mariadb/include/my_pthread.h

946 lines
31 KiB
C
Raw Normal View History

/* Copyright (C) 2000-2008 MySQL AB, 2008-2009 Sun Microsystems, Inc.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
2000-07-31 21:29:14 +02:00
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */
2000-07-31 21:29:14 +02:00
/* Defines to make different thread packages compatible */
#ifndef _my_pthread_h
#define _my_pthread_h
#include "my_global.h" /* myf */
2000-07-31 21:29:14 +02:00
#ifndef ETIME
#define ETIME ETIMEDOUT /* For FreeBSD */
2000-07-31 21:29:14 +02:00
#endif
2001-08-22 14:15:01 +02:00
#ifdef __cplusplus
#define EXTERNC extern "C"
extern "C" {
#else
#define EXTERNC
2001-08-22 14:15:01 +02:00
#endif /* __cplusplus */
#if defined(__WIN__)
2000-07-31 21:29:14 +02:00
typedef CRITICAL_SECTION pthread_mutex_t;
typedef DWORD pthread_t;
2000-07-31 21:29:14 +02:00
typedef struct thread_attr {
DWORD dwStackSize ;
DWORD dwCreatingFlag ;
} pthread_attr_t ;
typedef struct { int dummy; } pthread_condattr_t;
/* Implementation of posix conditions */
typedef struct st_pthread_link {
DWORD thread_id;
struct st_pthread_link *next;
} pthread_link;
A follow-up to the patch for bug #56405 "Deadlock in the MDL deadlock detector". This patch addresses performance regression in OLTP_RO/MyISAM test on Windows introduced by the fix for bug #56405. Thus it makes original patch acceptable as a solution for bug #56585 "Slowdown of readonly sysbench benchmarks (e.g point_select) on Windows 5.5". With this patch, MySQL will use native Windows condition variables and reader-writer locks if they are supported by the OS. This speeds up MyISAM and the effect comes mostly from using native rwlocks. Native conditions improve scalability with higher number of concurrent users in other situations, e.g for prlocks. Benchmark numbers for this patch as measured on Win2008R2 quad core machine are attached to the bug report. ( direct link http://bugs.mysql.com/file.php?id=15883 ) Note that currently we require at least Windows7/WS2008R2 for reader-writer locks, even though native rwlock is available also on Vista. Reason is that "trylock" APIs are missing on Vista, and trylock is used in the server (in a single place in query cache). While this patch could have been written differently, to enable the native rwlock optimization also on Vista/WS2008 (e.g using native locks everywhere but portable implementation in query cache), this would come at the expense of the code clarity, as it would introduce a new "try-able" rwlock type, to handle Vista case. Another way to improve performance for the special case (OLTP_RO/MYISAM/Vista) would be to eliminate "trylock" usage from server, but this is outside of the scope here. Native conditions variables are used beginning with Vista though the effect of using condition variables alone is not measurable in this benchmark. But when used together with native rwlocks on Win7, native conditions improve performance in high-concurrency OLTP_RO/MyISAM (128 and more sysbench users).
2010-10-04 13:03:11 +02:00
/**
Implementation of Windows condition variables.
We use native conditions on Vista and later, and fallback to own
implementation on earlier OS version.
*/
typedef union
{
/* Native condition (used on Vista and later) */
CONDITION_VARIABLE native_cond;
/* Own implementation (used on XP) */
struct
{
uint32 waiting;
CRITICAL_SECTION lock_waiting;
enum
{
SIGNAL= 0,
BROADCAST= 1,
MAX_EVENTS= 2
} EVENTS;
HANDLE events[MAX_EVENTS];
HANDLE broadcast_block_event;
};
2000-07-31 21:29:14 +02:00
} pthread_cond_t;
typedef int pthread_mutexattr_t;
#define pthread_self() GetCurrentThreadId()
#define pthread_handler_t EXTERNC void * __cdecl
typedef void * (__cdecl *pthread_handler)(void *);
2000-07-31 21:29:14 +02:00
typedef volatile LONG my_pthread_once_t;
#define MY_PTHREAD_ONCE_INIT 0
#define MY_PTHREAD_ONCE_INPROGRESS 1
#define MY_PTHREAD_ONCE_DONE 2
/*
Struct and macros to be used in combination with the
windows implementation of pthread_cond_timedwait
*/
/*
Declare a union to make sure FILETIME is properly aligned
so it can be used directly as a 64 bit value. The value
stored is in 100ns units.
*/
union ft64 {
FILETIME ft;
__int64 i64;
};
struct timespec {
union ft64 tv;
/* The max timeout value in millisecond for pthread_cond_timedwait */
long max_timeout_msec;
};
#define set_timespec(ABSTIME,SEC) { \
GetSystemTimeAsFileTime(&((ABSTIME).tv.ft)); \
(ABSTIME).tv.i64+= (__int64)(SEC)*10000000; \
(ABSTIME).max_timeout_msec= (long)((SEC)*1000); \
}
#define set_timespec_nsec(ABSTIME,NSEC) { \
GetSystemTimeAsFileTime(&((ABSTIME).tv.ft)); \
(ABSTIME).tv.i64+= (__int64)(NSEC)/100; \
(ABSTIME).max_timeout_msec= (long)((NSEC)/1000000); \
}
Bug #45225 Locking: hang if drop table with no timeout This patch introduces timeouts for metadata locks. The timeout is specified in seconds using the new dynamic system variable "lock_wait_timeout" which has both GLOBAL and SESSION scopes. Allowed values range from 1 to 31536000 seconds (= 1 year). The default value is 1 year. The new server parameter "lock-wait-timeout" can be used to set the default value parameter upon server startup. "lock_wait_timeout" applies to all statements that use metadata locks. These include DML and DDL operations on tables, views, stored procedures and stored functions. They also include LOCK TABLES, FLUSH TABLES WITH READ LOCK and HANDLER statements. The patch also changes thr_lock.c code (table data locks used by MyISAM and other simplistic engines) to use the same system variable. InnoDB row locks are unaffected. One exception to the handling of the "lock_wait_timeout" variable is delayed inserts. All delayed inserts are executed with a timeout of 1 year regardless of the setting for the global variable. As the connection issuing the delayed insert gets no notification of delayed insert timeouts, we want to avoid unnecessary timeouts. It's important to note that the timeout value is used for each lock acquired and that one statement can take more than one lock. A statement can therefore block for longer than the lock_wait_timeout value before reporting a timeout error. When lock timeout occurs, ER_LOCK_WAIT_TIMEOUT is reported. Test case added to lock_multi.test.
2010-02-11 11:23:39 +01:00
/**
Compare two timespec structs.
@retval 1 If TS1 ends after TS2.
@retval 0 If TS1 is equal to TS2.
@retval -1 If TS1 ends before TS2.
