Subject: Re: MOL crash, Pismo + YDL 2.0 [Re: ELF ?]
From: Samuel Rydh (samuel@ibrium.se)
Date: Wed Jun 20 2001 - 12:30:59 MDT
On Wed, Jun 20, 2001 at 11:23:05AM -0400, Stefan Jeglinski wrote:
> I'm not running Pismo or YDL, and I get virtually the same thing with 0.59:
>
> /usr/bin/startmol: line 329: 3555 Segmentation fault sleep 3
> ***** SIGNAL Illegal instruction in thread main-thread *****
> si_signo = 4, si_errno 0, si_code 00000080
> si_addr (nil). Last RVEC: 16411
> NIP 30025f6c
> cleaning up...
> Terminating threads...
> DONE
>
> This is on a kfukui 2.4.4-pre3 kernel on a G3-PowerTower.
This problem seems to be strongly related to the kernel version.
In particular, MOL does not seem to suffer from this problem
on the 2.4.2 kernel. I reverted to the 2.4.2 kernel some time
ago in order to test whether the occasional segmentation faults
would go away - and they did.
The (very similar) problem occuring on 2.2.18 kernels
may or may not be the same problem. I did isolate conditions
under which it occured - it is almost certain a signal related
kernel bug (I did write a small program which triggered
the bug).
YDL mol-0.9.59 and the rsync tree contains a workaround
for the 2.2.18 bug. However, I'd not be surprised if MOL
under 2.2.18 is still affected by the signaling problem
to some extent.
I don't know what cause the problem with the 2.4.4 kernels.
It might be a bug in MOL, but it could also be a kernel
issue. One of these days, I'll track down the kernel
version under which these segfaults started to occur.
I'd appreciate it if folks experiencing this bug on
2.4.4/mol-rsync could try older 2.4 versions and verify
that the segfault problem is not there.
Regards,
/Samuel
----------------------------------------------------------
E-mail <samuel@ibrium.se> WWW: <http://www.ibrium.se>
Phone/fax: (home) +46 8 4418431, (work) +46 8 7908470
----------------------------------------------------------
This archive was generated by hypermail 2a24 : Wed Jun 20 2001 - 11:37:02 MDT