Scrub an individual inode's block mappings to make sure they make sense.
Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
---
fs/xfs/libxfs/xfs_bmap_btree.c | 26 +++
fs/xfs/libxfs/xfs_fs.h | 5 +
fs/xfs/xfs_scrub.c | 301 ++++++++++++++++++++++++++++++++++++++++
fs/xfs/xfs_trace.h | 5 +
4 files changed, 331 insertions(+), 6 deletions(-)
diff --git a/fs/xfs/libxfs/xfs_bmap_btree.c b/fs/xfs/libxfs/xfs_bmap_btree.c
index 049fa59..54dd7a7 100644
--- a/fs/xfs/libxfs/xfs_bmap_btree.c
+++ b/fs/xfs/libxfs/xfs_bmap_btree.c
@@ -623,6 +623,16 @@ xfs_bmbt_init_key_from_rec(
}
STATIC void
+xfs_bmbt_init_high_key_from_rec(
+ union xfs_btree_key *key,
+ union xfs_btree_rec *rec)
+{
+ key->bmbt.br_startoff = cpu_to_be64(
+ xfs_bmbt_disk_get_startoff(&rec->bmbt) +
+ xfs_bmbt_disk_get_blockcount(&rec->bmbt) - 1);
+}
+
+STATIC void
xfs_bmbt_init_rec_from_cur(
struct xfs_btree_cur *cur,
union xfs_btree_rec *rec)
@@ -647,6 +657,16 @@ xfs_bmbt_key_diff(
cur->bc_rec.b.br_startoff;
}
+STATIC __int64_t
+xfs_bmbt_diff_two_keys(
+ struct xfs_btree_cur *cur,
+ union xfs_btree_key *k1,
+ union xfs_btree_key *k2)
+{
+ return (__int64_t)be64_to_cpu(k1->bmbt.br_startoff) -
+ be64_to_cpu(k2->bmbt.br_startoff);
+}
+
static bool
xfs_bmbt_verify(
struct xfs_buf *bp)
@@ -737,7 +757,6 @@ const struct xfs_buf_ops xfs_bmbt_buf_ops = {
};
-#if defined(DEBUG) || defined(XFS_WARN)
STATIC int
xfs_bmbt_keys_inorder(
struct xfs_btree_cur *cur,
@@ -758,7 +777,6 @@ xfs_bmbt_recs_inorder(
xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
xfs_bmbt_disk_get_startoff(&r2->bmbt);
}
-#endif /* DEBUG */
static const struct xfs_btree_ops xfs_bmbt_ops = {
.rec_len = sizeof(xfs_bmbt_rec_t),
@@ -772,14 +790,14 @@ static const struct xfs_btree_ops xfs_bmbt_ops = {
.get_minrecs = xfs_bmbt_get_minrecs,
.get_dmaxrecs = xfs_bmbt_get_dmaxrecs,
.init_key_from_rec = xfs_bmbt_init_key_from_rec,
+ .init_high_key_from_rec = xfs_bmbt_init_high_key_from_rec,
.init_rec_from_cur = xfs_bmbt_init_rec_from_cur,
.init_ptr_from_cur = xfs_bmbt_init_ptr_from_cur,
.key_diff = xfs_bmbt_key_diff,
+ .diff_two_keys = xfs_bmbt_diff_two_keys,
.buf_ops = &xfs_bmbt_buf_ops,
-#if defined(DEBUG) || defined(XFS_WARN)
.keys_inorder = xfs_bmbt_keys_inorder,
.recs_inorder = xfs_bmbt_recs_inorder,
-#endif
};
/*
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 251db32..5c80264 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -586,7 +586,10 @@ struct xfs_scrub_metadata {
#define XFS_SCRUB_TYPE_RMAPBT 9 /* reverse mapping btree */
#define XFS_SCRUB_TYPE_REFCNTBT 10 /* reference count btree */
#define XFS_SCRUB_TYPE_INODE 11 /* inode record */
-#define XFS_SCRUB_TYPE_MAX 11
+#define XFS_SCRUB_TYPE_BMBTD 12 /* data fork block mapping */
+#define XFS_SCRUB_TYPE_BMBTA 13 /* attr fork block mapping */
+#define XFS_SCRUB_TYPE_BMBTC 14 /* CoW fork block mapping */
+#define XFS_SCRUB_TYPE_MAX 14
#define XFS_SCRUB_FLAG_REPAIR 0x1 /* i: repair this metadata */
#define XFS_SCRUB_FLAG_CORRUPT 0x2 /* o: needs repair */
diff --git a/fs/xfs/xfs_scrub.c b/fs/xfs/xfs_scrub.c
index 3723c06..4e7b4b4 100644
--- a/fs/xfs/xfs_scrub.c
+++ b/fs/xfs/xfs_scrub.c
@@ -1426,6 +1426,26 @@ xfs_scrub_setup_inode(
return error;
}
+/* Set us up with an inode and AG headers, if needed. */
+STATIC int
+xfs_scrub_setup_inode_bmap(
+ struct xfs_scrub_context *sc,
+ struct xfs_inode *ip,
+ struct xfs_scrub_metadata *sm,
+ bool retry_deadlocked)
+{
+ int error;
+
+ error = xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked);
+ if (error || !retry_deadlocked)
+ return error;
+
+ error = xfs_scrub_ag_lock_all(sc);
+ if (error)
+ return xfs_scrub_teardown(sc, ip, error);
+ return 0;
+}
+
/* Metadata scrubbers */
#define XFS_SCRUB_SB_CHECK(fs_ok) \
@@ -2161,6 +2181,284 @@ xfs_scrub_inode(
#undef XFS_SCRUB_INODE_GOTO
#undef XFS_SCRUB_INODE_CHECK
+/*
+ * Inode fork block mapping (BMBT) scrubber.