*/
#define cmp_timespec(TS1, TS2) \
((TS1.tv.i64 > TS2.tv.i64) ? 1 : \
((TS1.tv.i64 < TS2.tv.i64) ? -1 : 0))
int win_pthread_mutex_trylock(pthread_mutex_t *mutex);
int pthread_create(pthread_t *, const pthread_attr_t *, pthread_handler, void *);
2000-07-31 21:29:14 +02:00
int pthread_cond_init(pthread_cond_t *cond, const pthread_condattr_t *attr);
int pthread_cond_wait(pthread_cond_t *cond, pthread_mutex_t *mutex);
int pthread_cond_timedwait(pthread_cond_t *cond, pthread_mutex_t *mutex,
struct timespec *abstime);
int pthread_cond_signal(pthread_cond_t *cond);
int pthread_cond_broadcast(pthread_cond_t *cond);
int pthread_cond_destroy(pthread_cond_t *cond);
int pthread_attr_init(pthread_attr_t *connect_att);
int pthread_attr_setstacksize(pthread_attr_t *connect_att,DWORD stack);
int pthread_attr_destroy(pthread_attr_t *connect_att);
int my_pthread_once(my_pthread_once_t *once_control,void (*init_routine)(void));
2000-07-31 21:29:14 +02:00
struct tm *localtime_r(const time_t *timep,struct tm *tmp);
struct tm *gmtime_r(const time_t *timep,struct tm *tmp);
void pthread_exit(void *a);
int pthread_join(pthread_t thread, void **value_ptr);
int pthread_cancel(pthread_t thread);
2000-07-31 21:29:14 +02:00
#ifndef ETIMEDOUT
2000-07-31 21:29:14 +02:00
#define ETIMEDOUT 145 /* Win32 doesn't have this */
#endif
#define HAVE_LOCALTIME_R 1
#define _REENTRANT 1
#define HAVE_PTHREAD_ATTR_SETSTACKSIZE 1
2000-07-31 21:29:14 +02:00
2000-09-20 19:37:07 +03:00
#undef SAFE_MUTEX /* This will cause conflicts */
2000-07-31 21:29:14 +02:00
#define pthread_key(T,V) DWORD V
#define pthread_key_create(A,B) ((*A=TlsAlloc())==0xFFFFFFFF)
#define pthread_key_delete(A) TlsFree(A)
#define my_pthread_setspecific_ptr(T,V) (!TlsSetValue((T),(V)))
#define pthread_setspecific(A,B) (!TlsSetValue((A),(B)))
2000-07-31 21:29:14 +02:00
#define pthread_getspecific(A) (TlsGetValue(A))
#define my_pthread_getspecific(T,A) ((T) TlsGetValue(A))
2000-10-11 00:48:03 +03:00
#define my_pthread_getspecific_ptr(T,V) ((T) TlsGetValue(V))
2000-07-31 21:29:14 +02:00
#define pthread_equal(A,B) ((A) == (B))
#define pthread_mutex_init(A,B) (InitializeCriticalSection(A),0)
2000-07-31 21:29:14 +02:00
#define pthread_mutex_lock(A) (EnterCriticalSection(A),0)
#define pthread_mutex_trylock(A) win_pthread_mutex_trylock((A))
#define pthread_mutex_unlock(A) (LeaveCriticalSection(A), 0)
#define pthread_mutex_destroy(A) (DeleteCriticalSection(A), 0)
#define pthread_kill(A,B) pthread_dummy((A) ? 0 : ESRCH)
/* Dummy defines for easier code */
2000-07-31 21:29:14 +02:00
#define pthread_attr_setdetachstate(A,B) pthread_dummy(0)
#define pthread_attr_setscope(A,B)
#define pthread_detach_this_thread()
#define pthread_condattr_init(A)
#define pthread_condattr_destroy(A)
2009-11-25 15:20:14 +01:00
#define pthread_yield() SwitchToThread()
#define my_sigset(A,B) signal(A,B)
2000-07-31 21:29:14 +02:00
#else /* Normal threads */
#ifdef HAVE_rts_threads
#define sigwait org_sigwait
#include <signal.h>
#undef sigwait
#endif
#include <pthread.h>
#ifndef _REENTRANT
#define _REENTRANT
#endif
#ifdef HAVE_THR_SETCONCURRENCY
#include <thread.h> /* Probably solaris */
#endif
#ifdef HAVE_SCHED_H
#include <sched.h>
#endif
#ifdef HAVE_SYNCH_H
#include <synch.h>
#endif
#define pthread_key(T,V) pthread_key_t V
#define my_pthread_getspecific_ptr(T,V) my_pthread_getspecific(T,(V))
#define my_pthread_setspecific_ptr(T,V) pthread_setspecific(T,(void*) (V))
#define pthread_detach_this_thread()
#define pthread_handler_t EXTERNC void *
2000-07-31 21:29:14 +02:00
typedef void *(* pthread_handler)(void *);
#define my_pthread_once_t pthread_once_t
#define MY_PTHREAD_ONCE_INIT PTHREAD_ONCE_INIT
#define my_pthread_once(C,F) pthread_once(C,F)
2000-07-31 21:29:14 +02:00
/* Test first for RTS or FSU threads */
#if defined(PTHREAD_SCOPE_GLOBAL) && !defined(PTHREAD_SCOPE_SYSTEM)
#define HAVE_rts_threads
extern int my_pthread_create_detached;
#define pthread_sigmask(A,B,C) sigprocmask((A),(B),(C))
#define PTHREAD_CREATE_DETACHED &my_pthread_create_detached
#define PTHREAD_SCOPE_SYSTEM PTHREAD_SCOPE_GLOBAL
#define PTHREAD_SCOPE_PROCESS PTHREAD_SCOPE_LOCAL
#define USE_ALARM_THREAD
#endif /* defined(PTHREAD_SCOPE_GLOBAL) && !defined(PTHREAD_SCOPE_SYSTEM) */
#if defined(_BSDI_VERSION) && _BSDI_VERSION < 199910
2000-08-31 22:00:09 +03:00
int sigwait(sigset_t *set, int *sig);
2000-07-31 21:29:14 +02:00
#endif
#ifndef HAVE_NONPOSIX_SIGWAIT
#define my_sigwait(A,B) sigwait((A),(B))
#else
2000-08-31 17:25:30 +03:00
int my_sigwait(const sigset_t *set,int *sig);
2000-07-31 21:29:14 +02:00
#endif
#ifdef HAVE_NONPOSIX_PTHREAD_MUTEX_INIT
#ifndef SAFE_MUTEX
#define pthread_mutex_init(a,b) my_pthread_mutex_init((a),(b))
extern int my_pthread_mutex_init(pthread_mutex_t *mp,
const pthread_mutexattr_t *attr);
#endif /* SAFE_MUTEX */
#define pthread_cond_init(a,b) my_pthread_cond_init((a),(b))
extern int my_pthread_cond_init(pthread_cond_t *mp,
const pthread_condattr_t *attr);
#endif /* HAVE_NONPOSIX_PTHREAD_MUTEX_INIT */
#if defined(HAVE_SIGTHREADMASK) && !defined(HAVE_PTHREAD_SIGMASK)
#define pthread_sigmask(A,B,C) sigthreadmask((A),(B),(C))
#endif
#if !defined(HAVE_SIGWAIT) && !defined(HAVE_rts_threads) && !defined(sigwait) && !defined(alpha_linux_port) && !defined(HAVE_NONPOSIX_SIGWAIT) && !defined(HAVE_DEC_3_2_THREADS) && !defined(_AIX)
2000-08-31 22:00:09 +03:00
int sigwait(sigset_t *setp, int *sigp); /* Use our implemention */
2000-07-31 21:29:14 +02:00
#endif
/*
We define my_sigset() and use that instead of the system sigset() so that
we can favor an implementation based on sigaction(). On some systems, such
as Mac OS X, sigset() results in flags such as SA_RESTART being set, and
we want to make sure that no such flags are set.
*/
#if defined(HAVE_SIGACTION) && !defined(my_sigset)
#define my_sigset(A,B) do { struct sigaction l_s; sigset_t l_set; \
DBUG_ASSERT((A) != 0); \
sigemptyset(&l_set); \
l_s.sa_handler = (B); \
l_s.sa_mask = l_set; \
l_s.sa_flags = 0; \
sigaction((A), &l_s, NULL); \
} while (0)
#elif defined(HAVE_SIGSET) && !defined(my_sigset)
#define my_sigset(A,B) sigset((A),(B))
#elif !defined(my_sigset)
#define my_sigset(A,B) signal((A),(B))
2000-07-31 21:29:14 +02:00
#endif
#if !defined(HAVE_PTHREAD_ATTR_SETSCOPE) || defined(HAVE_DEC_3_2_THREADS)
#define pthread_attr_setscope(A,B)
#undef HAVE_GETHOSTBYADDR_R /* No definition */
#endif
#if defined(HAVE_BROKEN_PTHREAD_COND_TIMEDWAIT) && !defined(SAFE_MUTEX)
extern int my_pthread_cond_timedwait(pthread_cond_t *cond,
pthread_mutex_t *mutex,
struct timespec *abstime);
#define pthread_cond_timedwait(A,B,C) my_pthread_cond_timedwait((A),(B),(C))
#endif
#if !defined( HAVE_NONPOSIX_PTHREAD_GETSPECIFIC)
2000-07-31 21:29:14 +02:00
#define my_pthread_getspecific(A,B) ((A) pthread_getspecific(B))
#else
#define my_pthread_getspecific(A,B) ((A) my_pthread_getspecific_imp(B))
void *my_pthread_getspecific_imp(pthread_key_t key);
#endif
2000-07-31 21:29:14 +02:00
#ifndef HAVE_LOCALTIME_R
struct tm *localtime_r(const time_t *clock, struct tm *res);
#endif
#ifndef HAVE_GMTIME_R
struct tm *gmtime_r(const time_t *clock, struct tm *res);
#endif
2000-07-31 21:29:14 +02:00
#ifdef HAVE_PTHREAD_CONDATTR_CREATE
/* DCE threads on HPUX 10.20 */
#define pthread_condattr_init pthread_condattr_create
#define pthread_condattr_destroy pthread_condattr_delete
#endif
/* FSU THREADS */
#if !defined(HAVE_PTHREAD_KEY_DELETE) && !defined(pthread_key_delete)
#define pthread_key_delete(A) pthread_dummy(0)
#endif
