On Thu, 2008-05-08 at 11:39 -0700, Rafael J. Wysocki wrote:

I get a similar warning with RAID1 on one of my test boxes:

WARNING: at /home/rafael/src/linux-2.6/include/linux/blkdev.h:443 blk_remove_plug+0x85/0xa0()
Modules linked in: raid456 async_xor async_memcpy async_tx xor raid0 ehci_hcd ohci_hcd sd_mod edd raid1 ext3 jbd fan sata_uli pata_ali thermal processor
Pid: 2159, comm: md1_raid1 Not tainted 2.6.26-rc1 #158

Call Trace:
 [<ffffffff80238bbf>] warn_on_slowpath+0x5f/0x80
 [<ffffffff8025e8d8>] ? __lock_acquire+0x748/0x10d0
 [<ffffffff80348f55>] blk_remove_plug+0x85/0xa0
 [<ffffffffa004df64>] :raid1:flush_pending_writes+0x44/0xb0
 [<ffffffffa004e649>] :raid1:raid1d+0x59/0xfe0
 [<ffffffff8025e8d8>] ? __lock_acquire+0x748/0x10d0
 [<ffffffff8025dc4f>] ? trace_hardirqs_on+0xbf/0x150
 [<ffffffff8043ea8c>] md_thread+0x3c/0x110
 [<ffffffff8024f6a0>] ? autoremove_wake_function+0x0/0x40
 [<ffffffff8043ea50>] ? md_thread+0x0/0x110
 [<ffffffff8024f23d>] kthread+0x4d/0x80
 [<ffffffff8020c548>] child_rip+0xa/0x12
 [<ffffffff8020bc5f>] ? restore_args+0x0/0x30
 [<ffffffff8024f1f0>] ? kthread+0x0/0x80
 [<ffffffff8020c53e>] ? child_rip+0x0/0x12

---[ end trace 05d4e0844c61f45d ]---

This is the WARN_ON_ONCE(!queue_is_locked(q)) in queue_flag_clear(),
apparently.

Yes, it triggers on all RAID levels.  The patch in this message:

http://marc.info/?l=linux-raid&m=121001065404056&w=2

...fixes the raid 0/1/10/5/6 cases, but I am still trying to isolate an
issue (potentially unrelated) with linear arrays.

--
Dan