btrfs: fix direct I/O read repair for split bios
authorChristoph Hellwig <hch@lst.de>
Thu, 24 Mar 2022 16:06:27 +0000 (17:06 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 9 May 2022 07:16:30 +0000 (09:16 +0200)
commit 00d825258bcc09c0e1b99aa7f9ad7d2c2fad41fa upstream.

When a bio is split in btrfs_submit_direct, dip->file_offset contains
the file offset for the first bio.  But this means the start value used
in btrfs_check_read_dio_bio is incorrect for subsequent bios.  Add
a file_offset field to struct btrfs_bio to pass along the correct offset.

Given that check_data_csum only uses start of an error message this
means problems with this miscalculation will only show up when I/O fails
or checksums mismatch.

The logic was removed in f4f39fc5dc30 ("btrfs: remove btrfs_bio::logical
member") but we need it due to the bio splitting.

CC: stable@vger.kernel.org # 5.16+
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Naohiro Aota <naohiro.aota@wdc.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Sweet Tea Dorminy <sweettea-kernel@dorminy.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/btrfs/extent_io.c
fs/btrfs/inode.c
fs/btrfs/volumes.h

index e93526d86a9226fb041acec887e9e544b857079f..9b488a737f8ae29e498a17d91ffd1991faa92eda 100644 (file)
@@ -2657,6 +2657,7 @@ int btrfs_repair_one_sector(struct inode *inode,
 
        repair_bio = btrfs_bio_alloc(1);
        repair_bbio = btrfs_bio(repair_bio);
+       repair_bbio->file_offset = start;
        repair_bio->bi_opf = REQ_OP_READ;
        repair_bio->bi_end_io = failed_bio->bi_end_io;
        repair_bio->bi_iter.bi_sector = failrec->logical >> 9;
index 9547088a930667cbd682992b4b039f5b93b09d19..86577d87eef5df53d8a710e41396ccd5728ba2cb 100644 (file)
@@ -7789,8 +7789,6 @@ static blk_status_t btrfs_check_read_dio_bio(struct btrfs_dio_private *dip,
        const bool csum = !(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM);
        struct bio_vec bvec;
        struct bvec_iter iter;
-       const u64 orig_file_offset = dip->file_offset;
-       u64 start = orig_file_offset;
        u32 bio_offset = 0;
        blk_status_t err = BLK_STS_OK;
 
@@ -7800,6 +7798,8 @@ static blk_status_t btrfs_check_read_dio_bio(struct btrfs_dio_private *dip,
                nr_sectors = BTRFS_BYTES_TO_BLKS(fs_info, bvec.bv_len);
                pgoff = bvec.bv_offset;
                for (i = 0; i < nr_sectors; i++) {
+                       u64 start = bbio->file_offset + bio_offset;
+
                        ASSERT(pgoff < PAGE_SIZE);
                        if (uptodate &&
                            (!csum || !check_data_csum(inode, bbio,
@@ -7812,17 +7812,13 @@ static blk_status_t btrfs_check_read_dio_bio(struct btrfs_dio_private *dip,
                        } else {
                                int ret;
 
-                               ASSERT((start - orig_file_offset) < UINT_MAX);
-                               ret = btrfs_repair_one_sector(inode,
-                                               &bbio->bio,
-                                               start - orig_file_offset,
-                                               bvec.bv_page, pgoff,
+                               ret = btrfs_repair_one_sector(inode, &bbio->bio,
+                                               bio_offset, bvec.bv_page, pgoff,
                                                start, bbio->mirror_num,
                                                submit_dio_repair_bio);
                                if (ret)
                                        err = errno_to_blk_status(ret);
                        }
-                       start += sectorsize;
                        ASSERT(bio_offset + sectorsize > bio_offset);
                        bio_offset += sectorsize;
                        pgoff += sectorsize;
@@ -8025,6 +8021,7 @@ static void btrfs_submit_direct(const struct iomap_iter *iter,
                bio = btrfs_bio_clone_partial(dio_bio, clone_offset, clone_len);
                bio->bi_private = dip;
                bio->bi_end_io = btrfs_end_dio_bio;
+               btrfs_bio(bio)->file_offset = file_offset;
 
                if (bio_op(bio) == REQ_OP_ZONE_APPEND) {
                        status = extract_ordered_extent(BTRFS_I(inode), bio,
index 005c9e2a491a1ff7d413f11f42ee44407af6e805..c22148bebc2f5cf231e1db76248279bedcb6034f 100644 (file)
@@ -323,6 +323,9 @@ struct btrfs_fs_devices {
 struct btrfs_bio {
        unsigned int mirror_num;
 
+       /* for direct I/O */
+       u64 file_offset;
+
        /* @device is for stripe IO submission. */
        struct btrfs_device *device;
        u8 *csum;