mariadb/sql/event_queue.cc

825 lines
22 KiB
C++
Raw Normal View History

2011-06-30 17:46:53 +02:00
/* Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
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
2011-06-30 17:46:53 +02:00
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */
#include "sql_priv.h"
#include "unireg.h"
#include "event_queue.h"
#include "event_data_objects.h"
Patch that refactors global read lock implementation and fixes bug #57006 "Deadlock between HANDLER and FLUSH TABLES WITH READ LOCK" and bug #54673 "It takes too long to get readlock for 'FLUSH TABLES WITH READ LOCK'". The first bug manifested itself as a deadlock which occurred when a connection, which had some table open through HANDLER statement, tried to update some data through DML statement while another connection tried to execute FLUSH TABLES WITH READ LOCK concurrently. What happened was that FTWRL in the second connection managed to perform first step of GRL acquisition and thus blocked all upcoming DML. After that it started to wait for table open through HANDLER statement to be flushed. When the first connection tried to execute DML it has started to wait for GRL/the second connection creating deadlock. The second bug manifested itself as starvation of FLUSH TABLES WITH READ LOCK statements in cases when there was a constant stream of concurrent DML statements (in two or more connections). This has happened because requests for protection against GRL which were acquired by DML statements were ignoring presence of pending GRL and thus the latter was starved. This patch solves both these problems by re-implementing GRL using metadata locks. Similar to the old implementation acquisition of GRL in new implementation is two-step. During the first step we block all concurrent DML and DDL statements by acquiring global S metadata lock (each DML and DDL statement acquires global IX lock for its duration). During the second step we block commits by acquiring global S lock in COMMIT namespace (commit code acquires global IX lock in this namespace). Note that unlike in old implementation acquisition of protection against GRL in DML and DDL is semi-automatic. We assume that any statement which should be blocked by GRL will either open and acquires write-lock on tables or acquires metadata locks on objects it is going to modify. For any such statement global IX metadata lock is automatically acquired for its duration. The first problem is solved because waits for GRL become visible to deadlock detector in metadata locking subsystem and thus deadlocks like one in the first bug become impossible. The second problem is solved because global S locks which are used for GRL implementation are given preference over IX locks which are acquired by concurrent DML (and we can switch to fair scheduling in future if needed). Important change: FTWRL/GRL no longer blocks DML and DDL on temporary tables. Before this patch behavior was not consistent in this respect: in some cases DML/DDL statements on temporary tables were blocked while in others they were not. Since the main use cases for FTWRL are various forms of backups and temporary tables are not preserved during backups we have opted for consistently allowing DML/DDL on temporary tables during FTWRL/GRL. Important change: This patch changes thread state names which are used when DML/DDL of FTWRL is waiting for global read lock. It is now either "Waiting for global read lock" or "Waiting for commit lock" depending on the stage on which FTWRL is. Incompatible change: To solve deadlock in events code which was exposed by this patch we have to replace LOCK_event_metadata mutex with metadata locks on events. As result we have to prohibit DDL on events under LOCK TABLES. This patch also adds extensive test coverage for interaction of DML/DDL and FTWRL. Performance of new and old global read lock implementations in sysbench tests were compared. There were no significant difference between new and old implementations.
2010-11-11 20:11:05 +03:00
#include "event_db_repository.h"
#include "events.h"
#include "sql_audit.h"
#include "tztime.h" // my_tz_find, my_tz_OFFSET0, struct Time_zone
#include "log.h" // sql_print_error
#include "sql_class.h" // struct THD
/**
@addtogroup Event_Scheduler
@{
*/
#define EVENT_QUEUE_INITIAL_SIZE 30
#define EVENT_QUEUE_EXTENT 30
#ifdef __GNUC__
#if __GNUC__ >= 2
#define SCHED_FUNC __FUNCTION__
#endif
#else
#define SCHED_FUNC "<unknown>"
#endif
#define LOCK_QUEUE_DATA() lock_data(SCHED_FUNC, __LINE__)
#define UNLOCK_QUEUE_DATA() unlock_data(SCHED_FUNC, __LINE__)
/*
Compares the execute_at members of two Event_queue_element instances.
Used as callback for the prioritized queue when shifting
elements inside.
