Bug#463808: daily O2 d-i image not booting
Giuseppe Sacco a écrit :
> Il giorno mar, 04/03/2008 alle 09.41 +0100, Aurelien Jarno ha scritto:
>> Giuseppe Sacco a écrit :
>>> Il giorno mar, 04/03/2008 alle 08.51 +0100, Martin Michlmayr ha scritto:
>>>> reassign 463808 libc6
>>>>
>>>> * Giuseppe Sacco <giuseppe@eppesuigoccas.homedns.org> [2008-03-04 02:31]:
>>>>> I think the problem is libc6_2.7-6, as explained in #463808, since it
>>>>> does not hangs always at the same point, but it always hangs when
>>>>> starting configuring packages (so, when running all sort of programs
>>>>> that use libc6.
>>>> Maybe this should be reassigned then.
>>> Right, I have planned to install etch on a different SGI O2 machine,
>>> then only update libc6 (and maybe locales), to check if this show the
>>> problem.
>> What are the symtoms exactly? Looking at the bug report I only see notes
>> about a kernel crash. I fail to see how is it related to libc6.
>
> You are right, but from the stack frame it seems a pointer was shifted
> up by 32 bits, and may be the problem.
Well there may be a bug in the glibc causing a problem with the stack
frame, but anyway the fact that you can crash a kernel from userland is
a bug in the kernel, and even a security bug as it is basically a local DoS.
> More information is available at
> http://www.linux-mips.org/archives/linux-mips/2008-02/msg00011.html
>From the kernel dump, it looks like the crash occurs in kernel mode (bit
KX of the CP0 Status enabled). A bad address from userland should kill
the process with a SIGBUS.
--
.''`. Aurelien Jarno | GPG: 1024D/F1BCDB73
: :' : Debian developer | Electrical Engineer
`. `' aurel32@debian.org | aurelien@aurel32.net
`- people.debian.org/~aurel32 | www.aurel32.net
Reply to: