[xfs-masters] [Bug 976] New: xfs_repair v3.1.11 -- segfault in Phase 6 Badness key lookup (length) : get_inode_offset at incore.h:319

[bugzilla-daemon at oss.sgi.com]

http://oss.sgi.com/bugzilla/show_bug.cgi?id=976

            Bug ID: 976
           Summary: xfs_repair v3.1.11 -- segfault in Phase 6 Badness key
                    lookup (length) : get_inode_offset at incore.h:319
           Product: XFS
           Version: Current
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: major
          Priority: P5
         Component: xfsprogs
          Assignee: xfs-masters at oss.sgi.com
          Reporter: miller at yoyo.ORG
                CC: gtr33m at gmail.com
    Classification: Unclassified
                CC: gtr33m at gmail.com

Created attachment 310
  --> http://oss.sgi.com/bugzilla/attachment.cgi?id=310&action=edit
gdb backtrace of segfault for Phase 6 Badness key lookup (length) :
get_inode_offset at incore.h:319

PRETTY_NAME="Ubuntu quantal (12.10)"
VERSION="12.10, Quantal Quetzal"

Bug is present in xfs_repair in Debian xfsprogs 3.1.7+b1 (armel), Ubuntu
xfsprogs  3.1.7build1 (i386), and latest source code 3.1.11 (compiled on i386).

A mounted xfs file system (on an external USB disk) created with mkfs.xfs
version 3.1.7 from Ubuntu 12.10 package was accidentally powered down whilst
data was being written on an arm5tel machine under Debian 7.1

The file system could be mounted again but obviously some data was corrupted.  

Trying xfs_repair on the arm5tel machine under Debian 7.1 resulted in
xfs_repair suffering from a segmentation fault.

The disk was then reconnected to the Ubuntu system and xfs_repair was tried
there with xfs_repair version 3.1.7 from Ubuntu 12.10 package, but again a
segmentation violation.  The binary did not have symbols so no meaningful trace
could be done.

The latest source code xfsprogs source code xfsprogs-3.1.11 was downloaded and
compiled and xfsrepair version 3.1.11 was run with gdb to get a backtrace which
is included as an attachment.

Trying the XFS FAQ suggested manual fix of

xfs_db -x -c 'inode 731' -c 'write core.nextents 0' -c 'write core.size 0'
/dev/sdd1
core.nextents = 0
core.size = 0

still results in

...
rebuilding directory inode 2193283622
        - traversal finished ...
        - moving disconnected inodes to lost+found ...
disconnected inode 731, b75ba700: Badness in key lookup (length)
bp=(bno 384, len 4096 bytes) key=(bno 384, len 8192 bytes)
b75ba700: Badness in key lookup (length)
bp=(bno 400, len 4096 bytes) key=(bno 400, len 8192 bytes)
Segmentation fault (core dumped)

--- Comment #1 from Mark Hawkins <gtr33m at gmail.com> ---
I can also confirm an almost identical error.

3.8.0-26-generic #38-Ubuntu SMP Mon Jun 17 21:43:33 UTC 2013 x86_64 x86_64
x86_64 GNU/Linux
Description:    Ubuntu 13.04
Release:    13.04
Codename:    raring

Compiled xfsprogs from git to 3.1.11

Same segmentation fault, tail of log:

rebuilding directory inode 4294965796
        - traversal finished ...
        - moving disconnected inodes to lost+found ...
disconnected dir inode 16777994, 7f8b872fd780: Badness in key lookup (length)
bp=(bno 0xa0, len 4096 bytes) key=(bno 0xa0, len 8192 bytes)
7f8b872fd780: Badness in key lookup (length)
bp=(bno 0xb0, len 4096 bytes) key=(bno 0xb0, len 8192 bytes)
Segmentation fault (core dumped)

I haven't tried the manual fix to inode 16777994 as I'm currently trying to
backup any retrievable files, however the similarity of the error suggests that
this will fail.

If anyone has a work around to this, I'd be grateful.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://oss.sgi.com/pipermail/xfs-masters/attachments/20130720/855a76bf/attachment.html>