]> err.no Git - linux-2.6/commitdiff
[SCSI] limit recursion when flushing shost->starved_list
authorAndreas Herrmann <aherrman@de.ibm.com>
Wed, 9 Aug 2006 15:31:16 +0000 (17:31 +0200)
committerJames Bottomley <jejb@mulgrave.il.steeleye.com>
Sat, 19 Aug 2006 20:31:53 +0000 (13:31 -0700)
Attached is a patch that should limit a possible recursion that can
lead to a stack overflow like follows:

Kernel stack overflow.
CPU:    3    Not tainted
Process zfcperp0.0.d819
(pid: 13897, task: 000000003e0d8cc8, ksp: 000000003499dbb8)
Krnl PSW : 0404000180000000 000000000030f8b2 (get_device+0x12/0x48)
Krnl GPRS: 00000000135a1980 000000000030f758 000000003ed6c1e8 0000000000000005
           0000000000000000 000000000044a780 000000003dbf7000 0000000034e15800
           000000003621c048 070000003499c108 000000003499c1a0 000000003ed6c000
           0000000040895000 00000000408ab630 000000003499c0a0 000000003499c0a0
Krnl Code: a7 fb ff e8 a7 19 00 00 b9 02 00 22 e3 e0 f0 98 00 24 a7 84
Call Trace:
([<000000004089edc2>] scsi_request_fn+0x13e/0x650 [scsi_mod])
 [<00000000002c5ff4>] blk_run_queue+0xd4/0x1a4
 [<000000004089ff8c>] scsi_queue_insert+0x22c/0x2a4 [scsi_mod]
 [<000000004089779a>] scsi_dispatch_cmd+0x8a/0x3d0 [scsi_mod]
 [<000000004089f1ec>] scsi_request_fn+0x568/0x650 [scsi_mod]
...
 [<000000004089f1ec>] scsi_request_fn+0x568/0x650 [scsi_mod]
 [<00000000002c5ff4>] blk_run_queue+0xd4/0x1a4
 [<000000004089ff8c>] scsi_queue_insert+0x22c/0x2a4 [scsi_mod]
 [<000000004089779a>] scsi_dispatch_cmd+0x8a/0x3d0 [scsi_mod]
 [<000000004089f1ec>] scsi_request_fn+0x568/0x650 [scsi_mod]
 [<00000000002c5ff4>] blk_run_queue+0xd4/0x1a4
 [<000000004089fa9e>] scsi_run_host_queues+0x196/0x230 [scsi_mod]
 [<00000000409eba28>] zfcp_erp_thread+0x2638/0x3080 [zfcp]
 [<0000000000107462>] kernel_thread_starter+0x6/0xc
 [<000000000010745c>] kernel_thread_starter+0x0/0xc
<0>Kernel panic - not syncing: Corrupt kernel stack, can't continue.

This stack overflow occurred during tests on s390 using zfcp.
Recursion depth for this panic was 19.

Usually recursion between blk_run_queue and a request_fn is avoided
using QUEUE_FLAG_REENTER. But this does not help if the scsi stack
tries to flush the starved_list of a scsi_host.

Limit recursion depth when flushing the starved_list
of a scsi_host.

Signed-off-by: Andreas Herrmann <aherrman@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
drivers/scsi/scsi_lib.c

index 077c1c691210284428d416fb73392c89f1cdc92b..d6743b959a72b151f1d6ade6d70b6d3b70c44d92 100644 (file)
@@ -551,7 +551,15 @@ static void scsi_run_queue(struct request_queue *q)
                list_del_init(&sdev->starved_entry);
                spin_unlock_irqrestore(shost->host_lock, flags);
 
-               blk_run_queue(sdev->request_queue);
+
+               if (test_bit(QUEUE_FLAG_REENTER, &q->queue_flags) &&
+                   !test_and_set_bit(QUEUE_FLAG_REENTER,
+                                     &sdev->request_queue->queue_flags)) {
+                       blk_run_queue(sdev->request_queue);
+                       clear_bit(QUEUE_FLAG_REENTER,
+                                 &sdev->request_queue->queue_flags);
+               } else
+                       blk_run_queue(sdev->request_queue);
 
                spin_lock_irqsave(shost->host_lock, flags);
                if (unlikely(!list_empty(&sdev->starved_entry)))