Subject: Re: multiprocessor i386 1.6ZH system crash
To: None <shnek@tiscali.cz>
From: Daniel Carosone <dan@geek.com.au>
List: current-users
Date: 02/11/2004 10:04:07
--bjuZg6miEcdLYP6q
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Tue, Feb 10, 2004 at 11:55:29PM +0100, shnek@tiscali.cz wrote:

> unfortunately, i don't have any serial console.=20

You just need another machine with a serial port running
cu/tip/kermit/whatever, and a "cross over" cable.

> another problem is that
> the freeze appears quite randomly, so it isn't easy to catch.

Which is why leaving the console attached waiting for messages is good :)

> crash completely locks the computer, so it is not possible for example
> to switch to "physical" console from X... and i expect debug info to be
> print to this console... or am i wrong? is it print to some file, too?

You are correct. The serial console can be printed to "real time" by
the kernel, and the messages remembered by the machine at the other
end of the cable.  Writing files to local disk rather requires that
the kernel (filesystem, buffer cache, disk drivers, interrupts, etc)
stays working long enough to record the message, which obviously it
won't in this case.

Now, as to the actual problem - I have seen it too. Please try a
truly-current kernel, there was at least one locking fix made by yamt
recently, though I haven't yet had time to try it myself and learn if
its relevant to this lockup.

--
Dan.
--bjuZg6miEcdLYP6q
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (NetBSD)

iD8DBQFAKWNnEAVxvV4N66cRAgstAKCbQsa/fpJvCYVnX05sD8ZBJU5mCgCg7+St
jB+JXnKCEJzfZTyrmP/Rb94=
=46iZ
-----END PGP SIGNATURE-----

--bjuZg6miEcdLYP6q--