xfs: create a shadow rmap btree during rmap repair

Create an in-memory btree of rmap records instead of an array.  This
enables us to do live record collection instead of freezing the fs.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>

fs/xfs/libxfs/xfs_rmap.c

@@ -269,6 +269,16 @@ xfs_rmap_check_irec(
 	return NULL;
 }
 
+static inline xfs_failaddr_t
+xfs_rmap_check_btrec(
+	struct xfs_btree_cur		*cur,
+	const struct xfs_rmap_irec	*irec)
+{
+	if (xfs_btree_is_mem_rmap(cur->bc_ops))
+		return xfs_rmap_check_irec(cur->bc_mem.pag, irec);
+	return xfs_rmap_check_irec(cur->bc_ag.pag, irec);
+}
+
 static inline int
 xfs_rmap_complain_bad_rec(
 	struct xfs_btree_cur		*cur,
@@ -277,6 +287,10 @@ xfs_rmap_complain_bad_rec(
 {
 	struct xfs_mount		*mp = cur->bc_mp;
 
+	if (xfs_btree_is_mem_rmap(cur->bc_ops))
+		xfs_warn(mp,
+ "In-Memory Reverse Mapping BTree record corruption detected at %pS!", fa);
+	else
 		xfs_warn(mp,
  "Reverse Mapping BTree record corruption in AG %d detected at %pS!",
 			cur->bc_ag.pag->pag_agno, fa);
@@ -307,7 +321,7 @@ xfs_rmap_get_rec(
 
 	fa = xfs_rmap_btrec_to_irec(rec, irec);
 	if (!fa)
-		fa = xfs_rmap_check_irec(cur->bc_ag.pag, irec);
+		fa = xfs_rmap_check_btrec(cur, irec);
 	if (fa)
 		return xfs_rmap_complain_bad_rec(cur, fa, irec);
 
@@ -2404,15 +2418,12 @@ xfs_rmap_map_raw(
 {
 	struct xfs_owner_info	oinfo;
 
-	oinfo.oi_owner = rmap->rm_owner;
-	oinfo.oi_offset = rmap->rm_offset;
-	oinfo.oi_flags = 0;
-	if (rmap->rm_flags & XFS_RMAP_ATTR_FORK)
-		oinfo.oi_flags |= XFS_OWNER_INFO_ATTR_FORK;
-	if (rmap->rm_flags & XFS_RMAP_BMBT_BLOCK)
-		oinfo.oi_flags |= XFS_OWNER_INFO_BMBT_BLOCK;
+	xfs_owner_info_pack(&oinfo, rmap->rm_owner, rmap->rm_offset,
+			rmap->rm_flags);
 
-	if (rmap->rm_flags || XFS_RMAP_NON_INODE_OWNER(rmap->rm_owner))
+	if ((rmap->rm_flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK |
+			       XFS_RMAP_UNWRITTEN)) ||
+	    XFS_RMAP_NON_INODE_OWNER(rmap->rm_owner))
 		return xfs_rmap_map(cur, rmap->rm_startblock,
 				rmap->rm_blockcount,
 				rmap->rm_flags & XFS_RMAP_UNWRITTEN,
@@ -2442,7 +2453,7 @@ xfs_rmap_query_range_helper(
 
 	fa = xfs_rmap_btrec_to_irec(rec, &irec);
 	if (!fa)
-		fa = xfs_rmap_check_irec(cur->bc_ag.pag, &irec);
+		fa = xfs_rmap_check_btrec(cur, &irec);
 	if (fa)
 		return xfs_rmap_complain_bad_rec(cur, fa, &irec);
 

fs/xfs/libxfs/xfs_rmap_btree.c

@@ -22,6 +22,8 @@
 #include "xfs_extent_busy.h"
 #include "xfs_ag.h"
 #include "xfs_ag_resv.h"
+#include "xfs_buf_mem.h"
+#include "xfs_btree_mem.h"
 
 static struct kmem_cache	*xfs_rmapbt_cur_cache;
 
