1.å¦ä½è®© Qt ç¨åº Sleep
2.LockSupportçparkçå¾
çåºå±å®ç°
3.多线程死锁检测的源码分析与实现(linux c)-有向图的应用
4.å¦ä½è®© Qt çç¨åºä½¿ç¨ Sleep
5.如何杀掉空闲事务
å¦ä½è®© Qt ç¨åº Sleep
使ç¨å¹³å°ç¸å ³ç Sleep æ nanosleep 以åï¼çé¢ä¸ºä»ä¹æ²¡æååºï¼QThread ä¸æä¾äºprotected æéç sleep å½æ°ï¼å¦ä½ç¨å°ä¸»çº¿ç¨ä¸ï¼ä½¿ç¨QTest ä¸ç qSleepï¼å¨windowsä¸å¦ä½éèæ§å¶å°ï¼è¿äºé®é¢å ¶å®å½ç»ä¸ºä¸ç¹ï¼å¨ä¸»çº¿ç¨ä¸ä½¿ç¨è¿äºå½æ°æ¯ä¸ç§é误ï¼è¿ä¼ç´æ¥å¯¼è´çé¢æ æ³å·æ°ï¼ç¨æ·ä¸ç¨åºæ æ³äº¤äºãQtä¸æä¾ï¼æ¯å ä¸ºä½ ä¸éè¦å¨ä¸»çº¿ç¨ä¸ä½¿ç¨ sleep å½æ°ãå¦ä½è®©ç¨åºçå¾ ä¸æ®µæ¶é´QTimeQTime t; t.start(); while(t.elapsed()<);è¿ç§æ»å¾ªç¯ä¹æ¯ä¸ç§å¸¸è§é误ç¨æ³ãä½æ¹ææ£ç¡®çè¿æ¯æ¯è¾ç®åçï¼QTime t; t.start(); while(t.elapsed()<) QCoreApplication::processEvents();ä¸åå°å¤çäºä»¶ï¼ä»¥ä½¿å¾ç¨åºä¿æååºãQElapsedTimerè¿æ¯Qt4.7å¼å ¥çæ°çç±»ï¼åQTimeç¸æ¯ï¼å®æä¾äºæ´å¿«çè®¡ç® elapsed æ¶é´çæ¹æ³ãQElapsedTimer t; t.start(); while(t.elapsed()<) QCoreApplication::processEvents();QTest::qWaitè¿æ¯QTest模åæä¾ççå¾ å½æ°ä¸é¢æ¯å ¶æºä»£ç ï¼åæ们åé¢ç代ç å¾åå§ï¼ï¼ï¼namespace QTest { inline static void qWait(int ms) { Q_ASSERT(QCoreApplication::instance()); QElapsedTimer timer; timer.start(); do { QCoreApplication::processEvents(QEventLoop::AllEvents, ms); QTest::qSleep(); } while (timer.elapsed() < ms); } ...å ¶å®æ²¡ä»ä¹éå,对å§ï¼ä½æ¯å 为å®QTest模åï¼æ以å¨ç¨åºä¸æ们ä¸è¦ä½¿ç¨å®ãQEventLoopé åQTimer使ç¨å±é¨ç eventLoop ä¹æ¯ä¸ä¸ªä¸éçéæ©ãä¾åï¼ QEventLoop eventloop; QTimer::singleShot(, &eventloop, SLOT(quit())); eventloop.exec();QTimer å QBasicTimerè¿ä¸¤ä¸ªåæ¬æ没æä»ä¹ç´æ¥å ³ç³»ï¼QTimer估计大家é½å¾çäºãèQBasicTimer估计å¾å°æ人ç¨ãä¸QTimerç¸æ¯ï¼QBasicTimeræ´å¿«éãè½»éãåºå±ãä¸QTimerç¸æ¯ï¼å®ä¸æ¯QObjectçæ´¾çç±»ã跨平å°çsleep尽管ä¸å¼å§æ们就说äºï¼ä¸éè¦è¿ä¸ªä¸è¥¿ãä½ä¸æé¤æç§åºåä¸ï¼ä½ ç¡®å®éè¦è¿ä¸ªä¸è¥¿ã橹械æ®��èè»ã£â³indowsä¸è°ç¨Sleepï¼å ¶ä»å¹³å°è°ç¨ nanosleepï¼ï¼void QTest::qSleep(int ms) { QTEST_ASSERT(ms > 0); #ifdef Q_OS_WIN Sleep(uint(ms)); #else struct timespec ts = { ms / , (ms % ) * * }; nanosleep(&ts, NULL); #endif }çQThreadçæºç ï¼windowsä¸åæ ·ç´æ¥è°ç¨Sleepï¼ä½éwindowsçå®ç°æ¯è¿ä¸ªå°±å¤æå¤äºï¼/* \internal helper function to do thread sleeps, since usleep()/nanosleep() aren't reliable enough (in terms of behavior and availability)*/staticvoidthread_sleep(structtimespec *ti){ pthread_mutex_tmtx;pthread_cond_tcnd;pthread_mutex_init(&mtx, 0);pthread_cond_init(&cnd, 0);pthread_mutex_lock(&mtx); (void) pthread_cond_timedwait(&cnd, &mtx, ti);pthread_mutex_unlock(&mtx);pthread_cond_destroy(&cnd);pthread_mutex_destroy(&mtx);}voidQThread::sleep(unsignedlongsecs){ structtimevaltv;gettimeofday(&tv, 0);structtimespecti;ti.tv_sec = tv.tv_sec + secs;ti.tv_nsec = (tv.tv_usec * );thread_sleep(&ti);}
LockSupportçparkçå¾ çåºå±å®ç°
