Merge tag 'xfs-perag-conv-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs into xfs-5.14-merge2

xfs: initial agnumber -> perag conversions for shrink

If we want to use active references to the perag to be able to gate
shrink removing AGs and hence perags safely, we've got a fair bit of
work to do actually use perags in all the places we need to.

There's a lot of code that iterates ag numbers and then
looks up perags from that, often multiple times for the same perag
in the one operation. If we want to use reference counted perags for
access control, then we need to convert all these uses to perag
iterators, not agno iterators.

[Patches 1-4]

The first step of this is consolidating all the perag management -
init, free, get, put, etc into a common location. THis is spread all
over the place right now, so move it all into libxfs/xfs_ag.[ch].
This does expose kernel only bits of the perag to libxfs and hence
userspace, so the structures and code is rearranged to minimise the
number of ifdefs that need to be added to the userspace codebase.
The perag iterator in xfs_icache.c is promoted to a first class API
and expanded to the needs of the code as required.

[Patches 5-10]

These are the first basic perag iterator conversions and changes to
pass the perag down the stack from those iterators where
appropriate. A lot of this is obvious, simple changes, though in
some places we stop passing the perag down the stack because the
code enters into an as yet unconverted subsystem that still uses raw
AGs.

[Patches 11-16]

These replace the agno passed in the btree cursor for per-ag btree
operations with a perag that is passed to the cursor init function.
The cursor takes it's own reference to the perag, and the reference
is dropped when the cursor is deleted. Hence we get reference
coverage for the entire time the cursor is active, even if the code
that initialised the cursor drops it's reference before the cursor
or any of it's children (duplicates) have been deleted.

The first patch adds the perag infrastructure for the cursor, the
next four patches convert a btree cursor at a time, and the last
removes the agno from the cursor once it is unused.

[Patches 17-21]

These patches are a demonstration of the simplifications and
cleanups that come from plumbing the perag through interfaces that
select and then operate on a specific AG. In this case the inode
allocation algorithm does up to three walks across all AGs before it
either allocates an inode or fails. Two of these walks are purely
just to select the AG, and even then it doesn't guarantee inode
allocation success so there's a third walk if the selected AG
allocation fails.

These patches collapse the selection and allocation into a single
loop, simplifies the error handling because xfs_dir_ialloc() always
returns ENOSPC if no AG was selected for inode allocation or we fail
to allocate an inode in any AG, gets rid of xfs_dir_ialloc()
wrapper, converts inode allocation to run entirely from a single
perag instance, and then factors xfs_dialloc() into a much, much
simpler loop which is easy to understand.

Hence we end up with the same inode allocation logic, but it only
needs two complete iterations at worst, makes AG selection and
allocation atomic w.r.t. shrink and chops out out over 100 lines of
code from this hot code path.

[Patch 22]

Converts the unlink path to pass perags through it.

There's more conversion work to be done, but this patchset gets
through a large chunk of it in one hit. Most of the iterators are
converted, so once this is solidified we can move on to converting
these to active references for being able to free perags while the
fs is still active.

* tag 'xfs-perag-conv-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs: (23 commits)
  xfs: remove xfs_perag_t
  xfs: use perag through unlink processing
  xfs: clean up and simplify xfs_dialloc()
  xfs: inode allocation can use a single perag instance
  xfs: get rid of xfs_dir_ialloc()
  xfs: collapse AG selection for inode allocation
  xfs: simplify xfs_dialloc_select_ag() return values
  xfs: remove agno from btree cursor
  xfs: use perag for ialloc btree cursors
  xfs: convert allocbt cursors to use perags
  xfs: convert refcount btree cursor to use perags
  xfs: convert rmap btree cursor to using a perag
  xfs: add a perag to the btree cursor
  xfs: pass perags around in fsmap data dev functions
  xfs: push perags through the ag reservation callouts
  xfs: pass perags through to the busy extent code
  xfs: convert secondary superblock walk to use perags
  xfs: convert xfs_iwalk to use perag references
  xfs: convert raw ag walks to use for_each_perag
  xfs: make for_each_perag... a first class citizen
  ...

60 files changed, 1408 insertions, 1389 deletions

diff --git a/fs/xfs/libxfs/xfs_ag.c b/fs/xfs/libxfs/xfs_ag.c
index be0087825ae0..5315e3f57207 100644
--- a/fs/xfs/libxfs/xfs_ag.c
+++ b/fs/xfs/libxfs/xfs_ag.c
@@ -27,6 +27,276 @@
 #include "xfs_defer.h"
 #include "xfs_log_format.h"
 #include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_inode.h"
+#include "xfs_icache.h"
+
+
+/*
+ * Passive reference counting access wrappers to the perag structures.  If the
+ * per-ag structure is to be freed, the freeing code is responsible for cleaning
+ * up objects with passive references before freeing the structure. This is
+ * things like cached buffers.
+ */
+struct xfs_perag *
+xfs_perag_get(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno)
+{
+	struct xfs_perag	*pag;
+	int			ref = 0;
+
+	rcu_read_lock();
+	pag = radix_tree_lookup(&mp->m_perag_tree, agno);
+	if (pag) {
+		ASSERT(atomic_read(&pag->pag_ref) >= 0);
+		ref = atomic_inc_return(&pag->pag_ref);
+	}
+	rcu_read_unlock();
+	trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
+	return pag;
+}
+
+/*
+ * search from @first to find the next perag with the given tag set.
+ */
+struct xfs_perag *
+xfs_perag_get_tag(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		first,
+	int			tag)
+{
+	struct xfs_perag	*pag;
+	int			found;
+	int			ref;
+
+	rcu_read_lock();
+	found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
+					(void **)&pag, first, 1, tag);
+	if (found <= 0) {
+		rcu_read_unlock();
+		return NULL;
+	}
+	ref = atomic_inc_return(&pag->pag_ref);
+	rcu_read_unlock();
+	trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
+	return pag;
+}
+
+void
+xfs_perag_put(
+	struct xfs_perag	*pag)
+{
+	int	ref;
+
+	ASSERT(atomic_read(&pag->pag_ref) > 0);
+	ref = atomic_dec_return(&pag->pag_ref);
+	trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
+}
+
+/*
+ * xfs_initialize_perag_data
+ *
+ * Read in each per-ag structure so we can count up the number of
+ * allocated inodes, free inodes and used filesystem blocks as this
+ * information is no longer persistent in the superblock. Once we have
+ * this information, write it into the in-core superblock structure.
+ */
+int
+xfs_initialize_perag_data(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agcount)
+{
+	xfs_agnumber_t		index;
+	struct xfs_perag	*pag;
+	struct xfs_sb		*sbp = &mp->m_sb;
+	uint64_t		ifree = 0;
+	uint64_t		ialloc = 0;
+	uint64_t		bfree = 0;
+	uint64_t		bfreelst = 0;
+	uint64_t		btree = 0;
+	uint64_t		fdblocks;
+	int			error = 0;
+
+	for (index = 0; index < agcount; index++) {
+		/*
+		 * read the agf, then the agi. This gets us
+		 * all the information we need and populates the
+		 * per-ag structures for us.
+		 */
+		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
+		if (error)
+			return error;
+
+		error = xfs_ialloc_pagi_init(mp, NULL, index);
+		if (error)
+			return error;
+		pag = xfs_perag_get(mp, index);
+		ifree += pag->pagi_freecount;
+		ialloc += pag->pagi_count;
+		bfree += pag->pagf_freeblks;
+		bfreelst += pag->pagf_flcount;
+		btree += pag->pagf_btreeblks;
+		xfs_perag_put(pag);
+	}
+	fdblocks = bfree + bfreelst + btree;
+
+	/*
+	 * If the new summary counts are obviously incorrect, fail the
+	 * mount operation because that implies the AGFs are also corrupt.
+	 * Clear FS_COUNTERS so that we don't unmount with a dirty log, which
+	 * will prevent xfs_repair from fixing anything.
+	 */
+	if (fdblocks > sbp->sb_dblocks || ifree > ialloc) {
+		xfs_alert(mp, "AGF corruption. Please run xfs_repair.");
+		error = -EFSCORRUPTED;
+		goto out;
+	}
+
+	/* Overwrite incore superblock counters with just-read data */
+	spin_lock(&mp->m_sb_lock);
+	sbp->sb_ifree = ifree;
+	sbp->sb_icount = ialloc;
+	sbp->sb_fdblocks = fdblocks;
+	spin_unlock(&mp->m_sb_lock);
+
+	xfs_reinit_percpu_counters(mp);
+out:
+	xfs_fs_mark_healthy(mp, XFS_SICK_FS_COUNTERS);
+	return error;
+}
+
+STATIC void
+__xfs_free_perag(
+	struct rcu_head	*head)
+{
+	struct xfs_perag *pag = container_of(head, struct xfs_perag, rcu_head);
+
+	ASSERT(!delayed_work_pending(&pag->pag_blockgc_work));
+	ASSERT(atomic_read(&pag->pag_ref) == 0);
+	kmem_free(pag);
+}
+
+/*
+ * Free up the per-ag resources associated with the mount structure.
+ */
+void
+xfs_free_perag(
+	struct xfs_mount	*mp)
+{
+	struct xfs_perag	*pag;
+	xfs_agnumber_t		agno;
+
+	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
+		spin_lock(&mp->m_perag_lock);
+		pag = radix_tree_delete(&mp->m_perag_tree, agno);
+		spin_unlock(&mp->m_perag_lock);
+		ASSERT(pag);
+		ASSERT(atomic_read(&pag->pag_ref) == 0);
+
+		cancel_delayed_work_sync(&pag->pag_blockgc_work);
+		xfs_iunlink_destroy(pag);
+		xfs_buf_hash_destroy(pag);
+
+		call_rcu(&pag->rcu_head, __xfs_free_perag);
+	}
+}
+
+int
+xfs_initialize_perag(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agcount,
+	xfs_agnumber_t		*maxagi)
+{
+	struct xfs_perag	*pag;
+	xfs_agnumber_t		index;
+	xfs_agnumber_t		first_initialised = NULLAGNUMBER;
+	int			error;
+
+	/*
+	 * Walk the current per-ag tree so we don't try to initialise AGs
+	 * that already exist (growfs case). Allocate and insert all the
+	 * AGs we don't find ready for initialisation.
+	 */
+	for (index = 0; index < agcount; index++) {
+		pag = xfs_perag_get(mp, index);
+		if (pag) {
+			xfs_perag_put(pag);
+			continue;
+		}
+
+		pag = kmem_zalloc(sizeof(*pag), KM_MAYFAIL);
+		if (!pag) {
+			error = -ENOMEM;
+			goto out_unwind_new_pags;
+		}
+		pag->pag_agno = index;
+		pag->pag_mount = mp;
+
+		error = radix_tree_preload(GFP_NOFS);
+		if (error)
+			goto out_free_pag;
+
+		spin_lock(&mp->m_perag_lock);
+		if (radix_tree_insert(&mp->m_perag_tree, index, pag)) {
+			WARN_ON_ONCE(1);
+			spin_unlock(&mp->m_perag_lock);
+			radix_tree_preload_end();
+			error = -EEXIST;
+			goto out_free_pag;
+		}
+		spin_unlock(&mp->m_perag_lock);
+		radix_tree_preload_end();
+
+		/* Place kernel structure only init below this point. */
+		spin_lock_init(&pag->pag_ici_lock);
+		spin_lock_init(&pag->pagb_lock);
+		spin_lock_init(&pag->pag_state_lock);
+		INIT_DELAYED_WORK(&pag->pag_blockgc_work, xfs_blockgc_worker);
+		INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC);
+		init_waitqueue_head(&pag->pagb_wait);
+		pag->pagb_count = 0;
+		pag->pagb_tree = RB_ROOT;
+
+		error = xfs_buf_hash_init(pag);
+		if (error)
+			goto out_remove_pag;
+
+		error = xfs_iunlink_init(pag);
+		if (error)
+			goto out_hash_destroy;
+
+		/* first new pag is fully initialized */
+		if (first_initialised == NULLAGNUMBER)
+			first_initialised = index;
+	}
+
+	index = xfs_set_inode_alloc(mp, agcount);
+
+	if (maxagi)
+		*maxagi = index;
+
+	mp->m_ag_prealloc_blocks = xfs_prealloc_blocks(mp);
+	return 0;
+
+out_hash_destroy:
+	xfs_buf_hash_destroy(pag);
+out_remove_pag:
+	radix_tree_delete(&mp->m_perag_tree, index);
+out_free_pag:
+	kmem_free(pag);
+out_unwind_new_pags:
+	/* unwind any prior newly initialized pags */
+	for (index = first_initialised; index < agcount; index++) {
+		pag = radix_tree_delete(&mp->m_perag_tree, index);
+		if (!pag)
+			break;
+		xfs_buf_hash_destroy(pag);
+		xfs_iunlink_destroy(pag);
+		kmem_free(pag);
+	}
+	return error;
+}
 
 static int
 xfs_get_aghdr_buf(
@@ -645,7 +915,7 @@ xfs_ag_extend_space(
 	 * XFS_RMAP_OINFO_SKIP_UPDATE is used here to tell the rmap btree that
 	 * this doesn't actually exist in the rmap btree.
 	 */
-	error = xfs_rmap_free(tp, bp, id->agno,
+	error = xfs_rmap_free(tp, bp, bp->b_pag,
 				be32_to_cpu(agf->agf_length) - len,
 				len, &XFS_RMAP_OINFO_SKIP_UPDATE);
 	if (error)
diff --git a/fs/xfs/libxfs/xfs_ag.h b/fs/xfs/libxfs/xfs_ag.h
index 4535de1d88ea..6006b4329026 100644
--- a/fs/xfs/libxfs/xfs_ag.h
+++ b/fs/xfs/libxfs/xfs_ag.h
@@ -9,6 +9,142 @@
 
 struct xfs_mount;
 struct xfs_trans;
+struct xfs_perag;
+
+/*
+ * Per-ag infrastructure
+ */
+
+/* per-AG block reservation data structures*/
+struct xfs_ag_resv {
+	/* number of blocks originally reserved here */
+	xfs_extlen_t			ar_orig_reserved;
+	/* number of blocks reserved here */
+	xfs_extlen_t			ar_reserved;
+	/* number of blocks originally asked for */
+	xfs_extlen_t			ar_asked;
+};
+
+/*
+ * Per-ag incore structure, copies of information in agf and agi, to improve the
+ * performance of allocation group selection.
+ */
+struct xfs_perag {
+	struct xfs_mount *pag_mount;	/* owner filesystem */
+	xfs_agnumber_t	pag_agno;	/* AG this structure belongs to */
+	atomic_t	pag_ref;	/* perag reference count */
+	char		pagf_init;	/* this agf's entry is initialized */
+	char		pagi_init;	/* this agi's entry is initialized */
+	char		pagf_metadata;	/* the agf is preferred to be metadata */
+	char		pagi_inodeok;	/* The agi is ok for inodes */
+	uint8_t		pagf_levels[XFS_BTNUM_AGF];
+					/* # of levels in bno & cnt btree */
+	bool		pagf_agflreset; /* agfl requires reset before use */
+	uint32_t	pagf_flcount;	/* count of blocks in freelist */
+	xfs_extlen_t	pagf_freeblks;	/* total free blocks */
+	xfs_extlen_t	pagf_longest;	/* longest free space */
+	uint32_t	pagf_btreeblks;	/* # of blocks held in AGF btrees */
+	xfs_agino_t	pagi_freecount;	/* number of free inodes */
+	xfs_agino_t	pagi_count;	/* number of allocated inodes */
+
+	/*
+	 * Inode allocation search lookup optimisation.
+	 * If the pagino matches, the search for new inodes
+	 * doesn't need to search the near ones again straight away
+	 */
+	xfs_agino_t	pagl_pagino;
+	xfs_agino_t	pagl_leftrec;
+	xfs_agino_t	pagl_rightrec;
+
+	int		pagb_count;	/* pagb slots in use */
+	uint8_t		pagf_refcount_level; /* recount btree height */
+
+	/* Blocks reserved for all kinds of metadata. */
+	struct xfs_ag_resv	pag_meta_resv;
+	/* Blocks reserved for the reverse mapping btree. */
+	struct xfs_ag_resv	pag_rmapbt_resv;
+
+	/* -- kernel only structures below this line -- */
+
+	/*
+	 * Bitsets of per-ag metadata that have been checked and/or are sick.
+	 * Callers should hold pag_state_lock before accessing this field.
+	 */
+	uint16_t	pag_checked;
+	uint16_t	pag_sick;
+	spinlock_t	pag_state_lock;
+
+	spinlock_t	pagb_lock;	/* lock for pagb_tree */
+	struct rb_root	pagb_tree;	/* ordered tree of busy extents */
+	unsigned int	pagb_gen;	/* generation count for pagb_tree */
+	wait_queue_head_t pagb_wait;	/* woken when pagb_gen changes */
+
+	atomic_t        pagf_fstrms;    /* # of filestreams active in this AG */
+
+	spinlock_t	pag_ici_lock;	/* incore inode cache lock */
+	struct radix_tree_root pag_ici_root;	/* incore inode cache root */
+	int		pag_ici_reclaimable;	/* reclaimable inodes */
+	unsigned long	pag_ici_reclaim_cursor;	/* reclaim restart point */
+
+	/* buffer cache index */
+	spinlock_t	pag_buf_lock;	/* lock for pag_buf_hash */
+	struct rhashtable pag_buf_hash;
+
+	/* for rcu-safe freeing */
+	struct rcu_head	rcu_head;
+
+	/* background prealloc block trimming */
+	struct delayed_work	pag_blockgc_work;
+
+	/*
+	 * Unlinked inode information.  This incore information reflects
+	 * data stored in the AGI, so callers must hold the AGI buffer lock
+	 * or have some other means to control concurrency.
+	 */
+	struct rhashtable	pagi_unlinked_hash;
+};
+
+int xfs_initialize_perag(struct xfs_mount *mp, xfs_agnumber_t agcount,
+			xfs_agnumber_t *maxagi);
+int xfs_initialize_perag_data(struct xfs_mount *, xfs_agnumber_t);
+void xfs_free_perag(struct xfs_mount *mp);
+
+struct xfs_perag *xfs_perag_get(struct xfs_mount *, xfs_agnumber_t);
+struct xfs_perag *xfs_perag_get_tag(struct xfs_mount *, xfs_agnumber_t,
+				   int tag);
+void	xfs_perag_put(struct xfs_perag *pag);
+
+/*
+ * Perag iteration APIs
+ *
+ * XXX: for_each_perag_range() usage really needs an iterator to clean up when
+ * we terminate at end_agno because we may have taken a reference to the perag
+ * beyond end_agno. Right now callers have to be careful to catch and clean that
+ * up themselves. This is not necessary for the callers of for_each_perag() and
+ * for_each_perag_from() because they terminate at sb_agcount where there are
+ * no perag structures in tree beyond end_agno.
+ */
+#define for_each_perag_range(mp, next_agno, end_agno, pag) \
+	for ((pag) = xfs_perag_get((mp), (next_agno)); \
+		(pag) != NULL && (next_agno) <= (end_agno); \
+		(next_agno) = (pag)->pag_agno + 1, \
+		xfs_perag_put(pag), \
+		(pag) = xfs_perag_get((mp), (next_agno)))
+
+#define for_each_perag_from(mp, next_agno, pag) \
+	for_each_perag_range((mp), (next_agno), (mp)->m_sb.sb_agcount, (pag))
+
+
+#define for_each_perag(mp, agno, pag) \
+	(agno) = 0; \
+	for_each_perag_from((mp), (agno), (pag))
+
+#define for_each_perag_tag(mp, agno, pag, tag) \
+	for ((agno) = 0, (pag) = xfs_perag_get_tag((mp), 0, (tag)); \
+		(pag) != NULL; \
+		(agno) = (pag)->pag_agno + 1, \
+		xfs_perag_put(pag), \
+		(pag) = xfs_perag_get_tag((mp), (agno), (tag)))
 