2000-07-31 21:29:14 +02:00
#ifdef HAVE_CTHREADS_WRAPPER /* For MacOSX */
#define pthread_cond_destroy(A) pthread_dummy(0)
#define pthread_mutex_destroy(A) pthread_dummy(0)
#define pthread_attr_delete(A) pthread_dummy(0)
#define pthread_condattr_delete(A) pthread_dummy(0)
#define pthread_attr_setstacksize(A,B) pthread_dummy(0)
#define pthread_equal(A,B) ((A) == (B))
#define pthread_cond_timedwait(a,b,c) pthread_cond_wait((a),(b))
#define pthread_attr_init(A) pthread_attr_create(A)
#define pthread_attr_destroy(A) pthread_attr_delete(A)
#define pthread_attr_setdetachstate(A,B) pthread_dummy(0)
#define pthread_create(A,B,C,D) pthread_create((A),*(B),(C),(D))
#define pthread_sigmask(A,B,C) sigprocmask((A),(B),(C))
#define pthread_kill(A,B) pthread_dummy((A) ? 0 : ESRCH)
2000-07-31 21:29:14 +02:00
#undef pthread_detach_this_thread
#define pthread_detach_this_thread() { pthread_t tmp=pthread_self() ; pthread_detach(&tmp); }
#endif
#ifdef HAVE_DARWIN5_THREADS
2000-10-04 23:20:16 +03:00
#define pthread_sigmask(A,B,C) sigprocmask((A),(B),(C))
#define pthread_kill(A,B) pthread_dummy((A) ? 0 : ESRCH)
2000-10-04 23:20:16 +03:00
#define pthread_condattr_init(A) pthread_dummy(0)
#define pthread_condattr_destroy(A) pthread_dummy(0)
#undef pthread_detach_this_thread
#define pthread_detach_this_thread() { pthread_t tmp=pthread_self() ; pthread_detach(tmp); }
2000-10-04 23:20:16 +03:00
#endif
2000-07-31 21:29:14 +02:00
#if ((defined(HAVE_PTHREAD_ATTR_CREATE) && !defined(HAVE_SIGWAIT)) || defined(HAVE_DEC_3_2_THREADS)) && !defined(HAVE_CTHREADS_WRAPPER)
/* This is set on AIX_3_2 and Siemens unix (and DEC OSF/1 3.2 too) */
#define pthread_key_create(A,B) \
pthread_keycreate(A,(B) ?\
(pthread_destructor_t) (B) :\
(pthread_destructor_t) pthread_dummy)
#define pthread_attr_init(A) pthread_attr_create(A)
#define pthread_attr_destroy(A) pthread_attr_delete(A)
#define pthread_attr_setdetachstate(A,B) pthread_dummy(0)
#define pthread_create(A,B,C,D) pthread_create((A),*(B),(C),(D))
#ifndef pthread_sigmask
#define pthread_sigmask(A,B,C) sigprocmask((A),(B),(C))
#endif
#define pthread_kill(A,B) pthread_dummy((A) ? 0 : ESRCH)
2000-07-31 21:29:14 +02:00
#undef pthread_detach_this_thread
#define pthread_detach_this_thread() { pthread_t tmp=pthread_self() ; pthread_detach(&tmp); }
#else /* HAVE_PTHREAD_ATTR_CREATE && !HAVE_SIGWAIT */
2000-07-31 21:29:14 +02:00
#define HAVE_PTHREAD_KILL
#endif
#endif /* defined(__WIN__) */
#if defined(HPUX10) && !defined(DONT_REMAP_PTHREAD_FUNCTIONS)
#undef pthread_cond_timedwait
#define pthread_cond_timedwait(a,b,c) my_pthread_cond_timedwait((a),(b),(c))
2001-12-17 15:00:14 +02:00
int my_pthread_cond_timedwait(pthread_cond_t *cond, pthread_mutex_t *mutex,
struct timespec *abstime);
#endif
2003-02-25 21:43:24 +02:00
#if defined(HPUX10)
#define pthread_attr_getstacksize(A,B) my_pthread_attr_getstacksize(A,B)
void my_pthread_attr_getstacksize(pthread_attr_t *attrib, size_t *size);
#endif
#if defined(HAVE_POSIX1003_4a_MUTEX) && !defined(DONT_REMAP_PTHREAD_FUNCTIONS)
#undef pthread_mutex_trylock
#define pthread_mutex_trylock(a) my_pthread_mutex_trylock((a))
int my_pthread_mutex_trylock(pthread_mutex_t *mutex);
#endif
#if !defined(HAVE_PTHREAD_YIELD_ONE_ARG) && !defined(HAVE_PTHREAD_YIELD_ZERO_ARG)
/* no pthread_yield() available */
#ifdef HAVE_SCHED_YIELD
#define pthread_yield() sched_yield()
#elif defined(HAVE_PTHREAD_YIELD_NP) /* can be Mac OS X */
#define pthread_yield() pthread_yield_np()
#elif defined(HAVE_THR_YIELD)
#define pthread_yield() thr_yield()
#endif
#endif
/*
The defines set_timespec and set_timespec_nsec should be used
for calculating an absolute time at which
pthread_cond_timedwait should timeout
*/
#ifdef HAVE_TIMESPEC_TS_SEC
#ifndef set_timespec
#define set_timespec(ABSTIME,SEC) \
{ \
(ABSTIME).ts_sec=time(0) + (time_t) (SEC); \
(ABSTIME).ts_nsec=0; \
}
#endif /* !set_timespec */
#ifndef set_timespec_nsec
#define set_timespec_nsec(ABSTIME,NSEC) \
{ \
ulonglong now= my_getsystime() + (NSEC/100); \
(ABSTIME).ts_sec= (now / ULL(10000000)); \
(ABSTIME).ts_nsec= (now % ULL(10000000) * 100 + ((NSEC) % 100)); \
}
#endif /* !set_timespec_nsec */
#else
#ifndef set_timespec
#define set_timespec(ABSTIME,SEC) \
{\
struct timeval tv;\
gettimeofday(&tv,0);\
(ABSTIME).tv_sec=tv.tv_sec+(time_t) (SEC);\
(ABSTIME).tv_nsec=tv.tv_usec*1000;\
}
#endif /* !set_timespec */
#ifndef set_timespec_nsec
#define set_timespec_nsec(ABSTIME,NSEC) \
{\
ulonglong now= my_getsystime() + (NSEC/100); \
(ABSTIME).tv_sec= (time_t) (now / ULL(10000000)); \
(ABSTIME).tv_nsec= (long) (now % ULL(10000000) * 100 + ((NSEC) % 100)); \
}
#endif /* !set_timespec_nsec */
Bug #45225 Locking: hang if drop table with no timeout This patch introduces timeouts for metadata locks. The timeout is specified in seconds using the new dynamic system variable "lock_wait_timeout" which has both GLOBAL and SESSION scopes. Allowed values range from 1 to 31536000 seconds (= 1 year). The default value is 1 year. The new server parameter "lock-wait-timeout" can be used to set the default value parameter upon server startup. "lock_wait_timeout" applies to all statements that use metadata locks. These include DML and DDL operations on tables, views, stored procedures and stored functions. They also include LOCK TABLES, FLUSH TABLES WITH READ LOCK and HANDLER statements. The patch also changes thr_lock.c code (table data locks used by MyISAM and other simplistic engines) to use the same system variable. InnoDB row locks are unaffected. One exception to the handling of the "lock_wait_timeout" variable is delayed inserts. All delayed inserts are executed with a timeout of 1 year regardless of the setting for the global variable. As the connection issuing the delayed insert gets no notification of delayed insert timeouts, we want to avoid unnecessary timeouts. It's important to note that the timeout value is used for each lock acquired and that one statement can take more than one lock. A statement can therefore block for longer than the lock_wait_timeout value before reporting a timeout error. When lock timeout occurs, ER_LOCK_WAIT_TIMEOUT is reported. Test case added to lock_multi.test.
2010-02-11 11:23:39 +01:00
#endif /* HAVE_TIMESPEC_TS_SEC */
/**
Compare two timespec structs.
@retval 1 If TS1 ends after TS2.
@retval 0 If TS1 is equal to TS2.
@retval -1 If TS1 ends before TS2.
*/
#ifdef HAVE_TIMESPEC_TS_SEC
#ifndef cmp_timespec
#define cmp_timespec(TS1, TS2) \
((TS1.ts_sec > TS2.ts_sec || \
(TS1.ts_sec == TS2.ts_sec && TS1.ts_nsec > TS2.ts_nsec)) ? 1 : \
((TS1.ts_sec < TS2.ts_sec || \
(TS1.ts_sec == TS2.ts_sec && TS1.ts_nsec < TS2.ts_nsec)) ? -1 : 0))
#endif /* !cmp_timespec */
#else
#ifndef cmp_timespec
#define cmp_timespec(TS1, TS2) \
((TS1.tv_sec > TS2.tv_sec || \
(TS1.tv_sec == TS2.tv_sec && TS1.tv_nsec > TS2.tv_nsec)) ? 1 : \
((TS1.tv_sec < TS2.tv_sec || \
(TS1.tv_sec == TS2.tv_sec && TS1.tv_nsec < TS2.tv_nsec)) ? -1 : 0))
#endif /* !cmp_timespec */
#endif /* HAVE_TIMESPEC_TS_SEC */
/* safe_mutex adds checking to mutex for easier debugging */
typedef struct st_safe_mutex_t
{
pthread_mutex_t global,mutex;
const char *file;
uint line,count;
pthread_t thread;
#ifdef SAFE_MUTEX_DETECT_DESTROY
struct st_safe_mutex_info_t *info; /* to track destroying of mutexes */
#endif
} safe_mutex_t;
#ifdef SAFE_MUTEX_DETECT_DESTROY
/*
Used to track the destroying of mutexes. This needs to be a seperate
structure because the safe_mutex_t structure could be freed before
the mutexes are destroyed.
