Re: [Valgrind-users] lockup using valgrind on powerpc with eglibc-2.16

[Chris P. <jud...@gm...>]

On Thu, Sep 18, 2014 at 9:35 PM, Chris Packham <jud...@gm...> wrote:
> Hi,
>
> I'm trying to use valgrind on an embedded powerpc system. When I start
> a process under valgrind the it seems to lock up. I can connect to the
> process using gdb/vgdb and the the following backtrace
>
>   #0  0x0ffbaba8 in _vgr20190ZU_ldZdsoZd1_bcmp (s1V=0x401d000,
> s2V=0x401a2e7, n=11) at ../shared/vg_replace_strmem.c:974
>   #1  0x0ffbabac in _vgr20190ZU_ldZdsoZd1_bcmp (s1V=<optimized out>,
> s2V=<optimized out>, n=<error reading variable: value has been
> optimized out>) at ../shared/vg_replace_strmem.c:974
>
> Originally I thought it was a problem with 3.10.0 but I've tried a few
> versions and they all seem to have the same type of issue so I think
> it's something about our system causing the problem.
>
> We have successfully run valgrind 3.7.0 on an earlier version of our
> system using uclibc (after a few patches). I did a quick test with
> 3.7.0 after convincing configure to allow eglibc-2.16 and got the same
> behaviour but given the fact that 3.7.0 doesn't claim to support
> glib-2.16 I'm not sure how valid that result is.
>
> Has anyone got any clues as to what I need to do to get valgrind
> running on this system?
>
> Thanks,
> Chris

Update. I was able to try a mips based target and valgrind works fine.
So this appears to be specific to powerpc. If it helps here is some
output from a powerpc target vs a mips target

POWERPC:

$ valgrind --verbose more /proc/self/cmdline
==6250== Memcheck, a memory error detector
==6250== Copyright (C) 2002-2013, and GNU GPL'd, by Julian Seward et al.
==6250== Using Valgrind-3.10.0 and LibVEX; rerun with -h for copyright info
==6250== Command: more /proc/self/cmdline
==6250==
--6250-- Valgrind options:
--6250--    --verbose
--6250-- Contents of /proc/version:
--6250--   Linux version 3.6.11 (@chrisp-dl) (gcc version 4.6.3
(Gentoo 4.6.3-r1 p1.9, pie-0.5.2) ) #1 SMP Thu Sep 18 14:22:06 NZST 2
014
--6250-- Arch and hwcaps: PPC32, BigEndian, ppc32-int-flt-GX
--6250-- Page sizes: currently 4096, max supported 65536
--6250-- Valgrind library directory: /usr/lib/valgrind
--6250-- Reading syms from /lib/ld-2.16.so
--6250-- Reading syms from /bin/more
--6250--    object doesn't have a symbol table
--6250-- Reading syms from /usr/lib/valgrind/memcheck-ppc32-linux
--6250--    object doesn't have a symbol table
--6250--    object doesn't have a dynamic symbol table
--6250-- Scheduler: using generic scheduler lock implementation.
--6250-- Reading suppressions file: /usr/lib/valgrind/default.supp
==6250== embedded gdbserver: reading from
/tmp/vgdb-pipe-from-vgdb-to-6250-by-root-on-AT_SBx81CFC960
==6250== embedded gdbserver: writing to
/tmp/vgdb-pipe-to-vgdb-from-6250-by-root-on-AT_SBx81CFC960
==6250== embedded gdbserver: shared mem
/tmp/vgdb-pipe-shared-mem-vgdb-6250-by-root-on-AT_SBx81CFC960
==6250==
==6250== TO CONTROL THIS PROCESS USING vgdb (which you probably
==6250== don't want to do, unless you know exactly what you're doing,
==6250== or are doing some strange experiment):
==6250==   /usr/lib/valgrind/../../bin/vgdb --pid=6250 ...command...
==6250==
==6250== TO DEBUG THIS PROCESS USING GDB: start GDB like this
==6250==   /path/to/gdb more
==6250== and then give GDB the following command
==6250==   target remote | /usr/lib/valgrind/../../bin/vgdb --pid=6250
==6250== --pid is optional if only one valgrind process is running
==6250==
--6250-- REDIR: 0x4016c58 (ld.so.1:strlen) redirected to 0x3806c5a0 (???)
--6250-- REDIR: 0x4016a90 (ld.so.1:strcmp) redirected to 0x3806c5c8 (???)
--6250-- REDIR: 0x40169b4 (ld.so.1:index) redirected to 0x3806c63c (???)
--6250-- Reading syms from /usr/lib/valgrind/vgpreload_core-ppc32-linux.so
--6250--    object doesn't have a symbol table
--6250-- Reading syms from /usr/lib/valgrind/vgpreload_memcheck-ppc32-linux.so
--6250--    object doesn't have a symbol table
--6250-- REDIR: 0x40176f8 (ld.so.1:memcpy) redirected to 0xffb9fa4 (memcpy)
<hang>