ä»ä¸ä¸ç¯æç« ä¸çJDKç延è¿éåä¸,æç»æ¯éè¿LockSupport.parkå®ç°çº¿ç¨ççå¾ ï¼é£ä¹åºå±æ¯å¦ä½å®ç°çå¾ åè¶ æ¶çå¾ çï¼æ¬ææ们æ¥æ¢è®¨ä¸ä¸ãLockSupportçparkåunparkçæ¹æ³publicstaticvoidpark(){ UNSAFE.park(false,0L);}publicstaticvoidparkNanos(longnanos){ if(nanos>0)UNSAFE.park(false,nanos);}publicstaticvoidunpark(Threadthread){ if(thread!=null)UNSAFE.unpark(thread);}ä»ä¸é¢å¯ä»¥çå°å®é LockSupport.parkæ¯éè¿Unsafeççparkæ¹æ³å®ç°ï¼ä»ä¸é¢çæ¹æ³å¯ä»¥çåºè¿ä¸ªæ¯ä¸ä¸ªnativeæ¹æ³.
/***Blockscurrentthread,returningwhenabalancing*{ @codeunpark}occurs,orabalancing{ @codeunpark}has*alreadyoccurred,orthethreadisinterrupted,or,ifnot*absoluteandtimeisnotzero,thegiventimenanosecondshave*elapsed,orifabsolute,thegivendeadlineinmilliseconds*sinceEpochhaspassed,orspuriously(i.e.,returningforno*"reason").Note:ThisoperationisintheUnsafeclassonly*because{ @codeunpark}is,soitwouldbestrangetoplaceit*elsewhere.*/publicnativevoidpark(booleanisAbsolute,longtime);JVMçUnsafeçparkæ¹æ³ä»ä¸é¢JDKä¸ä»£ç ä¸å¯ä»¥threadçParkerç对象çparkæ¹æ³è¿è¡ä¸æ®µæ¶é´ççå¾ ã
UNSAFE_ENTRY(void,Unsafe_Park(JNIEnv*env,jobjectunsafe,jbooleanisAbsolute,jlongtime)){ HOTSPOT_THREAD_PARK_BEGIN((uintptr_t)thread->parker(),(int)isAbsolute,time);EventThreadParkevent;JavaThreadParkedStatejtps(thread,time!=0);thread->parker()->park(isAbsolute!=0,time);if(event.should_commit()){ constoopobj=thread->current_park_blocker();if(time==0){ post_thread_park_event(&event,obj,min_jlong,min_jlong);}else{ if(isAbsolute!=0){ post_thread_park_event(&event,obj,min_jlong,time);}else{ post_thread_park_event(&event,obj,time,min_jlong);}}}HOTSPOT_THREAD_PARK_END((uintptr_t)thread->parker());}UNSAFE_ENDThread.hppçæ件ä¸å é¨å®ä¹çPark对象
private:Parker_parker;public:Parker*parker(){ return&_parker;}ä¸é¢æ¯Os_posix.cppä¸æ¯Linuxä¸å®ç°çParkçparkçå®ç°æ¹å¼
é¦å å°_counterçåééè¿CAS设置为0ï¼è¿åå°±æ§çå¼ï¼å¦æä¹åæ¯å¤§äº0ï¼å说ææ¯å 许访é®ï¼ä¸ç¨é»å¡ï¼ç´æ¥è¿åã
è·åå½å线ç¨ã
å¤æ线ç¨æ¯å¦æ¯ä¸æä¸ï¼å¦ææ¯ï¼åç´æ¥è¿åï¼(ä¹å°±æ¯è¯´çº¿ç¨å¤äºä¸æç¶æä¸ä¼å¿½ç¥parkï¼ä¸ä¼é»å¡çå¾ )
å¤æå¦æä¼ å ¥çtimeåæ°å°äº0 æè æ¯ç»å¯¹æ¶é´å¹¶ä¸timeæ¯0,åç´æ¥è¿å,(ä¸é¢çUnsafeè°ç¨parkä¼ å ¥çåæ°æ¯ falseã0ï¼æ以ä¸æ»¡è¶³è¿ç§æ åµ)
å¦ætime大äº0ï¼å转æ¢æç»å¯¹æ¶é´ã
å建ThreadBlockInVM对象ï¼å¹¶ä¸è°ç¨pthread_mutex_trylockè·å线ç¨äºæ¥éï¼å¦æ没æè·åå°éï¼åç´æ¥è¿åï¼
å¤æ_counteråéæ¯å¦å¤§äº0ï¼å¦ææ¯ï¼åéç½®_counter为0ï¼éæ¾çº¿ç¨éï¼ç´æ¥è¿åã
è°ç¨ OrderAccess::fence(); å å ¥å åå±éï¼ç¦æ¢æ令éæåºï¼ç¡®ä¿å éåéæ¾éçæ令ç顺åºã
å建OSThreadWaitState对象ï¼
å¤ætimeæ¯å¦å¤§äº0ï¼å¦ææ¯0ï¼åè°ç¨pthread_cond_waitè¿è¡çå¾ ï¼å¦æä¸æ¯0ï¼ç¶åè°ç¨pthread_cond_timedwaitè¿è¡æ¶é´åæ°ä¸ºabsTimeççå¾ ï¼
è°ç¨pthread_mutex_unlockè¿è¡éæ¾_mutexéï¼
å次è°ç¨OrderAccess::fence()ç¦æ¢æ令éæåºã
