2001-12-06 13:10:51 +01:00
|
|
|
/* Copyright (C) 2000 MySQL 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; either version 2 of the License, or
|
|
|
|
(at your option) any later version.
|
|
|
|
|
|
|
|
This program is distributed in the hope that it will be useful,
|
2000-07-31 21:29:14 +02:00
|
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
2001-12-06 13:10:51 +01:00
|
|
|
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */
|
2000-07-31 21:29:14 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
Read and write locks for Posix threads. All tread must acquire
|
|
|
|
all locks it needs through thr_multi_lock() to avoid dead-locks.
|
|
|
|
A lock consists of a master lock (THR_LOCK), and lock instances
|
|
|
|
(THR_LOCK_DATA).
|
|
|
|
Any thread can have any number of lock instances (read and write:s) on
|
|
|
|
any lock. All lock instances must be freed.
|
|
|
|
Locks are prioritized according to:
|
|
|
|
|
|
|
|
The current lock types are:
|
|
|
|
|
|
|
|
TL_READ # Low priority read
|
2001-02-17 23:03:37 +01:00
|
|
|
TL_READ_WITH_SHARED_LOCKS
|
2000-07-31 21:29:14 +02:00
|
|
|
TL_READ_HIGH_PRIORITY # High priority read
|
|
|
|
TL_READ_NO_INSERT # Read without concurrent inserts
|
|
|
|
TL_WRITE_ALLOW_WRITE # Write lock that allows other writers
|
|
|
|
TL_WRITE_ALLOW_READ # Write lock, but allow reading
|
|
|
|
TL_WRITE_CONCURRENT_INSERT
|
|
|
|
# Insert that can be mixed when selects
|
|
|
|
TL_WRITE_DELAYED # Used by delayed insert
|
|
|
|
# Allows lower locks to take over
|
|
|
|
TL_WRITE_LOW_PRIORITY # Low priority write
|
|
|
|
TL_WRITE # High priority write
|
|
|
|
TL_WRITE_ONLY # High priority write
|
|
|
|
# Abort all new lock request with an error
|
|
|
|
|
|
|
|
Locks are prioritized according to:
|
|
|
|
|
|
|
|
WRITE_ALLOW_WRITE, WRITE_ALLOW_READ, WRITE_CONCURRENT_INSERT, WRITE_DELAYED,
|
|
|
|
WRITE_LOW_PRIORITY, READ, WRITE, READ_HIGH_PRIORITY and WRITE_ONLY
|
|
|
|
|
|
|
|
Locks in the same privilege level are scheduled in first-in-first-out order.
|
|
|
|
|
|
|
|
To allow concurrent read/writes locks, with 'WRITE_CONCURRENT_INSERT' one
|
|
|
|
should put a pointer to the following functions in the lock structure:
|
|
|
|
(If the pointer is zero (default), the function is not called)
|
|
|
|
|
|
|
|
check_status:
|
|
|
|
Before giving a lock of type TL_WRITE_CONCURRENT_INSERT,
|
|
|
|
we check if this function exists and returns 0.
|
|
|
|
If not, then the lock is upgraded to TL_WRITE_LOCK
|
|
|
|
In MyISAM this is a simple check if the insert can be done
|
|
|
|
at the end of the datafile.
|
|
|
|
update_status:
|
|
|
|
Before a write lock is released, this function is called.
|
|
|
|
In MyISAM this functions updates the count and length of the datafile
|
|
|
|
get_status:
|
|
|
|
When one gets a lock this functions is called.
|
|
|
|
In MyISAM this stores the number of rows and size of the datafile
|
|
|
|
for concurrent reads.
|
|
|
|
|
|
|
|
The lock algorithm allows one to have one TL_WRITE_ALLOW_READ,
|
|
|
|
TL_WRITE_CONCURRENT_INSERT or one TL_WRITE_DELAYED lock at the same time as
|
|
|
|
multiple read locks.
|
|
|
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
#if !defined(MAIN) && !defined(DBUG_OFF) && !defined(EXTRA_DEBUG)
|
|
|
|
#define DBUG_OFF
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#include "mysys_priv.h"
|
2000-08-21 23:18:32 +02:00
|
|
|
|
|
|
|
#ifdef THREAD
|
2000-07-31 21:29:14 +02:00
|
|
|
#include "thr_lock.h"
|
|
|
|
#include <m_string.h>
|
|
|
|
#include <errno.h>
|
|
|
|
|
|
|
|
my_bool thr_lock_inited=0;
|
2001-01-27 04:00:42 +01:00
|
|
|
ulong locks_immediate = 0L, locks_waited = 0L;
|
2005-07-19 20:21:12 +02:00
|
|
|
ulong table_lock_wait_timeout;
|
2001-07-18 22:34:04 +02:00
|
|
|
enum thr_lock_type thr_upgraded_concurrent_insert_lock = TL_WRITE;
|
2000-07-31 21:29:14 +02:00
|
|
|
|
|
|
|
/* The following constants are only for debug output */
|
|
|
|
#define MAX_THREADS 100
|
|
|
|
#define MAX_LOCKS 100
|
|
|
|
|
|
|
|
|
2002-10-30 15:52:12 +01:00
|
|
|
LIST *thr_lock_thread_list; /* List of threads in use */
|
2000-07-31 21:29:14 +02:00
|
|
|
ulong max_write_lock_count= ~(ulong) 0L;
|
|
|
|
|
|
|
|
static inline pthread_cond_t *get_cond(void)
|
|
|
|
{
|
|
|
|
return &my_thread_var->suspend;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
** For the future (now the thread specific cond is alloced by my_pthread.c)
|
|
|
|
*/
|
|
|
|
|
|
|
|
my_bool init_thr_lock()
|
|
|
|
{
|
|
|
|
thr_lock_inited=1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-07-19 20:21:12 +02:00
|
|
|
static inline my_bool
|
|
|
|
thr_lock_owner_equal(THR_LOCK_OWNER *rhs, THR_LOCK_OWNER *lhs)
|
|
|
|
{
|
|
|
|
return rhs == lhs;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2000-07-31 21:29:14 +02:00
|
|
|
#ifdef EXTRA_DEBUG
|
2001-07-17 20:04:01 +02:00
|
|
|
#define MAX_FOUND_ERRORS 10 /* Report 10 first errors */
|
|
|
|
static uint found_errors=0;
|
2000-07-31 21:29:14 +02:00
|
|
|
|
|
|
|
static int check_lock(struct st_lock_list *list, const char* lock_type,
|
2005-07-19 20:21:12 +02:00
|
|
|
const char *where, my_bool same_owner, bool no_cond)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
THR_LOCK_DATA *data,**prev;
|
|
|
|
uint count=0;
|
2005-07-19 20:21:12 +02:00
|
|
|
THR_LOCK_OWNER *first_owner;
|
|
|
|
LINT_INIT(first_owner);
|
2000-07-31 21:29:14 +02:00
|
|
|
|
|
|
|
prev= &list->data;
|
|
|
|
if (list->data)
|
|
|
|
{
|
|
|
|
enum thr_lock_type last_lock_type=list->data->type;
|
|
|
|
|
2005-07-19 20:21:12 +02:00
|
|
|
if (same_owner && list->data)
|
|
|
|
first_owner= list->data->owner;
|
2000-07-31 21:29:14 +02:00
|
|
|
for (data=list->data; data && count++ < MAX_LOCKS ; data=data->next)
|
|
|
|
{
|
|
|
|
if (data->type != last_lock_type)
|
|
|
|
last_lock_type=TL_IGNORE;
|
|
|
|
if (data->prev != prev)
|
|
|
|
{
|
|
|
|
fprintf(stderr,
|
|
|
|
"Warning: prev link %d didn't point at previous lock at %s: %s\n",
|
|
|
|
count, lock_type, where);
|
|
|
|
return 1;
|
|
|
|
}
|
2005-07-19 20:21:12 +02:00
|
|
|
if (same_owner &&
|
|
|
|
!thr_lock_owner_equal(data->owner, first_owner) &&
|
2000-07-31 21:29:14 +02:00
|
|
|
last_lock_type != TL_WRITE_ALLOW_WRITE)
|
|
|
|
{
|
|
|
|
fprintf(stderr,
|
|
|
|
"Warning: Found locks from different threads in %s: %s\n",
|
|
|
|
lock_type,where);
|
|
|
|
return 1;
|
|
|
|
}
|
2001-08-31 22:02:09 +02:00
|
|
|
if (no_cond && data->cond)
|
|
|
|
{
|
|
|
|
fprintf(stderr,
|
|
|
|
"Warning: Found active lock with not reset cond %s: %s\n",
|
|
|
|
lock_type,where);
|
|
|
|
return 1;
|
|
|
|
}
|
2000-07-31 21:29:14 +02:00
|
|
|
prev= &data->next;
|
|
|
|
}
|
|
|
|
if (data)
|
|
|
|
{
|
|
|
|
fprintf(stderr,"Warning: found too many locks at %s: %s\n",
|
|
|
|
lock_type,where);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (prev != list->last)
|
|
|
|
{
|
|
|
|
fprintf(stderr,"Warning: last didn't point at last lock at %s: %s\n",
|
|
|
|
lock_type, where);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2003-07-18 03:37:37 +02:00
|
|
|
|
2000-07-31 21:29:14 +02:00
|
|
|
static void check_locks(THR_LOCK *lock, const char *where,
|
|
|
|
my_bool allow_no_locks)
|
|
|
|
{
|
2001-07-17 20:04:01 +02:00
|
|
|
uint old_found_errors=found_errors;
|
2003-07-18 03:37:37 +02:00
|
|
|
DBUG_ENTER("check_locks");
|
|
|
|
|
2001-07-17 20:04:01 +02:00
|
|
|
if (found_errors < MAX_FOUND_ERRORS)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
2001-08-31 22:02:09 +02:00
|
|
|
if (check_lock(&lock->write,"write",where,1,1) |
|
|
|
|
check_lock(&lock->write_wait,"write_wait",where,0,0) |
|
|
|
|
check_lock(&lock->read,"read",where,0,1) |
|
|
|
|
check_lock(&lock->read_wait,"read_wait",where,0,0))
|
2001-07-17 20:04:01 +02:00
|
|
|
found_errors++;
|
2000-07-31 21:29:14 +02:00
|
|
|
|
2001-07-17 20:04:01 +02:00
|
|
|
if (found_errors < MAX_FOUND_ERRORS)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
uint count=0;
|
|
|
|
THR_LOCK_DATA *data;
|
|
|
|
for (data=lock->read.data ; data ; data=data->next)
|
|
|
|
{
|
|
|
|
if ((int) data->type == (int) TL_READ_NO_INSERT)
|
|
|
|
count++;
|
2006-06-26 19:14:35 +02:00
|
|
|
/* Protect against infinite loop. */
|
|
|
|
DBUG_ASSERT(count <= lock->read_no_write_count);
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
if (count != lock->read_no_write_count)
|
|
|
|
{
|
2001-07-17 20:04:01 +02:00
|
|
|
found_errors++;
|
2000-07-31 21:29:14 +02:00
|
|
|
fprintf(stderr,
|
|
|
|
"Warning at '%s': Locks read_no_write_count was %u when it should have been %u\n", where, lock->read_no_write_count,count);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!lock->write.data)
|
|
|
|
{
|
|
|
|
if (!allow_no_locks && !lock->read.data &&
|
|
|
|
(lock->write_wait.data || lock->read_wait.data))
|
|
|
|
{
|
2001-07-17 20:04:01 +02:00
|
|
|
found_errors++;
|
2000-07-31 21:29:14 +02:00
|
|
|
fprintf(stderr,
|
|
|
|
"Warning at '%s': No locks in use but locks are in wait queue\n",
|
|
|
|
where);
|
|
|
|
}
|
|
|
|
if (!lock->write_wait.data)
|
|
|
|
{
|
|
|
|
if (!allow_no_locks && lock->read_wait.data)
|
|
|
|
{
|
2001-07-17 20:04:01 +02:00
|
|
|
found_errors++;
|
2000-07-31 21:29:14 +02:00
|
|
|
fprintf(stderr,
|
|
|
|
"Warning at '%s': No write locks and waiting read locks\n",
|
|
|
|
where);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
if (!allow_no_locks &&
|
|
|
|
(((lock->write_wait.data->type == TL_WRITE_CONCURRENT_INSERT ||
|
|
|
|
lock->write_wait.data->type == TL_WRITE_ALLOW_WRITE) &&
|
|
|
|
!lock->read_no_write_count) ||
|
|
|
|
lock->write_wait.data->type == TL_WRITE_ALLOW_READ ||
|
|
|
|
(lock->write_wait.data->type == TL_WRITE_DELAYED &&
|
|
|
|
!lock->read.data)))
|
|
|
|
{
|
2001-07-17 20:04:01 +02:00
|
|
|
found_errors++;
|
2000-07-31 21:29:14 +02:00
|
|
|
fprintf(stderr,
|
|
|
|
"Warning at '%s': Write lock %d waiting while no exclusive read locks\n",where,(int) lock->write_wait.data->type);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{ /* Have write lock */
|
|
|
|
if (lock->write_wait.data)
|
|
|
|
{
|
|
|
|
if (!allow_no_locks &&
|
|
|
|
lock->write.data->type == TL_WRITE_ALLOW_WRITE &&
|
|
|
|
lock->write_wait.data->type == TL_WRITE_ALLOW_WRITE)
|
|
|
|
{
|
2001-07-17 20:04:01 +02:00
|
|
|
found_errors++;
|
2000-07-31 21:29:14 +02:00
|
|
|
fprintf(stderr,
|
|
|
|
"Warning at '%s': Found WRITE_ALLOW_WRITE lock waiting for WRITE_ALLOW_WRITE lock\n",
|
|
|
|
where);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (lock->read.data)
|
|
|
|
{
|
2005-07-19 20:21:12 +02:00
|
|
|
if (!thr_lock_owner_equal(lock->write.data->owner,
|
|
|
|
lock->read.data->owner) &&
|
2003-07-18 03:37:37 +02:00
|
|
|
((lock->write.data->type > TL_WRITE_DELAYED &&
|
|
|
|
lock->write.data->type != TL_WRITE_ONLY) ||
|
|
|
|
((lock->write.data->type == TL_WRITE_CONCURRENT_INSERT ||
|
|
|
|
lock->write.data->type == TL_WRITE_ALLOW_WRITE) &&
|
|
|
|
lock->read_no_write_count)))
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
2001-07-17 20:04:01 +02:00
|
|
|
found_errors++;
|
2000-07-31 21:29:14 +02:00
|
|
|
fprintf(stderr,
|
2001-07-17 20:04:01 +02:00
|
|
|
"Warning at '%s': Found lock of type %d that is write and read locked\n",
|
|
|
|
where, lock->write.data->type);
|
2003-07-18 03:37:37 +02:00
|
|
|
DBUG_PRINT("warning",("At '%s': Found lock of type %d that is write and read locked\n",
|
|
|
|
where, lock->write.data->type));
|
|
|
|
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
if (lock->read_wait.data)
|
|
|
|
{
|
|
|
|
if (!allow_no_locks && lock->write.data->type <= TL_WRITE_DELAYED &&
|
|
|
|
lock->read_wait.data->type <= TL_READ_HIGH_PRIORITY)
|
|
|
|
{
|
2001-07-17 20:04:01 +02:00
|
|
|
found_errors++;
|
2000-07-31 21:29:14 +02:00
|
|
|
fprintf(stderr,
|
|
|
|
"Warning at '%s': Found read lock of type %d waiting for write lock of type %d\n",
|
|
|
|
where,
|
|
|
|
(int) lock->read_wait.data->type,
|
|
|
|
(int) lock->write.data->type);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2001-07-17 20:04:01 +02:00
|
|
|
if (found_errors != old_found_errors)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
DBUG_PRINT("error",("Found wrong lock"));
|
|
|
|
}
|
|
|
|
}
|
2003-07-18 03:37:37 +02:00
|
|
|
DBUG_VOID_RETURN;
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
#else /* EXTRA_DEBUG */
|
|
|
|
#define check_locks(A,B,C)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
|
|
/* Initialize a lock */
|
|
|
|
|
|
|
|
void thr_lock_init(THR_LOCK *lock)
|
|
|
|
{
|
|
|
|
DBUG_ENTER("thr_lock_init");
|
|
|
|
bzero((char*) lock,sizeof(*lock));
|
2001-03-26 00:05:04 +02:00
|
|
|
VOID(pthread_mutex_init(&lock->mutex,MY_MUTEX_INIT_FAST));
|
2000-07-31 21:29:14 +02:00
|
|
|
lock->read.last= &lock->read.data;
|
|
|
|
lock->read_wait.last= &lock->read_wait.data;
|
|
|
|
lock->write_wait.last= &lock->write_wait.data;
|
|
|
|
lock->write.last= &lock->write.data;
|
|
|
|
|
|
|
|
pthread_mutex_lock(&THR_LOCK_lock); /* Add to locks in use */
|
|
|
|
lock->list.data=(void*) lock;
|
2002-10-30 15:52:12 +01:00
|
|
|
thr_lock_thread_list=list_add(thr_lock_thread_list,&lock->list);
|
2000-07-31 21:29:14 +02:00
|
|
|
pthread_mutex_unlock(&THR_LOCK_lock);
|
|
|
|
DBUG_VOID_RETURN;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
void thr_lock_delete(THR_LOCK *lock)
|
|
|
|
{
|
|
|
|
DBUG_ENTER("thr_lock_delete");
|
|
|
|
VOID(pthread_mutex_destroy(&lock->mutex));
|
|
|
|
pthread_mutex_lock(&THR_LOCK_lock);
|
2002-10-30 15:52:12 +01:00
|
|
|
thr_lock_thread_list=list_delete(thr_lock_thread_list,&lock->list);
|
2000-07-31 21:29:14 +02:00
|
|
|
pthread_mutex_unlock(&THR_LOCK_lock);
|
|
|
|
DBUG_VOID_RETURN;
|
|
|
|
}
|
|
|
|
|
2005-07-19 20:21:12 +02:00
|
|
|
|
|
|
|
void thr_lock_info_init(THR_LOCK_INFO *info)
|
|
|
|
{
|
|
|
|
info->thread= pthread_self();
|
|
|
|
info->thread_id= my_thread_id(); /* for debugging */
|
|
|
|
info->n_cursors= 0;
|
|
|
|
}
|
|
|
|
|
2000-07-31 21:29:14 +02:00
|
|
|
/* Initialize a lock instance */
|
|
|
|
|
|
|
|
void thr_lock_data_init(THR_LOCK *lock,THR_LOCK_DATA *data, void *param)
|
|
|
|
{
|
|
|
|
data->lock=lock;
|
|
|
|
data->type=TL_UNLOCK;
|
2005-07-19 20:21:12 +02:00
|
|
|
data->owner= 0; /* no owner yet */
|
2000-07-31 21:29:14 +02:00
|
|
|
data->status_param=param;
|
2001-08-31 22:02:09 +02:00
|
|
|
data->cond=0;
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2005-07-19 20:21:12 +02:00
|
|
|
static inline my_bool
|
|
|
|
have_old_read_lock(THR_LOCK_DATA *data, THR_LOCK_OWNER *owner)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
for ( ; data ; data=data->next)
|
|
|
|
{
|
2005-07-19 20:21:12 +02:00
|
|
|
if (thr_lock_owner_equal(data->owner, owner))
|
2000-07-31 21:29:14 +02:00
|
|
|
return 1; /* Already locked by thread */
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline my_bool have_specific_lock(THR_LOCK_DATA *data,
|
|
|
|
enum thr_lock_type type)
|
|
|
|
{
|
|
|
|
for ( ; data ; data=data->next)
|
|
|
|
{
|
|
|
|
if (data->type == type)
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2005-07-19 20:21:12 +02:00
|
|
|
static enum enum_thr_lock_result
|
|
|
|
wait_for_lock(struct st_lock_list *wait, THR_LOCK_DATA *data,
|
|
|
|
my_bool in_wait_list)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
2005-07-19 20:21:12 +02:00
|
|
|
struct st_my_thread_var *thread_var= my_thread_var;
|
|
|
|
pthread_cond_t *cond= &thread_var->suspend;
|
|
|
|
struct timespec wait_timeout;
|
|
|
|
enum enum_thr_lock_result result= THR_LOCK_ABORTED;
|
|
|
|
my_bool can_deadlock= test(data->owner->info->n_cursors);
|
2000-07-31 21:29:14 +02:00
|
|
|
|
|
|
|
if (!in_wait_list)
|
|
|
|
{
|
|
|
|
(*wait->last)=data; /* Wait for lock */
|
|
|
|
data->prev= wait->last;
|
|
|
|
wait->last= &data->next;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Set up control struct to allow others to abort locks */
|
|
|
|
thread_var->current_mutex= &data->lock->mutex;
|
|
|
|
thread_var->current_cond= cond;
|
2005-07-19 20:21:12 +02:00
|
|
|
data->cond= cond;
|
2000-07-31 21:29:14 +02:00
|
|
|
|
2005-07-19 20:21:12 +02:00
|
|
|
if (can_deadlock)
|
2005-07-22 21:01:42 +02:00
|
|
|
set_timespec(wait_timeout, table_lock_wait_timeout);
|
2001-09-01 09:38:16 +02:00
|
|
|
while (!thread_var->abort || in_wait_list)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
2005-10-11 23:58:22 +02:00
|
|
|
int rc= (can_deadlock ?
|
|
|
|
pthread_cond_timedwait(cond, &data->lock->mutex,
|
|
|
|
&wait_timeout) :
|
|
|
|
pthread_cond_wait(cond, &data->lock->mutex));
|
2005-07-19 20:21:12 +02:00
|
|
|
/*
|
|
|
|
We must break the wait if one of the following occurs:
|
|
|
|
- the connection has been aborted (!thread_var->abort), but
|
|
|
|
this is not a delayed insert thread (in_wait_list). For a delayed
|
|
|
|
insert thread the proper action at shutdown is, apparently, to
|
|
|
|
acquire the lock and complete the insert.
|
|
|
|
- the lock has been granted (data->cond is set to NULL by the granter),
|
|
|
|
or the waiting has been aborted (additionally data->type is set to
|
|
|
|
TL_UNLOCK).
