Skip to content

Conversation

@janvorli
Copy link
Member

@janvorli janvorli commented Dec 9, 2019

There is a race between DynamicMethodDesc::Destroy called from
the finalizer thread and the MethodDescs enumeration called from
ETW::MethodLog::SendEventsForJitMethods at process exit.
DynamicMethodDesc::Destroy cleanos up its members m_pSig and
m_pszMethodName and then it calls GetLCGMethodResolver()->Destroy();
That calls EEJitManager::FreeCodeMemory, which tries to take the
m_CodeHeapCritSec lock. But this lock is already held by
the ETW::MethodLog::SendEventsForJitMethods.
So the iterator can see half-destroyed DynamicMethodDesc and
a crash happens when trying to get the dynamic method name
from the m_pszMethodName for the ETW event purposes.

The fix is to call the GetLCGMethodResolver()->Destroy() before
destroying the m_pSig and m_pszMethodName.

There is a race between DynamicMethodDesc::Destroy called from
the finalizer thread and the MethodDescs enumeration called from
ETW::MethodLog::SendEventsForJitMethods at process exit.
DynamicMethodDesc::Destroy cleanos up its members m_pSig and
m_pszMethodName and then it calls GetLCGMethodResolver()->Destroy();
That calls EEJitManager::FreeCodeMemory, which tries to take the
m_CodeHeapCritSec lock. But this lock is already held by
the ETW::MethodLog::SendEventsForJitMethods.
So the iterator can see half-destroyed DynamicMethodDesc and
a crash happens when trying to get the dynamic method name
from the m_pszMethodName for the ETW event purposes.

The fix is to call the GetLCGMethodResolver()->Destroy() before
destroying the m_pSig and m_pszMethodName.
@janvorli janvorli added this to the 5.0 milestone Dec 9, 2019
@janvorli janvorli requested a review from jkotas December 9, 2019 22:53
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants