fw: [PATCH] fix instant oops with tracing enabled

fw: [PATCH] fix instant oops with tracing enabled

[Dave Chinner]

SGI folks,

Looks like Christoph is having problems posting to the list;
the spam filter is dropping all his mail. In the mean time,
here's a fix for an oops in the tracing code as a result of
the last check ins. I didn't see this because the "combine
inodes" patches removes xfs_icount altogether. If that
series is going to be included in the current round of checkins
then this patch probably isn't needed.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com


From: Christoph Hellwig <hch@lst.de>

We can only read inode->i_count if the inode is actually there and not
a NULL pointer.  This was introduced in one of the recent sync patches.

Signed-off-by: Christoph Hellwig <hch@lst.de>

Index: linux-2.6-xfs/fs/xfs/linux-2.6/xfs_vnode.c
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/linux-2.6/xfs_vnode.c	2008-10-13 12:07:38.000000000 -0400
+++ linux-2.6-xfs/fs/xfs/linux-2.6/xfs_vnode.c	2008-10-13 12:07:47.000000000 -0400
@@ -92,7 +92,7 @@ static inline int xfs_icount(struct xfs_
 {
 	struct inode *inode = VFS_I(ip);
 
-	if (!inode)
+	if (inode)
 		return atomic_read(&inode->i_count);
 	return -1;
 }

Re: fw: [PATCH] fix instant oops with tracing enabled

[Mark Goodwin]

Dave Chinner wrote:
> SGI folks,
>
> Looks like Christoph is having problems posting to the list;
> the spam filter is dropping all his mail. In the mean time,
> here's a fix for an oops in the tracing code as a result of
> the last check ins. I didn't see this because the "combine
> inodes" patches removes xfs_icount altogether.

Lachlan also saw some regressions after merging these patchsets :
. replace the mount inode list with radix tree traversals
. clean up sync code

> If that
> series is going to be included in the current round of checkins
> then this patch probably isn't needed.

The agreed plan for 2.6.28 still has the following patchsets to go in:

. Combine the XFS and Linux inode structures V2
. Track reclaimable inodes in inode cache
. AIL cleanup and bug fixes
. Account for allocated blocks when expanding directories
. Check for valid transaction headers in recovery
. fix remount rw with unrecognized options

It's starting to look like a pretty aggressive merge and QA schedule.
Dave, is it worth doing any testing until these are *all* merged?

Thanks
-- Mark

Re: fw: [PATCH] fix instant oops with tracing enabled

[Dave Chinner]

On Tue, Oct 14, 2008 at 10:40:15AM +1000, Mark Goodwin wrote:
>
>
> Dave Chinner wrote:
>> SGI folks,
>>
>> Looks like Christoph is having problems posting to the list;
>> the spam filter is dropping all his mail. In the mean time,
>> here's a fix for an oops in the tracing code as a result of
>> the last check ins. I didn't see this because the "combine
>> inodes" patches removes xfs_icount altogether.
>
> Lachlan also saw some regressions after merging these patchsets :
> . replace the mount inode list with radix tree traversals
> . clean up sync code

Can you share with us all what those problems are? I can't help
find and fix the problems if I don't get told about them. perhaps
you should be opening bugzilla bugs rather than internal bugworks
PVs for regressions as a result of merges of community patch
sets....

>> If that
>> series is going to be included in the current round of checkins
>> then this patch probably isn't needed.
>
> The agreed plan for 2.6.28 still has the following patchsets to go in:
>
> . Combine the XFS and Linux inode structures V2
> . Track reclaimable inodes in inode cache
> . AIL cleanup and bug fixes
> . Account for allocated blocks when expanding directories
> . Check for valid transaction headers in recovery
> . fix remount rw with unrecognized options
>
> It's starting to look like a pretty aggressive merge and QA schedule.

We've got all of the -rc series to address regressions.

> Dave, is it worth doing any testing until these are *all* merged?

