xfs: miscellaneous improvements for 4.10

index 5ba2dac5e67c..44cfcd03c451 100644
@@ -428,6 +428,10 @@ xfs_allocbt_init_cursor(
 	cur->bc_btnum = btnum;
 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
 	cur->bc_ops = &xfs_allocbt_ops;
+	if (btnum == XFS_BTNUM_BNO)
+		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtb_2);
+	else
+		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtc_2);
 
 	if (btnum == XFS_BTNUM_CNT) {
 		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]);

index 8007d2ba9aef..94ad31d372ab 100644
@@ -803,6 +803,7 @@ xfs_bmbt_init_cursor(
 	cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
 	cur->bc_btnum = XFS_BTNUM_BMAP;
 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
+	cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_bmbt_2);
 
 	cur->bc_ops = &xfs_bmbt_ops;
 	cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;

56 lines not on this page · raw message

index eab68ae2e011..e7ff8ef0e5a7 100644
@@ -365,9 +365,11 @@ xfs_inobt_init_cursor(
 	if (btnum == XFS_BTNUM_INO) {
 		cur->bc_nlevels = be32_to_cpu(agi->agi_level);
 		cur->bc_ops = &xfs_inobt_ops;
+		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_ibt_2);
 	} else {
 		cur->bc_nlevels = be32_to_cpu(agi->agi_free_level);
 		cur->bc_ops = &xfs_finobt_ops;
+		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_fibt_2);
 	}
 
 	cur->bc_blocklog = mp->m_sb.sb_blocklog;

index 453bb2757ec2..6fb2215f8ff7 100644
@@ -354,6 +354,7 @@ xfs_refcountbt_init_cursor(
 	cur->bc_btnum = XFS_BTNUM_REFC;
 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
 	cur->bc_ops = &xfs_refcountbt_ops;
+	cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_refcbt_2);
 
 	cur->bc_nlevels = be32_to_cpu(agf->agf_refcount_level);
 

index 83e672ff7577..de25771764ba 100644
@@ -484,6 +484,7 @@ xfs_rmapbt_init_cursor(
 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
 	cur->bc_ops = &xfs_rmapbt_ops;
 	cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]);
+	cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_rmap_2);
 
 	cur->bc_private.a.agbp = agbp;
 	cur->bc_private.a.agno = agno;

index 12d48cd8f8a4..f11282c96887 100644
@@ -80,9 +80,9 @@ int xfs_stats_format(struct xfsstats __percpu *stats, char *buf)
 	}
 	/* extra precision counters */
 	for_each_possible_cpu(i) {
-		xs_xstrat_bytes += per_cpu_ptr(stats, i)->xs_xstrat_bytes;
-		xs_write_bytes += per_cpu_ptr(stats, i)->xs_write_bytes;
-		xs_read_bytes += per_cpu_ptr(stats, i)->xs_read_bytes;
+		xs_xstrat_bytes += per_cpu_ptr(stats, i)->s.xs_xstrat_bytes;
+		xs_write_bytes += per_cpu_ptr(stats, i)->s.xs_write_bytes;
+		xs_read_bytes += per_cpu_ptr(stats, i)->s.xs_read_bytes;
 	}
 
 	len += snprintf(buf + len, PATH_MAX-len, "xpc %Lu %Lu %Lu\n",

@@ -106,9 +106,9 @@ void xfs_stats_clearall(struct xfsstats __percpu *stats)
 	for_each_possible_cpu(c) {
 		preempt_disable();
 		/* save vn_active, it's a universal truth! */
-		vn_active = per_cpu_ptr(stats, c)->vn_active;
+		vn_active = per_cpu_ptr(stats, c)->s.vn_active;
 		memset(per_cpu_ptr(stats, c), 0, sizeof(*stats));
-		per_cpu_ptr(stats, c)->vn_active = vn_active;
+		per_cpu_ptr(stats, c)->s.vn_active = vn_active;
 		preempt_enable();
 	}
 }

227 lines not on this page · raw message

index b5b9bffe3520..a2f0648743db 100644
@@ -219,7 +219,6 @@ _xfs_buf_alloc(
 	init_completion(&bp->b_iowait);
 	INIT_LIST_HEAD(&bp->b_lru);
 	INIT_LIST_HEAD(&bp->b_list);
-	RB_CLEAR_NODE(&bp->b_rbnode);
 	sema_init(&bp->b_sema, 0); /* held, no waiters */
 	spin_lock_init(&bp->b_lock);
 	XB_SET_OWNER(bp);

