xfs: add support for large btree blocks

Add support for larger btree blocks that contains a CRC32C checksum,
a filesystem uuid and block number for detecting filesystem
consistency and out of place writes.

[dchinner@redhat.com] Also include an owner field to allow reverse
mappings to be implemented for improved repairability and a LSN
field to so that log recovery can easily determine the last
modification that made it to disk for each buffer.

[dchinner@redhat.com] Add buffer log format flags to indicate the
type of buffer to recovery so that we don't have to do blind magic
number tests to determine what the buffer is.

[dchinner@redhat.com] Modified to fit into the verifier structure.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>

fs/xfs/xfs_alloc_btree.c

@@ -33,6 +33,7 @@
 #include "xfs_extent_busy.h"
 #include "xfs_error.h"
 #include "xfs_trace.h"
+#include "xfs_cksum.h"
 
 
 STATIC struct xfs_btree_cur *
@@ -272,74 +273,111 @@ xfs_allocbt_key_diff(
 	return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
 }
 
-static void
+static bool
 xfs_allocbt_verify(
 	struct xfs_buf		*bp)
 {
 	struct xfs_mount	*mp = bp->b_target->bt_mount;
 	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
 	struct xfs_perag	*pag = bp->b_pag;
 	unsigned int		level;
-	int			sblock_ok; /* block passes checks */
 
 	/*
 	 * magic number and level verification
 	 *
-	 * During growfs operations, we can't verify the exact level as the
-	 * perag is not fully initialised and hence not attached to the buffer.
-	 * In this case, check against the maximum tree depth.
+	 * During growfs operations, we can't verify the exact level or owner as
+	 * the perag is not fully initialised and hence not attached to the
+	 * buffer.  In this case, check against the maximum tree depth.
+	 *
+	 * Similarly, during log recovery we will have a perag structure
+	 * attached, but the agf information will not yet have been initialised
+	 * from the on disk AGF. Again, we can only check against maximum limits
+	 * in this case.
 	 */
 	level = be16_to_cpu(block->bb_level);
 	switch (block->bb_magic) {
+	case cpu_to_be32(XFS_ABTB_CRC_MAGIC):
+		if (!xfs_sb_version_hascrc(&mp->m_sb))
+			return false;
+		if (!uuid_equal(&block->bb_u.s.bb_uuid, &mp->m_sb.sb_uuid))
+			return false;
+		if (block->bb_u.s.bb_blkno != cpu_to_be64(bp->b_bn))
+			return false;
+		if (pag &&
+		    be32_to_cpu(block->bb_u.s.bb_owner) != pag->pag_agno)
+			return false;
+		/* fall through */
 	case cpu_to_be32(XFS_ABTB_MAGIC):
-		if (pag)
-			sblock_ok = level < pag->pagf_levels[XFS_BTNUM_BNOi];
-		else
-			sblock_ok = level < mp->m_ag_maxlevels;
+		if (pag && pag->pagf_init) {
+			if (level >= pag->pagf_levels[XFS_BTNUM_BNOi])
+				return false;
+		} else if (level >= mp->m_ag_maxlevels)
+			return false;
 		break;
+	case cpu_to_be32(XFS_ABTC_CRC_MAGIC):
+		if (!xfs_sb_version_hascrc(&mp->m_sb))
+			return false;
+		if (!uuid_equal(&block->bb_u.s.bb_uuid, &mp->m_sb.sb_uuid))
+			return false;
+		if (block->bb_u.s.bb_blkno != cpu_to_be64(bp->b_bn))
+			return false;
+		if (pag &&
+		    be32_to_cpu(block->bb_u.s.bb_owner) != pag->pag_agno)
+			return false;
+		/* fall through */
 	case cpu_to_be32(XFS_ABTC_MAGIC):
-		if (pag)
-			sblock_ok = level < pag->pagf_levels[XFS_BTNUM_CNTi];
-		else
-			sblock_ok = level < mp->m_ag_maxlevels;
+		if (pag && pag->pagf_init) {
+			if (level >= pag->pagf_levels[XFS_BTNUM_CNTi])
+				return false;
+		} else if (level >= mp->m_ag_maxlevels)
+			return false;
 		break;
 	default:
-		sblock_ok = 0;
-		break;
+		return false;
 	}
 
 	/* numrecs verification */
-	sblock_ok = sblock_ok &&
-		be16_to_cpu(block->bb_numrecs) <= mp->m_alloc_mxr[level != 0];
+	if (be16_to_cpu(block->bb_numrecs) > mp->m_alloc_mxr[level != 0])
+		return false;
 