|
|
|
|
- the wait has timed out (rc == ETIMEDOUT)
|
|
|
|
Order of checks below is important to not report about timeout
|
|
|
|
if the predicate is true.
|
|
|
|
*/
|
|
|
|
if (data->cond == 0)
|
|
|
|
break;
|
2005-10-11 23:58:22 +02:00
|
|
|
if (rc == ETIMEDOUT || rc == ETIME)
|
2005-07-19 20:21:12 +02:00
|
|
|
{
|
|
|
|
result= THR_LOCK_WAIT_TIMEOUT;
|
2001-09-01 09:38:16 +02:00
|
|
|
break;
|
2005-07-19 20:21:12 +02:00
|
|
|
}
|
2001-09-01 09:38:16 +02:00
|
|
|
}
|
2000-07-31 21:29:14 +02:00
|
|
|
|
|
|
|
if (data->cond || data->type == TL_UNLOCK)
|
|
|
|
{
|
2005-07-19 20:21:12 +02:00
|
|
|
if (data->cond) /* aborted or timed out */
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
if (((*data->prev)=data->next)) /* remove from wait-list */
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else
|
|
|
|
wait->last=data->prev;
|
2005-07-19 20:21:12 +02:00
|
|
|
data->type= TL_UNLOCK; /* No lock */
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
check_locks(data->lock,"failed wait_for_lock",0);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2005-07-19 20:21:12 +02:00
|
|
|
result= THR_LOCK_SUCCESS;
|
2001-01-27 23:33:31 +01:00
|
|
|
statistic_increment(locks_waited, &THR_LOCK_lock);
|
2000-07-31 21:29:14 +02:00
|
|
|
if (data->lock->get_status)
|
2005-05-13 11:08:08 +02:00
|
|
|
(*data->lock->get_status)(data->status_param, 0);
|
2000-07-31 21:29:14 +02:00
|
|
|
check_locks(data->lock,"got wait_for_lock",0);
|
|
|
|
}
|
|
|
|
pthread_mutex_unlock(&data->lock->mutex);
|
|
|
|
|
|
|
|
/* The following must be done after unlock of lock->mutex */
|
|
|
|
pthread_mutex_lock(&thread_var->mutex);
|
|
|
|
thread_var->current_mutex= 0;
|
|
|
|
thread_var->current_cond= 0;
|
|
|
|
pthread_mutex_unlock(&thread_var->mutex);
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2005-07-19 20:21:12 +02:00
|
|
|
enum enum_thr_lock_result
|
|
|
|
thr_lock(THR_LOCK_DATA *data, THR_LOCK_OWNER *owner,
|
|
|
|
enum thr_lock_type lock_type)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
THR_LOCK *lock=data->lock;
|
2005-07-19 20:21:12 +02:00
|
|
|
enum enum_thr_lock_result result= THR_LOCK_SUCCESS;
|
|
|
|
struct st_lock_list *wait_queue;
|
|
|
|
THR_LOCK_DATA *lock_owner;
|
2000-07-31 21:29:14 +02:00
|
|
|
DBUG_ENTER("thr_lock");
|
|
|
|
|
|
|
|
data->next=0;
|
2001-08-31 22:02:09 +02:00
|
|
|
data->cond=0; /* safety */
|
2000-07-31 21:29:14 +02:00
|
|
|
data->type=lock_type;
|
2005-07-19 20:21:12 +02:00
|
|
|
data->owner= owner; /* Must be reset ! */
|
2000-07-31 21:29:14 +02:00
|
|
|
VOID(pthread_mutex_lock(&lock->mutex));
|
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes
Changes that requires code changes in other code of other storage engines.
(Note that all changes are very straightforward and one should find all issues
by compiling a --debug build and fixing all compiler errors and all
asserts in field.cc while running the test suite),
- New optional handler function introduced: reset()
This is called after every DML statement to make it easy for a handler to
statement specific cleanups.
(The only case it's not called is if force the file to be closed)
- handler::extra(HA_EXTRA_RESET) is removed. Code that was there before
should be moved to handler::reset()
- table->read_set contains a bitmap over all columns that are needed
in the query. read_row() and similar functions only needs to read these
columns
- table->write_set contains a bitmap over all columns that will be updated
in the query. write_row() and update_row() only needs to update these
columns.
The above bitmaps should now be up to date in all context
(including ALTER TABLE, filesort()).
The handler is informed of any changes to the bitmap after
fix_fields() by calling the virtual function
handler::column_bitmaps_signal(). If the handler does caching of
these bitmaps (instead of using table->read_set, table->write_set),
it should redo the caching in this code. as the signal() may be sent
several times, it's probably best to set a variable in the signal
and redo the caching on read_row() / write_row() if the variable was
set.
- Removed the read_set and write_set bitmap objects from the handler class
- Removed all column bit handling functions from the handler class.
(Now one instead uses the normal bitmap functions in my_bitmap.c instead
of handler dedicated bitmap functions)
- field->query_id is removed. One should instead instead check
table->read_set and table->write_set if a field is used in the query.
- handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and
handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now
instead use table->read_set to check for which columns to retrieve.
- If a handler needs to call Field->val() or Field->store() on columns
that are not used in the query, one should install a temporary
all-columns-used map while doing so. For this, we provide the following
functions:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set);
field->val();
dbug_tmp_restore_column_map(table->read_set, old_map);
and similar for the write map:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set);
field->val();
dbug_tmp_restore_column_map(table->write_set, old_map);
If this is not done, you will sooner or later hit a DBUG_ASSERT
in the field store() / val() functions.
(For not DBUG binaries, the dbug_tmp_restore_column_map() and
dbug_tmp_restore_column_map() are inline dummy functions and should
be optimized away be the compiler).
- If one needs to temporary set the column map for all binaries (and not
just to avoid the DBUG_ASSERT() in the Field::store() / Field::val()
methods) one should use the functions tmp_use_all_columns() and
tmp_restore_column_map() instead of the above dbug_ variants.
- All 'status' fields in the handler base class (like records,
data_file_length etc) are now stored in a 'stats' struct. This makes
it easier to know what status variables are provided by the base
handler. This requires some trivial variable names in the extra()
function.
- New virtual function handler::records(). This is called to optimize
COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true.
(stats.records is not supposed to be an exact value. It's only has to
be 'reasonable enough' for the optimizer to be able to choose a good
optimization path).
- Non virtual handler::init() function added for caching of virtual
constants from engine.
- Removed has_transactions() virtual method. Now one should instead return
HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support
transactions.
- The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument
that is to be used with 'new handler_name()' to allocate the handler
in the right area. The xxxx_create_handler() function is also
responsible for any initialization of the object before returning.
For example, one should change:
static handler *myisam_create_handler(TABLE_SHARE *table)
{
return new ha_myisam(table);
}
->
static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root)
{
return new (mem_root) ha_myisam(table);
}
- New optional virtual function: use_hidden_primary_key().
This is called in case of an update/delete when
(table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined
but we don't have a primary key. This allows the handler to take precisions
in remembering any hidden primary key to able to update/delete any
found row. The default handler marks all columns to be read.
- handler::table_flags() now returns a ulonglong (to allow for more flags).
- New/changed table_flags()
- HA_HAS_RECORDS Set if ::records() is supported
- HA_NO_TRANSACTIONS Set if engine doesn't support transactions
- HA_PRIMARY_KEY_REQUIRED_FOR_DELETE
Set if we should mark all primary key columns for
read when reading rows as part of a DELETE
statement. If there is no primary key,
all columns are marked for read.
- HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some
cases (based on table->read_set)
- HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS
Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION.
- HA_DUPP_POS Renamed to HA_DUPLICATE_POS
- HA_REQUIRES_KEY_COLUMNS_FOR_DELETE
Set this if we should mark ALL key columns for
read when when reading rows as part of a DELETE
statement. In case of an update we will mark
all keys for read for which key part changed
value.
- HA_STATS_RECORDS_IS_EXACT
Set this if stats.records is exact.
(This saves us some extra records() calls
when optimizing COUNT(*))
- Removed table_flags()
- HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if
handler::records() gives an exact count() and
HA_STATS_RECORDS_IS_EXACT if stats.records is exact.
- HA_READ_RND_SAME Removed (no one supported this one)
- Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk()
- Renamed handler::dupp_pos to handler::dup_pos
- Removed not used variable handler::sortkey
Upper level handler changes:
- ha_reset() now does some overall checks and calls ::reset()
- ha_table_flags() added. This is a cached version of table_flags(). The
cache is updated on engine creation time and updated on open.
MySQL level changes (not obvious from the above):
- DBUG_ASSERT() added to check that column usage matches what is set
in the column usage bit maps. (This found a LOT of bugs in current
column marking code).
- In 5.1 before, all used columns was marked in read_set and only updated
columns was marked in write_set. Now we only mark columns for which we
need a value in read_set.
- Column bitmaps are created in open_binary_frm() and open_table_from_share().
(Before this was in table.cc)
- handler::table_flags() calls are replaced with handler::ha_table_flags()
- For calling field->val() you must have the corresponding bit set in
table->read_set. For calling field->store() you must have the
corresponding bit set in table->write_set. (There are asserts in
all store()/val() functions to catch wrong usage)
- thd->set_query_id is renamed to thd->mark_used_columns and instead
of setting this to an integer value, this has now the values:
MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE
Changed also all variables named 'set_query_id' to mark_used_columns.
- In filesort() we now inform the handler of exactly which columns are needed
doing the sort and choosing the rows.
- The TABLE_SHARE object has a 'all_set' column bitmap one can use
when one needs a column bitmap with all columns set.
(This is used for table->use_all_columns() and other places)
- The TABLE object has 3 column bitmaps:
- def_read_set Default bitmap for columns to be read
- def_write_set Default bitmap for columns to be written
- tmp_set Can be used as a temporary bitmap when needed.
The table object has also two pointer to bitmaps read_set and write_set
that the handler should use to find out which columns are used in which way.
- count() optimization now calls handler::records() instead of using
handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true).
- Added extra argument to Item::walk() to indicate if we should also
traverse sub queries.
- Added TABLE parameter to cp_buffer_from_ref()
- Don't close tables created with CREATE ... SELECT but keep them in
the table cache. (Faster usage of newly created tables).
New interfaces:
- table->clear_column_bitmaps() to initialize the bitmaps for tables
at start of new statements.
- table->column_bitmaps_set() to set up new column bitmaps and signal
the handler about this.
- table->column_bitmaps_set_no_signal() for some few cases where we need
to setup new column bitmaps but don't signal the handler (as the handler
has already been signaled about these before). Used for the momement
only in opt_range.cc when doing ROR scans.
- table->use_all_columns() to install a bitmap where all columns are marked
as use in the read and the write set.
- table->default_column_bitmaps() to install the normal read and write
column bitmaps, but not signaling the handler about this.
This is mainly used when creating TABLE instances.
- table->mark_columns_needed_for_delete(),
table->mark_columns_needed_for_delete() and
table->mark_columns_needed_for_insert() to allow us to put additional
columns in column usage maps if handler so requires.
(The handler indicates what it neads in handler->table_flags())
- table->prepare_for_position() to allow us to tell handler that it
needs to read primary key parts to be able to store them in
future table->position() calls.
(This replaces the table->file->ha_retrieve_all_pk function)
- table->mark_auto_increment_column() to tell handler are going to update
columns part of any auto_increment key.
- table->mark_columns_used_by_index() to mark all columns that is part of
an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow
it to quickly know that it only needs to read colums that are part
of the key. (The handler can also use the column map for detecting this,
but simpler/faster handler can just monitor the extra() call).
- table->mark_columns_used_by_index_no_reset() to in addition to other columns,
also mark all columns that is used by the given key.
- table->restore_column_maps_after_mark_index() to restore to default
column maps after a call to table->mark_columns_used_by_index().
- New item function register_field_in_read_map(), for marking used columns
in table->read_map. Used by filesort() to mark all used columns
- Maintain in TABLE->merge_keys set of all keys that are used in query.
(Simplices some optimization loops)
- Maintain Field->part_of_key_not_clustered which is like Field->part_of_key
but the field in the clustered key is not assumed to be part of all index.
(used in opt_range.cc for faster loops)
- dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map()
tmp_use_all_columns() and tmp_restore_column_map() functions to temporally
mark all columns as usable. The 'dbug_' version is primarily intended
inside a handler when it wants to just call Field:store() & Field::val()
functions, but don't need the column maps set for any other usage.
(ie:: bitmap_is_set() is never called)
- We can't use compare_records() to skip updates for handlers that returns
a partial column set and the read_set doesn't cover all columns in the
write set. The reason for this is that if we have a column marked only for
write we can't in the MySQL level know if the value changed or not.
The reason this worked before was that MySQL marked all to be written
columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden
bug'.
- open_table_from_share() does not anymore setup temporary MEM_ROOT
object as a thread specific variable for the handler. Instead we
send the to-be-used MEMROOT to get_new_handler().
(Simpler, faster code)
Bugs fixed:
- Column marking was not done correctly in a lot of cases.
(ALTER TABLE, when using triggers, auto_increment fields etc)
(Could potentially result in wrong values inserted in table handlers
relying on that the old column maps or field->set_query_id was correct)
Especially when it comes to triggers, there may be cases where the
old code would cause lost/wrong values for NDB and/or InnoDB tables.
- Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags:
OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG.
This allowed me to remove some wrong warnings about:
"Some non-transactional changed tables couldn't be rolled back"
- Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset
(thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose
some warnings about
"Some non-transactional changed tables couldn't be rolled back")
- Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table()
which could cause delete_table to report random failures.
- Fixed core dumps for some tests when running with --debug
- Added missing FN_LIBCHAR in mysql_rm_tmp_tables()
(This has probably caused us to not properly remove temporary files after
crash)
- slow_logs was not properly initialized, which could maybe cause
extra/lost entries in slow log.
- If we get an duplicate row on insert, change column map to read and
write all columns while retrying the operation. This is required by
the definition of REPLACE and also ensures that fields that are only
part of UPDATE are properly handled. This fixed a bug in NDB and
REPLACE where REPLACE wrongly copied some column values from the replaced
row.
- For table handler that doesn't support NULL in keys, we would give an error
when creating a primary key with NULL fields, even after the fields has been
automaticly converted to NOT NULL.
- Creating a primary key on a SPATIAL key, would fail if field was not
declared as NOT NULL.
Cleanups:
- Removed not used condition argument to setup_tables
- Removed not needed item function reset_query_id_processor().
- Field->add_index is removed. Now this is instead maintained in
(field->flags & FIELD_IN_ADD_INDEX)
- Field->fieldnr is removed (use field->field_index instead)
- New argument to filesort() to indicate that it should return a set of
row pointers (not used columns). This allowed me to remove some references
to sql_command in filesort and should also enable us to return column
results in some cases where we couldn't before.
- Changed column bitmap handling in opt_range.cc to be aligned with TABLE
bitmap, which allowed me to use bitmap functions instead of looping over
all fields to create some needed bitmaps. (Faster and smaller code)
- Broke up found too long lines
- Moved some variable declaration at start of function for better code
readability.
- Removed some not used arguments from functions.
(setup_fields(), mysql_prepare_insert_check_table())
- setup_fields() now takes an enum instead of an int for marking columns
usage.
- For internal temporary tables, use handler::write_row(),
handler::delete_row() and handler::update_row() instead of
handler::ha_xxxx() for faster execution.
- Changed some constants to enum's and define's.
- Using separate column read and write sets allows for easier checking
of timestamp field was set by statement.
- Remove calls to free_io_cache() as this is now done automaticly in ha_reset()
- Don't build table->normalized_path as this is now identical to table->path
(after bar's fixes to convert filenames)
- Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to
do comparision with the 'convert-dbug-for-diff' tool.
Things left to do in 5.1:
- We wrongly log failed CREATE TABLE ... SELECT in some cases when using
row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result)
Mats has promised to look into this.
- Test that my fix for CREATE TABLE ... SELECT is indeed correct.
(I added several test cases for this, but in this case it's better that
someone else also tests this throughly).
Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
|
|
|
DBUG_PRINT("lock",("data: 0x%lx thread: 0x%lx lock: 0x%lx type: %d",
|
2005-07-19 20:21:12 +02:00
|
|
|
data, data->owner->info->thread_id,
|
|
|
|
lock, (int) lock_type));
|
2000-07-31 21:29:14 +02:00
|
|
|
check_locks(lock,(uint) lock_type <= (uint) TL_READ_NO_INSERT ?
|
|
|
|
"enter read_lock" : "enter write_lock",0);
|
|
|
|
if ((int) lock_type <= (int) TL_READ_NO_INSERT)
|
|
|
|
{
|
|
|
|
/* Request for READ lock */
|
|
|
|
if (lock->write.data)
|
|
|
|
{
|
2000-11-17 01:36:46 +01:00
|
|
|
/* We can allow a read lock even if there is already a write lock
|
|
|
|
on the table in one the following cases:
|
2000-07-31 21:29:14 +02:00
|
|
|
- This thread alread have a write lock on the table
|
|
|
|
- The write lock is TL_WRITE_ALLOW_READ or TL_WRITE_DELAYED
|
|
|
|
and the read lock is TL_READ_HIGH_PRIORITY or TL_READ
|
|
|
|
- The write lock is TL_WRITE_CONCURRENT_INSERT or TL_WRITE_ALLOW_WRITE
|
|
|
|
and the read lock is not TL_READ_NO_INSERT
|
|
|
|
*/
|
|
|
|
|
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes
Changes that requires code changes in other code of other storage engines.
(Note that all changes are very straightforward and one should find all issues
by compiling a --debug build and fixing all compiler errors and all
asserts in field.cc while running the test suite),
- New optional handler function introduced: reset()
This is called after every DML statement to make it easy for a handler to
statement specific cleanups.
(The only case it's not called is if force the file to be closed)
- handler::extra(HA_EXTRA_RESET) is removed. Code that was there before
should be moved to handler::reset()
- table->read_set contains a bitmap over all columns that are needed
in the query. read_row() and similar functions only needs to read these
columns
- table->write_set contains a bitmap over all columns that will be updated
in the query. write_row() and update_row() only needs to update these
columns.
The above bitmaps should now be up to date in all context
(including ALTER TABLE, filesort()).
The handler is informed of any changes to the bitmap after
fix_fields() by calling the virtual function
handler::column_bitmaps_signal(). If the handler does caching of
these bitmaps (instead of using table->read_set, table->write_set),
it should redo the caching in this code. as the signal() may be sent
several times, it's probably best to set a variable in the signal
and redo the caching on read_row() / write_row() if the variable was
set.
- Removed the read_set and write_set bitmap objects from the handler class
- Removed all column bit handling functions from the handler class.
(Now one instead uses the normal bitmap functions in my_bitmap.c instead
of handler dedicated bitmap functions)
- field->query_id is removed. One should instead instead check
table->read_set and table->write_set if a field is used in the query.
- handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and
handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now
instead use table->read_set to check for which columns to retrieve.
- If a handler needs to call Field->val() or Field->store() on columns
that are not used in the query, one should install a temporary
all-columns-used map while doing so. For this, we provide the following
functions:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set);
field->val();
dbug_tmp_restore_column_map(table->read_set, old_map);
and similar for the write map:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set);
field->val();
dbug_tmp_restore_column_map(table->write_set, old_map);
If this is not done, you will sooner or later hit a DBUG_ASSERT
in the field store() / val() functions.
(For not DBUG binaries, the dbug_tmp_restore_column_map() and
dbug_tmp_restore_column_map() are inline dummy functions and should
be optimized away be the compiler).
- If one needs to temporary set the column map for all binaries (and not
just to avoid the DBUG_ASSERT() in the Field::store() / Field::val()
methods) one should use the functions tmp_use_all_columns() and
tmp_restore_column_map() instead of the above dbug_ variants.
- All 'status' fields in the handler base class (like records,
data_file_length etc) are now stored in a 'stats' struct. This makes
it easier to know what status variables are provided by the base
handler. This requires some trivial variable names in the extra()
function.
- New virtual function handler::records(). This is called to optimize
COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true.
(stats.records is not supposed to be an exact value. It's only has to
be 'reasonable enough' for the optimizer to be able to choose a good
optimization path).
- Non virtual handler::init() function added for caching of virtual
constants from engine.
- Removed has_transactions() virtual method. Now one should instead return
HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support
transactions.
- The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument
that is to be used with 'new handler_name()' to allocate the handler
in the right area. The xxxx_create_handler() function is also
responsible for any initialization of the object before returning.
For example, one should change:
static handler *myisam_create_handler(TABLE_SHARE *table)
{
return new ha_myisam(table);
}
->
static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root)
{
return new (mem_root) ha_myisam(table);
}
- New optional virtual function: use_hidden_primary_key().
This is called in case of an update/delete when
(table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined
but we don't have a primary key. This allows the handler to take precisions
in remembering any hidden primary key to able to update/delete any
found row. The default handler marks all columns to be read.
- handler::table_flags() now returns a ulonglong (to allow for more flags).
- New/changed table_flags()
- HA_HAS_RECORDS Set if ::records() is supported
- HA_NO_TRANSACTIONS Set if engine doesn't support transactions
- HA_PRIMARY_KEY_REQUIRED_FOR_DELETE
Set if we should mark all primary key columns for
read when reading rows as part of a DELETE
statement. If there is no primary key,
all columns are marked for read.
- HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some
cases (based on table->read_set)
- HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS
Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION.
- HA_DUPP_POS Renamed to HA_DUPLICATE_POS
- HA_REQUIRES_KEY_COLUMNS_FOR_DELETE
Set this if we should mark ALL key columns for
read when when reading rows as part of a DELETE
statement. In case of an update we will mark
all keys for read for which key part changed
value.
- HA_STATS_RECORDS_IS_EXACT
Set this if stats.records is exact.
(This saves us some extra records() calls
when optimizing COUNT(*))
- Removed table_flags()
- HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if
handler::records() gives an exact count() and
HA_STATS_RECORDS_IS_EXACT if stats.records is exact.
- HA_READ_RND_SAME Removed (no one supported this one)
- Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk()
- Renamed handler::dupp_pos to handler::dup_pos
- Removed not used variable handler::sortkey
Upper level handler changes:
- ha_reset() now does some overall checks and calls ::reset()
- ha_table_flags() added. This is a cached version of table_flags(). The
cache is updated on engine creation time and updated on open.
