Rainer Gerhards 3250b21566 core/wtp: potential hang during shutdown
when the wtp worker is cancelled in the final stage of shutting down
while the mutex is locked and there is one worker left, the system
will hang. The reason is that the cancelled thread could not free the
mutex that the other needs in order to shut down orderly.

Deteced with clang thread sanitizer.
2017-11-10 19:03:09 +01:00

601 lines
18 KiB
C

/* wtp.c
*
* This file implements the worker thread pool (wtp) class.
*
* File begun on 2008-01-20 by RGerhards
*
* There is some in-depth documentation available in doc/dev_queue.html
* (and in the web doc set on http://www.rsyslog.com/doc). Be sure to read it
* if you are getting aquainted to the object.
*
* Copyright 2008-2017 Rainer Gerhards and Adiscon GmbH.
*
* This file is part of the rsyslog runtime library.
*
* The rsyslog runtime library is free software: you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* The rsyslog runtime library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public License
* along with the rsyslog runtime library. If not, see <http://www.gnu.org/licenses/>.
*
* A copy of the GPL can be found in the file "COPYING" in this distribution.
* A copy of the LGPL can be found in the file "COPYING.LESSER" in this distribution.
*/
#include "config.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include <signal.h>
#include <pthread.h>
#include <fcntl.h>
#include <unistd.h>
#include <errno.h>
#include <atomic.h>
#ifdef HAVE_SYS_PRCTL_H
# include <sys/prctl.h>
#endif
/// TODO: check on solaris if this is any longer needed - I don't think so - rgerhards, 2009-09-20
//#ifdef OS_SOLARIS
//# include <sched.h>
//#endif
#include "rsyslog.h"
#include "stringbuf.h"
#include "srUtils.h"
#include "wtp.h"
#include "wti.h"
#include "obj.h"
#include "unicode-helper.h"
#include "glbl.h"
#include "errmsg.h"
/* static data */
DEFobjStaticHelpers
DEFobjCurrIf(glbl)
/* forward-definitions */
/* methods */
/* get the header for debug messages
* The caller must NOT free or otherwise modify the returned string!
*/
static uchar *
wtpGetDbgHdr(wtp_t *pThis)
{
ISOBJ_TYPE_assert(pThis, wtp);
if(pThis->pszDbgHdr == NULL)
return (uchar*) "wtp"; /* should not normally happen */
else
return pThis->pszDbgHdr;
}
/* Not implemented dummy function for constructor */
static rsRetVal NotImplementedDummy(void) { return RS_RET_NOT_IMPLEMENTED; }
/* Standard-Constructor for the wtp object
*/
BEGINobjConstruct(wtp) /* be sure to specify the object type also in END macro! */
pthread_mutex_init(&pThis->mutWtp, NULL);
pthread_cond_init(&pThis->condThrdInitDone, NULL);
pthread_cond_init(&pThis->condThrdTrm, NULL);
pthread_attr_init(&pThis->attrThrd);
/* Set thread scheduling policy to default */
#ifdef HAVE_PTHREAD_SETSCHEDPARAM
pthread_attr_setschedpolicy(&pThis->attrThrd, default_thr_sched_policy);
pthread_attr_setschedparam(&pThis->attrThrd, &default_sched_param);
pthread_attr_setinheritsched(&pThis->attrThrd, PTHREAD_EXPLICIT_SCHED);
#endif
pthread_attr_setdetachstate(&pThis->attrThrd, PTHREAD_CREATE_DETACHED);
/* set all function pointers to "not implemented" dummy so that we can safely call them */
pThis->pfChkStopWrkr = (rsRetVal (*)(void*,int))NotImplementedDummy;
pThis->pfGetDeqBatchSize = (rsRetVal (*)(void*,int*))NotImplementedDummy;
pThis->pfDoWork = (rsRetVal (*)(void*,void*))NotImplementedDummy;
pThis->pfObjProcessed = (rsRetVal (*)(void*,wti_t*))NotImplementedDummy;
INIT_ATOMIC_HELPER_MUT(pThis->mutCurNumWrkThrd);