IMO, no, but that's up to you guys. I'd just merge them, run some
basic QA then push them to linus.  We've still got the whole -rc
series to address regressions.  And if you tell us about
regressions, we can help track them down and get them fixed quickly.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: fw: [PATCH] fix instant oops with tracing enabled

[Christoph Hellwig]

On Tue, Oct 14, 2008 at 10:40:15AM +1000, Mark Goodwin wrote:
> Lachlan also saw some regressions after merging these patchsets :
> . replace the mount inode list with radix tree traversals
> . clean up sync code

What exactly?  I saw some softlookup in 042, but when applying Dave's
xfs_sync_inodeS_ag fix (or the hal of it applying without the del inodes
tracking in the radix tree) it goes away.

> >If that
> >series is going to be included in the current round of checkins
> >then this patch probably isn't needed.
> 
> The agreed plan for 2.6.28 still has the following patchsets to go in:
> 
> . Combine the XFS and Linux inode structures V2
> . Track reclaimable inodes in inode cache
> . AIL cleanup and bug fixes
> . Account for allocated blocks when expanding directories
> . Check for valid transaction headers in recovery
> . fix remount rw with unrecognized options


3-6 are small bug fixes and should go in ASAP.  I'd really like to see 1
and 2 and volunter to help sorting out any fallout.  Not entirely sure
about the AIL patches - they seem ready but at least they don't have
much impact on everything else.   So if you really want to reduce the
amount of patches those would be the ones.

Re: fw: [PATCH] fix instant oops with tracing enabled

[Dave Chinner]

On Wed, Oct 15, 2008 at 11:27:09AM +1000, Lachlan McIlroy wrote:
> Christoph Hellwig wrote:
>> On Tue, Oct 14, 2008 at 10:40:15AM +1000, Mark Goodwin wrote:
>>> Lachlan also saw some regressions after merging these patchsets :
>>> . replace the mount inode list with radix tree traversals
>>> . clean up sync code
>>
>> What exactly?  I saw some softlookup in 042, but when applying Dave's
>> xfs_sync_inodeS_ag fix (or the hal of it applying without the del inodes
>> tracking in the radix tree) it goes away.
>
> I saw this panic but I don't think it's related to the above patches:
>
> [252921.307588] BUG: unable to handle kernel <3>BUG: scheduling while atomic: dd/16976/0xf101da90

Isn't there another line with this ouutput that looks like:

	atomic = 1 in_interrupt = 0

To indicate the "atomic" reason?

> [252921.307908] Modules linked in:
> [252921.307911] Pid: 16976, comm: dd Not tainted 2.6.27-rc8 #183
> [252921.307913] [252921.307913] Call Trace:

[ snip exceedingly deep stack that'll blow a 4k ia32 stack
completely ]

In summary, the stack is:

	write
	  balance_dirty_pages
	    xfs_iomap_write_allocate
	      <enter memory reclaim>
	      try_to_free_pages
	        xfs_iomap_write_allocate
		   _xfs_trans_commit
		     xlog_write
		       xlog_state_get_iclog_space
		         <sleep>

The question is what is the reason for running in atomic mode?
The only place I can see a sleep happening in this function is
the call to sv_wait(), which means the atomic state must have come
from higher up.... Seems very strange.

> I saw sync get stuck in an infinite loop running test 042 - maybe the same
> problem you saw.

Yes, that's the lockup that the later patch I posted fixes.

> I saw the panic in _xfs_itrace_exit() which has now been fixed.
>
> And I also saw this assertion:
>
> <4>[34770.626472] Assertion failed: (index >= 0) && (index < ktp->kt_nentries), file: fs/xfs/support/ktrace.c, line: 173
> <0>[34770.626511] ------------[ cut here ]------------
> <2>[34770.627419] kernel BUG at fs/xfs/support/debug.c:81!

I can't see how that is related to the changes - it's a trace
buffer index overrun. That kind of implies that the ktrace_t
has been corrupted. Memory corruption of some kind?

