md/raid1: restore ability for check and repair to fix read errors.

commit 30bc9b53878a9921b02e3b5bc4283ac1c6de102a
    md/raid1: fix bio handling problems in process_checks()

Move the bio_reset() to a point before where BIO_UPTODATE is checked,
so that check now always report that the bio is uptodate, even if it is not.

This causes process_check() to sometimes treat read-errors as
successful matches so the good data isn't written out.

This patch preserves the flag until it is needed.

Bug was introduced in 3.11, but backported to 3.10-stable (as it fixed
an even worse bug).  So suitable for any -stable since 3.10.

Reported-and-tested-by: Michael Tokarev <mjt@tls.msk.ru>
Cc: stable@vger.kernel.org (3.10+)
Fixed: 30bc9b53878a9921b02e3b5bc4283ac1c6de102a
Signed-off-by: NeilBrown <neilb@suse.de>

diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index fd3a2a14b587da5e3bb5046b0017ed7bd46f67a1..4a6ca1cb2e78539679b96a00b89542f6f0eab8f0 100644 (file)
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -1953,11 +1953,15 @@ static int process_checks(struct r1bio *r1_bio)
        for (i = 0; i < conf->raid_disks * 2; i++) {
                int j;
                int size;
+               int uptodate;
                struct bio *b = r1_bio->bios[i];
                if (b->bi_end_io != end_sync_read)
                        continue;
-               /* fixup the bio for reuse */
+               /* fixup the bio for reuse, but preserve BIO_UPTODATE */
+               uptodate = test_bit(BIO_UPTODATE, &b->bi_flags);
                bio_reset(b);
+               if (!uptodate)
+                       clear_bit(BIO_UPTODATE, &b->bi_flags);
                b->bi_vcnt = vcnt;
                b->bi_iter.bi_size = r1_bio->sectors << 9;
                b->bi_iter.bi_sector = r1_bio->sector +
@@ -1990,11 +1994,14 @@ static int process_checks(struct r1bio *r1_bio)
                int j;
                struct bio *pbio = r1_bio->bios[primary];
                struct bio *sbio = r1_bio->bios[i];
+               int uptodate = test_bit(BIO_UPTODATE, &sbio->bi_flags);
 
                if (sbio->bi_end_io != end_sync_read)
                        continue;
+               /* Now we can 'fixup' the BIO_UPTODATE flag */
+               set_bit(BIO_UPTODATE, &sbio->bi_flags);
 
-               if (test_bit(BIO_UPTODATE, &sbio->bi_flags)) {
+               if (uptodate) {
                        for (j = vcnt; j-- ; ) {
                                struct page *p, *s;
                                p = pbio->bi_io_vec[j].bv_page;
@@ -2009,7 +2016,7 @@ static int process_checks(struct r1bio *r1_bio)
                if (j >= 0)
                        atomic64_add(r1_bio->sectors, &mddev->resync_mismatches);
                if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
-                             && test_bit(BIO_UPTODATE, &sbio->bi_flags))) {
+                             && uptodate)) {
                        /* No need to write to this device. */
                        sbio->bi_end_io = NULL;
                        rdev_dec_pending(conf->mirrors[i].rdev, mddev);