MIPS:
$  valgrind --verbose more /proc/self/cmdline
==13966== Memcheck, a memory error detector
==13966== Copyright (C) 2002-2013, and GNU GPL'd, by Julian Seward et al.
==13966== Using Valgrind-3.10.0 and LibVEX; rerun with -h for copyright info
==13966== Command: more /proc/self/cmdline
==13966==
--13966-- Valgrind options:
--13966--    --verbose
--13966-- Contents of /proc/version:
--13966--   Linux version 3.6.11-at1 (@chrisp-dl) (gcc version 4.6.3
(Gentoo 4.6.3-r1 p1.9, pie-0.5.2) ) #1 Thu Sep 18 17:23:47 NZST
2014
--13966-- Arch and hwcaps: MIPS32, BigEndian, Broadcom-baseline
--13966-- Page sizes: currently 4096, max supported 4096
--13966-- Valgrind library directory: /usr/lib/valgrind
--13966-- Reading syms from /bin/more
--13966--    object doesn't have a symbol table
--13966-- Reading syms from /lib/ld-2.16.so
--13966-- Reading syms from /usr/lib/valgrind/memcheck-mips32-linux
--13966--    object doesn't have a symbol table
--13966--    object doesn't have a dynamic symbol table
--13966-- Scheduler: using generic scheduler lock implementation.
--13966-- Reading suppressions file: /usr/lib/valgrind/default.supp
==13966== embedded gdbserver: reading from
/tmp/vgdb-pipe-from-vgdb-to-13966-by-root-on-x510_52GTX
==13966== embedded gdbserver: writing to
/tmp/vgdb-pipe-to-vgdb-from-13966-by-root-on-x510_52GTX
==13966== embedded gdbserver: shared mem
/tmp/vgdb-pipe-shared-mem-vgdb-13966-by-root-on-x510_52GTX
==13966==
==13966== TO CONTROL THIS PROCESS USING vgdb (which you probably
==13966== don't want to do, unless you know exactly what you're doing,
==13966== or are doing some strange experiment):
==13966==   /usr/lib/valgrind/../../bin/vgdb --pid=13966 ...command...
==13966==
==13966== TO DEBUG THIS PROCESS USING GDB: start GDB like this
==13966==   /path/to/gdb more
==13966== and then give GDB the following command
==13966==   target remote | /usr/lib/valgrind/../../bin/vgdb --pid=13966
==13966== --pid is optional if only one valgrind process is running
==13966==
--13966-- Reading syms from /usr/lib/valgrind/vgpreload_core-mips32-linux.so
--13966--    object doesn't have a symbol table
--13966-- Reading syms from /usr/lib/valgrind/vgpreload_memcheck-mips32-linux.so
--13966--    object doesn't have a symbol table
--13966-- REDIR: 0x4019980 (ld.so.1:bcmp) redirected to 0x484c8f8 (bcmp)
--13966-- REDIR: 0x4019fc0 (ld.so.1:memcpy) redirected to 0x484bef0 (memcpy)
--13966-- REDIR: 0x4019ea0 (ld.so.1:mempcpy) redirected to 0x484da94 (mempcpy)
--13966-- Reading syms from /usr/lib/libncurses.so.5.9
--13966--    object doesn't have a symbol table