@@ -473,6 +472,63 @@ _xfs_buf_map_pages(
 /*
  *	Finding and Reading Buffers
  */
+static int
+_xfs_buf_cmp(
+	struct rhashtable_compare_arg *arg,
+	const void *obj)
+{
+	const struct xfs_buf_map *map = arg->key;
+	const struct xfs_buf *bp = obj;
+
+	/*
+	 * The key hashing in the lookup path depends on the key being the
+	 * first element of the compare_arg, make sure to assert this.
+	 */
+	BUILD_BUG_ON(offsetof(struct xfs_buf_map, bm_bn) != 0);
+
+	if (bp->b_bn == map->bm_bn) {
+		if (unlikely(bp->b_length != map->bm_len)) {
+			/*
+			 * found a block number match. If the range doesn't
+			 * match, the only way this is allowed is if the buffer
+			 * in the cache is stale and the transaction that made
+			 * it stale has not yet committed. i.e. we are
+			 * reallocating a busy extent. Skip this buffer and
+			 * continue searching for an exact match.
+			 */
+			ASSERT(bp->b_flags & XBF_STALE);
+			return 1;
+		}
+
+		return 0;
+	}
+
+	return 1;
+}
+
+static const struct rhashtable_params xfs_buf_hash_params = {
+	.min_size = 32,		/* empty/unused AGs have minimal footprint */
+	.nelem_hint = 16,
+	.key_len = sizeof(xfs_daddr_t),
+	.key_offset = offsetof(struct xfs_buf, b_bn),
+	.head_offset = offsetof(struct xfs_buf, b_rhash_head),
+	.automatic_shrinking = true,
+	.obj_cmpfn = _xfs_buf_cmp,
+};
+
+int xfs_buf_hash_init(
+	struct xfs_perag *pag)
+{
+	spin_lock_init(&pag->pag_buf_lock);
+	return rhashtable_init(&pag->pag_buf_hash, &xfs_buf_hash_params);
+}
+
+void
+xfs_buf_hash_destroy(
+	struct xfs_perag *pag)
+{
+	rhashtable_destroy(&pag->pag_buf_hash);
+}
 
 /*
  *	Look up, and creates if absent, a lockable buffer for

@@ -488,27 +544,24 @@ _xfs_buf_find(
 	xfs_buf_t		*new_bp)
 {
 	struct xfs_perag	*pag;
-	struct rb_node		**rbp;
-	struct rb_node		*parent;
 	xfs_buf_t		*bp;
-	xfs_daddr_t		blkno = map[0].bm_bn;
+	struct xfs_buf_map	cmap = { .bm_bn = map[0].bm_bn };
 	xfs_daddr_t		eofs;
-	int			numblks = 0;
 	int			i;
 
 	for (i = 0; i < nmaps; i++)
-		numblks += map[i].bm_len;
+		cmap.bm_len += map[i].bm_len;
 
 	/* Check for IOs smaller than the sector size / not sector aligned */
-	ASSERT(!(BBTOB(numblks) < btp->bt_meta_sectorsize));
-	ASSERT(!(BBTOB(blkno) & (xfs_off_t)btp->bt_meta_sectormask));
+	ASSERT(!(BBTOB(cmap.bm_len) < btp->bt_meta_sectorsize));
+	ASSERT(!(BBTOB(cmap.bm_bn) & (xfs_off_t)btp->bt_meta_sectormask));
 
 	/*
 	 * Corrupted block numbers can get through to here, unfortunately, so we
 	 * have to check that the buffer falls within the filesystem bounds.
 	 */
 	eofs = XFS_FSB_TO_BB(btp->bt_mount, btp->bt_mount->m_sb.sb_dblocks);
-	if (blkno < 0 || blkno >= eofs) {
+	if (cmap.bm_bn < 0 || cmap.bm_bn >= eofs) {
 		/*
 		 * XXX (dgc): we should really be returning -EFSCORRUPTED here,
 		 * but none of the higher level infrastructure supports