SYNOPSIS
event_queue_element_data_compare_q()
vptr Not used (set it to NULL)
a First Event_queue_element object
b Second Event_queue_element object
RETURN VALUE
-1 a->execute_at < b->execute_at
0 a->execute_at == b->execute_at
1 a->execute_at > b->execute_at
NOTES
execute_at.second_part is not considered during comparison
*/
extern "C" int event_queue_element_compare_q(void *, uchar *, uchar *);
int event_queue_element_compare_q(void *vptr, uchar* a, uchar *b)
{
Event_queue_element *left = (Event_queue_element *)a;
Event_queue_element *right = (Event_queue_element *)b;
my_time_t lhs = left->execute_at;
my_time_t rhs = right->execute_at;
if (left->status == Event_parse_data::DISABLED)
return right->status != Event_parse_data::DISABLED;
if (right->status == Event_parse_data::DISABLED)
return 1;
return (lhs < rhs ? -1 : (lhs > rhs ? 1 : 0));
}
/*
Constructor of class Event_queue.
SYNOPSIS
Event_queue::Event_queue()
*/
Event_queue::Event_queue()
:next_activation_at(0),
mutex_last_locked_at_line(0),
mutex_last_unlocked_at_line(0),
2007-05-22 16:19:36 -06:00
mutex_last_attempted_lock_at_line(0),
mutex_last_locked_in_func("n/a"),
mutex_last_unlocked_in_func("n/a"),
mutex_last_attempted_lock_in_func("n/a"),
mutex_queue_data_locked(FALSE),
2007-05-22 16:19:36 -06:00
mutex_queue_data_attempting_lock(FALSE),
waiting_on_cond(FALSE)
{
mysql_mutex_init(key_LOCK_event_queue, &LOCK_event_queue, MY_MUTEX_INIT_FAST);
mysql_cond_init(key_COND_queue_state, &COND_queue_state, NULL);
}
Event_queue::~Event_queue()
{
deinit_queue();
mysql_mutex_destroy(&LOCK_event_queue);
mysql_cond_destroy(&COND_queue_state);
}
/*
This is a queue's constructor. Until this method is called, the
queue is unusable. We don't use a C++ constructor instead in
order to be able to check the return value. The queue is
initialized once at server startup. Initialization can fail in
case of a failure reading events from the database or out of
memory.
SYNOPSIS
Event_queue::init()
RETURN VALUE
FALSE OK
TRUE Error
*/
bool
Event_queue::init_queue(THD *thd)
{
DBUG_ENTER("Event_queue::init_queue");
DBUG_PRINT("enter", ("this: 0x%lx", (long) this));
LOCK_QUEUE_DATA();
if (init_queue_ex(&queue, EVENT_QUEUE_INITIAL_SIZE , 0 /*offset*/,
0 /*max_on_top*/, event_queue_element_compare_q,
NULL, EVENT_QUEUE_EXTENT))
{
sql_print_error("Event Scheduler: Can't initialize the execution queue");
goto err;
}
UNLOCK_QUEUE_DATA();
DBUG_RETURN(FALSE);
err:
UNLOCK_QUEUE_DATA();
DBUG_RETURN(TRUE);
}
/*
Deinits the queue. Remove all elements from it and destroys them
too.
SYNOPSIS
Event_queue::deinit_queue()
*/
void
Event_queue::deinit_queue()
{
DBUG_ENTER("Event_queue::deinit_queue");
LOCK_QUEUE_DATA();
empty_queue();
delete_queue(&queue);
UNLOCK_QUEUE_DATA();
DBUG_VOID_RETURN;
}
/**
Adds an event to the queue.
Compute the next execution time for an event, and if it is still
active, add it to the queue. Otherwise delete it.
The object is left intact in case of an error. Otherwise
the queue container assumes ownership of it.
@param[in] thd thread handle
@param[in] new_element a new element to add to the queue
@param[out] created set to TRUE if no error and the element is
added to the queue, FALSE otherwise
@retval TRUE an error occured. The value of created is undefined,
the element was not deleted.
