Bug 203588

Summary:	[ufs] fsck_ufs segfaults during journals after powerloss
Product:	Base System	Reporter:	Julien Cigar <julien>
Component:	kern	Assignee:	freebsd-fs (Nobody) <fs>
Status:	Closed Not Enough Information
Severity:	Affects Many People	CC:	amvandemore, mckusick
Priority:	---
Version:	10.2-RELEASE
Hardware:	amd64
OS:	Any

Description Julien Cigar 2015-10-06 13:32:55 UTC

Hello,

I just had a powerloss with y desktop machine (which is not on an UPS). The machine is running:

FreeBSD frodon 10.2-RELEASE-p2 FreeBSD 10.2-RELEASE-p2 #2 r287854: Thu Sep 17 11:37:46 CEST 2015     root@frodon:/usr/obj/usr/src/sys/CUSTOM  amd64

After reboot SU+J was unable to recover the journal (segfaulting). A full fsck hopefully worked ..

I took a couple of screenshots which are available from here: http://home.bebif.be/~jcigar/ufs/
I can provide more information if needed ..

Comment 1 amvandemore 2015-11-17 19:57:45 UTC

Yes this is a bug.  Unfortunately, this type of PR is almost useless since there is no actionable information included.  I don't know any devs who love wild goose chases.

If it happens again, better ways to handle this are:

1.  Snapshot the underlying block device so there is always access to the problematic state.
2.  Run "fsck_ffs -d" and "fsck_ffs -f -d" and ensure the output is saved eg script(1).

See bug #187221 for an example of an actionable PR.

Also what does this "A full fsck hopefully worked .." mean?  Did a full fsck work?  If yes then there is no need to mention it and if no it needs to be in the PR.

Comment 2 Kirk McKusick freebsd_committer

2016-08-24 20:43:17 UTC

As noted this report does not have enough information to reproduce the problem.

Several bug fixes have been made to fsck that may have been relevant to this report.