 struct aghdr_init_data {
 	/* per ag data */
diff --git a/fs/xfs/libxfs/xfs_ag_resv.c b/fs/xfs/libxfs/xfs_ag_resv.c
index bbfea8022a3b..f6f868020e2e 100644
--- a/fs/xfs/libxfs/xfs_ag_resv.c
+++ b/fs/xfs/libxfs/xfs_ag_resv.c
@@ -19,7 +19,7 @@
 #include "xfs_btree.h"
 #include "xfs_refcount_btree.h"
 #include "xfs_ialloc_btree.h"
-#include "xfs_sb.h"
+#include "xfs_ag.h"
 #include "xfs_ag_resv.h"
 
 /*
@@ -250,7 +250,6 @@ xfs_ag_resv_init(
 	struct xfs_trans		*tp)
 {
 	struct xfs_mount		*mp = pag->pag_mount;
-	xfs_agnumber_t			agno = pag->pag_agno;
 	xfs_extlen_t			ask;
 	xfs_extlen_t			used;
 	int				error = 0, error2;
@@ -260,11 +259,11 @@ xfs_ag_resv_init(
 	if (pag->pag_meta_resv.ar_asked == 0) {
 		ask = used = 0;
 
-		error = xfs_refcountbt_calc_reserves(mp, tp, agno, &ask, &used);
+		error = xfs_refcountbt_calc_reserves(mp, tp, pag, &ask, &used);
 		if (error)
 			goto out;
 
-		error = xfs_finobt_calc_reserves(mp, tp, agno, &ask, &used);
+		error = xfs_finobt_calc_reserves(mp, tp, pag, &ask, &used);
 		if (error)
 			goto out;
 
@@ -282,7 +281,7 @@ xfs_ag_resv_init(
 
 			mp->m_finobt_nores = true;
 
-			error = xfs_refcountbt_calc_reserves(mp, tp, agno, &ask,
+			error = xfs_refcountbt_calc_reserves(mp, tp, pag, &ask,
 					&used);
 			if (error)
 				goto out;
@@ -300,7 +299,7 @@ xfs_ag_resv_init(
 	if (pag->pag_rmapbt_resv.ar_asked == 0) {
 		ask = used = 0;
 
-		error = xfs_rmapbt_calc_reserves(mp, tp, agno, &ask, &used);
+		error = xfs_rmapbt_calc_reserves(mp, tp, pag, &ask, &used);
 		if (error)
 			goto out;
 
diff --git a/fs/xfs/libxfs/xfs_ag_resv.h b/fs/xfs/libxfs/xfs_ag_resv.h
index 8a8eb4bc48bb..b74b210008ea 100644
--- a/fs/xfs/libxfs/xfs_ag_resv.h
+++ b/fs/xfs/libxfs/xfs_ag_resv.h
@@ -18,6 +18,21 @@ void xfs_ag_resv_alloc_extent(struct xfs_perag *pag, enum xfs_ag_resv_type type,
 void xfs_ag_resv_free_extent(struct xfs_perag *pag, enum xfs_ag_resv_type type,
 		struct xfs_trans *tp, xfs_extlen_t len);
 
+static inline struct xfs_ag_resv *
+xfs_perag_resv(
+	struct xfs_perag	*pag,
+	enum xfs_ag_resv_type	type)
+{
+	switch (type) {
+	case XFS_AG_RESV_METADATA:
+		return &pag->pag_meta_resv;
+	case XFS_AG_RESV_RMAPBT:
+		return &pag->pag_rmapbt_resv;
+	default:
+		return NULL;
+	}
+}
+
 /*
  * RMAPBT reservation accounting wrappers. Since rmapbt blocks are sourced from
  * the AGFL, they are allocated one at a time and the reservation updates don't
diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
index 82b7cbb1f24f..00bb34251829 100644
--- a/fs/xfs/libxfs/xfs_alloc.c
+++ b/fs/xfs/libxfs/xfs_alloc.c
@@ -10,7 +10,6 @@
 #include "xfs_shared.h"
 #include "xfs_trans_resv.h"
 #include "xfs_bit.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_defer.h"
 #include "xfs_btree.h"
@@ -24,6 +23,7 @@
 #include "xfs_trans.h"
 #include "xfs_buf_item.h"
 #include "xfs_log.h"
+#include "xfs_ag.h"
 #include "xfs_ag_resv.h"
 #include "xfs_bmap.h"
 
@@ -230,7 +230,7 @@ xfs_alloc_get_rec(
 	int			*stat)	/* output: success/failure */
 {
 	struct xfs_mount	*mp = cur->bc_mp;
-	xfs_agnumber_t		agno = cur->bc_ag.agno;
+	xfs_agnumber_t		agno = cur->bc_ag.pag->pag_agno;
 	union xfs_btree_rec	*rec;
 	int			error;
 
@@ -776,7 +776,7 @@ xfs_alloc_cur_setup(
 	 */
 	if (!acur->cnt)
 		acur->cnt = xfs_allocbt_init_cursor(args->mp, args->tp,
-					args->agbp, args->agno, XFS_BTNUM_CNT);
+					args->agbp, args->pag, XFS_BTNUM_CNT);
 	error = xfs_alloc_lookup_ge(acur->cnt, 0, args->maxlen, &i);
 	if (error)
 		return error;
@@ -786,10 +786,10 @@ xfs_alloc_cur_setup(
 	 */
 	if (!acur->bnolt)
 		acur->bnolt = xfs_allocbt_init_cursor(args->mp, args->tp,
-					args->agbp, args->agno, XFS_BTNUM_BNO);
+					args->agbp, args->pag, XFS_BTNUM_BNO);
 	if (!acur->bnogt)
 		acur->bnogt = xfs_allocbt_init_cursor(args->mp, args->tp,
-					args->agbp, args->agno, XFS_BTNUM_BNO);
+					args->agbp, args->pag, XFS_BTNUM_BNO);
 	return i == 1 ? 0 : -ENOSPC;
 }
 
@@ -1063,7 +1063,7 @@ xfs_alloc_ag_vextent_small(
 	if (fbno == NULLAGBLOCK)
 		goto out;
 
-	xfs_extent_busy_reuse(args->mp, args->agno, fbno, 1,
+	xfs_extent_busy_reuse(args->mp, args->pag, fbno, 1,
 			      (args->datatype & XFS_ALLOC_NOBUSY));
 
 	if (args->datatype & XFS_ALLOC_USERDATA) {
@@ -1089,7 +1089,7 @@ xfs_alloc_ag_vextent_small(
 	 * If we're feeding an AGFL block to something that doesn't live in the
 	 * free space, we need to clear out the OWN_AG rmap.
 	 */
-	error = xfs_rmap_free(args->tp, args->agbp, args->agno, fbno, 1,
+	error = xfs_rmap_free(args->tp, args->agbp, args->pag, fbno, 1,
 			      &XFS_RMAP_OINFO_AG);
 	if (error)
 		goto error;
@@ -1166,7 +1166,7 @@ xfs_alloc_ag_vextent(
 
 	/* if not file data, insert new block into the reverse map btree */
 	if (!xfs_rmap_should_skip_owner_update(&args->oinfo)) {
-		error = xfs_rmap_alloc(args->tp, args->agbp, args->agno,
+		error = xfs_rmap_alloc(args->tp, args->agbp, args->pag,
 				       args->agbno, args->len, &args->oinfo);
 		if (error)
 			return error;
@@ -1178,7 +1178,7 @@ xfs_alloc_ag_vextent(
 		if (error)
 			return error;
 
-		ASSERT(!xfs_extent_busy_search(args->mp, args->agno,
+		ASSERT(!xfs_extent_busy_search(args->mp, args->pag,
 					      args->agbno, args->len));
 	}
 
@@ -1217,7 +1217,7 @@ xfs_alloc_ag_vextent_exact(
 	 * Allocate/initialize a cursor for the by-number freespace btree.
 	 */
 	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
-					  args->agno, XFS_BTNUM_BNO);
+					  args->pag, XFS_BTNUM_BNO);
 
 	/*
 	 * Lookup bno and minlen in the btree (minlen is irrelevant, really).
@@ -1277,7 +1277,7 @@ xfs_alloc_ag_vextent_exact(
 	 * Allocate/initialize a cursor for the by-size btree.
 	 */
 	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
-		args->agno, XFS_BTNUM_CNT);
+					args->pag, XFS_BTNUM_CNT);
 	ASSERT(args->agbno + args->len <= be32_to_cpu(agf->agf_length));
 	error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen, args->agbno,
 				      args->len, XFSA_FIXUP_BNO_OK);
@@ -1674,7 +1674,7 @@ restart:
 	 * Allocate and initialize a cursor for the by-size btree.
 	 */
 	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
-		args->agno, XFS_BTNUM_CNT);
+					args->pag, XFS_BTNUM_CNT);
 	bno_cur = NULL;
 	busy = false;
 
@@ -1837,7 +1837,7 @@ restart:
 	 * Allocate and initialize a cursor for the by-block tree.
 	 */
 	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
-		args->agno, XFS_BTNUM_BNO);
+					args->pag, XFS_BTNUM_BNO);
 	if ((error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen,
 			rbno, rlen, XFSA_FIXUP_CNT_OK)))
 		goto error0;
@@ -1896,12 +1896,13 @@ xfs_free_ag_extent(
 	int				haveright; /* have a right neighbor */
 	int				i;
 	int				error;
+	struct xfs_perag		*pag = agbp->b_pag;
 
 	bno_cur = cnt_cur = NULL;
 	mp = tp->t_mountp;
 
 	if (!xfs_rmap_should_skip_owner_update(oinfo)) {
-		error = xfs_rmap_free(tp, agbp, agno, bno, len, oinfo);
+		error = xfs_rmap_free(tp, agbp, pag, bno, len, oinfo);
 		if (error)
 			goto error0;
 	}
@@ -1909,7 +1910,7 @@ xfs_free_ag_extent(
 	/*
 	 * Allocate and initialize a cursor for the by-block btree.
 	 */
-	bno_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_BNO);
+	bno_cur = xfs_allocbt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_BNO);
 	/*
 	 * Look for a neighboring block on the left (lower block numbers)
 	 * that is contiguous with this space.
@@ -1979,7 +1980,7 @@ xfs_free_ag_extent(
 	/*
 	 * Now allocate and initialize a cursor for the by-size tree.
 	 */
-	cnt_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_CNT);
+	cnt_cur = xfs_allocbt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_CNT);
 	/*
 	 * Have both left and right contiguous neighbors.
 	 * Merge all three into a single free block.
@@ -2490,7 +2491,7 @@ xfs_exact_minlen_extent_available(
 	int			error = 0;
 
 	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, agbp,
-			args->agno, XFS_BTNUM_CNT);
+					args->pag, XFS_BTNUM_CNT);
 	error = xfs_alloc_lookup_ge(cnt_cur, 0, args->minlen, stat);
 	if (error)
 		goto out;
@@ -2693,21 +2694,21 @@ out_no_agbp:
  * Get a block from the freelist.
  * Returns with the buffer for the block gotten.
  */
-int				/* error */
+int
 xfs_alloc_get_freelist(
-	xfs_trans_t	*tp,	/* transaction pointer */
-	struct xfs_buf	*agbp,	/* buffer containing the agf structure */
-	xfs_agblock_t	*bnop,	/* block address retrieved from freelist */
-	int		btreeblk) /* destination is a AGF btree */
+	struct xfs_trans	*tp,
+	struct xfs_buf		*agbp,
+	xfs_agblock_t		*bnop,
+	int			btreeblk)
 {
-	struct xfs_agf	*agf = agbp->b_addr;
-	struct xfs_buf	*agflbp;/* buffer for a.g. freelist structure */
-	xfs_agblock_t	bno;	/* block number returned */
-	__be32		*agfl_bno;
-	int		error;
-	int		logflags;
-	xfs_mount_t	*mp = tp->t_mountp;
-	xfs_perag_t	*pag;	/* per allocation group data */
+	struct xfs_agf		*agf = agbp->b_addr;
+	struct xfs_buf		*agflbp;
+	xfs_agblock_t		bno;
+	__be32			*agfl_bno;
+	int			error;
+	int			logflags;
+	struct xfs_mount	*mp = tp->t_mountp;
+	struct xfs_perag	*pag;
 
 	/*
 	 * Freelist is empty, give up.
@@ -2817,20 +2818,20 @@ xfs_alloc_pagf_init(
 /*
  * Put the block on the freelist for the allocation group.
  */
-int					/* error */
+int
 xfs_alloc_put_freelist(
-	xfs_trans_t		*tp,	/* transaction pointer */
-	struct xfs_buf		*agbp,	/* buffer for a.g. freelist header */
-	struct xfs_buf		*agflbp,/* buffer for a.g. free block array */
-	xfs_agblock_t		bno,	/* block being freed */
-	int			btreeblk) /* block came from a AGF btree */
+	struct xfs_trans	*tp,
+	struct xfs_buf		*agbp,
+	struct xfs_buf		*agflbp,
+	xfs_agblock_t		bno,
+	int			btreeblk)
 {
 	struct xfs_mount	*mp = tp->t_mountp;
 	struct xfs_agf		*agf = agbp->b_addr;
-	__be32			*blockp;/* pointer to array entry */
+	struct xfs_perag	*pag;
+	__be32			*blockp;
 	int			error;
 	int			logflags;
-	xfs_perag_t		*pag;	/* per allocation group data */
 	__be32			*agfl_bno;
 	int			startoff;
 
@@ -3292,7 +3293,7 @@ error0:
 int
 xfs_free_extent_fix_freelist(
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	struct xfs_buf		**agbp)
 {
 	struct xfs_alloc_arg	args;
@@ -3301,7 +3302,8 @@ xfs_free_extent_fix_freelist(
 	memset(&args, 0, sizeof(struct xfs_alloc_arg));
 	args.tp = tp;
 	args.mp = tp->t_mountp;
-	args.agno = agno;
+	args.agno = pag->pag_agno;
+	args.pag = pag;
 
 	/*
 	 * validate that the block number is legal - the enables us to detect
@@ -3310,17 +3312,12 @@ xfs_free_extent_fix_freelist(
 	if (args.agno >= args.mp->m_sb.sb_agcount)
 		return -EFSCORRUPTED;
 
-	args.pag = xfs_perag_get(args.mp, args.agno);
-	ASSERT(args.pag);
-
 	error = xfs_alloc_fix_freelist(&args, XFS_ALLOC_FLAG_FREEING);
 	if (error)
-		goto out;
+		return error;
 
 	*agbp = args.agbp;
-out:
-	xfs_perag_put(args.pag);
-	return error;
+	return 0;
 }
 
 /*
@@ -3344,6 +3341,7 @@ __xfs_free_extent(
 	struct xfs_agf			*agf;
 	int				error;
 	unsigned int			busy_flags = 0;
+	struct xfs_perag		*pag;
 
 	ASSERT(len != 0);
 	ASSERT(type != XFS_AG_RESV_AGFL);
@@ -3352,33 +3350,37 @@ __xfs_free_extent(
 			XFS_ERRTAG_FREE_EXTENT))
 		return -EIO;
 
-	error = xfs_free_extent_fix_freelist(tp, agno, &agbp);
+	pag = xfs_perag_get(mp, agno);
+	error = xfs_free_extent_fix_freelist(tp, pag, &agbp);
 	if (error)
-		return error;
+		goto err;
 	agf = agbp->b_addr;
 
 	if (XFS_IS_CORRUPT(mp, agbno >= mp->m_sb.sb_agblocks)) {
 		error = -EFSCORRUPTED;
-		goto err;
+		goto err_release;
 	}
 
 	/* validate the extent size is legal now we have the agf locked */
 	if (XFS_IS_CORRUPT(mp, agbno + len > be32_to_cpu(agf->agf_length))) {
 		error = -EFSCORRUPTED;
-		goto err;
+		goto err_release;
 	}
 
 	error = xfs_free_ag_extent(tp, agbp, agno, agbno, len, oinfo, type);
 	if (error)
-		goto err;
+		goto err_release;
 
 	if (skip_discard)
 		busy_flags |= XFS_EXTENT_BUSY_SKIP_DISCARD;
-	xfs_extent_busy_insert(tp, agno, agbno, len, busy_flags);
+	xfs_extent_busy_insert(tp, pag, agbno, len, busy_flags);
+	xfs_perag_put(pag);
 	return 0;
 
-err:
+err_release:
 	xfs_trans_brelse(tp, agbp);
+err:
+	xfs_perag_put(pag);
 	return error;
 }
 
diff --git a/fs/xfs/libxfs/xfs_alloc.h b/fs/xfs/libxfs/xfs_alloc.h
index a4427c5775c2..e30900b6f8ba 100644
--- a/fs/xfs/libxfs/xfs_alloc.h
+++ b/fs/xfs/libxfs/xfs_alloc.h
@@ -214,7 +214,7 @@ int xfs_alloc_read_agfl(struct xfs_mount *mp, struct xfs_trans *tp,
 int xfs_free_agfl_block(struct xfs_trans *, xfs_agnumber_t, xfs_agblock_t,
 			struct xfs_buf *, struct xfs_owner_info *);
 int xfs_alloc_fix_freelist(struct xfs_alloc_arg *args, int flags);
-int xfs_free_extent_fix_freelist(struct xfs_trans *tp, xfs_agnumber_t agno,
+int xfs_free_extent_fix_freelist(struct xfs_trans *tp, struct xfs_perag *pag,
 		struct xfs_buf **agbp);
 
 xfs_extlen_t xfs_prealloc_blocks(struct xfs_mount *mp);
diff --git a/fs/xfs/libxfs/xfs_alloc_btree.c b/fs/xfs/libxfs/xfs_alloc_btree.c
index a43e4c50e69b..6b363f78cfa2 100644
--- a/fs/xfs/libxfs/xfs_alloc_btree.c
+++ b/fs/xfs/libxfs/xfs_alloc_btree.c
@@ -9,7 +9,6 @@
 #include "xfs_format.h"
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_btree.h"
 #include "xfs_btree_staging.h"
@@ -19,6 +18,7 @@
 #include "xfs_error.h"
 #include "xfs_trace.h"
 #include "xfs_trans.h"
+#include "xfs_ag.h"
 
 
 STATIC struct xfs_btree_cur *
@@ -26,8 +26,7 @@ xfs_allocbt_dup_cursor(
 	struct xfs_btree_cur	*cur)
 {
 	return xfs_allocbt_init_cursor(cur->bc_mp, cur->bc_tp,
-			cur->bc_ag.agbp, cur->bc_ag.agno,
-			cur->bc_btnum);
+			cur->bc_ag.agbp, cur->bc_ag.pag, cur->bc_btnum);
 }
 
 STATIC void
@@ -39,13 +38,12 @@ xfs_allocbt_set_root(
 	struct xfs_buf		*agbp = cur->bc_ag.agbp;
 	struct xfs_agf		*agf = agbp->b_addr;
 	int			btnum = cur->bc_btnum;
-	struct xfs_perag	*pag = agbp->b_pag;
 
 	ASSERT(ptr->s != 0);
 
 	agf->agf_roots[btnum] = ptr->s;
 	be32_add_cpu(&agf->agf_levels[btnum], inc);
-	pag->pagf_levels[btnum] += inc;
+	cur->bc_ag.pag->pagf_levels[btnum] += inc;
 
 	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
 }
@@ -72,7 +70,7 @@ xfs_allocbt_alloc_block(
 	}
 
 	atomic64_inc(&cur->bc_mp->m_allocbt_blks);
-	xfs_extent_busy_reuse(cur->bc_mp, cur->bc_ag.agno, bno, 1, false);
+	xfs_extent_busy_reuse(cur->bc_mp, cur->bc_ag.agbp->b_pag, bno, 1, false);
 
 	new->s = cpu_to_be32(bno);
 
@@ -86,7 +84,6 @@ xfs_allocbt_free_block(
 	struct xfs_buf		*bp)
 {
 	struct xfs_buf		*agbp = cur->bc_ag.agbp;
-	struct xfs_agf		*agf = agbp->b_addr;
 	xfs_agblock_t		bno;
 	int			error;
 
@@ -96,7 +93,7 @@ xfs_allocbt_free_block(
 		return error;
 
 	atomic64_dec(&cur->bc_mp->m_allocbt_blks);
-	xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1,
+	xfs_extent_busy_insert(cur->bc_tp, agbp->b_pag, bno, 1,
 			      XFS_EXTENT_BUSY_SKIP_DISCARD);
 	return 0;
 }
@@ -225,7 +222,7 @@ xfs_allocbt_init_ptr_from_cur(
 {
 	struct xfs_agf		*agf = cur->bc_ag.agbp->b_addr;
 
-	ASSERT(cur->bc_ag.agno == be32_to_cpu(agf->agf_seqno));
+	ASSERT(cur->bc_ag.pag->pag_agno == be32_to_cpu(agf->agf_seqno));
 
 	ptr->s = agf->agf_roots[cur->bc_btnum];
 }
@@ -473,7 +470,7 @@ STATIC struct xfs_btree_cur *
 xfs_allocbt_init_common(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_btnum_t		btnum)
 {
 	struct xfs_btree_cur	*cur;
@@ -486,6 +483,7 @@ xfs_allocbt_init_common(
 	cur->bc_mp = mp;
 	cur->bc_btnum = btnum;
 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
+	cur->bc_ag.abt.active = false;
 
 	if (btnum == XFS_BTNUM_CNT) {
 		cur->bc_ops = &xfs_cntbt_ops;
@@ -496,8 +494,9 @@ xfs_allocbt_init_common(
 		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtb_2);
 	}
 
-	cur->bc_ag.agno = agno;
-	cur->bc_ag.abt.active = false;
+	/* take a reference for the cursor */
+	atomic_inc(&pag->pag_ref);
+	cur->bc_ag.pag = pag;
 
 	if (xfs_sb_version_hascrc(&mp->m_sb))
 		cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
@@ -513,13 +512,13 @@ xfs_allocbt_init_cursor(
 	struct xfs_mount	*mp,		/* file system mount point */
 	struct xfs_trans	*tp,		/* transaction pointer */
 	struct xfs_buf		*agbp,		/* buffer for agf structure */
-	xfs_agnumber_t		agno,		/* allocation group number */
+	struct xfs_perag	*pag,
 	xfs_btnum_t		btnum)		/* btree identifier */
 {
 	struct xfs_agf		*agf = agbp->b_addr;
 	struct xfs_btree_cur	*cur;
 
-	cur = xfs_allocbt_init_common(mp, tp, agno, btnum);
+	cur = xfs_allocbt_init_common(mp, tp, pag, btnum);
 	if (btnum == XFS_BTNUM_CNT)
 		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]);
 	else
@@ -535,12 +534,12 @@ struct xfs_btree_cur *
 xfs_allocbt_stage_cursor(
 	struct xfs_mount	*mp,
 	struct xbtree_afakeroot	*afake,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_btnum_t		btnum)
 {
 	struct xfs_btree_cur	*cur;
 
-	cur = xfs_allocbt_init_common(mp, NULL, agno, btnum);
+	cur = xfs_allocbt_init_common(mp, NULL, pag, btnum);
 	xfs_btree_stage_afakeroot(cur, afake);
 	return cur;
 }
diff --git a/fs/xfs/libxfs/xfs_alloc_btree.h b/fs/xfs/libxfs/xfs_alloc_btree.h
index a5b998e950fe..9eb4c667a6b8 100644
--- a/fs/xfs/libxfs/xfs_alloc_btree.h
+++ b/fs/xfs/libxfs/xfs_alloc_btree.h
@@ -13,6 +13,7 @@
 struct xfs_buf;
 struct xfs_btree_cur;
 struct xfs_mount;
+struct xfs_perag;
 struct xbtree_afakeroot;
 
 /*
@@ -46,11 +47,11 @@ struct xbtree_afakeroot;
 		 (maxrecs) * sizeof(xfs_alloc_key_t) + \
 		 ((index) - 1) * sizeof(xfs_alloc_ptr_t)))
 
-extern struct xfs_btree_cur *xfs_allocbt_init_cursor(struct xfs_mount *,
-		struct xfs_trans *, struct xfs_buf *,
-		xfs_agnumber_t, xfs_btnum_t);
+extern struct xfs_btree_cur *xfs_allocbt_init_cursor(struct xfs_mount *mp,
+		struct xfs_trans *tp, struct xfs_buf *bp,
+		struct xfs_perag *pag, xfs_btnum_t btnum);
 struct xfs_btree_cur *xfs_allocbt_stage_cursor(struct xfs_mount *mp,
-		struct xbtree_afakeroot *afake, xfs_agnumber_t agno,
+		struct xbtree_afakeroot *afake, struct xfs_perag *pag,
 		xfs_btnum_t btnum);
 extern int xfs_allocbt_maxrecs(struct xfs_mount *, int, int);
 extern xfs_extlen_t xfs_allocbt_calc_size(struct xfs_mount *mp,
diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c
index 556184b63061..aa371d005131 100644
--- a/fs/xfs/libxfs/xfs_attr_leaf.c
+++ b/fs/xfs/libxfs/xfs_attr_leaf.c
@@ -27,6 +27,7 @@
 #include "xfs_buf_item.h"
 #include "xfs_dir2.h"
 #include "xfs_log.h"
+#include "xfs_ag.h"
 
 
 /*
diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
index a3e0e6f672d6..7ae826cfe5a5 100644
--- a/fs/xfs/libxfs/xfs_bmap.c
+++ b/fs/xfs/libxfs/xfs_bmap.c
@@ -31,6 +31,7 @@
 #include "xfs_attr_leaf.h"
 #include "xfs_filestream.h"
 #include "xfs_rmap.h"
+#include "xfs_ag.h"
 #include "xfs_ag_resv.h"
 #include "xfs_refcount.h"
 #include "xfs_icache.h"
diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c
index 5b6fcb9b44e2..be74a6b53689 100644
--- a/fs/xfs/libxfs/xfs_btree.c
+++ b/fs/xfs/libxfs/xfs_btree.c
@@ -21,6 +21,7 @@
 #include "xfs_alloc.h"
 #include "xfs_log.h"
 #include "xfs_btree_staging.h"
+#include "xfs_ag.h"
 
 /*
  * Cursor allocation zone.
@@ -215,7 +216,7 @@ xfs_btree_check_sptr(
 {
 	if (level <= 0)
 		return false;
-	return xfs_verify_agbno(cur->bc_mp, cur->bc_ag.agno, agbno);
+	return xfs_verify_agbno(cur->bc_mp, cur->bc_ag.pag->pag_agno, agbno);
 }
 
 /*
@@ -244,7 +245,7 @@ xfs_btree_check_ptr(
 			return 0;
 		xfs_err(cur->bc_mp,
 "AG %u: Corrupt btree %d pointer at level %d index %d.",
-				cur->bc_ag.agno, cur->bc_btnum,
+				cur->bc_ag.pag->pag_agno, cur->bc_btnum,
 				level, index);
 	}
 
@@ -376,6 +377,8 @@ xfs_btree_del_cursor(
 	       XFS_FORCED_SHUTDOWN(cur->bc_mp));
 	if (unlikely(cur->bc_flags & XFS_BTREE_STAGING))
 		kmem_free(cur->bc_ops);
+	if (!(cur->bc_flags & XFS_BTREE_LONG_PTRS) && cur->bc_ag.pag)
+		xfs_perag_put(cur->bc_ag.pag);
 	kmem_cache_free(xfs_btree_cur_zone, cur);
 }
 
@@ -885,13 +888,13 @@ xfs_btree_readahead_sblock(
 
 
 	if ((lr & XFS_BTCUR_LEFTRA) && left != NULLAGBLOCK) {
-		xfs_btree_reada_bufs(cur->bc_mp, cur->bc_ag.agno,
+		xfs_btree_reada_bufs(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 				     left, 1, cur->bc_ops->buf_ops);
 		rval++;
 	}
 
 	if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLAGBLOCK) {
-		xfs_btree_reada_bufs(cur->bc_mp, cur->bc_ag.agno,
+		xfs_btree_reada_bufs(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 				     right, 1, cur->bc_ops->buf_ops);
 		rval++;
 	}
@@ -949,7 +952,7 @@ xfs_btree_ptr_to_daddr(
 		*daddr = XFS_FSB_TO_DADDR(cur->bc_mp, fsbno);
 	} else {
 		agbno = be32_to_cpu(ptr->s);
-		*daddr = XFS_AGB_TO_DADDR(cur->bc_mp, cur->bc_ag.agno,
+		*daddr = XFS_AGB_TO_DADDR(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 				agbno);
 	}
 
@@ -1150,7 +1153,7 @@ xfs_btree_init_block_cur(
 	if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
 		owner = cur->bc_ino.ip->i_ino;
 	else
-		owner = cur->bc_ag.agno;
+		owner = cur->bc_ag.pag->pag_agno;
 
 	xfs_btree_init_block_int(cur->bc_mp, XFS_BUF_TO_BLOCK(bp), bp->b_bn,
 				 cur->bc_btnum, level, numrecs,
diff --git a/fs/xfs/libxfs/xfs_btree.h b/fs/xfs/libxfs/xfs_btree.h
index 10e50cbacacf..4dbdc659c396 100644
--- a/fs/xfs/libxfs/xfs_btree.h
+++ b/fs/xfs/libxfs/xfs_btree.h
@@ -11,6 +11,7 @@ struct xfs_inode;
 struct xfs_mount;
 struct xfs_trans;
 struct xfs_ifork;
+struct xfs_perag;
 
 extern kmem_zone_t	*xfs_btree_cur_zone;
 
@@ -180,11 +181,11 @@ union xfs_btree_irec {
 
 /* Per-AG btree information. */
 struct xfs_btree_cur_ag {
+	struct xfs_perag	*pag;
 	union {
 		struct xfs_buf		*agbp;
 		struct xbtree_afakeroot	*afake;	/* for staging cursor */
 	};
-	xfs_agnumber_t		agno;
 	union {
 		struct {
 			unsigned long nr_ops;	/* # record updates */
@@ -231,6 +232,13 @@ typedef struct xfs_btree_cur
 	uint8_t		bc_blocklog;	/* log2(blocksize) of btree blocks */
 	xfs_btnum_t	bc_btnum;	/* identifies which btree type */
 	int		bc_statoff;	/* offset of btre stats array */
+
+	/*
+	 * Short btree pointers need an agno to be able to turn the pointers
+	 * into physical addresses for IO, so the btree cursor switches between
+	 * bc_ino and bc_ag based on whether XFS_BTREE_LONG_PTRS is set for the
+	 * cursor.
+	 */
 	union {
 		struct xfs_btree_cur_ag	bc_ag;
 		struct xfs_btree_cur_ino bc_ino;
diff --git a/fs/xfs/libxfs/xfs_ialloc.c b/fs/xfs/libxfs/xfs_ialloc.c
index eefdb518fe64..2ed6de6faf8a 100644
--- a/fs/xfs/libxfs/xfs_ialloc.c
+++ b/fs/xfs/libxfs/xfs_ialloc.c
@@ -10,7 +10,6 @@
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
 #include "xfs_bit.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_inode.h"
 #include "xfs_btree.h"
@@ -27,6 +26,7 @@
 #include "xfs_trace.h"
 #include "xfs_log.h"
 #include "xfs_rmap.h"
+#include "xfs_ag.h"
 
