Re: [PATCH] Mips: Enable asynchronous unwind tables with both ASAN and TSAN

[Fangrui Song <maskray@google.com>]

On Mon, Jul 4, 2022 at 6:54 PM Xi Ruoyao via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> On Mon, 2022-07-04 at 14:28 +0000, Dimitrije Milosevic wrote:
> > On Saturday, June 11, 2022 2:03 PM, Xi wrote:
> > > Just tried TSAN_SUPPORTED=yes with asynchronous unwind tables
> > > enabled,
> > > but I got some strange test failures for tls_race.c:
> > >
> > > FAIL: c-c++-common/tsan/tls_race.c   -O0  output pattern test
> > > Output was:
> > > ThreadSanitizer: CHECK failed: tsan_platform_linux.cpp:452
> > > "((thr_end)) <= ((tls_addr + tls_size))" (0xffec35f8c0,
> > > 0xffec35f784) (tid=748216)
> > >     #0 __tsan::CheckUnwind()
> > > ../../../../gcc/libsanitizer/tsan/tsan_rtl.cpp:627
> > > (libtsan.so.2+0xa30ec)
> > >     #1 __sanitizer::CheckFailed(char const*, int, char const*,
> > > unsigned long long, unsigned long long)
> > > ../../../../gcc/libsanitizer/sanitizer_common/sanitizer_termination.
> > > cpp:86 (libtsan.so.2+0xeb8cc)
> > >     #2 __tsan::ImitateTlsWrite(__tsan::ThreadState*, unsigned long,
> > > unsigned long)
> > > ../../../../gcc/libsanitizer/tsan/tsan_platform_linux.cpp:452
> > > (libtsan.so.2+0xa0cac)
> > >     #3 __tsan::ThreadStart(__tsan::ThreadState*, unsigned int,
> > > unsigned long long, __sanitizer::ThreadType)
> > > ../../../../gcc/libsanitizer/tsan/tsan_rtl_thread.cpp:197
> > > (libtsan.so.2+0xc0e88)
> > >     #4 __tsan_thread_start_func
> > > ../../../../gcc/libsanitizer/tsan/tsan_interceptors_posix.cpp:1009
> > > (libtsan.so.2+0x3e5dc)
> > >     #5 start_thread /sources/glibc-2.35/nptl/pthread_create.c:442
> > > (libc.so.6+0xc75f4)
> > >
> > > I've tried to diagnose the root cause but failed.
> >
> > Hi Xi, thanks for looking into this. I've tried running the testsuite
> > on a cross-toolchain (as I do not currently have access to a physical
> > machine)
> > for a MIPS64R6 and the test passes successfully. Could you please
> > verify that the test fails solely based on this change?
>
> I guess you mean you are running MIPS64R6 target code with qemu?  I'm
> not 100% sure because maybe something is wrong in my system.  I'm now
> retrying on gcc230.fsffrance.org (an EdgeRouter 4 in Cfarm) but building
> GCC on it is really slow.
>
> The changes I've tested:
>
> diff --git a/gcc/config/mips/mips.cc b/gcc/config/mips/mips.cc
> index 5eb845960e1..a7f0580e9ba 100644
> --- a/gcc/config/mips/mips.cc
> +++ b/gcc/config/mips/mips.cc
> @@ -20115,10 +20115,11 @@ mips_option_override (void)
>         target_flags |= MASK_64BIT;
>      }
>
> -  /* -fsanitize=address needs to turn on -fasynchronous-unwind-tables in
> -     order for tracebacks to be complete but not if any
> -     -fasynchronous-unwind-table were already specified.  */
> -  if (flag_sanitize & SANITIZE_USER_ADDRESS
> +  /* For -fsanitize=address or -fsanitize=thread, it's needed to turn
> +     on -fasynchronous-unwind-tables in order for tracebacks
> +     to be complete but not if any -fasynchronous-unwind-table
> +     were already specified.  */
> +  if (flag_sanitize & (SANITIZE_USER_ADDRESS | SANITIZE_THREAD)
>        && !global_options_set.x_flag_asynchronous_unwind_tables)
>      flag_asynchronous_unwind_tables = 1;
>
> diff --git a/libsanitizer/configure.tgt b/libsanitizer/configure.tgt
> index fb89df4935c..52546bbe4e3 100644
> --- a/libsanitizer/configure.tgt
> +++ b/libsanitizer/configure.tgt
> @@ -55,6 +55,9 @@ case "${target}" in
>    arm*-*-linux*)
>         ;;
>    mips*-*-linux*)
> +       if test x$ac_cv_sizeof_void_p = x8; then
> +               TSAN_SUPPORTED=yes
> +       fi
>         ;;
>    aarch64*-*-linux*)
>         if test x$ac_cv_sizeof_void_p = x8; then
> --
> Xi Ruoyao <xry111@xry111.site>
> School of Aerospace Science and Technology, Xidian University

Drive-by comment.

Clang considers that asan/msan/tsan/dataflow/etc enables
-fasynchronous-unwind-tables by default so I assume this GCC change is
fine.

With https://reviews.llvm.org/D102046 ("[sanitizer] Fall back to fast
unwinder"), compiler-rt may fall back to the frame pointer based
unwinder. There is not strong need to have the default
-fasynchronous-unwind-tables or -funwind-tables behavior.
However, most targets still default to omit frame pointer, so it's a
bit unfortunately that we need to enable unwind tables to get good
diagnostics.