM;
 CREATE TABLE t2 (a int PRIMARY KEY) ENGINE=InnoDB;
 INSERT INTO t1 VALUES (1);
@@ -257,6 +258,78 @@ a	b
 include/stop_slave.inc
 SET GLOBAL binlog_format=@old_format;
 SET GLOBAL slave_parallel_threads=0;
+SET GLOBAL slave_parallel_threads=10;
+include/start_slave.inc
+*** Test killing slave threads at various wait points ***
+*** 1. Test killing transaction waiting in commit for previous transaction to commit ***
+SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1';
+SET binlog_format=statement;
+INSERT INTO t3 VALUES (31, foo(31,
+'commit_before_prepare_ordered WAIT_FOR t2_waiting',
+'commit_after_prepare_ordered SIGNAL t1_ready WAIT_FOR t1_cont'));
+SET debug_sync='now WAIT_FOR master_queued1';
+SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2';
+SET binlog_format=statement;
+BEGIN;
+INSERT INTO t3 VALUES (32, foo(32,
+'ha_write_row_end SIGNAL t2_query WAIT_FOR t2_cont',
+''));
+INSERT INTO t3 VALUES (33, foo(33,
+'group_commit_waiting_for_prior SIGNAL t2_waiting',
+'group_commit_waiting_for_prior_killed SIGNAL t2_killed'));
+COMMIT;
+SET debug_sync='now WAIT_FOR master_queued2';
+SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued3';
+SET binlog_format=statement;
+INSERT INTO t3 VALUES (34, foo(34,
+'',
+''));
+SET debug_sync='now WAIT_FOR master_queued3';
+SET debug_sync='now SIGNAL master_cont1';
+SELECT * FROM t3 WHERE a >= 30 ORDER BY a;
+a       b
+31      31
+32      32
+33      33
+34      34
+SET sql_log_bin=0;
+CALL mtr.add_suppression("Query execution was interrupted");
+CALL mtr.add_suppression("Commit failed due to failure of an earlier commit on which this one depends");
+SET sql_log_bin=1;
+SET debug_sync='now WAIT_FOR t2_query';
+SET debug_sync='now SIGNAL t2_cont';
+SET debug_sync='now WAIT_FOR t1_ready';
+KILL THD_ID;
+SET debug_sync='now WAIT_FOR t2_killed';
+SET debug_sync='now SIGNAL t1_cont';
+include/wait_for_slave_sql_error.inc [errno=1317,1963]
+STOP SLAVE IO_THREAD;
+SELECT * FROM t3 WHERE a >= 30 ORDER BY a;
+a       b
+31      31
+SET GLOBAL slave_parallel_threads=0;
+SET GLOBAL slave_parallel_threads=10;
+SET sql_log_bin=0;
+DROP FUNCTION foo;
+CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500))
+RETURNS INT DETERMINISTIC
+BEGIN
+RETURN x;
+END
+||
+SET sql_log_bin=1;
+INSERT INTO t3 VALUES (39,0);
+include/start_slave.inc
+SELECT * FROM t3 WHERE a >= 30 ORDER BY a;
+a       b
+31      31
+32      32
+33      33
+34      34
+39      0
+include/stop_slave.inc
+SET GLOBAL binlog_format=@old_format;
+SET GLOBAL slave_parallel_threads=0;
 SET GLOBAL slave_parallel_threads=10;
 include/start_slave.inc
 include/stop_slave.inc

=== modified file 'mysql-test/suite/rpl/t/rpl_parallel.test'
--- a/mysql-test/suite/rpl/t/rpl_parallel.test	2013-10-31 13:11:41 +0000
+++ b/mysql-test/suite/rpl/t/rpl_parallel.test	2013-12-13 13:26:51 +0000
@@ -19,6 +19,7 @@ CHANGE MASTER TO master_use_gtid=slave_p
 --echo *** Test long-running query in domain 1 can run in parallel with short queries in domain 0 ***
 
 --connection server_1
+ALTER TABLE mysql.gtid_slave_pos ENGINE=InnoDB;
 CREATE TABLE t1 (a int PRIMARY KEY) ENGINE=MyISAM;
 CREATE TABLE t2 (a int PRIMARY KEY) ENGINE=InnoDB;
 INSERT INTO t1 VALUES (1);
@@ -334,6 +335,126 @@ SELECT * FROM t3 WHERE a >= 20 ORDER BY
 
 
 --connection server_2
+# Respawn all worker threads to clear any left-over debug_sync or other stuff.
+--source include/stop_slave.inc
+SET GLOBAL binlog_format=@old_format;
+SET GLOBAL slave_parallel_threads=0;
+SET GLOBAL slave_parallel_threads=10;
+--source include/start_slave.inc
+
+
+--echo *** Test killing slave threads at various wait points ***
+--echo *** 1. Test killing transaction waiting in commit for previous transaction to commit ***
+
+# Set up three transactions on the master that will be group-committed
+# together so they can be replicated in parallel on the slave.
+--connection con_temp3
+SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1';
+SET binlog_format=statement;
+send INSERT INTO t3 VALUES (31, foo(31,
+    'commit_before_prepare_ordered WAIT_FOR t2_waiting',
+    'commit_after_prepare_ordered SIGNAL t1_ready WAIT_FOR t1_cont'));
+
+--connection server_1
+SET debug_sync='now WAIT_FOR master_queued1';
+
+--connection con_temp4
+SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2';
+SET binlog_format=statement;
+BEGIN;
+# This insert is just so we can get T2 to wait while a query is running that we
+# can see in SHOW PROCESSLIST so we can get its thread_id to kill later.
+INSERT INTO t3 VALUES (32, foo(32,
+    'ha_write_row_end SIGNAL t2_query WAIT_FOR t2_cont',
+    ''));
+# This insert sets up debug_sync points so that T2 will tell when it is at its
+# wait point where we want to kill it - and when it has been killed.
+INSERT INTO t3 VALUES (33, foo(33,
+    'group_commit_waiting_for_prior SIGNAL t2_waiting',
+    'group_commit_waiting_for_prior_killed SIGNAL t2_killed'));
+send COMMIT;
+
+--connection server_1
+SET debug_sync='now WAIT_FOR master_queued2';
+
+--connection con_temp5
+SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued3';
+SET binlog_format=statement;
+send INSERT INTO t3 VALUES (34, foo(34,
+    '',
+    ''));
+
+--connection server_1
+SET debug_sync='now WAIT_FOR master_queued3';
+SET debug_sync='now SIGNAL master_cont1';
+
+--connection con_temp3
+REAP;
+--connection con_temp4
+REAP;
+--connection con_temp5
+REAP;
+
+--connection server_1
+SELECT * FROM t3 WHERE a >= 30 ORDER BY a;
+
+--connection server_2
+SET sql_log_bin=0;
+CALL mtr.add_suppression("Query execution was interrupted");
+CALL mtr.add_suppression("Commit failed due to failure of an earlier commit on which this one depends");
+SET sql_log_bin=1;
+# Wait until T2 is inside executing its insert of 32, then find it in SHOW
+# PROCESSLIST to know its thread id for KILL later.
+SET debug_sync='now WAIT_FOR t2_query';
+--let $thd_id= `SELECT ID FROM INFORMATION_SCHEMA.PROCESSLIST WHERE INFO LIKE '%foo(32%' AND INFO NOT LIKE '%LIKE%'`
+SET debug_sync='now SIGNAL t2_cont';
+
+# Wait until T2 has entered its wait for T1 to commit, and T1 has
+# progressed into its commit phase.
+SET debug_sync='now WAIT_FOR t1_ready';
+
+# Now kill the transaction T2.
+--replace_result $thd_id THD_ID
+eval KILL $thd_id;
+
+# Wait until T2 has reacted on the kill.
+SET debug_sync='now WAIT_FOR t2_killed';
+
+# Now we can allow T1 to proceed.
+SET debug_sync='now SIGNAL t1_cont';
+
+--let $slave_sql_errno= 1317,1963
+--source include/wait_for_slave_sql_error.inc
+STOP SLAVE IO_THREAD;
+SELECT * FROM t3 WHERE a >= 30 ORDER BY a;
+
+# Now we have to disable the debug_sync statements, so they do not trigger
+# when the events are retried.
+SET GLOBAL slave_parallel_threads=0;
+SET GLOBAL slave_parallel_threads=10;
+SET sql_log_bin=0;
+DROP FUNCTION foo;
+--delimiter ||
+CREATE FUNCTION foo(x INT, d1 VARCHAR(500), d2 VARCHAR(500))
+  RETURNS INT DETERMINISTIC
+  BEGIN
+    RETURN x;
+  END
+||
+--delimiter ;
+SET sql_log_bin=1;
+
+--connection server_1
+INSERT INTO t3 VALUES (39,0);
+--save_master_pos
+
+--connection server_2
+--source include/start_slave.inc
+--sync_with_master
+SELECT * FROM t3 WHERE a >= 30 ORDER BY a;
+
+
+--connection server_2
 --source include/stop_slave.inc
 SET GLOBAL binlog_format=@old_format;
 SET GLOBAL slave_parallel_threads=0;