//Parker::parkdecrementscountif>0,elsedoesacondvarwait.Unpark//setscountto1andsignalscondvar.Onlyonethreadeverwaits//onthecondvar.Contentionseenwhentryingtoparkimpliesthatsomeone//isunparkingyou,sodon'twait.Andspuriousreturnsarefine,sothere//isnoneedtotracknotifications.voidParker::park(boolisAbsolute,jlongtime){ //Optionalfast-pathcheck://Returnimmediatelyifapermitisavailable.//WedependonAtomic::xchg()havingfullbarriersemantics//sincewearedoingalock-freeupdateto_counter.if(Atomic::xchg(&_counter,0)>0)return;JavaThread*jt=JavaThread::current();//Optionaloptimization--avoidstatetransitionsifthere's//aninterruptpending.if(jt->is_interrupted(false)){ return;}//Next,demultiplex/decodetimeargumentsstructtimespecabsTime;if(time<0||(isAbsolute&&time==0)){ //don'twaitatallreturn;}if(time>0){ to_abstime(&absTime,time,isAbsolute,false);}//Entersafepointregion//Bewareofdeadlockssuchas.//Theper-threadParker::mutexisaclassicleaf-lock.//InparticularathreadmustneverblockontheThreads_lockwhile//holdingtheParker::mutex.Ifsafepointsarependingboththe//theThreadBlockInVM()CTORandDTORmaygrabThreads_lock.ThreadBlockInVMtbivm(jt);//Can'taccessinterruptstatenowthatweare_thread_blocked.Ifwe've//beeninterruptedsincewecheckedabovethen_counterwillbe>0.//Don'twaitifcannotgetlocksinceinterferencearisesfrom//unparking.if(pthread_mutex_trylock(_mutex)!=0){ return;}intstatus;if(_counter>0){ //nowaitneeded_counter=0;status=pthread_mutex_unlock(_mutex);assert_status(status==0,status,"invariant");//Paranoiatoensureourlockedandlock-freepathsinteract//correctlywitheachotherandJava-levelaccesses.OrderAccess::fence();return;}OSThreadWaitStateosts(jt->osthread(),false/*notObject.wait()*/);assert(_cur_index==-1,"invariant");if(time==0){ _cur_index=REL_INDEX;//arbitrarychoicewhennottimedstatus=pthread_cond_wait(&_cond[_cur_index],_mutex);assert_status(status==0MACOS_ONLY(||status==ETIMEDOUT),status,"cond_wait");}else{ _cur_index=isAbsolute?ABS_INDEX:REL_INDEX;status=pthread_cond_timedwait(&_cond[_cur_index],_mutex,&absTime);assert_status(status==0||status==ETIMEDOUT,status,"cond_timedwait");}_cur_index=-1;_counter=0;status=pthread_mutex_unlock(_mutex);assert_status(status==0,status,"invariant");//Paranoiatoensureourlockedandlock-freepathsinteract//correctlywitheachotherandJava-levelaccesses.OrderAccess::fence();Linuxæä½ç³»ç»æ¯å¦ä½å®ç°pthread_cond_timedwaitè¿è¡æ¶é´çå¾ çpthread_cond_timedwaitå½æ°ä½äºglibcä¸pthread_cond_wait.c, å¯ä»¥çå°æ¯è°ç¨__pthread_cond_wait_commonå®ç°
