Commit 3a9f28a5 authored by NeilBrown's avatar NeilBrown

md/raid1: improve handling of read failure during recovery.

If we cannot read a block from anywhere during recovery, there is
now a better approach than just giving up.
We can record a bad block on each device and keep going - being
careful not to clear the bad block when a write succeeds as it might -
it will be a write of incorrect data.

We have now reached the state where - for raid1 - we only call
md_error if md_set_badblocks has failed.
Signed-off-by: default avatarNeilBrown <neilb@suse.de>
Reviewed-by: default avatarNamhyung Kim <namhyung@gmail.com>
parent d8f05d29
...@@ -1392,7 +1392,12 @@ static void end_sync_write(struct bio *bio, int error) ...@@ -1392,7 +1392,12 @@ static void end_sync_write(struct bio *bio, int error)
} else if (is_badblock(conf->mirrors[mirror].rdev, } else if (is_badblock(conf->mirrors[mirror].rdev,
r1_bio->sector, r1_bio->sector,
r1_bio->sectors, r1_bio->sectors,
&first_bad, &bad_sectors)) &first_bad, &bad_sectors) &&
!is_badblock(conf->mirrors[r1_bio->read_disk].rdev,
r1_bio->sector,
r1_bio->sectors,
&first_bad, &bad_sectors)
)
set_bit(R1BIO_MadeGood, &r1_bio->state); set_bit(R1BIO_MadeGood, &r1_bio->state);
update_head_pos(mirror, r1_bio); update_head_pos(mirror, r1_bio);
...@@ -1473,16 +1478,36 @@ static int fix_sync_read_error(r1bio_t *r1_bio) ...@@ -1473,16 +1478,36 @@ static int fix_sync_read_error(r1bio_t *r1_bio)
if (!success) { if (!success) {
char b[BDEVNAME_SIZE]; char b[BDEVNAME_SIZE];
/* Cannot read from anywhere, array is toast */ int abort = 0;
md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev); /* Cannot read from anywhere, this block is lost.
* Record a bad block on each device. If that doesn't
* work just disable and interrupt the recovery.
* Don't fail devices as that won't really help.
*/
printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O read error" printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O read error"
" for block %llu\n", " for block %llu\n",
mdname(mddev), mdname(mddev),
bdevname(bio->bi_bdev, b), bdevname(bio->bi_bdev, b),
(unsigned long long)r1_bio->sector); (unsigned long long)r1_bio->sector);
md_done_sync(mddev, r1_bio->sectors, 0); for (d = 0; d < conf->raid_disks; d++) {
put_buf(r1_bio); rdev = conf->mirrors[d].rdev;
return 0; if (!rdev || test_bit(Faulty, &rdev->flags))
continue;
if (!rdev_set_badblocks(rdev, sect, s, 0))
abort = 1;
}
if (abort) {
mddev->recovery_disabled = 1;
set_bit(MD_RECOVERY_INTR, &mddev->recovery);
md_done_sync(mddev, r1_bio->sectors, 0);
put_buf(r1_bio);
return 0;
}
/* Try next page */
sectors -= s;
sect += s;
idx++;
continue;
} }
start = d; start = d;
...@@ -1879,7 +1904,9 @@ static void raid1d(mddev_t *mddev) ...@@ -1879,7 +1904,9 @@ static void raid1d(mddev_t *mddev)
if (bio->bi_end_io == NULL) if (bio->bi_end_io == NULL)
continue; continue;
if (test_bit(BIO_UPTODATE, if (test_bit(BIO_UPTODATE,
&bio->bi_flags)) { &bio->bi_flags) &&
test_bit(R1BIO_MadeGood,
&r1_bio->state)) {
rdev_clear_badblocks( rdev_clear_badblocks(
rdev, rdev,
r1_bio->sector, r1_bio->sector,
......
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