On Wed, 13 Sep 2000, Marco d'Itri wrote:
> On Sep 13, tytso@mit.edu wrote:
>
> > * Innd data corruption, probably caused by bug truncation bug (Rik
> > van Riel)
> This bug has not been fixed, I can still reproduce it (but not every
> time). This is how it happens:
>
> - INN (1.7.2+insync+other patches, the debian package I maintain) is
> running
> - active is correct
> - I post an article, which is filed with the correct number
> - for *this group only* the high value is wrong and equal to the one
> in the active file I precedently saved (in some cases this does not
> happen and I can't notice anything wrong in the active file)
> - I stop INN
> - the whole active file is now 100% identical to the saved copy
Ugh... How about relevant subset of strace?
> Right now it happened after the daily expire run: I stopped INN and the
> file on disk changed to the copy I saved before expire started.
Wait a minute. I don't believe in on-disk file being restored by magic,
but I could believe in page(s) being never written to disk and giving the
impression of "update that doesn't stick". You have a file shorter than
one page, so in principle it seems to point to the handling of partially
truncated pages. Hmm...
BTW, how does test8+patch to block_truncate_page() behave? And what is the
block size on your fs?
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Sep 15 2000 - 21:00:23 EST