--13966-- Reading syms from /lib/libpthread-2.16.so
--13966-- Reading syms from /lib/libc-2.16.so
==13966== Invalid write of size 4
==13966==    at 0x4001348: _dl_start_user (in /lib/ld-2.16.so)
==13966==    by 0x40012DC: __start (in /lib/ld-2.16.so)
==13966==  Address 0x7eab38cc is on thread 1's stack
==13966==  4 bytes below stack pointer
==13966==
--13966-- REDIR: 0x49483d0 (libc.so.6:memset) redirected to 0x484d0b8 (memset)
--13966-- REDIR: 0x4948b30 (libc.so.6:memcpy) redirected to 0x484b650 (memcpy)
--13966-- REDIR: 0x49470e0 (libc.so.6:rindex) redirected to 0x48493b0 (rindex)
--13966-- REDIR: 0x4946780 (libc.so.6:strcmp) redirected to 0x484ae1c (strcmp)
--13966-- REDIR: 0x4946d30 (libc.so.6:strlen) redirected to 0x4849ad8 (strlen)
--13966-- REDIR: 0x4946f60 (libc.so.6:strncmp) redirected to 0x484a328 (strncmp)
--13966-- REDIR: 0x4946690 (libc.so.6:index) redirected to 0x4849608 (index)
--13966-- REDIR: 0x4942134 (libc.so.6:free) redirected to 0x4847330 (free)
--13966-- REDIR: 0x4941a30 (libc.so.6:malloc) redirected to 0x4848d60 (malloc)
--13966-- REDIR: 0x4948460 (libc.so.6:mempcpy) redirected to 0x484d95c (mempcpy)
--13966-- REDIR: 0x494296c (libc.so.6:calloc) redirected to 0x4845fc0 (calloc)
--13966-- REDIR: 0x494a620 (libc.so.6:strchrnul) redirected to
0x484d7b0 (strchrnul)
--13966-- REDIR: 0x4946de0 (libc.so.6:strnlen) redirected to 0x4849a70 (strnlen)
--13966-- REDIR: 0x4947978 (libc.so.6:strstr) redirected to 0x484df98 (strstr)
--13966-- REDIR: 0x4947010 (libc.so.6:strncpy) redirected to 0x4849e28 (strncpy)
more/proc/self/cmdline
==13966==
==13966== HEAP SUMMARY:
==13966==     in use at exit: 8,514 bytes in 8 blocks
==13966==   total heap usage: 13 allocs, 5 frees, 9,288 bytes allocated
==13966==
==13966== Searching for pointers to 8 not-freed blocks
==13966== Checked 80,244 bytes
==13966==
==13966== LEAK SUMMARY:
==13966==    definitely lost: 0 bytes in 0 blocks
==13966==    indirectly lost: 0 bytes in 0 blocks
==13966==      possibly lost: 0 bytes in 0 blocks
==13966==    still reachable: 8,514 bytes in 8 blocks
==13966==         suppressed: 0 bytes in 0 blocks
==13966== Rerun with --leak-check=full to see details of leaked memory
==13966==
==13966== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
==13966==
==13966== 1 errors in context 1 of 1:
==13966== Invalid write of size 4
==13966==    at 0x4001348: _dl_start_user (in /lib/ld-2.16.so)
==13966==    by 0x40012DC: __start (in /lib/ld-2.16.so)
==13966==  Address 0x7eab38cc is on thread 1's stack
==13966==  4 bytes below stack pointer
==13966==
==13966== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)