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: fw: [PATCH] fix instant oops with tracing enabled

[Lachlan McIlroy]

Christoph Hellwig wrote:
> On Tue, Oct 14, 2008 at 10:40:15AM +1000, Mark Goodwin wrote:
>> Lachlan also saw some regressions after merging these patchsets :
>> . replace the mount inode list with radix tree traversals
>> . clean up sync code
> 
> What exactly?  I saw some softlookup in 042, but when applying Dave's
> xfs_sync_inodeS_ag fix (or the hal of it applying without the del inodes
> tracking in the radix tree) it goes away.

I saw this panic but I don't think it's related to the above patches:

[252921.307588] BUG: unable to handle kernel <3>BUG: scheduling while atomic: dd/16976/0xf101da90
[252921.307908] Modules linked in:
[252921.307911] Pid: 16976, comm: dd Not tainted 2.6.27-rc8 #183
[252921.307913] 
[252921.307913] Call Trace:
[252921.307920]  [<ffffffff8102fe22>] __schedule_bug+0x62/0x66
[252921.307923]  [<ffffffff8153dce1>] schedule+0x99/0x7c7
[252921.307925]  [<ffffffff8153e890>] schedule_timeout+0x22/0xb4
[252921.307929]  [<ffffffff810473f9>] ? add_wait_queue_exclusive+0x3c/0x41
[252921.307932]  [<ffffffff81198bc9>] xlog_state_get_iclog_space+0xe8/0x278
[252921.307934]  [<ffffffff8102de2d>] ? default_wake_function+0x0/0xf
[252921.307936]  [<ffffffff81198e6d>] xlog_write+0x114/0x579
[252921.307938]  [<ffffffff811761d5>] ? xfs_buf_item_pin+0x76/0x7b
[252921.307940]  [<ffffffff811993a7>] xfs_log_write+0x38/0x62
[252921.307943]  [<ffffffff811a4f6b>] _xfs_trans_commit+0x1fd/0x3c6
[252921.307945]  [<ffffffff81193e93>] xfs_iomap_write_allocate+0x2d5/0x387
[252921.307947]  [<ffffffff81194e07>] xfs_iomap+0x331/0x3ba
[252921.307950]  [<ffffffff811b0930>] xfs_map_blocks+0x30/0x69
[252921.307952]  [<ffffffff811b1a00>] xfs_page_state_convert+0x2e5/0x594
[252921.307953]  [<ffffffff811b1f1b>] ? xfs_vm_releasepage+0xae/0xbd
[252921.307955]  [<ffffffff811b1ff1>] xfs_vm_writepage+0xc7/0x109
[252921.307958]  [<ffffffff81080835>] shrink_page_list+0x31a/0x57c
[252921.307960]  [<ffffffff81080be3>] shrink_inactive_list+0x126/0x39d
[252921.307962]  [<ffffffff81080f3f>] shrink_zone+0xe5/0x10a
[252921.307964]  [<ffffffff81081436>] try_to_free_pages+0x248/0x3cf
[252921.307966]  [<ffffffff8108042f>] ? isolate_pages_global+0x0/0x34
[252921.307967]  [<ffffffff8107cc3c>] __alloc_pages_internal+0x262/0x3b6
[252921.307969]  [<ffffffff811b4284>] ? xfs_buf_get_flags+0x6b/0x165
[252921.307972]  [<ffffffff8109709f>] alloc_pages_current+0xb9/0xc2
[252921.307974]  [<ffffffff8109d66b>] new_slab+0x57/0x283
[252921.307975]  [<ffffffff8109daeb>] __slab_alloc+0x1e8/0x3dd
[252921.307977]  [<ffffffff811b0220>] ? kmem_zone_alloc+0x58/0xaa
[252921.307980]  [<ffffffff811638c1>] ? xfs_bmap_search_multi_extents+0x9a/0xda