 	/* sibling pointer verification */
-	sblock_ok = sblock_ok &&
-		(block->bb_u.s.bb_leftsib == cpu_to_be32(NULLAGBLOCK) ||
-		 be32_to_cpu(block->bb_u.s.bb_leftsib) < mp->m_sb.sb_agblocks) &&
-		block->bb_u.s.bb_leftsib &&
-		(block->bb_u.s.bb_rightsib == cpu_to_be32(NULLAGBLOCK) ||
-		 be32_to_cpu(block->bb_u.s.bb_rightsib) < mp->m_sb.sb_agblocks) &&
-		block->bb_u.s.bb_rightsib;
-
-	if (!sblock_ok) {
-		trace_xfs_btree_corrupt(bp, _RET_IP_);
-		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, block);
-		xfs_buf_ioerror(bp, EFSCORRUPTED);
-	}
+	if (!block->bb_u.s.bb_leftsib ||
+	    (be32_to_cpu(block->bb_u.s.bb_leftsib) >= mp->m_sb.sb_agblocks &&
+	     block->bb_u.s.bb_leftsib != cpu_to_be32(NULLAGBLOCK)))
+		return false;
+	if (!block->bb_u.s.bb_rightsib ||
+	    (be32_to_cpu(block->bb_u.s.bb_rightsib) >= mp->m_sb.sb_agblocks &&
+	     block->bb_u.s.bb_rightsib != cpu_to_be32(NULLAGBLOCK)))
+		return false;
+
+	return true;
 }
 
 static void
 xfs_allocbt_read_verify(
 	struct xfs_buf	*bp)
 {
-	xfs_allocbt_verify(bp);
+	if (!(xfs_btree_sblock_verify_crc(bp) &&
+	      xfs_allocbt_verify(bp))) {
+		trace_xfs_btree_corrupt(bp, _RET_IP_);
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW,
+				     bp->b_target->bt_mount, bp->b_addr);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
 }
 
 static void
 xfs_allocbt_write_verify(
 	struct xfs_buf	*bp)
 {
-	xfs_allocbt_verify(bp);
+	if (!xfs_allocbt_verify(bp)) {
+		trace_xfs_btree_corrupt(bp, _RET_IP_);
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW,
+				     bp->b_target->bt_mount, bp->b_addr);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+	xfs_btree_sblock_calc_crc(bp);
+
 }
 
 const struct xfs_buf_ops xfs_allocbt_buf_ops = {
@@ -444,6 +482,9 @@ xfs_allocbt_init_cursor(
 	cur->bc_private.a.agbp = agbp;
 	cur->bc_private.a.agno = agno;
 
+	if (xfs_sb_version_hascrc(&mp->m_sb))
+		cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
+
 	return cur;
 }
 

fs/xfs/xfs_alloc_btree.h

@@ -31,8 +31,10 @@ struct xfs_mount;
  * by blockcount and blockno.  All blocks look the same to make the code
  * simpler; if we have time later, we'll make the optimizations.
  */
-#define	XFS_ABTB_MAGIC	0x41425442	/* 'ABTB' for bno tree */
-#define	XFS_ABTC_MAGIC	0x41425443	/* 'ABTC' for cnt tree */
+#define	XFS_ABTB_MAGIC		0x41425442	/* 'ABTB' for bno tree */
+#define	XFS_ABTB_CRC_MAGIC	0x41423342	/* 'AB3B' */
+#define	XFS_ABTC_MAGIC		0x41425443	/* 'ABTC' for cnt tree */
+#define	XFS_ABTC_CRC_MAGIC	0x41423343	/* 'AB3C' */
 
 /*
  * Data record/key structure
@@ -59,10 +61,10 @@ typedef __be32 xfs_alloc_ptr_t;
 
 /*
  * Btree block header size depends on a superblock flag.
- *
- * (not quite yet, but soon)
  */
-#define XFS_ALLOC_BLOCK_LEN(mp)	XFS_BTREE_SBLOCK_LEN
+#define XFS_ALLOC_BLOCK_LEN(mp) \
+	(xfs_sb_version_hascrc(&((mp)->m_sb)) ? \
+		XFS_BTREE_SBLOCK_CRC_LEN : XFS_BTREE_SBLOCK_LEN)
 
 /*
  * Record, key, and pointer address macros for btree blocks.

fs/xfs/xfs_attr_leaf.c

@@ -232,7 +232,7 @@ xfs_attr_shortform_bytesfit(xfs_inode_t *dp, int bytes)
 				return 0;
 			return dp->i_d.di_forkoff;
 		}
-		dsize = XFS_BMAP_BROOT_SPACE(dp->i_df.if_broot);
+		dsize = XFS_BMAP_BROOT_SPACE(mp, dp->i_df.if_broot);
 		break;
 	}
 

fs/xfs/xfs_bmap.c

@@ -439,11 +439,15 @@ xfs_bmap_sanity_check(
 {
 	struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
 
-	if (block->bb_magic != cpu_to_be32(XFS_BMAP_MAGIC) ||
-	    be16_to_cpu(block->bb_level) != level ||
+	if (block->bb_magic != cpu_to_be32(XFS_BMAP_CRC_MAGIC) &&
+	    block->bb_magic != cpu_to_be32(XFS_BMAP_MAGIC))
+		return 0;
+
+	if (be16_to_cpu(block->bb_level) != level ||
 	    be16_to_cpu(block->bb_numrecs) == 0 ||
 	    be16_to_cpu(block->bb_numrecs) > mp->m_bmap_dmxr[level != 0])
 		return 0;
+
 	return 1;
 }
 
