Skip to contentSkip to conversation

CIFS & Lockdep warnings

DISCUSSION Srihari Vijayaraghavan 3 participants 4 messages Aug 14, 2006

lore
Share & tools
lore
#0 lore raw

This is observed on 2.6.18-rc4 on SUSE 10.1 x86 on P-IV. The volume is question was mounted from a Windows 2003 server.

=======================================================
[ INFO: possible circular locking dependency detected ]
-------------------------------------------------------
cp/11790 is trying to acquire lock:
(iprune_mutex){--..}, at: [<c029e364>]
mutex_lock+0x19/0x20

but task is already holding lock:
(&inode->i_mutex){--..}, at: [<c029e364>]
mutex_lock+0x19/0x20

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&inode->i_mutex){--..}:

       [<c012a1c0>] lock_acquire+0x56/0x73
       [<c029e205>] __mutex_lock_slowpath+0xa6/0x1ec
       [<c029e364>] mutex_lock+0x19/0x20
       [<e3cd2b7a>] ntfs_put_inode+0x3e/0x79 [ntfs]
       [<c0169ac1>] iput+0x33/0x70
       [<c017738e>] inotify_unmount_inodes+0x12e/0x15f
       [<c016a55c>] invalidate_inodes+0x38/0xd1
       [<c01599b3>] generic_shutdown_super+0x5a/0x108
       [<c0159a81>] kill_block_super+0x20/0x36
       [<c0159b56>] deactivate_super+0x61/0x78
       [<c016c561>] mntput_no_expire+0x44/0x78
       [<c015f536>] path_release_on_umount+0x16/0x1d
       [<c016d692>] sys_umount+0x1d2/0x208
       [<c016d6d5>] sys_oldumount+0xd/0xf
       [<c0102cfb>] syscall_call+0x7/0xb

-> #0 (iprune_mutex){--..}:

       [<c012a1c0>] lock_acquire+0x56/0x73
       [<c029e205>] __mutex_lock_slowpath+0xa6/0x1ec
       [<c029e364>] mutex_lock+0x19/0x20
       [<c016a3a2>] shrink_icache_memory+0x33/0x1b5
       [<c0142b56>] shrink_slab+0xce/0x125
       [<c0143385>] try_to_free_pages+0x125/0x1cc
       [<c013f8a8>] __alloc_pages+0x184/0x26d
       [<c013ca52>]
generic_file_buffered_write+0x15a/0x53d
       [<c013d175>]
__generic_file_aio_write_nolock+0x340/0x38d
       [<c013d223>] generic_file_aio_write+0x61/0xb3
       [<e3d8a326>] cifs_file_aio_write+0x23/0x43
[cifs]
       [<c0153bb3>] do_sync_write+0x9d/0xce
       [<c0154437>] vfs_write+0xaa/0x14e
       [<c015496e>] sys_write+0x3a/0x61
       [<c0102cfb>] syscall_call+0x7/0xb

other info that might help us debug this:

2 locks held by cp/11790:
#0: (&inode->i_mutex){--..}, at: [<c029e364>]

mutex_lock+0x19/0x20
 #1:  (shrinker_rwsem){----}, at: [<c0142aa8>]
shrink_slab+0x20/0x125