@retval FALSE success
*/
bool
Event_queue::create_event(THD *thd, Event_queue_element *new_element,
bool *created)
{
DBUG_ENTER("Event_queue::create_event");
2007-03-01 16:39:00 -07:00
DBUG_PRINT("enter", ("thd: 0x%lx et=%s.%s", (long) thd,
new_element->dbname.str, new_element->name.str));
/* Will do nothing if the event is disabled */
new_element->compute_next_execution_time();
if (new_element->status != Event_parse_data::ENABLED)
{
delete new_element;
*created= FALSE;
DBUG_RETURN(FALSE);
}
DBUG_PRINT("info", ("new event in the queue: 0x%lx", (long) new_element));
LOCK_QUEUE_DATA();
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 12:59:39 +03:00
*created= (queue_insert_safe(&queue, (uchar *) new_element) == FALSE);
dbug_dump_queue(thd->query_start());
mysql_cond_broadcast(&COND_queue_state);
UNLOCK_QUEUE_DATA();
DBUG_RETURN(!*created);
}
/*
Updates an event from the scheduler queue
SYNOPSIS
Event_queue::update_event()
thd Thread
dbname Schema of the event
name Name of the event
new_schema New schema, in case of RENAME TO, otherwise NULL
new_name New name, in case of RENAME TO, otherwise NULL
*/
void
Event_queue::update_event(THD *thd, LEX_STRING dbname, LEX_STRING name,
Event_queue_element *new_element)
{
DBUG_ENTER("Event_queue::update_event");
DBUG_PRINT("enter", ("thd: 0x%lx et=[%s.%s]", (long) thd, dbname.str, name.str));
if ((new_element->status == Event_parse_data::DISABLED) ||
(new_element->status == Event_parse_data::SLAVESIDE_DISABLED))
{
DBUG_PRINT("info", ("The event is disabled."));
/*
Destroy the object but don't skip to end: because we may have to remove
object from the cache.
*/
delete new_element;
new_element= NULL;
}
else
new_element->compute_next_execution_time();
LOCK_QUEUE_DATA();
find_n_remove_event(dbname, name);
/* If not disabled event */
if (new_element)
{
DBUG_PRINT("info", ("new event in the queue: 0x%lx", (long) new_element));
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 12:59:39 +03:00
queue_insert_safe(&queue, (uchar *) new_element);
mysql_cond_broadcast(&COND_queue_state);
}
dbug_dump_queue(thd->query_start());
UNLOCK_QUEUE_DATA();
DBUG_VOID_RETURN;
}
/*
Drops an event from the queue
SYNOPSIS
Event_queue::drop_event()
thd Thread
dbname Schema of the event to drop
name Name of the event to drop
*/
void
Event_queue::drop_event(THD *thd, LEX_STRING dbname, LEX_STRING name)
{
DBUG_ENTER("Event_queue::drop_event");
DBUG_PRINT("enter", ("thd: 0x%lx db :%s name: %s", (long) thd,
dbname.str, name.str));
LOCK_QUEUE_DATA();
find_n_remove_event(dbname, name);
dbug_dump_queue(thd->query_start());
UNLOCK_QUEUE_DATA();
/*
We don't signal here because the scheduler will catch the change
next time it wakes up.
*/
DBUG_VOID_RETURN;
}
/*
Drops all events from the in-memory queue and disk that match
certain pattern evaluated by a comparator function
SYNOPSIS
Event_queue::drop_matching_events()
thd THD
pattern A pattern string
comparator The function to use for comparing
RETURN VALUE
>=0 Number of dropped events
NOTE
Expected is the caller to acquire lock on LOCK_event_queue
*/
void
Event_queue::drop_matching_events(THD *thd, LEX_STRING pattern,
bool (*comparator)(LEX_STRING, Event_basic *))
{
uint i= 0;
DBUG_ENTER("Event_queue::drop_matching_events");
DBUG_PRINT("enter", ("pattern=%s", pattern.str));
while (i < queue.elements)
{
Event_queue_element *et= (Event_queue_element *) queue_element(&queue, i);
DBUG_PRINT("info", ("[%s.%s]?", et->dbname.str, et->name.str));
if (comparator(pattern, et))
{
/*
The queue is ordered. If we remove an element, then all elements
after it will shift one position to the left, if we imagine it as
an array from left to the right. In this case we should not
increment the counter and the (i < queue.elements) condition is ok.
*/
queue_remove(&queue, i);
delete et;
}
else
i++;
}
/*
We don't call mysql_cond_broadcast(&COND_queue_state);
If we remove the top event:
1. The queue is empty. The scheduler will wake up at some time and
realize that the queue is empty. If create_event() comes inbetween
it will signal the scheduler
2. The queue is not empty, but the next event after the previous top,
won't be executed any time sooner than the element we removed. Hence,
we may not notify the scheduler and it will realize the change when it
wakes up from timedwait.
*/
DBUG_VOID_RETURN;
}
/*
Drops all events from the in-memory queue and disk that are from
certain schema.
SYNOPSIS
Event_queue::drop_schema_events()
thd HD
schema The schema name
*/
void
Event_queue::drop_schema_events(THD *thd, LEX_STRING schema)
{
DBUG_ENTER("Event_queue::drop_schema_events");
LOCK_QUEUE_DATA();
drop_matching_events(thd, schema, event_basic_db_equal);
UNLOCK_QUEUE_DATA();
DBUG_VOID_RETURN;
}
/*
Searches for an event in the queue
SYNOPSIS
Event_queue::find_n_remove_event()
db The schema of the event to find
name The event to find
NOTE
The caller should do the locking also the caller is responsible for
actual signalling in case an event is removed from the queue.
