From owner-freebsd-current  Sun Jul 26 21:14:42 1998
Return-Path: <owner-freebsd-current@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id VAA04204
          for freebsd-current-outgoing; Sun, 26 Jul 1998 21:14:42 -0700 (PDT)
          (envelope-from owner-freebsd-current@FreeBSD.ORG)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.15.68.22])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id VAA04184
          for <current@FreeBSD.ORG>; Sun, 26 Jul 1998 21:14:35 -0700 (PDT)
          (envelope-from bde@godzilla.zeta.org.au)
Received: (from bde@localhost)
	by godzilla.zeta.org.au (8.8.7/8.8.7) id OAA08697;
	Mon, 27 Jul 1998 14:14:04 +1000
Date: Mon, 27 Jul 1998 14:14:04 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199807270414.OAA08697@godzilla.zeta.org.au>
To: current@FreeBSD.ORG, mitch@pa.dec.com
Subject: Re: Hard hangs of -current under heavy load - how to debug?
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

>I've been experiencing some random hangs on -current
>releases over the past few months (I'm currently at 
>3.0-19980723, but I've seen this since last December).
>The systems operate under heavy load for about 24 hours, 
>then one or two randomly hang.  The hangs are hard (no 
>console messages, no dumps/traps, can't escape to 
>the debugger).  It looks like interrupts are disabled.   

I just fixed a hang with most interrupts disabled in the swap pager.
An spl nesting botch caused the entire pageout daemon to run
at splvm() after the first pageout.  This bug dates from 23 Feb.
Now the hang occurs without interrupts disabled :-).  There is a
near deadlock if swap fills up during core dumping of a huge (mostly
swapped out) process (an 87MB core dump would have taken a day or
two to complete).

>Generally, how do you debug a hang like this?  Are there

I removed the tty mask bit from splhigh() and other places so that
the keyboard debugger hotkey worked.  The BREAK_TO_DEBUGGER option
should allow the serial console debugger entry (break) to work in
even more cases.

Bruce

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message