SU+J deadlock

[Steve Kargl]

My system locked up without panicking.  Neither access from
the console nor via ssh from another terminal worked.  The
only recourse was a power cycle.

% uname -a
FreeBSD troutmask.apl.washington.edu 9.0-CURRENT
r211766M: Tue Aug 24 14:52:25 PDT 2010
kargl at troutmask.apl.washington.edu:/usr/obj/usr/src/sys/SPEW  amd64

Upon rebooting, I entered single user mode.  The hand
transcribed session is

# fsck -y

** SU+J Recovering /dev/ad6s1f
** Reading 33554422 byte journal from inode 4

RECOVER? yes

** Building recovery table
** Resolving unreferenced inode list
** Processing journal entries
Bad cg number 6296367
UNEXPECTED SU+J INCONSISTENCY

FALLBACK TO FULL FSCK? yes

** Skipping journal, falling through to full fsck

** Last Mounted on /usr
** Phase 1 - Check Block and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
UNREF FILE I=38437143 OWNER=sgk MODE=100600
SIZE=536576 MTIME=Sep 9 14:30 2010
RECONNECT? yes

No lost+found DIRECTORY
CREATE? yes

** Phase 5 - Check cyl groups
FREE BLK COUNT(S) WRONG IN SUPERBLK
SALVAGE? yes

SUMMARY INFORMATION BAD
SALVAGE? yes
1188338 files, 76416941 used, 150441439 free

At point, 'fsck -y' proceeded to clean up the other
filesystems and I rebooted.

Note, OWNER=sgk was hammering the filesystem by running
the GCC testsuite to test the recent libelf changes.
The only file moved to lost+found is

troutmask:root[204] cd lost+found/
troutmask:root[205] ls
#38437143
troutmask:root[206] file #38437143 
#38437143: ELF 64-bit LSB core file x86-64, version 1 (FreeBSD), FreeBSD-style, from '-1.exe'

which is from GCC testing.

This is the 2nd such lock up in the past 2 weeks.  If
there are any kernel options that will help aid in 
debugging this problem, I'll turn them on.  Just ask.

-- 
Steve