Interesting lockdep message coming out of blktap

[Jeremy Fitzhardinge <jeremy@goop.org>]

I'm getting this:

blktap_validate_params: aio:/dev/vg_lilith-raid/xen-f13-64: capacity: 20971520, sector-size: 512
blktap_validate_params: aio:/dev/vg_lilith-raid/xen-f13-64: capacity: 20971520, sector-size: 512
blktap_device_create: minor 0 sectors 20971520 sector-size 512
blktap_device_create: creation of 253:0: 0
INFO: trying to register non-static key.
the code is fine but needs lockdep annotation.
turning off the locking correctness validator.
Pid: 4042, comm: blkid Not tainted 2.6.32 #75
Call Trace:
  [<ffffffff8107711b>] __lock_acquire+0x16d0/0x1767
  [<ffffffff8100f465>] ? xen_force_evtchn_callback+0xd/0xf
  [<ffffffff8100fd52>] ? check_events+0x12/0x20
  [<ffffffff810da9af>] ? apply_to_page_range+0x2ba/0x3c8
  [<ffffffff810772a4>] lock_acquire+0xf2/0x116
  [<ffffffff810da9af>] ? apply_to_page_range+0x2ba/0x3c8
  [<ffffffff810d02af>] ? ftrace_format_kmalloc+0x63/0xdd
  [<ffffffff814e9d47>] _spin_lock+0x36/0x45
  [<ffffffff810da9af>] ? apply_to_page_range+0x2ba/0x3c8
  [<ffffffff810da9af>] apply_to_page_range+0x2ba/0x3c8
  [<ffffffff81288d7c>] ? blktap_map_uaddr_fn+0x0/0x50
  [<ffffffff81289963>] blktap_device_process_request+0x457/0x989
  [<ffffffff810c5305>] ? get_page_from_freelist+0x49b/0x804
  [<ffffffff8100fd3f>] ? xen_restore_fl_direct_end+0x0/0x1
  [<ffffffff8107f323>] ? __module_text_address+0xd/0x53
  [<ffffffff81074d9d>] ? trace_hardirqs_on_caller+0x111/0x135
  [<ffffffff814e9b38>] ? _spin_unlock_irq+0x3c/0x5a
  [<ffffffff814e9540>] ? __down_read+0x38/0xad
  [<ffffffff812802d0>] ? evtchn_interrupt+0xaa/0x112
  [<ffffffff8128a0de>] blktap_device_do_request+0x1dc/0x298
  [<ffffffff814e9bac>] ? _spin_unlock_irqrestore+0x56/0x74
  [<ffffffff8105848b>] ? del_timer+0xd7/0xe5
  [<ffffffff810bf104>] ? sync_page_killable+0x0/0x30
  [<ffffffff81202143>] __generic_unplug_device+0x30/0x35
  [<ffffffff81202171>] generic_unplug_device+0x29/0x3a
  [<ffffffff811fb5dc>] blk_unplug+0x71/0x76
  [<ffffffff811fb5ee>] blk_backing_dev_unplug+0xd/0xf
  [<ffffffff8111a1ad>] block_sync_page+0x42/0x44
  [<ffffffff810bf0fb>] sync_page+0x3f/0x48
  [<ffffffff810bf10d>] sync_page_killable+0x9/0x30
  [<ffffffff814e7a2f>] __wait_on_bit_lock+0x41/0x8a
  [<ffffffff810bf040>] __lock_page_killable+0x61/0x68
  [<ffffffff8106486b>] ? wake_bit_function+0x0/0x2e
  [<ffffffff8103e0af>] ? __might_sleep+0x3d/0x127
  [<ffffffff810c0b1f>] generic_file_aio_read+0x3db/0x594
  [<ffffffff810763f0>] ? __lock_acquire+0x9a5/0x1767
  [<ffffffff8100fd52>] ? check_events+0x12/0x20
  [<ffffffff810f8e16>] do_sync_read+0xe3/0x120
  [<ffffffff81064837>] ? autoremove_wake_function+0x0/0x34
  [<ffffffff811d8da4>] ? selinux_file_permission+0x5d/0x10f
  [<ffffffff811d0d7c>] ? security_file_permission+0x11/0x13
  [<ffffffff810f997a>] vfs_read+0xaa/0x16f
  [<ffffffff81074d9d>] ? trace_hardirqs_on_caller+0x111/0x135
  [<ffffffff810f9af8>] sys_read+0x45/0x6c
  [<ffffffff81013b82>] system_call_fastpath+0x16/0x1b


The lock in question appears to be the pte spinlock, taken in 
apply_to_page_range() at:

0xffffffff810da9af is in apply_to_page_range (/home/jeremy/git/linux/mm/memory.c:1855).
1850		spinlock_t *uninitialized_var(ptl);
1851	
1852		pte = (mm ==&init_mm) ?
1853			pte_alloc_kernel(pmd, addr) :
1854			pte_alloc_map_lock(mm, pmd, addr,&ptl);
1855		if (!pte)
1856			return -ENOMEM;
1857	
1858		BUG_ON(pmd_huge(*pmd));
1859	

(I'm pretty sure its really 1854, the usermode mm case.)

I have split PTE locks enabled, so this is a per-page pte lock rather 
than the global mm one.  It seems highly unlikely this is not being 
initialized properly in general, or every pte lock would end up 
triggering this message.

I wonder if something else is going wrong here?  I'm not really sure 
what the blktap code is trying to do here.

Any thoughts?

Thanks,
     J