[252921.307982]  [<ffffffff8109e07e>] kmem_cache_alloc+0x43/0x76
[252921.307983]  [<ffffffff811b0220>] kmem_zone_alloc+0x58/0xaa
[252921.307985]  [<ffffffff811b0281>] kmem_zone_zalloc+0xf/0x31
[252921.307986]  [<ffffffff811a555c>] _xfs_trans_alloc+0x25/0x5f
[252921.307988]  [<ffffffff811a562c>] xfs_trans_alloc+0x96/0xa1
[252921.307990]  [<ffffffff81193d05>] xfs_iomap_write_allocate+0x147/0x387
[252921.307991]  [<ffffffff81194db4>] ? xfs_iomap+0x2de/0x3ba
[252921.307993]  [<ffffffff81194e07>] xfs_iomap+0x331/0x3ba
[252921.307995]  [<ffffffff811b0930>] xfs_map_blocks+0x30/0x69
[252921.307996]  [<ffffffff811b1a00>] xfs_page_state_convert+0x2e5/0x594
[252921.307998]  [<ffffffff811b1ff1>] xfs_vm_writepage+0xc7/0x109
[252921.307999]  [<ffffffff8107cec6>] __writepage+0x12/0x2b
[252921.308001]  [<ffffffff8107d39a>] write_cache_pages+0x1b3/0x317
[252921.308003]  [<ffffffff8107ceb4>] ? __writepage+0x0/0x2b
[252921.308004]  [<ffffffff8107d51d>] generic_writepages+0x1f/0x25
[252921.308006]  [<ffffffff811b20ca>] xfs_vm_writepages+0x43/0x4b
[252921.308007]  [<ffffffff8107d54b>] do_writepages+0x28/0x37
[252921.308011]  [<ffffffff810bfd82>] __writeback_single_inode+0x145/0x29f
[252921.308015]  [<ffffffff812283c5>] ? prop_fraction_single+0x3d/0x5f
[252921.308017]  [<ffffffff810c0294>] generic_sync_sb_inodes+0x1d0/0x2ba
[252921.308019]  [<ffffffff810c0387>] sync_sb_inodes+0x9/0xb
[252921.308021]  [<ffffffff810c06f3>] writeback_inodes+0x64/0xad
[252921.308023]  [<ffffffff8107da26>] balance_dirty_pages_ratelimited_nr+0x16b/0x2dd
[252921.308027]  [<ffffffff8107769f>] generic_file_buffered_write+0x203/0x625
[252921.308028]  [<ffffffff8107c16d>] ? get_page_from_freelist+0x45e/0x5d0
[252921.308031]  [<ffffffff811b8b80>] ? xfs_rw_enter_trace+0xbf/0xca
[252921.308032]  [<ffffffff811b9641>] xfs_write+0x64f/0x9cf
[252921.308035]  [<ffffffff81076b4e>] ? find_lock_page+0x2b/0x61
[252921.308037]  [<ffffffff811b50c3>] __xfs_file_write+0x4c/0x4e
[252921.308038]  [<ffffffff811b50e9>] xfs_file_aio_write+0x11/0x13
[252921.308040]  [<ffffffff810a2f94>] do_sync_write+0xe2/0x126
[252921.308042]  [<ffffffff81084935>] ? __do_fault+0x326/0x36c
[252921.308044]  [<ffffffff810471d3>] ? autoremove_wake_function+0x0/0x38
[252921.308047]  [<ffffffff811e8618>] ? selinux_file_permission+0x10d/0x116
[252921.308050]  [<ffffffff811e1321>] ? security_file_permission+0x11/0x13
[252921.308052]  [<ffffffff810a3790>] vfs_write+0xae/0x157
[252921.308053]  [<ffffffff810a3c9e>] sys_write+0x47/0x6f
[252921.308055]  [<ffffffff8100bf3b>] system_call_fastpath+0x16/0x1b
[252921.308056] 
[252921.308125] paging request at ffff881829c85a78
[252921.308125] IP: [<ffffffff810297a3>] cpuacct_charge+0x2b/0x34
[252921.308125] PGD 202063 PUD 0 
[252921.308125] Oops: 0000 [1] SMP 