INIT_ATOMIC_HELPER_MUT(pThis->mutWtpState);
ENDobjConstruct(wtp)
/* Construction finalizer
* rgerhards, 2008-01-17
*/
rsRetVal
wtpConstructFinalize(wtp_t *pThis)
{
DEFiRet;
int i;
uchar pszBuf[64];
size_t lenBuf;
wti_t *pWti;
ISOBJ_TYPE_assert(pThis, wtp);
DBGPRINTF("%s: finalizing construction of worker thread pool (numworkerThreads %d)\n",
wtpGetDbgHdr(pThis), pThis->iNumWorkerThreads);
/* alloc and construct workers - this can only be done in finalizer as we previously do
* not know the max number of workers
*/
CHKmalloc(pThis->pWrkr = MALLOC(sizeof(wti_t*) * pThis->iNumWorkerThreads));
for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i) {
CHKiRet(wtiConstruct(&pThis->pWrkr[i]));
pWti = pThis->pWrkr[i];
lenBuf = snprintf((char*)pszBuf, sizeof(pszBuf), "%s/w%d", wtpGetDbgHdr(pThis), i);
CHKiRet(wtiSetDbgHdr(pWti, pszBuf, lenBuf));
CHKiRet(wtiSetpWtp(pWti, pThis));
CHKiRet(wtiConstructFinalize(pWti));
}
finalize_it:
RETiRet;
}
/* Destructor */
BEGINobjDestruct(wtp) /* be sure to specify the object type also in END and CODESTART macros! */
int i;
CODESTARTobjDestruct(wtp)
d_pthread_mutex_lock(&pThis->mutWtp); /* make sure nobody is still using the mutex */
assert(pThis->iCurNumWrkThrd == 0);
/* destruct workers */
for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i)
wtiDestruct(&pThis->pWrkr[i]);
free(pThis->pWrkr);
pThis->pWrkr = NULL;
/* actual destruction */
d_pthread_mutex_unlock(&pThis->mutWtp);
pthread_cond_destroy(&pThis->condThrdTrm);
pthread_cond_destroy(&pThis->condThrdInitDone);
pthread_mutex_destroy(&pThis->mutWtp);
pthread_attr_destroy(&pThis->attrThrd);
DESTROY_ATOMIC_HELPER_MUT(pThis->mutCurNumWrkThrd);
DESTROY_ATOMIC_HELPER_MUT(pThis->mutWtpState);
free(pThis->pszDbgHdr);
ENDobjDestruct(wtp)
/* Sent a specific state for the worker thread pool. -- rgerhards, 2008-01-21
* We do not need to do atomic instructions as set operations are only
* called when terminating the pool, and then in strict sequence. So we
* can never overwrite each other. On the other hand, it also doesn't
* matter if the read operation obtains an older value, as we then simply
* do one more iteration, what is perfectly legal (during shutdown
* they are awoken in any case). -- rgerhards, 2009-07-20
*/
rsRetVal
wtpSetState(wtp_t *pThis, wtpState_t iNewState)
{
ISOBJ_TYPE_assert(pThis, wtp);
pThis->wtpState = iNewState; // TODO: do we need a mutex here? 2010-04-26
return RS_RET_OK;
}
/* check if the worker shall shutdown (1 = yes, 0 = no)
* Note: there may be two mutexes locked, the bLockUsrMutex is the one in our "user"
* (e.g. the queue clas)
* rgerhards, 2008-01-21
*/
rsRetVal
wtpChkStopWrkr(wtp_t *pThis, int bLockUsrMutex)
{
DEFiRet;
wtpState_t wtpState;
ISOBJ_TYPE_assert(pThis, wtp);
/* we need a consistent value, but it doesn't really matter if it is changed
* right after the fetch - then we simply do one more iteration in the worker
*/
wtpState = (wtpState_t) ATOMIC_FETCH_32BIT((int*)&pThis->wtpState, &pThis->mutWtpState);
if(wtpState == wtpState_SHUTDOWN_IMMEDIATE) {
ABORT_FINALIZE(RS_RET_TERMINATE_NOW);
} else if(wtpState == wtpState_SHUTDOWN) {
ABORT_FINALIZE(RS_RET_TERMINATE_WHEN_IDLE);
}
/* try customer handler if one was set and we do not yet have a definite result */
if(pThis->pfChkStopWrkr != NULL) {
iRet = pThis->pfChkStopWrkr(pThis->pUsr, bLockUsrMutex);
}
finalize_it:
RETiRet;
}
#if !defined(_AIX)
#pragma GCC diagnostic ignored "-Wempty-body"
#endif
/* Send a shutdown command to all workers and see if they terminate.
