umem: fix request_queue lock warning
authorSage Weil <sage@newdream.ne>
Thu, 23 Apr 2009 06:37:58 +0000 (08:37 +0200)
committerJens Axboe <jens.axboe@oracle.com>
Fri, 24 Apr 2009 06:54:21 +0000 (08:54 +0200)
The umem driver issues two warnings on boot, due to blk_plug_device() and
blk_remove_plug() being called without q->queue_lock held.  Starting with
e48ec690 (block: extend queue_flag bitops), the queue_flag_* functions
warn if q->queue_lock doesn't appear to be locked.  In fact, q->queue_lock
is NULL (though that apparently isn't otherwise a problem as the driver is
using card->lock for everything).

Although blk_init_queue() with take a request_fn_proc and spinlock_t*,
there isn't a corresponding init helper that takes a make_request_fn.
Setting queue_lock to &card->lock explicitly seems to work fine for me.
The warning goes away and the device appears to behave.

[    1.531881] v2.3 : Micro Memory(tm) PCI memory board block driver
[    1.538136] umem 0000:02:01.0: PCI INT A -> GSI 20 (level, low) -> IRQ 20
[    1.545018] umem 0000:02:01.0: Micro Memory(tm) controller found (PCI Mem Module (Battery Backup))
[    1.554176] umem 0000:02:01.0: CSR 0xfc9ffc00 -> 0xffffc200013d0c00 (0x100)
[    1.561279] umem 0000:02:01.0: Size 1048576 KB, Battery 1 Disabled (FAILURE), Battery 2 Disabled (FAILURE)
[    1.571114] umem 0000:02:01.0: Window size 16777216 bytes, IRQ 20
[    1.577304] umem 0000:02:01.0: memory NOT initialized. Consider over-writing whole device.
[    1.585989]  umema:<4>------------[ cut here ]------------
[    1.591775] WARNING: at include/linux/blkdev.h:492 blk_plug_device+0x6d/0x106()
[    1.592025] Hardware name: H8SSL
[    1.592025] Modules linked in:
[    1.592025] Pid: 1, comm: swapper Not tainted 2.6.29 #8
[    1.592025] Call Trace:
[    1.592025]  [<ffffffff8023c994>] warn_slowpath+0xd3/0xf2
[    1.592025]  [<ffffffff8025a5b5>] ? save_trace+0x3f/0x9b
[    1.592025]  [<ffffffff8025a68b>] ? add_lock_to_list+0x7a/0xba
[    1.592025]  [<ffffffff8025e609>] ? validate_chain+0xb3b/0xce8
[    1.592025]  [<ffffffff80441556>] ? mm_make_request+0x27/0x59
[    1.592025]  [<ffffffff80441556>] ? mm_make_request+0x27/0x59
[    1.592025]  [<ffffffff8025ef04>] ? __lock_acquire+0x74e/0x7b9
[    1.592025]  [<ffffffff8025a70e>] ? get_lock_stats+0x34/0x5e
[    1.592025]  [<ffffffff8025a746>] ? put_lock_stats+0xe/0x27
[    1.592025]  [<ffffffff80441556>] ? mm_make_request+0x27/0x59
[    1.592025]  [<ffffffff803ad165>] blk_plug_device+0x6d/0x106
[    1.592025]  [<ffffffff80441575>] mm_make_request+0x46/0x59
[    1.592025]  [<ffffffff803ac2d9>] generic_make_request+0x335/0x3cf
[    1.592025]  [<ffffffff8027fcc7>] ? mempool_alloc_slab+0x11/0x13
[    1.592025]  [<ffffffff8027fdce>] ? mempool_alloc+0x45/0x101
[    1.592025]  [<ffffffff8025a746>] ? put_lock_stats+0xe/0x27
