From 5883ea2555b2ae8dd84a256532f7abb2d4837fc1 Mon Sep 17 00:00:00 2001
From: Dave Chinner <dchinner@redhat.com>
Date: Tue, 20 Jul 2010 09:43:39 +1000
Subject: xfs: track AGs with reclaimable inodes in per-ag radix tree

https://bugzilla.kernel.org/show_bug.cgi?id=16348

When the filesystem grows to a large number of allocation groups,
the summing of recalimable inodes gets expensive. In many cases,
most AGs won't have any reclaimable inodes and so we are wasting CPU
time aggregating over these AGs. This is particularly important for
the inode shrinker that gets called frequently under memory
pressure.

To avoid the overhead, track AGs with reclaimable inodes in the
per-ag radix tree so that we can find all the AGs with reclaimable
inodes via a simple gang tag lookup. This involves setting the tag
when the first reclaimable inode is tracked in the AG, and removing
the tag when the last reclaimable inode is removed from the tree.
Then the summation process becomes a loop walking the radix tree
summing AGs with the reclaim tag set.

This significantly reduces the overhead of scanning - a 6400 AG
filesystea now only uses about 25% of a cpu in kswapd while slab
reclaim progresses instead of being permanently stuck at 100% CPU
and making little progress. Clean filesystems filesystems will see
no overhead and the overhead only increases linearly with the number
of dirty AGs.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
 fs/xfs/linux-2.6/xfs_sync.c  |   68 +++++++++++++++++++++++++++++++++++++----
 fs/xfs/linux-2.6/xfs_trace.h |   61 +++++++++++++++++++++----------------
 2 files changed, 95 insertions(+), 34 deletions(-)

diff --git a/fs/xfs/linux-2.6/xfs_sync.c b/fs/xfs/linux-2.6/xfs_sync.c
index a427c63..b927a54 100644
--- a/fs/xfs/linux-2.6/xfs_sync.c
+++ b/fs/xfs/linux-2.6/xfs_sync.c
@@ -144,6 +144,41 @@ restart:
 	return last_error;
 }
 
+/*
+ * Select the next per-ag structure to iterate during the walk. The reclaim
+ * walk is optimised only to walk AGs with reclaimable inodes in them.
+ */
+static struct xfs_perag *
+xfs_inode_ag_iter_next_pag(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		*first,
+	int			tag)
+{
+	struct xfs_perag	*pag = NULL;
+
+	if (tag == XFS_ICI_RECLAIM_TAG) {
+		int found;
+		int ref;
+
+		spin_lock(&mp->m_perag_lock);
+		found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
+				(void **)&pag, *first, 1, tag);
+		if (found <= 0) {
+			spin_unlock(&mp->m_perag_lock);
+			return NULL;
+		}
+		*first = pag->pag_agno + 1;
+		/* open coded pag reference increment */
+		ref = atomic_inc_return(&pag->pag_ref);
+		spin_unlock(&mp->m_perag_lock);
+		trace_xfs_perag_get_reclaim(mp, pag->pag_agno, ref, _RET_IP_);
+	} else {
+		pag = xfs_perag_get(mp, *first);
+		(*first)++;
+	}
+	return pag;
+}
+
 int
 xfs_inode_ag_iterator(
 	struct xfs_mount	*mp,
@@ -154,16 +189,15 @@ xfs_inode_ag_iterator(
 	int			exclusive,
 	int			*nr_to_scan)
 {
+	struct xfs_perag	*pag;
 	int			error = 0;
 	int			last_error = 0;
 	xfs_agnumber_t		ag;
 	int			nr;
 
 	nr = nr_to_scan ? *nr_to_scan : INT_MAX;
-	for (ag = 0; ag < mp->m_sb.sb_agcount; ag++) {
-		struct xfs_perag	*pag;
-
-		pag = xfs_perag_get(mp, ag);
+	ag = 0;
+	while ((pag = xfs_inode_ag_iter_next_pag(mp, &ag, tag))) {
 		if (!pag->pag_ici_init) {
 			xfs_perag_put(pag);
 			continue;
@@ -681,6 +715,17 @@ __xfs_inode_set_reclaim_tag(
 	radix_tree_tag_set(&pag->pag_ici_root,
 			   XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino),
 			   XFS_ICI_RECLAIM_TAG);
+
+	if (!pag->pag_ici_reclaimable) {
+		/* propagate the reclaim tag up into the perag radix tree */
+		spin_lock(&ip->i_mount->m_perag_lock);
+		radix_tree_tag_set(&ip->i_mount->m_perag_tree,
+				XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino),
+				XFS_ICI_RECLAIM_TAG);
+		spin_unlock(&ip->i_mount->m_perag_lock);
+		trace_xfs_perag_set_reclaim(ip->i_mount, pag->pag_agno,
+							-1, _RET_IP_);
+	}
 	pag->pag_ici_reclaimable++;
 }
 
