Files
SingularityViewer/indra/llcommon/llqueuedthread.cpp
Aleric Inglewood 2830b35aa6 Avoid dead lock in LLQueuedThread::generateHandle / LLTextureFetchWorker::callbackDecoded
Thead 1:

indra/llcommon/llqueuedthread.cpp:456:

452                     if (complete)
453                     {
454                             lockData();             // This locks LLThread::mRunCondition
455                             req->setStatus(STATUS_COMPLETE);
456                             req->finishRequest(true);

LLImageDecodeThread::ImageRequest::finishRequest calls:
    mResponder->completed(success, mDecodedImageRaw, mDecodedImageAux);

LLTextureFetchWorker::DecodeResponder::completed calls:
    worker->callbackDecoded(success, raw, aux);

LLTextureFetchWorker::callbackDecoded calls:
    LLMutexLock lock(&mWorkMutex);                      // This locks LLTextureFetchWorker::mWorkMutex

Thread 2:

LLTextureFetchWorker::doWork calls:
    LLMutexLock lock(&mWorkMutex);                      // This locks LLTextureFetchWorker::mWorkMutex
    .
    .
    .
    mDecodeHandle = mFetcher->mImageDecodeThread->decodeImage(mFormattedImage, image_priority, discard, mNeedsAux, new DecodeResponder(mFetcher, mID, this));

LLImageDecodeThread::decodeImage calls:
    handle_t handle = generateHandle();

LLQueuedThread::generateHandle calls:
    lockData();                                         // This locks LLThread::mRunCondition
2012-07-22 04:13:23 +02:00