@@ -541,6 +543,151 @@ xfs_rmapbt_init_cursor(
 	return cur;
 }
 
+#ifdef CONFIG_XFS_BTREE_IN_MEM
+static inline unsigned int
+xfs_rmapbt_mem_block_maxrecs(
+	unsigned int		blocklen,
+	bool			leaf)
+{
+	if (leaf)
+		return blocklen / sizeof(struct xfs_rmap_rec);
+	return blocklen /
+		(2 * sizeof(struct xfs_rmap_key) + sizeof(__be64));
+}
+
+/*
+ * Validate an in-memory rmap btree block.  Callers are allowed to generate an
+ * in-memory btree even if the ondisk feature is not enabled.
+ */
+static xfs_failaddr_t
+xfs_rmapbt_mem_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
+	xfs_failaddr_t		fa;
+	unsigned int		level;
+	unsigned int		maxrecs;
+
+	if (!xfs_verify_magic(bp, block->bb_magic))
+		return __this_address;
+
+	fa = xfs_btree_fsblock_v5hdr_verify(bp, XFS_RMAP_OWN_UNKNOWN);
+	if (fa)
+		return fa;
+
+	level = be16_to_cpu(block->bb_level);
+	if (level >= xfs_rmapbt_maxlevels_ondisk())
+		return __this_address;
+
+	maxrecs = xfs_rmapbt_mem_block_maxrecs(
+			XFBNO_BLOCKSIZE - XFS_BTREE_LBLOCK_CRC_LEN, level == 0);
+	return xfs_btree_memblock_verify(bp, maxrecs);
+}
+
+static void
+xfs_rmapbt_mem_rw_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_failaddr_t	fa = xfs_rmapbt_mem_verify(bp);
+
+	if (fa)
+		xfs_verifier_error(bp, -EFSCORRUPTED, fa);
+}
+
+/* skip crc checks on in-memory btrees to save time */
+static const struct xfs_buf_ops xfs_rmapbt_mem_buf_ops = {
+	.name			= "xfs_rmapbt_mem",
+	.magic			= { 0, cpu_to_be32(XFS_RMAP_CRC_MAGIC) },
+	.verify_read		= xfs_rmapbt_mem_rw_verify,
+	.verify_write		= xfs_rmapbt_mem_rw_verify,
+	.verify_struct		= xfs_rmapbt_mem_verify,
+};
+
+const struct xfs_btree_ops xfs_rmapbt_mem_ops = {
+	.name			= "mem_rmap",
+	.type			= XFS_BTREE_TYPE_MEM,
+	.geom_flags		= XFS_BTGEO_OVERLAPPING,
+
+	.rec_len		= sizeof(struct xfs_rmap_rec),
+	/* Overlapping btree; 2 keys per pointer. */
+	.key_len		= 2 * sizeof(struct xfs_rmap_key),
+	.ptr_len		= XFS_BTREE_LONG_PTR_LEN,
+
+	.lru_refs		= XFS_RMAP_BTREE_REF,
+	.statoff		= XFS_STATS_CALC_INDEX(xs_rmap_mem_2),
+
+	.dup_cursor		= xfbtree_dup_cursor,
+	.set_root		= xfbtree_set_root,
+	.alloc_block		= xfbtree_alloc_block,
+	.free_block		= xfbtree_free_block,
+	.get_minrecs		= xfbtree_get_minrecs,
+	.get_maxrecs		= xfbtree_get_maxrecs,
+	.init_key_from_rec	= xfs_rmapbt_init_key_from_rec,
+	.init_high_key_from_rec	= xfs_rmapbt_init_high_key_from_rec,
+	.init_rec_from_cur	= xfs_rmapbt_init_rec_from_cur,
+	.init_ptr_from_cur	= xfbtree_init_ptr_from_cur,
+	.key_diff		= xfs_rmapbt_key_diff,
+	.buf_ops		= &xfs_rmapbt_mem_buf_ops,
+	.diff_two_keys		= xfs_rmapbt_diff_two_keys,
+	.keys_inorder		= xfs_rmapbt_keys_inorder,
+	.recs_inorder		= xfs_rmapbt_recs_inorder,
+	.keys_contiguous	= xfs_rmapbt_keys_contiguous,
+};
+
+/* Create a cursor for an in-memory btree. */
+struct xfs_btree_cur *
+xfs_rmapbt_mem_cursor(
+	struct xfs_perag	*pag,
+	struct xfs_trans	*tp,
+	struct xfbtree		*xfbt)
+{
+	struct xfs_btree_cur	*cur;
+	struct xfs_mount	*mp = pag->pag_mount;
+
+	cur = xfs_btree_alloc_cursor(mp, tp, &xfs_rmapbt_mem_ops,
+			xfs_rmapbt_maxlevels_ondisk(), xfs_rmapbt_cur_cache);
+	cur->bc_mem.xfbtree = xfbt;
+	cur->bc_nlevels = xfbt->nlevels;
+
+	cur->bc_mem.pag = xfs_perag_hold(pag);
+	return cur;
+}
+
+/* Create an in-memory rmap btree. */
+int
+xfs_rmapbt_mem_init(
+	struct xfs_mount	*mp,
+	struct xfbtree		*xfbt,
+	struct xfs_buftarg	*btp,
+	xfs_agnumber_t		agno)
+{
+	xfbt->owner = agno;
+	return xfbtree_init(mp, xfbt, btp, &xfs_rmapbt_mem_ops);
+}
+
+/* Compute the max possible height for reverse mapping btrees in memory. */
+static unsigned int
+xfs_rmapbt_mem_maxlevels(void)
+{
+	unsigned int		minrecs[2];
+	unsigned int		blocklen;
+
+	blocklen = XFBNO_BLOCKSIZE - XFS_BTREE_LBLOCK_CRC_LEN;
+
+	minrecs[0] = xfs_rmapbt_mem_block_maxrecs(blocklen, true) / 2;
+	minrecs[1] = xfs_rmapbt_mem_block_maxrecs(blocklen, false) / 2;
+
+	/*
+	 * How tall can an in-memory rmap btree become if we filled the entire
+	 * AG with rmap records?
+	 */
+	return xfs_btree_compute_maxlevels(minrecs,
+			XFS_MAX_AG_BYTES / sizeof(struct xfs_rmap_rec));
+}
+#else
+# define xfs_rmapbt_mem_maxlevels()	(0)
+#endif /* CONFIG_XFS_BTREE_IN_MEM */
+
 /*
  * Install a new reverse mapping btree root.  Caller is responsible for
  * invalidating and freeing the old btree blocks.
@@ -611,7 +758,8 @@ xfs_rmapbt_maxlevels_ondisk(void)
 	 * like if it consumes almost all the blocks in the AG due to maximal
 	 * sharing factor.
 	 */
-	return xfs_btree_space_to_height(minrecs, XFS_MAX_CRC_AG_BLOCKS);
+	return max(xfs_btree_space_to_height(minrecs, XFS_MAX_CRC_AG_BLOCKS),
+		   xfs_rmapbt_mem_maxlevels());
 }
 