+ * More complex than the others because we have to scrub
+ * all the extents regardless of whether or not the fork
+ * is in btree format.
+ */
+
+struct xfs_scrub_bmap_info {
+ struct xfs_scrub_context *sc;
+ const char *type;
+ xfs_daddr_t eofs;
+ xfs_fileoff_t lastoff;
+ bool is_rt;
+ bool is_shared;
+ bool scrub_btrec;
+ int whichfork;
+};
+
+#define XFS_SCRUB_BMAP_CHECK(fs_ok) \
+ XFS_SCRUB_INO_CHECK(info->sc, bp, info->type, fs_ok)
+#define XFS_SCRUB_BMAP_GOTO(fs_ok, label) \
+ XFS_SCRUB_INO_GOTO(info->sc, bp, info->type, fs_ok, label)
+#define XFS_SCRUB_BMAP_OP_ERROR_GOTO(label) \
+ XFS_SCRUB_OP_ERROR_GOTO(info->sc, agno, 0, "bmap", &error, label);
+/* Scrub a single extent record. */
+STATIC int
+xfs_scrub_bmap_extent(
+ struct xfs_inode *ip,
+ struct xfs_btree_cur *cur,
+ struct xfs_scrub_bmap_info *info,
+ struct xfs_bmbt_irec *irec)
+{
+ struct xfs_scrub_ag sa = {0};
+ struct xfs_mount *mp = ip->i_mount;
+ struct xfs_buf *bp = NULL;
+ xfs_daddr_t daddr;
+ xfs_daddr_t dlen;
+ xfs_agnumber_t agno;
+ int error = 0;
+
+ if (cur)
+ xfs_btree_get_block(cur, 0, &bp);
+
+ XFS_SCRUB_BMAP_CHECK(irec->br_startoff >= info->lastoff);
+ XFS_SCRUB_BMAP_CHECK(irec->br_startblock != HOLESTARTBLOCK);
+
+ if (isnullstartblock(irec->br_startblock)) {
+ XFS_SCRUB_BMAP_CHECK(irec->br_state == XFS_EXT_NORM);
+ goto out;
+ }
+
+ /* Actual mapping, so check the block ranges. */
+ if (info->is_rt) {
+ daddr = XFS_FSB_TO_BB(mp, irec->br_startblock);
+ agno = NULLAGNUMBER;
+ } else {
+ daddr = XFS_FSB_TO_DADDR(mp, irec->br_startblock);
+ agno = XFS_FSB_TO_AGNO(mp, irec->br_startblock);
+ }
+ dlen = XFS_FSB_TO_BB(mp, irec->br_blockcount);
+ XFS_SCRUB_BMAP_CHECK(daddr < info->eofs);
+ XFS_SCRUB_BMAP_CHECK(daddr + dlen < info->eofs);
+ XFS_SCRUB_BMAP_CHECK(irec->br_state != XFS_EXT_UNWRITTEN ||
+ xfs_sb_version_hasextflgbit(&mp->m_sb));
+ if (error)
+ goto out;
+
+ /* Set ourselves up for cross-referencing later. */
+ if (!info->is_rt) {
+ if (!xfs_scrub_ag_can_lock(info->sc, agno))
+ return -EDEADLOCK;
+ error = xfs_scrub_ag_init(info->sc, agno, &sa);
+ XFS_SCRUB_BMAP_OP_ERROR_GOTO(out);
+ }
+
+ xfs_scrub_ag_free(&sa);
+out:
+ info->lastoff = irec->br_startoff + irec->br_blockcount;
+ return error;
+}
+#undef XFS_SCRUB_BMAP_OP_ERROR_GOTO
+#undef XFS_SCRUB_BMAP_GOTO
+#undef XFS_SCRUB_BMAP_CHECK
+
+/* Scrub a bmbt record. */
+STATIC int
+xfs_scrub_bmapbt_helper(
+ struct xfs_scrub_btree *bs,
+ union xfs_btree_rec *rec)
+{
+ struct xfs_bmbt_rec_host ihost;
+ struct xfs_bmbt_irec irec;
+ struct xfs_scrub_bmap_info *info = bs->private;
+
