2.6.22-rc7: known regressions with patches v2

Hi all,

Here is a list of some known regressions in 2.6.22-rc7 with patches available.

Feel free to add new regressions/remove fixed etc.
http://kernelnewbies.org/known_regressions

List of Aces

Name Regressions fixed since 21-Jun-2007

Hugh Dickins                           2
Andi Kleen                             2
Andrew Morton                          1
Benjamin Herrenschmidt                 1
Björn Steinbrink                       1
Bjorn Helgaas                          1
Christoph Lameter                      1
David Woodhouse                        1
Jean Delvare                           1
Olaf Hering                            1
Siddha, Suresh B                       1
Trent Piepho                           1

Ville Syrjälä 1

Unclassified

Subject : [2.6.22 regression fix] DLM must depend on SYSFS
References : http://lkml.org/lkml/2007/7/7/4
Last known good : ?
Submitter : Adrian Bunk <[email protected]>

Caused-By       : ?
Handled-By      : Adrian Bunk <[email protected]>
Status          : patch available

Subject : [2.6.22 regression fix] include/linux/kallsyms.h must #include <linux/errno.h>
References : http://lkml.org/lkml/2007/7/6/359
Last known good : ?
Submitter : Adrian Bunk <[email protected]>

Caused-By       : ?
Handled-By      : Adrian Bunk <[email protected]>
Status          : patch available

Bluetooth

Subject : problem with a bluetooth mouse
References : http://lkml.org/lkml/2007/7/4/321
Last known good : ?
Submitter : David Woodhouse <[email protected]>

Dmitry Torokhov <[email protected]>
Patch : http://lkml.org/lkml/2007/7/6/391
Status : patch available

Regards,
Michal

--
LOG
http://www.stardust.webpages.pl/log/

If you're looking for someone to blame, consider the fact that this has gone away now I've switched back to slab. :)

That patch was a straw man for Marcel to comment on, not necessarily a real fix. I don't know the code that well, and naïvely reordering bits of it could have introduced another race which I just haven't triggered yet.

Strictly speaking, it wasn't a 2.6.22 regression either -- I actually saw it for the first time in 2.6.21-rc iirc.

Thanks for tracking these, btw. It's very useful.

dwmw2

Yup, I forgot to add Christoph to CC :)

Regards,
Michal

LOG
http://www.stardust.webpages.pl/log/

Hmmmm.. Cannot see the corruption in the dump since it only includes 128 bytes and it seems to have occurred way into the object. If you want to dump 32k to your syslog to see the corruption then look at the print_trailer() function and change the parameter of the call to print_section

	print_section("Object", p, min(s->objsize, 128));

This is a 32k sized slab and SLAB does not allow debugging on power of two slabs > pagesize.... No surprise that SLAB does not complain.

What are these functions "malloc" and "free"? Would be good if those would be using kmalloc_track_caller() so that we can figure out who called them.

There's not a lot of those around. Looks like it's probably one of

init/do_mounts_rd.c:static void __init *malloc(size_t size) init/initramfs.c:static void __init *malloc(size_t size)

both of which exist just because those things include "lib/inflate.c".

It's probably do_mounts_rd.c, because initramfs.c defines malloc/free before the #include, and I would have expected gcc to just inline them in that case.

That's probably the right thign to do in do_mounts_rd.c too, rather than adding any kmalloc_track_caller() stuff.

Linus