tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

wapbl corruption (was: Re: netbsd-5: panic: ureadc: non-positive resid)

On Sun, Oct 11, 2009 at 09:31:28PM -0400, Matthew Mondor wrote:
 > As a side note, since this panic occurred a few times unexpectedly
 > before I could investigate and isolate it to readv related tests (and
 > stop running the application causing it), several file corruption
 > problems occurred, using WAPBL.
 > Any very recently written file would have some unrelated data (which
 > seems to also be related to the application or another block recently
 > written by another application).  For instance, [...]
 > This suggests that after WAPBL replay a metadata reference might exist
 > for a block which shouldn't be kept, or that it referenced to a wrong
 > block...  I'm not familiar with FFS or WAPBL enough to really know
 > what's going on without further investigation.

This behavior is expected, the same as in traditional FFS, and for the
same reason: the FS does not bother to preserve referential integrity
for data. If you don't like it, use softupdates.

David A. Holland

Home | Main Index | Thread Index | Old Index