xz(1) keeps SEGFAULT-ing
Mikhail T.
mi+thun at aldan.algebra.com
Sat Dec 22 15:07:55 UTC 2012
Hello!
I've set up several nightly backups all using the pipe-chain of dump |
xz -9 | ccrypt > /remote/backups/fs.xz.cpt
On one system these just work every night without a problem. On another
I see xz SEGFAULT-ing about 90% through almost every night for one of
the filesystems (the bigger of the two). This is, what cron emails me:
DUMP: WARNING: should use -L when dumping live read-write filesystems!
DUMP: Date of this level 1 dump: Sat Dec 22 03:23:00 2012
DUMP: Date of last level 0 dump: Thu Dec 6 01:23:02 2012
DUMP: Dumping /dev/ad4s1g (/home) to standard output
DUMP: mapping (Pass I) [regular files]
DUMP: Cache 16 MB, blocksize = 65536
DUMP: mapping (Pass II) [directories]
DUMP: estimated 2157823 tape blocks.
DUMP: dumping (Pass III) [directories]
DUMP: dumping (Pass IV) [regular files]
DUMP: 11.50% done, finished in 0:38 at Sat Dec 22 04:07:55 2012
DUMP: 17.80% done, finished in 0:46 at Sat Dec 22 04:20:37 2012
DUMP: 24.32% done, finished in 0:46 at Sat Dec 22 04:26:05 2012
DUMP: 31.23% done, finished in 0:44 at Sat Dec 22 04:28:27 2012
DUMP: 36.16% done, finished in 0:44 at Sat Dec 22 04:33:34 2012
DUMP: 43.21% done, finished in 0:39 at Sat Dec 22 04:33:51 2012
DUMP: 54.86% done, finished in 0:28 at Sat Dec 22 04:28:14 2012
DUMP: 60.63% done, finished in 0:25 at Sat Dec 22 04:30:24 2012
DUMP: 65.83% done, finished in 0:23 at Sat Dec 22 04:32:47 2012
DUMP: 70.54% done, finished in 0:20 at Sat Dec 22 04:35:18 2012
DUMP: 75.15% done, finished in 0:18 at Sat Dec 22 04:37:37 2012
DUMP: 80.28% done, finished in 0:14 at Sat Dec 22 04:39:10 2012
DUMP: 85.57% done, finished in 0:10 at Sat Dec 22 04:40:23 2012
DUMP: 90.50% done, finished in 0:07 at Sat Dec 22 04:41:46 2012
DUMP: SIGSEGV: ABORTING!
DUMP: DUMP: DUMP: SIGSEGV: ABORTING!
SIGSEGV: ABORTING!
SIGSEGV: ABORTING!
DUMP: SIGSEGV: ABORTING!
Following xz's fault, all other processes (dump, ccrypt, sh) dump their
cores to, but, according to dmesg, the culprit is always xz. According
to core, the fault is somewhere inside liblzma.so.5. I recompiled the
library to enable debugging and am trying to investigate, but, perhaps,
someone has already run into this?
Both systems used to run FreeBSD-8/amd64, when I first set the jobs up.
I have since upgraded the troublesome server to 9.1, but that did not
fix the problem.
The remote filesystem (where ccrypt writes encrypted output) is mounted
via smbfs on both servers, but I doubt that matters...
Any ideas? Thanks,
-mi
More information about the freebsd-stable
mailing list