*/
void
Event_queue::find_n_remove_event(LEX_STRING db, LEX_STRING name)
{
uint i;
DBUG_ENTER("Event_queue::find_n_remove_event");
for (i= 0; i < queue.elements; ++i)
{
Event_queue_element *et= (Event_queue_element *) queue_element(&queue, i);
DBUG_PRINT("info", ("[%s.%s]==[%s.%s]?", db.str, name.str,
et->dbname.str, et->name.str));
if (event_basic_identifier_equal(db, name, et))
{
queue_remove(&queue, i);
delete et;
break;
}
}
DBUG_VOID_RETURN;
}
/*
Recalculates activation times in the queue. There is one reason for
that. Because the values (execute_at) by which the queue is ordered are
changed by calls to compute_next_execution_time() on a request from the
scheduler thread, if it is not running then the values won't be updated.
Once the scheduler is started again the values has to be recalculated
so they are right for the current time.
SYNOPSIS
Event_queue::recalculate_activation_times()
thd Thread
*/
void
Event_queue::recalculate_activation_times(THD *thd)
{
uint i;
DBUG_ENTER("Event_queue::recalculate_activation_times");
LOCK_QUEUE_DATA();
DBUG_PRINT("info", ("%u loaded events to be recalculated", queue.elements));
for (i= 0; i < queue.elements; i++)
{
((Event_queue_element*)queue_element(&queue, i))->compute_next_execution_time();
}
queue_fix(&queue);
/*
The disabled elements are moved to the end during the `fix`.
Start from the end and remove all of the elements which are
disabled. When we find the first non-disabled one we break, as we
have removed all. The queue has been ordered in a way the disabled
events are at the end.
*/
for (i= queue.elements; i > 0; i--)
{
Event_queue_element *element = (Event_queue_element*)queue_element(&queue, i - 1);
if (element->status != Event_parse_data::DISABLED)
break;
/*
This won't cause queue re-order, because we remove
always the last element.
*/
queue_remove(&queue, i - 1);
delete element;
}
UNLOCK_QUEUE_DATA();
/*
XXX: The events are dropped only from memory and not from disk
even if `drop_list[j]->dropped` is TRUE. There will be still on the
disk till next server restart.
Please add code here to do it.
*/
DBUG_VOID_RETURN;
}
/*
Empties the queue and destroys the Event_queue_element objects in the
queue.
SYNOPSIS
Event_queue::empty_queue()
NOTE
Should be called with LOCK_event_queue locked
*/
void
Event_queue::empty_queue()
{
uint i;
DBUG_ENTER("Event_queue::empty_queue");
DBUG_PRINT("enter", ("Purging the queue. %u element(s)", queue.elements));
sql_print_information("Event Scheduler: Purging the queue. %u events",
queue.elements);
/* empty the queue */
for (i= 0; i < queue.elements; ++i)
{
Event_queue_element *et= (Event_queue_element *) queue_element(&queue, i);
delete et;
}
resize_queue(&queue, 0);
DBUG_VOID_RETURN;
}
/*
Dumps the queue to the trace log.
SYNOPSIS
Event_queue::dbug_dump_queue()
now Current timestamp
*/
void
Event_queue::dbug_dump_queue(time_t now)
{
#ifndef DBUG_OFF
Event_queue_element *et;
uint i;
DBUG_ENTER("Event_queue::dbug_dump_queue");
DBUG_PRINT("info", ("Dumping queue . Elements=%u", queue.elements));
for (i = 0; i < queue.elements; i++)
{
et= ((Event_queue_element*)queue_element(&queue, i));
DBUG_PRINT("info", ("et: 0x%lx name: %s.%s", (long) et,
et->dbname.str, et->name.str));
DBUG_PRINT("info", ("exec_at: %lu starts: %lu ends: %lu execs_so_far: %u "
"expr: %ld et.exec_at: %ld now: %ld "
"(et.exec_at - now): %d if: %d",
(long) et->execute_at, (long) et->starts,
(long) et->ends, et->execution_count,
(long) et->expression, (long) et->execute_at,
(long) now, (int) (et->execute_at - now),
et->execute_at <= now));
}
DBUG_VOID_RETURN;
#endif
}
static const char *queue_empty_msg= "Waiting on empty queue";
static const char *queue_wait_msg= "Waiting for next activation";
/*
Checks whether the top of the queue is elligible for execution and
returns an Event_job_data instance in case it should be executed.
