ext4: lock the xattr block before checksuming it
authorTheodore Ts'o <tytso@mit.edu>
Sat, 25 Mar 2017 21:22:47 +0000 (17:22 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Sat, 25 Mar 2017 21:22:47 +0000 (17:22 -0400)
We must lock the xattr block before calculating or verifying the
checksum in order to avoid spurious checksum failures.

https://bugzilla.kernel.org/show_bug.cgi?id=193661

Reported-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
fs/ext4/xattr.c

index 67636ac..996e790 100644 (file)
@@ -131,31 +131,26 @@ static __le32 ext4_xattr_block_csum(struct inode *inode,
 }
 
 static int ext4_xattr_block_csum_verify(struct inode *inode,
-                                       sector_t block_nr,
-                                       struct ext4_xattr_header *hdr)
+                                       struct buffer_head *bh)
 {
-       if (ext4_has_metadata_csum(inode->i_sb) &&
-           (hdr->h_checksum != ext4_xattr_block_csum(inode, block_nr, hdr)))
-               return 0;
-       return 1;
-}
-
-static void ext4_xattr_block_csum_set(struct inode *inode,
-                                     sector_t block_nr,
-                                     struct ext4_xattr_header *hdr)
-{
-       if (!ext4_has_metadata_csum(inode->i_sb))
-               return;
+       struct ext4_xattr_header *hdr = BHDR(bh);
+       int ret = 1;
 
-       hdr->h_checksum = ext4_xattr_block_csum(inode, block_nr, hdr);
+       if (ext4_has_metadata_csum(inode->i_sb)) {
+               lock_buffer(bh);
+               ret = (hdr->h_checksum == ext4_xattr_block_csum(inode,
+                                                       bh->b_blocknr, hdr));
+               unlock_buffer(bh);
+       }
+       return ret;
 }
 
-static inline int ext4_handle_dirty_xattr_block(handle_t *handle,
-                                               struct inode *inode,
-                                               struct buffer_head *bh)
+static void ext4_xattr_block_csum_set(struct inode *inode,
+                                     struct buffer_head *bh)
 {
-       ext4_xattr_block_csum_set(inode, bh->b_blocknr, BHDR(bh));
-       return ext4_handle_dirty_metadata(handle, inode, bh);
+       if (ext4_has_metadata_csum(inode->i_sb))
+               BHDR(bh)->h_checksum = ext4_xattr_block_csum(inode,
+                                               bh->b_blocknr, BHDR(bh));
 }
 
 static inline const struct xattr_handler *
@@ -233,7 +228,7 @@ ext4_xattr_check_block(struct inode *inode, struct buffer_head *bh)
        if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
            BHDR(bh)->h_blocks != cpu_to_le32(1))
                return -EFSCORRUPTED;
-       if (!ext4_xattr_block_csum_verify(inode, bh->b_blocknr, BHDR(bh)))
+       if (!ext4_xattr_block_csum_verify(inode, bh))
                return -EFSBADCRC;
        error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size,
                                       bh->b_data);
@@ -618,23 +613,22 @@ ext4_xattr_release_block(handle_t *handle, struct inode *inode,
                        }
                }
 
+               ext4_xattr_block_csum_set(inode, bh);
                /*
                 * Beware of this ugliness: Releasing of xattr block references
                 * from different inodes can race and so we have to protect
                 * from a race where someone else frees the block (and releases
                 * its journal_head) before we are done dirtying the buffer. In
                 * nojournal mode this race is harmless and we actually cannot
-                * call ext4_handle_dirty_xattr_block() with locked buffer as
+                * call ext4_handle_dirty_metadata() with locked buffer as
                 * that function can call sync_dirty_buffer() so for that case
                 * we handle the dirtying after unlocking the buffer.
                 */
                if (ext4_handle_valid(handle))
-                       error = ext4_handle_dirty_xattr_block(handle, inode,
-                                                             bh);
+                       error = ext4_handle_dirty_metadata(handle, inode, bh);
                unlock_buffer(bh);
                if (!ext4_handle_valid(handle))
-                       error = ext4_handle_dirty_xattr_block(handle, inode,
-                                                             bh);
+                       error = ext4_handle_dirty_metadata(handle, inode, bh);
                if (IS_SYNC(inode))
                        ext4_handle_sync(handle);
                dquot_free_block(inode, EXT4_C2B(EXT4_SB(inode->i_sb), 1));
@@ -863,13 +857,14 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
                                ext4_xattr_cache_insert(ext4_mb_cache,
                                        bs->bh);
                        }
+                       ext4_xattr_block_csum_set(inode, bs->bh);
                        unlock_buffer(bs->bh);
                        if (error == -EFSCORRUPTED)
                                goto bad_block;
                        if (!error)
-                               error = ext4_handle_dirty_xattr_block(handle,
-                                                                     inode,
-                                                                     bs->bh);
+                               error = ext4_handle_dirty_metadata(handle,
+                                                                  inode,
+                                                                  bs->bh);
                        if (error)
                                goto cleanup;
                        goto inserted;
@@ -967,10 +962,11 @@ inserted:
                                        ce->e_reusable = 0;
                                ea_bdebug(new_bh, "reusing; refcount now=%d",
                                          ref);
+                               ext4_xattr_block_csum_set(inode, new_bh);
                                unlock_buffer(new_bh);
-                               error = ext4_handle_dirty_xattr_block(handle,
-                                                                     inode,
-                                                                     new_bh);
+                               error = ext4_handle_dirty_metadata(handle,
+                                                                  inode,
+                                                                  new_bh);
                                if (error)
                                        goto cleanup_dquot;
                        }
@@ -1020,11 +1016,12 @@ getblk_failed:
                                goto getblk_failed;
                        }
                        memcpy(new_bh->b_data, s->base, new_bh->b_size);
+                       ext4_xattr_block_csum_set(inode, new_bh);
                        set_buffer_uptodate(new_bh);
                        unlock_buffer(new_bh);
                        ext4_xattr_cache_insert(ext4_mb_cache, new_bh);
-                       error = ext4_handle_dirty_xattr_block(handle,
-                                                             inode, new_bh);
+                       error = ext4_handle_dirty_metadata(handle, inode,
+                                                          new_bh);
                        if (error)
                                goto cleanup;
                }