On Tue, 23 May 2000, Matt Yourst wrote:
> As you may know, there were some deadlock problems with ReiserFS 3.6.5 and
> kernel >= 2.3.99-pre7-4. I submitted a patch for this problem earlier but it
> also broke the swap-out code in some other cases. Fortunately, whatever was
> originally causing the race condition with ReiserFS and writing out dirty mmaped
> pages now appears to be fixed as of 2.3.99-pre9-5. I have not stress-tested this
> prepatch though, so others might want to confirm this too.
The -pre9 + reiserfs 3.6.5 combo is still definitely *DANGEROUS*.
In particular, this breaks badly when you do a simple
apt-get dist-upgrade
When that happens, apt gets locked. Calling top, ps, killall, or even
ls -l /proc/<pid>
gives you a stuck, unkillable process.
strace points its fingers at sys_mlock(), but I don't know how reliable
this is.
This was still quite bearable. But there's worse. Once, I didn't reboot
after I stuck an apt-get (circa pre9-1), and worked with the machine. Next
reboot, bang ! My noffle database was clobbered... with copies of stuff
from /var/lib/dpkg/* (fortunately, it was not the other way around).
(currently, I've got only /var on reiserfs ; I had some plans to move the
other partitions (I've got a really bad pollen, acarians and e2fsck
allergy, the latter being absolutely not seasonal now) ; this will have to
wait for the 3.5.21 bugs to be squashed, I'm afraid was announced
recently).
Traces and tests (of new patches) welcome, on request.
-- Cyrille
------------------------------------------------------------------------------
Grumpf.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed May 31 2000 - 21:00:12 EST