@@ -715,6 +760,16 @@ __xfs_inode_clear_reclaim_tag(
 	radix_tree_tag_clear(&pag->pag_ici_root,
 			XFS_INO_TO_AGINO(mp, ip->i_ino), XFS_ICI_RECLAIM_TAG);
 	pag->pag_ici_reclaimable--;
+	if (!pag->pag_ici_reclaimable) {
+		/* clear the reclaim tag from the perag radix tree */
+		spin_lock(&ip->i_mount->m_perag_lock);
+		radix_tree_tag_clear(&ip->i_mount->m_perag_tree,
+				XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino),
+				XFS_ICI_RECLAIM_TAG);
+		spin_unlock(&ip->i_mount->m_perag_lock);
+		trace_xfs_perag_clear_reclaim(ip->i_mount, pag->pag_agno,
+							-1, _RET_IP_);
+	}
 }
 
 /*
@@ -903,9 +958,8 @@ xfs_reclaim_inode_shrink(
 
 	down_read(&xfs_mount_list_lock);
 	list_for_each_entry(mp, &xfs_mount_list, m_mplist) {
-		for (ag = 0; ag < mp->m_sb.sb_agcount; ag++) {
-
-			pag = xfs_perag_get(mp, ag);
+		while ((pag = xfs_inode_ag_iter_next_pag(mp, &ag,
+					XFS_ICI_RECLAIM_TAG))) {
 			if (!pag->pag_ici_init) {
 				xfs_perag_put(pag);
 				continue;
diff --git a/fs/xfs/linux-2.6/xfs_trace.h b/fs/xfs/linux-2.6/xfs_trace.h
index fcaa62f..072d581 100644
--- a/fs/xfs/linux-2.6/xfs_trace.h
+++ b/fs/xfs/linux-2.6/xfs_trace.h
@@ -78,33 +78,6 @@ DECLARE_EVENT_CLASS(xfs_attr_list_class,
 	)
 )
 
-#define DEFINE_PERAG_REF_EVENT(name) \
-TRACE_EVENT(name, \
-	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, int refcount, \
-		 unsigned long caller_ip), \
-	TP_ARGS(mp, agno, refcount, caller_ip), \
-	TP_STRUCT__entry( \
-		__field(dev_t, dev) \
-		__field(xfs_agnumber_t, agno) \
-		__field(int, refcount) \
-		__field(unsigned long, caller_ip) \
-	), \
-	TP_fast_assign( \
-		__entry->dev = mp->m_super->s_dev; \
-		__entry->agno = agno; \
-		__entry->refcount = refcount; \
-		__entry->caller_ip = caller_ip; \
-	), \
-	TP_printk("dev %d:%d agno %u refcount %d caller %pf", \
-		  MAJOR(__entry->dev), MINOR(__entry->dev), \
-		  __entry->agno, \
-		  __entry->refcount, \
-		  (char *)__entry->caller_ip) \
-);
-
-DEFINE_PERAG_REF_EVENT(xfs_perag_get)
-DEFINE_PERAG_REF_EVENT(xfs_perag_put)
-
 #define DEFINE_ATTR_LIST_EVENT(name) \
 DEFINE_EVENT(xfs_attr_list_class, name, \
 	TP_PROTO(struct xfs_attr_list_context *ctx), \
@@ -118,6 +91,40 @@ DEFINE_ATTR_LIST_EVENT(xfs_attr_list_add);
 DEFINE_ATTR_LIST_EVENT(xfs_attr_list_wrong_blk);
 DEFINE_ATTR_LIST_EVENT(xfs_attr_list_notfound);
 
+DECLARE_EVENT_CLASS(xfs_perag_class,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, int refcount,
+		 unsigned long caller_ip),
+	TP_ARGS(mp, agno, refcount, caller_ip),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, agno)
+		__field(int, refcount)
+		__field(unsigned long, caller_ip)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->agno = agno;
+		__entry->refcount = refcount;
+		__entry->caller_ip = caller_ip;
+	),
+	TP_printk("dev %d:%d agno %u refcount %d caller %pf",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->agno,
+		  __entry->refcount,
+		  (char *)__entry->caller_ip)
+);
+
+#define DEFINE_PERAG_REF_EVENT(name)	\
+DEFINE_EVENT(xfs_perag_class, name,	\
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, int refcount,	\
+		 unsigned long caller_ip),					\
+	TP_ARGS(mp, agno, refcount, caller_ip))
+DEFINE_PERAG_REF_EVENT(xfs_perag_get);
+DEFINE_PERAG_REF_EVENT(xfs_perag_get_reclaim);
+DEFINE_PERAG_REF_EVENT(xfs_perag_put);
+DEFINE_PERAG_REF_EVENT(xfs_perag_set_reclaim);
+DEFINE_PERAG_REF_EVENT(xfs_perag_clear_reclaim);
+
 TRACE_EVENT(xfs_attr_list_node_descend,
 	TP_PROTO(struct xfs_attr_list_context *ctx,
 		 struct xfs_da_node_entry *btree),
-- 
1.7.3.2