Hmmm .. there are only slight differences between 17july, 22july, and
1august patch. Are there any other patches included during the time?
The only real change between 22july and 1augus patch is: the debugging hack
in try_to_free_page() of mm/vmscan.c is removed and an paranoid setting for
swap_out() is added.
This change would only fail if try_to_swap_out() has at most return values
greater than 1 ... which would produce a "Couldn't get free page"
or something similar.
... it would very exiting if the old hack with the run_task_queue(&tq_disk)
is a ``solution''. IMHO this would indicate a real bug im the memory
management.
Do you have a 2.0.29 available for testing (LILO boot: linux.old)?
Only for a hardware check up.
Werner
>
> I use extensively 2.0.29 + some patches (I'm writing this under 2.0.29),
> not any problem doing the exact same simple thing: as root, cd /usr/src,
> tar xzvf /archive/ftp/pub/KPatches/linux-2.0.27.tar.gz.
>
> I just booted with 31-2 + your patch, trashed my /usr partition by doing
> it again under /usr/src.
> Thanks to Ted for the 1.10 e2fsck, it works much better than previous release
> :).
>
> I've now set up a spare partition that i mount under /CRASH,
> and reproduced in once again.
>
> I can untar anything I want with 2.0.29 witout any glitch.
>
> The two last oopses were more straightforward than yesterday, so i don't have
> anything in the log. I've just written down the last message on the console,
> that led to the total freeze (no VCs, no ping, nothing).
>
> It was trigered by doing sync on another VC after the latest oops.
>
> BTW: all my partition are formatted with a 2kbytes block size.
>
> scsi: aborting command due to timeout: pid 2508 (btw, this pid is rather high,
> the box was booted 1 min ago.. strange), scsi0, channel0, id0, lun0 0x2a 00
> 00 38 50 7a 00 00 78 00
>
> will recompile a bare bone pre31-2 with only your patch to see.
> I have this kernel packaged as a .deb, can give it to anyone intersted.