2.6.20.3 - possible recursive locking detected - in XFS

2.6.20.3 - possible recursive locking detected - in XFS

[Jesper Juhl]

Hi,

For your information : 

Once in a while I see the message below after I've just created a new XFS filesystem, mount it and then start copying data to it.

It doesn't happen every time - if I should try to make a guess at the frequency I'd say about 1 in 30.

What I do is this.

1. boot the server (an IBM x336) via the network
2. log in via ssh
3. load scsi modules
4. create raid1 of two disks
5. fdisk /dev/sda and create partitions
6. mkfs.xfs /dev/sda<some_partition>
7. mount -t xfs /dev/sda<some_partition> /mnt/mountpoint
8. ssh -x root@otherserver.mydomain.tld "tar --create --gzip --one-file-system --file - / 2>/dev/null" | tar --extract --gzip --preserve-permissions --numeric-owner --directory /mnt/mountpoint --file - 

since I do the above via a remote ssh session to the net-booted server, I don't usually notice 
kernel output to the console. But recently I've started looking via dmesg in a second ssh session and 
I can say for sure that most of the time there's no problem, but when there is, it is the same 
"recursive locking" dump I get.

If more info is needed, just let me know.


...
Ending clean XFS mount for filesystem: sda4

=============================================
[ INFO: possible recursive locking detected ]
2.6.20.3generic #1
---------------------------------------------
xfs_fsr/6117 is trying to acquire lock:
 (&(&ip->i_lock)->mr_lock){----}, at: [<f929422d>] xfs_ilock+0x7d/0xa0 [xfs]

but task is already holding lock:
 (&(&ip->i_lock)->mr_lock){----}, at: [<f929422d>] xfs_ilock+0x7d/0xa0 [xfs]

other info that might help us debug this:
2 locks held by xfs_fsr/6117:
 #0:  (&inode->i_mutex/1){--..}, at: [<c016d8b5>] lookup_create+0x25/0x90
 #1:  (&(&ip->i_lock)->mr_lock){----}, at: [<f929422d>] xfs_ilock+0x7d/0xa0 [xfs]

stack backtrace:
 [<c010404a>] show_trace_log_lvl+0x1a/0x30
 [<c0104732>] show_trace+0x12/0x20
 [<c01047e6>] dump_stack+0x16/0x20
 [<c0139311>] __lock_acquire+0xb01/0xdf0
 [<c0139670>] lock_acquire+0x70/0x90
 [<c0133f8b>] down_write+0x3b/0x60
 [<f929422d>] xfs_ilock+0x7d/0xa0 [xfs]
 [<f9294e17>] xfs_iget+0x467/0x7b0 [xfs]
 [<f92aeeb8>] xfs_trans_iget+0x108/0x180 [xfs]
 [<f92997db>] xfs_ialloc+0xab/0x520 [xfs]
 [<f92afadc>] xfs_dir_ialloc+0x6c/0x2b0 [xfs]
 [<f92b82b9>] xfs_mkdir+0x399/0x650 [xfs]
 [<f92c0d99>] xfs_vn_mknod+0x119/0x2d0 [xfs]
 [<f92c0f68>] xfs_vn_mkdir+0x18/0x20 [xfs]
 [<c016ceb8>] vfs_mkdir+0x98/0xe0
 [<c016f57e>] sys_mkdirat+0x8e/0xd0
 [<c016f5e0>] sys_mkdir+0x20/0x30
 [<c0102f7e>] sysenter_past_esp+0x5f/0x99
 =======================
...


-- 
Jesper Juhl <jesper.juhl@gmail.com>

Re: 2.6.20.3 - possible recursive locking detected - in XFS

[David Chinner]

On Wed, Apr 25, 2007 at 11:16:57AM +0200, Jesper Juhl wrote:
> 
> Hi,
> 
> For your information : 
> 
> Once in a while I see the message below after I've just created a new XFS filesystem, mount it and then start copying data to it.

