Created attachment 143848 [details] Valgrind 3.18.1 crash trachback in debug info on RHEL 6.7 SUMMARY *** NOTE: If you are reporting a crash, please try to attach a backtrace with debug symbols. See https://community.kde.org/Guidelines_and_HOWTOs/Debugging/How_to_create_useful_crash_reports *** STEPS TO REPRODUCE 1. login to RHEL 6.7 box. 2. export VALGRIND_OPTS=-v --trace-symtab=yes --trace-symtab-patt=*libmagic* 3. valgrind <binary> OBSERVED RESULT valgrind crashes with the following error: valgrind: ../../coregrind/m_debuginfo/readelf.c:717 (get_elf_symbol_info): Assertion 'in_rx' failed. EXPECTED RESULT valgrind does not crash SOFTWARE/OS VERSIONS RHEL 6.7 - running valgrind in an xterm window ADDITIONAL INFORMATION Valgrind 3.18.1 was compiled on RHEL 6.7. Valgrind 3.17.0 does NOT exhibit this problem, nor do earlier versions of valgrind. It seems specific to 3.18.1 on RHEL6. The same valgrind 3.18.1 binary and same profiled binary run on RHEL7 and RHEL8 does NOT crash. The problem only occurs when running on RHEL6 with no other changes.
Which compiler are you using?
(In reply to Paul Floyd from comment #1) > Which compiler are you using? (In reply to Paul Floyd from comment #1) > Which compiler are you using? I can confirm the same issue on RHEL6.8 (Santiago) with both code under test and valgrind compiled with GCC 11.1.0.
(In reply to Paul Floyd from comment #1) > Which compiler are you using? gcc 4.4.7; however, I also tried using gcc 4.8, 4.9 and 7.4 with the same results.
I'll see if I can access a RHEL 6.7 machine.
I tried this on a machine Red Hat Enterprise Linux Workstation release 6.7 (Santiago) $ gcc --version gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-16) I had difficulty building Valgrind due to as not understanding rdseed (https://bugs.kde.org/show_bug.cgi?id=440502). I made a few changes to hard code out rdseed. The result was no problem running valgrind.
(In reply to Paul Floyd from comment #5) > I tried this on a machine > Red Hat Enterprise Linux Workstation release 6.7 (Santiago) > > $ gcc --version > gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-16) > > I had difficulty building Valgrind due to as not understanding rdseed > (https://bugs.kde.org/show_bug.cgi?id=440502). I made a few changes to hard > code out rdseed. > > The result was no problem running valgrind. Agreed, I can run valgrind with some binaries just fine, but on a more complicated program that accesses libmagic, I get the failure. A simple test program using libmagic succeeds. (Like the "file" command). I'm not a liberty to provide you with the binary that produces the problem which is why I attached the debug log from valgrind running the problematic program.
Is there any further information that I can provide to help track down this issue? We are unable to use the latest version of valgrind due to this problem. Thank you.
Again I can't reproduce this with 'file'. git HEAD with this small change $ git diff diff --git a/VEX/priv/guest_amd64_helpers.c b/VEX/priv/guest_amd64_helpers.c index ba71c1b..5269d78 100644 --- a/VEX/priv/guest_amd64_helpers.c +++ b/VEX/priv/guest_amd64_helpers.c @@ -3951,7 +3951,7 @@ ULong amd64g_dirtyhelper_RDRAND ( void ) { } ULong amd64g_dirtyhelper_RDSEED ( void ) { -# if defined(__x86_64__) +# if defined(__x86_64__) && 0 ULong res = 0; ULong cflag = 0; __asm__ __volatile__( $ cat /etc/redhat-release Red Hat Enterprise Linux Workstation release 6.7 (Santiago) $ ./vg-in-place -q file ./build.ksh ./build.ksh: Korn shell script text executable $ ls -al /usr/lib64/libmagic.so.1.0.0 -rwxr-xr-x 1 root root 118112 Aug 4 2014 /usr/lib64/libmagic.so.1.0.0 $ rpm -q --whatprovides /usr/lib64/libmagic.so.1.0.0 file-libs-5.04-21.el6.x86_64