`now` is compared against `execute_at` of the top element in the queue.
SYNOPSIS
Event_queue::get_top_for_execution_if_time()
thd [in] Thread
event_name [out] The object to execute
RETURN VALUE
FALSE No error. event_name != NULL
TRUE Serious error
*/
bool
Event_queue::get_top_for_execution_if_time(THD *thd,
Event_queue_element_for_exec **event_name)
{
bool ret= FALSE;
*event_name= NULL;
my_time_t UNINIT_VAR(last_executed);
int UNINIT_VAR(status);
DBUG_ENTER("Event_queue::get_top_for_execution_if_time");
LOCK_QUEUE_DATA();
for (;;)
{
Event_queue_element *top= NULL;
/* Break loop if thd has been killed */
if (thd->killed)
{
DBUG_PRINT("info", ("thd->killed=%d", thd->killed));
goto end;
}
if (!queue.elements)
{
/* There are no events in the queue */
next_activation_at= 0;
/* Release any held audit resources before waiting */
mysql_audit_release(thd);
/* Wait on condition until signaled. Release LOCK_queue while waiting. */
cond_wait(thd, NULL, queue_empty_msg, SCHED_FUNC, __LINE__);
continue;
}
top= ((Event_queue_element*) queue_element(&queue, 0));
thd->set_current_time(); /* Get current time */
next_activation_at= top->execute_at;
if (next_activation_at > thd->query_start())
{
/*
Not yet time for top event, wait on condition with
time or until signaled. Release LOCK_queue while waiting.
*/
struct timespec top_time;
set_timespec(top_time, next_activation_at - thd->query_start());
/* Release any held audit resources before waiting */
mysql_audit_release(thd);
cond_wait(thd, &top_time, queue_wait_msg, SCHED_FUNC, __LINE__);
continue;
}
if (!(*event_name= new Event_queue_element_for_exec()) ||
(*event_name)->init(top->dbname, top->name))
{
ret= TRUE;
break;
}
DBUG_PRINT("info", ("Ready for execution"));
top->mark_last_executed(thd);
if (top->compute_next_execution_time())
top->status= Event_parse_data::DISABLED;
DBUG_PRINT("info", ("event %s status is %d", top->name.str, top->status));
top->execution_count++;
(*event_name)->dropped= top->dropped;
Patch that refactors global read lock implementation and fixes bug #57006 "Deadlock between HANDLER and FLUSH TABLES WITH READ LOCK" and bug #54673 "It takes too long to get readlock for 'FLUSH TABLES WITH READ LOCK'". The first bug manifested itself as a deadlock which occurred when a connection, which had some table open through HANDLER statement, tried to update some data through DML statement while another connection tried to execute FLUSH TABLES WITH READ LOCK concurrently. What happened was that FTWRL in the second connection managed to perform first step of GRL acquisition and thus blocked all upcoming DML. After that it started to wait for table open through HANDLER statement to be flushed. When the first connection tried to execute DML it has started to wait for GRL/the second connection creating deadlock. The second bug manifested itself as starvation of FLUSH TABLES WITH READ LOCK statements in cases when there was a constant stream of concurrent DML statements (in two or more connections). This has happened because requests for protection against GRL which were acquired by DML statements were ignoring presence of pending GRL and thus the latter was starved. This patch solves both these problems by re-implementing GRL using metadata locks. Similar to the old implementation acquisition of GRL in new implementation is two-step. During the first step we block all concurrent DML and DDL statements by acquiring global S metadata lock (each DML and DDL statement acquires global IX lock for its duration). During the second step we block commits by acquiring global S lock in COMMIT namespace (commit code acquires global IX lock in this namespace). Note that unlike in old implementation acquisition of protection against GRL in DML and DDL is semi-automatic. We assume that any statement which should be blocked by GRL will either open and acquires write-lock on tables or acquires metadata locks on objects it is going to modify. For any such statement global IX metadata lock is automatically acquired for its duration. The first problem is solved because waits for GRL become visible to deadlock detector in metadata locking subsystem and thus deadlocks like one in the first bug become impossible. The second problem is solved because global S locks which are used for GRL implementation are given preference over IX locks which are acquired by concurrent DML (and we can switch to fair scheduling in future if needed). Important change: FTWRL/GRL no longer blocks DML and DDL on temporary tables. Before this patch behavior was not consistent in this respect: in some cases DML/DDL statements on temporary tables were blocked while in others they were not. Since the main use cases for FTWRL are various forms of backups and temporary tables are not preserved during backups we have opted for consistently allowing DML/DDL on temporary tables during FTWRL/GRL. Important change: This patch changes thread state names which are used when DML/DDL of FTWRL is waiting for global read lock. It is now either "Waiting for global read lock" or "Waiting for commit lock" depending on the stage on which FTWRL is. Incompatible change: To solve deadlock in events code which was exposed by this patch we have to replace LOCK_event_metadata mutex with metadata locks on events. As result we have to prohibit DDL on events under LOCK TABLES. This patch also adds extensive test coverage for interaction of DML/DDL and FTWRL. Performance of new and old global read lock implementations in sysbench tests were compared. There were no significant difference between new and old implementations.