 /* Compute the maximum height of an rmap btree. */

fs/xfs/libxfs/xfs_rmap_btree.h

@@ -10,6 +10,7 @@ struct xfs_buf;
 struct xfs_btree_cur;
 struct xfs_mount;
 struct xbtree_afakeroot;
+struct xfbtree;
 
 /* rmaps only exist on crc enabled filesystems */
 #define XFS_RMAP_BLOCK_LEN	XFS_BTREE_SBLOCK_CRC_LEN
@@ -62,4 +63,9 @@ unsigned int xfs_rmapbt_maxlevels_ondisk(void);
 int __init xfs_rmapbt_init_cur_cache(void);
 void xfs_rmapbt_destroy_cur_cache(void);
 
+struct xfs_btree_cur *xfs_rmapbt_mem_cursor(struct xfs_perag *pag,
+		struct xfs_trans *tp, struct xfbtree *xfbtree);
+int xfs_rmapbt_mem_init(struct xfs_mount *mp, struct xfbtree *xfbtree,
+		struct xfs_buftarg *btp, xfs_agnumber_t agno);
+
 #endif /* __XFS_RMAP_BTREE_H__ */

fs/xfs/libxfs/xfs_shared.h

@@ -51,6 +51,7 @@ extern const struct xfs_btree_ops xfs_finobt_ops;
 extern const struct xfs_btree_ops xfs_bmbt_ops;
 extern const struct xfs_btree_ops xfs_refcountbt_ops;
 extern const struct xfs_btree_ops xfs_rmapbt_ops;
+extern const struct xfs_btree_ops xfs_rmapbt_mem_ops;
 
 static inline bool xfs_btree_is_bno(const struct xfs_btree_ops *ops)
 {
@@ -87,6 +88,15 @@ static inline bool xfs_btree_is_rmap(const struct xfs_btree_ops *ops)
 	return ops == &xfs_rmapbt_ops;
 }
 
