fsr stucks in D-state

fsr stucks in D-state

[Marvin]

Hi,

I observed a hang in D-state when using xfs_fsr.

- xfsprogs is from git (3.0.2)
- kernel also (2.6.30-rc7...)

I tried 2.6.29.4 and it worked flawless. Machine is powerpc32.

Here is the sysrq-t trace:

xfs_fsr       D 480f7f3c     0  4211   4210                  
Call Trace:                                                  
[d1adfc40] [00001032] 0x1032 (unreliable)                    
[d1adfd00] [c0008dd8] __switch_to+0x4c/0x6c                  
[d1adfd20] [c02c5294] __schedule+0x230/0x360                 
[d1adfd70] [c02c53ec] schedule+0x28/0x54                     
[d1adfd80] [c02c7964] rwsem_down_failed_common+0x104/0x25c   
[d1adfdb0] [c02c7ae4] rwsem_down_write_failed+0x28/0x40      
[d1adfde0] [c02c6994] down_write+0x30/0x40                   
[d1adfdf0] [c012399c] xfs_ilock+0x9c/0xa0                    
[d1adfe00] [c014375c] xfs_inactive+0x2e8/0x48c               
[d1adfe40] [c015113c] xfs_fs_clear_inode+0x40/0x50           
[d1adfe50] [c00ab764] clear_inode+0x68/0xec                  
[d1adfe60] [c00ac02c] generic_delete_inode+0x16c/0x170       
[d1adfe80] [c00aaf20] iput+0x70/0x90                         
[d1adfe90] [c00a7424] dentry_iput+0xc8/0x154                 
[d1adfeb0] [c00a75e0] d_kill+0x50/0x8c                       
[d1adfed0] [c00a80cc] dput+0x110/0x210                       
[d1adfef0] [c0095b50] __fput+0x178/0x204                     
[d1adff10] [c0091eec] filp_close+0x6c/0xa0                   
[d1adff30] [c0091fd4] sys_close+0xb4/0x124                   
[d1adff40] [c0011fc4] ret_from_syscall+0x0/0x38              
--- Exception: c01 at 0x480f7f3c
    LR = 0x1000397c

Greetings

Marvin

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: fsr stucks in D-state

[Marvin]

Hi again,

On Monday 01 June 2009 12:33:39 Marvin wrote:
> Hi,
>
> I observed a hang in D-state when using xfs_fsr.
>
> - xfsprogs is from git (3.0.2)
> - kernel also (2.6.30-rc7...)

I pulled from "git://oss.sgi.com/xfs/xfs for-linus", which seems to include 
some deadlock fixes. Now, fsr no longer deadlocks, so this seems to be fixed. 
Sorry for the noise.

Marvin


> I tried 2.6.29.4 and it worked flawless. Machine is powerpc32.
>
> Here is the sysrq-t trace:
>
> xfs_fsr       D 480f7f3c     0  4211   4210
> Call Trace:
> [d1adfc40] [00001032] 0x1032 (unreliable)
> [d1adfd00] [c0008dd8] __switch_to+0x4c/0x6c
> [d1adfd20] [c02c5294] __schedule+0x230/0x360
> [d1adfd70] [c02c53ec] schedule+0x28/0x54
> [d1adfd80] [c02c7964] rwsem_down_failed_common+0x104/0x25c
> [d1adfdb0] [c02c7ae4] rwsem_down_write_failed+0x28/0x40
> [d1adfde0] [c02c6994] down_write+0x30/0x40
> [d1adfdf0] [c012399c] xfs_ilock+0x9c/0xa0
> [d1adfe00] [c014375c] xfs_inactive+0x2e8/0x48c
> [d1adfe40] [c015113c] xfs_fs_clear_inode+0x40/0x50
> [d1adfe50] [c00ab764] clear_inode+0x68/0xec
> [d1adfe60] [c00ac02c] generic_delete_inode+0x16c/0x170
> [d1adfe80] [c00aaf20] iput+0x70/0x90
> [d1adfe90] [c00a7424] dentry_iput+0xc8/0x154
> [d1adfeb0] [c00a75e0] d_kill+0x50/0x8c
> [d1adfed0] [c00a80cc] dput+0x110/0x210
> [d1adfef0] [c0095b50] __fput+0x178/0x204
> [d1adff10] [c0091eec] filp_close+0x6c/0xa0
> [d1adff30] [c0091fd4] sys_close+0xb4/0x124
> [d1adff40] [c0011fc4] ret_from_syscall+0x0/0x38
> --- Exception: c01 at 0x480f7f3c
>     LR = 0x1000397c
>
> Greetings
>
> Marvin

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: fsr stucks in D-state

[Christoph Hellwig]

On Tue, Jun 02, 2009 at 08:30:18AM +0200, Marvin wrote:
> I pulled from "git://oss.sgi.com/xfs/xfs for-linus", which seems to include 
> some deadlock fixes. Now, fsr no longer deadlocks, so this seems to be fixed. 
> Sorry for the noise.

I'm pretty sure the fix was:

	xfs: fix double unlock in xfs_swap_extents()

from Felix.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs