Re: Crashes in x86_64 math functions

[Allan McRae <allan at archlinux dot org>]

On 12/01/12 07:07, PaweÅ Sikora wrote:
> On Tuesday 10 of January 2012 17:38:05 Allan McRae wrote:
> 
>> I have not been able to debug this further yet due to a combined lack of
>> time and ability...
> 
> Hi,
> 
> attached testcase (compiled with: gcc -Wall t.c -g2 -o t -fno-builtin -lm)
> runs on glibc-2.14.1/valgrind without errors but on glibc-2.15 the valgrind reports:
> 
> $ valgrind ./t                                  
> ==4233== Memcheck, a memory error detector
> ==4233== Copyright (C) 2002-2011, and GNU GPL'd, by Julian Seward et al.
> ==4233== Using Valgrind-3.7.0 and LibVEX; rerun with -h for copyright info
> ==4233== Command: ./t
> ==4233==
> ==4233== Conditional jump or move depends on uninitialised value(s)
> ==4233==    at 0x40174B6: index (strchr.S:56)
> ==4233==    by 0x4007862: expand_dynamic_string_token (dl-load.c:431)
> ==4233==    by 0x400812F: _dl_map_object (dl-load.c:2281)
> ==4233==    by 0x400185D: map_doit (rtld.c:632)
> ==4233==    by 0x400E2F5: _dl_catch_error (dl-error.c:178)
> ==4233==    by 0x4000F0A: do_preload (rtld.c:821)
> ==4233==    by 0x400487E: dl_main (rtld.c:1714)
> ==4233==    by 0x4014576: _dl_sysdep_start (dl-sysdep.c:244)
> ==4233==    by 0x4004EED: _dl_start (rtld.c:337)
> ==4233==    by 0x4001687: ??? (in /lib64/ld-2.15.so)
> ==4233==
> ==4233== Conditional jump or move depends on uninitialised value(s)
> ==4233==    at 0x40174BB: index (strchr.S:59)
> ==4233==    by 0x4007862: expand_dynamic_string_token (dl-load.c:431)
> ==4233==    by 0x400812F: _dl_map_object (dl-load.c:2281)
> ==4233==    by 0x400185D: map_doit (rtld.c:632)
> ==4233==    by 0x400E2F5: _dl_catch_error (dl-error.c:178)
> ==4233==    by 0x4000F0A: do_preload (rtld.c:821)
> ==4233==    by 0x400487E: dl_main (rtld.c:1714)
> ==4233==    by 0x4014576: _dl_sysdep_start (dl-sysdep.c:244)
> ==4233==    by 0x4004EED: _dl_start (rtld.c:337)
> ==4233==    by 0x4001687: ??? (in /lib64/ld-2.15.so)
> 
> 
> i have an intel-i7-2630QM mobile cpu (family: 0x06, model: 0x2a) with avx.
> 


That appears to be a different issue.  These errors start further back
in the glibc git history than the commits that cause the crashes.

FYI, there is a bug report open about the crashes which details what the
actual issue is:
http://sourceware.org/bugzilla/show_bug.cgi?id=13580

Allan

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.