md/raid1: clear bad-block record when write succeeds.
authorNeilBrown <neilb@suse.de>
Thu, 28 Jul 2011 01:31:49 +0000 (11:31 +1000)
committerNeilBrown <neilb@suse.de>
Thu, 28 Jul 2011 01:31:49 +0000 (11:31 +1000)
If we succeed in writing to a block that was recorded as
being bad, we clear the bad-block record.

This requires some delayed handling as the bad-block-list update has
to happen in process-context.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Namhyung Kim <namhyung@gmail.com>
drivers/md/raid1.c
drivers/md/raid1.h

index 3214606..141de75 100644 (file)
@@ -163,7 +163,7 @@ static void put_all_bios(conf_t *conf, r1bio_t *r1_bio)
 
        for (i = 0; i < conf->raid_disks; i++) {
                struct bio **bio = r1_bio->bios + i;
-               if (*bio && *bio != IO_BLOCKED)
+               if (!BIO_SPECIAL(*bio))
                        bio_put(*bio);
                *bio = NULL;
        }
@@ -337,7 +337,10 @@ static void r1_bio_write_done(r1bio_t *r1_bio)
                                !test_bit(R1BIO_Degraded, &r1_bio->state),
                                test_bit(R1BIO_BehindIO, &r1_bio->state));
                md_write_end(r1_bio->mddev);
-               raid_end_bio_io(r1_bio);
+               if (test_bit(R1BIO_MadeGood, &r1_bio->state))
+                       reschedule_retry(r1_bio);
+               else
+                       raid_end_bio_io(r1_bio);
        }
 }
 
@@ -363,7 +366,7 @@ static void raid1_end_write_request(struct bio *bio, int error)
                md_error(r1_bio->mddev, conf->mirrors[mirror].rdev);
                /* an I/O failed, we can't clear the bitmap */
                set_bit(R1BIO_Degraded, &r1_bio->state);
-       } else
+       } else {
                /*
                 * Set R1BIO_Uptodate in our master bio, so that we
                 * will return a good error code for to the higher
@@ -374,8 +377,20 @@ static void raid1_end_write_request(struct bio *bio, int error)
                 * to user-side. So if something waits for IO, then it
                 * will wait for the 'master' bio.
                 */
+               sector_t first_bad;
+               int bad_sectors;
+
                set_bit(R1BIO_Uptodate, &r1_bio->state);
 
+               /* Maybe we can clear some bad blocks. */
+               if (is_badblock(conf->mirrors[mirror].rdev,
+                               r1_bio->sector, r1_bio->sectors,
+                               &first_bad, &bad_sectors)) {
+                       r1_bio->bios[mirror] = IO_MADE_GOOD;
+                       set_bit(R1BIO_MadeGood, &r1_bio->state);
+               }
+       }
+
        update_head_pos(mirror, r1_bio);
 
        if (behind) {
@@ -402,7 +417,9 @@ static void raid1_end_write_request(struct bio *bio, int error)
                        }
                }
        }
-       rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
+       if (r1_bio->bios[mirror] == NULL)
+               rdev_dec_pending(conf->mirrors[mirror].rdev,
+                                conf->mddev);
 
        /*
         * Let's see if all mirrored write operations have finished
@@ -1340,6 +1357,8 @@ static void end_sync_write(struct bio *bio, int error)
        conf_t *conf = mddev->private;
        int i;
        int mirror=0;
+       sector_t first_bad;
+       int bad_sectors;
 
        for (i = 0; i < conf->raid_disks; i++)
                if (r1_bio->bios[i] == bio) {
@@ -1358,14 +1377,22 @@ static void end_sync_write(struct bio *bio, int error)
                        sectors_to_go -= sync_blocks;
                } while (sectors_to_go > 0);
                md_error(mddev, conf->mirrors[mirror].rdev);
-       }
+       } else if (is_badblock(conf->mirrors[mirror].rdev,
+                              r1_bio->sector,
+                              r1_bio->sectors,
+                              &first_bad, &bad_sectors))
+               set_bit(R1BIO_MadeGood, &r1_bio->state);
 
        update_head_pos(mirror, r1_bio);
 
        if (atomic_dec_and_test(&r1_bio->remaining)) {
-               sector_t s = r1_bio->sectors;
-               put_buf(r1_bio);
-               md_done_sync(mddev, s, uptodate);
+               int s = r1_bio->sectors;
+               if (test_bit(R1BIO_MadeGood, &r1_bio->state))
+                       reschedule_retry(r1_bio);
+               else {
+                       put_buf(r1_bio);
+                       md_done_sync(mddev, s, uptodate);
+               }
        }
 }
 
@@ -1727,9 +1754,39 @@ static void raid1d(mddev_t *mddev)
 
                mddev = r1_bio->mddev;
                conf = mddev->private;
-               if (test_bit(R1BIO_IsSync, &r1_bio->state))
-                       sync_request_write(mddev, r1_bio);
-               else if (test_bit(R1BIO_ReadError, &r1_bio->state)) {
+               if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
+                       if (test_bit(R1BIO_MadeGood, &r1_bio->state)) {
+                               int m;
+                               int s = r1_bio->sectors;
+                               for (m = 0; m < conf->raid_disks ; m++) {
+                                       struct bio *bio = r1_bio->bios[m];
+                                       if (bio->bi_end_io != NULL &&
+                                           test_bit(BIO_UPTODATE,
+                                                    &bio->bi_flags)) {
+                                               rdev = conf->mirrors[m].rdev;
+                                               rdev_clear_badblocks(
+                                                       rdev,
+                                                       r1_bio->sector,
+                                                       r1_bio->sectors);
+                                       }
+                               }
+                               put_buf(r1_bio);
+                               md_done_sync(mddev, s, 1);
+                       } else
+                               sync_request_write(mddev, r1_bio);
+               } else if (test_bit(R1BIO_MadeGood, &r1_bio->state)) {
+                       int m;
+                       for (m = 0; m < conf->raid_disks ; m++)
+                               if (r1_bio->bios[m] == IO_MADE_GOOD) {
+                                       rdev = conf->mirrors[m].rdev;
+                                       rdev_clear_badblocks(
+                                               rdev,
+                                               r1_bio->sector,
+                                               r1_bio->sectors);
+                                       rdev_dec_pending(rdev, mddev);
+                               }
+                       raid_end_bio_io(r1_bio);
+               } else if (test_bit(R1BIO_ReadError, &r1_bio->state)) {
                        int disk;
                        int max_sectors;
 
index aa6af37..f81360d 100644 (file)
@@ -116,7 +116,14 @@ struct r1bio_s {
  * correct the read error.  To keep track of bad blocks on a per-bio
  * level, we store IO_BLOCKED in the appropriate 'bios' pointer
  */
-#define IO_BLOCKED ((struct bio*)1)
+#define IO_BLOCKED ((struct bio *)1)
+/* When we successfully write to a known bad-block, we need to remove the
+ * bad-block marking which must be done from process context.  So we record
+ * the success by setting bios[n] to IO_MADE_GOOD
+ */
+#define IO_MADE_GOOD ((struct bio *)2)
+
+#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)
 
 /* bits for r1bio.state */
 #define        R1BIO_Uptodate  0
@@ -135,6 +142,10 @@ struct r1bio_s {
  * Record that bi_end_io was called with this flag...
  */
 #define        R1BIO_Returned 6
+/* If a write for this request means we can clear some
+ * known-bad-block records, we set this flag
+ */
+#define R1BIO_MadeGood 7
 
 extern int md_raid1_congested(mddev_t *mddev, int bits);