stack backtrace:
 [<c01033da>] show_trace_log_lvl+0x57/0x157
 [<c0103a28>] show_trace+0x16/0x19
 [<c0103ac9>] dump_stack+0x1a/0x1f
 [<c0129551>] print_circular_bug_tail+0x59/0x64
 [<c0129d6e>] __lock_acquire+0x812/0x9a3
 [<c012a1c0>] lock_acquire+0x56/0x73
 [<c029e205>] __mutex_lock_slowpath+0xa6/0x1ec
 [<c029e364>] mutex_lock+0x19/0x20
 [<c016a3a2>] shrink_icache_memory+0x33/0x1b5
 [<c0142b56>] shrink_slab+0xce/0x125
 [<c0143385>] try_to_free_pages+0x125/0x1cc
 [<c013f8a8>] __alloc_pages+0x184/0x26d
 [<c013ca52>] generic_file_buffered_write+0x15a/0x53d
 [<c013d175>]
__generic_file_aio_write_nolock+0x340/0x38d
 [<c013d223>] generic_file_aio_write+0x61/0xb3
 [<e3d8a326>] cifs_file_aio_write+0x23/0x43 [cifs]
 [<c0153bb3>] do_sync_write+0x9d/0xce
 [<c0154437>] vfs_write+0xaa/0x14e
 [<c015496e>] sys_write+0x3a/0x61
 [<c0102cfb>] syscall_call+0x7/0xb
DWARF2 unwinder stuck at syscall_call+0x7/0xb
Leftover inexact backtrace:
 [<c0103a28>] show_trace+0x16/0x19
 [<c0103ac9>] dump_stack+0x1a/0x1f
 [<c0129551>] print_circular_bug_tail+0x59/0x64
 [<c0129d6e>] __lock_acquire+0x812/0x9a3
 [<c012a1c0>] lock_acquire+0x56/0x73
 [<c029e205>] __mutex_lock_slowpath+0xa6/0x1ec
 [<c029e364>] mutex_lock+0x19/0x20
 [<c016a3a2>] shrink_icache_memory+0x33/0x1b5
 [<c0142b56>] shrink_slab+0xce/0x125
 [<c0143385>] try_to_free_pages+0x125/0x1cc
 [<c013f8a8>] __alloc_pages+0x184/0x26d
 [<c013ca52>] generic_file_buffered_write+0x15a/0x53d
 [<c013d175>]
__generic_file_aio_write_nolock+0x340/0x38d
 [<c013d223>] generic_file_aio_write+0x61/0xb3
 [<e3d8a326>] cifs_file_aio_write+0x23/0x43 [cifs]
 [<c0153bb3>] do_sync_write+0x9d/0xce
 [<c0154437>] vfs_write+0xaa/0x14e
 [<c015496e>] sys_write+0x3a/0x61
 [<c0102cfb>] syscall_call+0x7/0xb

More info on request.

Thanks

____________________________________________________
On Yahoo!7
360°: Your own space to share what you want with who you want!
http://www.yahoo7.com.au/360

#1 lore raw
Quoted 22 lines · On Mon, 14 Aug 2006 11:05:03 +1000 (EST) Srihari Vijayaraghavan…

This is observed on 2.6.18-rc4 on SUSE 10.1 x86 on P-IV. The volume is question was mounted from a Windows 2003 server.

=======================================================
[ INFO: possible circular locking dependency detected ]
-------------------------------------------------------
cp/11790 is trying to acquire lock:
(iprune_mutex){--..}, at: [<c029e364>]
mutex_lock+0x19/0x20

but task is already holding lock:
(&inode->i_mutex){--..}, at: [<c029e364>]
mutex_lock+0x19/0x20

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&inode->i_mutex){--..}:

Quoted 15 lines
       [<c012a1c0>] lock_acquire+0x56/0x73
       [<c029e205>] __mutex_lock_slowpath+0xa6/0x1ec
       [<c029e364>] mutex_lock+0x19/0x20
       [<e3cd2b7a>] ntfs_put_inode+0x3e/0x79 [ntfs]
       [<c0169ac1>] iput+0x33/0x70
       [<c017738e>] inotify_unmount_inodes+0x12e/0x15f
       [<c016a55c>] invalidate_inodes+0x38/0xd1
       [<c01599b3>] generic_shutdown_super+0x5a/0x108
       [<c0159a81>] kill_block_super+0x20/0x36
       [<c0159b56>] deactivate_super+0x61/0x78
       [<c016c561>] mntput_no_expire+0x44/0x78
       [<c015f536>] path_release_on_umount+0x16/0x1d
       [<c016d692>] sys_umount+0x1d2/0x208
       [<c016d6d5>] sys_oldumount+0xd/0xf
       [<c0102cfb>] syscall_call+0x7/0xb

NTFS takes i_mutex inside iprune_mutex. NTFS _should_ be deadlocking because of this (iprune_mutex nests inside i_mutex on the write() path) but somehow it gets away with it.

Quoted 11 lines

-> #0 (iprune_mutex){--..}:

       [<c012a1c0>] lock_acquire+0x56/0x73
       [<c029e205>] __mutex_lock_slowpath+0xa6/0x1ec
       [<c029e364>] mutex_lock+0x19/0x20
       [<c016a3a2>] shrink_icache_memory+0x33/0x1b5
       [<c0142b56>] shrink_slab+0xce/0x125
       [<c0143385>] try_to_free_pages+0x125/0x1cc
       [<c013f8a8>] __alloc_pages+0x184/0x26d
       [<c013ca52>]
generic_file_buffered_write+0x15a/0x53d
       [<c013d175>]
Quoted 8 lines
__generic_file_aio_write_nolock+0x340/0x38d
       [<c013d223>] generic_file_aio_write+0x61/0xb3
       [<e3d8a326>] cifs_file_aio_write+0x23/0x43
[cifs]
       [<c0153bb3>] do_sync_write+0x9d/0xce
       [<c0154437>] vfs_write+0xaa/0x14e
       [<c015496e>] sys_write+0x3a/0x61
       [<c0102cfb>] syscall_call+0x7/0xb

CIFS takes i_prune_mutex inside i_mutex.

