On 2010-03-08, at 03:40, Sergey Poznyakoff wrote:
Frankly speaking I cannot imagine a real-life situation when (BILLION <= ns || ns < 0). Nevertheless, a couple of lines won't hurt, so I've installed the patch. Thank you.
We also hit this in production with a patched ext3 (proto-ext4). Whether it was originally due to a software bug, or on-disk/memory corruption, I don't know, but it caused tar to segfault repeatedly.
I don't think that e2fsck actually validates the nanoseconds fields, so once such a timestamp is on disk it will not go away.
Thanks for applying the patch. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc.