MySQL level changes (not obvious from the above):
- DBUG_ASSERT() added to check that column usage matches what is set
in the column usage bit maps. (This found a LOT of bugs in current
column marking code).
- In 5.1 before, all used columns was marked in read_set and only updated
columns was marked in write_set. Now we only mark columns for which we
need a value in read_set.
- Column bitmaps are created in open_binary_frm() and open_table_from_share().
(Before this was in table.cc)
- handler::table_flags() calls are replaced with handler::ha_table_flags()
- For calling field->val() you must have the corresponding bit set in
table->read_set. For calling field->store() you must have the
corresponding bit set in table->write_set. (There are asserts in
all store()/val() functions to catch wrong usage)
- thd->set_query_id is renamed to thd->mark_used_columns and instead
of setting this to an integer value, this has now the values:
MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE
Changed also all variables named 'set_query_id' to mark_used_columns.
- In filesort() we now inform the handler of exactly which columns are needed
doing the sort and choosing the rows.
- The TABLE_SHARE object has a 'all_set' column bitmap one can use
when one needs a column bitmap with all columns set.
(This is used for table->use_all_columns() and other places)
- The TABLE object has 3 column bitmaps:
- def_read_set Default bitmap for columns to be read
- def_write_set Default bitmap for columns to be written
- tmp_set Can be used as a temporary bitmap when needed.
The table object has also two pointer to bitmaps read_set and write_set
that the handler should use to find out which columns are used in which way.
- count() optimization now calls handler::records() instead of using
handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true).
- Added extra argument to Item::walk() to indicate if we should also
traverse sub queries.
- Added TABLE parameter to cp_buffer_from_ref()
- Don't close tables created with CREATE ... SELECT but keep them in
the table cache. (Faster usage of newly created tables).
New interfaces:
- table->clear_column_bitmaps() to initialize the bitmaps for tables
at start of new statements.
- table->column_bitmaps_set() to set up new column bitmaps and signal
the handler about this.
- table->column_bitmaps_set_no_signal() for some few cases where we need
to setup new column bitmaps but don't signal the handler (as the handler
has already been signaled about these before). Used for the momement
only in opt_range.cc when doing ROR scans.
- table->use_all_columns() to install a bitmap where all columns are marked
as use in the read and the write set.
- table->default_column_bitmaps() to install the normal read and write
column bitmaps, but not signaling the handler about this.
This is mainly used when creating TABLE instances.
- table->mark_columns_needed_for_delete(),
table->mark_columns_needed_for_delete() and
table->mark_columns_needed_for_insert() to allow us to put additional
columns in column usage maps if handler so requires.
(The handler indicates what it neads in handler->table_flags())
- table->prepare_for_position() to allow us to tell handler that it
needs to read primary key parts to be able to store them in
future table->position() calls.
(This replaces the table->file->ha_retrieve_all_pk function)
- table->mark_auto_increment_column() to tell handler are going to update
columns part of any auto_increment key.
- table->mark_columns_used_by_index() to mark all columns that is part of
an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow
it to quickly know that it only needs to read colums that are part
of the key. (The handler can also use the column map for detecting this,
but simpler/faster handler can just monitor the extra() call).
- table->mark_columns_used_by_index_no_reset() to in addition to other columns,
also mark all columns that is used by the given key.
- table->restore_column_maps_after_mark_index() to restore to default
column maps after a call to table->mark_columns_used_by_index().
- New item function register_field_in_read_map(), for marking used columns
in table->read_map. Used by filesort() to mark all used columns
- Maintain in TABLE->merge_keys set of all keys that are used in query.
(Simplices some optimization loops)
- Maintain Field->part_of_key_not_clustered which is like Field->part_of_key
but the field in the clustered key is not assumed to be part of all index.
(used in opt_range.cc for faster loops)
- dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map()
tmp_use_all_columns() and tmp_restore_column_map() functions to temporally
mark all columns as usable. The 'dbug_' version is primarily intended
inside a handler when it wants to just call Field:store() & Field::val()
functions, but don't need the column maps set for any other usage.
(ie:: bitmap_is_set() is never called)
- We can't use compare_records() to skip updates for handlers that returns
a partial column set and the read_set doesn't cover all columns in the
write set. The reason for this is that if we have a column marked only for
write we can't in the MySQL level know if the value changed or not.
The reason this worked before was that MySQL marked all to be written
columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden
bug'.
- open_table_from_share() does not anymore setup temporary MEM_ROOT
object as a thread specific variable for the handler. Instead we
send the to-be-used MEMROOT to get_new_handler().
(Simpler, faster code)
Bugs fixed:
- Column marking was not done correctly in a lot of cases.
(ALTER TABLE, when using triggers, auto_increment fields etc)
(Could potentially result in wrong values inserted in table handlers
relying on that the old column maps or field->set_query_id was correct)
Especially when it comes to triggers, there may be cases where the
old code would cause lost/wrong values for NDB and/or InnoDB tables.
- Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags:
OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG.
This allowed me to remove some wrong warnings about:
"Some non-transactional changed tables couldn't be rolled back"
- Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset
(thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose
some warnings about
"Some non-transactional changed tables couldn't be rolled back")
- Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table()
which could cause delete_table to report random failures.
- Fixed core dumps for some tests when running with --debug
- Added missing FN_LIBCHAR in mysql_rm_tmp_tables()
(This has probably caused us to not properly remove temporary files after
crash)
- slow_logs was not properly initialized, which could maybe cause
extra/lost entries in slow log.
- If we get an duplicate row on insert, change column map to read and
write all columns while retrying the operation. This is required by
the definition of REPLACE and also ensures that fields that are only
part of UPDATE are properly handled. This fixed a bug in NDB and
REPLACE where REPLACE wrongly copied some column values from the replaced
row.
- For table handler that doesn't support NULL in keys, we would give an error
when creating a primary key with NULL fields, even after the fields has been
automaticly converted to NOT NULL.
- Creating a primary key on a SPATIAL key, would fail if field was not
declared as NOT NULL.
Cleanups:
- Removed not used condition argument to setup_tables
- Removed not needed item function reset_query_id_processor().
- Field->add_index is removed. Now this is instead maintained in
(field->flags & FIELD_IN_ADD_INDEX)
- Field->fieldnr is removed (use field->field_index instead)
- New argument to filesort() to indicate that it should return a set of
row pointers (not used columns). This allowed me to remove some references
to sql_command in filesort and should also enable us to return column
results in some cases where we couldn't before.
- Changed column bitmap handling in opt_range.cc to be aligned with TABLE
bitmap, which allowed me to use bitmap functions instead of looping over
all fields to create some needed bitmaps. (Faster and smaller code)
- Broke up found too long lines
- Moved some variable declaration at start of function for better code
readability.
- Removed some not used arguments from functions.
(setup_fields(), mysql_prepare_insert_check_table())
- setup_fields() now takes an enum instead of an int for marking columns
usage.
- For internal temporary tables, use handler::write_row(),
handler::delete_row() and handler::update_row() instead of
handler::ha_xxxx() for faster execution.
- Changed some constants to enum's and define's.
- Using separate column read and write sets allows for easier checking
of timestamp field was set by statement.
- Remove calls to free_io_cache() as this is now done automaticly in ha_reset()
- Don't build table->normalized_path as this is now identical to table->path
(after bar's fixes to convert filenames)
- Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to
do comparision with the 'convert-dbug-for-diff' tool.
Things left to do in 5.1:
- We wrongly log failed CREATE TABLE ... SELECT in some cases when using
row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result)
Mats has promised to look into this.
- Test that my fix for CREATE TABLE ... SELECT is indeed correct.
(I added several test cases for this, but in this case it's better that
someone else also tests this throughly).
Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
|
|
|
DBUG_PRINT("lock",("write locked by thread: 0x%lx",
|
2005-07-19 20:21:12 +02:00
|
|
|
lock->write.data->owner->info->thread_id));
|
|
|
|
if (thr_lock_owner_equal(data->owner, lock->write.data->owner) ||
|
2000-07-31 21:29:14 +02:00
|
|
|
(lock->write.data->type <= TL_WRITE_DELAYED &&
|
|
|
|
(((int) lock_type <= (int) TL_READ_HIGH_PRIORITY) ||
|
|
|
|
(lock->write.data->type != TL_WRITE_CONCURRENT_INSERT &&
|
|
|
|
lock->write.data->type != TL_WRITE_ALLOW_READ))))
|
|
|
|
{ /* Already got a write lock */
|
|
|
|
(*lock->read.last)=data; /* Add to running FIFO */
|
|
|
|
data->prev=lock->read.last;
|
|
|
|
lock->read.last= &data->next;
|
2005-07-31 11:49:55 +02:00
|
|
|
if (lock_type == TL_READ_NO_INSERT)
|
2000-07-31 21:29:14 +02:00
|
|
|
lock->read_no_write_count++;
|
|
|
|
check_locks(lock,"read lock with old write lock",0);
|
|
|
|
if (lock->get_status)
|
2005-05-13 11:08:08 +02:00
|
|
|
(*lock->get_status)(data->status_param, 0);
|
2001-05-23 01:30:17 +02:00
|
|
|
statistic_increment(locks_immediate,&THR_LOCK_lock);
|
2000-07-31 21:29:14 +02:00
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
if (lock->write.data->type == TL_WRITE_ONLY)
|
|
|
|
{
|
|
|
|
/* We are not allowed to get a READ lock in this case */
|
|
|
|
data->type=TL_UNLOCK;
|
2005-07-19 20:21:12 +02:00
|
|
|
result= THR_LOCK_ABORTED; /* Can't wait for this one */
|
2000-07-31 21:29:14 +02:00
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else if (!lock->write_wait.data ||
|
|
|
|
lock->write_wait.data->type <= TL_WRITE_LOW_PRIORITY ||
|
|
|
|
lock_type == TL_READ_HIGH_PRIORITY ||
|
2005-07-19 20:21:12 +02:00
|
|
|
have_old_read_lock(lock->read.data, data->owner))
|
2000-07-31 21:29:14 +02:00
|
|
|
{ /* No important write-locks */
|
|
|
|
(*lock->read.last)=data; /* Add to running FIFO */
|
|
|
|
data->prev=lock->read.last;
|
|
|
|
lock->read.last= &data->next;
|
|
|
|
if (lock->get_status)
|
2005-05-13 11:08:08 +02:00
|
|
|
(*lock->get_status)(data->status_param, 0);
|
2005-07-31 11:49:55 +02:00
|
|
|
if (lock_type == TL_READ_NO_INSERT)
|
2000-07-31 21:29:14 +02:00
|
|
|
lock->read_no_write_count++;
|
|
|
|
check_locks(lock,"read lock with no write locks",0);
|
2001-05-23 01:30:17 +02:00
|
|
|
statistic_increment(locks_immediate,&THR_LOCK_lock);
|
2000-07-31 21:29:14 +02:00
|
|
|
goto end;
|
|
|
|
}
|
2005-07-19 20:21:12 +02:00
|
|
|
/*
|
2005-07-31 11:49:55 +02:00
|
|
|
We're here if there is an active write lock or no write
|
2005-07-19 20:21:12 +02:00
|
|
|
lock but a high priority write waiting in the write_wait queue.
|
|
|
|
In the latter case we should yield the lock to the writer.
|
|
|
|
*/
|
|
|
|
wait_queue= &lock->read_wait;
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
else /* Request for WRITE lock */
|
|
|
|
{
|
|
|
|
if (lock_type == TL_WRITE_DELAYED)
|
|
|
|
{
|
|
|
|
if (lock->write.data && lock->write.data->type == TL_WRITE_ONLY)
|
|
|
|
{
|
|
|
|
data->type=TL_UNLOCK;
|
2005-07-19 20:21:12 +02:00
|
|
|
result= THR_LOCK_ABORTED; /* Can't wait for this one */
|
2000-07-31 21:29:14 +02:00
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
if there is a TL_WRITE_ALLOW_READ lock, we have to wait for a lock
|
|
|
|
(TL_WRITE_ALLOW_READ is used for ALTER TABLE in MySQL)
|
|
|
|
*/
|
|
|
|
if ((!lock->write.data ||
|
|
|
|
lock->write.data->type != TL_WRITE_ALLOW_READ) &&
|
|
|
|
!have_specific_lock(lock->write_wait.data,TL_WRITE_ALLOW_READ) &&
|
|
|
|
(lock->write.data || lock->read.data))
|
|
|
|
{
|
|
|
|
/* Add delayed write lock to write_wait queue, and return at once */
|
|
|
|
(*lock->write_wait.last)=data;
|
|
|
|
data->prev=lock->write_wait.last;
|
|
|
|
lock->write_wait.last= &data->next;
|
|
|
|
data->cond=get_cond();
|
2005-01-15 11:28:38 +01:00
|
|
|
/*
|
|
|
|
We don't have to do get_status here as we will do it when we change
|
|
|
|
the delayed lock to a real write lock
|
|
|
|
*/
|
2001-05-23 01:30:17 +02:00
|
|
|
statistic_increment(locks_immediate,&THR_LOCK_lock);
|
2000-07-31 21:29:14 +02:00
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else if (lock_type == TL_WRITE_CONCURRENT_INSERT && ! lock->check_status)
|
2001-07-18 22:34:04 +02:00
|
|
|
data->type=lock_type= thr_upgraded_concurrent_insert_lock;
|
2000-07-31 21:29:14 +02:00
|
|
|
|
|
|
|
if (lock->write.data) /* If there is a write lock */
|
|
|
|
{
|
|
|
|
if (lock->write.data->type == TL_WRITE_ONLY)
|
|
|
|
{
|
|
|
|
/* We are not allowed to get a lock in this case */
|
|
|
|
data->type=TL_UNLOCK;
|
2005-07-19 20:21:12 +02:00
|
|
|
result= THR_LOCK_ABORTED; /* Can't wait for this one */
|
2000-07-31 21:29:14 +02:00
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
The following test will not work if the old lock was a
|
|
|
|
TL_WRITE_ALLOW_WRITE, TL_WRITE_ALLOW_READ or TL_WRITE_DELAYED in
|
|
|
|
the same thread, but this will never happen within MySQL.
|
|
|
|
*/
|
2005-07-19 20:21:12 +02:00
|
|
|
if (thr_lock_owner_equal(data->owner, lock->write.data->owner) ||
|
2000-07-31 21:29:14 +02:00
|
|
|
(lock_type == TL_WRITE_ALLOW_WRITE &&
|
|
|
|
!lock->write_wait.data &&
|
|
|
|
lock->write.data->type == TL_WRITE_ALLOW_WRITE))
|
|
|
|
{
|
2004-10-20 15:04:28 +02:00
|
|
|
/*
|
|
|
|
We have already got a write lock or all locks are
|
|
|
|
TL_WRITE_ALLOW_WRITE
|
|
|
|
*/
|
|
|
|
DBUG_PRINT("info", ("write_wait.data: 0x%lx old_type: %d",
|
|
|
|
(ulong) lock->write_wait.data,
|
|
|
|
lock->write.data->type));
|
|
|
|
|
2000-07-31 21:29:14 +02:00
|
|
|
(*lock->write.last)=data; /* Add to running fifo */
|
|
|
|
data->prev=lock->write.last;
|
|
|
|
lock->write.last= &data->next;
|
|
|
|
check_locks(lock,"second write lock",0);
|
|
|
|
if (data->lock->get_status)
|
2005-05-13 11:08:08 +02:00
|
|
|
(*data->lock->get_status)(data->status_param, 0);
|
2001-05-23 01:30:17 +02:00
|
|
|
statistic_increment(locks_immediate,&THR_LOCK_lock);
|
2000-07-31 21:29:14 +02:00
|
|
|
goto end;
|
|
|
|
}
|
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes
Changes that requires code changes in other code of other storage engines.
(Note that all changes are very straightforward and one should find all issues
by compiling a --debug build and fixing all compiler errors and all
asserts in field.cc while running the test suite),
- New optional handler function introduced: reset()
This is called after every DML statement to make it easy for a handler to
statement specific cleanups.
(The only case it's not called is if force the file to be closed)
- handler::extra(HA_EXTRA_RESET) is removed. Code that was there before
should be moved to handler::reset()
- table->read_set contains a bitmap over all columns that are needed
in the query. read_row() and similar functions only needs to read these
columns
- table->write_set contains a bitmap over all columns that will be updated
in the query. write_row() and update_row() only needs to update these
columns.
The above bitmaps should now be up to date in all context
(including ALTER TABLE, filesort()).
The handler is informed of any changes to the bitmap after
fix_fields() by calling the virtual function
handler::column_bitmaps_signal(). If the handler does caching of
these bitmaps (instead of using table->read_set, table->write_set),
it should redo the caching in this code. as the signal() may be sent
several times, it's probably best to set a variable in the signal
and redo the caching on read_row() / write_row() if the variable was
set.
- Removed the read_set and write_set bitmap objects from the handler class
- Removed all column bit handling functions from the handler class.
(Now one instead uses the normal bitmap functions in my_bitmap.c instead
of handler dedicated bitmap functions)
- field->query_id is removed. One should instead instead check
table->read_set and table->write_set if a field is used in the query.
- handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and
handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now
instead use table->read_set to check for which columns to retrieve.
- If a handler needs to call Field->val() or Field->store() on columns
that are not used in the query, one should install a temporary
all-columns-used map while doing so. For this, we provide the following
functions:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set);
field->val();
dbug_tmp_restore_column_map(table->read_set, old_map);
and similar for the write map:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set);
field->val();
dbug_tmp_restore_column_map(table->write_set, old_map);
If this is not done, you will sooner or later hit a DBUG_ASSERT
in the field store() / val() functions.
(For not DBUG binaries, the dbug_tmp_restore_column_map() and
dbug_tmp_restore_column_map() are inline dummy functions and should
be optimized away be the compiler).
- If one needs to temporary set the column map for all binaries (and not
just to avoid the DBUG_ASSERT() in the Field::store() / Field::val()
methods) one should use the functions tmp_use_all_columns() and
tmp_restore_column_map() instead of the above dbug_ variants.
- All 'status' fields in the handler base class (like records,
data_file_length etc) are now stored in a 'stats' struct. This makes
it easier to know what status variables are provided by the base
handler. This requires some trivial variable names in the extra()
function.
- New virtual function handler::records(). This is called to optimize
COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true.
(stats.records is not supposed to be an exact value. It's only has to
be 'reasonable enough' for the optimizer to be able to choose a good
optimization path).
- Non virtual handler::init() function added for caching of virtual
constants from engine.
- Removed has_transactions() virtual method. Now one should instead return
HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support
transactions.
- The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument
that is to be used with 'new handler_name()' to allocate the handler
in the right area. The xxxx_create_handler() function is also
responsible for any initialization of the object before returning.
For example, one should change:
static handler *myisam_create_handler(TABLE_SHARE *table)
{
return new ha_myisam(table);
}
->
static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root)
{
return new (mem_root) ha_myisam(table);
}
- New optional virtual function: use_hidden_primary_key().
This is called in case of an update/delete when
(table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined
but we don't have a primary key. This allows the handler to take precisions
in remembering any hidden primary key to able to update/delete any
found row. The default handler marks all columns to be read.
- handler::table_flags() now returns a ulonglong (to allow for more flags).
- New/changed table_flags()
- HA_HAS_RECORDS Set if ::records() is supported
- HA_NO_TRANSACTIONS Set if engine doesn't support transactions
- HA_PRIMARY_KEY_REQUIRED_FOR_DELETE
Set if we should mark all primary key columns for
read when reading rows as part of a DELETE
statement. If there is no primary key,
all columns are marked for read.
- HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some
cases (based on table->read_set)
- HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS
Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION.
- HA_DUPP_POS Renamed to HA_DUPLICATE_POS
- HA_REQUIRES_KEY_COLUMNS_FOR_DELETE
Set this if we should mark ALL key columns for
read when when reading rows as part of a DELETE
statement. In case of an update we will mark
all keys for read for which key part changed
value.
- HA_STATS_RECORDS_IS_EXACT
Set this if stats.records is exact.
(This saves us some extra records() calls
when optimizing COUNT(*))
- Removed table_flags()
- HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if
handler::records() gives an exact count() and
HA_STATS_RECORDS_IS_EXACT if stats.records is exact.
- HA_READ_RND_SAME Removed (no one supported this one)
- Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk()
- Renamed handler::dupp_pos to handler::dup_pos
- Removed not used variable handler::sortkey
Upper level handler changes:
- ha_reset() now does some overall checks and calls ::reset()
- ha_table_flags() added. This is a cached version of table_flags(). The
cache is updated on engine creation time and updated on open.
MySQL level changes (not obvious from the above):
- DBUG_ASSERT() added to check that column usage matches what is set
in the column usage bit maps. (This found a LOT of bugs in current
column marking code).
- In 5.1 before, all used columns was marked in read_set and only updated
columns was marked in write_set. Now we only mark columns for which we
need a value in read_set.
- Column bitmaps are created in open_binary_frm() and open_table_from_share().
(Before this was in table.cc)
- handler::table_flags() calls are replaced with handler::ha_table_flags()
- For calling field->val() you must have the corresponding bit set in
table->read_set. For calling field->store() you must have the
corresponding bit set in table->write_set. (There are asserts in
all store()/val() functions to catch wrong usage)
- thd->set_query_id is renamed to thd->mark_used_columns and instead
of setting this to an integer value, this has now the values:
MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE
Changed also all variables named 'set_query_id' to mark_used_columns.
- In filesort() we now inform the handler of exactly which columns are needed
doing the sort and choosing the rows.
- The TABLE_SHARE object has a 'all_set' column bitmap one can use
when one needs a column bitmap with all columns set.
(This is used for table->use_all_columns() and other places)
- The TABLE object has 3 column bitmaps:
- def_read_set Default bitmap for columns to be read
- def_write_set Default bitmap for columns to be written
- tmp_set Can be used as a temporary bitmap when needed.
The table object has also two pointer to bitmaps read_set and write_set
that the handler should use to find out which columns are used in which way.
- count() optimization now calls handler::records() instead of using
handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true).
- Added extra argument to Item::walk() to indicate if we should also
traverse sub queries.
- Added TABLE parameter to cp_buffer_from_ref()
- Don't close tables created with CREATE ... SELECT but keep them in
the table cache. (Faster usage of newly created tables).