2010-11-11 20:11:05 +03:00
/*
Save new values of last_executed timestamp and event status on stack
in order to be able to update event description in system table once
QUEUE_DATA lock is released.
*/
last_executed= top->last_executed;
status= top->status;
if (top->status == Event_parse_data::DISABLED)
{
DBUG_PRINT("info", ("removing from the queue"));
sql_print_information("Event Scheduler: Last execution of %s.%s. %s",
top->dbname.str, top->name.str,
top->dropped? "Dropping.":"");
delete top;
queue_remove(&queue, 0);
}
else
queue_replaced(&queue);
dbug_dump_queue(thd->query_start());
break;
}
end:
UNLOCK_QUEUE_DATA();
DBUG_PRINT("info", ("returning %d et_new: 0x%lx ",
ret, (long) *event_name));
if (*event_name)
Patch that refactors global read lock implementation and fixes bug #57006 "Deadlock between HANDLER and FLUSH TABLES WITH READ LOCK" and bug #54673 "It takes too long to get readlock for 'FLUSH TABLES WITH READ LOCK'". The first bug manifested itself as a deadlock which occurred when a connection, which had some table open through HANDLER statement, tried to update some data through DML statement while another connection tried to execute FLUSH TABLES WITH READ LOCK concurrently. What happened was that FTWRL in the second connection managed to perform first step of GRL acquisition and thus blocked all upcoming DML. After that it started to wait for table open through HANDLER statement to be flushed. When the first connection tried to execute DML it has started to wait for GRL/the second connection creating deadlock. The second bug manifested itself as starvation of FLUSH TABLES WITH READ LOCK statements in cases when there was a constant stream of concurrent DML statements (in two or more connections). This has happened because requests for protection against GRL which were acquired by DML statements were ignoring presence of pending GRL and thus the latter was starved. This patch solves both these problems by re-implementing GRL using metadata locks. Similar to the old implementation acquisition of GRL in new implementation is two-step. During the first step we block all concurrent DML and DDL statements by acquiring global S metadata lock (each DML and DDL statement acquires global IX lock for its duration). During the second step we block commits by acquiring global S lock in COMMIT namespace (commit code acquires global IX lock in this namespace). Note that unlike in old implementation acquisition of protection against GRL in DML and DDL is semi-automatic. We assume that any statement which should be blocked by GRL will either open and acquires write-lock on tables or acquires metadata locks on objects it is going to modify. For any such statement global IX metadata lock is automatically acquired for its duration. The first problem is solved because waits for GRL become visible to deadlock detector in metadata locking subsystem and thus deadlocks like one in the first bug become impossible. The second problem is solved because global S locks which are used for GRL implementation are given preference over IX locks which are acquired by concurrent DML (and we can switch to fair scheduling in future if needed). Important change: FTWRL/GRL no longer blocks DML and DDL on temporary tables. Before this patch behavior was not consistent in this respect: in some cases DML/DDL statements on temporary tables were blocked while in others they were not. Since the main use cases for FTWRL are various forms of backups and temporary tables are not preserved during backups we have opted for consistently allowing DML/DDL on temporary tables during FTWRL/GRL. Important change: This patch changes thread state names which are used when DML/DDL of FTWRL is waiting for global read lock. It is now either "Waiting for global read lock" or "Waiting for commit lock" depending on the stage on which FTWRL is. Incompatible change: To solve deadlock in events code which was exposed by this patch we have to replace LOCK_event_metadata mutex with metadata locks on events. As result we have to prohibit DDL on events under LOCK TABLES. This patch also adds extensive test coverage for interaction of DML/DDL and FTWRL. Performance of new and old global read lock implementations in sysbench tests were compared. There were no significant difference between new and old implementations.