 /*
  * Lookup a record by ino in the btree given by cur.
@@ -105,7 +105,7 @@ xfs_inobt_get_rec(
 	int				*stat)
 {
 	struct xfs_mount		*mp = cur->bc_mp;
-	xfs_agnumber_t			agno = cur->bc_ag.agno;
+	xfs_agnumber_t			agno = cur->bc_ag.pag->pag_agno;
 	union xfs_btree_rec		*rec;
 	int				error;
 	uint64_t			realfree;
@@ -172,18 +172,17 @@ xfs_inobt_insert(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
 	struct xfs_buf		*agbp,
+	struct xfs_perag	*pag,
 	xfs_agino_t		newino,
 	xfs_agino_t		newlen,
 	xfs_btnum_t		btnum)
 {
 	struct xfs_btree_cur	*cur;
-	struct xfs_agi		*agi = agbp->b_addr;
-	xfs_agnumber_t		agno = be32_to_cpu(agi->agi_seqno);
 	xfs_agino_t		thisino;
 	int			i;
 	int			error;
 
-	cur = xfs_inobt_init_cursor(mp, tp, agbp, agno, btnum);
+	cur = xfs_inobt_init_cursor(mp, tp, agbp, pag, btnum);
 
 	for (thisino = newino;
 	     thisino < newino + newlen;
@@ -520,18 +519,17 @@ xfs_inobt_insert_sprec(
 	struct xfs_mount		*mp,
 	struct xfs_trans		*tp,
 	struct xfs_buf			*agbp,
+	struct xfs_perag		*pag,
 	int				btnum,
 	struct xfs_inobt_rec_incore	*nrec,	/* in/out: new/merged rec. */
 	bool				merge)	/* merge or replace */
 {
 	struct xfs_btree_cur		*cur;
-	struct xfs_agi			*agi = agbp->b_addr;
-	xfs_agnumber_t			agno = be32_to_cpu(agi->agi_seqno);
 	int				error;
 	int				i;
 	struct xfs_inobt_rec_incore	rec;
 
-	cur = xfs_inobt_init_cursor(mp, tp, agbp, agno, btnum);
+	cur = xfs_inobt_init_cursor(mp, tp, agbp, pag, btnum);
 
 	/* the new record is pre-aligned so we know where to look */
 	error = xfs_inobt_lookup(cur, nrec->ir_startino, XFS_LOOKUP_EQ, &i);
@@ -578,14 +576,14 @@ xfs_inobt_insert_sprec(
 			goto error;
 		}
 
-		trace_xfs_irec_merge_pre(mp, agno, rec.ir_startino,
+		trace_xfs_irec_merge_pre(mp, pag->pag_agno, rec.ir_startino,
 					 rec.ir_holemask, nrec->ir_startino,
 					 nrec->ir_holemask);
 
 		/* merge to nrec to output the updated record */
 		__xfs_inobt_rec_merge(nrec, &rec);
 
-		trace_xfs_irec_merge_post(mp, agno, nrec->ir_startino,
+		trace_xfs_irec_merge_post(mp, pag->pag_agno, nrec->ir_startino,
 					  nrec->ir_holemask);
 
 		error = xfs_inobt_rec_check_count(mp, nrec);
@@ -606,28 +604,28 @@ error:
 }
 
 /*
- * Allocate new inodes in the allocation group specified by agbp.
- * Returns 0 if inodes were allocated in this AG; 1 if there was no space
- * in this AG; or the usual negative error code.
+ * Allocate new inodes in the allocation group specified by agbp.  Returns 0 if
+ * inodes were allocated in this AG; -EAGAIN if there was no space in this AG so
+ * the caller knows it can try another AG, a hard -ENOSPC when over the maximum
+ * inode count threshold, or the usual negative error code for other errors.
  */
 STATIC int
 xfs_ialloc_ag_alloc(
 	struct xfs_trans	*tp,
-	struct xfs_buf		*agbp)
+	struct xfs_buf		*agbp,
+	struct xfs_perag	*pag)
 {
 	struct xfs_agi		*agi;
 	struct xfs_alloc_arg	args;
-	xfs_agnumber_t		agno;
 	int			error;
 	xfs_agino_t		newino;		/* new first inode's number */
 	xfs_agino_t		newlen;		/* new number of inodes */
 	int			isaligned = 0;	/* inode allocation at stripe */
 						/* unit boundary */
 	/* init. to full chunk */
-	uint16_t		allocmask = (uint16_t) -1;
 	struct xfs_inobt_rec_incore rec;
-	struct xfs_perag	*pag;
 	struct xfs_ino_geometry	*igeo = M_IGEO(tp->t_mountp);
+	uint16_t		allocmask = (uint16_t) -1;
 	int			do_sparse = 0;
 
 	memset(&args, 0, sizeof(args));
@@ -660,14 +658,13 @@ xfs_ialloc_ag_alloc(
 	 */
 	agi = agbp->b_addr;
 	newino = be32_to_cpu(agi->agi_newino);
-	agno = be32_to_cpu(agi->agi_seqno);
 	args.agbno = XFS_AGINO_TO_AGBNO(args.mp, newino) +
 		     igeo->ialloc_blks;
 	if (do_sparse)
 		goto sparse_alloc;
 	if (likely(newino != NULLAGINO &&
 		  (args.agbno < be32_to_cpu(agi->agi_length)))) {
-		args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno);
+		args.fsbno = XFS_AGB_TO_FSB(args.mp, pag->pag_agno, args.agbno);
 		args.type = XFS_ALLOCTYPE_THIS_BNO;
 		args.prod = 1;
 
@@ -727,7 +724,7 @@ xfs_ialloc_ag_alloc(
 		 * For now, just allocate blocks up front.
 		 */
 		args.agbno = be32_to_cpu(agi->agi_root);
-		args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno);
+		args.fsbno = XFS_AGB_TO_FSB(args.mp, pag->pag_agno, args.agbno);
 		/*
 		 * Allocate a fixed-size extent of inodes.
 		 */
@@ -748,7 +745,7 @@ xfs_ialloc_ag_alloc(
 	if (isaligned && args.fsbno == NULLFSBLOCK) {
 		args.type = XFS_ALLOCTYPE_NEAR_BNO;
 		args.agbno = be32_to_cpu(agi->agi_root);
-		args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno);
+		args.fsbno = XFS_AGB_TO_FSB(args.mp, pag->pag_agno, args.agbno);
 		args.alignment = igeo->cluster_align;
 		if ((error = xfs_alloc_vextent(&args)))
 			return error;
@@ -764,7 +761,7 @@ xfs_ialloc_ag_alloc(
 sparse_alloc:
 		args.type = XFS_ALLOCTYPE_NEAR_BNO;
 		args.agbno = be32_to_cpu(agi->agi_root);
-		args.fsbno = XFS_AGB_TO_FSB(args.mp, agno, args.agbno);
+		args.fsbno = XFS_AGB_TO_FSB(args.mp, pag->pag_agno, args.agbno);
 		args.alignment = args.mp->m_sb.sb_spino_align;
 		args.prod = 1;
 
@@ -796,7 +793,7 @@ sparse_alloc:
 	}
 
 	if (args.fsbno == NULLFSBLOCK)
-		return 1;
+		return -EAGAIN;
 
 	ASSERT(args.len == args.minlen);
 
@@ -809,7 +806,7 @@ sparse_alloc:
 	 * rather than a linear progression to prevent the next generation
 	 * number from being easily guessable.
 	 */
-	error = xfs_ialloc_inode_init(args.mp, tp, NULL, newlen, agno,
+	error = xfs_ialloc_inode_init(args.mp, tp, NULL, newlen, pag->pag_agno,
 			args.agbno, args.len, prandom_u32());
 
 	if (error)
@@ -836,12 +833,12 @@ sparse_alloc:
 		 * if necessary. If a merge does occur, rec is updated to the
 		 * merged record.
 		 */
-		error = xfs_inobt_insert_sprec(args.mp, tp, agbp, XFS_BTNUM_INO,
-					       &rec, true);
+		error = xfs_inobt_insert_sprec(args.mp, tp, agbp, pag,
+				XFS_BTNUM_INO, &rec, true);
 		if (error == -EFSCORRUPTED) {
 			xfs_alert(args.mp,
 	"invalid sparse inode record: ino 0x%llx holemask 0x%x count %u",
-				  XFS_AGINO_TO_INO(args.mp, agno,
+				  XFS_AGINO_TO_INO(args.mp, pag->pag_agno,
 						   rec.ir_startino),
 				  rec.ir_holemask, rec.ir_count);
 			xfs_force_shutdown(args.mp, SHUTDOWN_CORRUPT_INCORE);
@@ -861,21 +858,20 @@ sparse_alloc:
 		 * existing record with this one.
 		 */
 		if (xfs_sb_version_hasfinobt(&args.mp->m_sb)) {
-			error = xfs_inobt_insert_sprec(args.mp, tp, agbp,
-						       XFS_BTNUM_FINO, &rec,
-						       false);
+			error = xfs_inobt_insert_sprec(args.mp, tp, agbp, pag,
+				       XFS_BTNUM_FINO, &rec, false);
 			if (error)
 				return error;
 		}
 	} else {
 		/* full chunk - insert new records to both btrees */
-		error = xfs_inobt_insert(args.mp, tp, agbp, newino, newlen,
+		error = xfs_inobt_insert(args.mp, tp, agbp, pag, newino, newlen,
 					 XFS_BTNUM_INO);
 		if (error)
 			return error;
 
 		if (xfs_sb_version_hasfinobt(&args.mp->m_sb)) {
-			error = xfs_inobt_insert(args.mp, tp, agbp, newino,
+			error = xfs_inobt_insert(args.mp, tp, agbp, pag, newino,
 						 newlen, XFS_BTNUM_FINO);
 			if (error)
 				return error;
@@ -887,7 +883,6 @@ sparse_alloc:
 	 */
 	be32_add_cpu(&agi->agi_count, newlen);
 	be32_add_cpu(&agi->agi_freecount, newlen);
-	pag = agbp->b_pag;
 	pag->pagi_freecount += newlen;
 	pag->pagi_count += newlen;
 	agi->agi_newino = cpu_to_be32(newino);
@@ -905,139 +900,6 @@ sparse_alloc:
 	return 0;
 }
 
-STATIC xfs_agnumber_t
-xfs_ialloc_next_ag(
-	xfs_mount_t	*mp)
-{
-	xfs_agnumber_t	agno;
-
-	spin_lock(&mp->m_agirotor_lock);
-	agno = mp->m_agirotor;
-	if (++mp->m_agirotor >= mp->m_maxagi)
-		mp->m_agirotor = 0;
-	spin_unlock(&mp->m_agirotor_lock);
-
-	return agno;
-}
-
-/*
- * Select an allocation group to look for a free inode in, based on the parent
- * inode and the mode.  Return the allocation group buffer.
- */
-STATIC xfs_agnumber_t
-xfs_ialloc_ag_select(
-	xfs_trans_t	*tp,		/* transaction pointer */
-	xfs_ino_t	parent,		/* parent directory inode number */
-	umode_t		mode)		/* bits set to indicate file type */
-{
-	xfs_agnumber_t	agcount;	/* number of ag's in the filesystem */
-	xfs_agnumber_t	agno;		/* current ag number */
-	int		flags;		/* alloc buffer locking flags */
-	xfs_extlen_t	ineed;		/* blocks needed for inode allocation */
-	xfs_extlen_t	longest = 0;	/* longest extent available */
-	xfs_mount_t	*mp;		/* mount point structure */
-	int		needspace;	/* file mode implies space allocated */
-	xfs_perag_t	*pag;		/* per allocation group data */
-	xfs_agnumber_t	pagno;		/* parent (starting) ag number */
-	int		error;
-
-	/*
-	 * Files of these types need at least one block if length > 0
-	 * (and they won't fit in the inode, but that's hard to figure out).
-	 */
-	needspace = S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode);
-	mp = tp->t_mountp;
-	agcount = mp->m_maxagi;
-	if (S_ISDIR(mode))
-		pagno = xfs_ialloc_next_ag(mp);
-	else {
-		pagno = XFS_INO_TO_AGNO(mp, parent);
-		if (pagno >= agcount)
-			pagno = 0;
-	}
-
-	ASSERT(pagno < agcount);
-
-	/*
-	 * Loop through allocation groups, looking for one with a little
-	 * free space in it.  Note we don't look for free inodes, exactly.
-	 * Instead, we include whether there is a need to allocate inodes
-	 * to mean that blocks must be allocated for them,
-	 * if none are currently free.
-	 */
-	agno = pagno;
-	flags = XFS_ALLOC_FLAG_TRYLOCK;
-	for (;;) {
-		pag = xfs_perag_get(mp, agno);
-		if (!pag->pagi_inodeok) {
-			xfs_ialloc_next_ag(mp);
-			goto nextag;
-		}
-
-		if (!pag->pagi_init) {
-			error = xfs_ialloc_pagi_init(mp, tp, agno);
-			if (error)
-				goto nextag;
-		}
-
-		if (pag->pagi_freecount) {
-			xfs_perag_put(pag);
-			return agno;
-		}
-
-		if (!pag->pagf_init) {
-			error = xfs_alloc_pagf_init(mp, tp, agno, flags);
-			if (error)
-				goto nextag;
-		}
-
-		/*
-		 * Check that there is enough free space for the file plus a
-		 * chunk of inodes if we need to allocate some. If this is the
-		 * first pass across the AGs, take into account the potential
-		 * space needed for alignment of inode chunks when checking the
-		 * longest contiguous free space in the AG - this prevents us
-		 * from getting ENOSPC because we have free space larger than
-		 * ialloc_blks but alignment constraints prevent us from using
-		 * it.
-		 *
-		 * If we can't find an AG with space for full alignment slack to
-		 * be taken into account, we must be near ENOSPC in all AGs.
-		 * Hence we don't include alignment for the second pass and so
-		 * if we fail allocation due to alignment issues then it is most
-		 * likely a real ENOSPC condition.
-		 */
-		ineed = M_IGEO(mp)->ialloc_min_blks;
-		if (flags && ineed > 1)
-			ineed += M_IGEO(mp)->cluster_align;
-		longest = pag->pagf_longest;
-		if (!longest)
-			longest = pag->pagf_flcount > 0;
-
-		if (pag->pagf_freeblks >= needspace + ineed &&
-		    longest >= ineed) {
-			xfs_perag_put(pag);
-			return agno;
-		}
-nextag:
-		xfs_perag_put(pag);
-		/*
-		 * No point in iterating over the rest, if we're shutting
-		 * down.
-		 */
-		if (XFS_FORCED_SHUTDOWN(mp))
-			return NULLAGNUMBER;
-		agno++;
-		if (agno >= agcount)
-			agno = 0;
-		if (agno == pagno) {
-			if (flags == 0)
-				return NULLAGNUMBER;
-			flags = 0;
-		}
-	}
-}
-
 /*
  * Try to retrieve the next record to the left/right from the current one.
  */
@@ -1123,15 +985,14 @@ STATIC int
 xfs_dialloc_ag_inobt(
 	struct xfs_trans	*tp,
 	struct xfs_buf		*agbp,
+	struct xfs_perag	*pag,
 	xfs_ino_t		parent,
 	xfs_ino_t		*inop)
 {
 	struct xfs_mount	*mp = tp->t_mountp;
 	struct xfs_agi		*agi = agbp->b_addr;
-	xfs_agnumber_t		agno = be32_to_cpu(agi->agi_seqno);
 	xfs_agnumber_t		pagno = XFS_INO_TO_AGNO(mp, parent);
 	xfs_agino_t		pagino = XFS_INO_TO_AGINO(mp, parent);
-	struct xfs_perag	*pag = agbp->b_pag;
 	struct xfs_btree_cur	*cur, *tcur;
 	struct xfs_inobt_rec_incore rec, trec;
 	xfs_ino_t		ino;
@@ -1145,7 +1006,7 @@ xfs_dialloc_ag_inobt(
 	ASSERT(pag->pagi_freecount > 0);
 
  restart_pagno:
-	cur = xfs_inobt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_INO);
+	cur = xfs_inobt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_INO);
 	/*
 	 * If pagino is 0 (this is the root inode allocation) use newino.
 	 * This must work because we've just allocated some.
@@ -1160,7 +1021,7 @@ xfs_dialloc_ag_inobt(
 	/*
 	 * If in the same AG as the parent, try to get near the parent.
 	 */
-	if (pagno == agno) {
+	if (pagno == pag->pag_agno) {
 		int		doneleft;	/* done, to the left */
 		int		doneright;	/* done, to the right */
 