*/
typedef struct st_safe_mutex_info_t
{
struct st_safe_mutex_info_t *next;
struct st_safe_mutex_info_t *prev;
const char *init_file;
uint32 init_line;
} safe_mutex_info_t;
#endif /* SAFE_MUTEX_DETECT_DESTROY */
int safe_mutex_init(safe_mutex_t *mp, const pthread_mutexattr_t *attr,
const char *file, uint line);
int safe_mutex_lock(safe_mutex_t *mp, my_bool try_lock, const char *file, uint line);
int safe_mutex_unlock(safe_mutex_t *mp,const char *file, uint line);
int safe_mutex_destroy(safe_mutex_t *mp,const char *file, uint line);
int safe_cond_wait(pthread_cond_t *cond, safe_mutex_t *mp,const char *file,
uint line);
int safe_cond_timedwait(pthread_cond_t *cond, safe_mutex_t *mp,
const struct timespec *abstime,
const char *file, uint line);
void safe_mutex_global_init(void);
void safe_mutex_end(FILE *file);
/* Wrappers if safe mutex is actually used */
#ifdef SAFE_MUTEX
#undef pthread_mutex_init
#undef pthread_mutex_lock
#undef pthread_mutex_unlock
#undef pthread_mutex_destroy
#undef pthread_mutex_wait
#undef pthread_mutex_timedwait
#undef pthread_mutex_t
#undef pthread_cond_wait
#undef pthread_cond_timedwait
#undef pthread_mutex_trylock
#define pthread_mutex_init(A,B) safe_mutex_init((A),(B),__FILE__,__LINE__)
#define pthread_mutex_lock(A) safe_mutex_lock((A), FALSE, __FILE__, __LINE__)
#define pthread_mutex_unlock(A) safe_mutex_unlock((A),__FILE__,__LINE__)
#define pthread_mutex_destroy(A) safe_mutex_destroy((A),__FILE__,__LINE__)
#define pthread_cond_wait(A,B) safe_cond_wait((A),(B),__FILE__,__LINE__)
#define pthread_cond_timedwait(A,B,C) safe_cond_timedwait((A),(B),(C),__FILE__,__LINE__)
#define pthread_mutex_trylock(A) safe_mutex_lock((A), TRUE, __FILE__, __LINE__)
#define pthread_mutex_t safe_mutex_t
#define safe_mutex_assert_owner(mp) \
DBUG_ASSERT((mp)->count > 0 && \
pthread_equal(pthread_self(), (mp)->thread))
#define safe_mutex_assert_not_owner(mp) \
DBUG_ASSERT(! (mp)->count || \
! pthread_equal(pthread_self(), (mp)->thread))
#else
#define safe_mutex_assert_owner(mp)
#define safe_mutex_assert_not_owner(mp)
#endif /* SAFE_MUTEX */
#if defined(MY_PTHREAD_FASTMUTEX) && !defined(SAFE_MUTEX)
typedef struct st_my_pthread_fastmutex_t
{
pthread_mutex_t mutex;
uint spins;
uint rng_state;
} my_pthread_fastmutex_t;
void fastmutex_global_init(void);
int my_pthread_fastmutex_init(my_pthread_fastmutex_t *mp,
const pthread_mutexattr_t *attr);
int my_pthread_fastmutex_lock(my_pthread_fastmutex_t *mp);
#undef pthread_mutex_init
#undef pthread_mutex_lock
#undef pthread_mutex_unlock
#undef pthread_mutex_destroy
#undef pthread_mutex_wait
#undef pthread_mutex_timedwait
#undef pthread_mutex_t
#undef pthread_cond_wait
#undef pthread_cond_timedwait
#undef pthread_mutex_trylock
#define pthread_mutex_init(A,B) my_pthread_fastmutex_init((A),(B))
#define pthread_mutex_lock(A) my_pthread_fastmutex_lock(A)
#define pthread_mutex_unlock(A) pthread_mutex_unlock(&(A)->mutex)
#define pthread_mutex_destroy(A) pthread_mutex_destroy(&(A)->mutex)
#define pthread_cond_wait(A,B) pthread_cond_wait((A),&(B)->mutex)
#define pthread_cond_timedwait(A,B,C) pthread_cond_timedwait((A),&(B)->mutex,(C))
#define pthread_mutex_trylock(A) pthread_mutex_trylock(&(A)->mutex)
#define pthread_mutex_t my_pthread_fastmutex_t
#endif /* defined(MY_PTHREAD_FASTMUTEX) && !defined(SAFE_MUTEX) */
/* READ-WRITE thread locking */
2000-07-31 21:29:14 +02:00
#ifdef HAVE_BROKEN_RWLOCK /* For OpenUnix */
#undef HAVE_PTHREAD_RWLOCK_RDLOCK
#undef HAVE_RWLOCK_INIT
#undef HAVE_RWLOCK_T
#endif
2000-07-31 21:29:14 +02:00
#if defined(USE_MUTEX_INSTEAD_OF_RW_LOCKS)
/* use these defs for simple mutex locking */
#define rw_lock_t pthread_mutex_t
#define my_rwlock_init(A,B) pthread_mutex_init((A),(B))
#define rw_rdlock(A) pthread_mutex_lock((A))
#define rw_wrlock(A) pthread_mutex_lock((A))
#define rw_tryrdlock(A) pthread_mutex_trylock((A))
#define rw_trywrlock(A) pthread_mutex_trylock((A))
2000-07-31 21:29:14 +02:00
#define rw_unlock(A) pthread_mutex_unlock((A))
#define rwlock_destroy(A) pthread_mutex_destroy((A))
#elif defined(HAVE_PTHREAD_RWLOCK_RDLOCK)
#define rw_lock_t pthread_rwlock_t
#define my_rwlock_init(A,B) pthread_rwlock_init((A),(B))
#define rw_rdlock(A) pthread_rwlock_rdlock(A)
#define rw_wrlock(A) pthread_rwlock_wrlock(A)
2002-06-30 12:08:58 +03:00
#define rw_tryrdlock(A) pthread_rwlock_tryrdlock((A))
#define rw_trywrlock(A) pthread_rwlock_trywrlock((A))
2000-07-31 21:29:14 +02:00
#define rw_unlock(A) pthread_rwlock_unlock(A)
#define rwlock_destroy(A) pthread_rwlock_destroy(A)
#elif defined(HAVE_RWLOCK_INIT)
#ifdef HAVE_RWLOCK_T /* For example Solaris 2.6-> */
#define rw_lock_t rwlock_t
#endif
#define my_rwlock_init(A,B) rwlock_init((A),USYNC_THREAD,0)
#else
/* Use our own version of read/write locks */
#define NEED_MY_RW_LOCK 1
2000-07-31 21:29:14 +02:00
#define rw_lock_t my_rw_lock_t
A better fix for bug #56405 "Deadlock in the MDL deadlock detector" that doesn't introduce bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors". Deadlock could have occurred when workload containing a mix of DML, DDL and FLUSH TABLES statements affecting the same set of tables was executed in a heavily concurrent environment. This deadlock occurred when several connections tried to perform deadlock detection in the metadata locking subsystem. The first connection started traversing wait-for graph, encountered a sub-graph representing a wait for flush, acquired LOCK_open and dived into sub-graph inspection. Then it encountered sub-graph corresponding to wait for metadata lock and blocked while trying to acquire a rd-lock on MDL_lock::m_rwlock, since some,other thread had a wr-lock on it. When this wr-lock was released it could have happened (if there was another pending wr-lock against this rwlock) that the rd-lock from the first connection was left unsatisfied but at the same time the new rd-lock request from the second connection sneaked in and was satisfied (for this to be possible the second rd-request should come exactly after the wr-lock is released but before pending the wr-lock manages to grab rwlock, which is possible both on Linux and in our own rwlock implementation). If this second connection continued traversing the wait-for graph and encountered a sub-graph representing a wait for flush it tried to acquire LOCK_open and thus the deadlock was created. The previous patch tried to workaround this problem by not allowing the deadlock detector to lock LOCK_open mutex if some other thread doing deadlock detection already owns it and current search depth is greater than 0. Instead deadlock was reported. As a result it has introduced bug #56715. This patch solves this problem in a different way. It introduces a new rw_pr_lock_t implementation to be used by MDL subsystem instead of one based on Linux rwlocks or our own rwlock implementation. This new implementation never allows situation in which an rwlock is rd-locked and there is a blocked pending rd-lock. Thus the situation which has caused this bug becomes impossible with this implementation. Due to fact that this implementation is optimized for wr-lock/unlock scenario which is most common in the MDL subsystem it doesn't introduce noticeable performance regressions in sysbench tests. Moreover it significantly improves situation for POINT_SELECT test when many connections are used. No test case is provided as this bug is very hard to repeat in MTR environment but is repeatable with the help of RQG tests. This patch also doesn't include a test for bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors" as it takes too much time to be run as part of normal test-suite runs.
2010-09-29 16:09:07 +04:00
#define my_rwlock_init(A,B) my_rw_init((A))
2000-07-31 21:29:14 +02:00
#define rw_rdlock(A) my_rw_rdlock((A))
#define rw_wrlock(A) my_rw_wrlock((A))
#define rw_tryrdlock(A) my_rw_tryrdlock((A))
#define rw_trywrlock(A) my_rw_trywrlock((A))
2000-07-31 21:29:14 +02:00
#define rw_unlock(A) my_rw_unlock((A))
#define rwlock_destroy(A) my_rw_destroy((A))
#define rw_lock_assert_write_owner(A) my_rw_lock_assert_write_owner((A))
#define rw_lock_assert_not_write_owner(A) my_rw_lock_assert_not_write_owner((A))
#endif /* USE_MUTEX_INSTEAD_OF_RW_LOCKS */
2000-07-31 21:29:14 +02:00
A better fix for bug #56405 "Deadlock in the MDL deadlock detector" that doesn't introduce bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors". Deadlock could have occurred when workload containing a mix of DML, DDL and FLUSH TABLES statements affecting the same set of tables was executed in a heavily concurrent environment. This deadlock occurred when several connections tried to perform deadlock detection in the metadata locking subsystem. The first connection started traversing wait-for graph, encountered a sub-graph representing a wait for flush, acquired LOCK_open and dived into sub-graph inspection. Then it encountered sub-graph corresponding to wait for metadata lock and blocked while trying to acquire a rd-lock on MDL_lock::m_rwlock, since some,other thread had a wr-lock on it. When this wr-lock was released it could have happened (if there was another pending wr-lock against this rwlock) that the rd-lock from the first connection was left unsatisfied but at the same time the new rd-lock request from the second connection sneaked in and was satisfied (for this to be possible the second rd-request should come exactly after the wr-lock is released but before pending the wr-lock manages to grab rwlock, which is possible both on Linux and in our own rwlock implementation). If this second connection continued traversing the wait-for graph and encountered a sub-graph representing a wait for flush it tried to acquire LOCK_open and thus the deadlock was created. The previous patch tried to workaround this problem by not allowing the deadlock detector to lock LOCK_open mutex if some other thread doing deadlock detection already owns it and current search depth is greater than 0. Instead deadlock was reported. As a result it has introduced bug #56715. This patch solves this problem in a different way. It introduces a new rw_pr_lock_t implementation to be used by MDL subsystem instead of one based on Linux rwlocks or our own rwlock implementation. This new implementation never allows situation in which an rwlock is rd-locked and there is a blocked pending rd-lock. Thus the situation which has caused this bug becomes impossible with this implementation. Due to fact that this implementation is optimized for wr-lock/unlock scenario which is most common in the MDL subsystem it doesn't introduce noticeable performance regressions in sysbench tests. Moreover it significantly improves situation for POINT_SELECT test when many connections are used. No test case is provided as this bug is very hard to repeat in MTR environment but is repeatable with the help of RQG tests. This patch also doesn't include a test for bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors" as it takes too much time to be run as part of normal test-suite runs.
2010-09-29 16:09:07 +04:00
/**
Portable implementation of special type of read-write locks.
These locks have two properties which are unusual for rwlocks:
1) They "prefer readers" in the sense that they do not allow
situations in which rwlock is rd-locked and there is a
pending rd-lock which is blocked (e.g. due to pending
request for wr-lock).
This is a stronger guarantee than one which is provided for
PTHREAD_RWLOCK_PREFER_READER_NP rwlocks in Linux.
MDL subsystem deadlock detector relies on this property for
its correctness.
2) They are optimized for uncontended wr-lock/unlock case.
This is scenario in which they are most oftenly used
within MDL subsystem. Optimizing for it gives significant
performance improvements in some of tests involving many
connections.
Another important requirement imposed on this type of rwlock
by the MDL subsystem is that it should be OK to destroy rwlock
object which is in unlocked state even though some threads might
have not yet fully left unlock operation for it (of course there
is an external guarantee that no thread will try to lock rwlock
which is destroyed).
Putting it another way the unlock operation should not access
rwlock data after changing its state to unlocked.
TODO/FIXME: We should consider alleviating this requirement as
it blocks us from doing certain performance optimizations.
*/
A better fix for bug #56405 "Deadlock in the MDL deadlock detector" that doesn't introduce bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors". Deadlock could have occurred when workload containing a mix of DML, DDL and FLUSH TABLES statements affecting the same set of tables was executed in a heavily concurrent environment. This deadlock occurred when several connections tried to perform deadlock detection in the metadata locking subsystem. The first connection started traversing wait-for graph, encountered a sub-graph representing a wait for flush, acquired LOCK_open and dived into sub-graph inspection. Then it encountered sub-graph corresponding to wait for metadata lock and blocked while trying to acquire a rd-lock on MDL_lock::m_rwlock, since some,other thread had a wr-lock on it. When this wr-lock was released it could have happened (if there was another pending wr-lock against this rwlock) that the rd-lock from the first connection was left unsatisfied but at the same time the new rd-lock request from the second connection sneaked in and was satisfied (for this to be possible the second rd-request should come exactly after the wr-lock is released but before pending the wr-lock manages to grab rwlock, which is possible both on Linux and in our own rwlock implementation). If this second connection continued traversing the wait-for graph and encountered a sub-graph representing a wait for flush it tried to acquire LOCK_open and thus the deadlock was created. The previous patch tried to workaround this problem by not allowing the deadlock detector to lock LOCK_open mutex if some other thread doing deadlock detection already owns it and current search depth is greater than 0. Instead deadlock was reported. As a result it has introduced bug #56715. This patch solves this problem in a different way. It introduces a new rw_pr_lock_t implementation to be used by MDL subsystem instead of one based on Linux rwlocks or our own rwlock implementation. This new implementation never allows situation in which an rwlock is rd-locked and there is a blocked pending rd-lock. Thus the situation which has caused this bug becomes impossible with this implementation. Due to fact that this implementation is optimized for wr-lock/unlock scenario which is most common in the MDL subsystem it doesn't introduce noticeable performance regressions in sysbench tests. Moreover it significantly improves situation for POINT_SELECT test when many connections are used. No test case is provided as this bug is very hard to repeat in MTR environment but is repeatable with the help of RQG tests. This patch also doesn't include a test for bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors" as it takes too much time to be run as part of normal test-suite runs.
2010-09-29 16:09:07 +04:00
typedef struct st_rw_pr_lock_t {
/**
Lock which protects the structure.
Also held for the duration of wr-lock.
*/
pthread_mutex_t lock;
/**
Condition variable which is used to wake-up
writers waiting for readers to go away.
*/
pthread_cond_t no_active_readers;
/** Number of active readers. */
uint active_readers;
/** Number of writers waiting for readers to go away. */
uint writers_waiting_readers;
/** Indicates whether there is an active writer. */
my_bool active_writer;
#ifdef SAFE_MUTEX
/** Thread holding wr-lock (for debug purposes only). */
pthread_t writer_thread;
#endif
} rw_pr_lock_t;
extern int rw_pr_init(rw_pr_lock_t *);
A better fix for bug #56405 "Deadlock in the MDL deadlock detector" that doesn't introduce bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors". Deadlock could have occurred when workload containing a mix of DML, DDL and FLUSH TABLES statements affecting the same set of tables was executed in a heavily concurrent environment. This deadlock occurred when several connections tried to perform deadlock detection in the metadata locking subsystem. The first connection started traversing wait-for graph, encountered a sub-graph representing a wait for flush, acquired LOCK_open and dived into sub-graph inspection. Then it encountered sub-graph corresponding to wait for metadata lock and blocked while trying to acquire a rd-lock on MDL_lock::m_rwlock, since some,other thread had a wr-lock on it. When this wr-lock was released it could have happened (if there was another pending wr-lock against this rwlock) that the rd-lock from the first connection was left unsatisfied but at the same time the new rd-lock request from the second connection sneaked in and was satisfied (for this to be possible the second rd-request should come exactly after the wr-lock is released but before pending the wr-lock manages to grab rwlock, which is possible both on Linux and in our own rwlock implementation). If this second connection continued traversing the wait-for graph and encountered a sub-graph representing a wait for flush it tried to acquire LOCK_open and thus the deadlock was created. The previous patch tried to workaround this problem by not allowing the deadlock detector to lock LOCK_open mutex if some other thread doing deadlock detection already owns it and current search depth is greater than 0. Instead deadlock was reported. As a result it has introduced bug #56715. This patch solves this problem in a different way. It introduces a new rw_pr_lock_t implementation to be used by MDL subsystem instead of one based on Linux rwlocks or our own rwlock implementation. This new implementation never allows situation in which an rwlock is rd-locked and there is a blocked pending rd-lock. Thus the situation which has caused this bug becomes impossible with this implementation. Due to fact that this implementation is optimized for wr-lock/unlock scenario which is most common in the MDL subsystem it doesn't introduce noticeable performance regressions in sysbench tests. Moreover it significantly improves situation for POINT_SELECT test when many connections are used. No test case is provided as this bug is very hard to repeat in MTR environment but is repeatable with the help of RQG tests. This patch also doesn't include a test for bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors" as it takes too much time to be run as part of normal test-suite runs.
