[SCSI] hpsa: fix warning with smp_processor_id() in preemptible
authorJohn Kacur <jkacur@redhat.com>
Fri, 26 Jul 2013 14:06:18 +0000 (16:06 +0200)
committerJames Bottomley <JBottomley@Parallels.com>
Mon, 26 Aug 2013 14:53:50 +0000 (18:53 +0400)
commit804a5cb526c121226830c686132b1b82aa12b76c
tree4fb48f69d1a2cea7e3e4fdc4eed050de8eee0cc7
parent6a06a4b8cff89b91cb04d3b44242658df1b3e851
[SCSI] hpsa: fix warning with smp_processor_id() in preemptible

 section Signed-off-by: John Kacur <jkacur@redhat.com>

On a 3.6-rt (real-time patch) kernel we are seeing the following BUG
However, it appears to be relevant for non-realtime (mainline) as well.

[   49.688847] hpsa 0000:03:00.0: hpsa0: <0x323a> at IRQ 67 using DAC
[   49.749928] scsi0 : hpsa
[   49.784437] BUG: using smp_processor_id() in preemptible [00000000
00000000] code: kworker/u:0/6
[   49.784465] caller is enqueue_cmd_and_start_io+0x5a/0x100 [hpsa]
[   49.784468] Pid: 6, comm: kworker/u:0 Not tainted
3.6.11.5-rt37.52.el6rt.x86_64.debug #1
[   49.784471] Call Trace:
[   49.784512]  [<ffffffff812abe83>] debug_smp_processor_id+0x123/0x150
[   49.784520]  [<ffffffffa009043a>] enqueue_cmd_and_start_io+0x5a/0x100
[hpsa]
[   49.784529]  [<ffffffffa00905cb>]
hpsa_scsi_do_simple_cmd_core+0xeb/0x110 [hpsa]
[   49.784537]  [<ffffffff812b09c8>] ? swiotlb_dma_mapping_error+0x18/0x30
[   49.784544]  [<ffffffff812b09c8>] ? swiotlb_dma_mapping_error+0x18/0x30
[   49.784553]  [<ffffffffa0090701>]
hpsa_scsi_do_simple_cmd_with_retry+0x91/0x280 [hpsa]
[   49.784562]  [<ffffffffa0093558>]
hpsa_scsi_do_report_luns.clone.2+0xd8/0x130 [hpsa]
[   49.784571]  [<ffffffffa00935ea>]
hpsa_gather_lun_info.clone.3+0x3a/0x1a0 [hpsa]
[   49.784580]  [<ffffffffa00963df>] hpsa_update_scsi_devices+0x11f/0x4f0
[hpsa]
[   49.784592]  [<ffffffff81592019>] ? sub_preempt_count+0xa9/0xe0
[   49.784601]  [<ffffffffa00968ad>] hpsa_scan_start+0xfd/0x150 [hpsa]
[   49.784613]  [<ffffffff8158cba8>] ? rt_spin_lock_slowunlock+0x78/0x90
[   49.784626]  [<ffffffff813b04d7>] do_scsi_scan_host+0x37/0xa0
[   49.784632]  [<ffffffff813b05da>] do_scan_async+0x1a/0x30
[   49.784643]  [<ffffffff8107c4ab>] async_run_entry_fn+0x9b/0x1d0
[   49.784655]  [<ffffffff8106ae92>] process_one_work+0x1f2/0x620
[   49.784661]  [<ffffffff8106ae20>] ? process_one_work+0x180/0x620
[   49.784668]  [<ffffffff8106d4fe>] ? worker_thread+0x5e/0x3a0
[   49.784674]  [<ffffffff8107c410>] ? async_schedule+0x20/0x20
[   49.784681]  [<ffffffff8106d5d3>] worker_thread+0x133/0x3a0
[   49.784688]  [<ffffffff8106d4a0>] ? manage_workers+0x190/0x190
[   49.784696]  [<ffffffff81073236>] kthread+0xa6/0xb0
[   49.784707]  [<ffffffff815970a4>] kernel_thread_helper+0x4/0x10
[   49.784715]  [<ffffffff81082a7c>] ? finish_task_switch+0x8c/0x110
[   49.784721]  [<ffffffff8158e44b>] ? _raw_spin_unlock_irq+0x3b/0x70
[   49.784727]  [<ffffffff8158e85d>] ? retint_restore_args+0xe/0xe
[   49.784734]  [<ffffffff81073190>] ? kthreadd+0x1e0/0x1e0
[   49.784739]  [<ffffffff815970a0>] ? gs_change+0xb/0xb

-------------------

This is caused by
enqueue_cmd_and_start_io()->
        set_performant_mode()->
                smp_processor_id()
Which if you have debugging enabled calls debug_processor_id() and triggers the warning.

The code here is
 c->Header.ReplyQueue = smp_processor_id() % h->nreply_queues;

Since it is not critical that the code complete on the same processor,
but the cpu is a hint used in generating the ReplyQueue and will still work if
the cpu migrates or is preempted, it is safe to use the raw_smp_processor_id()
to surpress the false positve warning.

Signed-off-by: John Kacur <jkacur@redhat.com>
Acked-by: Stephen Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
drivers/scsi/hpsa.c