@@ -1363,7 +1224,7 @@ alloc_inode:
 	ASSERT(offset < XFS_INODES_PER_CHUNK);
 	ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) %
 				   XFS_INODES_PER_CHUNK) == 0);
-	ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino + offset);
+	ino = XFS_AGINO_TO_INO(mp, pag->pag_agno, rec.ir_startino + offset);
 	rec.ir_free &= ~XFS_INOBT_MASK(offset);
 	rec.ir_freecount--;
 	error = xfs_inobt_update(cur, &rec);
@@ -1568,16 +1429,16 @@ xfs_dialloc_ag_update_inobt(
  * The caller selected an AG for us, and made sure that free inodes are
  * available.
  */
-int
+static int
 xfs_dialloc_ag(
 	struct xfs_trans	*tp,
 	struct xfs_buf		*agbp,
+	struct xfs_perag	*pag,
 	xfs_ino_t		parent,
 	xfs_ino_t		*inop)
 {
 	struct xfs_mount		*mp = tp->t_mountp;
 	struct xfs_agi			*agi = agbp->b_addr;
-	xfs_agnumber_t			agno = be32_to_cpu(agi->agi_seqno);
 	xfs_agnumber_t			pagno = XFS_INO_TO_AGNO(mp, parent);
 	xfs_agino_t			pagino = XFS_INO_TO_AGINO(mp, parent);
 	struct xfs_btree_cur		*cur;	/* finobt cursor */
@@ -1589,7 +1450,7 @@ xfs_dialloc_ag(
 	int				i;
 
 	if (!xfs_sb_version_hasfinobt(&mp->m_sb))
-		return xfs_dialloc_ag_inobt(tp, agbp, parent, inop);
+		return xfs_dialloc_ag_inobt(tp, agbp, pag, parent, inop);
 
 	/*
 	 * If pagino is 0 (this is the root inode allocation) use newino.
@@ -1598,7 +1459,7 @@ xfs_dialloc_ag(
 	if (!pagino)
 		pagino = be32_to_cpu(agi->agi_newino);
 
-	cur = xfs_inobt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_FINO);
+	cur = xfs_inobt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_FINO);
 
 	error = xfs_check_agi_freecount(cur, agi);
 	if (error)
@@ -1609,7 +1470,7 @@ xfs_dialloc_ag(
 	 * parent. If so, find the closest available inode to the parent. If
 	 * not, consider the agi hint or find the first free inode in the AG.
 	 */
-	if (agno == pagno)
+	if (pag->pag_agno == pagno)
 		error = xfs_dialloc_ag_finobt_near(pagino, &cur, &rec);
 	else
 		error = xfs_dialloc_ag_finobt_newino(agi, cur, &rec);
@@ -1621,7 +1482,7 @@ xfs_dialloc_ag(
 	ASSERT(offset < XFS_INODES_PER_CHUNK);
 	ASSERT((XFS_AGINO_TO_OFFSET(mp, rec.ir_startino) %
 				   XFS_INODES_PER_CHUNK) == 0);
-	ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino + offset);
+	ino = XFS_AGINO_TO_INO(mp, pag->pag_agno, rec.ir_startino + offset);
 
 	/*
 	 * Modify or remove the finobt record.
@@ -1641,7 +1502,7 @@ xfs_dialloc_ag(
 	 * the original freecount. If all is well, make the equivalent update to
 	 * the inobt using the finobt record and offset information.
 	 */
-	icur = xfs_inobt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_INO);
+	icur = xfs_inobt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_INO);
 
 	error = xfs_check_agi_freecount(icur, agi);
 	if (error)
@@ -1657,7 +1518,7 @@ xfs_dialloc_ag(
 	 */
 	be32_add_cpu(&agi->agi_freecount, -1);
 	xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT);
-	agbp->b_pag->pagi_freecount--;
+	pag->pagi_freecount--;
 
 	xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -1);
 
@@ -1708,54 +1569,195 @@ xfs_dialloc_roll(
 	/* Re-attach the quota info that we detached from prev trx. */
 	tp->t_dqinfo = dqinfo;
 
+	/*
+	 * Join the buffer even on commit error so that the buffer is released
+	 * when the caller cancels the transaction and doesn't have to handle
+	 * this error case specially.
+	 */
+	xfs_trans_bjoin(tp, agibp);
 	*tpp = tp;
+	return error;
+}
+
+static xfs_agnumber_t
+xfs_ialloc_next_ag(
+	xfs_mount_t	*mp)
+{
+	xfs_agnumber_t	agno;
+
+	spin_lock(&mp->m_agirotor_lock);
+	agno = mp->m_agirotor;
+	if (++mp->m_agirotor >= mp->m_maxagi)
+		mp->m_agirotor = 0;
+	spin_unlock(&mp->m_agirotor_lock);
+
+	return agno;
+}
+
+static bool
+xfs_dialloc_good_ag(
+	struct xfs_trans	*tp,
+	struct xfs_perag	*pag,
+	umode_t			mode,
+	int			flags,
+	bool			ok_alloc)
+{
+	struct xfs_mount	*mp = tp->t_mountp;
+	xfs_extlen_t		ineed;
+	xfs_extlen_t		longest = 0;
+	int			needspace;
+	int			error;
+
+	if (!pag->pagi_inodeok)
+		return false;
+
+	if (!pag->pagi_init) {
+		error = xfs_ialloc_pagi_init(mp, tp, pag->pag_agno);
+		if (error)
+			return false;
+	}
+
+	if (pag->pagi_freecount)
+		return true;
+	if (!ok_alloc)
+		return false;
+
+	if (!pag->pagf_init) {
+		error = xfs_alloc_pagf_init(mp, tp, pag->pag_agno, flags);
+		if (error)
+			return false;
+	}
+
+	/*
+	 * Check that there is enough free space for the file plus a chunk of
+	 * inodes if we need to allocate some. If this is the first pass across
+	 * the AGs, take into account the potential space needed for alignment
+	 * of inode chunks when checking the longest contiguous free space in
+	 * the AG - this prevents us from getting ENOSPC because we have free
+	 * space larger than ialloc_blks but alignment constraints prevent us
+	 * from using it.
+	 *
+	 * If we can't find an AG with space for full alignment slack to be
+	 * taken into account, we must be near ENOSPC in all AGs.  Hence we
+	 * don't include alignment for the second pass and so if we fail
+	 * allocation due to alignment issues then it is most likely a real
+	 * ENOSPC condition.
+	 *
+	 * XXX(dgc): this calculation is now bogus thanks to the per-ag
+	 * reservations that xfs_alloc_fix_freelist() now does via
+	 * xfs_alloc_space_available(). When the AG fills up, pagf_freeblks will
+	 * be more than large enough for the check below to succeed, but
+	 * xfs_alloc_space_available() will fail because of the non-zero
+	 * metadata reservation and hence we won't actually be able to allocate
+	 * more inodes in this AG. We do soooo much unnecessary work near ENOSPC
+	 * because of this.
+	 */
+	ineed = M_IGEO(mp)->ialloc_min_blks;
+	if (flags && ineed > 1)
+		ineed += M_IGEO(mp)->cluster_align;
+	longest = pag->pagf_longest;
+	if (!longest)
+		longest = pag->pagf_flcount > 0;
+	needspace = S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode);
+
+	if (pag->pagf_freeblks < needspace + ineed || longest < ineed)
+		return false;
+	return true;
+}
+
+static int
+xfs_dialloc_try_ag(
+	struct xfs_trans	**tpp,
+	struct xfs_perag	*pag,
+	xfs_ino_t		parent,
+	xfs_ino_t		*new_ino,
+	bool			ok_alloc)
+{
+	struct xfs_buf		*agbp;
+	xfs_ino_t		ino;
+	int			error;
+
+	/*
+	 * Then read in the AGI buffer and recheck with the AGI buffer
+	 * lock held.
+	 */
+	error = xfs_ialloc_read_agi(pag->pag_mount, *tpp, pag->pag_agno, &agbp);
 	if (error)
 		return error;
-	xfs_trans_bjoin(tp, agibp);
-	return 0;
+
+	if (!pag->pagi_freecount) {
+		if (!ok_alloc) {
+			error = -EAGAIN;
+			goto out_release;
+		}
+
+		error = xfs_ialloc_ag_alloc(*tpp, agbp, pag);
+		if (error < 0)
+			goto out_release;
+
+		/*
+		 * We successfully allocated space for an inode cluster in this
+		 * AG.  Roll the transaction so that we can allocate one of the
+		 * new inodes.
+		 */
+		ASSERT(pag->pagi_freecount > 0);
+		error = xfs_dialloc_roll(tpp, agbp);
+		if (error)
+			goto out_release;
+	}
+
+	/* Allocate an inode in the found AG */
+	error = xfs_dialloc_ag(*tpp, agbp, pag, parent, &ino);
+	if (!error)
+		*new_ino = ino;
+	return error;
+
+out_release:
+	xfs_trans_brelse(*tpp, agbp);
+	return error;
 }
 
 /*
- * Select and prepare an AG for inode allocation.
+ * Allocate an on-disk inode.
  *
  * Mode is used to tell whether the new inode is a directory and hence where to
- * locate it.
- *
- * This function will ensure that the selected AG has free inodes available to
- * allocate from. The selected AGI will be returned locked to the caller, and it
- * will allocate more free inodes if required. If no free inodes are found or
- * can be allocated, no AGI will be returned.
+ * locate it. The on-disk inode that is allocated will be returned in @new_ino
+ * on success, otherwise an error will be set to indicate the failure (e.g.
+ * -ENOSPC).
  */
 int
-xfs_dialloc_select_ag(
+xfs_dialloc(
 	struct xfs_trans	**tpp,
 	xfs_ino_t		parent,
 	umode_t			mode,
-	struct xfs_buf		**IO_agbp)
+	xfs_ino_t		*new_ino)
 {
 	struct xfs_mount	*mp = (*tpp)->t_mountp;
-	struct xfs_buf		*agbp;
 	xfs_agnumber_t		agno;
-	int			error;
-	bool			noroom = false;
+	int			error = 0;
 	xfs_agnumber_t		start_agno;
 	struct xfs_perag	*pag;
 	struct xfs_ino_geometry	*igeo = M_IGEO(mp);
-	bool			okalloc = true;
-
-	*IO_agbp = NULL;
+	bool			ok_alloc = true;
+	int			flags;
+	xfs_ino_t		ino;
 
 	/*
-	 * We do not have an agbp, so select an initial allocation
-	 * group for inode allocation.
+	 * Directories, symlinks, and regular files frequently allocate at least
+	 * one block, so factor that potential expansion when we examine whether
+	 * an AG has enough space for file creation.
 	 */
-	start_agno = xfs_ialloc_ag_select(*tpp, parent, mode);
-	if (start_agno == NULLAGNUMBER)
-		return 0;
+	if (S_ISDIR(mode))
+		start_agno = xfs_ialloc_next_ag(mp);
+	else {
+		start_agno = XFS_INO_TO_AGNO(mp, parent);
+		if (start_agno >= mp->m_maxagi)
+			start_agno = 0;
+	}
 
 	/*
 	 * If we have already hit the ceiling of inode blocks then clear
-	 * okalloc so we scan all available agi structures for a free
+	 * ok_alloc so we scan all available agi structures for a free
 	 * inode.
 	 *
 	 * Read rough value of mp->m_icount by percpu_counter_read_positive,
@@ -1764,8 +1766,7 @@ xfs_dialloc_select_ag(
 	if (igeo->maxicount &&
 	    percpu_counter_read_positive(&mp->m_icount) + igeo->ialloc_inos
 							> igeo->maxicount) {
-		noroom = true;
-		okalloc = false;
+		ok_alloc = false;
 	}
 
 	/*
@@ -1774,82 +1775,36 @@ xfs_dialloc_select_ag(
 	 * allocation groups upward, wrapping at the end.
 	 */
 	agno = start_agno;
+	flags = XFS_ALLOC_FLAG_TRYLOCK;
 	for (;;) {
 		pag = xfs_perag_get(mp, agno);
-		if (!pag->pagi_inodeok) {
-			xfs_ialloc_next_ag(mp);
-			goto nextag;
-		}
-
-		if (!pag->pagi_init) {
-			error = xfs_ialloc_pagi_init(mp, *tpp, agno);
-			if (error)
+		if (xfs_dialloc_good_ag(*tpp, pag, mode, flags, ok_alloc)) {
+			error = xfs_dialloc_try_ag(tpp, pag, parent,
+					&ino, ok_alloc);
+			if (error != -EAGAIN)
 				break;
 		}
 
-		/*
-		 * Do a first racy fast path check if this AG is usable.
-		 */
-		if (!pag->pagi_freecount && !okalloc)
-			goto nextag;
-
-		/*
-		 * Then read in the AGI buffer and recheck with the AGI buffer
-		 * lock held.
-		 */
-		error = xfs_ialloc_read_agi(mp, *tpp, agno, &agbp);
-		if (error)
-			break;
-
-		if (pag->pagi_freecount) {
-			xfs_perag_put(pag);
-			goto found_ag;
-		}
-
-		if (!okalloc)
-			goto nextag_relse_buffer;
-
-		error = xfs_ialloc_ag_alloc(*tpp, agbp);
-		if (error < 0) {
-			xfs_trans_brelse(*tpp, agbp);
-
-			if (error == -ENOSPC)
-				error = 0;
+		if (XFS_FORCED_SHUTDOWN(mp)) {
+			error = -EFSCORRUPTED;
 			break;
 		}
-
-		if (error == 0) {
-			/*
-			 * We successfully allocated space for an inode cluster
-			 * in this AG.  Roll the transaction so that we can
-			 * allocate one of the new inodes.
-			 */
-			ASSERT(pag->pagi_freecount > 0);
-			xfs_perag_put(pag);
-
-			error = xfs_dialloc_roll(tpp, agbp);
-			if (error) {
-				xfs_buf_relse(agbp);
-				return error;
+		if (++agno == mp->m_maxagi)
+			agno = 0;
+		if (agno == start_agno) {
+			if (!flags) {
+				error = -ENOSPC;
+				break;
 			}
-			goto found_ag;
+			flags = 0;
 		}
-
-nextag_relse_buffer:
-		xfs_trans_brelse(*tpp, agbp);
-nextag:
 		xfs_perag_put(pag);
-		if (++agno == mp->m_sb.sb_agcount)
-			agno = 0;
-		if (agno == start_agno)
-			return noroom ? -ENOSPC : 0;
 	}
 
+	if (!error)
+		*new_ino = ino;
 	xfs_perag_put(pag);
 	return error;
-found_ag:
-	*IO_agbp = agbp;
-	return 0;
 }
 
 /*
@@ -1935,12 +1890,12 @@ xfs_difree_inobt(
 	struct xfs_mount		*mp,
 	struct xfs_trans		*tp,
 	struct xfs_buf			*agbp,
+	struct xfs_perag		*pag,
 	xfs_agino_t			agino,
 	struct xfs_icluster		*xic,
 	struct xfs_inobt_rec_incore	*orec)
 {
 	struct xfs_agi			*agi = agbp->b_addr;
-	xfs_agnumber_t			agno = be32_to_cpu(agi->agi_seqno);
 	struct xfs_btree_cur		*cur;
 	struct xfs_inobt_rec_incore	rec;
 	int				ilen;
@@ -1954,7 +1909,7 @@ xfs_difree_inobt(
 	/*
 	 * Initialize the cursor.
 	 */
-	cur = xfs_inobt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_INO);
+	cur = xfs_inobt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_INO);
 
 	error = xfs_check_agi_freecount(cur, agi);
 	if (error)
@@ -2005,7 +1960,8 @@ xfs_difree_inobt(
 		struct xfs_perag	*pag = agbp->b_pag;
 
 		xic->deleted = true;
-		xic->first_ino = XFS_AGINO_TO_INO(mp, agno, rec.ir_startino);
+		xic->first_ino = XFS_AGINO_TO_INO(mp, pag->pag_agno,
+				rec.ir_startino);
 		xic->alloc = xfs_inobt_irec_to_allocmask(&rec);
 
 		/*
@@ -2028,7 +1984,7 @@ xfs_difree_inobt(
 			goto error0;
 		}
 
-		xfs_difree_inode_chunk(tp, agno, &rec);
+		xfs_difree_inode_chunk(tp, pag->pag_agno, &rec);
 	} else {
 		xic->deleted = false;
 
@@ -2044,7 +2000,7 @@ xfs_difree_inobt(
 		 */
 		be32_add_cpu(&agi->agi_freecount, 1);
 		xfs_ialloc_log_agi(tp, agbp, XFS_AGI_FREECOUNT);
-		agbp->b_pag->pagi_freecount++;
+		pag->pagi_freecount++;
 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, 1);
 	}
 
@@ -2069,18 +2025,18 @@ xfs_difree_finobt(
 	struct xfs_mount		*mp,
 	struct xfs_trans		*tp,
 	struct xfs_buf			*agbp,
+	struct xfs_perag		*pag,
 	xfs_agino_t			agino,
 	struct xfs_inobt_rec_incore	*ibtrec) /* inobt record */
 {
 	struct xfs_agi			*agi = agbp->b_addr;
-	xfs_agnumber_t			agno = be32_to_cpu(agi->agi_seqno);
 	struct xfs_btree_cur		*cur;
 	struct xfs_inobt_rec_incore	rec;
 	int				offset = agino - ibtrec->ir_startino;
 	int				error;
 	int				i;
 