[    1.592025]  [<ffffffff803adda5>] submit_bio+0x10a/0x119
[    1.592025]  [<ffffffff802c8d00>] submit_bh+0xe5/0x109
[    1.592025]  [<ffffffff802cbf43>] block_read_full_page+0x2aa/0x2cb
[    1.592025]  [<ffffffff802cf4c4>] ? blkdev_get_block+0x0/0x4c
[    1.592025]  [<ffffffff805c90a8>] ? _spin_unlock_irq+0x36/0x51
[    1.592025]  [<ffffffff80286836>] ? __lru_cache_add+0x92/0xb2
[    1.592025]  [<ffffffff802cf008>] blkdev_readpage+0x13/0x15
[    1.592025]  [<ffffffff8027de06>] read_cache_page_async+0x90/0x134
[    1.592025]  [<ffffffff802ceff5>] ? blkdev_readpage+0x0/0x15
[    1.592025]  [<ffffffff802f5f1c>] ? adfspart_check_ICS+0x0/0x16c
[    1.592025]  [<ffffffff8027deb8>] read_cache_page+0xe/0x45
[    1.592025]  [<ffffffff802f5170>] read_dev_sector+0x2e/0x93
[    1.592025]  [<ffffffff802f5f44>] adfspart_check_ICS+0x28/0x16c
[    1.592025]  [<ffffffff8025d427>] ? trace_hardirqs_on+0xd/0xf
[    1.592025]  [<ffffffff802f5f1c>] ? adfspart_check_ICS+0x0/0x16c
[    1.592025]  [<ffffffff802f59c5>] rescan_partitions+0x168/0x2fb
[    1.592025]  [<ffffffff802ceae9>] __blkdev_get+0x259/0x336
[    1.592025]  [<ffffffff803ca1e2>] ? kobject_put+0x47/0x4b
[    1.592025]  [<ffffffff802cebd1>] blkdev_get+0xb/0xd
[    1.592025]  [<ffffffff802f5773>] register_disk+0xc4/0x12b
[    1.592025]  [<ffffffff803b2a7b>] add_disk+0xc3/0x12d
[    1.592025]  [<ffffffff808a1d4a>] ? mm_init+0x0/0x1a5
[    1.592025]  [<ffffffff808a1e73>] mm_init+0x129/0x1a5
[    1.592025]  [<ffffffff808a1d4a>] ? mm_init+0x0/0x1a5
[    1.592025]  [<ffffffff80209056>] _stext+0x56/0x130
[    1.592025]  [<ffffffff80274932>] ? register_irq_proc+0xae/0xca
[    1.592025]  [<ffffffff802f0000>] ? proc_pid_lookup+0xb4/0x18b
[    1.592025]  [<ffffffff8087f975>] kernel_init+0x132/0x18b
[    1.592025]  [<ffffffff8020d17a>] child_rip+0xa/0x20
[    1.592025]  [<ffffffff8020cb40>] ? restore_args+0x0/0x30
[    1.592025]  [<ffffffff8087f843>] ? kernel_init+0x0/0x18b
[    1.592025]  [<ffffffff8020d170>] ? child_rip+0x0/0x20
[    1.592025] ---[ end trace 7150b3b86da74e1e ]---
[    1.889858] ------------[ cut here ]------------[ve_plug+0x5f/0x91()
[    1.893848] Hardware name: H8SSL
[    1.893848] Modules linked in:
[    1.893848] Pid: 1, comm: swapper Tainted: G        W  2.6.29 #8
[    1.893848] Call Trace:
[    1.893848]  [<ffffffff8023c994>] warn_slowpath+0xd3/0xf2
[    1.893848]  [<ffffffff805c8411>] ? trace_hardirqs_on_thunk+0x3a/0x3f
[    1.893848]  [<ffffffff8020cb40>] ? restore_args+0x0/0x30
[    1.893848]  [<ffffffff80254245>] ? __atomic_notifier_call_chain+0x0/0xb2
[    1.893848]  [<ffffffff805c90a3>] ? _spin_unlock_irq+0x31/0x51
[    1.893848]  [<ffffffff805c90bf>] ? _spin_unlock_irq+0x4d/0x51
[    1.893848]  [<ffffffff8044157d>] ? mm_make_request+0x4e/0x59
[    1.893848]  [<ffffffff8025a70e>] ? get_lock_stats+0x34/0x5e