=== modified file 'sql/handler.cc'
--- a/sql/handler.cc	2013-11-22 23:50:54 +0000
+++ b/sql/handler.cc	2013-12-13 13:26:51 +0000
@@ -1300,7 +1300,10 @@ int ha_commit_trans(THD *thd, bool all)
   {
     /* Free resources and perform other cleanup even for 'empty' transactions. */
     if (is_real_trans)
+    {
       thd->transaction.cleanup();
+      thd->wakeup_subsequent_commits(error);
+    }
     DBUG_RETURN(0);
   }
 
@@ -1334,6 +1337,7 @@ int ha_commit_trans(THD *thd, bool all)
                                       thd->variables.lock_wait_timeout))
     {
       ha_rollback_trans(thd, all);
+      thd->wakeup_subsequent_commits(1);
       DBUG_RETURN(1);
     }
 
@@ -1421,6 +1425,7 @@ int ha_commit_trans(THD *thd, bool all)
 err:
   error= 1;                                  /* Transaction was rolled back */
   ha_rollback_trans(thd, all);
+  thd->wakeup_subsequent_commits(error);
 
 end:
   if (rw_trans && mdl_request.ticket)
@@ -1591,10 +1596,7 @@ int ha_rollback_trans(THD *thd, bool all
 
   /* Always cleanup. Even if nht==0. There may be savepoints. */
   if (is_real_trans)
-  {
-    thd->wakeup_subsequent_commits(error);
     thd->transaction.cleanup();
-  }
   if (all)
     thd->transaction_rollback_request= FALSE;
 

=== modified file 'sql/log.cc'
--- a/sql/log.cc	2013-12-05 13:36:09 +0000
+++ b/sql/log.cc	2013-12-13 13:26:51 +0000
@@ -6622,7 +6622,7 @@ int
 MYSQL_BIN_LOG::queue_for_group_commit(group_commit_entry *orig_entry)
 {
   group_commit_entry *entry, *orig_queue;
-  wait_for_commit *list, *cur, *last;
+  wait_for_commit *cur, *last;
   wait_for_commit *wfc;
   DBUG_ENTER("MYSQL_BIN_LOG::queue_for_group_commit");
 