New interfaces:
- table->clear_column_bitmaps() to initialize the bitmaps for tables
at start of new statements.
- table->column_bitmaps_set() to set up new column bitmaps and signal
the handler about this.
- table->column_bitmaps_set_no_signal() for some few cases where we need
to setup new column bitmaps but don't signal the handler (as the handler
has already been signaled about these before). Used for the momement
only in opt_range.cc when doing ROR scans.
- table->use_all_columns() to install a bitmap where all columns are marked
as use in the read and the write set.
- table->default_column_bitmaps() to install the normal read and write
column bitmaps, but not signaling the handler about this.
This is mainly used when creating TABLE instances.
- table->mark_columns_needed_for_delete(),
table->mark_columns_needed_for_delete() and
table->mark_columns_needed_for_insert() to allow us to put additional
columns in column usage maps if handler so requires.
(The handler indicates what it neads in handler->table_flags())
- table->prepare_for_position() to allow us to tell handler that it
needs to read primary key parts to be able to store them in
future table->position() calls.
(This replaces the table->file->ha_retrieve_all_pk function)
- table->mark_auto_increment_column() to tell handler are going to update
columns part of any auto_increment key.
- table->mark_columns_used_by_index() to mark all columns that is part of
an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow
it to quickly know that it only needs to read colums that are part
of the key. (The handler can also use the column map for detecting this,
but simpler/faster handler can just monitor the extra() call).
- table->mark_columns_used_by_index_no_reset() to in addition to other columns,
also mark all columns that is used by the given key.
- table->restore_column_maps_after_mark_index() to restore to default
column maps after a call to table->mark_columns_used_by_index().
- New item function register_field_in_read_map(), for marking used columns
in table->read_map. Used by filesort() to mark all used columns
- Maintain in TABLE->merge_keys set of all keys that are used in query.
(Simplices some optimization loops)
- Maintain Field->part_of_key_not_clustered which is like Field->part_of_key
but the field in the clustered key is not assumed to be part of all index.
(used in opt_range.cc for faster loops)
- dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map()
tmp_use_all_columns() and tmp_restore_column_map() functions to temporally
mark all columns as usable. The 'dbug_' version is primarily intended
inside a handler when it wants to just call Field:store() & Field::val()
functions, but don't need the column maps set for any other usage.
(ie:: bitmap_is_set() is never called)
- We can't use compare_records() to skip updates for handlers that returns
a partial column set and the read_set doesn't cover all columns in the
write set. The reason for this is that if we have a column marked only for
write we can't in the MySQL level know if the value changed or not.
The reason this worked before was that MySQL marked all to be written
columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden
bug'.
- open_table_from_share() does not anymore setup temporary MEM_ROOT
object as a thread specific variable for the handler. Instead we
send the to-be-used MEMROOT to get_new_handler().
(Simpler, faster code)
Bugs fixed:
- Column marking was not done correctly in a lot of cases.
(ALTER TABLE, when using triggers, auto_increment fields etc)
(Could potentially result in wrong values inserted in table handlers
relying on that the old column maps or field->set_query_id was correct)
Especially when it comes to triggers, there may be cases where the
old code would cause lost/wrong values for NDB and/or InnoDB tables.
- Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags:
OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG.
This allowed me to remove some wrong warnings about:
"Some non-transactional changed tables couldn't be rolled back"
- Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset
(thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose
some warnings about
"Some non-transactional changed tables couldn't be rolled back")
- Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table()
which could cause delete_table to report random failures.
- Fixed core dumps for some tests when running with --debug
- Added missing FN_LIBCHAR in mysql_rm_tmp_tables()
(This has probably caused us to not properly remove temporary files after
crash)
- slow_logs was not properly initialized, which could maybe cause
extra/lost entries in slow log.
- If we get an duplicate row on insert, change column map to read and
write all columns while retrying the operation. This is required by
the definition of REPLACE and also ensures that fields that are only
part of UPDATE are properly handled. This fixed a bug in NDB and
REPLACE where REPLACE wrongly copied some column values from the replaced
row.
- For table handler that doesn't support NULL in keys, we would give an error
when creating a primary key with NULL fields, even after the fields has been
automaticly converted to NOT NULL.
- Creating a primary key on a SPATIAL key, would fail if field was not
declared as NOT NULL.
Cleanups:
- Removed not used condition argument to setup_tables
- Removed not needed item function reset_query_id_processor().
- Field->add_index is removed. Now this is instead maintained in
(field->flags & FIELD_IN_ADD_INDEX)
- Field->fieldnr is removed (use field->field_index instead)
- New argument to filesort() to indicate that it should return a set of
row pointers (not used columns). This allowed me to remove some references
to sql_command in filesort and should also enable us to return column
results in some cases where we couldn't before.
- Changed column bitmap handling in opt_range.cc to be aligned with TABLE
bitmap, which allowed me to use bitmap functions instead of looping over
all fields to create some needed bitmaps. (Faster and smaller code)
- Broke up found too long lines
- Moved some variable declaration at start of function for better code
readability.
- Removed some not used arguments from functions.
(setup_fields(), mysql_prepare_insert_check_table())
- setup_fields() now takes an enum instead of an int for marking columns
usage.
- For internal temporary tables, use handler::write_row(),
handler::delete_row() and handler::update_row() instead of
handler::ha_xxxx() for faster execution.
- Changed some constants to enum's and define's.
- Using separate column read and write sets allows for easier checking
of timestamp field was set by statement.
- Remove calls to free_io_cache() as this is now done automaticly in ha_reset()
- Don't build table->normalized_path as this is now identical to table->path
(after bar's fixes to convert filenames)
- Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to
do comparision with the 'convert-dbug-for-diff' tool.
Things left to do in 5.1:
- We wrongly log failed CREATE TABLE ... SELECT in some cases when using
row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result)
Mats has promised to look into this.
- Test that my fix for CREATE TABLE ... SELECT is indeed correct.
(I added several test cases for this, but in this case it's better that
someone else also tests this throughly).
Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
|
|
|
DBUG_PRINT("lock",("write locked by thread: 0x%lx",
|
2005-07-19 20:21:12 +02:00
|
|
|
lock->write.data->owner->info->thread_id));
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2004-10-20 15:04:28 +02:00
|
|
|
DBUG_PRINT("info", ("write_wait.data: 0x%lx",
|
|
|
|
(ulong) lock->write_wait.data));
|
2000-07-31 21:29:14 +02:00
|
|
|
if (!lock->write_wait.data)
|
|
|
|
{ /* no scheduled write locks */
|
2005-05-13 11:08:08 +02:00
|
|
|
my_bool concurrent_insert= 0;
|
|
|
|
if (lock_type == TL_WRITE_CONCURRENT_INSERT)
|
|
|
|
{
|
|
|
|
concurrent_insert= 1;
|
|
|
|
if ((*lock->check_status)(data->status_param))
|
|
|
|
{
|
|
|
|
concurrent_insert= 0;
|
|
|
|
data->type=lock_type= thr_upgraded_concurrent_insert_lock;
|
|
|
|
}
|
|
|
|
}
|
2000-07-31 21:29:14 +02:00
|
|
|
|
|
|
|
if (!lock->read.data ||
|
|
|
|
(lock_type <= TL_WRITE_DELAYED &&
|
|
|
|
((lock_type != TL_WRITE_CONCURRENT_INSERT &&
|
|
|
|
lock_type != TL_WRITE_ALLOW_WRITE) ||
|
|
|
|
!lock->read_no_write_count)))
|
|
|
|
{
|
|
|
|
(*lock->write.last)=data; /* Add as current write lock */
|
|
|
|
data->prev=lock->write.last;
|
|
|
|
lock->write.last= &data->next;
|
|
|
|
if (data->lock->get_status)
|
2005-05-13 11:08:08 +02:00
|
|
|
(*data->lock->get_status)(data->status_param, concurrent_insert);
|
2000-07-31 21:29:14 +02:00
|
|
|
check_locks(lock,"only write lock",0);
|
2001-05-23 01:30:17 +02:00
|
|
|
statistic_increment(locks_immediate,&THR_LOCK_lock);
|
2000-07-31 21:29:14 +02:00
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
}
|
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes
Changes that requires code changes in other code of other storage engines.
(Note that all changes are very straightforward and one should find all issues
by compiling a --debug build and fixing all compiler errors and all
asserts in field.cc while running the test suite),
- New optional handler function introduced: reset()
This is called after every DML statement to make it easy for a handler to
statement specific cleanups.
(The only case it's not called is if force the file to be closed)
- handler::extra(HA_EXTRA_RESET) is removed. Code that was there before
should be moved to handler::reset()
- table->read_set contains a bitmap over all columns that are needed
in the query. read_row() and similar functions only needs to read these
columns
- table->write_set contains a bitmap over all columns that will be updated
in the query. write_row() and update_row() only needs to update these
columns.
The above bitmaps should now be up to date in all context
(including ALTER TABLE, filesort()).
The handler is informed of any changes to the bitmap after
fix_fields() by calling the virtual function
handler::column_bitmaps_signal(). If the handler does caching of
these bitmaps (instead of using table->read_set, table->write_set),
it should redo the caching in this code. as the signal() may be sent
several times, it's probably best to set a variable in the signal
and redo the caching on read_row() / write_row() if the variable was
set.
- Removed the read_set and write_set bitmap objects from the handler class
- Removed all column bit handling functions from the handler class.
(Now one instead uses the normal bitmap functions in my_bitmap.c instead
of handler dedicated bitmap functions)
- field->query_id is removed. One should instead instead check
table->read_set and table->write_set if a field is used in the query.
- handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and
handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now
instead use table->read_set to check for which columns to retrieve.
- If a handler needs to call Field->val() or Field->store() on columns
that are not used in the query, one should install a temporary
all-columns-used map while doing so. For this, we provide the following
functions:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set);
field->val();
dbug_tmp_restore_column_map(table->read_set, old_map);
and similar for the write map:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set);
field->val();
dbug_tmp_restore_column_map(table->write_set, old_map);
If this is not done, you will sooner or later hit a DBUG_ASSERT
in the field store() / val() functions.
(For not DBUG binaries, the dbug_tmp_restore_column_map() and
dbug_tmp_restore_column_map() are inline dummy functions and should
be optimized away be the compiler).
- If one needs to temporary set the column map for all binaries (and not
just to avoid the DBUG_ASSERT() in the Field::store() / Field::val()
methods) one should use the functions tmp_use_all_columns() and
tmp_restore_column_map() instead of the above dbug_ variants.
- All 'status' fields in the handler base class (like records,
data_file_length etc) are now stored in a 'stats' struct. This makes
it easier to know what status variables are provided by the base
handler. This requires some trivial variable names in the extra()
function.
- New virtual function handler::records(). This is called to optimize
COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true.
(stats.records is not supposed to be an exact value. It's only has to
be 'reasonable enough' for the optimizer to be able to choose a good
optimization path).
- Non virtual handler::init() function added for caching of virtual
constants from engine.
- Removed has_transactions() virtual method. Now one should instead return
HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support
transactions.
- The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument
that is to be used with 'new handler_name()' to allocate the handler
in the right area. The xxxx_create_handler() function is also
responsible for any initialization of the object before returning.
For example, one should change:
static handler *myisam_create_handler(TABLE_SHARE *table)
{
return new ha_myisam(table);
}
->
static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root)
{
return new (mem_root) ha_myisam(table);
}
- New optional virtual function: use_hidden_primary_key().
This is called in case of an update/delete when
(table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined
but we don't have a primary key. This allows the handler to take precisions
in remembering any hidden primary key to able to update/delete any
found row. The default handler marks all columns to be read.
- handler::table_flags() now returns a ulonglong (to allow for more flags).
- New/changed table_flags()
- HA_HAS_RECORDS Set if ::records() is supported
- HA_NO_TRANSACTIONS Set if engine doesn't support transactions
- HA_PRIMARY_KEY_REQUIRED_FOR_DELETE
Set if we should mark all primary key columns for
read when reading rows as part of a DELETE
statement. If there is no primary key,
all columns are marked for read.
- HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some
cases (based on table->read_set)
- HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS
Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION.
- HA_DUPP_POS Renamed to HA_DUPLICATE_POS
- HA_REQUIRES_KEY_COLUMNS_FOR_DELETE
Set this if we should mark ALL key columns for
read when when reading rows as part of a DELETE
statement. In case of an update we will mark
all keys for read for which key part changed
value.
- HA_STATS_RECORDS_IS_EXACT
Set this if stats.records is exact.
(This saves us some extra records() calls
when optimizing COUNT(*))
- Removed table_flags()
- HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if
handler::records() gives an exact count() and
HA_STATS_RECORDS_IS_EXACT if stats.records is exact.
- HA_READ_RND_SAME Removed (no one supported this one)
- Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk()
- Renamed handler::dupp_pos to handler::dup_pos
- Removed not used variable handler::sortkey
Upper level handler changes:
- ha_reset() now does some overall checks and calls ::reset()
- ha_table_flags() added. This is a cached version of table_flags(). The
cache is updated on engine creation time and updated on open.
MySQL level changes (not obvious from the above):
- DBUG_ASSERT() added to check that column usage matches what is set
in the column usage bit maps. (This found a LOT of bugs in current
column marking code).
- In 5.1 before, all used columns was marked in read_set and only updated
columns was marked in write_set. Now we only mark columns for which we
need a value in read_set.
- Column bitmaps are created in open_binary_frm() and open_table_from_share().
(Before this was in table.cc)
- handler::table_flags() calls are replaced with handler::ha_table_flags()
- For calling field->val() you must have the corresponding bit set in
table->read_set. For calling field->store() you must have the
corresponding bit set in table->write_set. (There are asserts in
all store()/val() functions to catch wrong usage)
- thd->set_query_id is renamed to thd->mark_used_columns and instead
of setting this to an integer value, this has now the values:
MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE
Changed also all variables named 'set_query_id' to mark_used_columns.
- In filesort() we now inform the handler of exactly which columns are needed
doing the sort and choosing the rows.
- The TABLE_SHARE object has a 'all_set' column bitmap one can use
when one needs a column bitmap with all columns set.
(This is used for table->use_all_columns() and other places)
- The TABLE object has 3 column bitmaps:
- def_read_set Default bitmap for columns to be read
- def_write_set Default bitmap for columns to be written
- tmp_set Can be used as a temporary bitmap when needed.
The table object has also two pointer to bitmaps read_set and write_set
that the handler should use to find out which columns are used in which way.
- count() optimization now calls handler::records() instead of using
handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true).
- Added extra argument to Item::walk() to indicate if we should also
traverse sub queries.
- Added TABLE parameter to cp_buffer_from_ref()
- Don't close tables created with CREATE ... SELECT but keep them in
the table cache. (Faster usage of newly created tables).
New interfaces:
- table->clear_column_bitmaps() to initialize the bitmaps for tables
at start of new statements.
- table->column_bitmaps_set() to set up new column bitmaps and signal
the handler about this.
- table->column_bitmaps_set_no_signal() for some few cases where we need
to setup new column bitmaps but don't signal the handler (as the handler
has already been signaled about these before). Used for the momement
only in opt_range.cc when doing ROR scans.
- table->use_all_columns() to install a bitmap where all columns are marked
as use in the read and the write set.
- table->default_column_bitmaps() to install the normal read and write
column bitmaps, but not signaling the handler about this.
This is mainly used when creating TABLE instances.
- table->mark_columns_needed_for_delete(),
table->mark_columns_needed_for_delete() and
table->mark_columns_needed_for_insert() to allow us to put additional
columns in column usage maps if handler so requires.
(The handler indicates what it neads in handler->table_flags())
- table->prepare_for_position() to allow us to tell handler that it
needs to read primary key parts to be able to store them in
future table->position() calls.
(This replaces the table->file->ha_retrieve_all_pk function)
- table->mark_auto_increment_column() to tell handler are going to update
columns part of any auto_increment key.
- table->mark_columns_used_by_index() to mark all columns that is part of
an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow
it to quickly know that it only needs to read colums that are part
of the key. (The handler can also use the column map for detecting this,
but simpler/faster handler can just monitor the extra() call).
- table->mark_columns_used_by_index_no_reset() to in addition to other columns,
also mark all columns that is used by the given key.
- table->restore_column_maps_after_mark_index() to restore to default
column maps after a call to table->mark_columns_used_by_index().
- New item function register_field_in_read_map(), for marking used columns
in table->read_map. Used by filesort() to mark all used columns
- Maintain in TABLE->merge_keys set of all keys that are used in query.
(Simplices some optimization loops)
- Maintain Field->part_of_key_not_clustered which is like Field->part_of_key
but the field in the clustered key is not assumed to be part of all index.
(used in opt_range.cc for faster loops)
- dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map()
tmp_use_all_columns() and tmp_restore_column_map() functions to temporally
mark all columns as usable. The 'dbug_' version is primarily intended
inside a handler when it wants to just call Field:store() & Field::val()
functions, but don't need the column maps set for any other usage.
(ie:: bitmap_is_set() is never called)
- We can't use compare_records() to skip updates for handlers that returns
a partial column set and the read_set doesn't cover all columns in the
write set. The reason for this is that if we have a column marked only for
write we can't in the MySQL level know if the value changed or not.
The reason this worked before was that MySQL marked all to be written
columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden
bug'.
- open_table_from_share() does not anymore setup temporary MEM_ROOT
object as a thread specific variable for the handler. Instead we
send the to-be-used MEMROOT to get_new_handler().
(Simpler, faster code)
Bugs fixed:
- Column marking was not done correctly in a lot of cases.
(ALTER TABLE, when using triggers, auto_increment fields etc)
(Could potentially result in wrong values inserted in table handlers
relying on that the old column maps or field->set_query_id was correct)
Especially when it comes to triggers, there may be cases where the
old code would cause lost/wrong values for NDB and/or InnoDB tables.
- Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags:
OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG.
This allowed me to remove some wrong warnings about:
"Some non-transactional changed tables couldn't be rolled back"
- Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset
(thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose
some warnings about
"Some non-transactional changed tables couldn't be rolled back")
- Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table()
which could cause delete_table to report random failures.
- Fixed core dumps for some tests when running with --debug
- Added missing FN_LIBCHAR in mysql_rm_tmp_tables()
(This has probably caused us to not properly remove temporary files after
crash)
- slow_logs was not properly initialized, which could maybe cause
extra/lost entries in slow log.
- If we get an duplicate row on insert, change column map to read and
write all columns while retrying the operation. This is required by
the definition of REPLACE and also ensures that fields that are only
part of UPDATE are properly handled. This fixed a bug in NDB and
REPLACE where REPLACE wrongly copied some column values from the replaced
row.
- For table handler that doesn't support NULL in keys, we would give an error
when creating a primary key with NULL fields, even after the fields has been
automaticly converted to NOT NULL.
- Creating a primary key on a SPATIAL key, would fail if field was not
declared as NOT NULL.
Cleanups:
- Removed not used condition argument to setup_tables
- Removed not needed item function reset_query_id_processor().
- Field->add_index is removed. Now this is instead maintained in
(field->flags & FIELD_IN_ADD_INDEX)
- Field->fieldnr is removed (use field->field_index instead)
- New argument to filesort() to indicate that it should return a set of
row pointers (not used columns). This allowed me to remove some references
to sql_command in filesort and should also enable us to return column
results in some cases where we couldn't before.
- Changed column bitmap handling in opt_range.cc to be aligned with TABLE
bitmap, which allowed me to use bitmap functions instead of looping over
all fields to create some needed bitmaps. (Faster and smaller code)
- Broke up found too long lines
- Moved some variable declaration at start of function for better code
readability.
- Removed some not used arguments from functions.
(setup_fields(), mysql_prepare_insert_check_table())
- setup_fields() now takes an enum instead of an int for marking columns
usage.
- For internal temporary tables, use handler::write_row(),
handler::delete_row() and handler::update_row() instead of
handler::ha_xxxx() for faster execution.
- Changed some constants to enum's and define's.
- Using separate column read and write sets allows for easier checking
of timestamp field was set by statement.
- Remove calls to free_io_cache() as this is now done automaticly in ha_reset()
- Don't build table->normalized_path as this is now identical to table->path
(after bar's fixes to convert filenames)
- Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to
do comparision with the 'convert-dbug-for-diff' tool.
Things left to do in 5.1:
- We wrongly log failed CREATE TABLE ... SELECT in some cases when using
row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result)
Mats has promised to look into this.
- Test that my fix for CREATE TABLE ... SELECT is indeed correct.
(I added several test cases for this, but in this case it's better that
someone else also tests this throughly).
Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
|
|
|
DBUG_PRINT("lock",("write locked by thread: 0x%lx, type: %ld",
|
2005-07-19 20:21:12 +02:00
|
|
|
lock->read.data->owner->info->thread_id, data->type));
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
2005-07-19 20:21:12 +02:00
|
|
|
wait_queue= &lock->write_wait;
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
2005-07-19 20:21:12 +02:00
|
|
|
/*
|
|
|
|
Try to detect a trivial deadlock when using cursors: attempt to
|
|
|
|
lock a table that is already locked by an open cursor within the
|
|
|
|
same connection. lock_owner can be zero if we succumbed to a high
|
|
|
|
priority writer in the write_wait queue.
|
|
|
|
*/
|
|
|
|
lock_owner= lock->read.data ? lock->read.data : lock->write.data;
|
|
|
|
if (lock_owner && lock_owner->owner->info == owner->info)
|
|
|
|
{
|
|
|
|
result= THR_LOCK_DEADLOCK;
|
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
/* Can't get lock yet; Wait for it */
|
|
|
|
DBUG_RETURN(wait_for_lock(wait_queue, data, 0));
|
2000-07-31 21:29:14 +02:00
|
|
|
end:
|
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
|
|
|
DBUG_RETURN(result);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static inline void free_all_read_locks(THR_LOCK *lock,
|
|
|
|
bool using_concurrent_insert)
|
|
|
|
{
|
|
|
|
THR_LOCK_DATA *data=lock->read_wait.data;
|
|
|
|
|
|
|
|
check_locks(lock,"before freeing read locks",1);
|
|
|
|
|
|
|
|
/* move all locks from read_wait list to read list */
|
|
|
|
(*lock->read.last)=data;
|
|
|
|
data->prev=lock->read.last;
|
|
|
|
lock->read.last=lock->read_wait.last;
|
|
|
|
|
|
|
|
/* Clear read_wait list */
|
|
|
|
lock->read_wait.last= &lock->read_wait.data;
|
|
|
|
|
|
|
|
do
|
|
|
|
{
|
|
|
|
pthread_cond_t *cond=data->cond;
|
|
|
|
if ((int) data->type == (int) TL_READ_NO_INSERT)
|
|
|
|
{
|
|
|
|
if (using_concurrent_insert)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
We can't free this lock;
|
|
|
|
Link lock away from read chain back into read_wait chain
|
|
|
|
*/
|
|
|
|
if (((*data->prev)=data->next))
|
|
|
|
data->next->prev=data->prev;
|
|
|
|
else
|
|
|
|
lock->read.last=data->prev;
|
|
|
|
*lock->read_wait.last= data;
|
|
|
|
data->prev= lock->read_wait.last;
|
|
|
|
lock->read_wait.last= &data->next;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
lock->read_no_write_count++;
|
|
|
|
}
|
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes
Changes that requires code changes in other code of other storage engines.