[    1.893848]  [<ffffffff8025a75d>] ? put_lock_stats+0x25/0x27
[    1.893848]  [<ffffffff80441504>] ? mm_unplug_device+0x25/0x50
[    1.893848]  [<ffffffff803acf23>] blk_remove_plug+0x5f/0x91
[    1.893848]  [<ffffffff8044150f>] mm_unplug_device+0x30/0x50
[    1.893848]  [<ffffffff803ab74a>] blk_unplug+0x78/0x7d
[    1.893848]  [<ffffffff803ab75c>] blk_backing_dev_unplug+0xd/0xf
[    1.893848]  [<ffffffff802c853c>] block_sync_page+0x4a/0x4c
[    1.893848]  [<ffffffff8027da1c>] sync_page+0x44/0x4d
[    1.893848]  [<ffffffff805c66fd>] __wait_on_bit_lock+0x42/0x8a
[    1.893848]  [<ffffffff8027d9d8>] ? sync_page+0x0/0x4d
[    1.893848]  [<ffffffff8027d9c4>] __lock_page+0x64/0x6b
[    1.893848]  [<ffffffff802508db>] ? wake_bit_function+0x0/0x2a
[    1.893848]  [<ffffffff8027de4a>] read_cache_page_async+0xd4/0x134
[    1.893848]  [<ffffffff802ceff5>] ? blkdev_readpage+0x0/0x15
[    1.893848]  [<ffffffff802f5f1c>] ? adfspart_check_ICS+0x0/0x16c
[    1.893848]  [<ffffffff8027deb8>] read_cache_page+0xe/0x45
[    1.893848]  [<ffffffff802f5170>] read_dev_sector+0x2e/0x93
[    1.893848]  [<ffffffff802f5f44>] adfspart_check_ICS+0x28/0x16c
[    1.893848]  [<ffffffff8025d427>] ? trace_hardirqs_on+0xd/0xf
[    1.893848]  [<ffffffff802f5f1c>] ? adfspart_check_ICS+0x0/0x16c
[    1.893848]  [<ffffffff802f59c5>] rescan_partitions+0x168/0x2fb
[    1.893848]  [<ffffffff802ceae9>] __blkdev_get+0x259/0x336
[    1.893848]  [<ffffffff803ca1e2>] ? kobject_put+0x47/0x4b
[    1.893848]  [<ffffffff802cebd1>] blkdev_get+0xb/0xd
[    1.893848]  [<ffffffff802f5773>] register_disk+0xc4/0x12b
[    1.893848]  [<ffffffff803b2a7b>] add_disk+0xc3/0x12d
[    1.893848]  [<ffffffff808a1d4a>] ? mm_init+0x0/0x1a5
[    1.893848]  [<ffffffff808a1e73>] mm_init+0x129/0x1a5
[    1.893848]  [<ffffffff808a1d4a>] ? mm_init+0x0/0x1a5
[    1.893848]  [<ffffffff80209056>] _stext+0x56/0x130
[    1.893848]  [<ffffffff80274932>] ? register_irq_proc+0xae/0xca
[    1.893848]  [<ffffffff802f0000>] ? proc_pid_lookup+0xb4/0x18b
[    1.893848]  [<ffffffff8087f975>] kernel_init+0x132/0x18b
[    1.893848]  [<ffffffff8020d17a>] child_rip+0xa/0x20
[    1.893848]  [<ffffffff8020cb40>] ? restore_args+0x0/0x30
[    1.893848]  [<ffffffff8087f843>] ? kernel_init+0x0/0x18b
[    1.893848]  [<ffffffff8020d170>] ? child_rip+0x0/0x20
[    1.893848] ---[ end trace 7150b3b86da74e1f ]---

Signed-off-by: Sage Weil <sage@newdream.net>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
drivers/block/umem.c

index 9744d59..858c34d 100644 (file)
@@ -906,6 +906,7 @@ static int __devinit mm_pci_probe(struct pci_dev *dev,
                goto failed_alloc;
 
        blk_queue_make_request(card->queue, mm_make_request);
+       card->queue->queue_lock = &card->lock;
        card->queue->queuedata = card;
        card->queue->unplug_fn = mm_unplug_device;