-	cur = xfs_inobt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_FINO);
+	cur = xfs_inobt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_FINO);
 
 	error = xfs_inobt_lookup(cur, ibtrec->ir_startino, XFS_LOOKUP_EQ, &i);
 	if (error)
@@ -2178,36 +2134,33 @@ error:
  */
 int
 xfs_difree(
-	struct xfs_trans	*tp,		/* transaction pointer */
-	xfs_ino_t		inode,		/* inode to be freed */
-	struct xfs_icluster	*xic)	/* cluster info if deleted */
+	struct xfs_trans	*tp,
+	struct xfs_perag	*pag,
+	xfs_ino_t		inode,
+	struct xfs_icluster	*xic)
 {
 	/* REFERENCED */
 	xfs_agblock_t		agbno;	/* block number containing inode */
 	struct xfs_buf		*agbp;	/* buffer for allocation group header */
 	xfs_agino_t		agino;	/* allocation group inode number */
-	xfs_agnumber_t		agno;	/* allocation group number */
 	int			error;	/* error return value */
-	struct xfs_mount	*mp;	/* mount structure for filesystem */
+	struct xfs_mount	*mp = tp->t_mountp;
 	struct xfs_inobt_rec_incore rec;/* btree record */
 
-	mp = tp->t_mountp;
-
 	/*
 	 * Break up inode number into its components.
 	 */
-	agno = XFS_INO_TO_AGNO(mp, inode);
-	if (agno >= mp->m_sb.sb_agcount)  {
-		xfs_warn(mp, "%s: agno >= mp->m_sb.sb_agcount (%d >= %d).",
-			__func__, agno, mp->m_sb.sb_agcount);
+	if (pag->pag_agno != XFS_INO_TO_AGNO(mp, inode)) {
+		xfs_warn(mp, "%s: agno != pag->pag_agno (%d != %d).",
+			__func__, XFS_INO_TO_AGNO(mp, inode), pag->pag_agno);
 		ASSERT(0);
 		return -EINVAL;
 	}
 	agino = XFS_INO_TO_AGINO(mp, inode);
-	if (inode != XFS_AGINO_TO_INO(mp, agno, agino))  {
+	if (inode != XFS_AGINO_TO_INO(mp, pag->pag_agno, agino))  {
 		xfs_warn(mp, "%s: inode != XFS_AGINO_TO_INO() (%llu != %llu).",
 			__func__, (unsigned long long)inode,
-			(unsigned long long)XFS_AGINO_TO_INO(mp, agno, agino));
+			(unsigned long long)XFS_AGINO_TO_INO(mp, pag->pag_agno, agino));
 		ASSERT(0);
 		return -EINVAL;
 	}
@@ -2221,7 +2174,7 @@ xfs_difree(
 	/*
 	 * Get the allocation group header.
 	 */
-	error = xfs_ialloc_read_agi(mp, tp, agno, &agbp);
+	error = xfs_ialloc_read_agi(mp, tp, pag->pag_agno, &agbp);
 	if (error) {
 		xfs_warn(mp, "%s: xfs_ialloc_read_agi() returned error %d.",
 			__func__, error);
@@ -2231,7 +2184,7 @@ xfs_difree(
 	/*
 	 * Fix up the inode allocation btree.
 	 */
-	error = xfs_difree_inobt(mp, tp, agbp, agino, xic, &rec);
+	error = xfs_difree_inobt(mp, tp, agbp, pag, agino, xic, &rec);
 	if (error)
 		goto error0;
 
@@ -2239,7 +2192,7 @@ xfs_difree(
 	 * Fix up the free inode btree.
 	 */
 	if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
-		error = xfs_difree_finobt(mp, tp, agbp, agino, &rec);
+		error = xfs_difree_finobt(mp, tp, agbp, pag, agino, &rec);
 		if (error)
 			goto error0;
 	}
@@ -2254,7 +2207,7 @@ STATIC int
 xfs_imap_lookup(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_agino_t		agino,
 	xfs_agblock_t		agbno,
 	xfs_agblock_t		*chunk_agbno,
@@ -2267,11 +2220,11 @@ xfs_imap_lookup(
 	int			error;
 	int			i;
 
-	error = xfs_ialloc_read_agi(mp, tp, agno, &agbp);
+	error = xfs_ialloc_read_agi(mp, tp, pag->pag_agno, &agbp);
 	if (error) {
 		xfs_alert(mp,
 			"%s: xfs_ialloc_read_agi() returned error %d, agno %d",
-			__func__, error, agno);
+			__func__, error, pag->pag_agno);
 		return error;
 	}
 
@@ -2281,7 +2234,7 @@ xfs_imap_lookup(
 	 * we have a record, we need to ensure it contains the inode number
 	 * we are looking up.
 	 */
-	cur = xfs_inobt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_INO);
+	cur = xfs_inobt_init_cursor(mp, tp, agbp, pag, XFS_BTNUM_INO);
 	error = xfs_inobt_lookup(cur, agino, XFS_LOOKUP_LE, &i);
 	if (!error) {
 		if (i)
@@ -2315,42 +2268,44 @@ xfs_imap_lookup(
  */
 int
 xfs_imap(
-	xfs_mount_t	 *mp,	/* file system mount structure */
-	xfs_trans_t	 *tp,	/* transaction pointer */
-	xfs_ino_t	ino,	/* inode to locate */
-	struct xfs_imap	*imap,	/* location map structure */
-	uint		flags)	/* flags for inode btree lookup */
+	struct xfs_mount	 *mp,	/* file system mount structure */
+	struct xfs_trans	 *tp,	/* transaction pointer */
+	xfs_ino_t		ino,	/* inode to locate */
+	struct xfs_imap		*imap,	/* location map structure */
+	uint			flags)	/* flags for inode btree lookup */
 {
-	xfs_agblock_t	agbno;	/* block number of inode in the alloc group */
-	xfs_agino_t	agino;	/* inode number within alloc group */
-	xfs_agnumber_t	agno;	/* allocation group number */
-	xfs_agblock_t	chunk_agbno;	/* first block in inode chunk */
-	xfs_agblock_t	cluster_agbno;	/* first block in inode cluster */
-	int		error;	/* error code */
-	int		offset;	/* index of inode in its buffer */
-	xfs_agblock_t	offset_agbno;	/* blks from chunk start to inode */
+	xfs_agblock_t		agbno;	/* block number of inode in the alloc group */
+	xfs_agino_t		agino;	/* inode number within alloc group */
+	xfs_agblock_t		chunk_agbno;	/* first block in inode chunk */
+	xfs_agblock_t		cluster_agbno;	/* first block in inode cluster */
+	int			error;	/* error code */
+	int			offset;	/* index of inode in its buffer */
+	xfs_agblock_t		offset_agbno;	/* blks from chunk start to inode */
+	struct xfs_perag	*pag;
 
 	ASSERT(ino != NULLFSINO);
 
 	/*
 	 * Split up the inode number into its parts.
 	 */
-	agno = XFS_INO_TO_AGNO(mp, ino);
+	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ino));
 	agino = XFS_INO_TO_AGINO(mp, ino);
 	agbno = XFS_AGINO_TO_AGBNO(mp, agino);
-	if (agno >= mp->m_sb.sb_agcount || agbno >= mp->m_sb.sb_agblocks ||
-	    ino != XFS_AGINO_TO_INO(mp, agno, agino)) {
+	if (!pag || agbno >= mp->m_sb.sb_agblocks ||
+	    ino != XFS_AGINO_TO_INO(mp, pag->pag_agno, agino)) {
+		error = -EINVAL;
 #ifdef DEBUG
 		/*
 		 * Don't output diagnostic information for untrusted inodes
 		 * as they can be invalid without implying corruption.
 		 */
 		if (flags & XFS_IGET_UNTRUSTED)
-			return -EINVAL;
-		if (agno >= mp->m_sb.sb_agcount) {
+			goto out_drop;
+		if (!pag) {
 			xfs_alert(mp,
 				"%s: agno (%d) >= mp->m_sb.sb_agcount (%d)",
-				__func__, agno, mp->m_sb.sb_agcount);
+				__func__, XFS_INO_TO_AGNO(mp, ino),
+				mp->m_sb.sb_agcount);
 		}
 		if (agbno >= mp->m_sb.sb_agblocks) {
 			xfs_alert(mp,
@@ -2358,15 +2313,15 @@ xfs_imap(
 				__func__, (unsigned long long)agbno,
 				(unsigned long)mp->m_sb.sb_agblocks);
 		}
-		if (ino != XFS_AGINO_TO_INO(mp, agno, agino)) {
+		if (pag && ino != XFS_AGINO_TO_INO(mp, pag->pag_agno, agino)) {
 			xfs_alert(mp,
 		"%s: ino (0x%llx) != XFS_AGINO_TO_INO() (0x%llx)",
 				__func__, ino,
-				XFS_AGINO_TO_INO(mp, agno, agino));
+				XFS_AGINO_TO_INO(mp, pag->pag_agno, agino));
 		}
 		xfs_stack_trace();
 #endif /* DEBUG */
-		return -EINVAL;
+		goto out_drop;
 	}
 
 	/*
@@ -2377,10 +2332,10 @@ xfs_imap(
 	 * in all cases where an untrusted inode number is passed.
 	 */
 	if (flags & XFS_IGET_UNTRUSTED) {
-		error = xfs_imap_lookup(mp, tp, agno, agino, agbno,
+		error = xfs_imap_lookup(mp, tp, pag, agino, agbno,
 					&chunk_agbno, &offset_agbno, flags);
 		if (error)
-			return error;
+			goto out_drop;
 		goto out_map;
 	}
 
@@ -2392,11 +2347,12 @@ xfs_imap(
 		offset = XFS_INO_TO_OFFSET(mp, ino);
 		ASSERT(offset < mp->m_sb.sb_inopblock);
 
-		imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, agbno);
+		imap->im_blkno = XFS_AGB_TO_DADDR(mp, pag->pag_agno, agbno);
 		imap->im_len = XFS_FSB_TO_BB(mp, 1);
 		imap->im_boffset = (unsigned short)(offset <<
 							mp->m_sb.sb_inodelog);
-		return 0;
+		error = 0;
+		goto out_drop;
 	}
 
 	/*
@@ -2408,10 +2364,10 @@ xfs_imap(
 		offset_agbno = agbno & M_IGEO(mp)->inoalign_mask;
 		chunk_agbno = agbno - offset_agbno;
 	} else {
-		error = xfs_imap_lookup(mp, tp, agno, agino, agbno,
+		error = xfs_imap_lookup(mp, tp, pag, agino, agbno,
 					&chunk_agbno, &offset_agbno, flags);
 		if (error)
-			return error;
+			goto out_drop;
 	}
 
 out_map:
@@ -2422,7 +2378,7 @@ out_map:
 	offset = ((agbno - cluster_agbno) * mp->m_sb.sb_inopblock) +
 		XFS_INO_TO_OFFSET(mp, ino);
 
-	imap->im_blkno = XFS_AGB_TO_DADDR(mp, agno, cluster_agbno);
+	imap->im_blkno = XFS_AGB_TO_DADDR(mp, pag->pag_agno, cluster_agbno);
 	imap->im_len = XFS_FSB_TO_BB(mp, M_IGEO(mp)->blocks_per_cluster);
 	imap->im_boffset = (unsigned short)(offset << mp->m_sb.sb_inodelog);
 
@@ -2439,9 +2395,13 @@ out_map:
 			__func__, (unsigned long long) imap->im_blkno,
 			(unsigned long long) imap->im_len,
 			XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks));
-		return -EINVAL;
+		error = -EINVAL;
+		goto out_drop;
 	}
-	return 0;
+	error = 0;
+out_drop:
+	xfs_perag_put(pag);
+	return error;
 }
 
 /*
diff --git a/fs/xfs/libxfs/xfs_ialloc.h b/fs/xfs/libxfs/xfs_ialloc.h
index 3511086a7ae1..9df7c80408ff 100644
--- a/fs/xfs/libxfs/xfs_ialloc.h
+++ b/fs/xfs/libxfs/xfs_ialloc.h
@@ -33,42 +33,14 @@ xfs_make_iptr(struct xfs_mount *mp, struct xfs_buf *b, int o)
 }
 
 /*
- * Allocate an inode on disk.
- * Mode is used to tell whether the new inode will need space, and whether
- * it is a directory.
- *
- * There are two phases to inode allocation: selecting an AG and ensuring
- * that it contains free inodes, followed by allocating one of the free
- * inodes. xfs_dialloc_select_ag() does the former and returns a locked AGI
- * to the caller, ensuring that followup call to xfs_dialloc_ag() will
- * have free inodes to allocate from. xfs_dialloc_ag() will return the inode
- * number of the free inode we allocated.
+ * Allocate an inode on disk.  Mode is used to tell whether the new inode will
+ * need space, and whether it is a directory.
  */
-int					/* error */
-xfs_dialloc_select_ag(
-	struct xfs_trans **tpp,		/* double pointer of transaction */
-	xfs_ino_t	parent,		/* parent inode (directory) */
-	umode_t		mode,		/* mode bits for new inode */
-	struct xfs_buf	**IO_agbp);
-
-int
-xfs_dialloc_ag(
-	struct xfs_trans	*tp,
-	struct xfs_buf		*agbp,
-	xfs_ino_t		parent,
-	xfs_ino_t		*inop);
+int xfs_dialloc(struct xfs_trans **tpp, xfs_ino_t parent, umode_t mode,
+		xfs_ino_t *new_ino);
 
-/*
- * Free disk inode.  Carefully avoids touching the incore inode, all
- * manipulations incore are the caller's responsibility.
- * The on-disk inode is not changed by this operation, only the
- * btree (free inode mask) is changed.
- */
-int					/* error */
-xfs_difree(
-	struct xfs_trans *tp,		/* transaction pointer */
-	xfs_ino_t	inode,		/* inode to be freed */
-	struct xfs_icluster *ifree);	/* cluster info if deleted */
+int xfs_difree(struct xfs_trans *tp, struct xfs_perag *pag,
+		xfs_ino_t ino, struct xfs_icluster *ifree);
 
 /*
  * Return the location of the inode in imap, for mapping it into a buffer.
diff --git a/fs/xfs/libxfs/xfs_ialloc_btree.c b/fs/xfs/libxfs/xfs_ialloc_btree.c
index 4c5831646bd9..823a038939f8 100644
--- a/fs/xfs/libxfs/xfs_ialloc_btree.c
+++ b/fs/xfs/libxfs/xfs_ialloc_btree.c
@@ -20,6 +20,7 @@
 #include "xfs_trace.h"
 #include "xfs_trans.h"
 #include "xfs_rmap.h"
+#include "xfs_ag.h"
 
 STATIC int
 xfs_inobt_get_minrecs(
@@ -34,8 +35,7 @@ xfs_inobt_dup_cursor(
 	struct xfs_btree_cur	*cur)
 {
 	return xfs_inobt_init_cursor(cur->bc_mp, cur->bc_tp,
-			cur->bc_ag.agbp, cur->bc_ag.agno,
-			cur->bc_btnum);
+			cur->bc_ag.agbp, cur->bc_ag.pag, cur->bc_btnum);
 }
 
 STATIC void
@@ -102,7 +102,7 @@ __xfs_inobt_alloc_block(
 	args.tp = cur->bc_tp;
 	args.mp = cur->bc_mp;
 	args.oinfo = XFS_RMAP_OINFO_INOBT;
-	args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_ag.agno, sbno);
+	args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_ag.pag->pag_agno, sbno);
 	args.minlen = 1;
 	args.maxlen = 1;
 	args.prod = 1;
@@ -235,7 +235,7 @@ xfs_inobt_init_ptr_from_cur(
 {
 	struct xfs_agi		*agi = cur->bc_ag.agbp->b_addr;
 
-	ASSERT(cur->bc_ag.agno == be32_to_cpu(agi->agi_seqno));
+	ASSERT(cur->bc_ag.pag->pag_agno == be32_to_cpu(agi->agi_seqno));
 
 	ptr->s = agi->agi_root;
 }
@@ -247,7 +247,7 @@ xfs_finobt_init_ptr_from_cur(
 {
 	struct xfs_agi		*agi = cur->bc_ag.agbp->b_addr;
 
-	ASSERT(cur->bc_ag.agno == be32_to_cpu(agi->agi_seqno));
+	ASSERT(cur->bc_ag.pag->pag_agno == be32_to_cpu(agi->agi_seqno));
 	ptr->s = agi->agi_free_root;
 }
 