2010-11-11 20:11:05 +03:00
{
DBUG_PRINT("info", ("db: %s name: %s",
(*event_name)->dbname.str, (*event_name)->name.str));
Patch that refactors global read lock implementation and fixes bug #57006 "Deadlock between HANDLER and FLUSH TABLES WITH READ LOCK" and bug #54673 "It takes too long to get readlock for 'FLUSH TABLES WITH READ LOCK'". The first bug manifested itself as a deadlock which occurred when a connection, which had some table open through HANDLER statement, tried to update some data through DML statement while another connection tried to execute FLUSH TABLES WITH READ LOCK concurrently. What happened was that FTWRL in the second connection managed to perform first step of GRL acquisition and thus blocked all upcoming DML. After that it started to wait for table open through HANDLER statement to be flushed. When the first connection tried to execute DML it has started to wait for GRL/the second connection creating deadlock. The second bug manifested itself as starvation of FLUSH TABLES WITH READ LOCK statements in cases when there was a constant stream of concurrent DML statements (in two or more connections). This has happened because requests for protection against GRL which were acquired by DML statements were ignoring presence of pending GRL and thus the latter was starved. This patch solves both these problems by re-implementing GRL using metadata locks. Similar to the old implementation acquisition of GRL in new implementation is two-step. During the first step we block all concurrent DML and DDL statements by acquiring global S metadata lock (each DML and DDL statement acquires global IX lock for its duration). During the second step we block commits by acquiring global S lock in COMMIT namespace (commit code acquires global IX lock in this namespace). Note that unlike in old implementation acquisition of protection against GRL in DML and DDL is semi-automatic. We assume that any statement which should be blocked by GRL will either open and acquires write-lock on tables or acquires metadata locks on objects it is going to modify. For any such statement global IX metadata lock is automatically acquired for its duration. The first problem is solved because waits for GRL become visible to deadlock detector in metadata locking subsystem and thus deadlocks like one in the first bug become impossible. The second problem is solved because global S locks which are used for GRL implementation are given preference over IX locks which are acquired by concurrent DML (and we can switch to fair scheduling in future if needed). Important change: FTWRL/GRL no longer blocks DML and DDL on temporary tables. Before this patch behavior was not consistent in this respect: in some cases DML/DDL statements on temporary tables were blocked while in others they were not. Since the main use cases for FTWRL are various forms of backups and temporary tables are not preserved during backups we have opted for consistently allowing DML/DDL on temporary tables during FTWRL/GRL. Important change: This patch changes thread state names which are used when DML/DDL of FTWRL is waiting for global read lock. It is now either "Waiting for global read lock" or "Waiting for commit lock" depending on the stage on which FTWRL is. Incompatible change: To solve deadlock in events code which was exposed by this patch we have to replace LOCK_event_metadata mutex with metadata locks on events. As result we have to prohibit DDL on events under LOCK TABLES. This patch also adds extensive test coverage for interaction of DML/DDL and FTWRL. Performance of new and old global read lock implementations in sysbench tests were compared. There were no significant difference between new and old implementations.
2010-11-11 20:11:05 +03:00
Event_db_repository *db_repository= Events::get_db_repository();
(void) db_repository->update_timing_fields_for_event(thd,
(*event_name)->dbname, (*event_name)->name,
last_executed, (ulonglong) status);
}
DBUG_RETURN(ret);
}
/*
Auxiliary function for locking LOCK_event_queue. Used by the
LOCK_QUEUE_DATA macro
SYNOPSIS
Event_queue::lock_data()
func Which function is requesting mutex lock
line On which line mutex lock is requested
*/
void
Event_queue::lock_data(const char *func, uint line)
{
DBUG_ENTER("Event_queue::lock_data");
DBUG_PRINT("enter", ("func=%s line=%u", func, line));
mutex_last_attempted_lock_in_func= func;
mutex_last_attempted_lock_at_line= line;
mutex_queue_data_attempting_lock= TRUE;
mysql_mutex_lock(&LOCK_event_queue);
mutex_last_attempted_lock_in_func= "";
mutex_last_attempted_lock_at_line= 0;
mutex_queue_data_attempting_lock= FALSE;