618 lines
14 KiB
C++

/**
* @file llqueuedthread.cpp
*
* $LicenseInfo:firstyear=2004&license=viewergpl$
*
* Copyright (c) 2004-2009, Linden Research, Inc.
*
* Second Life Viewer Source Code
* The source code in this file ("Source Code") is provided by Linden Lab
* to you under the terms of the GNU General Public License, version 2.0
* ("GPL"), unless you have obtained a separate licensing agreement
* ("Other License"), formally executed by you and Linden Lab. Terms of
* the GPL can be found in doc/GPL-license.txt in this distribution, or
* online at http://secondlifegrid.net/programs/open_source/licensing/gplv2
*
* There are special exceptions to the terms and conditions of the GPL as
* it is applied to this Source Code. View the full text of the exception
* in the file doc/FLOSS-exception.txt in this software distribution, or
* online at
* http://secondlifegrid.net/programs/open_source/licensing/flossexception
*
* By copying, modifying or distributing this software, you acknowledge
* that you have read and understood your obligations described above,
* and agree to abide by those obligations.
*
* ALL LINDEN LAB SOURCE CODE IS PROVIDED "AS IS." LINDEN LAB MAKES NO
* WARRANTIES, EXPRESS, IMPLIED OR OTHERWISE, REGARDING ITS ACCURACY,
* COMPLETENESS OR PERFORMANCE.
* $/LicenseInfo$
*/
#include "linden_common.h"
#include "llqueuedthread.h"
#include "llstl.h"
#include "lltimer.h" // ms_sleep()
//============================================================================
// MAIN THREAD
LLQueuedThread::LLQueuedThread(const std::string& name, bool threaded, bool should_pause) :
LLThread(name),
mThreaded(threaded),
mIdleThread(TRUE),
mNextHandle(0),
mStarted(FALSE)
{
if (mThreaded)
{
if(should_pause)
{
pause() ; //call this before start the thread.
}
start();
}
}
// MAIN THREAD
LLQueuedThread::~LLQueuedThread()
{
if (!mThreaded)
{
endThread();
}
shutdown();
// ~LLThread() will be called here
}
void LLQueuedThread::shutdown()
{
setQuitting();
unpause(); // MAIN THREAD
if (mThreaded)
{
S32 timeout = 100;
for ( ; timeout>0; timeout--)
{
if (isStopped())
{
break;
}
ms_sleep(100);
LLThread::yield();
}
if (timeout == 0)
{
llwarns << "~LLQueuedThread (" << mName << ") timed out!" << llendl;
}
}
else
{
mStatus = STOPPED;
}
QueuedRequest* req;
S32 active_count = 0;
while ( (req = (QueuedRequest*)mRequestHash.pop_element()) )
{
if (req->getStatus() == STATUS_QUEUED || req->getStatus() == STATUS_INPROGRESS)
{
++active_count;
req->setStatus(STATUS_ABORTED); // avoid assert in deleteRequest
}
req->deleteRequest();
}
if (active_count)
{
llwarns << "~LLQueuedThread() called with active requests: " << active_count << llendl;
}
}
//----------------------------------------------------------------------------
// MAIN THREAD
// virtual
S32 LLQueuedThread::update(F32 max_time_ms)
{
if (!mStarted)
{
if (!mThreaded)
{
startThread();
mStarted = TRUE;
}
}
return updateQueue(max_time_ms);
}
S32 LLQueuedThread::updateQueue(F32 max_time_ms)
{
F64 max_time = (F64)max_time_ms * .001;
LLTimer timer;
S32 pending = 1;
// Frame Update
if (mThreaded)
{
pending = getPending();
if(pending > 0)
{
unpause();
}
}
else
{
while (pending > 0)
{
pending = processNextRequest();
if (max_time && timer.getElapsedTimeF64() > max_time)
break;
}
}
return pending;
}
void LLQueuedThread::incQueue()
{
// Something has been added to the queue
if (!isPaused())
{
if (mThreaded)
{
wake(); // Wake the thread up if necessary.
}
}
}
//virtual
// May be called from any thread
S32 LLQueuedThread::getPending()
{
S32 res;
lockData();
res = mRequestQueue.size();
unlockData();
return res;
}
// MAIN thread
void LLQueuedThread::waitOnPending()
{
while(1)
{
update(0);
if (mIdleThread)
{
break;
}
if (mThreaded)
{
yield();
}
}
return;
}
// MAIN thread
void LLQueuedThread::printQueueStats()
{
lockData();
if (!mRequestQueue.empty())
{
QueuedRequest *req = *mRequestQueue.begin();
llinfos << llformat("Pending Requests:%d Current status:%d", mRequestQueue.size(), req->getStatus()) << llendl;
}
else
{
llinfos << "Queued Thread Idle" << llendl;
}
unlockData();
}
// MAIN thread
LLQueuedThread::handle_t LLQueuedThread::generateHandle()
{
lockData();
while ((mNextHandle == nullHandle()) || (mRequestHash.find(mNextHandle)))
{
mNextHandle++;
}
const LLQueuedThread::handle_t res = mNextHandle++;
unlockData();
return res;
}
// MAIN thread
bool LLQueuedThread::addRequest(QueuedRequest* req)
{
if (mStatus == QUITTING)
{
return false;
}
lockData();
req->setStatus(STATUS_QUEUED);
mRequestQueue.insert(req);
mRequestHash.insert(req);
#if _DEBUG
// llinfos << llformat("LLQueuedThread::Added req [%08d]",handle) << llendl;
#endif
unlockData();
incQueue();
return true;
}
// MAIN thread
bool LLQueuedThread::waitForResult(LLQueuedThread::handle_t handle, bool auto_complete)
{
llassert (handle != nullHandle())
bool res = false;
bool waspaused = isPaused();
bool done = false;
while(!done)
{
update(0); // unpauses
lockData();
QueuedRequest* req = (QueuedRequest*)mRequestHash.find(handle);
if (!req)
{
done = true; // request does not exist
}
else if (req->getStatus() == STATUS_COMPLETE && !(req->getFlags() & FLAG_LOCKED))
{
res = true;
if (auto_complete)
{
mRequestHash.erase(handle);
req->deleteRequest();
// check();
}
done = true;
}
unlockData();
if (!done && mThreaded)
{
yield();
}
}
if (waspaused)
{
pause();
}
return res;
}
// MAIN thread
LLQueuedThread::QueuedRequest* LLQueuedThread::getRequest(handle_t handle)
{
if (handle == nullHandle())
{
return 0;
}
lockData();
QueuedRequest* res = (QueuedRequest*)mRequestHash.find(handle);
unlockData();
return res;
}
LLQueuedThread::status_t LLQueuedThread::getRequestStatus(handle_t handle)
{
status_t res = STATUS_EXPIRED;
lockData();
QueuedRequest* req = (QueuedRequest*)mRequestHash.find(handle);
if (req)
{
res = req->getStatus();
}
unlockData();
return res;
}
void LLQueuedThread::abortRequest(handle_t handle, bool autocomplete)
{
lockData();
QueuedRequest* req = (QueuedRequest*)mRequestHash.find(handle);
if (req)
{
req->setFlags(FLAG_ABORT | (autocomplete ? FLAG_AUTO_COMPLETE : 0));
}
unlockData();
}
// MAIN thread
void LLQueuedThread::setFlags(handle_t handle, U32 flags)
{
lockData();
QueuedRequest* req = (QueuedRequest*)mRequestHash.find(handle);
if (req)
{
req->setFlags(flags);
}
unlockData();
}
void LLQueuedThread::setPriority(handle_t handle, U32 priority)
{
lockData();
QueuedRequest* req = (QueuedRequest*)mRequestHash.find(handle);
if (req)
{
if(req->getStatus() == STATUS_INPROGRESS)
{
// not in list
req->setPriority(priority);
}
else if(req->getStatus() == STATUS_QUEUED)
{
// remove from list then re-insert
llverify(mRequestQueue.erase(req) == 1);
req->setPriority(priority);
mRequestQueue.insert(req);
}
}
unlockData();
}
bool LLQueuedThread::completeRequest(handle_t handle)
{
bool res = false;
lockData();
QueuedRequest* req = (QueuedRequest*)mRequestHash.find(handle);
if (req)
{
llassert_always(req->getStatus() != STATUS_QUEUED);
llassert_always(req->getStatus() != STATUS_INPROGRESS);
#if _DEBUG
// llinfos << llformat("LLQueuedThread::Completed req [%08d]",handle) << llendl;
#endif
if (!(req->getFlags() & FLAG_LOCKED))
{
mRequestHash.erase(handle);
req->deleteRequest();
// check();
}
else
{
// Cause deletion immediately after FLAG_LOCKED is released.
req->setFlags(FLAG_AUTO_COMPLETE);
}
res = true;
}
unlockData();
return res;
}
bool LLQueuedThread::check()
{
#if 0 // not a reliable check once mNextHandle wraps, just for quick and dirty debugging
for (int i=0; i<REQUEST_HASH_SIZE; i++)
{
LLSimpleHashEntry<handle_t>* entry = mRequestHash.get_element_at_index(i);
while (entry)
{
if (entry->getHashKey() > mNextHandle)
{
llerrs << "Hash Error" << llendl;
return false;
}
entry = entry->getNextEntry();
}
}
#endif
return true;
}
//============================================================================
// Runs on its OWN thread
S32 LLQueuedThread::processNextRequest()
{
QueuedRequest *req;
// Get next request from pool
lockData();
while(1)
{
req = NULL;
if (mRequestQueue.empty())
{
break;
}
req = *mRequestQueue.begin();
mRequestQueue.erase(mRequestQueue.begin());
if ((req->getFlags() & FLAG_ABORT) || (mStatus == QUITTING))
{
req->setStatus(STATUS_ABORTED);
// Unlock, because we can't call finishRequest() while keeping this lock:
// generateHandle() takes this lock too and is called while holding a lock
// (ie LLTextureFetchWorker::mWorkMutex) that finishRequest will lock too,
// causing a dead lock.
// Although a complete rewrite of LLQueuedThread is in order because it's
// far from robust the way it handles it's locking; the following rationale
// at least makes plausible that releasing the lock here SHOULD work if
// the original coder didn't completely fuck up: if before the QueuedRequest
// req was only accessed while keeping the lock, then it still should
