[patch] cciss: unregister from SCSI before tearing down device resources

[Bjorn Helgaas <bjorn.helgaas@hp.com>]

We must unregister from SCSI before we unmap device resources and
unhook the IRQ handler.  Otherwise, SCSI may send us more requests,
and we won't be able to handle them.

I'd like to see this patch in 2.6.21.  In 2.6.21-rc6, I see the
following oops during every reboot of my HP DL360:

    ...
    Unmounting local filesystems...done.
    Rebooting... Completed flushing cache on controller 0
    BUG: unable to handle kernel paging request at virtual address f8808040
     printing eip:
    c02dc72b
    *pde = 02120067
    *pte = 00000000
    Oops: 0002 [#1]
    SMP 
    Modules linked in:
    CPU:    1
    EIP:    0060:[<c02dc72b>]    Not tainted VLI
    EFLAGS: 00010046   (2.6.21-rc6 #1)
    EIP is at SA5_submit_command+0xb/0x20
    eax: f8808000   ebx: f7a00000   ecx: f79f0000   edx: 37a00000
    esi: f79f0000   edi: 00000000   ebp: 00000000   esp: dd717a44
    ds: 007b   es: 007b   fs: 00d8  gs: 0000  ss: 0068
    Process khelper (pid: 1427, ti=dd716000 task=c2260a70 task.ti=dd716000)
    Stack: c02df2c0 f7a00000 f7a00000 00d41008 c02df691 00000000 00000010 00000002 
	   00000001 f79f0000 f7fff844 c1398420 00000000 00000000 00001000 230a3020 
	   69666564 5420656e 50434f49 465f544b 4853554c 44414552 0a312009 66656423 
    Call Trace:
     [<c02df2c0>] start_io+0x80/0x120
     [<c02df691>] do_cciss_request+0x331/0x350
     [<c014242a>] mempool_alloc+0x2a/0xe0
     [<c020ad71>] blk_alloc_request+0x61/0x80
     [<c020b02e>] get_request+0x15e/0x1e0
     [<c01595e0>] cache_alloc_refill+0xb0/0x1e0
     [<c021049d>] as_update_rq+0x2d/0x80
     [<c0210d28>] as_add_request+0x68/0x90
     [<c0207f99>] elv_insert+0x119/0x160
     [<c020bd0b>] __make_request+0xcb/0x320
     [<c0122ee0>] lock_timer_base+0x20/0x50
     [<c0123096>] del_timer+0x56/0x60
     [<c020a7b8>] blk_remove_plug+0x38/0x70
     [<c020a815>] __generic_unplug_device+0x25/0x30
     [<c020a835>] generic_unplug_device+0x15/0x30
    ...

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>

Index: cciss/drivers/block/cciss.c
===================================================================
--- cciss.orig/drivers/block/cciss.c	2007-04-10 09:46:09.000000000 -0600
+++ cciss/drivers/block/cciss.c	2007-04-10 10:11:06.000000000 -0600
@@ -3423,6 +3423,25 @@
 		       "already be removed \n");
 		return;
 	}
+
+	remove_proc_entry(hba[i]->devname, proc_cciss);
+	unregister_blkdev(hba[i]->major, hba[i]->devname);
+
+	/* remove it from the disk list */
+	for (j = 0; j < CISS_MAX_LUN; j++) {
+		struct gendisk *disk = hba[i]->gendisk[j];
+		if (disk) {
+			request_queue_t *q = disk->queue;
+
+			if (disk->flags & GENHD_FL_UP)
+				del_gendisk(disk);
+			if (q)
+				blk_cleanup_queue(q);
+		}
+	}
+
+	cciss_unregister_scsi(i);	/* unhook from SCSI subsystem */
+
 	/* Turn board interrupts off  and send the flush cache command */
 	/* sendcmd will turn off interrupt, and send the flush...
 	 * To write all data in the battery backed cache to disks */
@@ -3444,22 +3463,6 @@
 #endif				/* CONFIG_PCI_MSI */
 
 	iounmap(hba[i]->vaddr);
-	cciss_unregister_scsi(i);	/* unhook from SCSI subsystem */
-	unregister_blkdev(hba[i]->major, hba[i]->devname);
-	remove_proc_entry(hba[i]->devname, proc_cciss);
-
-	/* remove it from the disk list */
-	for (j = 0; j < CISS_MAX_LUN; j++) {
-		struct gendisk *disk = hba[i]->gendisk[j];
-		if (disk) {
-			request_queue_t *q = disk->queue;
-
-			if (disk->flags & GENHD_FL_UP)
-				del_gendisk(disk);
-			if (q)
-				blk_cleanup_queue(q);
-		}
-	}
 
 	pci_free_consistent(hba[i]->pdev, hba[i]->nr_cmds * sizeof(CommandList_struct),
 			    hba[i]->cmd_pool, hba[i]->cmd_pool_dhandle);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/