On 12.03.2021 23:41, Nick Desaulniers wrote: > LLVM changed the expected function signatures for llvm_gcda_start_file() > and llvm_gcda_emit_function() in the clang-11 release. Users of clang-11 > or newer may have noticed their kernels failing to boot due to a panic > when enabling CONFIG_GCOV_KERNEL=y +CONFIG_GCOV_PROFILE_ALL=y. Fix up > the function signatures so calling these functions doesn't panic the > kernel. > > Link: https://reviews.llvm.org/rGcdd683b516d147925212724b09ec6fb792a40041 > Link: https://reviews.llvm.org/rG13a633b438b6500ecad9e4f936ebadf3411d0f44 > Cc: stable@xxxxxxxxxxxxxxx # 5.4 > Reported-by: Prasad Sodagudi <psodagud@xxxxxxxxxxx> > Suggested-by: Nathan Chancellor <nathan@xxxxxxxxxx> > Reviewed-by: Fangrui Song <maskray@xxxxxxxxxx> > Signed-off-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx> > Tested-by: Nathan Chancellor <nathan@xxxxxxxxxx> Looks good to me (minus the code duplication - but that's IMO acceptable since it's cleaned up again with patch 2). Acked-by: Peter Oberparleiter <oberpar@xxxxxxxxxxxxx> That said, I'm currently thinking of adding a compile time check that performs a dry-run gcov_info => gcda conversion in user space to detect these kind of issues before kernels fail unpredictably [1]. I'm confident that this could work for the GCC gcov kernel code, not sure about the Clang version though. But if it's possible I guess it would make sense to extend this to include the Clang code as well. Note that this check wouldn't work for cross-compiles since the build machine must be able to run code for the target machine. [1] https://lore.kernel.org/lkml/1c7a49e7-0e27-561b-a2f9-d42a83dc4c29@xxxxxxxxxxxxx/ Regards, Peter -- Peter Oberparleiter Linux on Z Development - IBM Germany