2010-09-29 16:09:07 +04:00
extern int rw_pr_rdlock(rw_pr_lock_t *);
extern int rw_pr_wrlock(rw_pr_lock_t *);
extern int rw_pr_unlock(rw_pr_lock_t *);
extern int rw_pr_destroy(rw_pr_lock_t *);
#ifdef SAFE_MUTEX
#define rw_pr_lock_assert_write_owner(A) \
DBUG_ASSERT((A)->active_writer && pthread_equal(pthread_self(), \
(A)->writer_thread))
#define rw_pr_lock_assert_not_write_owner(A) \
DBUG_ASSERT(! (A)->active_writer || ! pthread_equal(pthread_self(), \
(A)->writer_thread))
#else
#define rw_pr_lock_assert_write_owner(A)
#define rw_pr_lock_assert_not_write_owner(A)
A better fix for bug #56405 "Deadlock in the MDL deadlock detector" that doesn't introduce bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors". Deadlock could have occurred when workload containing a mix of DML, DDL and FLUSH TABLES statements affecting the same set of tables was executed in a heavily concurrent environment. This deadlock occurred when several connections tried to perform deadlock detection in the metadata locking subsystem. The first connection started traversing wait-for graph, encountered a sub-graph representing a wait for flush, acquired LOCK_open and dived into sub-graph inspection. Then it encountered sub-graph corresponding to wait for metadata lock and blocked while trying to acquire a rd-lock on MDL_lock::m_rwlock, since some,other thread had a wr-lock on it. When this wr-lock was released it could have happened (if there was another pending wr-lock against this rwlock) that the rd-lock from the first connection was left unsatisfied but at the same time the new rd-lock request from the second connection sneaked in and was satisfied (for this to be possible the second rd-request should come exactly after the wr-lock is released but before pending the wr-lock manages to grab rwlock, which is possible both on Linux and in our own rwlock implementation). If this second connection continued traversing the wait-for graph and encountered a sub-graph representing a wait for flush it tried to acquire LOCK_open and thus the deadlock was created. The previous patch tried to workaround this problem by not allowing the deadlock detector to lock LOCK_open mutex if some other thread doing deadlock detection already owns it and current search depth is greater than 0. Instead deadlock was reported. As a result it has introduced bug #56715. This patch solves this problem in a different way. It introduces a new rw_pr_lock_t implementation to be used by MDL subsystem instead of one based on Linux rwlocks or our own rwlock implementation. This new implementation never allows situation in which an rwlock is rd-locked and there is a blocked pending rd-lock. Thus the situation which has caused this bug becomes impossible with this implementation. Due to fact that this implementation is optimized for wr-lock/unlock scenario which is most common in the MDL subsystem it doesn't introduce noticeable performance regressions in sysbench tests. Moreover it significantly improves situation for POINT_SELECT test when many connections are used. No test case is provided as this bug is very hard to repeat in MTR environment but is repeatable with the help of RQG tests. This patch also doesn't include a test for bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors" as it takes too much time to be run as part of normal test-suite runs.
2010-09-29 16:09:07 +04:00
#endif /* SAFE_MUTEX */
#ifdef NEED_MY_RW_LOCK
A follow-up to the patch for bug #56405 "Deadlock in the MDL deadlock detector". This patch addresses performance regression in OLTP_RO/MyISAM test on Windows introduced by the fix for bug #56405. Thus it makes original patch acceptable as a solution for bug #56585 "Slowdown of readonly sysbench benchmarks (e.g point_select) on Windows 5.5". With this patch, MySQL will use native Windows condition variables and reader-writer locks if they are supported by the OS. This speeds up MyISAM and the effect comes mostly from using native rwlocks. Native conditions improve scalability with higher number of concurrent users in other situations, e.g for prlocks. Benchmark numbers for this patch as measured on Win2008R2 quad core machine are attached to the bug report. ( direct link http://bugs.mysql.com/file.php?id=15883 ) Note that currently we require at least Windows7/WS2008R2 for reader-writer locks, even though native rwlock is available also on Vista. Reason is that "trylock" APIs are missing on Vista, and trylock is used in the server (in a single place in query cache). While this patch could have been written differently, to enable the native rwlock optimization also on Vista/WS2008 (e.g using native locks everywhere but portable implementation in query cache), this would come at the expense of the code clarity, as it would introduce a new "try-able" rwlock type, to handle Vista case. Another way to improve performance for the special case (OLTP_RO/MYISAM/Vista) would be to eliminate "trylock" usage from server, but this is outside of the scope here. Native conditions variables are used beginning with Vista though the effect of using condition variables alone is not measurable in this benchmark. But when used together with native rwlocks on Win7, native conditions improve performance in high-concurrency OLTP_RO/MyISAM (128 and more sysbench users).
2010-10-04 13:03:11 +02:00
#ifdef _WIN32
/**
Implementation of Windows rwlock.
We use native (slim) rwlocks on Win7 and later, and fallback to portable
implementation on earlier Windows.
slim rwlock are also available on Vista/WS2008, but we do not use it
("trylock" APIs are missing on Vista)
*/
typedef union
{
/* Native rwlock (is_srwlock == TRUE) */
struct
{
SRWLOCK srwlock; /* native reader writer lock */
BOOL have_exclusive_srwlock; /* used for unlock */
};
/*
Portable implementation (is_srwlock == FALSE)
Fields are identical with Unix my_rw_lock_t fields.
*/
struct
{
pthread_mutex_t lock; /* lock for structure */
pthread_cond_t readers; /* waiting readers */
pthread_cond_t writers; /* waiting writers */
int state; /* -1:writer,0:free,>0:readers */
int waiters; /* number of waiting writers */
#ifdef SAFE_MUTEX
pthread_t write_thread;
#endif
};
} my_rw_lock_t;
#else /* _WIN32 */
/*
A better fix for bug #56405 "Deadlock in the MDL deadlock detector" that doesn't introduce bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors". Deadlock could have occurred when workload containing a mix of DML, DDL and FLUSH TABLES statements affecting the same set of tables was executed in a heavily concurrent environment. This deadlock occurred when several connections tried to perform deadlock detection in the metadata locking subsystem. The first connection started traversing wait-for graph, encountered a sub-graph representing a wait for flush, acquired LOCK_open and dived into sub-graph inspection. Then it encountered sub-graph corresponding to wait for metadata lock and blocked while trying to acquire a rd-lock on MDL_lock::m_rwlock, since some,other thread had a wr-lock on it. When this wr-lock was released it could have happened (if there was another pending wr-lock against this rwlock) that the rd-lock from the first connection was left unsatisfied but at the same time the new rd-lock request from the second connection sneaked in and was satisfied (for this to be possible the second rd-request should come exactly after the wr-lock is released but before pending the wr-lock manages to grab rwlock, which is possible both on Linux and in our own rwlock implementation). If this second connection continued traversing the wait-for graph and encountered a sub-graph representing a wait for flush it tried to acquire LOCK_open and thus the deadlock was created. The previous patch tried to workaround this problem by not allowing the deadlock detector to lock LOCK_open mutex if some other thread doing deadlock detection already owns it and current search depth is greater than 0. Instead deadlock was reported. As a result it has introduced bug #56715. This patch solves this problem in a different way. It introduces a new rw_pr_lock_t implementation to be used by MDL subsystem instead of one based on Linux rwlocks or our own rwlock implementation. This new implementation never allows situation in which an rwlock is rd-locked and there is a blocked pending rd-lock. Thus the situation which has caused this bug becomes impossible with this implementation. Due to fact that this implementation is optimized for wr-lock/unlock scenario which is most common in the MDL subsystem it doesn't introduce noticeable performance regressions in sysbench tests. Moreover it significantly improves situation for POINT_SELECT test when many connections are used. No test case is provided as this bug is very hard to repeat in MTR environment but is repeatable with the help of RQG tests. This patch also doesn't include a test for bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors" as it takes too much time to be run as part of normal test-suite runs.