* A timeout may be specified. This function may also be called with
* the current number of workers being 0, in which case it does not
* shut down any worker.
* rgerhards, 2008-01-14
*/
rsRetVal ATTR_NONNULL()
wtpShutdownAll(wtp_t *pThis, wtpState_t tShutdownCmd, struct timespec *ptTimeout)
{
DEFiRet;
int bTimedOut;
int i;
ISOBJ_TYPE_assert(pThis, wtp);
/* lock mutex to prevent races (may otherwise happen during idle processing and such...) */
d_pthread_mutex_lock(pThis->pmutUsr);
wtpSetState(pThis, tShutdownCmd);
/* awake workers in retry loop */
for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i) {
pthread_cond_signal(&pThis->pWrkr[i]->pcondBusy);
wtiWakeupThrd(pThis->pWrkr[i]);
}
d_pthread_mutex_unlock(pThis->pmutUsr);
/* wait for worker thread termination */
d_pthread_mutex_lock(&pThis->mutWtp);
pthread_cleanup_push(mutexCancelCleanup, &pThis->mutWtp);
bTimedOut = 0;
while(pThis->iCurNumWrkThrd > 0 && !bTimedOut) {
DBGPRINTF("%s: waiting %ldms on worker thread termination, %d still running\n",
wtpGetDbgHdr(pThis), timeoutVal(ptTimeout),
ATOMIC_FETCH_32BIT(&pThis->iCurNumWrkThrd, &pThis->mutCurNumWrkThrd));
if(d_pthread_cond_timedwait(&pThis->condThrdTrm, &pThis->mutWtp, ptTimeout) != 0) {
DBGPRINTF("%s: timeout waiting on worker thread termination\n",
wtpGetDbgHdr(pThis));
bTimedOut = 1; /* we exit the loop on timeout */
}
/* awake workers in retry loop */
for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i) {
wtiWakeupThrd(pThis->pWrkr[i]);
}
}
pthread_cleanup_pop(1);
if(bTimedOut)
iRet = RS_RET_TIMED_OUT;
RETiRet;
}
#if !defined(_AIX)
#pragma GCC diagnostic warning "-Wempty-body"
#endif
/* Unconditionally cancel all running worker threads.
* rgerhards, 2008-01-14
*/
rsRetVal
wtpCancelAll(wtp_t *pThis)
{
DEFiRet;
int i;
ISOBJ_TYPE_assert(pThis, wtp);
/* go through all workers and cancel those that are active */
for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i) {
wtiCancelThrd(pThis->pWrkr[i]);
}
RETiRet;
}
/* this function contains shared code for both regular worker shutdown as
* well as shutdown via cancellation. We can not simply use pthread_cleanup_pop(1)
* as this introduces a race in the debug system (RETiRet system).