mutex_last_locked_in_func= func;
mutex_last_locked_at_line= line;
mutex_queue_data_locked= TRUE;
DBUG_VOID_RETURN;
}
/*
Auxiliary function for unlocking LOCK_event_queue. Used by the
UNLOCK_QUEUE_DATA macro
SYNOPSIS
Event_queue::unlock_data()
func Which function is requesting mutex unlock
line On which line mutex unlock is requested
*/
void
Event_queue::unlock_data(const char *func, uint line)
{
DBUG_ENTER("Event_queue::unlock_data");
DBUG_PRINT("enter", ("func=%s line=%u", func, line));
mutex_last_unlocked_at_line= line;
mutex_queue_data_locked= FALSE;
mutex_last_unlocked_in_func= func;
mysql_mutex_unlock(&LOCK_event_queue);
DBUG_VOID_RETURN;
}
/*
Wrapper for mysql_cond_wait/timedwait
SYNOPSIS
Event_queue::cond_wait()
thd Thread (Could be NULL during shutdown procedure)
msg Message for thd->proc_info
abstime If not null then call mysql_cond_timedwait()
func Which function is requesting cond_wait
line On which line cond_wait is requested
*/
void
Event_queue::cond_wait(THD *thd, struct timespec *abstime, const char* msg,
const char *func, uint line)
{
DBUG_ENTER("Event_queue::cond_wait");
waiting_on_cond= TRUE;
mutex_last_unlocked_at_line= line;
mutex_queue_data_locked= FALSE;
mutex_last_unlocked_in_func= func;
thd->enter_cond(&COND_queue_state, &LOCK_event_queue, msg);
Bug#37780: Make KILL reliable (main.kill fails randomly) - A prerequisite cleanup patch for making KILL reliable. The test case main.kill did not work reliably. The following problems have been identified: 1. A kill signal could go lost if it came in, short before a thread went reading on the client connection. 2. A kill signal could go lost if it came in, short before a thread went waiting on a condition variable. These problems have been solved as follows. Please see also added code comments for more details. 1. There is no safe way to detect, when a thread enters the blocking state of a read(2) or recv(2) system call, where it can be interrupted by a signal. Hence it is not possible to wait for the right moment to send a kill signal. It has been decided, not to fix it in the code. Instead, the test case repeats the KILL statement until the connection terminates. 2. Before waiting on a condition variable, we register it together with a synchronizating mutex in THD::mysys_var. After this, we need to test THD::killed again. At some places we did only test it in a loop condition before the registration. When THD::killed had been set between this test and the registration, we entered waiting without noticing the killed flag. Additional checks ahve been introduced where required. In addition to the above, a re-write of the main.kill test case has been done. All sleeps have been replaced by Debug Sync Facility synchronization. A couple of sync points have been added to the server code. To avoid further problems, if the test case fails in spite of the fixes, the test case has been added to the "experimental" list for now. - Most of the work on this patch is authored by Ingo Struewing
2010-10-22 09:58:09 -02:00
if (!thd->killed)
{
if (!abstime)
mysql_cond_wait(&COND_queue_state, &LOCK_event_queue);
else
mysql_cond_timedwait(&COND_queue_state, &LOCK_event_queue, abstime);
}
mutex_last_locked_in_func= func;
mutex_last_locked_at_line= line;
mutex_queue_data_locked= TRUE;
waiting_on_cond= FALSE;
/*
This will free the lock so we need to relock. Not the best thing to
do but we need to obey cond_wait()
*/
thd->exit_cond("");
lock_data(func, line);
DBUG_VOID_RETURN;
}
/*
Dumps the internal status of the queue
SYNOPSIS
Event_queue::dump_internal_status()
*/
void
Event_queue::dump_internal_status()
{
DBUG_ENTER("Event_queue::dump_internal_status");
/* element count */
puts("");
puts("Event queue status:");
printf("Element count : %u\n", queue.elements);
printf("Data locked : %s\n", mutex_queue_data_locked? "YES":"NO");
printf("Attempting lock : %s\n", mutex_queue_data_attempting_lock? "YES":"NO");
printf("LLA : %s:%u\n", mutex_last_locked_in_func,
mutex_last_locked_at_line);
printf("LUA : %s:%u\n", mutex_last_unlocked_in_func,
mutex_last_unlocked_at_line);
if (mutex_last_attempted_lock_at_line)
printf("Last lock attempt at: %s:%u\n", mutex_last_attempted_lock_in_func,
mutex_last_attempted_lock_at_line);
printf("WOC : %s\n", waiting_on_cond? "YES":"NO");
MYSQL_TIME time;
my_tz_OFFSET0->gmt_sec_to_TIME(&time, next_activation_at);
if (time.year != 1970)
printf("Next activation : %04d-%02d-%02d %02d:%02d:%02d\n",
time.year, time.month, time.day, time.hour, time.minute, time.second);
else
printf("Next activation : never");
DBUG_VOID_RETURN;
}
/**
@} (End of group Event_Scheduler)
*/