@@ -516,53 +569,29 @@ _xfs_buf_find(
 		 */
 		xfs_alert(btp->bt_mount,
 			  "%s: Block out of range: block 0x%llx, EOFS 0x%llx ",
-			  __func__, blkno, eofs);
+			  __func__, cmap.bm_bn, eofs);
 		WARN_ON(1);
 		return NULL;
 	}
 
-	/* get tree root */
 	pag = xfs_perag_get(btp->bt_mount,
-				xfs_daddr_to_agno(btp->bt_mount, blkno));
+			    xfs_daddr_to_agno(btp->bt_mount, cmap.bm_bn));
 
-	/* walk tree */
 	spin_lock(&pag->pag_buf_lock);
-	rbp = &pag->pag_buf_tree.rb_node;
-	parent = NULL;
-	bp = NULL;
-	while (*rbp) {
-		parent = *rbp;
-		bp = rb_entry(parent, struct xfs_buf, b_rbnode);
-
-		if (blkno < bp->b_bn)
-			rbp = &(*rbp)->rb_left;
-		else if (blkno > bp->b_bn)
-			rbp = &(*rbp)->rb_right;
-		else {
-			/*
-			 * found a block number match. If the range doesn't
-			 * match, the only way this is allowed is if the buffer
-			 * in the cache is stale and the transaction that made
-			 * it stale has not yet committed. i.e. we are
-			 * reallocating a busy extent. Skip this buffer and
-			 * continue searching to the right for an exact match.
-			 */
-			if (bp->b_length != numblks) {
-				ASSERT(bp->b_flags & XBF_STALE);
-				rbp = &(*rbp)->rb_right;
-				continue;
-			}
-			atomic_inc(&bp->b_hold);
-			goto found;
-		}
+	bp = rhashtable_lookup_fast(&pag->pag_buf_hash, &cmap,
+				    xfs_buf_hash_params);
+	if (bp) {
+		atomic_inc(&bp->b_hold);
+		goto found;
 	}
 
 	/* No match found */
 	if (new_bp) {
-		rb_link_node(&new_bp->b_rbnode, parent, rbp);
-		rb_insert_color(&new_bp->b_rbnode, &pag->pag_buf_tree);
 		/* the buffer keeps the perag reference until it is freed */
 		new_bp->b_pag = pag;
+		rhashtable_insert_fast(&pag->pag_buf_hash,
+				       &new_bp->b_rhash_head,
+				       xfs_buf_hash_params);
 		spin_unlock(&pag->pag_buf_lock);
 	} else {
 		XFS_STATS_INC(btp->bt_mount, xb_miss_locked);

@@ -930,7 +959,6 @@ xfs_buf_rele(
 
 	if (!pag) {
 		ASSERT(list_empty(&bp->b_lru));
-		ASSERT(RB_EMPTY_NODE(&bp->b_rbnode));
 		if (atomic_dec_and_test(&bp->b_hold)) {
 			xfs_buf_ioacct_dec(bp);
 			xfs_buf_free(bp);

@@ -938,8 +966,6 @@ xfs_buf_rele(
 		return;
 	}
 
-	ASSERT(!RB_EMPTY_NODE(&bp->b_rbnode));
-
 	ASSERT(atomic_read(&bp->b_hold) > 0);
 
 	release = atomic_dec_and_lock(&bp->b_hold, &pag->pag_buf_lock);

@@ -983,7 +1009,8 @@ xfs_buf_rele(
 		}
 
 		ASSERT(!(bp->b_flags & _XBF_DELWRI_Q));
-		rb_erase(&bp->b_rbnode, &pag->pag_buf_tree);
+		rhashtable_remove_fast(&pag->pag_buf_hash, &bp->b_rhash_head,
+				       xfs_buf_hash_params);
 		spin_unlock(&pag->pag_buf_lock);
 		xfs_perag_put(pag);
 		freebuf = true;

index 38c9a95bda7d..8a9d3a9599f0 100644
@@ -151,7 +151,7 @@ typedef struct xfs_buf {
 	 * which is the only bit that is touched if we hit the semaphore
 	 * fast-path on locking.
 	 */
-	struct rb_node		b_rbnode;	/* rbtree node */
+	struct rhash_head	b_rhash_head;	/* pag buffer hash node */
 	xfs_daddr_t		b_bn;		/* block number of buffer */
 	int			b_length;	/* size of buffer in BBs */
 	atomic_t		b_hold;		/* reference count */

index 68640fb63a54..a415f822f2c1 100644
@@ -78,6 +78,7 @@ typedef __u32			xfs_nlink_t;
 #include <linux/freezer.h>
 #include <linux/list_sort.h>
 #include <linux/ratelimit.h>
+#include <linux/rhashtable.h>
 