/*See__pthread_cond_wait_common.*/int___pthread_cond_timedwait(pthread_cond_t*cond,pthread_mutex_t*mutex,conststruct__timespec*abstime){ /*Checkparametervalidity.ThisshouldalsotellthecompilerthatitcanassumethatabstimeisnotNULL.*/if(!valid_nanoseconds(abstime->tv_nsec))returnEINVAL;/*RelaxedMOissufficebecauseclockIDbitisonlymodifiedinconditioncreation.*/unsignedintflags=atomic_load_relaxed(&cond->__data.__wrefs);clockid_tclockid=(flags&__PTHREAD_COND_CLOCK_MONOTONIC_MASK)?CLOCK_MONOTONIC:CLOCK_REALTIME;return__pthread_cond_wait_common(cond,mutex,clockid,abstime);}ä¸é¢__pthread_cond_wait_commonæ¯å®ç°éè¿__futex_abstimed_wait_cancelableå®ç°æ¶é´çå¾
static__always_inlineint__pthread_cond_wait_common(pthread_cond_t*cond,pthread_mutex_t*mutex,clockid_tclockid,conststruct__timespec*abstime){ ''çç¥''`err=__futex_abstimed_wait_cancelable(cond->__data.__g_signals+g,0,clockid,abstime,private);''çç¥''`}__futex_abstimed_wait_cancelableæ¯è°ç¨__futex_abstimed_wait_common
int__futex_abstimed_wait_cancelable(unsignedint*futex_word,unsignedintexpected,clockid_tclockid,conststruct__timespec*abstime,intprivate){ return__futex_abstimed_wait_common(futex_word,expected,clockid,abstime,private,true);}__futex_abstimed_wait_commonä¸é¢åæ¯éè¿å¤æå¹³å°æ¯ä½æè ä½,è°ç¨__futex_abstimed_wait_commonæè __futex_abstimed_wait_common
staticint__futex_abstimed_wait_common(unsignedint*futex_word,unsignedintexpected,clockid_tclockid,conststruct__timespec*abstime,intprivate,boolcancel){ interr;unsignedintclockbit;/*Workaroundthefactthatthekernelrejectsnegativetimeoutvaluesdespitethembeingvalid.*/if(__glibc_unlikely((abstime!=NULL)&&(abstime->tv_sec<0)))returnETIMEDOUT;if(!lll_futex_supported_clockid(clockid))returnEINVAL;clockbit=(clockid==CLOCK_REALTIME)?FUTEX_CLOCK_REALTIME:0;intop=__lll_private_flag(FUTEX_WAIT_BITSET|clockbit,private);#ifdef__ASSUME_TIME_SYSCALLSerr=__futex_abstimed_wait_common(futex_word,expected,op,abstime,private,cancel);#elseboolneed_time=abstime!=NULL&&!in_time_t_range(abstime->tv_sec);if(need_time){ err=__futex_abstimed_wait_common(futex_word,expected,op,abstime,private,cancel);if(err==-ENOSYS)err=-EOVERFLOW;}elseerr=__futex_abstimed_wait_common(futex_word,expected,op,abstime,private,cancel);#endifswitch(err){ case0:case-EAGAIN:case-EINTR:case-ETIMEDOUT:case-EINVAL:case-EOVERFLOW:/*Passedabsolutetimeoutusesbittime_ttype,butunderlyingkerneldoesnotsupportbittime_tfutexsyscalls.*/return-err;case-EFAULT:/*Musthavebeencausedbyaglibcorapplicationbug.*/case-ENOSYS:/*Musthavebeencausedbyaglibcbug.*//*Noothererrorsaredocumentedatthistime.*/default:futex_fatal_error();}}__futex_abstimed_wait_commonæ¯è°ç¨INTERNAL_SYSCALL_CANCELå®å®ä¹å®ç°
staticint__futex_abstimed_wait_common(unsignedint*futex_word,unsignedintexpected,intop,conststruct__timespec*abstime,intprivate,boolcancel){ if(cancel)returnINTERNAL_SYSCALL_CANCEL(futex_time,futex_word,op,expected,abstime,NULL/*Unused.*/,FUTEX_BITSET_MATCH_ANY);elsereturnINTERNAL_SYSCALL_CALL(futex_time,futex_word,op,expected,abstime,NULL/*Ununsed.*/,FUTEX_BITSET_MATCH_ANY);}ç³»ç»è°ç¨ççå®å®ä¹
/***Blockscurrentthread,returningwhenabalancing*{ @codeunpark}occurs,orabalancing{ @codeunpark}has*alreadyoccurred,orthethreadisinterrupted,or,ifnot*absoluteandtimeisnotzero,thegiventimenanosecondshave*elapsed,orifabsolute,thegivendeadlineinmilliseconds*sinceEpochhaspassed,orspuriously(i.e.,returningforno*"reason").Note:ThisoperationisintheUnsafeclassonly*because{ @codeunpark}is,soitwouldbestrangetoplaceit*elsewhere.*/publicnativevoidpark(booleanisAbsolute,longtime);0æ»ç»ä¸»è¦å¯¹LockSupportçparkçå¾ å®ç°çåºå±å®ç°çæµ æï¼é对äºLinuxçç³»ç»è°ç¨è¿æ²¡ææ¾å°æºç ï¼åç»ä¼ç»§ç»è·è¸ªï¼å¸ææ读è ç¥éç满å¸å¯ä»¥åç¥ä¸ï¼è°¢è°¢ã
é¾æ¥ï¼/post/多线程死锁检测的分析与实现(linux c)-有向图的应用
多线程死锁检测与有向图的应用分析与实现
死锁的检测与避免是多线程开发中关键的挑战。本文旨在深入理解死锁产生的源码条件、原因,源码以及如何通过有向图模型检测多线程程序中的源码死锁问题。
### 一、源码死锁的源码diypage源码产生与检测
死锁指的是多个进程在运行过程中因争夺资源而形成的一种僵局,当进程处于这种状态时,源码若无外力干预,源码它们将无法继续推进执行。源码通过分析死锁产生的源码条件,我们可以发现其根源主要在于竞争资源和进程间推进顺序的源码非法性。利用有向图模型,源码我们可以直观地检测多线程程序中是源码否存在死锁。
### 二、源码死锁条件与分析
死锁的源码面部识别源码下载条件包括:
- **互斥条件**:资源需要排它性控制,即一段时间内资源仅为一个进程所使用。
- **请求与保持条件**:进程请求资源时,对已获得的资源保持不释放。
- **不剥夺条件**:进程持有的资源在未使用完毕前不能被剥夺,只能在释放后由自己管理。
- **环路等待条件**:系统中存在一个进程-资源的环形依赖结构。
### 三、死锁检测方法
检测死锁的关键在于识别系统中是否存在环形依赖关系。当多线程程序中存在环形锁依赖关系时,即构成了死锁问题。我们可以通过维护有向图的状态,即在加锁前、加锁后、释放锁后进行状态更新,js解析不到源码以此来判断是否有死锁发生。
### 四、实现步骤与代码示例
在检测死锁时,可以通过自定义函数接口,如`pthread_mutex_lock`和`pthread_mutex_unlock`,在调用这些系统API之前和之后,实现加锁前和加锁后操作,以及解锁后清除对应关系。这不仅实现了检测功能,同时也使用户无需改变原有的编程习惯。
### 五、图结构与遍历
对于死锁问题,图模型提供了一种有效的可视化手段。图中的java list类源码节点代表线程和锁,边表示锁的占有关系。通过深度优先遍历(DFS)算法,我们可以检测图中是否存在环路,从而判断是否存在死锁。
### 六、源代码实现
通过编译`deadlock.c`文件,使用`gcc`命令链接`-lpthread`和`-ldl`库后执行,可以生成可执行文件`deadlock`。此步骤验证了死锁检测算法的正确性和有效性。
### 结论
本文通过深入分析死锁产生的原因、条件,以及如何利用有向图模型进行检测,为开发者提供了理论与实践相结合的指导。通过代码实现,下拉式分类源码不仅解决了多线程死锁检测的问题,也为避免死锁提供了实用的方法。未来,随着多线程应用的日益普及,这种检测方法将发挥越来越重要的作用。
å¦ä½è®© Qt çç¨åºä½¿ç¨ Sleep
Qt 为ä½æ²¡ææä¾ Sleep
论åä¸ä¸æ¶è§å°æ人é®ï¼
Qt 为ä»ä¹æ²¡ææä¾è·¨å¹³å°ç sleep å½æ°ï¼
使ç¨å¹³å°ç¸å ³ç Sleep æ nanosleep 以åï¼çé¢ä¸ºä»ä¹æ²¡æååºï¼
QThread ä¸æä¾äºprotected æéç sleep å½æ°ï¼å¦ä½ç¨å°ä¸»çº¿ç¨ä¸ï¼
ä½¿ç¨ QTest ä¸ç qSleepï¼å¨windowsä¸å¦ä½éèæ§å¶å°ï¼
è¿äºé®é¢å ¶å®å½ç»ä¸ºä¸ç¹ï¼å¨ä¸»çº¿ç¨ä¸ä½¿ç¨è¿äºå½æ°æ¯ä¸ç§é误ï¼è¿ä¼ç´æ¥å¯¼è´çé¢æ æ³å·æ°ï¼ç¨æ·ä¸ç¨åºæ æ³äº¤äºã
Qtä¸æä¾ï¼æ¯å ä¸ºä½ ä¸éè¦å¨ä¸»çº¿ç¨ä¸ä½¿ç¨ sleep å½æ°ã
å¦ä½è®©ç¨åºçå¾ ä¸æ®µæ¶é´
QTime
QTime t;
t.start();
while(t.elapsed()<);
è¿ç§æ»å¾ªç¯ä¹æ¯ä¸ç§å¸¸è§é误ç¨æ³ãä½æ¹ææ£ç¡®çè¿æ¯æ¯è¾ç®åçï¼