+ if (!info->scrub_btrec)
+ return 0;
+
+ /* Set up the in-core record and scrub it. */
+ ihost.l0 = be64_to_cpu(rec->bmbt.l0);
+ ihost.l1 = be64_to_cpu(rec->bmbt.l1);
+ xfs_bmbt_get_all(&ihost, &irec);
+ return xfs_scrub_bmap_extent(bs->cur->bc_private.b.ip, bs->cur,
+ info, &irec);
+}
+
+#define XFS_SCRUB_FORK_CHECK(fs_ok) \
+ XFS_SCRUB_INO_CHECK(sc, NULL, info.type, fs_ok);
+#define XFS_SCRUB_FORK_GOTO(fs_ok, label) \
+ XFS_SCRUB_INO_GOTO(sc, NULL, info.type, fs_ok, label);
+#define XFS_SCRUB_FORK_OP_ERROR_GOTO(label) \
+ XFS_SCRUB_OP_ERROR_GOTO(sc, \
+ XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino), \
+ XFS_INO_TO_AGBNO(ip->i_mount, ip->i_ino), \
+ info.type, &error, label)
+/* Scrub an inode fork's block mappings. */
+STATIC int
+xfs_scrub_bmap(
+ struct xfs_scrub_context *sc,
+ int whichfork)
+{
+ struct xfs_bmbt_irec irec;
+ struct xfs_scrub_bmap_info info = {0};
+ struct xfs_owner_info oinfo;
+ struct xfs_mount *mp = sc->tp->t_mountp;
+ struct xfs_inode *ip = sc->ip;
+ struct xfs_ifork *ifp;
+ struct xfs_btree_cur *cur;
+ xfs_fileoff_t off;
+ xfs_fileoff_t endoff;
+ int nmaps;
+ int flags = 0;
+ int error = 0;
+ int err2 = 0;
+
+ switch (whichfork) {
+ case XFS_DATA_FORK:
+ info.type = "data fork";
+ break;
+ case XFS_ATTR_FORK:
+ info.type = "attr fork";
+ break;
+ case XFS_COW_FORK:
+ info.type = "CoW fork";
+ break;
+ }
+ ifp = XFS_IFORK_PTR(ip, whichfork);
+
+ info.is_rt = whichfork == XFS_DATA_FORK && XFS_IS_REALTIME_INODE(ip);
+ info.eofs = XFS_FSB_TO_BB(mp, info.is_rt ? mp->m_sb.sb_rblocks :
+ mp->m_sb.sb_dblocks);
+ info.whichfork = whichfork;
+ info.is_shared = whichfork == XFS_DATA_FORK && xfs_is_reflink_inode(ip);
+ info.sc = sc;
+
+ switch (whichfork) {
+ case XFS_COW_FORK:
+ /* Non-existent CoW forks are ignorable. */
+ if (!ifp)
+ goto out_unlock;
+ /* No CoW forks on non-reflink inodes/filesystems. */
+ XFS_SCRUB_FORK_GOTO(xfs_is_reflink_inode(ip), out_unlock);
+ break;
+ case XFS_ATTR_FORK:
+ if (!ifp)
+ goto out_unlock;
+ XFS_SCRUB_FORK_CHECK(xfs_sb_version_hasattr(&mp->m_sb));
+ break;
+ }
+
+ /* Check the fork values */
+ switch (XFS_IFORK_FORMAT(ip, whichfork)) {
+ case XFS_DINODE_FMT_UUID:
+ case XFS_DINODE_FMT_DEV:
+ case XFS_DINODE_FMT_LOCAL:
+ /* No mappings to check. */
+ goto out_unlock;
+ case XFS_DINODE_FMT_EXTENTS:
+ XFS_SCRUB_FORK_GOTO(ifp->if_flags & XFS_IFEXTENTS, out_unlock);
+ break;
+ case XFS_DINODE_FMT_BTREE:
+ XFS_SCRUB_FORK_CHECK(whichfork != XFS_COW_FORK);
+ /*
+ * Scan the btree. If extents aren't loaded, have the btree
+ * scrub routine examine the extent records.