 #include <asm/page.h>
 #include <asm/div64.h>

index b341f10cf481..9b9540db17a6 100644
@@ -157,6 +157,7 @@ xfs_free_perag(
 		spin_unlock(&mp->m_perag_lock);
 		ASSERT(pag);
 		ASSERT(atomic_read(&pag->pag_ref) == 0);
+		xfs_buf_hash_destroy(pag);
 		call_rcu(&pag->rcu_head, __xfs_free_perag);
 	}
 }

@@ -212,8 +213,8 @@ xfs_initialize_perag(
 		spin_lock_init(&pag->pag_ici_lock);
 		mutex_init(&pag->pag_ici_reclaim_lock);
 		INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC);
-		spin_lock_init(&pag->pag_buf_lock);
-		pag->pag_buf_tree = RB_ROOT;
+		if (xfs_buf_hash_init(pag))
+			goto out_unwind;
 
 		if (radix_tree_preload(GFP_NOFS))
 			goto out_unwind;

@@ -239,9 +240,11 @@ xfs_initialize_perag(
 	return 0;
 
 out_unwind:
+	xfs_buf_hash_destroy(pag);
 	kmem_free(pag);
 	for (; index > first_initialised; index--) {
 		pag = radix_tree_delete(&mp->m_perag_tree, index);
+		xfs_buf_hash_destroy(pag);
 		kmem_free(pag);
 	}
 	return error;

index 819b80b15bfb..84f785218907 100644
@@ -393,8 +393,8 @@ typedef struct xfs_perag {
 	unsigned long	pag_ici_reclaim_cursor;	/* reclaim restart point */
 
 	/* buffer cache index */
-	spinlock_t	pag_buf_lock;	/* lock for pag_buf_tree */
-	struct rb_root	pag_buf_tree;	/* ordered tree of active buffers */
+	spinlock_t	pag_buf_lock;	/* lock for pag_buf_hash */
+	struct rhashtable pag_buf_hash;
 
 	/* for rcu-safe freeing */
 	struct rcu_head	rcu_head;

@@ -424,6 +424,9 @@ xfs_perag_resv(
 	}
 }
 
+int xfs_buf_hash_init(xfs_perag_t *pag);
+void xfs_buf_hash_destroy(xfs_perag_t *pag);
+
 extern void	xfs_uuid_table_free(void);
 extern int	xfs_log_sbcount(xfs_mount_t *);
 extern __uint64_t xfs_default_resblks(xfs_mount_t *mp);

index fad1676ad8cd..a416c7cb23ea 100644
@@ -6,10 +6,11 @@
 /*
  * Calculate the intermediate checksum for a buffer that has the CRC field
  * inside it.  The offset of the 32bit crc fields is passed as the
- * cksum_offset parameter.
+ * cksum_offset parameter. We do not modify the buffer during verification,
+ * hence we have to split the CRC calculation across the cksum_offset.
  */
 static inline __uint32_t
-xfs_start_cksum(char *buffer, size_t length, unsigned long cksum_offset)
+xfs_start_cksum_safe(char *buffer, size_t length, unsigned long cksum_offset)
 {
 	__uint32_t zero = 0;
 	__uint32_t crc;

@@ -26,6 +27,20 @@ xfs_start_cksum(char *buffer, size_t length, unsigned long cksum_offset)
 }
 
 /*
+ * Fast CRC method where the buffer is modified. Callers must have exclusive
+ * access to the buffer while the calculation takes place.
+ */
+static inline __uint32_t
+xfs_start_cksum_update(char *buffer, size_t length, unsigned long cksum_offset)
+{
+	/* zero the CRC field */
+	*(__le32 *)(buffer + cksum_offset) = 0;
+
+	/* single pass CRC calculation for the entire buffer */
+	return crc32c(XFS_CRC_SEED, buffer, length);
+}
+
+/*
  * Convert the intermediate checksum to the final ondisk format.
  *
  * The CRC32c calculation uses LE format even on BE machines, but returns the