QTime t;
t.start();
while(t.elapsed()<)
QCoreApplication::processEvents();
ä¸åå°å¤çäºä»¶ï¼ä»¥ä½¿å¾ç¨åºä¿æååºã
QElapsedTimer
è¿æ¯Qt4.7å¼å ¥çæ°çç±»ï¼åQTimeç¸æ¯ï¼å®æä¾äºæ´å¿«çè®¡ç® elapsed æ¶é´çæ¹æ³ã
QElapsedTimer t;
t.start();
while(t.elapsed()<)
QCoreApplication::processEvents();
QTest::qWait
è¿æ¯QTest模åæä¾ççå¾ å½æ°
ä¸é¢æ¯å ¶æºä»£ç ï¼åæ们åé¢ç代ç å¾åå§ï¼ï¼ï¼
namespace QTest
{
inline static void qWait(int ms)
{
Q_ASSERT(QCoreApplication::instance());
QElapsedTimer timer;
timer.start();
do {
QCoreApplication::processEvents(QEventLoop::AllEvents, ms);
QTest::qSleep();
} while (timer.elapsed() < ms);
}
...
å ¶å®æ²¡ä»ä¹éå,对å§ï¼ä½æ¯å 为å®QTest模åï¼æ以å¨ç¨åºä¸æ们ä¸è¦ä½¿ç¨å®ã
QEventLoop
é åQTimer使ç¨å±é¨ç eventLoop ä¹æ¯ä¸ä¸ªä¸éçéæ©ãä¾åï¼
QEventLoop eventloop;
QTimer::singleShot(, &eventloop, SLOT(quit()));
eventloop.exec();
QTimer å QBasicTimer
è¿ä¸¤ä¸ªåæ¬æ没æä»ä¹ç´æ¥å ³ç³»ï¼QTimer估计大家é½å¾çäºãèQBasicTimer估计å¾å°æ人ç¨ã
ä¸QTimerç¸æ¯ï¼QBasicTimeræ´å¿«éãè½»éãåºå±ã
ä¸QTimerç¸æ¯ï¼å®ä¸æ¯QObjectçæ´¾çç±»ã
跨平å°çsleep
尽管ä¸å¼å§æ们就说äºï¼ä¸éè¦è¿ä¸ªä¸è¥¿ãä½ä¸æé¤æç§åºåä¸ï¼ä½ ç¡®å®éè¦è¿ä¸ªä¸è¥¿ãå¦ä½å®ç°ä¸ä¸ªè·¨å¹³å°ç sleep å¢ï¼
æ们ä¸å¼å§ä¹æå°äºï¼QThreadç±» å QTest模åé½æä¾äºsleepå½æ°ï¼å ¶å®æ们åªéè¦ççä»ä»¬çæºç å°±å¤äºï¼
QTest 模åä¸çå½æ°å¾ç®åï¼windowsä¸è°ç¨Sleepï¼å ¶ä»å¹³å°è°ç¨ nanosleepï¼ï¼
void QTest::qSleep(int ms)
{
QTEST_ASSERT(ms > 0);
#ifdef Q_OS_WIN
Sleep(uint(ms));
#else
struct timespec ts = { ms / , (ms % ) * * };
nanosleep(&ts, NULL);
#endif
}
çQThreadçæºç ï¼windowsä¸åæ ·ç´æ¥è°ç¨Sleepï¼ä½éwindowsçå®ç°æ¯è¿ä¸ªå°±å¤æå¤äºï¼
[cpp] view plain copy
/* /internal
helper function to do thread sleeps, since usleep()/nanosleep()
aren't reliable enough (in terms of behavior and availability)
*/
static void thread_sleep(struct timespec *ti)
{
pthread_mutex_t mtx;
pthread_cond_t cnd;
pthread_mutex_init(&mtx, 0);
pthread_cond_init(&cnd, 0);
pthread_mutex_lock(&mtx);
(void) pthread_cond_timedwait(&cnd, &mtx, ti);
pthread_mutex_unlock(&mtx);
pthread_cond_destroy(&cnd);
pthread_mutex_destroy(&mtx);
}
void QThread::sleep(unsigned long secs)
{
struct timeval tv;
gettimeofday(&tv, 0);
struct timespec ti;
ti.tv_sec = tv.tv_sec + secs;
ti.tv_nsec = (tv.tv_usec * );
thread_sleep(&ti);
}
如何杀掉空闲事务
本文内容遵从CC版权协议, 可以随意转载, 但必须以超链接形式标明文章原始出处和作者信息及版权声明网址: /tech/database/how_to_kill_idle_trx.html 我们经常遇到一个情况,就是网络断开或程序Bug导致COMMIT/ROLLBACK语句没有传到数
本文内容遵从CC版权协议, 可以随意转载, 但必须以超链接形式标明文章原始出处和作者信息及版权声明网址: /tech/database/how_to_kill_idle_trx.html
我们经常遇到一个情况,就是网络断开或程序Bug导致COMMIT/ROLLBACK语句没有传到数据库,也没有释放线程,但是线上事务锁定等待严重,连接数暴涨,尤其在测试库这种情况很多,线上也偶有发生,于是想为MySQL增加一个杀掉空闲事务的功能。