@@ -6663,17 +6663,39 @@ MYSQL_BIN_LOG::queue_for_group_commit(gr
       wfc->opaque_pointer= NULL;
       DBUG_PRINT("info", ("After waiting for prior commit, queued_by_other=%d",
                  orig_entry->queued_by_other));
-      orig_entry->thd->exit_cond(old_msg);
 
       if (wfc->waiting_for_commit)
       {
-        /* Interrupted by kill. */
-        wfc->wakeup_error= orig_entry->thd->killed_errno();
-        if (wfc->wakeup_error)
-          wfc->wakeup_error= ER_QUERY_INTERRUPTED;
-        my_message(wfc->wakeup_error, ER(wfc->wakeup_error), MYF(0));
-        DBUG_RETURN(-1);
+        /* Wait terminated due to kill. */
+        wait_for_commit *loc_waitee= wfc->waitee;
+        mysql_mutex_lock(&loc_waitee->LOCK_wait_commit);
+        if (loc_waitee->wakeup_subsequent_commits_running ||
+            orig_entry->queued_by_other)
+        {
+          /* Our waitee is already waking us up, so ignore the kill. */
+          mysql_mutex_unlock(&loc_waitee->LOCK_wait_commit);
+          do
+          {
+            mysql_cond_wait(&wfc->COND_wait_commit, &wfc->LOCK_wait_commit);
+          } while (wfc->waiting_for_commit);
+        }
+        else
+        {
+          /* We were killed, so remove us from the list of waitee. */
+          wfc->remove_from_list(&loc_waitee->subsequent_commits_list);
+          mysql_mutex_unlock(&loc_waitee->LOCK_wait_commit);
+
+          orig_entry->thd->exit_cond(old_msg);
+          /* Interrupted by kill. */
+          DEBUG_SYNC(orig_entry->thd, "group_commit_waiting_for_prior_killed");
+          wfc->wakeup_error= orig_entry->thd->killed_errno();
+          if (wfc->wakeup_error)
+            wfc->wakeup_error= ER_QUERY_INTERRUPTED;
+          my_message(wfc->wakeup_error, ER(wfc->wakeup_error), MYF(0));
+          DBUG_RETURN(-1);
+        }
       }
+      orig_entry->thd->exit_cond(old_msg);
     }
     else
       mysql_mutex_unlock(&wfc->LOCK_wait_commit);
@@ -6729,9 +6751,8 @@ MYSQL_BIN_LOG::queue_for_group_commit(gr
     used by the caller or any other function.
   */
 
-  list= wfc;
-  cur= list;
-  last= list;
+  cur= wfc;
+  last= wfc;
   entry= orig_entry;
   for (;;)
   {
@@ -6757,11 +6778,11 @@ MYSQL_BIN_LOG::queue_for_group_commit(gr
     */
     if (cur->subsequent_commits_list)
     {
-      bool have_lock;
       wait_for_commit *waiter;
+      wait_for_commit *wakeup_list= NULL;
+      wait_for_commit **wakeup_next_ptr= &wakeup_list;
 
