Re: raid 1 bug with write-mostly and administrative failed disk

[NeilBrown <neilb@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 1640 bytes --]

On Thu, 5 Jan 2012 22:30:23 +0100 Art -kwaak- van Breemen
<ard@telegraafnet.nl> wrote:

> Hi,
> 
> Please Cc: me too as I am trying to subscribe to the list.
> 
> Anyway: I found a small bug in raid1, with write-behind and
> write-mostly, occuring at least on 3.1.4 and 3.2 .
> 
> This is the test setup:
> mdadm --stop /dev/md5
> mdadm --zero-superblock /dev/sda8
> mdadm --zero-superblock /dev/sdb8
> mdadm --create -l 1 -n 2 --metadata=0.90 --bitmap=internal --bitmap-chunk=1024 --write-behind=2048 /dev/md5 /dev/sdb8 -W /dev/sda8
> (wait until finished)
> mdadm --fail /dev/md5 /dev/sdb8
> # And this to trigger the bug:
> dd if=/dev/md5 of=/dev/null bs=10k count=1
> 

Thanks for the excellent bug report.

I never tested write-behind :-(

Please confirm that this patch fixes it for you.

NeilBrown


diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index cc24f0c..a368db2 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -531,8 +531,17 @@ static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sect
 		if (test_bit(WriteMostly, &rdev->flags)) {
 			/* Don't balance among write-mostly, just
 			 * use the first as a last resort */
-			if (best_disk < 0)
+			if (best_disk < 0) {
+				if (is_badblock(rdev, this_sector, sectors,
+						&first_bad, &bad_sectors)) {
+					if (first_bad < this_sector)
+						/* Cannot use this */
+						continue;
+					best_good_sectors = first_bad - this_sector;
+				} else
+					best_good_sectors = sectors;
 				best_disk = disk;
+			}
 			continue;
 		}
 		/* This is a reasonable device to use.  It might

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]