那么如何实现呢,通过MySQL Server层有很多不确定因素,最保险还是在存储引擎层实现,我们用的几乎都是InnoDB/XtraDB,所以就基于Percona来修改了,Oracle版的MySQL也可以照着修改。
需求:
1. 一个事务启动,如果事务内最后一个语句执行完超过一个时间(innodb_idle_trx_timeout),就应该关闭链接。
2. 如果事务是纯读事务,因为不加锁,所以无害,不需要关闭,保持即可。
虽然这个思路被Percona的指出Alexey Kopytov可能存在“Even though SELECT queries do not place row locks by default (there are exceptions), they can still block undo log records from being purged.”的问题,但是我们确实有场景SELECT是绝对不能kill的,除非之后的INSERT/UPDATE/DELETE发生了,所以我根据我们的业务特点来修改。
跟Percona的Yasufumi Kinoshita和Alexey Kopytov提出过纯SELECT事务不应被kill,但通过一个参数控制的方案还没有被Alexey Kopytov接受,作为通用处理我提出了用两个变量分别控制纯读事务的空闲超时时间和有锁事务的空闲超时时间,还在等待Percona的回复,因为这个方案还在测试,就先不开放修改了,当然如果你很熟悉MYSQL源码,我提出这个思路你肯定知道怎么分成这两个参数控制了。
根据这两个需求我们来设计方法,首先想到这个功能肯定是放在InnoDB Master Thread最方便,Master Thread每秒调度一次,可以顺便检查空闲事务,然后关闭,因为在事务中操作trx->mysql_thd并不安全,所以一般来说最好在InnoDB层换成Thread ID操作,并且InnoDB中除了ha_innodb.cc,其他地方不能饮用THD,所以Master Thread中需要的线程数值,都需要在ha_innodb中计算好传递整型或布尔型返回值给master thread调用。
首先,我们要增加一个参数:idle_trx_timeout,它表示事务多久没有下一条语句发生就超时关闭。
在storage/innodb_plugin/srv/srv0srv.c的“/* plugin options */”注释下增加如下代码注册idle_trx_timeout变量。
static MYSQL_SYSVAR_LONG(idle_trx_timeout, srv_idle_trx_timeout,
PLUGIN_VAR_RQCMDARG,
"If zero then this function no effect, if no-zero then wait idle_trx_timeout seconds this transaction will be closed",
"Seconds of Idle-Transaction timeout",
NULL, NULL, 0, 0, LONG_MAX, 0);
代码往下找在innobase_system_variables结构体内加上:
MYSQL_SYSVAR(idle_trx_timeout),
有了这个变量,我们需要在Master Thread(storage/innodb_plugin/srv/srv0srv.c )中执行检测函数查找空闲事务。在loop循环的if (sync_array_print_long_waits(&waiter, &sema)判断后加上这段判断
if (srv_idle_trx_timeout && trx_sys) {
trx_t* trx;
time_t now;
rescan_idle:
now = time(NULL);
mutex_enter(&kernel_mutex);
trx = UT_LIST_GET_FIRST(trx_sys->mysql_trx_list); # 从当前事务列表里获取第一个事务
while (trx) { # 依次循环每个事务进行检查
if (trx->conc_state == TRX_ACTIVE
&& trx->mysql_thd
&& innobase_thd_is_idle(trx->mysql_thd)) { # 如果事务还活着并且它的状态时空闲的
ib_int_t start_time = innobase_thd_get_start_time(trx->mysql_thd); # 获取线程最后一个语句的开始时间
ulong thd_id = innobase_thd_get_thread_id(trx->mysql_thd); #获取线程ID,因为存储引擎内直接操作THD不安全
if (trx->last_stmt_start != start_time) { # 如果事务最后语句起始时间不等于线程最后语句起始时间说明事务是新起的
trx->idle_start = now; # 更新事务的空闲起始时间
trx->last_stmt_start = start_time; # 更新事务的最后语句起始时间
} else if (difftime(now, trx->idle_start) # 如果事务不是新起的,已经执行了一部分则判断空闲时间有多长了
> srv_idle_trx_timeout) { # 如果空闲时间超过阈值则杀掉链接
/* kill the session */
mutex_exit(&kernel_mutex);
thd_kill(thd_id); # 杀链接
goto rescan_idle;
}
}
trx = UT_LIST_GET_NEXT(mysql_trx_list, trx); # 检查下一个事务
}
mutex_exit(&kernel_mutex);
}
其中trx中的变量是新加的,在storage/innodb_plugin/include/trx0trx.h的trx_truct加上需要的变量:
struct trx_struct{
...