       mysql_mutex_lock(&cur->LOCK_wait_commit);
-      have_lock= true;
       /*
         Grab the list, now safely under lock, and process it if still
         non-empty.
@@ -6802,18 +6823,68 @@ MYSQL_BIN_LOG::queue_for_group_commit(gr
             For this, we need to set the "wakeup running" flag and release
             the waitee lock to avoid a deadlock, see comments on
             THD::wakeup_subsequent_commits2() for details.
+
+            So we need to put these on a list and delay the wakeup until we
+            have released the lock.
+          */
+          *wakeup_next_ptr= waiter;
+          wakeup_next_ptr= &waiter->next_subsequent_commit;
+        }
+        waiter= next;
+      }
+      if (wakeup_list)
+      {
+        /* Now release our lock and do the wakeups that were delayed above. */
+        cur->wakeup_subsequent_commits_running= true;
+        mysql_mutex_unlock(&cur->LOCK_wait_commit);
+        for (;;)
+        {
+          wait_for_commit *next;
+
+          /*
+            ToDo: We wakeup the waiter here, so that it can have the chance to
+            reach its own commit state and queue up for this same group commit,
+            if it is still pending.
+
+            One problem with this is that if the waiter does not reach its own
+            commit state before this group commit starts, and then the group
+            commit fails (binlog write failure), we do not get to propagate
+            the error to the waiter.
+
+            A solution for this could be to delay the wakeup until commit is
+            successful. But then we need to set a flag in the waitee that it is
+            already queued for group commit, so that the waiter can check this
+            flag and queue itself if it _does_ reach the commit state in time.
+
+            (But error handling in case of binlog write failure is currently
+            broken in other ways, as well).
           */
-          if (have_lock)
+          if (&wakeup_list->next_subsequent_commit == wakeup_next_ptr)
           {
-            have_lock= false;
-            cur->wakeup_subsequent_commits_running= true;
-            mysql_mutex_unlock(&cur->LOCK_wait_commit);
+            /* The last one in the list. */
+            wakeup_list->wakeup(0);
+            break;
           }
-          waiter->wakeup(0);
+          /*
+            Important: don't access wakeup_list->next after the wakeup() call,
+            it may be invalidated by the other thread.
+          */
+          next= wakeup_list->next_subsequent_commit;
+          wakeup_list->wakeup(0);
+          wakeup_list= next;
         }
-        waiter= next;
+        /*
+          We need a full memory barrier between walking the list and clearing
+          the flag wakeup_subsequent_commits_running. This barrier is needed
+          to ensure that no other thread will start to modify the list
+          pointers before we are done traversing the list.
+
+          But wait_for_commit::wakeup(), which was called above, does a full
+          memory barrier already (it locks a mutex).
+        */
+        cur->wakeup_subsequent_commits_running= false;
       }
-      if (have_lock)
+      else
         mysql_mutex_unlock(&cur->LOCK_wait_commit);
     }
     if (cur == last)
@@ -6827,29 +6898,6 @@ MYSQL_BIN_LOG::queue_for_group_commit(gr
     DBUG_ASSERT(entry != NULL);
   }
 
-  /*
-    Now we need to clear the wakeup_subsequent_commits_running flags.
-
-    We need a full memory barrier between walking the list above, and clearing
-    the flag wakeup_subsequent_commits_running below. This barrier is needed
-    to ensure that no other thread will start to modify the list pointers
-    before we are done traversing the list.
-
-    But wait_for_commit::wakeup(), which was called above for any other thread
-    that might modify the list in parallel, does a full memory barrier already
-    (it locks a mutex).
-  */
-  if (list)
-  {
-    for (;;)
-    {
-      list->wakeup_subsequent_commits_running= false;
-      if (list == last)
-        break;
-      list= list->next_subsequent_commit;
-    }
-  }
-
   if (opt_binlog_commit_wait_count > 0)
     mysql_cond_signal(&COND_prepare_ordered);
   mysql_mutex_unlock(&LOCK_prepare_ordered);

=== modified file 'sql/log_event.cc'
--- a/sql/log_event.cc	2013-11-22 23:50:54 +0000
+++ b/sql/log_event.cc	2013-12-13 13:26:51 +0000
@@ -7158,7 +7158,7 @@ int Xid_log_event::do_apply_event(rpl_gr
   res= trans_commit(thd); /* Automatically rolls back on error. */
   thd->mdl_context.release_transactional_locks();
 
-  if (sub_id)
+  if (!res && sub_id)
     rpl_global_gtid_slave_state.update_state_hash(sub_id, &gtid);
 
   /*

=== modified file 'sql/rpl_parallel.cc'
--- a/sql/rpl_parallel.cc	2013-12-06 12:28:23 +0000
+++ b/sql/rpl_parallel.cc	2013-12-13 13:26:51 +0000
@@ -24,7 +24,7 @@ static int
 rpt_handle_event(rpl_parallel_thread::queued_event *qev,
                  struct rpl_parallel_thread *rpt)
 {
-  int err __attribute__((unused));
+  int err;
   rpl_group_info *rgi= qev->rgi;
   Relay_log_info *rli= rgi->rli;
   THD *thd= rgi->thd;
@@ -172,6 +172,18 @@ finish_event_group(THD *thd, int err, ui
 }
 
