CONTEXT is captured here:
|
ClrCaptureContext(&context); |
And then passed to RaiseFailFastException here:
|
pExceptionInfo == NULL ? NULL : pExceptionInfo->ContextRecord, |
Stacktrace:
KERNELBASE!RaiseFailFastException
coreclr!WatsonLastChance+0x1b3 [D:\a\_work\1\s\src\coreclr\vm\excep.cpp @ 3590]
coreclr!EEPolicy::LogFatalError+0x6b4 [D:\a\_work\1\s\src\coreclr\vm\eepolicy.cpp @ 589]
coreclr!EEPolicy::HandleFatalError+0x133 [D:\a\_work\1\s\src\coreclr\vm\eepolicy.cpp @ 828]
The problem is that RaiseFailFastException validates that the context matches a valid return address on the stack as part of CET. If the context is not valid, it rejects it, flags it as a potential attack and asserts in checked builds of Windows OS.
We should stop passing invalid CONTEXT to RaiseFailFastException and verify that the Watson bucketing of the crashes originating in EEPolicy::HandleFatalError works as expected.
CONTEXT is captured here:
runtime/src/coreclr/vm/eepolicy.cpp
Line 786 in 31b5ae2
And then passed to RaiseFailFastException here:
runtime/src/coreclr/vm/excep.cpp
Line 3574 in 31b5ae2
Stacktrace:
The problem is that RaiseFailFastException validates that the context matches a valid return address on the stack as part of CET. If the context is not valid, it rejects it, flags it as a potential attack and asserts in checked builds of Windows OS.
We should stop passing invalid CONTEXT to RaiseFailFastException and verify that the Watson bucketing of the crashes originating in EEPolicy::HandleFatalError works as expected.