There's no deadlock here. Arguably lockdep should be treating i_mutex in filesystem A as being a different class from i_mutex from filesystem B.

#2 lore raw
Quoted 9 lines · --- Andrew Morton

On Mon, 14 Aug 2006 11:05:03 +1000 (EST) Srihari Vijayaraghavan <[email protected]> wrote:
[...]
NTFS takes i_mutex inside iprune_mutex. NTFS _should_ be deadlocking
because of this (iprune_mutex nests inside i_mutex on the write() path) but
somehow it gets away with it.

Yup. An NTFS volume (thro' USB Storage) was also mounted at that time. (And the copy operation may have been between NFTS & CIFS volumes)

Thanks

____________________________________________________
On Yahoo!7
Photos: Unlimited free storage – keep all your photos in one place!
http://au.photos.yahoo.com

#3 lore raw
In reply to Andrew Morton
Quoted 41 lines · Andrew Morton
Quoted 22 lines · On Mon, 14 Aug 2006 11:05:03 +1000 (EST) Srihari Vijayaraghavan…

This is observed on 2.6.18-rc4 on SUSE 10.1 x86 on P-IV. The volume is question was mounted from a Windows 2003 server.

=======================================================
[ INFO: possible circular locking dependency detected ]
-------------------------------------------------------
cp/11790 is trying to acquire lock:
(iprune_mutex){--..}, at: [<c029e364>]
mutex_lock+0x19/0x20

but task is already holding lock:
(&inode->i_mutex){--..}, at: [<c029e364>]
mutex_lock+0x19/0x20

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&inode->i_mutex){--..}:

Quoted 15 lines
       [<c012a1c0>] lock_acquire+0x56/0x73
       [<c029e205>] __mutex_lock_slowpath+0xa6/0x1ec
       [<c029e364>] mutex_lock+0x19/0x20
       [<e3cd2b7a>] ntfs_put_inode+0x3e/0x79 [ntfs]
       [<c0169ac1>] iput+0x33/0x70
       [<c017738e>] inotify_unmount_inodes+0x12e/0x15f
       [<c016a55c>] invalidate_inodes+0x38/0xd1
       [<c01599b3>] generic_shutdown_super+0x5a/0x108
       [<c0159a81>] kill_block_super+0x20/0x36
       [<c0159b56>] deactivate_super+0x61/0x78
       [<c016c561>] mntput_no_expire+0x44/0x78
       [<c015f536>] path_release_on_umount+0x16/0x1d
       [<c016d692>] sys_umount+0x1d2/0x208
       [<c016d6d5>] sys_oldumount+0xd/0xf
       [<c0102cfb>] syscall_call+0x7/0xb
Quoted 3 lines

NTFS takes i_mutex inside iprune_mutex. NTFS _should_ be deadlocking because of this (iprune_mutex nests inside i_mutex on the write() path) but somehow it gets away with it.

Directories are not subject to write(2) or at least not last time I checked. And ntfs_put_inode() only operates on directories... To quote current 2.6 kernel/fs/ntfs/inode.c:

void ntfs_put_inode(struct inode *vi)

{
        if (S_ISDIR(vi->i_mode) && atomic_read(&vi->i_count) == 2) {
                ntfs_inode *ni = NTFS_I(vi);
                if (NInoIndexAllocPresent(ni)) {
                        struct inode *bvi = NULL;
                        mutex_lock(&vi->i_mutex);
                        if (atomic_read(&vi->i_count) == 2) {
                                bvi = ni->itype.index.bmp_ino;
                                if (bvi)
                                        ni->itype.index.bmp_ino = NULL;
                        }
                        mutex_unlock(&vi->i_mutex);
                        if (bvi)
                                iput(bvi);
                }
        }
}

Best regards,

Anton

Signature · 4 lines
Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
Unix Support, Computing Service, University of Cambridge, CB2 3QH, UK
Linux NTFS maintainer / IRC: #ntfs on irc.freenode.net
WWW: http://www.linux-ntfs.org/ & http://www-stu.christs.cam.ac.uk/~aia21/