@@ -40,11 +55,14 @@ xfs_end_cksum(__uint32_t crc)
 
 /*
  * Helper to generate the checksum for a buffer.
+ *
+ * This modifies the buffer temporarily - callers must have exclusive
+ * access to the buffer while the calculation takes place.
  */
 static inline void
 xfs_update_cksum(char *buffer, size_t length, unsigned long cksum_offset)
 {
-	__uint32_t crc = xfs_start_cksum(buffer, length, cksum_offset);
+	__uint32_t crc = xfs_start_cksum_update(buffer, length, cksum_offset);
 
 	*(__le32 *)(buffer + cksum_offset) = xfs_end_cksum(crc);
 }

@@ -55,7 +73,7 @@ xfs_update_cksum(char *buffer, size_t length, unsigned long cksum_offset)
 static inline int
 xfs_verify_cksum(char *buffer, size_t length, unsigned long cksum_offset)
 {
-	__uint32_t crc = xfs_start_cksum(buffer, length, cksum_offset);
+	__uint32_t crc = xfs_start_cksum_safe(buffer, length, cksum_offset);
 
 	return *(__le32 *)(buffer + cksum_offset) == xfs_end_cksum(crc);
 }

index 54817f82212c..5efeb4233387 100644
@@ -436,7 +436,7 @@ xfs_dinode_calc_crc(
 		return;
 
 	ASSERT(xfs_sb_version_hascrc(&mp->m_sb));
-	crc = xfs_start_cksum((char *)dip, mp->m_sb.sb_inodesize,
+	crc = xfs_start_cksum_update((char *)dip, mp->m_sb.sb_inodesize,
 			      XFS_DINODE_CRC_OFF);
 	dip->di_crc = xfs_end_cksum(crc);
 }

index 3b74fa011bb1..3ebe444eb60f 100644
@@ -1668,7 +1668,7 @@ xlog_cksum(
 	__uint32_t		crc;
 
 	/* first generate the crc for the record header ... */
-	crc = xfs_start_cksum((char *)rhead,
+	crc = xfs_start_cksum_update((char *)rhead,
 			      sizeof(struct xlog_rec_header),
 			      offsetof(struct xlog_rec_header, h_crc));
 

index cf754bcbcb1c..4a98762ec8b4 100644
@@ -5113,19 +5113,21 @@ xlog_recover_process(
 	struct list_head	*buffer_list)
 {
 	int			error;
+	__le32			old_crc = rhead->h_crc;
 	__le32			crc;
 
+
 	crc = xlog_cksum(log, rhead, dp, be32_to_cpu(rhead->h_len));
 
 	/*
 	 * Nothing else to do if this is a CRC verification pass. Just return
 	 * if this a record with a non-zero crc. Unfortunately, mkfs always
-	 * sets h_crc to 0 so we must consider this valid even on v5 supers.
+	 * sets old_crc to 0 so we must consider this valid even on v5 supers.
 	 * Otherwise, return EFSBADCRC on failure so the callers up the stack
 	 * know precisely what failed.
 	 */
 	if (pass == XLOG_RECOVER_CRCPASS) {
-		if (rhead->h_crc && crc != rhead->h_crc)
+		if (old_crc && crc != old_crc)
 			return -EFSBADCRC;
 		return 0;
 	}

@@ -5136,11 +5138,11 @@ xlog_recover_process(
 	 * zero CRC check prevents warnings from being emitted when upgrading
 	 * the kernel from one that does not add CRCs by default.
 	 */
-	if (crc != rhead->h_crc) {
-		if (rhead->h_crc || xfs_sb_version_hascrc(&log->l_mp->m_sb)) {
+	if (crc != old_crc) {
+		if (old_crc || xfs_sb_version_hascrc(&log->l_mp->m_sb)) {
 			xfs_alert(log->l_mp,
 		"log record CRC mismatch: found 0x%x, expected 0x%x.",
-					le32_to_cpu(rhead->h_crc),
+					le32_to_cpu(old_crc),
 					le32_to_cpu(crc));
 			xfs_hex_dump(dp, 32);
 		}