linux - 为 powerpc 运行 valgrind 时出错

Question

我已经成功地为 PPC85xx 架构交叉编译了 valgrid。我有以下文件

./coregrind/valgrind
./coregrind/vgpreload_core-ppc32-linux.so
./memcheck/memcheck-ppc32-linux
./memcheck/vgpreload_memcheck-ppc32-linux.so
./default.supp

我确实将这些文件放在目标文件夹的 /home/valg11 文件夹中，并将库导出为

export VALGRIND_LIB=/home/valg11

我正在运行 valgrind 并收到此错误。

[root@85xx valg11]# ./valgrind -v /home/val
==988== Memcheck, a memory error detector
==988== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==988== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright info
==988== Command: /home/val
==988== 
--988-- Valgrind options:
--988--    -v
--988-- Contents of /proc/version:
--988--   Linux version 2.6.32 (somya@somya-HP-Pro-3330-MT) (gcc version 4.2.2) #230 Mon Oct 12 14:02:51 IST 2015
--988-- 
--988-- Arch and hwcaps: PPC32, BigEndian, ppc32-int-flt-FX-GX
--988-- Page sizes: currently 4096, max supported 65536
--988-- Valgrind library directory: /home/valg11
--988-- Reading syms from /lib/ld-2.8.so
--988-- Reading syms from /home/val
--988-- Reading syms from /home/valg11/memcheck-ppc32-linux
--988--    object doesn't have a dynamic symbol table
--988-- Scheduler: using generic scheduler lock implementation.
--988-- Reading suppressions file: /home/valg11/default.supp
==988== embedded gdbserver: reading from /tmp/vgdb-pipe-from-vgdb-to-988-by-root-on-???
==988== embedded gdbserver: writing to   /tmp/vgdb-pipe-to-vgdb-from-988-by-root-on-???
==988== embedded gdbserver: shared mem   /tmp/vgdb-pipe-shared-mem-vgdb-988-by-root-on-???
==988== 
==988== TO CONTROL THIS PROCESS USING vgdb (which you probably
==988== don't want to do, unless you know exactly what you're doing,
==988== or are doing some strange experiment):
==988==   /home/valg11/../../bin/vgdb --pid=988 ...command...
==988== 
==988== TO DEBUG THIS PROCESS USING GDB: start GDB like this
==988==   /path/to/gdb /home/val
==988== and then give GDB the following command
==988==   target remote | /home/valg11/../../bin/vgdb --pid=988
==988== --pid is optional if only one valgrind process is running
==988== 
==988== error 22 Invalid argument
==988== error VG_(am_shared_mmap_file_float_valgrind) /tmp/vgdb-pipe-shared-mem-vgdb-988-by-root-on-???

正如某处所建议的那样，我确实在没有 vgdb 的情况下运行了 valgrind，但是 valgrind 没有检测到内存泄漏并给出了非法指令错误。

[root@85xx valg11]# ./valgrind -v --vgdb=no /home/val
==995== Memcheck, a memory error detector
==995== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==995== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright info
==995== Command: /home/val
==995== 
--995-- Valgrind options:
--995--    -v
--995--    --vgdb=no
--995-- Contents of /proc/version:
--995--   Linux version 2.6.32 (somya@somya-HP-Pro-3330-MT) (gcc version 4.2.2) #230 Mon Oct 12 14:02:51 IST 2015
--995-- 
--995-- Arch and hwcaps: PPC32, BigEndian, ppc32-int-flt-FX-GX
--995-- Page sizes: currently 4096, max supported 65536
--995-- Valgrind library directory: /home/valg11
--995-- Reading syms from /lib/ld-2.8.so
--995-- Reading syms from /home/val
--995-- Reading syms from /home/valg11/memcheck-ppc32-linux
--995--    object doesn't have a dynamic symbol table
--995-- Scheduler: using generic scheduler lock implementation.
--995-- Reading suppressions file: /home/valg11/default.supp
disInstr(ppc): unhandled instruction: 0x10E40301
                 primary 4(0x4), secondary 769(0x301)
==995== valgrind: Unrecognised instruction at address 0x4018ff0.
==995==    at 0x4018FF0: memcpy (in /lib/ld-2.8.so)
==995==    by 0x40021C3: _dl_start_final (in /lib/ld-2.8.so)
==995==    by 0x4015F8B: _start (in /lib/ld-2.8.so)
==995== Your program just tried to execute an instruction that Valgrind
==995== did not recognise.  There are two possible reasons for this.
==995== 1. Your program has a bug and erroneously jumped to a non-code
==995==    location.  If you are running Memcheck and you just saw a
==995==    warning about a bad jump, it's probably your program's fault.
==995== 2. The instruction is legitimate but Valgrind doesn't handle it,
==995==    i.e. it's Valgrind's fault.  If you think this is the case or
==995==    you are not sure, please let us know and we'll try to fix it.
==995== Either way, Valgrind will now raise a SIGILL signal which will
==995== probably kill your program.
==995== 
==995== Process terminating with default action of signal 4 (SIGILL)
==995==  Illegal opcode at address 0x4018FF0
==995==    at 0x4018FF0: memcpy (in /lib/ld-2.8.so)
==995==    by 0x40021C3: _dl_start_final (in /lib/ld-2.8.so)
==995==    by 0x4015F8B: _start (in /lib/ld-2.8.so)
==995== 
==995== HEAP SUMMARY:
==995==     in use at exit: 0 bytes in 0 blocks
==995==   total heap usage: 0 allocs, 0 frees, 0 bytes allocated
==995== 
==995== All heap blocks were freed -- no leaks are possible
==995== 
==995== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 4 from 2)
--995-- 
--995-- used_suppression:      4 dl-hack1 /home/valg11/default.supp:1205
==995== 
==995== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 4 from 2)
Illegal instruction

我不知道问题出在哪里，有人在 powerpc 架构上运行过 valgrind 吗？我应该遵循什么程序？

Answer 1

很可能是它记录的第二个原因:

The instruction is legitimate but Valgrind doesn't handle it, i.e. it's Valgrind's fault. If you think this is the case or you are not sure, please let us know and we'll try to fix it.

您的可执行文件似乎是使用您正在使用的 valgrind 版本不完全支持的目标指令集编译的。假设您从最新的来源构建了 valgrind，解决您的问题的最简单方法可能是使用目标 ISA 的更严格(通常是旧版本)版本重新编译您的可执行文件。如果您提供完整的编译器版本和命令行信息，我们也许能够帮助您弄清楚如何做到这一点。