In 7471fb77(md/raid6: Fix anomily when recovering a single device in RAID6.) It avoids to re-read P
when it can be computed from other members. But it misses the chance to re-write the right data
to P. Now it sets R5_ReadError if the re-read fails. Because it avoids the re-read, so it misses
the chance to set R5_ReadError. The re-write is submitted in state machine when r5dev has flag
R5_ReadError. So it doesn't re-write the right data to disk. We need to do this to keep the raid
having right data.
Because it don't send re-read, so it also misses the chance to reset rdev->read_erros to 0. It can
fail the disk when there are many read errors on P member disk(other disks don't have read error)
Fixes: 7471fb77(md/raid6: Fix anomily when recovering a single device in RAID6.)
Signed-off-by: Xiao Ni <xni@xxxxxxxxxx>
---
drivers/md/raid5.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 9c4f765..6cdd37f 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -2561,7 +2561,9 @@ static void raid5_end_read_request(struct bio * bi)
&& !test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
retry = 1;
if (retry)
- if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags)) {
+ if (sh->qd_idx >= 0 && sh->pd_idx == i)
+ set_bit(R5_ReadError, &sh->dev[i].flags);
+ else if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags)) {
set_bit(R5_ReadError, &sh->dev[i].flags);
clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);
} else
@@ -6128,7 +6130,10 @@ static int retry_aligned_read(struct r5conf *conf, struct bio *raid_bio,
return handled;
}
- set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
+ if (sh->qd_idx >= 0 && sh->pd_idx == i)
+ set_bit(R5_ReadError, &sh->dev[i].flags);
+ else
+ set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
handle_stripe(sh);
raid5_release_stripe(sh);
handled++;
--
2.7.5