time_t idle_start;
ib_int_t last_stmt_start;
...
}
这里有几个函数是自定义的:
ibool innobase_thd_is_idle(const void* thd);
ib_int_t innobase_thd_get_start_time(const void* thd);
ulong innobase_thd_get_thread_id(const void* thd);
这些函数在ha_innodb.cc中实现,需要在storage/innodb_plugin/srv/srv0srv.c头文件定义下加上这些函数的引用形势。
然后在storage/innodb_plugin/handler/ha_innodb.cc 中定义这些函数的实现:
extern "C"
ibool
innobase_thd_is_idle(
const void* thd) /*!{
return(((const THD*)thd)->command == COM_SLEEP);
}
extern "C"
ib_int_t
innobase_thd_get_start_time(
const void* thd) /*!{
return((ib_int_t)((const THD*)thd)->start_time);
}
extern "C"
ulong
innobase_thd_get_thread_id(
const void* thd)
{
return(thd_get_thread_id((const THD*) thd));
}
还有最重要的thd_kill函数负责杀线程的,在sql/sql_class.cc中,找个地方定义这个函数:
void thd_kill(ulong id)
{
THD *tmp;
VOID(pthread_mutex_lock(&LOCK_thread_count));
I_List_iterator it(threads);
while ((tmp=it++))
{
if (tmp->command == COM_DAEMON || tmp->is_have_lock_thd == 0 ) # 如果是DAEMON线程和不含锁的线程就不要kill了
continue;
if (tmp->thread_id == id)
{
pthread_mutex_lock(&tmp->LOCK_thd_data);
break;
}
}
VOID(pthread_mutex_unlock(&LOCK_thread_count));
if (tmp)
{
tmp->awake(THD::KILL_CONNECTION);
pthread_mutex_unlock(&tmp->LOCK_thd_data);
}
}
为了存储引擎能引用到这个函数,我们要把它定义到plugin中:
include/mysql/plugin.h和include/mysql/plugin.h中加上
void thd_kill(unsigned long id);
如何判定线程的is_have_lock_thd值?首先在THD中加上这个变量(sql/sql_class.cc):
class THD :public Statement,
public Open_tables_state
{
....
uint is_have_lock_thd;
....
}
然后在SQL的必经之路mysql_execute_command拦上一刀,判断是有锁操作发生了还是事务提交或新起事务。
switch (lex->sql_command) {
case SQLCOM_REPLACE:
case SQLCOM_REPLACE_SELECT:
case SQLCOM_UPDATE:
case SQLCOM_UPDATE_MULTI:
case SQLCOM_DELETE:
case SQLCOM_DELETE_MULTI:
case SQLCOM_INSERT:
case SQLCOM_INSERT_SELECT:
thd->is_have_lock_thd = 1;
break;
case SQLCOM_COMMIT:
case SQLCOM_ROLLBACK:
case SQLCOM_XA_START:
case SQLCOM_XA_END:
case SQLCOM_XA_PREPARE:
case SQLCOM_XA_COMMIT:
case SQLCOM_XA_ROLLBACK:
case SQLCOM_XA_RECOVER:
thd->is_have_lock_thd = 0;
break;
}
为了尽可能兼容Percona的补丁,能引用的都引用了Percona的操作,有些函数调用是在层次太多看不下去了就简化了。
另外还有一个版本是我自己弄的,在THD中增加了一个last_sql_end_time,在do_command结束后更新last_sql_end_time,然后在事务中拿到THD查看last_sql_end_time就可以得出idle时间,Oracle版我还是建议这么做,不要去改trx_struct结构体了,那个感觉更危险。
2024-11-06 13:12
2024-11-06 12:25
2024-11-06 12:18
2024-11-06 12:06
2024-11-06 11:48
2024-11-06 11:30