.....

> =============================================
> [ INFO: possible recursive locking detected ]
> 2.6.20.3generic #1
> ---------------------------------------------
> xfs_fsr/6117 is trying to acquire lock:
>  (&(&ip->i_lock)->mr_lock){----}, at: [<f929422d>] xfs_ilock+0x7d/0xa0 [xfs]
> 
> but task is already holding lock:
>  (&(&ip->i_lock)->mr_lock){----}, at: [<f929422d>] xfs_ilock+0x7d/0xa0 [xfs]

Known false positive - XFS doesn't have the annotations needed for
this yet; we've got a patch that will probably make it's way into 6.5.22 that
should fix most of these issues.

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

Re: 2.6.20.3 - possible recursive locking detected - in XFS

[Jesper Juhl]

On 26/04/07, David Chinner <dgc@sgi.com> wrote:
> On Wed, Apr 25, 2007 at 11:16:57AM +0200, Jesper Juhl wrote:
> >
> > Hi,
> >
> > For your information :
> >
> > Once in a while I see the message below after I've just created a new XFS filesystem, mount it and then start copying data to it.
>
> .....
>
> > =============================================
> > [ INFO: possible recursive locking detected ]
> > 2.6.20.3generic #1
> > ---------------------------------------------
> > xfs_fsr/6117 is trying to acquire lock:
> >  (&(&ip->i_lock)->mr_lock){----}, at: [<f929422d>] xfs_ilock+0x7d/0xa0 [xfs]
> >
> > but task is already holding lock:
> >  (&(&ip->i_lock)->mr_lock){----}, at: [<f929422d>] xfs_ilock+0x7d/0xa0 [xfs]
>
> Known false positive - XFS doesn't have the annotations needed for
> this yet; we've got a patch that will probably make it's way into 6.5.22 that
> should fix most of these issues.
>
Ok. Thanks for the feedback.

-- 
Jesper Juhl <jesper.juhl@gmail.com>
Don't top-post  http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please      http://www.expita.com/nomime.html

Re: 2.6.20.3 - possible recursive locking detected - in XFS

[Eric Sandeen]

Jesper Juhl wrote:
> On 26/04/07, David Chinner <dgc@sgi.com> wrote:

>> Known false positive - XFS doesn't have the annotations needed for
>> this yet; we've got a patch that will probably make it's way into 6.5.22 that
>> should fix most of these issues.

wow... 6.5.22.... do we really have to wait that long? :)

-Eric

Re: 2.6.20.3 - possible recursive locking detected - in XFS

[Klaus Strebel]

Eric Sandeen schrieb:
> Jesper Juhl wrote:
>> On 26/04/07, David Chinner <dgc@sgi.com> wrote:
> 
>>> Known false positive - XFS doesn't have the annotations needed for
>>> this yet; we've got a patch that will probably make it's way into 6.5.22 that
>>> should fix most of these issues.
> 
> wow... 6.5.22.... do we really have to wait that long? :)

Why wait? If i remember right ( it's about 3 years ago that it had to do
with it ), its a quite old patch-version of mother ... IRIX :-)

It's fun to see that your can't hide your roots Dave :-) ...

so now that .21 ( of the Linux 2.6 series ) i wonder, will there be
anybody remembering IRIX 6.5.22 when the Linux kernel reaches to that
version number? Hey Google, suck this message in and help us remember
this question !

Cheers to all friends of Linux und IRIX
Klaus

P.S.: Guess i will have to boot my old Indy at home, arrrggghhh,
clock-chip battery will be empty again ... 'll have to reset the MAC ...
must recover the lost root password ... oh Lord ( no, not you Steve ... )

-- 
Mit freundlichen Grüssen / best regards

Klaus Strebel, Dipl.-Inform. (FH), mailto:klaus.strebel@gmx.net

/"\
\ /     ASCII RIBBON CAMPAIGN
 X        AGAINST HTML MAIL
/ \