From owner-freebsd-stable Mon Feb 5 5: 3:38 2001 Delivered-To: freebsd-stable@freebsd.org Received: from sunu422.rz.ruhr-uni-bochum.de (sunu422.rz.ruhr-uni-bochum.de [134.147.64.14]) by hub.freebsd.org (Postfix) with SMTP id 912F437B6A2 for ; Mon, 5 Feb 2001 05:03:17 -0800 (PST) Received: (qmail 842 invoked from network); 5 Feb 2001 13:03:15 -0000 Received: from unknown (HELO ruhr-uni-bochum.de) (134.147.159.4) by mailhost.rz.ruhr-uni-bochum.de with SMTP; 5 Feb 2001 13:03:15 -0000 Message-ID: <3A7EA2FA.E08B958D@ruhr-uni-bochum.de> Date: Mon, 05 Feb 2001 13:56:27 +0100 From: Thomas Stratmann X-Mailer: Mozilla 4.7 [de] (Win95; I) X-Accept-Language: de MIME-Version: 1.0 To: freebsd-stable@freebsd.org Subject: ktrace sync'ing before panic Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Hi everyone, I had a problem once on a 4.1-stable which does not seem to be fixed by now. I had mount cause panic under certain circumstances and went to find the error, first doing ktrace on the call. The result was an empty ktrace.out as the output was not synchronized/written out do disk before the actual kernel panic. I went into kern_ktrace.c and inserted a VOP_FSync call inside ktrwrite and it worked: after the panic, one could see the last kernel action was a call to mount (into the kernel) - the first stage of success in debugging. I believe the issue of syncing is a tradeoff between speed and reliability in the (rare) case of a panic. One should consider introducing an option to the ktrace call for this - if the manpage will be elaborate enough about the issue, even non-hackers will be able to give more precise error mailings. Thanks Thomas Stratmann To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message