2010-09-29 16:09:07 +04:00
On systems which don't support native read/write locks we have
to use own implementation.
*/
typedef struct st_my_rw_lock_t {
pthread_mutex_t lock; /* lock for structure */
pthread_cond_t readers; /* waiting readers */
pthread_cond_t writers; /* waiting writers */
int state; /* -1:writer,0:free,>0:readers */
int waiters; /* number of waiting writers */
#ifdef SAFE_MUTEX
pthread_t write_thread;
#endif
} my_rw_lock_t;
A follow-up to the patch for bug #56405 "Deadlock in the MDL deadlock detector". This patch addresses performance regression in OLTP_RO/MyISAM test on Windows introduced by the fix for bug #56405. Thus it makes original patch acceptable as a solution for bug #56585 "Slowdown of readonly sysbench benchmarks (e.g point_select) on Windows 5.5". With this patch, MySQL will use native Windows condition variables and reader-writer locks if they are supported by the OS. This speeds up MyISAM and the effect comes mostly from using native rwlocks. Native conditions improve scalability with higher number of concurrent users in other situations, e.g for prlocks. Benchmark numbers for this patch as measured on Win2008R2 quad core machine are attached to the bug report. ( direct link http://bugs.mysql.com/file.php?id=15883 ) Note that currently we require at least Windows7/WS2008R2 for reader-writer locks, even though native rwlock is available also on Vista. Reason is that "trylock" APIs are missing on Vista, and trylock is used in the server (in a single place in query cache). While this patch could have been written differently, to enable the native rwlock optimization also on Vista/WS2008 (e.g using native locks everywhere but portable implementation in query cache), this would come at the expense of the code clarity, as it would introduce a new "try-able" rwlock type, to handle Vista case. Another way to improve performance for the special case (OLTP_RO/MYISAM/Vista) would be to eliminate "trylock" usage from server, but this is outside of the scope here. Native conditions variables are used beginning with Vista though the effect of using condition variables alone is not measurable in this benchmark. But when used together with native rwlocks on Win7, native conditions improve performance in high-concurrency OLTP_RO/MyISAM (128 and more sysbench users).
2010-10-04 13:03:11 +02:00
#endif /*! _WIN32 */
A better fix for bug #56405 "Deadlock in the MDL deadlock detector" that doesn't introduce bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors". Deadlock could have occurred when workload containing a mix of DML, DDL and FLUSH TABLES statements affecting the same set of tables was executed in a heavily concurrent environment. This deadlock occurred when several connections tried to perform deadlock detection in the metadata locking subsystem. The first connection started traversing wait-for graph, encountered a sub-graph representing a wait for flush, acquired LOCK_open and dived into sub-graph inspection. Then it encountered sub-graph corresponding to wait for metadata lock and blocked while trying to acquire a rd-lock on MDL_lock::m_rwlock, since some,other thread had a wr-lock on it. When this wr-lock was released it could have happened (if there was another pending wr-lock against this rwlock) that the rd-lock from the first connection was left unsatisfied but at the same time the new rd-lock request from the second connection sneaked in and was satisfied (for this to be possible the second rd-request should come exactly after the wr-lock is released but before pending the wr-lock manages to grab rwlock, which is possible both on Linux and in our own rwlock implementation). If this second connection continued traversing the wait-for graph and encountered a sub-graph representing a wait for flush it tried to acquire LOCK_open and thus the deadlock was created. The previous patch tried to workaround this problem by not allowing the deadlock detector to lock LOCK_open mutex if some other thread doing deadlock detection already owns it and current search depth is greater than 0. Instead deadlock was reported. As a result it has introduced bug #56715. This patch solves this problem in a different way. It introduces a new rw_pr_lock_t implementation to be used by MDL subsystem instead of one based on Linux rwlocks or our own rwlock implementation. This new implementation never allows situation in which an rwlock is rd-locked and there is a blocked pending rd-lock. Thus the situation which has caused this bug becomes impossible with this implementation. Due to fact that this implementation is optimized for wr-lock/unlock scenario which is most common in the MDL subsystem it doesn't introduce noticeable performance regressions in sysbench tests. Moreover it significantly improves situation for POINT_SELECT test when many connections are used. No test case is provided as this bug is very hard to repeat in MTR environment but is repeatable with the help of RQG tests. This patch also doesn't include a test for bug #56715 "Concurrent transactions + FLUSH result in sporadical unwarranted deadlock errors" as it takes too much time to be run as part of normal test-suite runs.
2010-09-29 16:09:07 +04:00
extern int my_rw_init(my_rw_lock_t *);
extern int my_rw_destroy(my_rw_lock_t *);
extern int my_rw_rdlock(my_rw_lock_t *);
extern int my_rw_wrlock(my_rw_lock_t *);
extern int my_rw_unlock(my_rw_lock_t *);
extern int my_rw_tryrdlock(my_rw_lock_t *);
extern int my_rw_trywrlock(my_rw_lock_t *);
#ifdef SAFE_MUTEX
#define my_rw_lock_assert_write_owner(A) \
DBUG_ASSERT((A)->state == -1 && pthread_equal(pthread_self(), \
(A)->write_thread))
#define my_rw_lock_assert_not_write_owner(A) \
DBUG_ASSERT((A)->state >= 0 || ! pthread_equal(pthread_self(), \
(A)->write_thread))
#else
#define my_rw_lock_assert_write_owner(A)
#define my_rw_lock_assert_not_write_owner(A)
#endif
#endif /* NEED_MY_RW_LOCK */
2000-07-31 21:29:14 +02:00
#define GETHOSTBYADDR_BUFF_SIZE 2048
#ifndef HAVE_THR_SETCONCURRENCY
#define thr_setconcurrency(A) pthread_dummy(0)
#endif
#if !defined(HAVE_PTHREAD_ATTR_SETSTACKSIZE) && ! defined(pthread_attr_setstacksize)
#define pthread_attr_setstacksize(A,B) pthread_dummy(0)
#endif
/* Define mutex types, see my_thr_init.c */
#define MY_MUTEX_INIT_SLOW NULL
#ifdef PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP
extern pthread_mutexattr_t my_fast_mutexattr;
#define MY_MUTEX_INIT_FAST &my_fast_mutexattr
#else
#define MY_MUTEX_INIT_FAST NULL
#endif
#ifdef PTHREAD_ERRORCHECK_MUTEX_INITIALIZER_NP
extern pthread_mutexattr_t my_errorcheck_mutexattr;
#define MY_MUTEX_INIT_ERRCHK &my_errorcheck_mutexattr
#else
#define MY_MUTEX_INIT_ERRCHK NULL
#endif
#ifndef ESRCH
/* Define it to something */
#define ESRCH 1
#endif
Fixed compiler warnings Fixed compile-pentium64 scripts Fixed wrong estimate of update_with_key_prefix in sql-bench Merge bk-internal.mysql.com:/home/bk/mysql-5.1 into mysql.com:/home/my/mysql-5.1 Fixed unsafe define of uint4korr() Fixed that --extern works with mysql-test-run.pl Small trivial cleanups This also fixes a bug in counting number of rows that are updated when we have many simultanous queries Move all connection handling and command exectuion main loop from sql_parse.cc to sql_connection.cc Split handle_one_connection() into reusable sub functions. Split create_new_thread() into reusable sub functions. Added thread_scheduler; Preliminary interface code for future thread_handling code. Use 'my_thread_id' for internal thread id's Make thr_alarm_kill() to depend on thread_id instead of thread Make thr_abort_locks_for_thread() depend on thread_id instead of thread In store_globals(), set my_thread_var->id to be thd->thread_id. Use my_thread_var->id as basis for my_thread_name() The above changes makes the connection we have between THD and threads more soft. Added a lot of DBUG_PRINT() and DBUG_ASSERT() functions Fixed compiler warnings Fixed core dumps when running with --debug Removed setting of signal masks (was never used) Made event code call pthread_exit() (portability fix) Fixed that event code doesn't call DBUG_xxx functions before my_thread_init() is called. Made handling of thread_id and thd->variables.pseudo_thread_id uniform. Removed one common 'not freed memory' warning from mysqltest Fixed a couple of usage of not initialized warnings (unlikely cases) Suppress compiler warnings from bdb and (for the moment) warnings from ndb
2007-02-23 13:13:55 +02:00
typedef ulong my_thread_id;
2000-07-31 21:29:14 +02:00
extern my_bool my_thread_global_init(void);
extern my_bool my_thread_basic_global_init(void);
extern void my_thread_basic_global_reinit(void);
2000-07-31 21:29:14 +02:00
extern void my_thread_global_end(void);
extern my_bool my_thread_init(void);
extern void my_thread_end(void);
extern const char *my_thread_name(void);
Fixed compiler warnings Fixed compile-pentium64 scripts Fixed wrong estimate of update_with_key_prefix in sql-bench Merge bk-internal.mysql.com:/home/bk/mysql-5.1 into mysql.com:/home/my/mysql-5.1 Fixed unsafe define of uint4korr() Fixed that --extern works with mysql-test-run.pl Small trivial cleanups This also fixes a bug in counting number of rows that are updated when we have many simultanous queries Move all connection handling and command exectuion main loop from sql_parse.cc to sql_connection.cc Split handle_one_connection() into reusable sub functions. Split create_new_thread() into reusable sub functions. Added thread_scheduler; Preliminary interface code for future thread_handling code. Use 'my_thread_id' for internal thread id's Make thr_alarm_kill() to depend on thread_id instead of thread Make thr_abort_locks_for_thread() depend on thread_id instead of thread In store_globals(), set my_thread_var->id to be thd->thread_id. Use my_thread_var->id as basis for my_thread_name() The above changes makes the connection we have between THD and threads more soft. Added a lot of DBUG_PRINT() and DBUG_ASSERT() functions Fixed compiler warnings Fixed core dumps when running with --debug Removed setting of signal masks (was never used) Made event code call pthread_exit() (portability fix) Fixed that event code doesn't call DBUG_xxx functions before my_thread_init() is called. Made handling of thread_id and thd->variables.pseudo_thread_id uniform. Removed one common 'not freed memory' warning from mysqltest Fixed a couple of usage of not initialized warnings (unlikely cases) Suppress compiler warnings from bdb and (for the moment) warnings from ndb
2007-02-23 13:13:55 +02:00
extern my_thread_id my_thread_dbug_id(void);
2000-07-31 21:29:14 +02:00
extern int pthread_dummy(int);
/* All thread specific variables are in the following struct */
#define THREAD_NAME_SIZE 10
#ifndef DEFAULT_THREAD_STACK
#if SIZEOF_CHARP > 4
/*
MySQL can survive with 32K, but some glibc libraries require > 128K stack
To resolve hostnames. Also recursive stored procedures needs stack.