* rgerhards, 2009-10-26
*/
static void
wtpWrkrExecCleanup(wti_t *pWti)
{
wtp_t *pThis;
BEGINfunc
ISOBJ_TYPE_assert(pWti, wti);
pThis = pWti->pWtp;
ISOBJ_TYPE_assert(pThis, wtp);
/* the order of the next two statements is important! */
wtiSetState(pWti, WRKTHRD_STOPPED);
ATOMIC_DEC(&pThis->iCurNumWrkThrd, &pThis->mutCurNumWrkThrd);
/* note: numWorkersNow is only for message generation, so we do not try
* hard to get it 100% accurate (as curently done, it is not).
*/
const int numWorkersNow = ATOMIC_FETCH_32BIT(&pThis->iCurNumWrkThrd, &pThis->mutCurNumWrkThrd);
DBGPRINTF("%s: Worker thread %lx, terminated, num workers now %d\n",
wtpGetDbgHdr(pThis), (unsigned long) pWti, numWorkersNow);
if(numWorkersNow > 0) {
LogMsg(0, RS_RET_OPERATION_STATUS, LOG_INFO,
"%s: worker thread %lx terminated, now %d active worker threads",
wtpGetDbgHdr(pThis), (unsigned long) pWti, numWorkersNow);
}
ENDfunc
}
/* cancellation cleanup handler for executing worker decrements the worker counter.
* rgerhards, 2009-07-20
*/
static void
wtpWrkrExecCancelCleanup(void *arg)
{
wti_t *pWti = (wti_t*) arg;
wtp_t *pThis;
BEGINfunc
ISOBJ_TYPE_assert(pWti, wti);
pThis = pWti->pWtp;
ISOBJ_TYPE_assert(pThis, wtp);
DBGPRINTF("%s: Worker thread %lx requested to be cancelled.\n",
wtpGetDbgHdr(pThis), (unsigned long) pWti);
wtpWrkrExecCleanup(pWti);
ENDfunc
/* NOTE: we must call ENDfunc FIRST, because otherwise the schedule may activate the main
* thread after the broadcast, which could destroy the debug class, resulting in a potential
* segfault. So we need to do the broadcast as actually the last action in our processing
*/
pthread_cond_broadcast(&pThis->condThrdTrm); /* activate anyone waiting on thread shutdown */
}
/* wtp worker shell. This is started and calls into the actual
* wti worker.
* rgerhards, 2008-01-21
*/
#if !defined(_AIX)
#pragma GCC diagnostic ignored "-Wempty-body"
#endif
static void *
wtpWorker(void *arg) /* the arg is actually a wti object, even though we are in wtp! */
{
wti_t *pWti = (wti_t*) arg;
wtp_t *pThis;
sigset_t sigSet;
# if defined(HAVE_PRCTL) && defined(PR_SET_NAME)
uchar *pszDbgHdr;
uchar thrdName[32] = "rs:";
# endif
BEGINfunc
ISOBJ_TYPE_assert(pWti, wti);
pThis = pWti->pWtp;
ISOBJ_TYPE_assert(pThis, wtp);
/* block all signals except SIGTTIN and SIGSEGV */
sigfillset(&sigSet);
sigdelset(&sigSet, SIGTTIN);
sigdelset(&sigSet, SIGSEGV);
pthread_sigmask(SIG_BLOCK, &sigSet, NULL);
# if defined(HAVE_PRCTL) && defined(PR_SET_NAME)
/* set thread name - we ignore if the call fails, has no harsh consequences... */
pszDbgHdr = wtpGetDbgHdr(pThis);
ustrncpy(thrdName+3, pszDbgHdr, 20);
if(prctl(PR_SET_NAME, thrdName, 0, 0, 0) != 0) {
DBGPRINTF("prctl failed, not setting thread name for '%s'\n", wtpGetDbgHdr(pThis));