I saw sync get stuck in an infinite loop running test 042 - maybe the same
problem you saw.

[1]kdb> btp 7356
Stack traceback for pid 7356
0xffff880071d10740     7356     7390  1    2   R  0xffff880071d10ba8  sync
sp                ip                Function (args)
0xffff880058cc3c88 0xffffffff81540566 kdb_interrupt+0x66 (0xffff8800720e4ac4, 0x202, 0x0, 0xffff88007119b810, 0xffff880058cc3d48, 0xffff88007213deb8)
0xffff880058cc3ce8 0xffffffff8153ff8e _spin_unlock_irqrestore+0x8 (0xffff8800720e4ac4, 0x202)
0xffff880058cc3d20 0xffffffff81229b96 __down_read_trylock+0x3f (invalid)
0xffff880058cc3d40 0xffffffff8104a34d down_read_trylock+0x9
0xffff880058cc3d50 0xffffffff8118bcd9 xfs_ilock_nowait+0xaf (0xffff8800720e4a00, invalid)
0xffff880058cc3d80 0xffffffff811bc3d9 xfs_sync_inodes_ag+0x12a (0xffff88007119b800, invalid, invalid, 0x0)
0xffff880058cc3e00 0xffffffff811bc6ee xfs_sync_inodes+0x65 (0xffff88007119b800, invalid, 0x0)
0xffff880058cc3e40 0xffffffff811bc785 xfs_syncsub+0x67 (0xffff88007119b800, invalid, 0x0)
0xffff880058cc3e80 0xffffffff811bc9d0 xfs_sync+0x7d (0xffff88007119b800, invalid)
0xffff880058cc3eb0 0xffffffff811ba6b9 xfs_fs_sync_super+0x38 (0xffff88007e056000)
0xffff880058cc3f20 0xffffffff810a5311 sync_filesystems+0xb7 (invalid)
0xffff880058cc3f50 0xffffffff810c2deb do_sync+0x37 (0x1)
0xffff880058cc3f70 0xffffffff810c2e25 sys_sync+0xe
  not matched: from 0xffffffff8100bfad to 0xffffffff8100c025 drop_through 0 bb_jmp[7


I saw the panic in _xfs_itrace_exit() which has now been fixed.

And I also saw this assertion:

<4>[34770.626472] Assertion failed: (index >= 0) && (index < ktp->kt_nentries), file: fs/xfs/support/ktrace.c, line: 173
<0>[34770.626511] ------------[ cut here ]------------
<2>[34770.627419] kernel BUG at fs/xfs/support/debug.c:81!

[2]kdb> bt
Stack traceback for pid 400
0xffff88007e883a00      400        2  1    2   R  0xffff88007e883e68 *xfslogd/2
sp                ip                Function (args)
0xffff88007e66fbf8 0xffffffff811bd5d5 assfail+0x1a (invalid, invalid, invalid)
0xffff88007e66fc28 0xffffffff811bdb24 ktrace_enter+0x8b (invalid, invalid, invalid, invalid, invalid, invalid, invalid, invalid, invalid)
0xffff88007e66fc78 0xffffffff81175b35 xfs_buf_item_trace+0xe6 (0xffffffff816d8948, 0xffff88007c47cb40)
0xffff88007e66fd18 0xffffffff81175b60 xfs_buf_item_committed+0x1c (0xffff88007c47cb40, 0x100000b1f)
0xffff88007e66fd38 0xffffffff811a4766 xfs_trans_chunk_committed+0x60 (0xffff880050124780, 0x100000b1f, invalid)
0xffff88007e66fd98 0xffffffff811a4873 xfs_trans_committed+0x43 (0xffff880050124670, invalid)
0xffff88007e66fdc8 0xffffffff81197b2a xlog_state_do_callback+0x19a (0xffff88007ef78400, invalid, 0xffff88007ef79000)
0xffff88007e66fe38 0xffffffff81197d6d xlog_state_done_syncing+0xda (0xffff88007ef79000, invalid)
0xffff88007e66fe68 0xffffffff81198587 xlog_iodone+0x154 (0xffff88006ac37c80)
0xffff88007e66fe98 0xffffffff811b3afb xfs_buf_iodone_work+0x65 (invalid)
0xffff88007e66feb8 0xffffffff81043cfb run_workqueue+0x7c (0xffff88007e866b80)
0xffff88007e66fed8 0xffffffff81044711 worker_thread+0xd8 (0xffff88007e866b80)
0xffff88007e66ff28 0xffffffff810470a3 kthread+0x49 (invalid)
0xffff88007e66ff48 0xffffffff8100ce89 child_rip+0xa (invalid, invalid)


