On Sat, May 18, 2013 at 09:05:15AM +0200, Jens Axboe wrote:
> On Fri, May 17 2013, Calvin Owens wrote:
> > Commit 2f6db2a7 was part of a series that cleaned up mdraid code by
> > replacing explicit re-initialization of struct bio with bio_reset().
> >
> > In raid5 it incorrectly assumed that a couple initializations of its
> > members was a full reset, erasing the existing data and unconditionally
> > triggering the following BUG when assembling arrays:
> >
> > [ 14.653072] kernel BUG at /home/calvinow/git/linux/drivers/scsi/scsi_lib.c:1196!
> > [ 14.653074] invalid opcode: 0000 [#1] PREEMPT SMP
> > [ 14.653076] CPU: 1 PID: 40 Comm: kworker/1:0H Not tainted 3.10.0-rc1-amd-00279-g8f710dd #3
> > [ 14.653077] Hardware name: System manufacturer System Product Name/M5A88-M, BIOS 0601 09/20/2011
> > [ 14.653082] Workqueue: kblockd blk_delay_work
> > <snip>
> > [ 14.653123] Call Trace:
> > [ 14.653126] [<ffffffff81477248>] sd_prep_fn+0x2c8/0xb70
> > [ 14.653129] [<ffffffff812c8b70>] ? deadline_remove_request.isra.9+0x50/0x90
> > [ 14.653132] [<ffffffff812b8f5b>] blk_peek_request+0xdb/0x210
> > [ 14.653134] [<ffffffff81465f15>] scsi_request_fn+0x45/0x4e0
> > [ 14.653136] [<ffffffff812b6a51>] __blk_run_queue+0x31/0x40
> > [ 14.653138] [<ffffffff812b6a84>] blk_delay_work+0x24/0x40
> > [ 14.653141] [<ffffffff8105dc2a>] process_one_work+0x1da/0x490
> > [ 14.653143] [<ffffffff8105dbcd>] ? process_one_work+0x17d/0x490
> > [ 14.653145] [<ffffffff8105e32a>] worker_thread+0x11a/0x370
> > [ 14.653147] [<ffffffff8105e210>] ? rescuer_thread+0x2f0/0x2f0
> > [ 14.653149] [<ffffffff81066296>] kthread+0xd6/0xe0
> > [ 14.653151] [<ffffffff810661c0>] ? __kthread_unpark+0x50/0x50
> > [ 14.653154] [<ffffffff816e4d6c>] ret_from_fork+0x7c/0xb0
> > [ 14.653156] [<ffffffff810661c0>] ? __kthread_unpark+0x50/0x50
> > [ 14.653172] Code: <snip>
> > [ 14.653174] RIP [<ffffffff81467329>] scsi_setup_fs_cmnd+0x89/0x90
> >
> > Signed-off-by: Calvin Owens <jcalvinowens@xxxxxxxxx>
>
> Kent, there was a report on this issue yesterday as well. We need to get
> this fixed up ASAP.
Sorry for the delay - been vacationing. Reproduced the original bug,
here's a patch that fixes it:
commit 402f5db3708b2062795a384a3d8397cf702e27bc
Author: Kent Overstreet <koverstreet@xxxxxxxxxx>
Date: Sun May 19 10:27:07 2013 -0700
raid5: Initialize bi_vcnt
The patch that converted raid5 to use bio_reset() forgot to initialize
bi_vcnt.
Signed-off-by: Kent Overstreet <koverstreet@xxxxxxxxxx>
Cc: NeilBrown <neilb@xxxxxxx>
Cc: Jens Axboe <axboe@xxxxxxxxx>
Cc: linux-raid@xxxxxxxxxxxxxxx
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 9359828..753f318 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -664,6 +664,7 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
bi->bi_rw |= REQ_FLUSH;
+ bi->bi_vcnt = 1;
bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
bi->bi_io_vec[0].bv_offset = 0;
bi->bi_size = STRIPE_SIZE;
@@ -701,6 +702,7 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
else
rbi->bi_sector = (sh->sector
+ rrdev->data_offset);
+ rbi->bi_vcnt = 1;
rbi->bi_io_vec[0].bv_len = STRIPE_SIZE;
rbi->bi_io_vec[0].bv_offset = 0;
rbi->bi_size = STRIPE_SIZE;
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
