Commit 53a983c4 authored by James Bottomley's avatar James Bottomley

[SCSI] mvsas: Fix kernel panic on tile due to unaligned data access

slot->response is a 64 bit quantity (and accessed as such), but its alignment
is only 32 bits.  This doesn't cause a problem on x86, but apparently causes a
kernel panic on Tile:

Stack dump complete Kernel panic - not syncing:
 Kernel unalign fault running the idle task!
 Starting stack dump of tid 0, pid 0 (swapper) on cpu 1 at cycle 341586172541
   frame 0: 0xfffffff700140ee0 dump_stack+0x0/0x20 (sp 0xfffffe43ffedf420)
   frame 1: 0xfffffff700283270 panic+0x150/0x3a0 (sp 0xfffffe43ffedf420)
   frame 2: 0xfffffff70012bff8 jit_bundle_gen+0xfd8/0x27e0 (sp 0xfffffe43ffedf4c8)
   frame 3: 0xfffffff7003b5b68 do_unaligned+0xc0/0x5a0 (sp 0xfffffe43ffedf710)
   frame 4: 0xfffffff70044ca78 handle_interrupt+0x270/0x278 (sp 0xfffffe43ffedf840)
   <interrupt 17 while in kernel mode>
   frame 5: 0xfffffff7002ac370 mvs_slot_complete+0x5f0/0x12a0 (sp 0xfffffe43ffedfa90)
   frame 6: 0xfffffff7002abec0 mvs_slot_complete+0x140/0x12a0 (sp 0xfffffe43ffedfa90)
   frame 7: 0xfffffff7005cc840 mvs_int_rx+0x140/0x2a0 (sp 0xfffffe43ffedfb00)
   frame 8: 0xfffffff7005bbaf0 mvs_94xx_isr+0xd8/0x2b8 (sp 0xfffffe43ffedfb68)
   frame 9: 0xfffffff700658ba0 mvs_tasklet+0x128/0x1f8 (sp 0xfffffe43ffedfba8)
   frame 10: 0xfffffff7003e8230 tasklet_action+0x178/0x2c8 (sp 0xfffffe43ffedfbe0)
   frame 11: 0xfffffff700103850 __do_softirq+0x210/0x398 (sp 0xfffffe43ffedfc40)
   frame 12: 0xfffffff700180308 do_softirq+0xc8/0x140 (sp 0xfffffe43ffedfcd8)
   frame 13: 0xfffffff7000bd7f0 irq_exit+0xb0/0x158 (sp 0xfffffe43ffedfcf0)
   frame 14: 0xfffffff70013fa58 tile_dev_intr+0x1d8/0x2f0 (sp 0xfffffe43ffedfd00)
   frame 15: 0xfffffff70044ca78 handle_interrupt+0x270/0x278 (sp 0xfffffe43ffedfd40)
   <interrupt 30 while in kernel mode>
   frame 16: 0xfffffff700143e68 _cpu_idle_nap+0x0/0x18 (sp 0xfffffe43ffedffb0)
   frame 17: 0xfffffff700482480 cpu_idle+0x310/0x428 (sp 0xfffffe43ffedffb0)

Since the check is just for non-zero, split it to be two 32 bit accesses
(preserving speed in the fast path) and do a get_unaligned() in the slow path.

This is a modification of a wholly get_unaligned patch submitted by Paul Guo
Reported-by: default avatarPaul Guo <ggang@tilera.com>
Signed-off-by: default avatarJames Bottomley <JBottomley@Parallels.com>
parent 085b513f
......@@ -1857,11 +1857,16 @@ int mvs_slot_complete(struct mvs_info *mvi, u32 rx_desc, u32 flags)
goto out;
}
/* error info record present */
if (unlikely((rx_desc & RXQ_ERR) && (*(u64 *) slot->response))) {
/*
* error info record present; slot->response is 32 bit aligned but may
* not be 64 bit aligned, so check for zero in two 32 bit reads
*/
if (unlikely((rx_desc & RXQ_ERR)
&& (*((u32 *)slot->response)
|| *(((u32 *)slot->response) + 1)))) {
mv_dprintk("port %d slot %d rx_desc %X has error info"
"%016llX.\n", slot->port->sas_port.id, slot_idx,
rx_desc, (u64)(*(u64 *)slot->response));
rx_desc, get_unaligned_le64(slot->response));
tstat->stat = mvs_slot_err(mvi, task, slot_idx);
tstat->resp = SAS_TASK_COMPLETE;
goto out;
......
......@@ -39,6 +39,7 @@
#include <linux/irq.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>
#include <asm/unaligned.h>
#include <scsi/libsas.h>
#include <scsi/scsi.h>
#include <scsi/scsi_tcq.h>
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment