From owner-freebsd-questions Thu Jun 17 13:14:14 1999 Delivered-To: freebsd-questions@freebsd.org Received: from veenet.value.net (value.net [209.182.128.4]) by hub.freebsd.org (Postfix) with ESMTP id C3499154E0 for ; Thu, 17 Jun 1999 13:14:06 -0700 (PDT) (envelope-from rfg@monkeys.com) Received: from monkeys.com (i180.value.net [206.14.136.180]) by veenet.value.net (8.8.7/8.7.4) with ESMTP id NAA16635 for ; Thu, 17 Jun 1999 13:17:41 -0700 (PDT) Received: from monkeys.com (LOCALHOST [127.0.0.1]) by monkeys.com (8.9.3/8.9.3) with ESMTP id NAA24431 for ; Thu, 17 Jun 1999 13:14:47 -0700 To: questions@freebsd.org Subject: A suggested kernel enhancement relating to kernel panics. Date: Thu, 17 Jun 1999 13:14:47 -0700 Message-ID: <24429.929650487@monkeys.com> From: "Ronald F. Guilmette" Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG I have already told my sad story about the kernel panics on my remote machine, so I will not belabor that anymore now. However I would like to just note that the *one* key piece of data that I cannot seem to lay my hands at the moment (and that would be most helpful to me in sorting this out) is the "panic string", i.e. the specific panic message that is being produced by the kernel when these panics occur. Just having thoe would probably help a lot. Right now, I don't even have the first clue as to what is causing the panics. These panic strings are presumably being written to the system console at panic time. That's swell, except for the fact that the machine in question normally has no monitor attached. :-( :-( And even if a monitor WAS attached, I believe that the system is set to perform an automatic reboot after a panic, so these message get cleared from the screen right after the panic/reboot anyway. :-( So this all prompts a suggestion to the developers... How about arranging for the panic string to be written *both* to the system console and also to the currently configured "dumpdev", if any? I have read about how a system panic dump gets preserved (by copying it into regular filesystem space) after a crash and reboot, and it seems to me that ideally, this same sort of preservation should be applied as well to any kernel panic message. After any kernel panic (and subsequent reboot), I should be able to go and look at a certain file and see the last kernel panic message. Is this too much to ask? Couldn't this be accomodated via a mechanism that borrows a lot (both code-wise and idea-wise) from the existing panic dump mechanism? Should we perhaps have a "panicdev" kernel configuration setting, just as we already have a kernel "dumpdev" configuration setting? The former could be used to specify where a copy of the last panic message should be written to. (That could perhaps even be a serial line or something.) To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message