(Note that all changes are very straightforward and one should find all issues
by compiling a --debug build and fixing all compiler errors and all
asserts in field.cc while running the test suite),
- New optional handler function introduced: reset()
This is called after every DML statement to make it easy for a handler to
statement specific cleanups.
(The only case it's not called is if force the file to be closed)
- handler::extra(HA_EXTRA_RESET) is removed. Code that was there before
should be moved to handler::reset()
- table->read_set contains a bitmap over all columns that are needed
in the query. read_row() and similar functions only needs to read these
columns
- table->write_set contains a bitmap over all columns that will be updated
in the query. write_row() and update_row() only needs to update these
columns.
The above bitmaps should now be up to date in all context
(including ALTER TABLE, filesort()).
The handler is informed of any changes to the bitmap after
fix_fields() by calling the virtual function
handler::column_bitmaps_signal(). If the handler does caching of
these bitmaps (instead of using table->read_set, table->write_set),
it should redo the caching in this code. as the signal() may be sent
several times, it's probably best to set a variable in the signal
and redo the caching on read_row() / write_row() if the variable was
set.
- Removed the read_set and write_set bitmap objects from the handler class
- Removed all column bit handling functions from the handler class.
(Now one instead uses the normal bitmap functions in my_bitmap.c instead
of handler dedicated bitmap functions)
- field->query_id is removed. One should instead instead check
table->read_set and table->write_set if a field is used in the query.
- handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and
handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now
instead use table->read_set to check for which columns to retrieve.
- If a handler needs to call Field->val() or Field->store() on columns
that are not used in the query, one should install a temporary
all-columns-used map while doing so. For this, we provide the following
functions:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set);
field->val();
dbug_tmp_restore_column_map(table->read_set, old_map);
and similar for the write map:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set);
field->val();
dbug_tmp_restore_column_map(table->write_set, old_map);
If this is not done, you will sooner or later hit a DBUG_ASSERT
in the field store() / val() functions.
(For not DBUG binaries, the dbug_tmp_restore_column_map() and
dbug_tmp_restore_column_map() are inline dummy functions and should
be optimized away be the compiler).
- If one needs to temporary set the column map for all binaries (and not
just to avoid the DBUG_ASSERT() in the Field::store() / Field::val()
methods) one should use the functions tmp_use_all_columns() and
tmp_restore_column_map() instead of the above dbug_ variants.
- All 'status' fields in the handler base class (like records,
data_file_length etc) are now stored in a 'stats' struct. This makes
it easier to know what status variables are provided by the base
handler. This requires some trivial variable names in the extra()
function.
- New virtual function handler::records(). This is called to optimize
COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true.
(stats.records is not supposed to be an exact value. It's only has to
be 'reasonable enough' for the optimizer to be able to choose a good
optimization path).
- Non virtual handler::init() function added for caching of virtual
constants from engine.
- Removed has_transactions() virtual method. Now one should instead return
HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support
transactions.
- The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument
that is to be used with 'new handler_name()' to allocate the handler
in the right area. The xxxx_create_handler() function is also
responsible for any initialization of the object before returning.
For example, one should change:
static handler *myisam_create_handler(TABLE_SHARE *table)
{
return new ha_myisam(table);
}
->
static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root)
{
return new (mem_root) ha_myisam(table);
}
- New optional virtual function: use_hidden_primary_key().
This is called in case of an update/delete when
(table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined
but we don't have a primary key. This allows the handler to take precisions
in remembering any hidden primary key to able to update/delete any
found row. The default handler marks all columns to be read.
- handler::table_flags() now returns a ulonglong (to allow for more flags).
- New/changed table_flags()
- HA_HAS_RECORDS Set if ::records() is supported
- HA_NO_TRANSACTIONS Set if engine doesn't support transactions
- HA_PRIMARY_KEY_REQUIRED_FOR_DELETE
Set if we should mark all primary key columns for
read when reading rows as part of a DELETE
statement. If there is no primary key,
all columns are marked for read.
- HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some
cases (based on table->read_set)
- HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS
Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION.
- HA_DUPP_POS Renamed to HA_DUPLICATE_POS
- HA_REQUIRES_KEY_COLUMNS_FOR_DELETE
Set this if we should mark ALL key columns for
read when when reading rows as part of a DELETE
statement. In case of an update we will mark
all keys for read for which key part changed
value.
- HA_STATS_RECORDS_IS_EXACT
Set this if stats.records is exact.
(This saves us some extra records() calls
when optimizing COUNT(*))
- Removed table_flags()
- HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if
handler::records() gives an exact count() and
HA_STATS_RECORDS_IS_EXACT if stats.records is exact.
- HA_READ_RND_SAME Removed (no one supported this one)
- Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk()
- Renamed handler::dupp_pos to handler::dup_pos
- Removed not used variable handler::sortkey
Upper level handler changes:
- ha_reset() now does some overall checks and calls ::reset()
- ha_table_flags() added. This is a cached version of table_flags(). The
cache is updated on engine creation time and updated on open.
MySQL level changes (not obvious from the above):
- DBUG_ASSERT() added to check that column usage matches what is set
in the column usage bit maps. (This found a LOT of bugs in current
column marking code).
- In 5.1 before, all used columns was marked in read_set and only updated
columns was marked in write_set. Now we only mark columns for which we
need a value in read_set.
- Column bitmaps are created in open_binary_frm() and open_table_from_share().
(Before this was in table.cc)
- handler::table_flags() calls are replaced with handler::ha_table_flags()
- For calling field->val() you must have the corresponding bit set in
table->read_set. For calling field->store() you must have the
corresponding bit set in table->write_set. (There are asserts in
all store()/val() functions to catch wrong usage)
- thd->set_query_id is renamed to thd->mark_used_columns and instead
of setting this to an integer value, this has now the values:
MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE
Changed also all variables named 'set_query_id' to mark_used_columns.
- In filesort() we now inform the handler of exactly which columns are needed
doing the sort and choosing the rows.
- The TABLE_SHARE object has a 'all_set' column bitmap one can use
when one needs a column bitmap with all columns set.
(This is used for table->use_all_columns() and other places)
- The TABLE object has 3 column bitmaps:
- def_read_set Default bitmap for columns to be read
- def_write_set Default bitmap for columns to be written
- tmp_set Can be used as a temporary bitmap when needed.
The table object has also two pointer to bitmaps read_set and write_set
that the handler should use to find out which columns are used in which way.
- count() optimization now calls handler::records() instead of using
handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true).
- Added extra argument to Item::walk() to indicate if we should also
traverse sub queries.
- Added TABLE parameter to cp_buffer_from_ref()
- Don't close tables created with CREATE ... SELECT but keep them in
the table cache. (Faster usage of newly created tables).
New interfaces:
- table->clear_column_bitmaps() to initialize the bitmaps for tables
at start of new statements.
- table->column_bitmaps_set() to set up new column bitmaps and signal
the handler about this.
- table->column_bitmaps_set_no_signal() for some few cases where we need
to setup new column bitmaps but don't signal the handler (as the handler
has already been signaled about these before). Used for the momement
only in opt_range.cc when doing ROR scans.
- table->use_all_columns() to install a bitmap where all columns are marked
as use in the read and the write set.
- table->default_column_bitmaps() to install the normal read and write
column bitmaps, but not signaling the handler about this.
This is mainly used when creating TABLE instances.
- table->mark_columns_needed_for_delete(),
table->mark_columns_needed_for_delete() and
table->mark_columns_needed_for_insert() to allow us to put additional
columns in column usage maps if handler so requires.
(The handler indicates what it neads in handler->table_flags())
- table->prepare_for_position() to allow us to tell handler that it
needs to read primary key parts to be able to store them in
future table->position() calls.
(This replaces the table->file->ha_retrieve_all_pk function)
- table->mark_auto_increment_column() to tell handler are going to update
columns part of any auto_increment key.
- table->mark_columns_used_by_index() to mark all columns that is part of
an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow
it to quickly know that it only needs to read colums that are part
of the key. (The handler can also use the column map for detecting this,
but simpler/faster handler can just monitor the extra() call).
- table->mark_columns_used_by_index_no_reset() to in addition to other columns,
also mark all columns that is used by the given key.
- table->restore_column_maps_after_mark_index() to restore to default
column maps after a call to table->mark_columns_used_by_index().
- New item function register_field_in_read_map(), for marking used columns
in table->read_map. Used by filesort() to mark all used columns
- Maintain in TABLE->merge_keys set of all keys that are used in query.
(Simplices some optimization loops)
- Maintain Field->part_of_key_not_clustered which is like Field->part_of_key
but the field in the clustered key is not assumed to be part of all index.
(used in opt_range.cc for faster loops)
- dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map()
tmp_use_all_columns() and tmp_restore_column_map() functions to temporally
mark all columns as usable. The 'dbug_' version is primarily intended
inside a handler when it wants to just call Field:store() & Field::val()
functions, but don't need the column maps set for any other usage.
(ie:: bitmap_is_set() is never called)
- We can't use compare_records() to skip updates for handlers that returns
a partial column set and the read_set doesn't cover all columns in the
write set. The reason for this is that if we have a column marked only for
write we can't in the MySQL level know if the value changed or not.
The reason this worked before was that MySQL marked all to be written
columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden
bug'.
- open_table_from_share() does not anymore setup temporary MEM_ROOT
object as a thread specific variable for the handler. Instead we
send the to-be-used MEMROOT to get_new_handler().
(Simpler, faster code)
Bugs fixed:
- Column marking was not done correctly in a lot of cases.
(ALTER TABLE, when using triggers, auto_increment fields etc)
(Could potentially result in wrong values inserted in table handlers
relying on that the old column maps or field->set_query_id was correct)
Especially when it comes to triggers, there may be cases where the
old code would cause lost/wrong values for NDB and/or InnoDB tables.
- Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags:
OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG.
This allowed me to remove some wrong warnings about:
"Some non-transactional changed tables couldn't be rolled back"
- Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset
(thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose
some warnings about
"Some non-transactional changed tables couldn't be rolled back")
- Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table()
which could cause delete_table to report random failures.
- Fixed core dumps for some tests when running with --debug
- Added missing FN_LIBCHAR in mysql_rm_tmp_tables()
(This has probably caused us to not properly remove temporary files after
crash)
- slow_logs was not properly initialized, which could maybe cause
extra/lost entries in slow log.
- If we get an duplicate row on insert, change column map to read and
write all columns while retrying the operation. This is required by
the definition of REPLACE and also ensures that fields that are only
part of UPDATE are properly handled. This fixed a bug in NDB and
REPLACE where REPLACE wrongly copied some column values from the replaced
row.
- For table handler that doesn't support NULL in keys, we would give an error
when creating a primary key with NULL fields, even after the fields has been
automaticly converted to NOT NULL.
- Creating a primary key on a SPATIAL key, would fail if field was not
declared as NOT NULL.
Cleanups:
- Removed not used condition argument to setup_tables
- Removed not needed item function reset_query_id_processor().
- Field->add_index is removed. Now this is instead maintained in
(field->flags & FIELD_IN_ADD_INDEX)
- Field->fieldnr is removed (use field->field_index instead)
- New argument to filesort() to indicate that it should return a set of
row pointers (not used columns). This allowed me to remove some references
to sql_command in filesort and should also enable us to return column
results in some cases where we couldn't before.
- Changed column bitmap handling in opt_range.cc to be aligned with TABLE
bitmap, which allowed me to use bitmap functions instead of looping over
all fields to create some needed bitmaps. (Faster and smaller code)
- Broke up found too long lines
- Moved some variable declaration at start of function for better code
readability.
- Removed some not used arguments from functions.
(setup_fields(), mysql_prepare_insert_check_table())
- setup_fields() now takes an enum instead of an int for marking columns
usage.
- For internal temporary tables, use handler::write_row(),
handler::delete_row() and handler::update_row() instead of
handler::ha_xxxx() for faster execution.
- Changed some constants to enum's and define's.
- Using separate column read and write sets allows for easier checking
of timestamp field was set by statement.
- Remove calls to free_io_cache() as this is now done automaticly in ha_reset()
- Don't build table->normalized_path as this is now identical to table->path
(after bar's fixes to convert filenames)
- Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to
do comparision with the 'convert-dbug-for-diff' tool.
Things left to do in 5.1:
- We wrongly log failed CREATE TABLE ... SELECT in some cases when using
row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result)
Mats has promised to look into this.
- Test that my fix for CREATE TABLE ... SELECT is indeed correct.
(I added several test cases for this, but in this case it's better that
someone else also tests this throughly).
Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
|
|
|
DBUG_PRINT("lock",("giving read lock to thread: 0x%lx",
|
2005-07-19 20:21:12 +02:00
|
|
|
data->owner->info->thread_id));
|
2000-07-31 21:29:14 +02:00
|
|
|
data->cond=0; /* Mark thread free */
|
|
|
|
VOID(pthread_cond_signal(cond));
|
|
|
|
} while ((data=data->next));
|
|
|
|
*lock->read_wait.last=0;
|
|
|
|
if (!lock->read_wait.data)
|
|
|
|
lock->write_lock_count=0;
|
|
|
|
check_locks(lock,"after giving read locks",0);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Unlock lock and free next thread on same lock */
|
|
|
|
|
|
|
|
void thr_unlock(THR_LOCK_DATA *data)
|
|
|
|
{
|
|
|
|
THR_LOCK *lock=data->lock;
|
|
|
|
enum thr_lock_type lock_type=data->type;
|
|
|
|
DBUG_ENTER("thr_unlock");
|
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes
Changes that requires code changes in other code of other storage engines.
(Note that all changes are very straightforward and one should find all issues
by compiling a --debug build and fixing all compiler errors and all
asserts in field.cc while running the test suite),
- New optional handler function introduced: reset()
This is called after every DML statement to make it easy for a handler to
statement specific cleanups.
(The only case it's not called is if force the file to be closed)
- handler::extra(HA_EXTRA_RESET) is removed. Code that was there before
should be moved to handler::reset()
- table->read_set contains a bitmap over all columns that are needed
in the query. read_row() and similar functions only needs to read these
columns
- table->write_set contains a bitmap over all columns that will be updated
in the query. write_row() and update_row() only needs to update these
columns.
The above bitmaps should now be up to date in all context
(including ALTER TABLE, filesort()).
The handler is informed of any changes to the bitmap after
fix_fields() by calling the virtual function
handler::column_bitmaps_signal(). If the handler does caching of
these bitmaps (instead of using table->read_set, table->write_set),
it should redo the caching in this code. as the signal() may be sent
several times, it's probably best to set a variable in the signal
and redo the caching on read_row() / write_row() if the variable was
set.
- Removed the read_set and write_set bitmap objects from the handler class
- Removed all column bit handling functions from the handler class.
(Now one instead uses the normal bitmap functions in my_bitmap.c instead
of handler dedicated bitmap functions)
- field->query_id is removed. One should instead instead check
table->read_set and table->write_set if a field is used in the query.
- handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and
handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now
instead use table->read_set to check for which columns to retrieve.
- If a handler needs to call Field->val() or Field->store() on columns
that are not used in the query, one should install a temporary
all-columns-used map while doing so. For this, we provide the following
functions:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set);
field->val();
dbug_tmp_restore_column_map(table->read_set, old_map);
and similar for the write map:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set);
field->val();
dbug_tmp_restore_column_map(table->write_set, old_map);
If this is not done, you will sooner or later hit a DBUG_ASSERT
in the field store() / val() functions.
(For not DBUG binaries, the dbug_tmp_restore_column_map() and
dbug_tmp_restore_column_map() are inline dummy functions and should
be optimized away be the compiler).
- If one needs to temporary set the column map for all binaries (and not
just to avoid the DBUG_ASSERT() in the Field::store() / Field::val()
methods) one should use the functions tmp_use_all_columns() and
tmp_restore_column_map() instead of the above dbug_ variants.
- All 'status' fields in the handler base class (like records,
data_file_length etc) are now stored in a 'stats' struct. This makes
it easier to know what status variables are provided by the base
handler. This requires some trivial variable names in the extra()
function.
- New virtual function handler::records(). This is called to optimize
COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true.
(stats.records is not supposed to be an exact value. It's only has to
be 'reasonable enough' for the optimizer to be able to choose a good
optimization path).
- Non virtual handler::init() function added for caching of virtual
constants from engine.
- Removed has_transactions() virtual method. Now one should instead return
HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support
transactions.
- The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument
that is to be used with 'new handler_name()' to allocate the handler
in the right area. The xxxx_create_handler() function is also
responsible for any initialization of the object before returning.
For example, one should change:
static handler *myisam_create_handler(TABLE_SHARE *table)
{
return new ha_myisam(table);
}
->
static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root)
{
return new (mem_root) ha_myisam(table);
}
- New optional virtual function: use_hidden_primary_key().
This is called in case of an update/delete when
(table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined
but we don't have a primary key. This allows the handler to take precisions
in remembering any hidden primary key to able to update/delete any
found row. The default handler marks all columns to be read.
- handler::table_flags() now returns a ulonglong (to allow for more flags).
- New/changed table_flags()
- HA_HAS_RECORDS Set if ::records() is supported
- HA_NO_TRANSACTIONS Set if engine doesn't support transactions
- HA_PRIMARY_KEY_REQUIRED_FOR_DELETE
Set if we should mark all primary key columns for
read when reading rows as part of a DELETE
statement. If there is no primary key,
all columns are marked for read.
- HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some
cases (based on table->read_set)
- HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS
Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION.
- HA_DUPP_POS Renamed to HA_DUPLICATE_POS
- HA_REQUIRES_KEY_COLUMNS_FOR_DELETE
Set this if we should mark ALL key columns for
read when when reading rows as part of a DELETE
statement. In case of an update we will mark
all keys for read for which key part changed
value.
- HA_STATS_RECORDS_IS_EXACT
Set this if stats.records is exact.
(This saves us some extra records() calls
when optimizing COUNT(*))
- Removed table_flags()
- HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if
handler::records() gives an exact count() and
HA_STATS_RECORDS_IS_EXACT if stats.records is exact.
- HA_READ_RND_SAME Removed (no one supported this one)
- Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk()
- Renamed handler::dupp_pos to handler::dup_pos
- Removed not used variable handler::sortkey
Upper level handler changes:
- ha_reset() now does some overall checks and calls ::reset()
- ha_table_flags() added. This is a cached version of table_flags(). The
cache is updated on engine creation time and updated on open.
MySQL level changes (not obvious from the above):
- DBUG_ASSERT() added to check that column usage matches what is set
in the column usage bit maps. (This found a LOT of bugs in current
column marking code).
- In 5.1 before, all used columns was marked in read_set and only updated
columns was marked in write_set. Now we only mark columns for which we
need a value in read_set.
- Column bitmaps are created in open_binary_frm() and open_table_from_share().
(Before this was in table.cc)
- handler::table_flags() calls are replaced with handler::ha_table_flags()
- For calling field->val() you must have the corresponding bit set in
table->read_set. For calling field->store() you must have the
corresponding bit set in table->write_set. (There are asserts in
all store()/val() functions to catch wrong usage)
- thd->set_query_id is renamed to thd->mark_used_columns and instead
of setting this to an integer value, this has now the values:
MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE
Changed also all variables named 'set_query_id' to mark_used_columns.
- In filesort() we now inform the handler of exactly which columns are needed
doing the sort and choosing the rows.
- The TABLE_SHARE object has a 'all_set' column bitmap one can use
when one needs a column bitmap with all columns set.
(This is used for table->use_all_columns() and other places)
- The TABLE object has 3 column bitmaps:
- def_read_set Default bitmap for columns to be read
- def_write_set Default bitmap for columns to be written
- tmp_set Can be used as a temporary bitmap when needed.
The table object has also two pointer to bitmaps read_set and write_set
that the handler should use to find out which columns are used in which way.
- count() optimization now calls handler::records() instead of using
handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true).
- Added extra argument to Item::walk() to indicate if we should also
traverse sub queries.
- Added TABLE parameter to cp_buffer_from_ref()
- Don't close tables created with CREATE ... SELECT but keep them in
the table cache. (Faster usage of newly created tables).
New interfaces:
- table->clear_column_bitmaps() to initialize the bitmaps for tables
at start of new statements.
- table->column_bitmaps_set() to set up new column bitmaps and signal
the handler about this.
- table->column_bitmaps_set_no_signal() for some few cases where we need
to setup new column bitmaps but don't signal the handler (as the handler
has already been signaled about these before). Used for the momement
only in opt_range.cc when doing ROR scans.
- table->use_all_columns() to install a bitmap where all columns are marked
as use in the read and the write set.
- table->default_column_bitmaps() to install the normal read and write
column bitmaps, but not signaling the handler about this.
This is mainly used when creating TABLE instances.
- table->mark_columns_needed_for_delete(),
table->mark_columns_needed_for_delete() and
table->mark_columns_needed_for_insert() to allow us to put additional
columns in column usage maps if handler so requires.
(The handler indicates what it neads in handler->table_flags())
- table->prepare_for_position() to allow us to tell handler that it
needs to read primary key parts to be able to store them in
future table->position() calls.
(This replaces the table->file->ha_retrieve_all_pk function)
- table->mark_auto_increment_column() to tell handler are going to update
columns part of any auto_increment key.
- table->mark_columns_used_by_index() to mark all columns that is part of
an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow
it to quickly know that it only needs to read colums that are part
of the key. (The handler can also use the column map for detecting this,
but simpler/faster handler can just monitor the extra() call).