@@ -1031,6 +1035,7 @@ xfs_bmap_extents_to_btree(
 	xfs_extnum_t		nextents;	/* number of file extents */
 	xfs_bmbt_ptr_t		*pp;		/* root block address pointer */
 
+	mp = ip->i_mount;
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS);
 
@@ -1044,16 +1049,18 @@ xfs_bmap_extents_to_btree(
 	 * Fill in the root.
 	 */
 	block = ifp->if_broot;
-	block->bb_magic = cpu_to_be32(XFS_BMAP_MAGIC);
-	block->bb_level = cpu_to_be16(1);
-	block->bb_numrecs = cpu_to_be16(1);
-	block->bb_u.l.bb_leftsib = cpu_to_be64(NULLDFSBNO);
-	block->bb_u.l.bb_rightsib = cpu_to_be64(NULLDFSBNO);
+	if (xfs_sb_version_hascrc(&mp->m_sb))
+		xfs_btree_init_block_int(mp, block, XFS_BUF_DADDR_NULL,
+				 XFS_BMAP_CRC_MAGIC, 1, 1, ip->i_ino,
+				 XFS_BTREE_LONG_PTRS | XFS_BTREE_CRC_BLOCKS);
+	else
+		xfs_btree_init_block_int(mp, block, XFS_BUF_DADDR_NULL,
+				 XFS_BMAP_MAGIC, 1, 1, ip->i_ino,
+				 XFS_BTREE_LONG_PTRS);
 
 	/*
 	 * Need a cursor.  Can't allocate until bb_level is filled in.
 	 */
-	mp = ip->i_mount;
 	cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
 	cur->bc_private.b.firstblock = *firstblock;
 	cur->bc_private.b.flist = flist;
@@ -1102,10 +1109,15 @@ xfs_bmap_extents_to_btree(
 	 */
 	abp->b_ops = &xfs_bmbt_buf_ops;
 	ablock = XFS_BUF_TO_BLOCK(abp);
-	ablock->bb_magic = cpu_to_be32(XFS_BMAP_MAGIC);
-	ablock->bb_level = 0;
-	ablock->bb_u.l.bb_leftsib = cpu_to_be64(NULLDFSBNO);
-	ablock->bb_u.l.bb_rightsib = cpu_to_be64(NULLDFSBNO);
+	if (xfs_sb_version_hascrc(&mp->m_sb))
+		xfs_btree_init_block_int(mp, ablock, abp->b_bn,
+				XFS_BMAP_CRC_MAGIC, 0, 0, ip->i_ino,
+				XFS_BTREE_LONG_PTRS | XFS_BTREE_CRC_BLOCKS);
+	else
+		xfs_btree_init_block_int(mp, ablock, abp->b_bn,
+				XFS_BMAP_MAGIC, 0, 0, ip->i_ino,
+				XFS_BTREE_LONG_PTRS);
+
 	arp = XFS_BMBT_REC_ADDR(mp, ablock, 1);
 	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
 	for (cnt = i = 0; i < nextents; i++) {
@@ -1155,7 +1167,8 @@ xfs_bmap_local_to_extents(
 	xfs_extlen_t	total,		/* total blocks needed by transaction */
 	int		*logflagsp,	/* inode logging flags */
 	int		whichfork,
-	void		(*init_fn)(struct xfs_buf *bp,
+	void		(*init_fn)(struct xfs_trans *tp,
+				   struct xfs_buf *bp,
 				   struct xfs_inode *ip,
 				   struct xfs_ifork *ifp))
 {
@@ -1207,7 +1220,7 @@ xfs_bmap_local_to_extents(
 		bp = xfs_btree_get_bufl(args.mp, tp, args.fsbno, 0);
 
 		/* initialise the block and copy the data */
-		init_fn(bp, ip, ifp);
+		init_fn(tp, bp, ip, ifp);
 
 		/* account for the change in fork size and log everything */
 		xfs_trans_log_buf(tp, bp, 0, ifp->if_bytes - 1);
@@ -1314,16 +1327,19 @@ xfs_bmap_add_attrfork_extents(
  */
 STATIC void
 xfs_bmap_local_to_extents_init_fn(
+	struct xfs_trans	*tp,
 	struct xfs_buf		*bp,
 	struct xfs_inode	*ip,
 	struct xfs_ifork	*ifp)
 {
 	bp->b_ops = &xfs_bmbt_buf_ops;
 	memcpy(bp->b_addr, ifp->if_u1.if_data, ifp->if_bytes);
+	xfs_trans_buf_set_type(tp, bp, XFS_BLF_BTREE_BUF);
 }
 
 STATIC void
 xfs_symlink_local_to_remote(
+	struct xfs_trans	*tp,
 	struct xfs_buf		*bp,
 	struct xfs_inode	*ip,
 	struct xfs_ifork	*ifp)
@@ -1342,8 +1358,7 @@ xfs_symlink_local_to_remote(
  *
  * XXX (dgc): investigate whether directory conversion can use the generic
  * formatting callout. It should be possible - it's just a very complex
- * formatter. it would also require passing the transaction through to the init
- * function.
+ * formatter.
  */
 STATIC int					/* error */
 xfs_bmap_add_attrfork_local(