STABILITY [Fwd: cvs commit: src/sys/kern kern_switch.c]

[Kevin Oberman]

> Date: Thu, 22 Jul 2004 09:13:55 -0600
> From: Scott Long <scottl at freebsd.org>
> Sender: owner-freebsd-current at freebsd.org
> 
> All,
> 
> This commit is another hack to try to improve stability some
> more.  Please let me know if it helps or hurts.  If it helps
> then I think that we are getting closer to at least one of the
> real culprits.
> 
> Scott
> 
> -------- Original Message --------
> Subject: cvs commit: src/sys/kern kern_switch.c
> Date: Thu, 22 Jul 2004 14:32:48 +0000 (UTC)
> From: Scott Long <scottl at FreeBSD.org>
> To: src-committers at FreeBSD.org, cvs-src at FreeBSD.org, cvs-all at FreeBSD.org
> 
> scottl      2004-07-22 14:32:48 UTC
> 
>    FreeBSD src repository
> 
>    Modified files:
>      sys/kern             kern_switch.c
>    Log:
>    Disable the PREEMPTION-enabled code in critical_exit() that encourages
>    switching to a different thread.  This is just a hack to try to improve
>    stability some more, but likely points closer to the real culprit.
> 
>    Revision  Changes    Path
>    1.72      +2 -0      src/sys/kern/kern_switch.c

Scott,

Sorry to be so late with this report, but I got stuck in Columbus and
Denver for way over 24 hours and am trying to catch up on
CURRENT. (Unusually heavy traffic.)

In any case, I still am seeing the freezes with PREEMPTION defined, but
they are far less frequent and the symptoms are slightly different. In
the past, the system (IBM T30) simply froze until power was cycled. The
display froze 

Now, after the freeze, the display starts to fade. The contrast is
reduced and the characters get fuzzy. Sometime random horizontal lines
appear on the display. This behavior is present with either a vty or
graphic (X) display.

It is possible that this is a totally different problem an results from
some other patch, but it's pretty close to what I was seeing in the past
in that it happens only under heavy CPU load and the only way to recover
is a power cycle.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634