- table->mark_columns_used_by_index_no_reset() to in addition to other columns,
also mark all columns that is used by the given key.
- table->restore_column_maps_after_mark_index() to restore to default
column maps after a call to table->mark_columns_used_by_index().
- New item function register_field_in_read_map(), for marking used columns
in table->read_map. Used by filesort() to mark all used columns
- Maintain in TABLE->merge_keys set of all keys that are used in query.
(Simplices some optimization loops)
- Maintain Field->part_of_key_not_clustered which is like Field->part_of_key
but the field in the clustered key is not assumed to be part of all index.
(used in opt_range.cc for faster loops)
- dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map()
tmp_use_all_columns() and tmp_restore_column_map() functions to temporally
mark all columns as usable. The 'dbug_' version is primarily intended
inside a handler when it wants to just call Field:store() & Field::val()
functions, but don't need the column maps set for any other usage.
(ie:: bitmap_is_set() is never called)
- We can't use compare_records() to skip updates for handlers that returns
a partial column set and the read_set doesn't cover all columns in the
write set. The reason for this is that if we have a column marked only for
write we can't in the MySQL level know if the value changed or not.
The reason this worked before was that MySQL marked all to be written
columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden
bug'.
- open_table_from_share() does not anymore setup temporary MEM_ROOT
object as a thread specific variable for the handler. Instead we
send the to-be-used MEMROOT to get_new_handler().
(Simpler, faster code)
Bugs fixed:
- Column marking was not done correctly in a lot of cases.
(ALTER TABLE, when using triggers, auto_increment fields etc)
(Could potentially result in wrong values inserted in table handlers
relying on that the old column maps or field->set_query_id was correct)
Especially when it comes to triggers, there may be cases where the
old code would cause lost/wrong values for NDB and/or InnoDB tables.
- Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags:
OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG.
This allowed me to remove some wrong warnings about:
"Some non-transactional changed tables couldn't be rolled back"
- Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset
(thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose
some warnings about
"Some non-transactional changed tables couldn't be rolled back")
- Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table()
which could cause delete_table to report random failures.
- Fixed core dumps for some tests when running with --debug
- Added missing FN_LIBCHAR in mysql_rm_tmp_tables()
(This has probably caused us to not properly remove temporary files after
crash)
- slow_logs was not properly initialized, which could maybe cause
extra/lost entries in slow log.
- If we get an duplicate row on insert, change column map to read and
write all columns while retrying the operation. This is required by
the definition of REPLACE and also ensures that fields that are only
part of UPDATE are properly handled. This fixed a bug in NDB and
REPLACE where REPLACE wrongly copied some column values from the replaced
row.
- For table handler that doesn't support NULL in keys, we would give an error
when creating a primary key with NULL fields, even after the fields has been
automaticly converted to NOT NULL.
- Creating a primary key on a SPATIAL key, would fail if field was not
declared as NOT NULL.
Cleanups:
- Removed not used condition argument to setup_tables
- Removed not needed item function reset_query_id_processor().
- Field->add_index is removed. Now this is instead maintained in
(field->flags & FIELD_IN_ADD_INDEX)
- Field->fieldnr is removed (use field->field_index instead)
- New argument to filesort() to indicate that it should return a set of
row pointers (not used columns). This allowed me to remove some references
to sql_command in filesort and should also enable us to return column
results in some cases where we couldn't before.
- Changed column bitmap handling in opt_range.cc to be aligned with TABLE
bitmap, which allowed me to use bitmap functions instead of looping over
all fields to create some needed bitmaps. (Faster and smaller code)
- Broke up found too long lines
- Moved some variable declaration at start of function for better code
readability.
- Removed some not used arguments from functions.
(setup_fields(), mysql_prepare_insert_check_table())
- setup_fields() now takes an enum instead of an int for marking columns
usage.
- For internal temporary tables, use handler::write_row(),
handler::delete_row() and handler::update_row() instead of
handler::ha_xxxx() for faster execution.
- Changed some constants to enum's and define's.
- Using separate column read and write sets allows for easier checking
of timestamp field was set by statement.
- Remove calls to free_io_cache() as this is now done automaticly in ha_reset()
- Don't build table->normalized_path as this is now identical to table->path
(after bar's fixes to convert filenames)
- Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to
do comparision with the 'convert-dbug-for-diff' tool.
Things left to do in 5.1:
- We wrongly log failed CREATE TABLE ... SELECT in some cases when using
row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result)
Mats has promised to look into this.
- Test that my fix for CREATE TABLE ... SELECT is indeed correct.
(I added several test cases for this, but in this case it's better that
someone else also tests this throughly).
Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
|
|
|
DBUG_PRINT("lock",("data: 0x%lx thread: 0x%lx lock: 0x%lx",
|
2005-07-19 20:21:12 +02:00
|
|
|
data, data->owner->info->thread_id, lock));
|
2000-07-31 21:29:14 +02:00
|
|
|
pthread_mutex_lock(&lock->mutex);
|
|
|
|
check_locks(lock,"start of release lock",0);
|
|
|
|
|
|
|
|
if (((*data->prev)=data->next)) /* remove from lock-list */
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else if (lock_type <= TL_READ_NO_INSERT)
|
|
|
|
lock->read.last=data->prev;
|
|
|
|
else if (lock_type == TL_WRITE_DELAYED && data->cond)
|
|
|
|
{
|
2005-06-05 16:01:20 +02:00
|
|
|
/*
|
|
|
|
This only happens in extreme circumstances when a
|
|
|
|
write delayed lock that is waiting for a lock
|
|
|
|
*/
|
2000-07-31 21:29:14 +02:00
|
|
|
lock->write_wait.last=data->prev; /* Put it on wait queue */
|
|
|
|
}
|
|
|
|
else
|
|
|
|
lock->write.last=data->prev;
|
|
|
|
if (lock_type >= TL_WRITE_CONCURRENT_INSERT && lock->update_status)
|
|
|
|
(*lock->update_status)(data->status_param);
|
|
|
|
if (lock_type == TL_READ_NO_INSERT)
|
|
|
|
lock->read_no_write_count--;
|
|
|
|
data->type=TL_UNLOCK; /* Mark unlocked */
|
|
|
|
check_locks(lock,"after releasing lock",1);
|
|
|
|
|
2000-11-20 21:25:59 +01:00
|
|
|
if (!lock->write.data) /* If no active write locks */
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
data=lock->write_wait.data;
|
|
|
|
if (!lock->read.data) /* If no more locks in use */
|
|
|
|
{
|
|
|
|
/* Release write-locks with TL_WRITE or TL_WRITE_ONLY priority first */
|
|
|
|
if (data &&
|
|
|
|
(data->type != TL_WRITE_LOW_PRIORITY || !lock->read_wait.data ||
|
2001-02-17 23:03:37 +01:00
|
|
|
lock->read_wait.data->type < TL_READ_HIGH_PRIORITY))
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
if (lock->write_lock_count++ > max_write_lock_count)
|
|
|
|
{
|
|
|
|
/* Too many write locks in a row; Release all waiting read locks */
|
|
|
|
lock->write_lock_count=0;
|
|
|
|
if (lock->read_wait.data)
|
|
|
|
{
|
|
|
|
DBUG_PRINT("info",("Freeing all read_locks because of max_write_lock_count"));
|
|
|
|
free_all_read_locks(lock,0);
|
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
for (;;)
|
|
|
|
{
|
|
|
|
if (((*data->prev)=data->next)) /* remove from wait-list */
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else
|
|
|
|
lock->write_wait.last=data->prev;
|
|
|
|
(*lock->write.last)=data; /* Put in execute list */
|
|
|
|
data->prev=lock->write.last;
|
|
|
|
data->next=0;
|
|
|
|
lock->write.last= &data->next;
|
|
|
|
if (data->type == TL_WRITE_CONCURRENT_INSERT &&
|
|
|
|
(*lock->check_status)(data->status_param))
|
|
|
|
data->type=TL_WRITE; /* Upgrade lock */
|
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes
Changes that requires code changes in other code of other storage engines.
(Note that all changes are very straightforward and one should find all issues
by compiling a --debug build and fixing all compiler errors and all
asserts in field.cc while running the test suite),
- New optional handler function introduced: reset()
This is called after every DML statement to make it easy for a handler to
statement specific cleanups.
(The only case it's not called is if force the file to be closed)
- handler::extra(HA_EXTRA_RESET) is removed. Code that was there before
should be moved to handler::reset()
- table->read_set contains a bitmap over all columns that are needed
in the query. read_row() and similar functions only needs to read these
columns
- table->write_set contains a bitmap over all columns that will be updated
in the query. write_row() and update_row() only needs to update these
columns.
The above bitmaps should now be up to date in all context
(including ALTER TABLE, filesort()).
The handler is informed of any changes to the bitmap after
fix_fields() by calling the virtual function
handler::column_bitmaps_signal(). If the handler does caching of
these bitmaps (instead of using table->read_set, table->write_set),
it should redo the caching in this code. as the signal() may be sent
several times, it's probably best to set a variable in the signal
and redo the caching on read_row() / write_row() if the variable was
set.
- Removed the read_set and write_set bitmap objects from the handler class
- Removed all column bit handling functions from the handler class.
(Now one instead uses the normal bitmap functions in my_bitmap.c instead
of handler dedicated bitmap functions)
- field->query_id is removed. One should instead instead check
table->read_set and table->write_set if a field is used in the query.
- handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and
handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now
instead use table->read_set to check for which columns to retrieve.
- If a handler needs to call Field->val() or Field->store() on columns
that are not used in the query, one should install a temporary
all-columns-used map while doing so. For this, we provide the following
functions:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set);
field->val();
dbug_tmp_restore_column_map(table->read_set, old_map);
and similar for the write map:
my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set);
field->val();
dbug_tmp_restore_column_map(table->write_set, old_map);
If this is not done, you will sooner or later hit a DBUG_ASSERT
in the field store() / val() functions.
(For not DBUG binaries, the dbug_tmp_restore_column_map() and
dbug_tmp_restore_column_map() are inline dummy functions and should
be optimized away be the compiler).
- If one needs to temporary set the column map for all binaries (and not
just to avoid the DBUG_ASSERT() in the Field::store() / Field::val()
methods) one should use the functions tmp_use_all_columns() and
tmp_restore_column_map() instead of the above dbug_ variants.
- All 'status' fields in the handler base class (like records,
data_file_length etc) are now stored in a 'stats' struct. This makes
it easier to know what status variables are provided by the base
handler. This requires some trivial variable names in the extra()
function.
- New virtual function handler::records(). This is called to optimize
COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true.
(stats.records is not supposed to be an exact value. It's only has to
be 'reasonable enough' for the optimizer to be able to choose a good
optimization path).
- Non virtual handler::init() function added for caching of virtual
constants from engine.
- Removed has_transactions() virtual method. Now one should instead return
HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support
transactions.
- The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument
that is to be used with 'new handler_name()' to allocate the handler
in the right area. The xxxx_create_handler() function is also
responsible for any initialization of the object before returning.
For example, one should change:
static handler *myisam_create_handler(TABLE_SHARE *table)
{
return new ha_myisam(table);
}
->
static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root)
{
return new (mem_root) ha_myisam(table);
}
- New optional virtual function: use_hidden_primary_key().
This is called in case of an update/delete when
(table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined
but we don't have a primary key. This allows the handler to take precisions
in remembering any hidden primary key to able to update/delete any
found row. The default handler marks all columns to be read.
- handler::table_flags() now returns a ulonglong (to allow for more flags).
- New/changed table_flags()
- HA_HAS_RECORDS Set if ::records() is supported
- HA_NO_TRANSACTIONS Set if engine doesn't support transactions
- HA_PRIMARY_KEY_REQUIRED_FOR_DELETE
Set if we should mark all primary key columns for
read when reading rows as part of a DELETE
statement. If there is no primary key,
all columns are marked for read.
- HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some
cases (based on table->read_set)
- HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS
Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION.
- HA_DUPP_POS Renamed to HA_DUPLICATE_POS
- HA_REQUIRES_KEY_COLUMNS_FOR_DELETE
Set this if we should mark ALL key columns for
read when when reading rows as part of a DELETE
statement. In case of an update we will mark
all keys for read for which key part changed
value.
- HA_STATS_RECORDS_IS_EXACT
Set this if stats.records is exact.
(This saves us some extra records() calls
when optimizing COUNT(*))
- Removed table_flags()
- HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if
handler::records() gives an exact count() and
HA_STATS_RECORDS_IS_EXACT if stats.records is exact.
- HA_READ_RND_SAME Removed (no one supported this one)
- Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk()
- Renamed handler::dupp_pos to handler::dup_pos
- Removed not used variable handler::sortkey
Upper level handler changes:
- ha_reset() now does some overall checks and calls ::reset()
- ha_table_flags() added. This is a cached version of table_flags(). The
cache is updated on engine creation time and updated on open.
MySQL level changes (not obvious from the above):
- DBUG_ASSERT() added to check that column usage matches what is set
in the column usage bit maps. (This found a LOT of bugs in current
column marking code).
- In 5.1 before, all used columns was marked in read_set and only updated
columns was marked in write_set. Now we only mark columns for which we
need a value in read_set.
- Column bitmaps are created in open_binary_frm() and open_table_from_share().
(Before this was in table.cc)
- handler::table_flags() calls are replaced with handler::ha_table_flags()
- For calling field->val() you must have the corresponding bit set in
table->read_set. For calling field->store() you must have the
corresponding bit set in table->write_set. (There are asserts in
all store()/val() functions to catch wrong usage)
- thd->set_query_id is renamed to thd->mark_used_columns and instead
of setting this to an integer value, this has now the values:
MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE
Changed also all variables named 'set_query_id' to mark_used_columns.
- In filesort() we now inform the handler of exactly which columns are needed
doing the sort and choosing the rows.
- The TABLE_SHARE object has a 'all_set' column bitmap one can use
when one needs a column bitmap with all columns set.
(This is used for table->use_all_columns() and other places)
- The TABLE object has 3 column bitmaps:
- def_read_set Default bitmap for columns to be read
- def_write_set Default bitmap for columns to be written
- tmp_set Can be used as a temporary bitmap when needed.
The table object has also two pointer to bitmaps read_set and write_set
that the handler should use to find out which columns are used in which way.
- count() optimization now calls handler::records() instead of using
handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true).
- Added extra argument to Item::walk() to indicate if we should also
traverse sub queries.
- Added TABLE parameter to cp_buffer_from_ref()
- Don't close tables created with CREATE ... SELECT but keep them in
the table cache. (Faster usage of newly created tables).
New interfaces:
- table->clear_column_bitmaps() to initialize the bitmaps for tables
at start of new statements.
- table->column_bitmaps_set() to set up new column bitmaps and signal
the handler about this.
- table->column_bitmaps_set_no_signal() for some few cases where we need
to setup new column bitmaps but don't signal the handler (as the handler
has already been signaled about these before). Used for the momement
only in opt_range.cc when doing ROR scans.
- table->use_all_columns() to install a bitmap where all columns are marked
as use in the read and the write set.
- table->default_column_bitmaps() to install the normal read and write
column bitmaps, but not signaling the handler about this.
This is mainly used when creating TABLE instances.
- table->mark_columns_needed_for_delete(),
table->mark_columns_needed_for_delete() and
table->mark_columns_needed_for_insert() to allow us to put additional
columns in column usage maps if handler so requires.
(The handler indicates what it neads in handler->table_flags())
- table->prepare_for_position() to allow us to tell handler that it
needs to read primary key parts to be able to store them in
future table->position() calls.
(This replaces the table->file->ha_retrieve_all_pk function)
- table->mark_auto_increment_column() to tell handler are going to update
columns part of any auto_increment key.
- table->mark_columns_used_by_index() to mark all columns that is part of
an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow
it to quickly know that it only needs to read colums that are part
of the key. (The handler can also use the column map for detecting this,
but simpler/faster handler can just monitor the extra() call).
- table->mark_columns_used_by_index_no_reset() to in addition to other columns,
also mark all columns that is used by the given key.
- table->restore_column_maps_after_mark_index() to restore to default
column maps after a call to table->mark_columns_used_by_index().
- New item function register_field_in_read_map(), for marking used columns
in table->read_map. Used by filesort() to mark all used columns
- Maintain in TABLE->merge_keys set of all keys that are used in query.
(Simplices some optimization loops)
- Maintain Field->part_of_key_not_clustered which is like Field->part_of_key
but the field in the clustered key is not assumed to be part of all index.
(used in opt_range.cc for faster loops)
- dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map()
tmp_use_all_columns() and tmp_restore_column_map() functions to temporally
mark all columns as usable. The 'dbug_' version is primarily intended
inside a handler when it wants to just call Field:store() & Field::val()
functions, but don't need the column maps set for any other usage.
(ie:: bitmap_is_set() is never called)
- We can't use compare_records() to skip updates for handlers that returns
a partial column set and the read_set doesn't cover all columns in the
write set. The reason for this is that if we have a column marked only for
write we can't in the MySQL level know if the value changed or not.
The reason this worked before was that MySQL marked all to be written
columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden
bug'.
- open_table_from_share() does not anymore setup temporary MEM_ROOT
object as a thread specific variable for the handler. Instead we
send the to-be-used MEMROOT to get_new_handler().
(Simpler, faster code)
Bugs fixed:
- Column marking was not done correctly in a lot of cases.
(ALTER TABLE, when using triggers, auto_increment fields etc)
(Could potentially result in wrong values inserted in table handlers
relying on that the old column maps or field->set_query_id was correct)
Especially when it comes to triggers, there may be cases where the
old code would cause lost/wrong values for NDB and/or InnoDB tables.
- Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags:
OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG.
This allowed me to remove some wrong warnings about:
"Some non-transactional changed tables couldn't be rolled back"
- Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset
(thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose
some warnings about
"Some non-transactional changed tables couldn't be rolled back")
- Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table()
which could cause delete_table to report random failures.
- Fixed core dumps for some tests when running with --debug
- Added missing FN_LIBCHAR in mysql_rm_tmp_tables()
(This has probably caused us to not properly remove temporary files after
crash)
- slow_logs was not properly initialized, which could maybe cause
extra/lost entries in slow log.
- If we get an duplicate row on insert, change column map to read and
write all columns while retrying the operation. This is required by
the definition of REPLACE and also ensures that fields that are only
part of UPDATE are properly handled. This fixed a bug in NDB and
REPLACE where REPLACE wrongly copied some column values from the replaced
row.
- For table handler that doesn't support NULL in keys, we would give an error
when creating a primary key with NULL fields, even after the fields has been
automaticly converted to NOT NULL.
- Creating a primary key on a SPATIAL key, would fail if field was not
declared as NOT NULL.
Cleanups:
- Removed not used condition argument to setup_tables
- Removed not needed item function reset_query_id_processor().
- Field->add_index is removed. Now this is instead maintained in
(field->flags & FIELD_IN_ADD_INDEX)
- Field->fieldnr is removed (use field->field_index instead)
- New argument to filesort() to indicate that it should return a set of
row pointers (not used columns). This allowed me to remove some references
to sql_command in filesort and should also enable us to return column
results in some cases where we couldn't before.
- Changed column bitmap handling in opt_range.cc to be aligned with TABLE
bitmap, which allowed me to use bitmap functions instead of looping over
all fields to create some needed bitmaps. (Faster and smaller code)
- Broke up found too long lines
- Moved some variable declaration at start of function for better code
readability.
- Removed some not used arguments from functions.
(setup_fields(), mysql_prepare_insert_check_table())
- setup_fields() now takes an enum instead of an int for marking columns
usage.
- For internal temporary tables, use handler::write_row(),
handler::delete_row() and handler::update_row() instead of
handler::ha_xxxx() for faster execution.
- Changed some constants to enum's and define's.
- Using separate column read and write sets allows for easier checking
of timestamp field was set by statement.
- Remove calls to free_io_cache() as this is now done automaticly in ha_reset()
- Don't build table->normalized_path as this is now identical to table->path
(after bar's fixes to convert filenames)
- Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to
do comparision with the 'convert-dbug-for-diff' tool.
Things left to do in 5.1:
- We wrongly log failed CREATE TABLE ... SELECT in some cases when using
row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result)
Mats has promised to look into this.
- Test that my fix for CREATE TABLE ... SELECT is indeed correct.
(I added several test cases for this, but in this case it's better that
someone else also tests this throughly).
Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
|
|
|
DBUG_PRINT("lock",("giving write lock of type %d to thread: 0x%lx",
|
2005-07-19 20:21:12 +02:00
|
|
|
data->type, data->owner->info->thread_id));
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
pthread_cond_t *cond=data->cond;
|
|
|
|
data->cond=0; /* Mark thread free */
|
|
|
|
VOID(pthread_cond_signal(cond)); /* Start waiting thread */
|
|
|
|
}
|
|
|
|
if (data->type != TL_WRITE_ALLOW_WRITE ||
|
|
|
|
!lock->write_wait.data ||
|
|
|
|
lock->write_wait.data->type != TL_WRITE_ALLOW_WRITE)
|
|
|
|
break;
|
|
|
|
data=lock->write_wait.data; /* Free this too */
|
|
|
|
}
|
|
|
|
if (data->type >= TL_WRITE_LOW_PRIORITY)
|
|
|
|
{
|
|
|
|
check_locks(lock,"giving write lock",0);
|
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
|
|
|
DBUG_VOID_RETURN;
|
|
|
|
}
|
|
|
|
/* Release possible read locks together with the write lock */
|
|
|
|
}
|
|
|
|
if (lock->read_wait.data)
|
|
|
|
free_all_read_locks(lock,
|
|
|
|
data &&
|
|
|
|
(data->type == TL_WRITE_CONCURRENT_INSERT ||
|
|
|
|
data->type == TL_WRITE_ALLOW_WRITE));
|
|
|
|
else
|
|
|
|
{
|
2001-09-08 10:47:34 +02:00
|
|
|
DBUG_PRINT("lock",("No waiting read locks to free"));
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
else if (data &&
|
|
|
|
(lock_type=data->type) <= TL_WRITE_DELAYED &&
|
|
|
|
((lock_type != TL_WRITE_CONCURRENT_INSERT &&
|
|
|
|
lock_type != TL_WRITE_ALLOW_WRITE) ||
|
|
|
|
!lock->read_no_write_count))
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
For DELAYED, ALLOW_READ, WRITE_ALLOW_WRITE or CONCURRENT_INSERT locks
|
|
|
|
start WRITE locks together with the READ locks
|
|
|
|
*/
|
|
|
|
if (lock_type == TL_WRITE_CONCURRENT_INSERT &&
|
|
|
|
(*lock->check_status)(data->status_param))
|
|
|
|
{
|
|
|
|
data->type=TL_WRITE; /* Upgrade lock */
|
|
|
|
if (lock->read_wait.data)
|
|
|
|
free_all_read_locks(lock,0);
|
|
|
|
goto end;
|
|
|
|
}
|
|
|
|
do {
|
|
|
|
pthread_cond_t *cond=data->cond;
|
|
|
|
if (((*data->prev)=data->next)) /* remove from wait-list */
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else
|
|
|
|
lock->write_wait.last=data->prev;
|
2000-11-20 21:25:59 +01:00
|
|
|
(*lock->write.last)=data; /* Put in execute list */
|
2000-07-31 21:29:14 +02:00
|
|
|
data->prev=lock->write.last;
|
|
|
|
lock->write.last= &data->next;
|
|
|
|
data->next=0; /* Only one write lock */
|
|
|
|
data->cond=0; /* Mark thread free */
|
|
|
|
VOID(pthread_cond_signal(cond)); /* Start waiting thread */
|
|
|
|
} while (lock_type == TL_WRITE_ALLOW_WRITE &&
|
|
|
|
(data=lock->write_wait.data) &&
|
|
|
|
data->type == TL_WRITE_ALLOW_WRITE);
|
|
|
|
if (lock->read_wait.data)
|
|
|
|
free_all_read_locks(lock,
|
|
|
|
(lock_type == TL_WRITE_CONCURRENT_INSERT ||
|
|
|
|
lock_type == TL_WRITE_ALLOW_WRITE));
|
|
|
|
}
|
2000-11-20 21:25:59 +01:00
|
|
|
else if (!data && lock->read_wait.data)
|
2000-07-31 21:29:14 +02:00
|
|
|
free_all_read_locks(lock,0);
|
|
|
|
}
|
|
|
|
end:
|
|
|
|
check_locks(lock,"thr_unlock",0);
|
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
|
|
|
DBUG_VOID_RETURN;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
** Get all locks in a specific order to avoid dead-locks
|
|
|
|
** Sort acording to lock position and put write_locks before read_locks if
|
|
|
|
** lock on same lock.
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
|
|
#define LOCK_CMP(A,B) ((byte*) (A->lock) - (uint) ((A)->type) < (byte*) (B->lock)- (uint) ((B)->type))
|
|
|
|
|
|
|
|
static void sort_locks(THR_LOCK_DATA **data,uint count)
|
|
|
|
{
|
|
|
|
THR_LOCK_DATA **pos,**end,**prev,*tmp;
|
|
|
|
|
|
|
|
/* Sort locks with insertion sort (fast because almost always few locks) */
|
|
|
|
|
|
|
|
for (pos=data+1,end=data+count; pos < end ; pos++)
|
|
|
|
{
|
|
|
|
tmp= *pos;
|
|
|
|
if (LOCK_CMP(tmp,pos[-1]))
|
|
|
|
{
|
|
|
|
prev=pos;
|
|
|
|
do {
|
|
|
|
prev[0]=prev[-1];
|
|
|
|
} while (--prev != data && LOCK_CMP(tmp,prev[-1]));
|
|
|
|
prev[0]=tmp;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2005-07-19 20:21:12 +02:00
|
|
|
enum enum_thr_lock_result
|
|
|
|
thr_multi_lock(THR_LOCK_DATA **data, uint count, THR_LOCK_OWNER *owner)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
THR_LOCK_DATA **pos,**end;
|
|
|
|
DBUG_ENTER("thr_multi_lock");
|
2004-08-23 12:46:51 +02:00
|
|
|
DBUG_PRINT("lock",("data: 0x%lx count: %d",data,count));
|
2000-07-31 21:29:14 +02:00
|
|
|
if (count > 1)
|
|
|
|
sort_locks(data,count);
|
|
|
|
/* lock everything */
|
|
|
|
for (pos=data,end=data+count; pos < end ; pos++)
|
|
|
|
{
|
2005-07-19 20:21:12 +02:00
|
|
|
enum enum_thr_lock_result result= thr_lock(*pos, owner, (*pos)->type);
|
|
|
|
if (result != THR_LOCK_SUCCESS)
|
2000-07-31 21:29:14 +02:00
|
|
|
{ /* Aborted */
|
|
|
|
thr_multi_unlock(data,(uint) (pos-data));
|
2005-07-19 20:21:12 +02:00
|
|
|
DBUG_RETURN(result);
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
#ifdef MAIN
|
2004-08-23 12:46:51 +02:00
|
|
|
printf("Thread: %s Got lock: 0x%lx type: %d\n",my_thread_name(),
|
2000-07-31 21:29:14 +02:00
|
|
|
(long) pos[0]->lock, pos[0]->type); fflush(stdout);
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
Ensure that all get_locks() have the same status
|
|
|
|
If we lock the same table multiple times, we must use the same
|
|
|
|
status_param!
|
|
|
|
*/
|
|
|
|
#if !defined(DONT_USE_RW_LOCKS)
|
|
|
|
if (count > 1)
|
|
|
|
{
|
|
|
|
THR_LOCK_DATA *last_lock= end[-1];
|
|
|
|
pos=end-1;
|
|
|
|
do
|
|
|
|
{
|
|
|
|
pos--;
|
|
|
|
if (last_lock->lock == (*pos)->lock &&
|
|
|
|
last_lock->lock->copy_status)
|
|
|
|
{
|
|
|
|
if (last_lock->type <= TL_READ_NO_INSERT)
|
|
|
|
{
|
|
|
|
THR_LOCK_DATA **read_lock;
|
|
|
|
/*
|
|
|
|
If we are locking the same table with read locks we must ensure
|
|
|
|
that all tables share the status of the last write lock or
|
|
|
|
the same read lock.
|
|
|
|
*/
|
|
|
|
for (;
|
|
|
|
(*pos)->type <= TL_READ_NO_INSERT &&
|
|
|
|
pos != data &&
|
|
|
|
pos[-1]->lock == (*pos)->lock ;
|
|
|
|
pos--) ;
|
|
|
|
|
|
|
|
read_lock = pos+1;
|
|
|
|
do
|
|
|
|
{
|
|
|
|
(last_lock->lock->copy_status)((*read_lock)->status_param,
|
|
|
|
(*pos)->status_param);
|
|
|
|
} while (*(read_lock++) != last_lock);
|
|
|
|
last_lock= (*pos); /* Point at last write lock */
|
|
|
|
}
|
|
|
|
else
|
|
|
|
(*last_lock->lock->copy_status)((*pos)->status_param,
|
|
|
|
last_lock->status_param);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
last_lock=(*pos);
|
|
|
|
} while (pos != data);
|
|
|
|
}
|
|
|
|
#endif
|
2005-07-19 20:21:12 +02:00
|
|
|
DBUG_RETURN(THR_LOCK_SUCCESS);
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* free all locks */
|
|
|
|
|
|
|
|
void thr_multi_unlock(THR_LOCK_DATA **data,uint count)
|
|
|
|
{
|
|
|
|
THR_LOCK_DATA **pos,**end;
|
|
|
|
DBUG_ENTER("thr_multi_unlock");
|
2004-08-23 12:46:51 +02:00
|
|
|
DBUG_PRINT("lock",("data: 0x%lx count: %d",data,count));
|
2000-07-31 21:29:14 +02:00
|
|
|
|
|
|
|
for (pos=data,end=data+count; pos < end ; pos++)
|
|
|
|
{
|
|
|
|
#ifdef MAIN
|
2004-08-23 12:46:51 +02:00
|
|
|
printf("Thread: %s Rel lock: 0x%lx type: %d\n",
|
2000-07-31 21:29:14 +02:00
|
|
|
my_thread_name(), (long) pos[0]->lock, pos[0]->type);
|
|
|
|
fflush(stdout);
|
|
|
|
#endif
|
|
|
|
if ((*pos)->type != TL_UNLOCK)
|
|
|
|
thr_unlock(*pos);
|
|
|
|
else
|
|
|
|
{
|
2004-08-23 12:46:51 +02:00
|
|
|
DBUG_PRINT("lock",("Free lock: data: 0x%lx thread: %ld lock: 0x%lx",
|
2005-07-19 20:21:12 +02:00
|
|
|
*pos, (*pos)->owner->info->thread_id, (*pos)->lock));
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
DBUG_VOID_RETURN;
|
|
|
|
}
|
|
|
|
|
2001-11-26 01:16:38 +01:00
|
|
|
/*
|
|
|
|
Abort all threads waiting for a lock. The lock will be upgraded to
|
2000-07-31 21:29:14 +02:00
|
|
|
TL_WRITE_ONLY to abort any new accesses to the lock
|
|
|
|
*/
|
|
|
|
|
2006-01-17 08:40:00 +01:00
|
|
|
void thr_abort_locks(THR_LOCK *lock, bool upgrade_lock)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
THR_LOCK_DATA *data;
|
|
|
|
DBUG_ENTER("thr_abort_locks");
|
|
|
|
pthread_mutex_lock(&lock->mutex);
|
|
|
|
|
|
|
|
for (data=lock->read_wait.data; data ; data=data->next)
|
|
|
|
{
|
|
|
|
data->type=TL_UNLOCK; /* Mark killed */
|
2005-07-19 20:21:12 +02:00
|
|
|
/* It's safe to signal the cond first: we're still holding the mutex. */
|
2000-07-31 21:29:14 +02:00
|
|
|
pthread_cond_signal(data->cond);
|
|
|
|
data->cond=0; /* Removed from list */
|
|
|
|
}
|
|
|
|
for (data=lock->write_wait.data; data ; data=data->next)
|
|
|
|
{
|
|
|
|
data->type=TL_UNLOCK;
|
|
|
|
pthread_cond_signal(data->cond);
|
|
|
|
data->cond=0;
|
|
|
|
}
|
|
|
|
lock->read_wait.last= &lock->read_wait.data;
|
|
|
|
lock->write_wait.last= &lock->write_wait.data;
|
|
|
|
lock->read_wait.data=lock->write_wait.data=0;
|
2006-01-17 08:40:00 +01:00
|
|
|
if (upgrade_lock && lock->write.data)
|
2000-07-31 21:29:14 +02:00
|
|
|
lock->write.data->type=TL_WRITE_ONLY;
|
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
|
|
|
DBUG_VOID_RETURN;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2003-03-04 11:22:35 +01:00
|
|
|
/*
|
|
|
|
Abort all locks for specific table/thread combination
|
|
|
|
|
|
|
|
This is used to abort all locks for a specific thread
|
|
|
|
*/
|
|
|
|
|
2005-07-26 16:55:58 +02:00
|
|
|
my_bool thr_abort_locks_for_thread(THR_LOCK *lock, pthread_t thread)
|
2003-03-04 11:22:35 +01:00
|
|
|
{
|
|
|
|
THR_LOCK_DATA *data;
|
2005-07-26 16:55:58 +02:00
|
|
|
my_bool found= FALSE;
|
2003-03-04 11:22:35 +01:00
|
|
|
DBUG_ENTER("thr_abort_locks_for_thread");
|
|
|
|
|
|
|
|
pthread_mutex_lock(&lock->mutex);
|
|
|
|
for (data= lock->read_wait.data; data ; data= data->next)
|
|
|
|
{
|
2005-07-19 20:21:12 +02:00
|
|
|
if (pthread_equal(thread, data->owner->info->thread))
|
2003-03-04 11:22:35 +01:00
|
|
|
{
|
|
|
|
DBUG_PRINT("info",("Aborting read-wait lock"));
|
|
|
|
data->type= TL_UNLOCK; /* Mark killed */
|
2005-07-19 20:21:12 +02:00
|
|
|
/* It's safe to signal the cond first: we're still holding the mutex. */
|
2005-07-19 00:29:19 +02:00
|
|
|
found= TRUE;
|
2003-03-04 11:22:35 +01:00
|
|
|
pthread_cond_signal(data->cond);
|
|
|
|
data->cond= 0; /* Removed from list */
|
|
|
|
|
|
|
|
if (((*data->prev)= data->next))
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else
|
|
|
|
lock->read_wait.last= data->prev;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
for (data= lock->write_wait.data; data ; data= data->next)
|
|
|
|
{
|
2005-07-19 20:21:12 +02:00
|
|
|
if (pthread_equal(thread, data->owner->info->thread))
|
2003-03-04 11:22:35 +01:00
|
|
|
{
|
|
|
|
DBUG_PRINT("info",("Aborting write-wait lock"));
|
|
|
|
data->type= TL_UNLOCK;
|
2005-07-19 00:29:19 +02:00
|
|
|
found= TRUE;
|
2003-03-04 11:22:35 +01:00
|
|
|
pthread_cond_signal(data->cond);
|
|
|
|
data->cond= 0;
|
|
|
|
|
|
|
|
if (((*data->prev)= data->next))
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else
|
|
|
|
lock->write_wait.last= data->prev;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
2005-07-19 00:29:19 +02:00
|
|
|
DBUG_RETURN(found);
|
2003-03-04 11:22:35 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2006-01-17 08:40:00 +01:00
|
|
|
/*
|
|
|
|
Downgrade a WRITE_* to a lower WRITE level
|
|
|
|
SYNOPSIS
|
|
|
|
thr_downgrade_write_lock()
|
|
|
|
in_data Lock data of thread downgrading its lock
|
|
|
|
new_lock_type New write lock type
|
|
|
|
RETURN VALUE
|
|
|
|
NONE
|
|
|
|
DESCRIPTION
|
|
|
|
This can be used to downgrade a lock already owned. When the downgrade
|
|
|
|
occurs also other waiters, both readers and writers can be allowed to
|
|
|
|
start.
|
|
|
|
The previous lock is often TL_WRITE_ONLY but can also be
|
|
|
|
TL_WRITE and TL_WRITE_ALLOW_READ. The normal downgrade variants are
|
|
|
|
TL_WRITE_ONLY => TL_WRITE_ALLOW_READ After a short exclusive lock
|
|
|
|
TL_WRITE_ALLOW_READ => TL_WRITE_ALLOW_WRITE After discovering that the
|
|
|
|
operation didn't need such a high lock.
|
|
|
|
TL_WRITE_ONLY => TL_WRITE after a short exclusive lock while holding a
|
|
|
|
write table lock
|
|
|
|
TL_WRITE_ONLY => TL_WRITE_ALLOW_WRITE After a short exclusive lock after
|
|
|
|
already earlier having dongraded lock to TL_WRITE_ALLOW_WRITE
|
|
|
|
The implementation is conservative and rather don't start rather than
|
|
|
|
go on unknown paths to start, the common cases are handled.
|
|
|
|
|
|
|
|
NOTE:
|
|
|
|
In its current implementation it is only allowed to downgrade from
|
|
|
|
TL_WRITE_ONLY. In this case there are no waiters. Thus no wake up
|
|
|
|
logic is required.
|
|
|
|
*/
|
|
|
|
|
|
|
|
void thr_downgrade_write_lock(THR_LOCK_DATA *in_data,
|
|
|
|
enum thr_lock_type new_lock_type)
|
|
|
|
{
|
|
|
|
THR_LOCK *lock=in_data->lock;
|
|
|
|
enum thr_lock_type old_lock_type= in_data->type;
|
2006-06-23 01:49:19 +02:00
|
|
|
#ifdef TO_BE_REMOVED
|
|
|
|
THR_LOCK_DATA *data, *next;
|
2006-01-17 08:40:00 +01:00
|
|
|
bool start_writers= FALSE;
|
|
|
|
bool start_readers= FALSE;
|
2006-06-23 01:49:19 +02:00
|
|
|
#endif
|
2006-01-17 08:40:00 +01:00
|
|
|
DBUG_ENTER("thr_downgrade_write_only_lock");
|
|
|
|
|
|
|
|
pthread_mutex_lock(&lock->mutex);
|
|
|
|
DBUG_ASSERT(old_lock_type == TL_WRITE_ONLY);
|
|
|
|
DBUG_ASSERT(old_lock_type > new_lock_type);
|
|
|
|
in_data->type= new_lock_type;
|
|
|
|
check_locks(lock,"after downgrading lock",0);
|
2006-06-23 01:49:19 +02:00
|
|
|
|
|
|
|
#if TO_BE_REMOVED
|
2006-01-17 08:40:00 +01:00
|
|
|
switch (old_lock_type)
|
|
|
|
{
|
|
|
|
case TL_WRITE_ONLY:
|
|
|
|
case TL_WRITE:
|
|
|
|
case TL_WRITE_LOW_PRIORITY:
|
|
|
|
/*
|
|
|
|
Previous lock was exclusive we are now ready to start up most waiting
|
|
|
|
threads.
|
|
|
|
*/
|
|
|
|
switch (new_lock_type)
|
|
|
|
{
|
|
|
|
case TL_WRITE_ALLOW_READ:
|
|
|
|
/* Still cannot start WRITE operations. Can only start readers. */
|
|
|
|
start_readers= TRUE;
|
|
|
|
break;
|
|
|
|
case TL_WRITE:
|
|
|
|
case TL_WRITE_LOW_PRIORITY:
|
|
|
|
/*
|
|
|
|
Still cannot start anything, but new requests are no longer
|
|
|
|
aborted.
|
|
|
|
*/
|
|
|
|
break;
|
|
|
|
case TL_WRITE_ALLOW_WRITE:
|
|
|
|
/*
|
|
|
|
We can start both writers and readers.
|
|
|
|
*/
|
|
|
|
start_writers= TRUE;
|
|
|
|
start_readers= TRUE;
|
|
|
|
break;
|
|
|
|
case TL_WRITE_CONCURRENT_INSERT:
|
|
|
|
case TL_WRITE_DELAYED:
|
|
|
|
/*
|
|
|
|
This routine is not designed for those. Lock will be downgraded
|
|
|
|
but no start of waiters will occur. This is not the optimal but
|
|
|
|
should be a correct behaviour.
|
|
|
|
*/
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
DBUG_ASSERT(0);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case TL_WRITE_DELAYED:
|
|
|
|
case TL_WRITE_CONCURRENT_INSERT:
|
|
|
|
/*
|
|
|
|
This routine is not designed for those. Lock will be downgraded
|
|
|
|
but no start of waiters will occur. This is not the optimal but
|
|
|
|
should be a correct behaviour.
|
|
|
|
*/
|
|
|
|
break;
|
|
|
|
case TL_WRITE_ALLOW_READ:
|
|
|
|
DBUG_ASSERT(new_lock_type == TL_WRITE_ALLOW_WRITE);
|
|
|
|
/*
|
|
|
|
Previously writers were not allowed to start, now it is ok to
|
|
|
|
start them again. Readers are already allowed so no reason to
|
|
|
|
handle them.
|
|
|
|
*/
|
|
|
|
start_writers= TRUE;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
DBUG_ASSERT(0);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (start_writers)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
At this time the only active writer can be ourselves. Thus we need
|
|
|
|
not worry about that there are other concurrent write operations
|
|
|
|
active on the table. Thus we only need to worry about starting
|
|
|
|
waiting operations.
|
|
|
|
We also only come here with TL_WRITE_ALLOW_WRITE as the new
|
|
|
|
lock type, thus we can start other writers also of the same type.
|
|
|
|
If we find a lock at exclusive level >= TL_WRITE_LOW_PRIORITY we
|
|
|
|
don't start any more operations that would be mean those operations
|
|
|
|
will have to wait for things started afterwards.
|
|
|
|
*/
|
|
|
|
DBUG_ASSERT(new_lock_type == TL_WRITE_ALLOW_WRITE);
|
|
|
|
for (data=lock->write_wait.data; data ; data= next)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
All WRITE requests compatible with new lock type are also
|
|
|
|
started
|
|
|
|
*/
|
|
|
|
next= data->next;
|
|
|
|
if (start_writers && data->type == new_lock_type)
|
|
|
|
{
|
|
|
|
pthread_cond_t *cond= data->cond;
|
|
|
|
/*
|
|
|
|
It is ok to start this waiter.
|
|
|
|
Move from being first in wait queue to be last in write queue.
|
|
|
|
*/
|
|
|
|
if (((*data->prev)= data->next))
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else
|
|
|
|
lock->write_wait.last= data->prev;
|
|
|
|
data->prev= lock->write.last;
|
|
|
|
lock->write.last= &data->next;
|
|
|
|
data->next= 0;
|
|
|
|
check_locks(lock, "Started write lock after downgrade",0);
|
|
|
|
data->cond= 0;
|
|
|
|
pthread_cond_signal(cond);
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
We found an incompatible lock, we won't start any more write
|
|
|
|
requests to avoid letting writers pass other writers in the
|
|
|
|
queue.
|
|
|
|
*/
|
|
|
|
start_writers= FALSE;
|
|
|
|
if (data->type >= TL_WRITE_LOW_PRIORITY)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
We have an exclusive writer in the queue so we won't start
|
|
|
|
readers either.
|
|
|
|
*/
|
|
|
|
start_readers= FALSE;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (start_readers)
|
|
|
|
{
|
|
|
|
DBUG_ASSERT(new_lock_type == TL_WRITE_ALLOW_WRITE ||
|
|
|
|
new_lock_type == TL_WRITE_ALLOW_READ);
|
|
|
|
/*
|
|
|
|
When we come here we know that the write locks are
|
|
|
|
TL_WRITE_ALLOW_WRITE or TL_WRITE_ALLOW_READ. This means that reads
|
|
|
|
are ok
|
|
|
|
*/
|
|
|
|
for (data=lock->read_wait.data; data ; data=next)
|
|
|
|
{
|
|
|
|
next= data->next;
|
|
|
|
/*
|
|
|
|
All reads are ok to start now except TL_READ_NO_INSERT when
|
|
|
|
write lock is TL_WRITE_ALLOW_READ.