*/
#define DEFAULT_THREAD_STACK (256*1024L)
2000-09-07 04:55:17 +03:00
#else
#define DEFAULT_THREAD_STACK (192*1024)
2000-09-07 04:55:17 +03:00
#endif
#endif
2000-07-31 21:29:14 +02:00
#include <mysql/psi/mysql_thread.h>
#define INSTRUMENT_ME 0
2000-07-31 21:29:14 +02:00
struct st_my_thread_var
{
int thr_errno;
mysql_cond_t suspend;
mysql_mutex_t mutex;
mysql_mutex_t * volatile current_mutex;
mysql_cond_t * volatile current_cond;
2000-07-31 21:29:14 +02:00
pthread_t pthread_self;
Fixed compiler warnings Fixed compile-pentium64 scripts Fixed wrong estimate of update_with_key_prefix in sql-bench Merge bk-internal.mysql.com:/home/bk/mysql-5.1 into mysql.com:/home/my/mysql-5.1 Fixed unsafe define of uint4korr() Fixed that --extern works with mysql-test-run.pl Small trivial cleanups This also fixes a bug in counting number of rows that are updated when we have many simultanous queries Move all connection handling and command exectuion main loop from sql_parse.cc to sql_connection.cc Split handle_one_connection() into reusable sub functions. Split create_new_thread() into reusable sub functions. Added thread_scheduler; Preliminary interface code for future thread_handling code. Use 'my_thread_id' for internal thread id's Make thr_alarm_kill() to depend on thread_id instead of thread Make thr_abort_locks_for_thread() depend on thread_id instead of thread In store_globals(), set my_thread_var->id to be thd->thread_id. Use my_thread_var->id as basis for my_thread_name() The above changes makes the connection we have between THD and threads more soft. Added a lot of DBUG_PRINT() and DBUG_ASSERT() functions Fixed compiler warnings Fixed core dumps when running with --debug Removed setting of signal masks (was never used) Made event code call pthread_exit() (portability fix) Fixed that event code doesn't call DBUG_xxx functions before my_thread_init() is called. Made handling of thread_id and thd->variables.pseudo_thread_id uniform. Removed one common 'not freed memory' warning from mysqltest Fixed a couple of usage of not initialized warnings (unlikely cases) Suppress compiler warnings from bdb and (for the moment) warnings from ndb
2007-02-23 13:13:55 +02:00
my_thread_id id;
2000-09-22 01:46:26 +03:00
int cmp_length;
2001-09-01 10:38:16 +03:00
int volatile abort;
my_bool init;
struct st_my_thread_var *next,**prev;
void *opt_info;
void *stack_ends_here;
2000-07-31 21:29:14 +02:00
#ifndef DBUG_OFF
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 12:59:39 +03:00
void *dbug;
2000-09-22 01:46:26 +03:00
char name[THREAD_NAME_SIZE+1];
2000-07-31 21:29:14 +02:00
#endif
};
extern struct st_my_thread_var *_my_thread_var(void) __attribute__ ((const));
extern void **my_thread_var_dbug();
extern uint my_thread_end_wait_time;
2000-07-31 21:29:14 +02:00
#define my_thread_var (_my_thread_var())
#define my_errno my_thread_var->thr_errno
/*
Keep track of shutdown,signal, and main threads so that my_end() will not
report errors with them
*/
2007-03-28 15:33:29 +02:00
/* Which kind of thread library is in use */
#define THD_LIB_OTHER 1
#define THD_LIB_NPTL 2
#define THD_LIB_LT 4
extern uint thd_lib_detected;
/*
thread_safe_xxx functions are for critical statistic or counters.
The implementation is guaranteed to be thread safe, on all platforms.
Note that the calling code should *not* assume the counter is protected
by the mutex given, as the implementation of these helpers may change
to use my_atomic operations instead.
*/
/*
Warning:
When compiling without threads, this file is not included.
See the *other* declarations of thread_safe_xxx in include/my_global.h
*/
#ifdef THREAD
2007-05-22 16:19:36 -06:00
#ifndef thread_safe_increment
#ifdef _WIN32
#define thread_safe_increment(V,L) InterlockedIncrement((long*) &(V))
#define thread_safe_decrement(V,L) InterlockedDecrement((long*) &(V))
#else
2000-07-31 21:29:14 +02:00
#define thread_safe_increment(V,L) \
(mysql_mutex_lock((L)), (V)++, mysql_mutex_unlock((L)))
#define thread_safe_decrement(V,L) \
(mysql_mutex_lock((L)), (V)--, mysql_mutex_unlock((L)))
2007-05-22 16:19:36 -06:00
#endif
#endif
2007-05-22 16:19:36 -06:00
#ifndef thread_safe_add
#ifdef _WIN32
#define thread_safe_add(V,C,L) InterlockedExchangeAdd((long*) &(V),(C))
#define thread_safe_sub(V,C,L) InterlockedExchangeAdd((long*) &(V),-(long) (C))
#else
#define thread_safe_add(V,C,L) \
(mysql_mutex_lock((L)), (V)+=(C), mysql_mutex_unlock((L)))
2000-07-31 21:29:14 +02:00
#define thread_safe_sub(V,C,L) \
(mysql_mutex_lock((L)), (V)-=(C), mysql_mutex_unlock((L)))
#endif
2007-05-22 16:19:36 -06:00
#endif
#endif
/*
statistics_xxx functions are for non critical statistic,
maintained in global variables.
When compiling with SAFE_STATISTICS:
- race conditions can not occur.
- some locking occurs, which may cause performance degradation.
When compiling without SAFE_STATISTICS:
- race conditions can occur, making the result slightly inaccurate.
- the lock given is not honored.
*/
2000-07-31 21:29:14 +02:00
#ifdef SAFE_STATISTICS
#define statistic_increment(V,L) thread_safe_increment((V),(L))
#define statistic_decrement(V,L) thread_safe_decrement((V),(L))
#define statistic_add(V,C,L) thread_safe_add((V),(C),(L))
#define statistic_sub(V,C,L) thread_safe_sub((V),(C),(L))
2000-07-31 21:29:14 +02:00
#else
#define statistic_decrement(V,L) (V)--
2000-07-31 21:29:14 +02:00
#define statistic_increment(V,L) (V)++
#define statistic_add(V,C,L) (V)+=(C)
#define statistic_sub(V,C,L) (V)-=(C)
2000-07-31 21:29:14 +02:00
#endif /* SAFE_STATISTICS */
/*
No locking needed, the counter is owned by the thread
*/
#define status_var_increment(V) (V)++
#define status_var_decrement(V) (V)--
#define status_var_add(V,C) (V)+=(C)
#define status_var_sub(V,C) (V)-=(C)
2001-08-22 14:15:01 +02:00
#ifdef __cplusplus
}
#endif
2000-07-31 21:29:14 +02:00
#endif /* _my_ptread_h */