From a232fa4d9f4bf8f29d2783da8c2e01f7c9f80fd9 Mon Sep 17 00:00:00 2001
From: Monty Brandenberg <monty@lindenlab.com>
Date: Thu, 19 Sep 2013 19:20:31 -0400
Subject: [PATCH] SH-4516 Fix a subset of thread races in llmeshrepository
Much earlier identified some thread races including two on mskininfoq and
mdecompositionq. I actually hit one in testing the other day so I'm fixing
them now. Put them under the mMutex lock and use the mutex in such a way
that main thread stalls are not added.
---
indra/newview/llmeshrepository.cpp | 49 ++++++++++++++++++++++--------
1 file changed, 36 insertions(+), 13 deletions(-)
diff --git a/indra/newview/llmeshrepository.cpp b/indra/newview/llmeshrepository.cpp
index c3d149db9ab..5174a7af00d 100755
--- a/indra/newview/llmeshrepository.cpp
+++ b/indra/newview/llmeshrepository.cpp
@@ -246,10 +246,10 @@
// mMeshHeader mHeaderMutex rw.repo.mHeaderMutex, ro.main.mHeaderMutex, ro.main.none [0]
// mMeshHeaderSize mHeaderMutex rw.repo.mHeaderMutex
// mSkinRequests mMutex rw.repo.mMutex, ro.repo.none [5]
-// mSkinInfoQ none rw.repo.none, rw.main.mMutex [0]
+// mSkinInfoQ mMutex rw.repo.mMutex, rw.main.mMutex [5] (was: [0])
// mDecompositionRequests mMutex rw.repo.mMutex, ro.repo.none [5]
// mPhysicsShapeRequests mMutex rw.repo.mMutex, ro.repo.none [5]
-// mDecompositionQ none rw.repo.none, rw.main.mMutex [0]
+// mDecompositionQ mMutex rw.repo.mMutex, rw.main.mMutex [5] (was: [0])
// mHeaderReqQ mMutex ro.repo.none [5], rw.repo.mMutex, rw.any.mMutex
// mLODReqQ mMutex ro.repo.none [5], rw.repo.mMutex, rw.any.mMutex
// mUnavailableQ mMutex rw.repo.none [0], ro.main.none [5], rw.main.mMutex
@@ -1713,7 +1713,10 @@ bool LLMeshRepoThread::skinInfoReceived(const LLUUID& mesh_id, U8* data, S32 dat
info.mMeshID = mesh_id;
// LL_DEBUGS(LOG_MESH) << "info pelvis offset" << info.mPelvisOffset << LL_ENDL;
- mSkinInfoQ.push(info);
+ {
+ LLMutexLock lock(mMutex);
+ mSkinInfoQ.push(info);
+ }
}
return true;
@@ -1740,7 +1743,10 @@ bool LLMeshRepoThread::decompositionReceived(const LLUUID& mesh_id, U8* data, S3
{
LLModel::Decomposition* d = new LLModel::Decomposition(decomp);
d->mMeshID = mesh_id;
- mDecompositionQ.push(d);
+ {
+ LLMutexLock lock(mMutex);
+ mDecompositionQ.push(d);
+ }
}
return true;
@@ -1799,7 +1805,10 @@ bool LLMeshRepoThread::physicsShapeReceived(const LLUUID& mesh_id, U8* data, S32
}
}
- mDecompositionQ.push(d);
+ {
+ LLMutexLock lock(mMutex);
+ mDecompositionQ.push(d);
+ }
return true;
}
@@ -2460,16 +2469,30 @@ void LLMeshRepoThread::notifyLoadedMeshes()
gMeshRepo.notifyMeshUnavailable(req.mMeshParams, req.mLOD);
}
- while (!mSkinInfoQ.empty())
+ if (! mSkinInfoQ.empty() || ! mDecompositionQ.empty())
{
- gMeshRepo.notifySkinInfoReceived(mSkinInfoQ.front());
- mSkinInfoQ.pop();
- }
+ std::queue<LLMeshSkinInfo> skin_info_q;
+ std::queue<LLModel::Decomposition*> decomp_q;
- while (!mDecompositionQ.empty())
- {
- gMeshRepo.notifyDecompositionReceived(mDecompositionQ.front());
- mDecompositionQ.pop();
+ if (mMutex->trylock())
+ {
+ // Make thread-shared data private with swap under lock.
+ skin_info_q.swap(mSkinInfoQ);
+ decomp_q.swap(mDecompositionQ);
+ mMutex->unlock();
+
+ while (! skin_info_q.empty())
+ {
+ gMeshRepo.notifySkinInfoReceived(skin_info_q.front());
+ skin_info_q.pop();
+ }
+
+ while (! decomp_q.empty())
+ {
+ gMeshRepo.notifyDecompositionReceived(decomp_q.front());
+ decomp_q.pop();
+ }
+ }
}
if (update_metrics)
--
GitLab