|
|
|
|
*/
|
|
|
|
if (new_lock_type != TL_WRITE_ALLOW_READ ||
|
|
|
|
data->type != TL_READ_NO_INSERT)
|
|
|
|
{
|
|
|
|
pthread_cond_t *cond= data->cond;
|
|
|
|
if (((*data->prev)= data->next))
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else
|
|
|
|
lock->read_wait.last= data->prev;
|
|
|
|
data->prev= lock->read.last;
|
|
|
|
lock->read.last= &data->next;
|
|
|
|
data->next= 0;
|
|
|
|
|
|
|
|
if (data->type == TL_READ_NO_INSERT)
|
|
|
|
lock->read_no_write_count++;
|
|
|
|
check_locks(lock, "Started read lock after downgrade",0);
|
|
|
|
data->cond= 0;
|
|
|
|
pthread_cond_signal(cond);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
check_locks(lock,"after starting waiters after downgrading lock",0);
|
|
|
|
#endif
|
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
|
|
|
DBUG_VOID_RETURN;
|
|
|
|
}
|
2003-03-04 11:22:35 +01:00
|
|
|
|
2000-07-31 21:29:14 +02:00
|
|
|
/* Upgrade a WRITE_DELAY lock to a WRITE_LOCK */
|
|
|
|
|
|
|
|
my_bool thr_upgrade_write_delay_lock(THR_LOCK_DATA *data)
|
|
|
|
{
|
|
|
|
THR_LOCK *lock=data->lock;
|
|
|
|
DBUG_ENTER("thr_upgrade_write_delay_lock");
|
|
|
|
|
|
|
|
pthread_mutex_lock(&lock->mutex);
|
2001-07-18 22:34:04 +02:00
|
|
|
if (data->type == TL_UNLOCK || data->type >= TL_WRITE_LOW_PRIORITY)
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
2001-07-18 22:34:04 +02:00
|
|
|
DBUG_RETURN(data->type == TL_UNLOCK); /* Test if Aborted */
|
2000-07-31 21:29:14 +02:00
|
|
|
}
|
|
|
|
check_locks(lock,"before upgrading lock",0);
|
|
|
|
/* TODO: Upgrade to TL_WRITE_CONCURRENT_INSERT in some cases */
|
|
|
|
data->type=TL_WRITE; /* Upgrade lock */
|
|
|
|
|
|
|
|
/* Check if someone has given us the lock */
|
|
|
|
if (!data->cond)
|
|
|
|
{
|
|
|
|
if (!lock->read.data) /* No read locks */
|
|
|
|
{ /* We have the lock */
|
|
|
|
if (data->lock->get_status)
|
2005-05-13 11:08:08 +02:00
|
|
|
(*data->lock->get_status)(data->status_param, 0);
|
2000-07-31 21:29:14 +02:00
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
|
|
|
DBUG_RETURN(0);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (((*data->prev)=data->next)) /* remove from lock-list */
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else
|
|
|
|
lock->write.last=data->prev;
|
|
|
|
|
|
|
|
if ((data->next=lock->write_wait.data)) /* Put first in lock_list */
|
|
|
|
data->next->prev= &data->next;
|
|
|
|
else
|
|
|
|
lock->write_wait.last= &data->next;
|
|
|
|
data->prev= &lock->write_wait.data;
|
|
|
|
lock->write_wait.data=data;
|
|
|
|
check_locks(lock,"upgrading lock",0);
|
|
|
|
}
|
2001-08-31 22:02:09 +02:00
|
|
|
else
|
|
|
|
{
|
|
|
|
check_locks(lock,"waiting for lock",0);
|
|
|
|
}
|
2000-07-31 21:29:14 +02:00
|
|
|
DBUG_RETURN(wait_for_lock(&lock->write_wait,data,1));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* downgrade a WRITE lock to a WRITE_DELAY lock if there is pending locks */
|
|
|
|
|
|
|
|
my_bool thr_reschedule_write_lock(THR_LOCK_DATA *data)
|
|
|
|
{
|
|
|
|
THR_LOCK *lock=data->lock;
|
|
|
|
DBUG_ENTER("thr_reschedule_write_lock");
|
|
|
|
|
|
|
|
pthread_mutex_lock(&lock->mutex);
|
|
|
|
if (!lock->read_wait.data) /* No waiting read locks */
|
|
|
|
{
|
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
|
|
|
DBUG_RETURN(0);
|
|
|
|
}
|
|
|
|
|
|
|
|
data->type=TL_WRITE_DELAYED;
|
|
|
|
if (lock->update_status)
|
|
|
|
(*lock->update_status)(data->status_param);
|
|
|
|
if (((*data->prev)=data->next)) /* remove from lock-list */
|
|
|
|
data->next->prev= data->prev;
|
|
|
|
else
|
|
|
|
lock->write.last=data->prev;
|
|
|
|
|
|
|
|
if ((data->next=lock->write_wait.data)) /* Put first in lock_list */
|
|
|
|
data->next->prev= &data->next;
|
|
|
|
else
|
|
|
|
lock->write_wait.last= &data->next;
|
|
|
|
data->prev= &lock->write_wait.data;
|
|
|
|
data->cond=get_cond(); /* This was zero */
|
|
|
|
lock->write_wait.data=data;
|
|
|
|
free_all_read_locks(lock,0);
|
|
|
|
|
|
|
|
pthread_mutex_unlock(&lock->mutex);
|
|
|
|
DBUG_RETURN(thr_upgrade_write_delay_lock(data));
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
#include <my_sys.h>
|
|
|
|
|
|
|
|
static void thr_print_lock(const char* name,struct st_lock_list *list)
|
|
|
|
{
|
|
|
|
THR_LOCK_DATA *data,**prev;
|
|
|
|
uint count=0;
|
|
|
|
|
|
|
|
if (list->data)
|
|
|
|
{
|
|
|
|
printf("%-10s: ",name);
|
|
|
|
prev= &list->data;
|
|
|
|
for (data=list->data; data && count++ < MAX_LOCKS ; data=data->next)
|
|
|
|
{
|
2005-07-19 20:21:12 +02:00
|
|
|
printf("0x%lx (%lu:%d); ", (ulong) data, data->owner->info->thread_id,
|
|
|
|
(int) data->type);
|
2000-07-31 21:29:14 +02:00
|
|
|
if (data->prev != prev)
|
|
|
|
printf("\nWarning: prev didn't point at previous lock\n");
|
|
|
|
prev= &data->next;
|
|
|
|
}
|
|
|
|
puts("");
|
|
|
|
if (prev != list->last)
|
|
|
|
printf("Warning: last didn't point at last lock\n");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void thr_print_locks(void)
|
|
|
|
{
|
|
|
|
LIST *list;
|
|
|
|
uint count=0;
|
|
|
|
|
|
|
|
pthread_mutex_lock(&THR_LOCK_lock);
|
|
|
|
puts("Current locks:");
|
2005-08-10 03:02:36 +02:00
|
|
|
for (list= thr_lock_thread_list; list && count++ < MAX_THREADS;
|
|
|
|
list= list_rest(list))
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
THR_LOCK *lock=(THR_LOCK*) list->data;
|
|
|
|
VOID(pthread_mutex_lock(&lock->mutex));
|
2004-08-23 12:46:51 +02:00
|
|
|
printf("lock: 0x%lx:",(ulong) lock);
|
2000-07-31 21:29:14 +02:00
|
|
|
if ((lock->write_wait.data || lock->read_wait.data) &&
|
|
|
|
(! lock->read.data && ! lock->write.data))
|
|
|
|
printf(" WARNING: ");
|
|
|
|
if (lock->write.data)
|
|
|
|
printf(" write");
|
|
|
|
if (lock->write_wait.data)
|
|
|
|
printf(" write_wait");
|
|
|
|
if (lock->read.data)
|
|
|
|
printf(" read");
|
|
|
|
if (lock->read_wait.data)
|
|
|
|
printf(" read_wait");
|
|
|
|
puts("");
|
|
|
|
thr_print_lock("write",&lock->write);
|
|
|
|
thr_print_lock("write_wait",&lock->write_wait);
|
|
|
|
thr_print_lock("read",&lock->read);
|
|
|
|
thr_print_lock("read_wait",&lock->read_wait);
|
|
|
|
VOID(pthread_mutex_unlock(&lock->mutex));
|
|
|
|
puts("");
|
|
|
|
}
|
|
|
|
fflush(stdout);
|
|
|
|
pthread_mutex_unlock(&THR_LOCK_lock);
|
|
|
|
}
|
|
|
|
|
2000-08-21 23:18:32 +02:00
|
|
|
#endif /* THREAD */
|
|
|
|
|
|
|
|
/*****************************************************************************
|
|
|
|
** Test of thread locks
|
|
|
|
****************************************************************************/
|
|
|
|
|
2000-07-31 21:29:14 +02:00
|
|
|
#ifdef MAIN
|
|
|
|
|
2000-08-21 23:18:32 +02:00
|
|
|
#ifdef THREAD
|
|
|
|
|
2000-07-31 21:29:14 +02:00
|
|
|
struct st_test {
|
|
|
|
uint lock_nr;
|
|
|
|
enum thr_lock_type lock_type;
|
|
|
|
};
|
|
|
|
|
|
|
|
THR_LOCK locks[5]; /* 4 locks */
|
|
|
|
|
|
|
|
struct st_test test_0[] = {{0,TL_READ}}; /* One lock */
|
|
|
|
struct st_test test_1[] = {{0,TL_READ},{0,TL_WRITE}}; /* Read and write lock of lock 0 */
|
|
|
|
struct st_test test_2[] = {{1,TL_WRITE},{0,TL_READ},{2,TL_READ}};
|
|
|
|
struct st_test test_3[] = {{2,TL_WRITE},{1,TL_READ},{0,TL_READ}}; /* Deadlock with test_2 ? */
|
|
|
|
struct st_test test_4[] = {{0,TL_WRITE},{0,TL_READ},{0,TL_WRITE},{0,TL_READ}};
|
|
|
|
struct st_test test_5[] = {{0,TL_READ},{1,TL_READ},{2,TL_READ},{3,TL_READ}}; /* Many reads */
|
|
|
|
struct st_test test_6[] = {{0,TL_WRITE},{1,TL_WRITE},{2,TL_WRITE},{3,TL_WRITE}}; /* Many writes */
|
|
|
|
struct st_test test_7[] = {{3,TL_READ}};
|
|
|
|
struct st_test test_8[] = {{1,TL_READ_NO_INSERT},{2,TL_READ_NO_INSERT},{3,TL_READ_NO_INSERT}}; /* Should be quick */
|
|
|
|
struct st_test test_9[] = {{4,TL_READ_HIGH_PRIORITY}};
|
|
|
|
struct st_test test_10[] ={{4,TL_WRITE}};
|
|
|
|
struct st_test test_11[] = {{0,TL_WRITE_LOW_PRIORITY},{1,TL_WRITE_LOW_PRIORITY},{2,TL_WRITE_LOW_PRIORITY},{3,TL_WRITE_LOW_PRIORITY}}; /* Many writes */
|
|
|
|
struct st_test test_12[] = {{0,TL_WRITE_ALLOW_READ},{1,TL_WRITE_ALLOW_READ},{2,TL_WRITE_ALLOW_READ},{3,TL_WRITE_ALLOW_READ}}; /* Many writes */
|
|
|
|
struct st_test test_13[] = {{0,TL_WRITE_CONCURRENT_INSERT},{1,TL_WRITE_CONCURRENT_INSERT},{2,TL_WRITE_CONCURRENT_INSERT},{3,TL_WRITE_CONCURRENT_INSERT}};
|
|
|
|
struct st_test test_14[] = {{0,TL_WRITE_CONCURRENT_INSERT},{1,TL_READ}};
|
|
|
|
struct st_test test_15[] = {{0,TL_WRITE_ALLOW_WRITE},{1,TL_READ}};
|
|
|
|
struct st_test test_16[] = {{0,TL_WRITE_ALLOW_WRITE},{1,TL_WRITE_ALLOW_WRITE}};
|
|
|
|
|
|
|
|
struct st_test *tests[] = {test_0,test_1,test_2,test_3,test_4,test_5,test_6,
|
|
|
|
test_7,test_8,test_9,test_10,test_11,test_12,
|
|
|
|
test_13,test_14,test_15,test_16};
|
|
|
|
int lock_counts[]= {sizeof(test_0)/sizeof(struct st_test),
|
|
|
|
sizeof(test_1)/sizeof(struct st_test),
|
|
|
|
sizeof(test_2)/sizeof(struct st_test),
|
|
|
|
sizeof(test_3)/sizeof(struct st_test),
|
|
|
|
sizeof(test_4)/sizeof(struct st_test),
|
|
|
|
sizeof(test_5)/sizeof(struct st_test),
|
|
|
|
sizeof(test_6)/sizeof(struct st_test),
|
|
|
|
sizeof(test_7)/sizeof(struct st_test),
|
|
|
|
sizeof(test_8)/sizeof(struct st_test),
|
|
|
|
sizeof(test_9)/sizeof(struct st_test),
|
|
|
|
sizeof(test_10)/sizeof(struct st_test),
|
|
|
|
sizeof(test_11)/sizeof(struct st_test),
|
|
|
|
sizeof(test_12)/sizeof(struct st_test),
|
|
|
|
sizeof(test_13)/sizeof(struct st_test),
|
|
|
|
sizeof(test_14)/sizeof(struct st_test),
|
|
|
|
sizeof(test_15)/sizeof(struct st_test),
|
|
|
|
sizeof(test_16)/sizeof(struct st_test)
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
static pthread_cond_t COND_thread_count;
|
|
|
|
static pthread_mutex_t LOCK_thread_count;
|
|
|
|
static uint thread_count;
|
|
|
|
static ulong sum=0;
|
|
|
|
|
|
|
|
#define MAX_LOCK_COUNT 8
|
|
|
|
|
|
|
|
/* The following functions is for WRITE_CONCURRENT_INSERT */
|
|
|
|
|
2005-05-13 11:08:08 +02:00
|
|
|
static void test_get_status(void* param __attribute__((unused)),
|
|
|
|
int concurrent_insert __attribute__((unused)))
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
static void test_copy_status(void* to __attribute__((unused)) ,
|
|
|
|
void *from __attribute__((unused)))
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
static my_bool test_check_status(void* param __attribute__((unused)))
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void *test_thread(void *arg)
|
|
|
|
{
|
|
|
|
int i,j,param=*((int*) arg);
|
|
|
|
THR_LOCK_DATA data[MAX_LOCK_COUNT];
|
2005-07-19 20:21:12 +02:00
|
|
|
THR_LOCK_OWNER owner;
|
|
|
|
THR_LOCK_INFO lock_info;
|
2000-07-31 21:29:14 +02:00
|
|
|
THR_LOCK_DATA *multi_locks[MAX_LOCK_COUNT];
|
|
|
|
my_thread_init();
|
|
|
|
|
|
|
|
printf("Thread %s (%d) started\n",my_thread_name(),param); fflush(stdout);
|
|
|
|
|
2005-07-19 20:21:12 +02:00
|
|
|
|
|
|
|
thr_lock_info_init(&lock_info);
|
|
|
|
thr_lock_owner_init(&owner, &lock_info);
|
2000-07-31 21:29:14 +02:00
|
|
|
for (i=0; i < lock_counts[param] ; i++)
|
|
|
|
thr_lock_data_init(locks+tests[param][i].lock_nr,data+i,NULL);
|
|
|
|
for (j=1 ; j < 10 ; j++) /* try locking 10 times */
|
|
|
|
{
|
|
|
|
for (i=0; i < lock_counts[param] ; i++)
|
|
|
|
{ /* Init multi locks */
|
|
|
|
multi_locks[i]= &data[i];
|
|
|
|
data[i].type= tests[param][i].lock_type;
|
|
|
|
}
|
2005-07-19 20:21:12 +02:00
|
|
|
thr_multi_lock(multi_locks, lock_counts[param], &owner);
|
2000-07-31 21:29:14 +02:00
|
|
|
pthread_mutex_lock(&LOCK_thread_count);
|
|
|
|
{
|
|
|
|
int tmp=rand() & 7; /* Do something from 0-2 sec */
|
|
|
|
if (tmp == 0)
|
|
|
|
sleep(1);
|
|
|
|
else if (tmp == 1)
|
|
|
|
sleep(2);
|
|
|
|
else
|
|
|
|
{
|
|
|
|
ulong k;
|
|
|
|
for (k=0 ; k < (ulong) (tmp-2)*100000L ; k++)
|
|
|
|
sum+=k;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
pthread_mutex_unlock(&LOCK_thread_count);
|
|
|
|
thr_multi_unlock(multi_locks,lock_counts[param]);
|
|
|
|
}
|
|
|
|
|
2001-02-01 13:25:51 +01:00
|
|
|
printf("Thread %s (%d) ended\n",my_thread_name(),param); fflush(stdout);
|
2000-07-31 21:29:14 +02:00
|
|
|
thr_print_locks();
|
|
|
|
pthread_mutex_lock(&LOCK_thread_count);
|
|
|
|
thread_count--;
|
|
|
|
VOID(pthread_cond_signal(&COND_thread_count)); /* Tell main we are ready */
|
|
|
|
pthread_mutex_unlock(&LOCK_thread_count);
|
|
|
|
free((gptr) arg);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
int main(int argc __attribute__((unused)),char **argv __attribute__((unused)))
|
|
|
|
{
|
|
|
|
pthread_t tid;
|
|
|
|
pthread_attr_t thr_attr;
|
|
|
|
int i,*param,error;
|
|
|
|
MY_INIT(argv[0]);
|
|
|
|
if (argc > 1 && argv[1][0] == '-' && argv[1][1] == '#')
|
|
|
|
DBUG_PUSH(argv[1]+2);
|
|
|
|
|
|
|
|
printf("Main thread: %s\n",my_thread_name());
|
|
|
|
|
|
|
|
if ((error=pthread_cond_init(&COND_thread_count,NULL)))
|
|
|
|
{
|
|
|
|
fprintf(stderr,"Got error: %d from pthread_cond_init (errno: %d)",
|
|
|
|
error,errno);
|
|
|
|
exit(1);
|
|
|
|
}
|
2001-03-26 00:05:04 +02:00
|
|
|
if ((error=pthread_mutex_init(&LOCK_thread_count,MY_MUTEX_INIT_FAST)))
|
2000-07-31 21:29:14 +02:00
|
|
|
{
|
|
|
|
fprintf(stderr,"Got error: %d from pthread_cond_init (errno: %d)",
|
|
|
|
error,errno);
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
|
|
|
|
for (i=0 ; i < (int) array_elements(locks) ; i++)
|
|
|
|
{
|
|
|
|
thr_lock_init(locks+i);
|
|
|
|
locks[i].check_status= test_check_status;
|
|
|
|
locks[i].update_status=test_get_status;
|
|
|
|
locks[i].copy_status= test_copy_status;
|
|
|
|
locks[i].get_status= test_get_status;
|
|
|
|
}
|
|
|
|
if ((error=pthread_attr_init(&thr_attr)))
|
|
|
|
{
|
|
|
|
fprintf(stderr,"Got error: %d from pthread_attr_init (errno: %d)",
|
|
|
|
error,errno);
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
if ((error=pthread_attr_setdetachstate(&thr_attr,PTHREAD_CREATE_DETACHED)))
|
|
|
|
{
|
|
|
|
fprintf(stderr,
|
|
|
|
"Got error: %d from pthread_attr_setdetachstate (errno: %d)",
|
|
|
|
error,errno);
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
#ifndef pthread_attr_setstacksize /* void return value */
|
|
|
|
if ((error=pthread_attr_setstacksize(&thr_attr,65536L)))
|
|
|
|
{
|
|
|
|
fprintf(stderr,"Got error: %d from pthread_attr_setstacksize (errno: %d)",
|
|
|
|
error,errno);
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
#ifdef HAVE_THR_SETCONCURRENCY
|
|
|
|
VOID(thr_setconcurrency(2));
|
|
|
|
#endif
|
|
|
|
for (i=0 ; i < (int) array_elements(lock_counts) ; i++)
|
|
|
|
{
|
|
|
|
param=(int*) malloc(sizeof(int));
|
|
|
|
*param=i;
|
|
|
|
|
|
|
|
if ((error=pthread_mutex_lock(&LOCK_thread_count)))
|
|
|
|
{
|
|
|
|
fprintf(stderr,"Got error: %d from pthread_mutex_lock (errno: %d)",
|
|
|
|
error,errno);
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
if ((error=pthread_create(&tid,&thr_attr,test_thread,(void*) param)))
|
|
|
|
{
|
|
|
|
fprintf(stderr,"Got error: %d from pthread_create (errno: %d)\n",
|
|
|
|
error,errno);
|
|
|
|
pthread_mutex_unlock(&LOCK_thread_count);
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
thread_count++;
|
|
|
|
pthread_mutex_unlock(&LOCK_thread_count);
|
|
|
|
}
|
|
|
|
|
|
|
|
pthread_attr_destroy(&thr_attr);
|
|
|
|
if ((error=pthread_mutex_lock(&LOCK_thread_count)))
|
|
|
|
fprintf(stderr,"Got error: %d from pthread_mutex_lock\n",error);
|
|
|
|
while (thread_count)
|
|
|
|
{
|
|
|
|
if ((error=pthread_cond_wait(&COND_thread_count,&LOCK_thread_count)))
|
|
|
|
fprintf(stderr,"Got error: %d from pthread_cond_wait\n",error);
|
|
|
|
}
|
|
|
|
if ((error=pthread_mutex_unlock(&LOCK_thread_count)))
|
|
|
|
fprintf(stderr,"Got error: %d from pthread_mutex_unlock\n",error);
|
|
|
|
for (i=0 ; i < (int) array_elements(locks) ; i++)
|
|
|
|
thr_lock_delete(locks+i);
|
|
|
|
#ifdef EXTRA_DEBUG
|
|
|
|
if (found_errors)
|
|
|
|
printf("Got %d warnings\n",found_errors);
|
|
|
|
else
|
|
|
|
#endif
|
|
|
|
printf("Test succeeded\n");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2000-08-21 23:18:32 +02:00
|
|
|
#else /* THREAD */
|
|
|
|
|
|
|
|
int main(int argc __attribute__((unused)),char **argv __attribute__((unused)))
|
|
|
|
{
|
|
|
|
printf("thr_lock disabled because we are not using threads\n");
|
|
|
|
exit(1);
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif /* THREAD */
|
|
|
|
#endif /* MAIN */
|