[Profiler] Avoid Recursive ThreadStoreLock in Profiling Thread Enumerator #110548
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
mdh1418
changed the title
|
Do we need a test for this? |
VSadov
reviewed
Dec 9, 2024
VSadov
reviewed
Dec 9, 2024
This was referenced Dec 10, 2024
Open
Profiling Enumerators look to acquire the ThreadStoreLock. In release config, re-acquiring the ThreadStoreLock and releasing it in ProfilerThreadEnum::Init will cause problems if the callback invoking EnumThread has logic that depends on the ThreadStoreLock being held. To avoid recursively acquiring the ThreadStoreLock, expand the condition when the profiling thread enumerator shouldn't acquire the ThreadStoreLock.
There was a potential race condition when setting the flag before suspending and resetting the flag after restarting. For example, if the thread restarting runtime is preempted right after resuming runtime, the flag could remain unset by the time another thread looks to suspend runtime, which would see that the flag as set.
mdh1418
force-pushed
the
profiler_thread_enum_avoid_recursive_thread_store_lock
branch
from
cd6f6f5 to
d56d8bb
Compare
|
Given that this test should run quickly when it passes, and its expected failure case is when it hangs/deadlocks, is there a straightforward way to set a quick timeout for this test instead of hitting CI's timeouts? And is there a way to hit such a timeout in a local run via corerun? @davmason @jkotas |
|
The problem with timeouts is you can't check them in because we run a lot of tests in various modes (GCStress, JITStress, etc) that can make them take 10+ minutes. For this kind of thing what I have done in the past is make the timeout short locally and run it in a loop on your dev machine to build confidence, but leave the timeout unchanged in the tree. |
mdh1418
deleted the
profiler_thread_enum_avoid_recursive_thread_store_lock
branch
|
/backport to release/9.0-staging |
|
Started backporting to release/9.0-staging: https://github.com/dotnet/runtime/actions/runs/12303767691 |
4 tasks
hez2010
pushed a commit
to hez2010/runtime
that referenced
this pull request
Dec 14, 2024
…ator (dotnet#110548) * [Profiler] Avoid Recursive ThreadStoreLock Profiling Enumerators look to acquire the ThreadStoreLock. In release config, re-acquiring the ThreadStoreLock and releasing it in ProfilerThreadEnum::Init will cause problems if the callback invoking EnumThread has logic that depends on the ThreadStoreLock being held. To avoid recursively acquiring the ThreadStoreLock, expand the condition when the profiling thread enumerator shouldn't acquire the ThreadStoreLock. * [Profiler] Change order to set fProfilerRequestedRuntimeSuspend There was a potential race condition when setting the flag before suspending and resetting the flag after restarting. For example, if the thread restarting runtime is preempted right after resuming runtime, the flag could remain unset by the time another thread looks to suspend runtime, which would see that the flag as set. * [Profiler][Tests] Add unit test for EnumThreads during suspension * [Profiler][Tests] Fixup EnumThreads test
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #110062
Profiling Enumerators look to acquire the ThreadStoreLock.
In release config, re-acquiring the ThreadStoreLock and releasing it in ProfilerThreadEnum::Init will cause problems if the callback invoking EnumThread has logic that depends on the ThreadStoreLock being held.
For example, invoking
EnumThreadwithinRuntimeSuspendFinishedwill violate the expectation that the ThreadStoreLock is held until RestartEE is called, demonstrated in #110062This PR aims to avoid recursively acquiring the ThreadStoreLock by expanding the known scnearios where the profiling thread enumerator shouldn't acquire the ThreadStoreLock.