Skip to contentSkip to conversation

wierd failures from -mm1

DISCUSSION Martin Bligh 2 participants 4 messages Apr 7, 2006 – Apr 8, 2006 1 message not in the archive

lore
Share & tools
lore
#0 lore raw

I hadn't mailed this out for a while, cause we weren't sure if it was -mm or a testing glitch, but there's been no -git releases, so Andy reran -mm to double check, and it still seems to be there. a subsequent test of rc1 + cons patches didn't hit this ... I think -mm has issues ;-)

Look at the 2.6.17-rc1-mm1 column from: http://test.kernel.org/

Drilling down into the console logs:

http://test.kernel.org/abat/27597/debug/console.log
Hangs after testing NMI watchdog.
http://test.kernel.org/abat/27596/debug/console.log
Hangs after bringing up cpus.

http://test.kernel.org/abat/27598/debug/console.log
http://test.kernel.org/abat/27593/debug/console.log
Both fail with reiserfs fsck errors; at first sight look like just dirty root partitions, but I don't think they are.

Filesystem is clean
Failed to lock the process to fsck the mounted ro partition. Bad address. fsck.reiserfs /dev/sda3 failed (status 0x8). Run manually!

Note that it's actually saying it's clean.

#2 lore raw
In reply to a message not in the archive
Quoted 7 lines · Dave Hansen
Quoted 2 lines · Martin Bligh

http://test.kernel.org/abat/27596/debug/console.log
Hangs after bringing up cpus.

See attached patch. It fixes curly.

Splendid -thanks. This may well fix the first two ... I think the reiser thing is likely still borked though.

M.

Quoted 55 lines

-- Dave

------------------------------------------------------------------------

Subject:
[PATCH 2.6.17-rc1-mm1] sched_domain-handle-kmalloc-failure-fix From:
Lee Schermerhorn <[email protected]>
Date:
Thu, 06 Apr 2006 15:58:47 -0400
To:
linux-kernel <[email protected]>

To:
linux-kernel <[email protected]>
CC:
Andrew Morton <[email protected]>, Eric Whitney <[email protected]>

[PATCH] sched_domain-handle-kmalloc-failure-fix

2.6.17-rc1-mm1 hangs during boot on HP rx8620 and dl585 -- both 4 node NUMA platforms. Problem is in build_sched_domains() setting up the sched_group_nodes[] lists, resulting from patch:
sched_domain-handle-kmalloc-failure.patch

The referenced patch does not propagate the "next" pointer from the head
of the list, resulting in a loop between the last 2 groups in the list.
This causes a tight loop/hang in init_numa_sched_groups_power() because 'sg->next' never == 'group_head' when you have > 2 nodes.

This patch seems to fix the problem.

Index: linux-2.6.17-rc1-mm1/kernel/sched.c
===================================================================

kernel/sched.c +1 -2
@@ -6360,7 +6360,7 @@ static int build_sched_domains(const cpu
 			}
 			sg->cpu_power = 0;
 			sg->cpumask = tmp;
-			sg->next = prev;
+			sg->next = prev->next;
 			cpus_or(covered, covered, tmp);
 			prev->next = sg;
 			prev = sg;
 
 
-

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#3 lore raw
Quoted 4 lines · Martin Bligh
Quoted 2 lines · Dave Hansen

On Fri, 2006-04-07 at 11:05 -0700, Martin Bligh wrote:

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

http://test.kernel.org/abat/27596/debug/console.log
Hangs after bringing up cpus.

See attached patch. It fixes curly.

Splendid -thanks. This may well fix the first two ... I think the reiser thing is likely still borked though.

The reiserfsck problem looks like a failed mlockall. Reverting mm-posix-memory-lock.patch should fix it.

#4 lore raw
Quoted 9 lines · Andrew Morton
Quoted 7 lines · Martin Bligh
Quoted 6 lines · Dave Hansen
Quoted 2 lines · Martin Bligh

http://test.kernel.org/abat/27596/debug/console.log
Hangs after bringing up cpus.

Quoted 8 lines
Quoted 4 lines
Quoted 1 line

See attached patch. It fixes curly.

Splendid -thanks. This may well fix the first two ... I think the reiser thing is likely still borked though.

The reiserfsck problem looks like a failed mlockall. Reverting mm-posix-memory-lock.patch should fix it.

Didn't manage to get that test kicked off before you released -mm2, which seems to work fine (across the boxes that still work, at least)

M.