 
+static void
+signal_error_to_sql_driver_thread(THD *thd, rpl_group_info *rgi)
+{
+  rgi->is_error= true;
+  rgi->cleanup_context(thd, true);
+  rgi->rli->abort_slave= true;
+  mysql_mutex_lock(rgi->rli->relay_log.get_log_lock());
+  mysql_mutex_unlock(rgi->rli->relay_log.get_log_lock());
+  rgi->rli->relay_log.signal_update();
+}
+
+
 pthread_handler_t
 handle_rpl_parallel_thread(void *arg)
 {
@@ -304,10 +316,8 @@ handle_rpl_parallel_thread(void *arg)
             {
               /* The thread got a kill signal. */
               thd->send_kill_message();
-              rgi->is_error= true;
               slave_output_error_info(rgi->rli, thd);
-              rgi->cleanup_context(thd, true);
-              rgi->rli->abort_slave= true;
+              signal_error_to_sql_driver_thread(thd, rgi);
             }
             rgi->wait_start_sub_id= 0;            /* No need to check again. */
           }
@@ -363,10 +373,8 @@ handle_rpl_parallel_thread(void *arg)
 
       if (err)
       {
-        rgi->is_error= true;
         slave_output_error_info(rgi->rli, thd);
-        rgi->cleanup_context(thd, true);
-        rgi->rli->abort_slave= true;
+        signal_error_to_sql_driver_thread(thd, rgi);
       }
       if (end_of_group)
       {
@@ -405,11 +413,9 @@ handle_rpl_parallel_thread(void *arg)
         half-processed event group.
       */
       mysql_mutex_unlock(&rpt->LOCK_rpl_thread);
-      group_rgi->is_error= true;
       finish_event_group(thd, 1, group_rgi->gtid_sub_id,
                          group_rgi->parallel_entry, &group_rgi->commit_orderer);
-      group_rgi->cleanup_context(thd, true);
-      group_rgi->rli->abort_slave= true;
+      signal_error_to_sql_driver_thread(thd, group_rgi);
       in_event_group= false;
       delete group_rgi;
       group_rgi= NULL;

=== modified file 'sql/slave.cc'
--- a/sql/slave.cc	2013-11-29 14:46:09 +0000
+++ b/sql/slave.cc	2013-12-13 13:26:51 +0000
@@ -6242,6 +6242,17 @@ static Log_event* next_event(rpl_group_i
         }
 
         /*
+          We have to check sql_slave_killed() here an extra time.
+          Otherwise we may miss a wakeup, since last check was done
+          without holding LOCK_log.
+        */
+        if (sql_slave_killed(rgi))
+        {
+          mysql_mutex_unlock(log_lock);
+          break;
+        }
+
+        /*
           If the I/O thread is blocked, unblock it.  Ok to broadcast
           after unlock, because the mutex is only destroyed in
           ~Relay_log_info(), i.e. when rli is destroyed, and rli will

=== modified file 'sql/sql_class.cc'
--- a/sql/sql_class.cc	2013-12-05 13:36:09 +0000
+++ b/sql/sql_class.cc	2013-12-13 13:26:51 +0000
@@ -5786,25 +5786,49 @@ int
 wait_for_commit::wait_for_prior_commit2(THD *thd)
 {
   const char *old_msg;
+  wait_for_commit *loc_waitee;
 
