From owner-freebsd-current Sat Oct 6 22:19:21 2001 Delivered-To: freebsd-current@freebsd.org Received: from mail-relay1.yahoo.com (mail-relay1.yahoo.com [216.145.48.34]) by hub.freebsd.org (Postfix) with ESMTP id 7463B37B407 for ; Sat, 6 Oct 2001 22:19:18 -0700 (PDT) Received: from DougBarton.net (db-cvad-2-tmp.yahoo.com [216.145.48.243]) by mail-relay1.yahoo.com (Postfix) with ESMTP id BA20C8B5BC; Sat, 6 Oct 2001 22:19:17 -0700 (PDT) Message-ID: <3BBFE5DB.2831C139@DougBarton.net> Date: Sat, 06 Oct 2001 22:19:23 -0700 From: Doug Barton Organization: Triborough Bridge & Tunnel Authority X-Mailer: Mozilla 4.78 [en] (X11; U; Linux 2.4.2 i386) X-Accept-Language: en MIME-Version: 1.0 To: Bruce Evans Cc: freebsd-current@FreeBSD.ORG Subject: Re: panic: blockable sleep lock (sx) allproc @ /usr/local/src/sys/kern/kern_proc.c:212 References: <20011007142123.A6012-100000@delplex.bde.org> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Bruce Evans wrote: > > On Sat, 6 Oct 2001, Doug Barton wrote: > > > Bruce Evans wrote: > > > > > This is a well-know bug in printf(9). The TIOCCONS ioctl always gave > > > non-deterministic crashes. Now it gives determinstic panics when > > > pintf() is called while sched_lock is held. Work-around: don't use > > > anything that uses TIOCCONS (xconsole?). > > > > Yes, I'm using xconsole, and I find it very useful. How hard would it be > > to actually fix this? I'm still crashing fairly often with this same exact > > problem. > > Thanks. No one actually acknowledged my diagnosis of this bug before (I > don't use xconsole and haven't got around to trying it). > > Is using xconsole significantly better than "tail -f /var/log/messages"? Yes... for a couple reasons. It's less processes, it has a clever little thing where it changes the icon title if something has written a message to it, I can do clever things like 'ln -s /dev/console $HOME/.xsession-errors', etc. Admittedly, there isn't anything here that I couldn't reproduce, or live without. But xconsole is enabled standard with xdm, so it's something that should work. > Both have the problem of not being able to show any output if userland > can't run, but I think this problem would be only slightly larger for > syslogd and tail combined than for xconsole. A fix would involve more > buffering in the kernel, or perhaps more elegantly, dropping TIOCCONS > and putting "tail -f"-like support in dmesg. Well, there is the console facility for syslog now, so some of this is done already. But as I said, in the end xconsole should work. BTW, I can confirm that without xconsole running, I haven't crashed yet today, that's about 12 hours worth... previously the longest I lasted was about 4 hours. Doug -- "We will not tire, we will not falter, and we will not fail." - George W. Bush, President of the United States September 20, 2001 Do YOU Yahoo!? To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message