}
dbgOutputTID((char*)thrdName);
# endif
/* let the parent know we're done with initialization */
d_pthread_mutex_lock(&pThis->mutWtp);
wtiSetState(pWti, WRKTHRD_RUNNING);
pthread_cond_broadcast(&pThis->condThrdInitDone);
d_pthread_mutex_unlock(&pThis->mutWtp);
pthread_cleanup_push(wtpWrkrExecCancelCleanup, pWti);
wtiWorker(pWti);
pthread_cleanup_pop(0);
d_pthread_mutex_lock(&pThis->mutWtp);
pthread_cleanup_push(mutexCancelCleanup, &pThis->mutWtp);
wtpWrkrExecCleanup(pWti);
ENDfunc
/* NOTE: we must call ENDfunc FIRST, because otherwise the schedule may activate the main
* thread after the broadcast, which could destroy the debug class, resulting in a potential
* segfault. So we need to do the broadcast as actually the last action in our processing
*/
pthread_cond_broadcast(&pThis->condThrdTrm); /* activate anyone waiting on thread shutdown */
pthread_cleanup_pop(1); /* unlock mutex */
pthread_exit(0);
}
#if !defined(_AIX)
#pragma GCC diagnostic warning "-Wempty-body"
#endif
/* start a new worker */
static rsRetVal
wtpStartWrkr(wtp_t *pThis)
{
wti_t *pWti;
int i;
int iState;
DEFiRet;
ISOBJ_TYPE_assert(pThis, wtp);
d_pthread_mutex_lock(&pThis->mutWtp);
/* find free spot in thread table. */
for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i) {
if(wtiGetState(pThis->pWrkr[i]) == WRKTHRD_STOPPED) {
break;
}
}
if(i == pThis->iNumWorkerThreads)
ABORT_FINALIZE(RS_RET_NO_MORE_THREADS);
if(i == 0 || pThis->toWrkShutdown == -1) {
wtiSetAlwaysRunning(pThis->pWrkr[i]);
}
pWti = pThis->pWrkr[i];
wtiSetState(pWti, WRKTHRD_INITIALIZING);
iState = pthread_create(&(pWti->thrdID), &pThis->attrThrd,wtpWorker, (void*) pWti);
ATOMIC_INC(&pThis->iCurNumWrkThrd, &pThis->mutCurNumWrkThrd); /* we got one more! */
DBGPRINTF("%s: started with state %d, num workers now %d\n",
wtpGetDbgHdr(pThis), iState,
ATOMIC_FETCH_32BIT(&pThis->iCurNumWrkThrd, &pThis->mutCurNumWrkThrd));
/* wait for the new thread to initialize its signal mask and
* cancelation cleanup handler before proceeding
*/
do {
d_pthread_cond_wait(&pThis->condThrdInitDone, &pThis->mutWtp);
} while(wtiGetState(pWti) != WRKTHRD_RUNNING);
finalize_it:
d_pthread_mutex_unlock(&pThis->mutWtp);
RETiRet;
}
/* set the number of worker threads that should be running. If less than currently running,
* a new worker may be started. Please note that there is no guarantee the number of workers
* said will be running after we exit this function. It is just a hint. If the number is
* higher than one, and no worker is started, the "busy" condition is signaled to awake a worker.
* So the caller can assume that there is at least one worker re-checking if there is "work to do"
* after this function call.