   mysql_mutex_lock(&LOCK_wait_commit);
   old_msg= thd->enter_cond(&COND_wait_commit, &LOCK_wait_commit,
                            "Waiting for prior transaction to commit");
   while (waiting_for_commit && !thd->check_killed())
     mysql_cond_wait(&COND_wait_commit, &LOCK_wait_commit);
-  thd->exit_cond(old_msg);
-  waitee= NULL;
   if (!waiting_for_commit)
   {
     if (wakeup_error)
       my_error(ER_PRIOR_COMMIT_FAILED, MYF(0));
-    return wakeup_error;
+    goto end;
   }
-  /* Wait was interrupted by kill, so give the error. */
+  /*
+    Wait was interrupted by kill. We need to unregister our wait and give the
+    error. But if a wakeup is already in progress, then we must ignore the
+    kill and not give error, otherwise we get inconsistency between waitee and
+    waiter as to whether we succeed or fail (eg. we may roll back but waitee
+    might attempt to commit both us and any subsequent commits waiting for us).
+  */
+  loc_waitee= this->waitee;
+  mysql_mutex_lock(&loc_waitee->LOCK_wait_commit);
+  if (loc_waitee->wakeup_subsequent_commits_running)
+  {
+    /* We are being woken up; ignore the kill and just wait. */
+    mysql_mutex_unlock(&loc_waitee->LOCK_wait_commit);
+    do
+    {
+      mysql_cond_wait(&COND_wait_commit, &LOCK_wait_commit);
+    } while (waiting_for_commit);
+    goto end;
+  }
+  remove_from_list(&loc_waitee->subsequent_commits_list);
+  mysql_mutex_unlock(&loc_waitee->LOCK_wait_commit);
+
   wakeup_error= thd->killed_errno();
   if (!wakeup_error)
     wakeup_error= ER_QUERY_INTERRUPTED;
   my_message(wakeup_error, ER(wakeup_error), MYF(0));
+
+end:
+  thd->exit_cond(old_msg);
+  waitee= NULL;
   return wakeup_error;
 }
 
@@ -5891,7 +5915,6 @@ wait_for_commit::unregister_wait_for_pri
   if (waiting_for_commit)
   {
     wait_for_commit *loc_waitee= this->waitee;
-    wait_for_commit **next_ptr_ptr, *cur;
     mysql_mutex_lock(&loc_waitee->LOCK_wait_commit);
     if (loc_waitee->wakeup_subsequent_commits_running)
     {
@@ -5909,17 +5932,7 @@ wait_for_commit::unregister_wait_for_pri
     else
     {
       /* Remove ourselves from the list in the waitee. */
-      next_ptr_ptr= &loc_waitee->subsequent_commits_list;
-      while ((cur= *next_ptr_ptr) != NULL)
-      {
-        if (cur == this)
-        {
-          *next_ptr_ptr= this->next_subsequent_commit;
-          break;
-        }
-        next_ptr_ptr= &cur->next_subsequent_commit;
-      }
-      waiting_for_commit= false;
+      remove_from_list(&loc_waitee->subsequent_commits_list);
       mysql_mutex_unlock(&loc_waitee->LOCK_wait_commit);
     }
   }

=== modified file 'sql/sql_class.h'
--- a/sql/sql_class.h	2013-12-05 13:36:09 +0000
+++ b/sql/sql_class.h	2013-12-13 13:26:51 +0000
@@ -1660,6 +1660,25 @@ struct wait_for_commit
     if (waiting_for_commit)
       unregister_wait_for_prior_commit2();
   }
+  /*
+    Remove a waiter from the list in the waitee. Used to unregister a wait.
+    The caller must be holding the locks of both waiter and waitee.
+  */
+  void remove_from_list(wait_for_commit **next_ptr_ptr)
+  {
+    wait_for_commit *cur;
+
+    while ((cur= *next_ptr_ptr) != NULL)
+    {
+      if (cur == this)
+      {
+        *next_ptr_ptr= this->next_subsequent_commit;
+        break;
+      }
+      next_ptr_ptr= &cur->next_subsequent_commit;
+    }
+    waiting_for_commit= false;
+  }
 
   void wakeup(int wakeup_error);
 



More information about the commits mailing list