btrfs: Reset sblock->xxx_error stats before calling scrub_recheck_block_checksum
authorZhao Lei <zhaolei@cn.fujitsu.com>
Mon, 24 Aug 2015 13:10:18 +0000 (21:10 +0800)
committerChris Mason <clm@fb.com>
Wed, 11 Nov 2015 03:27:03 +0000 (19:27 -0800)
We should reset sblock->xxx_error stats before calling
scrub_recheck_block_checksum().

Current code run correctly because all sblock are allocated by
k[cz]alloc(), and the error stats are not got changed.

Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
fs/btrfs/scrub.c

index 12ed8a7..95dc60a 100644 (file)
@@ -1495,6 +1495,7 @@ static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
        sblock->no_io_error_seen = 1;
        sblock->header_error = 0;
        sblock->checksum_error = 0;
+       sblock->generation_error = 0;
 
        for (page_num = 0; page_num < sblock->page_count; page_num++) {
                struct bio *bio;
@@ -2203,6 +2204,9 @@ static void scrub_missing_raid56_worker(struct btrfs_work *work)
        logical = sblock->pagev[0]->logical;
        dev = sblock->pagev[0]->dev;
 
+       sblock->header_error = 0;
+       sblock->checksum_error = 0;
+       sblock->generation_error = 0;
        if (sblock->no_io_error_seen) {
                scrub_recheck_block_checksum(fs_info, sblock, is_metadata,
                                             have_csum, csum, generation,