* rgerhards, 2008-01-21
*/
rsRetVal
wtpAdviseMaxWorkers(wtp_t *pThis, int nMaxWrkr)
{
DEFiRet;
int nMissing; /* number workers missing to run */
int i, nRunning;
ISOBJ_TYPE_assert(pThis, wtp);
if(nMaxWrkr == 0)
FINALIZE;
if(nMaxWrkr > pThis->iNumWorkerThreads) /* limit to configured maximum */
nMaxWrkr = pThis->iNumWorkerThreads;
nMissing = nMaxWrkr - ATOMIC_FETCH_32BIT(&pThis->iCurNumWrkThrd, &pThis->mutCurNumWrkThrd);
if(nMissing > 0) {
if(ATOMIC_FETCH_32BIT(&pThis->iCurNumWrkThrd, &pThis->mutCurNumWrkThrd) > 0) {
LogMsg(0, RS_RET_OPERATION_STATUS, LOG_INFO,
"%s: high activity - starting %d additional worker thread(s), "
"currently %d active worker threads.",
wtpGetDbgHdr(pThis), nMissing,
ATOMIC_FETCH_32BIT(&pThis->iCurNumWrkThrd,
&pThis->mutCurNumWrkThrd) );
}
/* start the rqtd nbr of workers */
for(i = 0 ; i < nMissing ; ++i) {
CHKiRet(wtpStartWrkr(pThis));
}
} else {
/* we have needed number of workers, but they may be sleeping */
for(i = 0, nRunning = 0; i < pThis->iNumWorkerThreads && nRunning < nMaxWrkr; ++i) {
if (wtiGetState(pThis->pWrkr[i]) != WRKTHRD_STOPPED) {
pthread_cond_signal(&pThis->pWrkr[i]->pcondBusy);
nRunning++;
}
}
}
finalize_it:
RETiRet;
}
/* some simple object access methods */
DEFpropSetMeth(wtp, toWrkShutdown, long)
DEFpropSetMeth(wtp, wtpState, wtpState_t)
DEFpropSetMeth(wtp, iNumWorkerThreads, int)
DEFpropSetMeth(wtp, pUsr, void*)
DEFpropSetMethPTR(wtp, pmutUsr, pthread_mutex_t)
DEFpropSetMethFP(wtp, pfChkStopWrkr, rsRetVal(*pVal)(void*, int))
DEFpropSetMethFP(wtp, pfRateLimiter, rsRetVal(*pVal)(void*))
DEFpropSetMethFP(wtp, pfGetDeqBatchSize, rsRetVal(*pVal)(void*, int*))
DEFpropSetMethFP(wtp, pfDoWork, rsRetVal(*pVal)(void*, void*))
DEFpropSetMethFP(wtp, pfObjProcessed, rsRetVal(*pVal)(void*, wti_t*))
/* set the debug header message
* The passed-in string is duplicated. So if the caller does not need
* it any longer, it must free it. Must be called only before object is finalized.
* rgerhards, 2008-01-09
*/
rsRetVal
wtpSetDbgHdr(wtp_t *pThis, uchar *pszMsg, size_t lenMsg)
{
DEFiRet;
ISOBJ_TYPE_assert(pThis, wtp);
assert(pszMsg != NULL);
if(lenMsg < 1)
ABORT_FINALIZE(RS_RET_PARAM_ERROR);
if(pThis->pszDbgHdr != NULL) {
free(pThis->pszDbgHdr);
pThis->pszDbgHdr = NULL;
}
if((pThis->pszDbgHdr = MALLOC(lenMsg + 1)) == NULL)
ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);
memcpy(pThis->pszDbgHdr, pszMsg, lenMsg + 1); /* always think about the \0! */
finalize_it:
RETiRet;
}
/* dummy */
static rsRetVal wtpQueryInterface(void) { return RS_RET_NOT_IMPLEMENTED; }
/* exit our class
*/
BEGINObjClassExit(wtp, OBJ_IS_CORE_MODULE) /* CHANGE class also in END MACRO! */
CODESTARTObjClassExit(nsdsel_gtls)
/* release objects we no longer need */
objRelease(glbl, CORE_COMPONENT);
ENDObjClassExit(wtp)
/* Initialize the stream class. Must be called as the very first method
* before anything else is called inside this class.
* rgerhards, 2008-01-09
*/
BEGINObjClassInit(wtp, 1, OBJ_IS_CORE_MODULE)
/* request objects we use */
CHKiRet(objUse(glbl, CORE_COMPONENT));
ENDObjClassInit(wtp)