// never happen that another thread is waiting for this lock in order to
// access the QueuedRequest: a few lines lower we delete it.
// In other words, if directly after releasing this lock another thread
// would access req, then that can only happen by finding it again in
// either mRequestQueue or mRequestHash. We already deleted it from the
// first, so this access would have to happen by finding it in mRequestHash.
// Such access happens in the following functions:
// 1) LLQueuedThread::shutdown -- but it does that anyway, as it doesn't use any locking.
// 2) LLQueuedThread::generateHandle -- might skip our handle while before it would block until we deleted it. Skipping it is actually better.
// 3) LLQueuedThread::waitForResult -- this now doesn't touch the req when it has the flag FLAG_LOCKED set.
// 4) LLQueuedThread::getRequest -- whereever this is used, FLAG_LOCKED is tested before using the req.
// 5) LLQueuedThread::getRequestStatus -- this is a read-only operation on the status, which should never be changed from finishRequest().
// 6) LLQueuedThread::abortRequest -- it doesn't seem to hurt to add flags (if this happens at all), while calling finishRequest().
// 7) LLQueuedThread::setFlags -- same.
// 8) LLQueuedThread::setPriority -- doesn't access req with status STATUS_ABORTED, STATUS_COMPLETE or STATUS_INPROGRESS.
// 9) LLQueuedThread::completeRequest -- now sets FLAG_AUTO_COMPLETE instead of deleting the req, if FLAG_LOCKED is set, so that deletion happens here when finishRequest returns.
req->setFlags(FLAG_LOCKED);
unlockData();
req->finishRequest(false);
lockData();
req->resetFlags(FLAG_LOCKED);
if ((req->getFlags() & FLAG_AUTO_COMPLETE))
{
req->resetFlags(FLAG_AUTO_COMPLETE);
mRequestHash.erase(req);
// check();
unlockData();
req->deleteRequest();
lockData();
}
continue;
}
llassert_always(req->getStatus() == STATUS_QUEUED);
break;
}
U32 start_priority = 0 ;
if (req)
{
req->setStatus(STATUS_INPROGRESS);
start_priority = req->getPriority();
}
unlockData();
// This is the only place we will call req->setStatus() after
// it has initially been seet to STATUS_QUEUED, so it is
// safe to access req.
if (req)
{
// process request
bool complete = req->processRequest();
if (complete)
{
lockData();
req->setStatus(STATUS_COMPLETE);
req->setFlags(FLAG_LOCKED);
unlockData();
req->finishRequest(true);
if ((req->getFlags() & FLAG_AUTO_COMPLETE))
{
lockData();
req->resetFlags(FLAG_AUTO_COMPLETE);
mRequestHash.erase(req);
// check();
req->resetFlags(FLAG_LOCKED);
unlockData();
req->deleteRequest();
}
else
{
req->resetFlags(FLAG_LOCKED);
}
}
else
{
lockData();
req->setStatus(STATUS_QUEUED);
mRequestQueue.insert(req);
unlockData();
if (mThreaded && start_priority < PRIORITY_NORMAL)
{
ms_sleep(1); // sleep the thread a little
}
}
}
S32 pending = getPending();
return pending;
}
// virtual
bool LLQueuedThread::runCondition()
{
// mRunCondition must be locked here
if (mRequestQueue.empty() && mIdleThread)
return false;
else
return true;
}
// virtual
void LLQueuedThread::run()
{
// call checPause() immediately so we don't try to do anything before the class is fully constructed
checkPause();
startThread();
mStarted = TRUE;
while (1)
{
// this will block on the condition until runCondition() returns true, the thread is unpaused, or the thread leaves the RUNNING state.
checkPause();
if (isQuitting())
{
endThread();
break;
}
mIdleThread = FALSE;
threadedUpdate();
int res = processNextRequest();
if (res == 0)
{
mIdleThread = TRUE;
ms_sleep(1);
}
//LLThread::yield(); // thread should yield after each request
}
llinfos << "LLQueuedThread " << mName << " EXITING." << llendl;
}
// virtual
void LLQueuedThread::startThread()
{
}
// virtual
void LLQueuedThread::endThread()
{
}
// virtual
void LLQueuedThread::threadedUpdate()
{
}
//============================================================================
LLQueuedThread::QueuedRequest::QueuedRequest(LLQueuedThread::handle_t handle, U32 priority, U32 flags) :
LLSimpleHashEntry<LLQueuedThread::handle_t>(handle),
mStatus(STATUS_UNKNOWN),
mPriority(priority),
mFlags(flags)
{
}
LLQueuedThread::QueuedRequest::~QueuedRequest()
{
llassert_always(mStatus == STATUS_DELETE);
}
//virtual
void LLQueuedThread::QueuedRequest::finishRequest(bool completed)
{
}
//virtual
void LLQueuedThread::QueuedRequest::deleteRequest()
{
llassert_always(mStatus != STATUS_INPROGRESS);
setStatus(STATUS_DELETE);
delete this;
}