On Mon, 4 Mar 2013 09:52:57 +0800 majianpeng [off-list ref] wrote:

quoted

On Thu, 28 Feb 2013 15:51:01 +0800 majianpeng [off-list ref] wrote:

quoted

For badsector,it took two steps:rewrite and reread.So for
write-operation,it make no sense to rewrite after write-operation.

Please please please please please try to explain yourself better.

Sorry for that. 

quoted

This patch doesn't make any sense.  The write succeeded.  We set the MadeGood
flag so that later code will remove the block from the bad-block list.
There will be no re-read or re-write.

For a striep which contain badsector is doing write-operation.
In analyse_stripe, it check badsector and set R5_ReadError.
 write--->MadeGood->clear badblock log. But it can't clear R5_ReadError flag.
So it will exec those code in func handle_stripe:

After you wrote that, did you read it again to see if it makes any sense?  It
does but only just.

I'd rather have a good explanation in the code so it is clear what is going
on.

diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 3e0ec16..277d9c2 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c

@@ -1911,8 +1911,15 @@ static void raid5_end_write_request(struct bio *bi, int error)
 					&rdev->mddev->recovery);
 		} else if (is_badblock(rdev, sh->sector,
 				       STRIPE_SECTORS,
-				       &first_bad, &bad_sectors))
+				       &first_bad, &bad_sectors)) {
 			set_bit(R5_MadeGood, &sh->dev[i].flags);
+			if (test_bit(R5_ReadError, &sh->dev[i].flags))
+				/* That was a successful write so make
+				 * sure it looks like we already did
+				 * a re-write.
+				 */
+				set_bit(R5_ReWrite, &sh->dev[i].flags);
+		}
 	}
 	rdev_dec_pending(rdev, conf->mddev);
 

Thanks,
NeilBrown