I've had similar problems with the 2.3.x series. I believe I posted a
comment saying so to this mailing list. Basically I'll get crashes on an
SMP machine when accessing two different IDE busses simultaneously. It
used to happen instantly with earlier kernels, now it sometimes takes half
an hour to occcur. It still makes these kernels unusable as 2.2.x never
did this. I'm using a Promise Ultra66 controller with one IBM Deskstar on
each channel.
This doesn't sound exactly like your problem but it sure does sound
suspiciously similar.
I doubt it's the IDE code for the Promise card itself as I did a diff
of a kernel that had this problem and a kernel that didn't and the actual
IDE drive C file changed in a totally trivial manner.
Nicholas Vinen
On Tue, 21 Mar 2000, Simon Kirby wrote:
> Woops, sent that last email too early. :)
>
> On Tue, Mar 21, 2000 at 08:37:53AM -0800, Simon Kirby wrote:
>
> > - hdparm -u 1 vs hdparm -u 0 on all drives does not seem to make a
> > difference.
> >
> > - hdparm -d 0 vs hdparm -d 1 on all drives tends to make the problem
> > happen even more often (guessing 5 times as often). Tested with
> > both -u 1 and -u 0.
>
> - Memory pressure does not appear to affect the problem.
>
> - Booting "nosmp" seems to make the problem disappear (the problem seems
> to be SMP specific).
>
> - The problem only seems to happen when copying files between drives
> (copying on the same filesystem does not seem to reproduce it).
>
> - The problem does not seem to occur when copying files from one drive
> to another on the same IDE bus. This may be a coincidence of my setup,
> however -- I only have hda, hdb, and hdc, an hdc contains only 1K block
> EXT2 filesystems. hda and hdb contain 2K and 4K filesystems.
>
> The problem occurs when copying from /dev/hdc2 (1K EXT2 filesystem) to
> /dev/hdb1 (2K EXT2 filesystem). I haven't been able to reproduce it
> copying from hda4 (4K) to hdb2 (2K). I also haven't been able to
> reproduce it copying from hda4 (4K) to hdc2 (1K). I'm not sure if this
> is a result of the direction of the copying or an indication of a
> filesystem blocksize dependency.
>
> I just reproduced it copying in the exact opposite direction (/dev/hdb1
> to /dev/hdc2) from the original way I reproduced it. This is from a 2K
> to a 1K. Aha! I just reproduced it going from /dev/hdb4 (4K) to /dev/hdc2
> (1K), so the source blocksize does not seem to matter. I should see if I
> can reformat hdc to have a 4K filesystem, I suppose... Although I'm
> getting the feeling the problem might be completely independent of block
> size...
>
> So, SMP definitely makes a difference and IDE bus definitely makes a
> difference.
>
> Simon-
>
> [ Stormix Technologies Inc. ][ NetNation Communcations Inc. ]
> [ sim@stormix.com ][ sim@netnation.com ]
> [ Opinions expressed are not necessarily those of my employers. ]
>
> -
> 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/
>
-
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 : Thu Mar 23 2000 - 21:00:34 EST