Menu
▾
▴
Re: [Valgrind-users] Valgrind throws assertion on powerpc while trying to run 'ls'
|
From: Vishal <vis...@gm...> - 2013-07-31 05:02:29
|
> > Hi Vishal, > > I do not have a PPC476 FP system, so I cannot attempt to reproduce the issue. > But, we need to understand what is going on. I tried this on the IBM POWER 7 > system I have here at work (see log below). Please could you try the corresponding > debug session on your PPC476 FP system? > > Note the comment in the code: "Either the value must not have been set yet (zero) or we can > tolerate it being set to the same value multiple times, ..." In the log below, > we see that vgPlain_machine_ppc64_set_clszB() is invoked twice before > Valgrind/memcheck exits, each time with szB=128. What happens in your case? > > When vgPlain_machine_ppc32_set_clszB() is invoked for the first time, at entry time: > Is vai.ppc_cache_line_szB == 0? What is the value of szB? > > Is vgPlain_machine_ppc32_set_clszB() invoked more than once? What value does > vai.ppc_cache_line_szB hold each time? What value does szB hold each time? > > Essentially, why does: vai.ppc_cache_line_szB == 0 || vai.ppc_cache_line_szB == szB fail? > Either at the first invocation vai.ppc_cache_line_szB != 0 or at a subsequent invocation, > szB changes what already exists in vai.ppc_cache_line_szB - why? > > One other try: Likely, you are compiling Valgrind optimized by default. What if you compile > it with optimization turned off: make CFLAGS="-O0 -g"? Does the assert still happen? > > Regards, > Anmol. > > anmol:/proj/ppc/DT/labhome/anmol/valgrind-3.8.1> gdb bin/valgrind > GNU gdb (GDB) Fedora (7.3.50.20110722-9.fc16) > Copyright (C) 2011 Free Software Foundation, Inc. > License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html> > This is free software: you are free to change and redistribute it. > There is NO WARRANTY, to the extent permitted by law. Type "show copying" > and "show warranty" for details. > This GDB was configured as "ppc64-redhat-linux-gnu". > For bug reporting instructions, please see: > <http://www.gnu.org/software/gdb/bugs/>... > Reading symbols from /proj/.ppc_DT_labhome/labhome/anmol/valgrind- 3.8.1/bin/valgrind...done. > (gdb) add-inferior -exec lib/valgrind/memcheck-ppc64-linux > Added inferior 2 > Reading symbols from /proj/.ppc_DT_labhome/labhome/anmol/valgrind- 3.8.1/lib/valgrind/memcheck-ppc64-linux...done. > (gdb) inferior 2 > [Switching to inferior 2 [process 0] (/proj/.ppc_DT_labhome/labhome/anmol/valgrind-3.8.1/lib/valgrind/memcheck- ppc64-linux)] > (gdb) break vgPlain_machine_ppc64_set_clszB > Breakpoint 1 at 0x38078354: file m_machine.c, line 1465. > (gdb) inferior 1 > [Switching to inferior 1 [process 0] (/proj/.ppc_DT_labhome/labhome/anmol/valgrind-3.8.1/bin/valgrind)] > (gdb) run ~/hello > Starting program: /proj/.ppc_DT_labhome/labhome/anmol/valgrind- 3.8.1/bin/valgrind ~/hello > process 7757 is executing new program: /proj/.ppc_DT_labhome/labhome/anmol/valgrind-3.8.1/lib/valgrind/memcheck- ppc64-linux > Missing separate debuginfos, use: debuginfo-install glibc-2.14.90- 24.fc16.7.ppc64 > > Breakpoint 1, vgPlain_machine_ppc64_set_clszB (szB=128) at m_machine.c:1465 > 1465 { > (gdb) bt > #0 vgPlain_machine_ppc64_set_clszB (szB=128) at m_machine.c:1465 > #1 0x00000000380c82c4 in setup_client_stack (clstack_max_size=<optimized out>, > clstack_end=34343026687, client_auxv=0x38f116e0, info=0x38f116e8, orig_envp=0x402010260, init_sp=0xfffffffeec0) > at m_initimg/initimg-linux.c:725 > #2 vgPlain_ii_create_image (iicii=...) at m_initimg/initimg-linux.c:930 > #3 0x000000003807b748 in valgrind_main (argc=<optimized out>, argv=0xfffffffeec8, > envp=0xfffffffeee0) at m_main.c:1852 > #4 0x000000003807fb14 in _start_in_C_linux (pArgc=0xfffffffeec0) at m_main.c:2994 > #5 0x0000000038078608 in ._start () > (gdb) c > Continuing. > > Breakpoint 1, vgPlain_machine_ppc64_set_clszB (szB=128) at m_machine.c:1465 > 1465 { > (gdb) bt > #0 vgPlain_machine_ppc64_set_clszB (szB=128) at m_machine.c:1465 > #1 0x00000000380c82c4 in setup_client_stack (clstack_max_size=<optimized out>, > clstack_end=34343026687, client_auxv=0x38f116e0, info=0x38f116e8, orig_envp=0x402010260, init_sp=0xfffffffeec0) > at m_initimg/initimg-linux.c:725 > #2 vgPlain_ii_create_image (iicii=...) at m_initimg/initimg-linux.c:930 > #3 0x000000003807b748 in valgrind_main (argc=<optimized out>, argv=0xfffffffeec8, > envp=0xfffffffeee0) at m_main.c:1852 > #4 0x000000003807fb14 in _start_in_C_linux (pArgc=0xfffffffeec0) at m_main.c:2994 > #5 0x0000000038078608 in ._start () > (gdb) c > Continuing. > ==7822== Memcheck, a memory error detector > ==7822== Copyright (C) 2002-2012, and GNU GPL'd, by Julian Seward et al. > ==7822== Using Valgrind-3.8.1-FSL-SDK-1.4-spe-Fri-May-24-080638-PDT-2013 and LibVEX; rerun with > -h for copyright info > ==7822== Command: /home/anmol/hello > ==7822== > ==7822== Invalid write of size 4 > ==7822== at 0x10000540: main (hello.c:8) > ==7822== Address 0x4040044 is 0 bytes after a block of size 4 alloc'd > ==7822== at 0x40281FC: malloc (vg_replace_malloc.c:270) > ==7822== by 0x10000523: main (hello.c:6) > ==7822== > ==7822== > ==7822== HEAP SUMMARY: > ==7822== in use at exit: 4 bytes in 1 blocks > ==7822== total heap usage: 1 allocs, 0 frees, 4 bytes allocated > ==7822== > ==7822== LEAK SUMMARY: > ==7822== definitely lost: 4 bytes in 1 blocks > ==7822== indirectly lost: 0 bytes in 0 blocks > ==7822== possibly lost: 0 bytes in 0 blocks > ==7822== still reachable: 0 bytes in 0 blocks > ==7822== suppressed: 0 bytes in 0 blocks > ==7822== Rerun with --leak-check=full to see details of leaked memory > ==7822== > ==7822== For counts of detected and suppressed errors, rerun with: -v > ==7822== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0) > [Inferior 1 (process 7822) exited normally] > (gdb) > Hi Anmol, I could not try GDB session as you suggested below. Unfortunately add- inferior command is not supported on our installation. I tried to put printf in the code but that gave vague errors. Next I tried to put some asserts at the same place like: 1. vg_assert(vai.ppc_cache_line_szB == 0) 2. vg_assert(vai.ppc_cache_line_szB == szB) 3. vg_assert(szB != 0) 4. vg_assert(szB != 32) 5. vg_assert(szB != 64) 6. vg_assert(szB != 128) The test failed again at line (2) above. Will this information help? Do you want me to try something else? Warm Regards, Vishal |
