Bug#626001: sigseg in memset (powerpc)

[Jonathan Nieder <jrnieder@gmail.com>]

retitle 626001 [powerpc] sigsegv in memset called by calloc
tags 626001 = moreinfo
quit

Hi,

Jörg Sommer wrote:

> last night, grep was killed by SIGSEG. It was run by a cron job and I
> can't reproduce the error. But I've a coredump.

What version of grep do you use?  I don't know of anything like
debug.debian.net for powerpc which would make it easy to get a full
backtrace, (though it should be possible to get with sufficient
interest and time) unfortunately.

Also, do you have the cron job and the /var/lib/rkhunter/db/i18n/en
file from the time of failure?  Can we take a look at them?

[...]
> Core was generated by `grep ^ROOTKIT_FILES_DIRS_FILE: /var/lib/rkhunter/db/i18n/en'.
[...]
> #0  memset () at ../sysdeps/powerpc/powerpc32/memset.S:100
> No locals.
> #1  0x0fed241c in __libc_calloc (n=<value optimized out>, elem_size=<value optimized out>) at malloc.c:4123
>         av = 0xffc9268
>         oldtop = 0x103fcd18
>         bytes = 4

Looks like calloc is being called for a 4-byte allocation.  Usually
for such small allocations calloc handles the clearing itself, so that
information might help to narrow things down.

Just to check: are there any MALLOC_* or LD_* environment variables
set in your cron jobs' environment?  (I'm asking probably stupid
questions just to get basic information to trace what happened,
nothing else.)

[...]
> #2  0x10018700 in ?? ()
> No symbol table info available.
> #3  0x1000f260 in ?? ()
> No symbol table info available.
[...]
> Dump of assembler code for function memset:

I'd also be interested in disassembly for frame 2 or 3, so we can get
some idea of what grep was doing.

Thanks,
Jonathan