+#ifdef CONFIG_XFS_BTREE_IN_MEM
+static inline bool xfs_btree_is_mem_rmap(const struct xfs_btree_ops *ops)
+{
+	return ops == &xfs_rmapbt_mem_ops;
+}
+#else
+# define xfs_btree_is_mem_rmap(...)	(false)
+#endif
+
 /* log size calculation functions */
 int	xfs_log_calc_unit_res(struct xfs_mount *mp, int unit_bytes);
 int	xfs_log_calc_minimum_size(struct xfs_mount *);

fs/xfs/scrub/repair.c

@@ -31,12 +31,14 @@
 #include "xfs_error.h"
 #include "xfs_reflink.h"
 #include "xfs_health.h"
+#include "xfs_buf_mem.h"
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/trace.h"
 #include "scrub/repair.h"
 #include "scrub/bitmap.h"
 #include "scrub/stats.h"
+#include "scrub/xfile.h"
 
 /*
  * Attempt to repair some metadata, if the metadata is corrupt and userspace
@@ -1147,3 +1149,19 @@ xrep_metadata_inode_forks(
 
 	return 0;
 }
+
+/*
+ * Set up an in-memory buffer cache so that we can use the xfbtree.  Allocating
+ * a shmem file might take loks, so we cannot be in transaction context.  Park
+ * our resources in the scrub context and let the teardown function take care
+ * of them at the right time.
+ */
+int
+xrep_setup_xfbtree(
+	struct xfs_scrub	*sc,
+	const char		*descr)
+{
+	ASSERT(sc->tp == NULL);
+
+	return xmbuf_alloc(sc->mp, descr, &sc->xmbtp);
+}

fs/xfs/scrub/repair.h

@@ -81,6 +81,8 @@ int xrep_ino_dqattach(struct xfs_scrub *sc);
 # define xrep_ino_dqattach(sc)			(0)
 #endif /* CONFIG_XFS_QUOTA */
 
+int xrep_setup_xfbtree(struct xfs_scrub *sc, const char *descr);
+
 int xrep_ino_ensure_extent_count(struct xfs_scrub *sc, int whichfork,
 		xfs_extnum_t nextents);
 int xrep_reset_perag_resv(struct xfs_scrub *sc);

fs/xfs/xfs_stats.c

@@ -50,7 +50,8 @@ int xfs_stats_format(struct xfsstats __percpu *stats, char *buf)
 		{ "ibt2",		xfsstats_offset(xs_fibt_2)	},
 		{ "fibt2",		xfsstats_offset(xs_rmap_2)	},
 		{ "rmapbt",		xfsstats_offset(xs_refcbt_2)	},
-		{ "refcntbt",		xfsstats_offset(xs_qm_dqreclaims)},
+		{ "refcntbt",		xfsstats_offset(xs_rmap_mem_2)	},
+		{ "rmapbt_mem",		xfsstats_offset(xs_qm_dqreclaims)},
 		/* we print both series of quota information together */
 		{ "qm",			xfsstats_offset(xs_xstrat_bytes)},
 	};

fs/xfs/xfs_stats.h

@@ -125,6 +125,7 @@ struct __xfsstats {
 	uint32_t		xs_fibt_2[__XBTS_MAX];
 	uint32_t		xs_rmap_2[__XBTS_MAX];
 	uint32_t		xs_refcbt_2[__XBTS_MAX];
+	uint32_t		xs_rmap_mem_2[__XBTS_MAX];
 	uint32_t		xs_qm_dqreclaims;
 	uint32_t		xs_qm_dqreclaim_misses;
 	uint32_t		xs_qm_dquot_dups;