+ */
+ info.scrub_btrec = !(ifp->if_flags & XFS_IFEXTENTS);
+
+ cur = xfs_bmbt_init_cursor(mp, sc->tp, ip, whichfork);
+ xfs_rmap_ino_bmbt_owner(&oinfo, ip->i_ino, whichfork);
+ err2 = xfs_scrub_btree(sc, cur, xfs_scrub_bmapbt_helper,
+ &oinfo, &info);
+ xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
+ if (err2 == -EDEADLOCK)
+ return err2;
+ else if (err2)
+ goto out_unlock;
+ /* Skip in-core extent checking if we did it in the btree */
+ if (info.scrub_btrec)
+ goto out_unlock;
+ break;
+ default:
+ XFS_SCRUB_FORK_GOTO(false, out_unlock);
+ break;
+ }
+
+ /* Extent data is in memory, so scrub that. */
+ switch (whichfork) {
+ case XFS_ATTR_FORK:
+ flags |= XFS_BMAPI_ATTRFORK;
+ break;
+ case XFS_COW_FORK:
+ flags |= XFS_BMAPI_COWFORK;
+ break;
+ default:
+ break;
+ }
+
+ /* Find the offset of the last extent in the mapping. */
+ error = xfs_bmap_last_offset(ip, &endoff, whichfork);
+ XFS_SCRUB_FORK_OP_ERROR_GOTO(out_unlock);
+
+ /* Scrub extent records. */
+ off = 0;
+ while (true) {
+ nmaps = 1;
+ err2 = xfs_bmapi_read(ip, off, endoff - off, &irec,
+ &nmaps, flags);
+ if (err2 || nmaps == 0 || irec.br_startoff > endoff)
+ break;
+ /* Scrub non-hole extent. */
+ if (irec.br_startblock != HOLESTARTBLOCK) {
+ err2 = xfs_scrub_bmap_extent(ip, NULL, &info, &irec);
+ if (err2 == -EDEADLOCK)
+ return err2;
+ else if (!error && err2)
+ error = err2;
+ if (xfs_scrub_should_terminate(&error))
+ break;
+ }
+
+ off += irec.br_blockcount;
+ }
+
+out_unlock:
+ if (error == 0 && err2 != 0)
+ error = err2;
+ return error;
+}
+#undef XFS_SCRUB_FORK_CHECK
+#undef XFS_SCRUB_FORK_GOTO
+
+/* Scrub an inode's data fork. */
+STATIC int
+xfs_scrub_bmap_data(
+ struct xfs_scrub_context *sc)
+{
+ return xfs_scrub_bmap(sc, XFS_DATA_FORK);
+}
+
+/* Scrub an inode's attr fork. */
+STATIC int
+xfs_scrub_bmap_attr(
+ struct xfs_scrub_context *sc)
+{
+ return xfs_scrub_bmap(sc, XFS_ATTR_FORK);
+}
+
+/* Scrub an inode's CoW fork. */
+STATIC int
+xfs_scrub_bmap_cow(
+ struct xfs_scrub_context *sc)
+{
+ if (!xfs_is_reflink_inode(sc->ip))
+ return -ENOENT;
+
+ return xfs_scrub_bmap(sc, XFS_COW_FORK);
+}
+
/* Scrubbing dispatch. */
struct xfs_scrub_meta_fns {
@@ -2184,6 +2482,9 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
{xfs_scrub_setup_ag_header, xfs_scrub_rmapbt, NULL, xfs_sb_version_hasrmapbt},
{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, NULL, xfs_sb_version_hasreflink},
{xfs_scrub_setup_inode, xfs_scrub_inode, NULL, NULL},
+ {xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_data, NULL, NULL},
+ {xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_attr, NULL, NULL},
+ {xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_cow, NULL, NULL},
};
/* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index e37c8bc..dd48b5a 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3473,7 +3473,10 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
{ XFS_SCRUB_TYPE_FINOBT, "finobt" }, \
{ XFS_SCRUB_TYPE_RMAPBT, "rmapbt" }, \
{ XFS_SCRUB_TYPE_REFCNTBT, "refcountbt" }, \
- { XFS_SCRUB_TYPE_INODE, "inode" }
+ { XFS_SCRUB_TYPE_INODE, "inode" }, \
+ { XFS_SCRUB_TYPE_BMBTD, "bmapbtd" }, \
+ { XFS_SCRUB_TYPE_BMBTA, "bmapbta" }, \
+ { XFS_SCRUB_TYPE_BMBTC, "bmapbtc" }
DECLARE_EVENT_CLASS(xfs_scrub_class,
TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
xfs_ino_t inum, unsigned int gen, unsigned int flags,
--
To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