> 
>>> If that
>>> series is going to be included in the current round of checkins
>>> then this patch probably isn't needed.
>> The agreed plan for 2.6.28 still has the following patchsets to go in:
>>
>> . Combine the XFS and Linux inode structures V2
>> . Track reclaimable inodes in inode cache
>> . AIL cleanup and bug fixes
>> . Account for allocated blocks when expanding directories
>> . Check for valid transaction headers in recovery
>> . fix remount rw with unrecognized options
> 
> 
> 3-6 are small bug fixes and should go in ASAP.  I'd really like to see 1
> and 2 and volunter to help sorting out any fallout.  Not entirely sure
> about the AIL patches - they seem ready but at least they don't have
> much impact on everything else.   So if you really want to reduce the
> amount of patches those would be the ones.
> 
> 
> 

Re: fw: [PATCH] fix instant oops with tracing enabled

[Lachlan McIlroy]

Dave Chinner wrote:
> On Wed, Oct 15, 2008 at 11:27:09AM +1000, Lachlan McIlroy wrote:
>> Christoph Hellwig wrote:
>>> On Tue, Oct 14, 2008 at 10:40:15AM +1000, Mark Goodwin wrote:
>>>> Lachlan also saw some regressions after merging these patchsets :
>>>> . replace the mount inode list with radix tree traversals
>>>> . clean up sync code
>>> What exactly?  I saw some softlookup in 042, but when applying Dave's
>>> xfs_sync_inodeS_ag fix (or the hal of it applying without the del inodes
>>> tracking in the radix tree) it goes away.
>> I saw this panic but I don't think it's related to the above patches:
>>
>> [252921.307588] BUG: unable to handle kernel <3>BUG: scheduling while atomic: dd/16976/0xf101da90
> 
> Isn't there another line with this ouutput that looks like:
> 
> 	atomic = 1 in_interrupt = 0
> 
> To indicate the "atomic" reason?
No, no other output.

> 
>> [252921.307908] Modules linked in:
>> [252921.307911] Pid: 16976, comm: dd Not tainted 2.6.27-rc8 #183
>> [252921.307913] [252921.307913] Call Trace:
> 
> [ snip exceedingly deep stack that'll blow a 4k ia32 stack
> completely ]
> 
> In summary, the stack is:
> 
> 	write
> 	  balance_dirty_pages
> 	    xfs_iomap_write_allocate
> 	      <enter memory reclaim>
> 	      try_to_free_pages
> 	        xfs_iomap_write_allocate
> 		   _xfs_trans_commit
> 		     xlog_write
> 		       xlog_state_get_iclog_space
> 		         <sleep>
> 
> The question is what is the reason for running in atomic mode?
> The only place I can see a sleep happening in this function is
> the call to sv_wait(), which means the atomic state must have come
> from higher up.... Seems very strange.
Yeah it's got me bugged too.  I had seen a similar problem in this
code before where the tracing code was allocating memory with a
spinlock held but that's gone now.

> 
>> I saw sync get stuck in an infinite loop running test 042 - maybe the same
>> problem you saw.
> 
> Yes, that's the lockup that the later patch I posted fixes.
Good.

> 
>> I saw the panic in _xfs_itrace_exit() which has now been fixed.
>>
>> And I also saw this assertion:
>>
>> <4>[34770.626472] Assertion failed: (index >= 0) && (index < ktp->kt_nentries), file: fs/xfs/support/ktrace.c, line: 173
>> <0>[34770.626511] ------------[ cut here ]------------
>> <2>[34770.627419] kernel BUG at fs/xfs/support/debug.c:81!
> 
> I can't see how that is related to the changes - it's a trace
> buffer index overrun. That kind of implies that the ktrace_t
> has been corrupted. Memory corruption of some kind?

Yeah or maybe a use after free.  The buffer and the buffer log
item look sane but the ktrace hanging off the buffer log item
is hosed.  It's the first time I've seen it and it may just be
a coincidence.