Re: 2.6.25: random stalls on certain hardware - regression?

[Marcin Slusarz]

On Wed, Jul 09, 2008 at 01:56:12AM +0400, Michael Tokarev wrote:
> No hope to resolve this but still - maybe someone has an idea...
>
> Hardware is -
>  AMD Athlon x2-64 system
>  Asus M2N-SLI DELUXE motherboard (nvidia MCP55)
>  AMD BE-2400 CPU
>  2x 2Gb mem
>  Adaptec 3950B Ultra2 SCSI adapter (*)
>  4x 36Gb Seagate SCSI disks
>  2x nvidia GbE ethernet
>  rtl8139 nic as 3rd one
>
> When booting 2.6.25 - either 32 or 64 bits - the system freezes/hangs
> at some random point.  So far there was about 10 hangs, some after
> several minutes after boot, some are within hours.
>
> The "hang" is a complete system freeze - on the console there's
> still "... login:" prompt, but nothing works - keyboard is stuck
> (numlock doesn't work), and server is not responding over network.
> Sometimes ping to the server itself works, but definitely not routing.
> (it's a server so no fancy stuff is loaded - X isn't even installed).
>
> The kernel is vanilla 2.6.25 - tried several - .5, .8, .10 now -
> the effect is the same.
>
> 2.6.24 and before worked without any glitch (2.6.24 is currently
> running).  We tried different versions of BIOS (was 12something
> before, tried 1304 and - currently - 1405, since 1502 is still
> beta) - no difference at all.
>
> The problem is that it is a production machine, and quite some
> people depend on it (it's a remote office with only one server),
> so I've very limited ability to try something.  Unfortunately not
> git bisect, -- or at least I'm afraid to try it, both because of
> possibility to have many reboots AND new freezes, and because
> unstable kernel (2.6.25pre stuff) with possibility to break something.

Definitely try 2.6.26-rc9 (which will soon become 2.6.26, so people will
probably ask you to test patches on top of this kernel) and if it will still 
lock up, take a look at Documentation/nmi_watchdog.txt and maybe
Documentation/networking/netconsole.txt in kernel sources.

Marcin
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/