@@ -427,7 +427,7 @@ static struct xfs_btree_cur *
 xfs_inobt_init_common(
 	struct xfs_mount	*mp,		/* file system mount point */
 	struct xfs_trans	*tp,		/* transaction pointer */
-	xfs_agnumber_t		agno,		/* allocation group number */
+	struct xfs_perag	*pag,
 	xfs_btnum_t		btnum)		/* ialloc or free ino btree */
 {
 	struct xfs_btree_cur	*cur;
@@ -449,7 +449,9 @@ xfs_inobt_init_common(
 	if (xfs_sb_version_hascrc(&mp->m_sb))
 		cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
 
-	cur->bc_ag.agno = agno;
+	/* take a reference for the cursor */
+	atomic_inc(&pag->pag_ref);
+	cur->bc_ag.pag = pag;
 	return cur;
 }
 
@@ -459,13 +461,13 @@ xfs_inobt_init_cursor(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
 	struct xfs_buf		*agbp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_btnum_t		btnum)
 {
 	struct xfs_btree_cur	*cur;
 	struct xfs_agi		*agi = agbp->b_addr;
 
-	cur = xfs_inobt_init_common(mp, tp, agno, btnum);
+	cur = xfs_inobt_init_common(mp, tp, pag, btnum);
 	if (btnum == XFS_BTNUM_INO)
 		cur->bc_nlevels = be32_to_cpu(agi->agi_level);
 	else
@@ -479,12 +481,12 @@ struct xfs_btree_cur *
 xfs_inobt_stage_cursor(
 	struct xfs_mount	*mp,
 	struct xbtree_afakeroot	*afake,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_btnum_t		btnum)
 {
 	struct xfs_btree_cur	*cur;
 
-	cur = xfs_inobt_init_common(mp, NULL, agno, btnum);
+	cur = xfs_inobt_init_common(mp, NULL, pag, btnum);
 	xfs_btree_stage_afakeroot(cur, afake);
 	return cur;
 }
@@ -656,7 +658,7 @@ int
 xfs_inobt_cur(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_btnum_t		which,
 	struct xfs_btree_cur	**curpp,
 	struct xfs_buf		**agi_bpp)
@@ -667,11 +669,11 @@ xfs_inobt_cur(
 	ASSERT(*agi_bpp == NULL);
 	ASSERT(*curpp == NULL);
 
-	error = xfs_ialloc_read_agi(mp, tp, agno, agi_bpp);
+	error = xfs_ialloc_read_agi(mp, tp, pag->pag_agno, agi_bpp);
 	if (error)
 		return error;
 
-	cur = xfs_inobt_init_cursor(mp, tp, *agi_bpp, agno, which);
+	cur = xfs_inobt_init_cursor(mp, tp, *agi_bpp, pag, which);
 	*curpp = cur;
 	return 0;
 }
@@ -680,7 +682,7 @@ static int
 xfs_inobt_count_blocks(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_btnum_t		btnum,
 	xfs_extlen_t		*tree_blocks)
 {
@@ -688,7 +690,7 @@ xfs_inobt_count_blocks(
 	struct xfs_btree_cur	*cur = NULL;
 	int			error;
 
-	error = xfs_inobt_cur(mp, tp, agno, btnum, &cur, &agbp);
+	error = xfs_inobt_cur(mp, tp, pag, btnum, &cur, &agbp);
 	if (error)
 		return error;
 
@@ -704,14 +706,14 @@ static int
 xfs_finobt_read_blocks(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_extlen_t		*tree_blocks)
 {
 	struct xfs_buf		*agbp;
 	struct xfs_agi		*agi;
 	int			error;
 
-	error = xfs_ialloc_read_agi(mp, tp, agno, &agbp);
+	error = xfs_ialloc_read_agi(mp, tp, pag->pag_agno, &agbp);
 	if (error)
 		return error;
 
@@ -728,7 +730,7 @@ int
 xfs_finobt_calc_reserves(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_extlen_t		*ask,
 	xfs_extlen_t		*used)
 {
@@ -739,14 +741,14 @@ xfs_finobt_calc_reserves(
 		return 0;
 
 	if (xfs_sb_version_hasinobtcounts(&mp->m_sb))
-		error = xfs_finobt_read_blocks(mp, tp, agno, &tree_len);
+		error = xfs_finobt_read_blocks(mp, tp, pag, &tree_len);
 	else
-		error = xfs_inobt_count_blocks(mp, tp, agno, XFS_BTNUM_FINO,
+		error = xfs_inobt_count_blocks(mp, tp, pag, XFS_BTNUM_FINO,
 				&tree_len);
 	if (error)
 		return error;
 
-	*ask += xfs_inobt_max_size(mp, agno);
+	*ask += xfs_inobt_max_size(mp, pag->pag_agno);
 	*used += tree_len;
 	return 0;
 }
diff --git a/fs/xfs/libxfs/xfs_ialloc_btree.h b/fs/xfs/libxfs/xfs_ialloc_btree.h
index 35bbd978c272..e530c82b2217 100644
--- a/fs/xfs/libxfs/xfs_ialloc_btree.h
+++ b/fs/xfs/libxfs/xfs_ialloc_btree.h
@@ -13,6 +13,7 @@
 struct xfs_buf;
 struct xfs_btree_cur;
 struct xfs_mount;
+struct xfs_perag;
 
 /*
  * Btree block header size depends on a superblock flag.
@@ -45,11 +46,11 @@ struct xfs_mount;
 		 (maxrecs) * sizeof(xfs_inobt_key_t) + \
 		 ((index) - 1) * sizeof(xfs_inobt_ptr_t)))
 
-extern struct xfs_btree_cur *xfs_inobt_init_cursor(struct xfs_mount *,
-		struct xfs_trans *, struct xfs_buf *, xfs_agnumber_t,
-		xfs_btnum_t);
+extern struct xfs_btree_cur *xfs_inobt_init_cursor(struct xfs_mount *mp,
+		struct xfs_trans *tp, struct xfs_buf *agbp,
+		struct xfs_perag *pag, xfs_btnum_t btnum);
 struct xfs_btree_cur *xfs_inobt_stage_cursor(struct xfs_mount *mp,
-		struct xbtree_afakeroot *afake, xfs_agnumber_t agno,
+		struct xbtree_afakeroot *afake, struct xfs_perag *pag,
 		xfs_btnum_t btnum);
 extern int xfs_inobt_maxrecs(struct xfs_mount *, int, int);
 
@@ -64,11 +65,11 @@ int xfs_inobt_rec_check_count(struct xfs_mount *,
 #endif	/* DEBUG */
 
 int xfs_finobt_calc_reserves(struct xfs_mount *mp, struct xfs_trans *tp,
-		xfs_agnumber_t agno, xfs_extlen_t *ask, xfs_extlen_t *used);
+		struct xfs_perag *pag, xfs_extlen_t *ask, xfs_extlen_t *used);
 extern xfs_extlen_t xfs_iallocbt_calc_size(struct xfs_mount *mp,
 		unsigned long long len);
 int xfs_inobt_cur(struct xfs_mount *mp, struct xfs_trans *tp,
-		xfs_agnumber_t agno, xfs_btnum_t btnum,
+		struct xfs_perag *pag, xfs_btnum_t btnum,
 		struct xfs_btree_cur **curpp, struct xfs_buf **agi_bpp);
 
 void xfs_inobt_commit_staged_btree(struct xfs_btree_cur *cur,
diff --git a/fs/xfs/libxfs/xfs_refcount.c b/fs/xfs/libxfs/xfs_refcount.c
index 2037b9f23069..860a0c9801ba 100644
--- a/fs/xfs/libxfs/xfs_refcount.c
+++ b/fs/xfs/libxfs/xfs_refcount.c
@@ -22,6 +22,7 @@
 #include "xfs_bit.h"
 #include "xfs_refcount.h"
 #include "xfs_rmap.h"
+#include "xfs_ag.h"
 
 /* Allowable refcount adjustment amounts. */
 enum xfs_refc_adjust_op {
@@ -46,7 +47,7 @@ xfs_refcount_lookup_le(
 	xfs_agblock_t		bno,
 	int			*stat)
 {
-	trace_xfs_refcount_lookup(cur->bc_mp, cur->bc_ag.agno, bno,
+	trace_xfs_refcount_lookup(cur->bc_mp, cur->bc_ag.pag->pag_agno, bno,
 			XFS_LOOKUP_LE);
 	cur->bc_rec.rc.rc_startblock = bno;
 	cur->bc_rec.rc.rc_blockcount = 0;
@@ -63,7 +64,7 @@ xfs_refcount_lookup_ge(
 	xfs_agblock_t		bno,
 	int			*stat)
 {
-	trace_xfs_refcount_lookup(cur->bc_mp, cur->bc_ag.agno, bno,
+	trace_xfs_refcount_lookup(cur->bc_mp, cur->bc_ag.pag->pag_agno, bno,
 			XFS_LOOKUP_GE);
 	cur->bc_rec.rc.rc_startblock = bno;
 	cur->bc_rec.rc.rc_blockcount = 0;
@@ -80,7 +81,7 @@ xfs_refcount_lookup_eq(
 	xfs_agblock_t		bno,
 	int			*stat)
 {
-	trace_xfs_refcount_lookup(cur->bc_mp, cur->bc_ag.agno, bno,
+	trace_xfs_refcount_lookup(cur->bc_mp, cur->bc_ag.pag->pag_agno, bno,
 			XFS_LOOKUP_LE);
 	cur->bc_rec.rc.rc_startblock = bno;
 	cur->bc_rec.rc.rc_blockcount = 0;
@@ -108,7 +109,7 @@ xfs_refcount_get_rec(
 	int				*stat)
 {
 	struct xfs_mount		*mp = cur->bc_mp;
-	xfs_agnumber_t			agno = cur->bc_ag.agno;
+	xfs_agnumber_t			agno = cur->bc_ag.pag->pag_agno;
 	union xfs_btree_rec		*rec;
 	int				error;
 	xfs_agblock_t			realstart;
@@ -119,7 +120,7 @@ xfs_refcount_get_rec(
 
 	xfs_refcount_btrec_to_irec(rec, irec);
 
-	agno = cur->bc_ag.agno;
+	agno = cur->bc_ag.pag->pag_agno;
 	if (irec->rc_blockcount == 0 || irec->rc_blockcount > MAXREFCEXTLEN)
 		goto out_bad_rec;
 
@@ -144,7 +145,7 @@ xfs_refcount_get_rec(
 	if (irec->rc_refcount == 0 || irec->rc_refcount > MAXREFCOUNT)
 		goto out_bad_rec;
 
-	trace_xfs_refcount_get(cur->bc_mp, cur->bc_ag.agno, irec);
+	trace_xfs_refcount_get(cur->bc_mp, cur->bc_ag.pag->pag_agno, irec);
 	return 0;
 
 out_bad_rec:
@@ -169,14 +170,14 @@ xfs_refcount_update(
 	union xfs_btree_rec	rec;
 	int			error;
 
-	trace_xfs_refcount_update(cur->bc_mp, cur->bc_ag.agno, irec);
+	trace_xfs_refcount_update(cur->bc_mp, cur->bc_ag.pag->pag_agno, irec);
 	rec.refc.rc_startblock = cpu_to_be32(irec->rc_startblock);
 	rec.refc.rc_blockcount = cpu_to_be32(irec->rc_blockcount);
 	rec.refc.rc_refcount = cpu_to_be32(irec->rc_refcount);
 	error = xfs_btree_update(cur, &rec);
 	if (error)
 		trace_xfs_refcount_update_error(cur->bc_mp,
-				cur->bc_ag.agno, error, _RET_IP_);
+				cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -193,7 +194,7 @@ xfs_refcount_insert(
 {
 	int				error;
 
-	trace_xfs_refcount_insert(cur->bc_mp, cur->bc_ag.agno, irec);
+	trace_xfs_refcount_insert(cur->bc_mp, cur->bc_ag.pag->pag_agno, irec);
 	cur->bc_rec.rc.rc_startblock = irec->rc_startblock;
 	cur->bc_rec.rc.rc_blockcount = irec->rc_blockcount;
 	cur->bc_rec.rc.rc_refcount = irec->rc_refcount;
@@ -208,7 +209,7 @@ xfs_refcount_insert(
 out_error:
 	if (error)
 		trace_xfs_refcount_insert_error(cur->bc_mp,
-				cur->bc_ag.agno, error, _RET_IP_);
+				cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -234,7 +235,7 @@ xfs_refcount_delete(
 		error = -EFSCORRUPTED;
 		goto out_error;
 	}
-	trace_xfs_refcount_delete(cur->bc_mp, cur->bc_ag.agno, &irec);
+	trace_xfs_refcount_delete(cur->bc_mp, cur->bc_ag.pag->pag_agno, &irec);
 	error = xfs_btree_delete(cur, i);
 	if (XFS_IS_CORRUPT(cur->bc_mp, *i != 1)) {
 		error = -EFSCORRUPTED;
@@ -246,7 +247,7 @@ xfs_refcount_delete(
 out_error:
 	if (error)
 		trace_xfs_refcount_delete_error(cur->bc_mp,
-				cur->bc_ag.agno, error, _RET_IP_);
+				cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -366,7 +367,7 @@ xfs_refcount_split_extent(
 		return 0;
 
 	*shape_changed = true;
-	trace_xfs_refcount_split_extent(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_refcount_split_extent(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			&rcext, agbno);
 
 	/* Establish the right extent. */
@@ -391,7 +392,7 @@ xfs_refcount_split_extent(
 
 out_error:
 	trace_xfs_refcount_split_extent_error(cur->bc_mp,
-			cur->bc_ag.agno, error, _RET_IP_);
+			cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -411,7 +412,7 @@ xfs_refcount_merge_center_extents(
 	int				found_rec;
 
 	trace_xfs_refcount_merge_center_extents(cur->bc_mp,
-			cur->bc_ag.agno, left, center, right);
+			cur->bc_ag.pag->pag_agno, left, center, right);
 
 	/*
 	 * Make sure the center and right extents are not in the btree.
@@ -468,7 +469,7 @@ xfs_refcount_merge_center_extents(
 
 out_error:
 	trace_xfs_refcount_merge_center_extents_error(cur->bc_mp,
-			cur->bc_ag.agno, error, _RET_IP_);
+			cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -487,7 +488,7 @@ xfs_refcount_merge_left_extent(
 	int				found_rec;
 
 	trace_xfs_refcount_merge_left_extent(cur->bc_mp,
-			cur->bc_ag.agno, left, cleft);
+			cur->bc_ag.pag->pag_agno, left, cleft);
 
 	/* If the extent at agbno (cleft) wasn't synthesized, remove it. */
 	if (cleft->rc_refcount > 1) {
@@ -530,7 +531,7 @@ xfs_refcount_merge_left_extent(
 
 out_error:
 	trace_xfs_refcount_merge_left_extent_error(cur->bc_mp,
-			cur->bc_ag.agno, error, _RET_IP_);
+			cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -548,7 +549,7 @@ xfs_refcount_merge_right_extent(
 	int				found_rec;
 
 	trace_xfs_refcount_merge_right_extent(cur->bc_mp,
-			cur->bc_ag.agno, cright, right);
+			cur->bc_ag.pag->pag_agno, cright, right);
 
 	/*
 	 * If the extent ending at agbno+aglen (cright) wasn't synthesized,
@@ -594,7 +595,7 @@ xfs_refcount_merge_right_extent(
 
 out_error:
 	trace_xfs_refcount_merge_right_extent_error(cur->bc_mp,
-			cur->bc_ag.agno, error, _RET_IP_);
+			cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -679,13 +680,13 @@ xfs_refcount_find_left_extents(
 		cleft->rc_blockcount = aglen;
 		cleft->rc_refcount = 1;
 	}
-	trace_xfs_refcount_find_left_extent(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_refcount_find_left_extent(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			left, cleft, agbno);
 	return error;
 
 out_error:
 	trace_xfs_refcount_find_left_extent_error(cur->bc_mp,
-			cur->bc_ag.agno, error, _RET_IP_);
+			cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -768,13 +769,13 @@ xfs_refcount_find_right_extents(
 		cright->rc_blockcount = aglen;
 		cright->rc_refcount = 1;
 	}
-	trace_xfs_refcount_find_right_extent(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_refcount_find_right_extent(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			cright, right, agbno + aglen);
 	return error;
 
 out_error:
 	trace_xfs_refcount_find_right_extent_error(cur->bc_mp,
-			cur->bc_ag.agno, error, _RET_IP_);
+			cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -952,7 +953,7 @@ xfs_refcount_adjust_extents(
 					ext.rc_startblock - *agbno);
 			tmp.rc_refcount = 1 + adj;
 			trace_xfs_refcount_modify_extent(cur->bc_mp,
-					cur->bc_ag.agno, &tmp);
+					cur->bc_ag.pag->pag_agno, &tmp);
 
 			/*
 			 * Either cover the hole (increment) or
@@ -971,7 +972,7 @@ xfs_refcount_adjust_extents(
 				cur->bc_ag.refc.nr_ops++;
 			} else {
 				fsbno = XFS_AGB_TO_FSB(cur->bc_mp,
-						cur->bc_ag.agno,
+						cur->bc_ag.pag->pag_agno,
 						tmp.rc_startblock);
 				xfs_bmap_add_free(cur->bc_tp, fsbno,
 						  tmp.rc_blockcount, oinfo);
@@ -998,7 +999,7 @@ xfs_refcount_adjust_extents(
 			goto skip;
 		ext.rc_refcount += adj;
 		trace_xfs_refcount_modify_extent(cur->bc_mp,
-				cur->bc_ag.agno, &ext);
+				cur->bc_ag.pag->pag_agno, &ext);
 		if (ext.rc_refcount > 1) {
 			error = xfs_refcount_update(cur, &ext);
 			if (error)
@@ -1016,7 +1017,7 @@ xfs_refcount_adjust_extents(
 			goto advloop;
 		} else {
 			fsbno = XFS_AGB_TO_FSB(cur->bc_mp,
-					cur->bc_ag.agno,
+					cur->bc_ag.pag->pag_agno,
 					ext.rc_startblock);
 			xfs_bmap_add_free(cur->bc_tp, fsbno, ext.rc_blockcount,
 					  oinfo);
@@ -1035,7 +1036,7 @@ advloop:
 	return error;
 out_error:
 	trace_xfs_refcount_modify_extent_error(cur->bc_mp,
-			cur->bc_ag.agno, error, _RET_IP_);
+			cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -1057,10 +1058,10 @@ xfs_refcount_adjust(
 	*new_agbno = agbno;
 	*new_aglen = aglen;
 	if (adj == XFS_REFCOUNT_ADJUST_INCREASE)
-		trace_xfs_refcount_increase(cur->bc_mp, cur->bc_ag.agno,
+		trace_xfs_refcount_increase(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 				agbno, aglen);
 	else
-		trace_xfs_refcount_decrease(cur->bc_mp, cur->bc_ag.agno,
+		trace_xfs_refcount_decrease(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 				agbno, aglen);
 
 	/*
@@ -1099,7 +1100,7 @@ xfs_refcount_adjust(
 	return 0;
 
 out_error:
-	trace_xfs_refcount_adjust_error(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_refcount_adjust_error(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			error, _RET_IP_);
 	return error;
 }
@@ -1142,30 +1143,30 @@ xfs_refcount_finish_one(
 	struct xfs_btree_cur		*rcur;
 	struct xfs_buf			*agbp = NULL;
 	int				error = 0;
-	xfs_agnumber_t			agno;
 	xfs_agblock_t			bno;
 	xfs_agblock_t			new_agbno;
 	unsigned long			nr_ops = 0;
 	int				shape_changes = 0;
+	struct xfs_perag		*pag;
 
-	agno = XFS_FSB_TO_AGNO(mp, startblock);
-	ASSERT(agno != NULLAGNUMBER);
+	pag = xfs_perag_get(mp, XFS_FSB_TO_AGNO(mp, startblock));
 	bno = XFS_FSB_TO_AGBNO(mp, startblock);
 
 	trace_xfs_refcount_deferred(mp, XFS_FSB_TO_AGNO(mp, startblock),
 			type, XFS_FSB_TO_AGBNO(mp, startblock),
 			blockcount);
 
-	if (XFS_TEST_ERROR(false, mp,
-			XFS_ERRTAG_REFCOUNT_FINISH_ONE))
-		return -EIO;
+	if (XFS_TEST_ERROR(false, mp, XFS_ERRTAG_REFCOUNT_FINISH_ONE)) {
+		error = -EIO;
+		goto out_drop;
+	}
 
 	/*
 	 * If we haven't gotten a cursor or the cursor AG doesn't match
 	 * the startblock, get one now.
 	 */
 	rcur = *pcur;
-	if (rcur != NULL && rcur->bc_ag.agno != agno) {
+	if (rcur != NULL && rcur->bc_ag.pag != pag) {
 		nr_ops = rcur->bc_ag.refc.nr_ops;
 		shape_changes = rcur->bc_ag.refc.shape_changes;
 		xfs_refcount_finish_one_cleanup(tp, rcur, 0);
@@ -1173,12 +1174,12 @@ xfs_refcount_finish_one(
 		*pcur = NULL;
 	}
 	if (rcur == NULL) {
-		error = xfs_alloc_read_agf(tp->t_mountp, tp, agno,
+		error = xfs_alloc_read_agf(tp->t_mountp, tp, pag->pag_agno,
 				XFS_ALLOC_FLAG_FREEING, &agbp);
 		if (error)
-			return error;
+			goto out_drop;
 
-		rcur = xfs_refcountbt_init_cursor(mp, tp, agbp, agno);
+		rcur = xfs_refcountbt_init_cursor(mp, tp, agbp, pag);
 		rcur->bc_ag.refc.nr_ops = nr_ops;
 		rcur->bc_ag.refc.shape_changes = shape_changes;
 	}
@@ -1188,12 +1189,12 @@ xfs_refcount_finish_one(
 	case XFS_REFCOUNT_INCREASE:
 		error = xfs_refcount_adjust(rcur, bno, blockcount, &new_agbno,
 			new_len, XFS_REFCOUNT_ADJUST_INCREASE, NULL);
-		*new_fsb = XFS_AGB_TO_FSB(mp, agno, new_agbno);
+		*new_fsb = XFS_AGB_TO_FSB(mp, pag->pag_agno, new_agbno);
 		break;
 	case XFS_REFCOUNT_DECREASE:
 		error = xfs_refcount_adjust(rcur, bno, blockcount, &new_agbno,
 			new_len, XFS_REFCOUNT_ADJUST_DECREASE, NULL);
-		*new_fsb = XFS_AGB_TO_FSB(mp, agno, new_agbno);
+		*new_fsb = XFS_AGB_TO_FSB(mp, pag->pag_agno, new_agbno);
 		break;
 	case XFS_REFCOUNT_ALLOC_COW:
 		*new_fsb = startblock + blockcount;
@@ -1210,8 +1211,10 @@ xfs_refcount_finish_one(
 		error = -EFSCORRUPTED;
 	}
 	if (!error && *new_len > 0)
-		trace_xfs_refcount_finish_one_leftover(mp, agno, type,
+		trace_xfs_refcount_finish_one_leftover(mp, pag->pag_agno, type,
 				bno, blockcount, new_agbno, *new_len);
+out_drop:
+	xfs_perag_put(pag);
 	return error;
 }
 
@@ -1294,7 +1297,7 @@ xfs_refcount_find_shared(
 	int				have;
 	int				error;
 
-	trace_xfs_refcount_find_shared(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_refcount_find_shared(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			agbno, aglen);
 
 	/* By default, skip the whole range */
@@ -1374,12 +1377,12 @@ xfs_refcount_find_shared(
 
 done:
 	trace_xfs_refcount_find_shared_result(cur->bc_mp,
-			cur->bc_ag.agno, *fbno, *flen);
+			cur->bc_ag.pag->pag_agno, *fbno, *flen);
 
 out_error:
 	if (error)
 		trace_xfs_refcount_find_shared_error(cur->bc_mp,
-				cur->bc_ag.agno, error, _RET_IP_);
+				cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -1476,7 +1479,7 @@ xfs_refcount_adjust_cow_extents(
 		tmp.rc_blockcount = aglen;
 		tmp.rc_refcount = 1;
 		trace_xfs_refcount_modify_extent(cur->bc_mp,
-				cur->bc_ag.agno, &tmp);
+				cur->bc_ag.pag->pag_agno, &tmp);
 
 		error = xfs_refcount_insert(cur, &tmp,
 				&found_tmp);
@@ -1504,7 +1507,7 @@ xfs_refcount_adjust_cow_extents(
 
 		ext.rc_refcount = 0;
 		trace_xfs_refcount_modify_extent(cur->bc_mp,
-				cur->bc_ag.agno, &ext);
+				cur->bc_ag.pag->pag_agno, &ext);
 		error = xfs_refcount_delete(cur, &found_rec);
 		if (error)
 			goto out_error;
@@ -1520,7 +1523,7 @@ xfs_refcount_adjust_cow_extents(
 	return error;
 out_error:
 	trace_xfs_refcount_modify_extent_error(cur->bc_mp,
-			cur->bc_ag.agno, error, _RET_IP_);
+			cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -1566,7 +1569,7 @@ xfs_refcount_adjust_cow(
 	return 0;
 
 out_error:
-	trace_xfs_refcount_adjust_cow_error(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_refcount_adjust_cow_error(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			error, _RET_IP_);
 	return error;
 }
@@ -1580,7 +1583,7 @@ __xfs_refcount_cow_alloc(
 	xfs_agblock_t		agbno,
 	xfs_extlen_t		aglen)
 {
-	trace_xfs_refcount_cow_increase(rcur->bc_mp, rcur->bc_ag.agno,
+	trace_xfs_refcount_cow_increase(rcur->bc_mp, rcur->bc_ag.pag->pag_agno,
 			agbno, aglen);
 
 	/* Add refcount btree reservation */
@@ -1597,7 +1600,7 @@ __xfs_refcount_cow_free(
 	xfs_agblock_t		agbno,
 	xfs_extlen_t		aglen)
 {
-	trace_xfs_refcount_cow_decrease(rcur->bc_mp, rcur->bc_ag.agno,
+	trace_xfs_refcount_cow_decrease(rcur->bc_mp, rcur->bc_ag.pag->pag_agno,
 			agbno, aglen);
 
 	/* Remove refcount btree reservation */
@@ -1672,7 +1675,7 @@ xfs_refcount_recover_extent(
 int
 xfs_refcount_recover_cow_leftovers(
 	struct xfs_mount		*mp,
-	xfs_agnumber_t			agno)
+	struct xfs_perag		*pag)
 {
 	struct xfs_trans		*tp;
 	struct xfs_btree_cur		*cur;
@@ -1704,10 +1707,10 @@ xfs_refcount_recover_cow_leftovers(
 	if (error)
 		return error;
 
-	error = xfs_alloc_read_agf(mp, tp, agno, 0, &agbp);
+	error = xfs_alloc_read_agf(mp, tp, pag->pag_agno, 0, &agbp);
 	if (error)
 		goto out_trans;
-	cur = xfs_refcountbt_init_cursor(mp, tp, agbp, agno);
+	cur = xfs_refcountbt_init_cursor(mp, tp, agbp, pag);
 
 	/* Find all the leftover CoW staging extents. */
 	memset(&low, 0, sizeof(low));
@@ -1729,11 +1732,12 @@ xfs_refcount_recover_cow_leftovers(
 		if (error)
 			goto out_free;
 
-		trace_xfs_refcount_recover_extent(mp, agno, &rr->rr_rrec);
+		trace_xfs_refcount_recover_extent(mp, pag->pag_agno,
+				&rr->rr_rrec);
 
 		/* Free the orphan record */
 		agbno = rr->rr_rrec.rc_startblock - XFS_REFC_COW_START;
-		fsb = XFS_AGB_TO_FSB(mp, agno, agbno);
+		fsb = XFS_AGB_TO_FSB(mp, pag->pag_agno, agbno);
 		xfs_refcount_free_cow_extent(tp, fsb,
 				rr->rr_rrec.rc_blockcount);
 
diff --git a/fs/xfs/libxfs/xfs_refcount.h b/fs/xfs/libxfs/xfs_refcount.h
index 209795539c8d..9f6e9aae4da0 100644
--- a/fs/xfs/libxfs/xfs_refcount.h
+++ b/fs/xfs/libxfs/xfs_refcount.h
@@ -6,6 +6,13 @@
 #ifndef __XFS_REFCOUNT_H__
 #define __XFS_REFCOUNT_H__
 
+struct xfs_trans;
+struct xfs_mount;
+struct xfs_perag;
+struct xfs_btree_cur;
+struct xfs_bmbt_irec;
+struct xfs_refcount_irec;
+
 extern int xfs_refcount_lookup_le(struct xfs_btree_cur *cur,
 		xfs_agblock_t bno, int *stat);
 extern int xfs_refcount_lookup_ge(struct xfs_btree_cur *cur,
@@ -50,7 +57,7 @@ void xfs_refcount_alloc_cow_extent(struct xfs_trans *tp, xfs_fsblock_t fsb,
 void xfs_refcount_free_cow_extent(struct xfs_trans *tp, xfs_fsblock_t fsb,
 		xfs_extlen_t len);
 extern int xfs_refcount_recover_cow_leftovers(struct xfs_mount *mp,
-		xfs_agnumber_t agno);
+		struct xfs_perag *pag);
 
 /*
  * While we're adjusting the refcounts records of an extent, we have
diff --git a/fs/xfs/libxfs/xfs_refcount_btree.c b/fs/xfs/libxfs/xfs_refcount_btree.c
index a6ac60ae9421..92d336c17e83 100644
--- a/fs/xfs/libxfs/xfs_refcount_btree.c
+++ b/fs/xfs/libxfs/xfs_refcount_btree.c
@@ -9,7 +9,6 @@
 #include "xfs_format.h"
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_btree.h"
 #include "xfs_btree_staging.h"
@@ -20,13 +19,14 @@
 #include "xfs_trans.h"
 #include "xfs_bit.h"
 #include "xfs_rmap.h"
+#include "xfs_ag.h"
 
 static struct xfs_btree_cur *
 xfs_refcountbt_dup_cursor(
 	struct xfs_btree_cur	*cur)
 {
 	return xfs_refcountbt_init_cursor(cur->bc_mp, cur->bc_tp,
-			cur->bc_ag.agbp, cur->bc_ag.agno);
+			cur->bc_ag.agbp, cur->bc_ag.pag);
 }
 
 STATIC void
@@ -65,7 +65,7 @@ xfs_refcountbt_alloc_block(
 	args.tp = cur->bc_tp;
 	args.mp = cur->bc_mp;
 	args.type = XFS_ALLOCTYPE_NEAR_BNO;
-	args.fsbno = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_ag.agno,
+	args.fsbno = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			xfs_refc_block(args.mp));
 	args.oinfo = XFS_RMAP_OINFO_REFC;
 	args.minlen = args.maxlen = args.prod = 1;
@@ -74,13 +74,13 @@ xfs_refcountbt_alloc_block(
 	error = xfs_alloc_vextent(&args);
 	if (error)
 		goto out_error;
-	trace_xfs_refcountbt_alloc_block(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_refcountbt_alloc_block(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			args.agbno, 1);
 	if (args.fsbno == NULLFSBLOCK) {
 		*stat = 0;
 		return 0;
 	}
-	ASSERT(args.agno == cur->bc_ag.agno);
+	ASSERT(args.agno == cur->bc_ag.pag->pag_agno);
 	ASSERT(args.len == 1);
 
 	new->s = cpu_to_be32(args.agbno);
@@ -105,7 +105,7 @@ xfs_refcountbt_free_block(
 	xfs_fsblock_t		fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
 	int			error;
 
-	trace_xfs_refcountbt_free_block(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_refcountbt_free_block(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			XFS_FSB_TO_AGBNO(cur->bc_mp, fsbno), 1);
 	be32_add_cpu(&agf->agf_refcount_blocks, -1);
 	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_REFCOUNT_BLOCKS);
@@ -170,7 +170,7 @@ xfs_refcountbt_init_ptr_from_cur(
 {
 	struct xfs_agf		*agf = cur->bc_ag.agbp->b_addr;
 
-	ASSERT(cur->bc_ag.agno == be32_to_cpu(agf->agf_seqno));
+	ASSERT(cur->bc_ag.pag->pag_agno == be32_to_cpu(agf->agf_seqno));
 
 	ptr->s = agf->agf_refcount_root;
 }
@@ -316,12 +316,11 @@ static struct xfs_btree_cur *
 xfs_refcountbt_init_common(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno)
+	struct xfs_perag	*pag)
 {
 	struct xfs_btree_cur	*cur;
 
-	ASSERT(agno != NULLAGNUMBER);
-	ASSERT(agno < mp->m_sb.sb_agcount);
+	ASSERT(pag->pag_agno < mp->m_sb.sb_agcount);
 
 	cur = kmem_cache_zalloc(xfs_btree_cur_zone, GFP_NOFS | __GFP_NOFAIL);
 	cur->bc_tp = tp;
@@ -330,9 +329,12 @@ xfs_refcountbt_init_common(
 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
 	cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_refcbt_2);
 
-	cur->bc_ag.agno = agno;
 	cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
 
+	/* take a reference for the cursor */
+	atomic_inc(&pag->pag_ref);
+	cur->bc_ag.pag = pag;
+
 	cur->bc_ag.refc.nr_ops = 0;
 	cur->bc_ag.refc.shape_changes = 0;
 	cur->bc_ops = &xfs_refcountbt_ops;
@@ -345,12 +347,12 @@ xfs_refcountbt_init_cursor(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
 	struct xfs_buf		*agbp,
-	xfs_agnumber_t		agno)
+	struct xfs_perag	*pag)
 {
 	struct xfs_agf		*agf = agbp->b_addr;
 	struct xfs_btree_cur	*cur;
 
-	cur = xfs_refcountbt_init_common(mp, tp, agno);
+	cur = xfs_refcountbt_init_common(mp, tp, pag);
 	cur->bc_nlevels = be32_to_cpu(agf->agf_refcount_level);
 	cur->bc_ag.agbp = agbp;
 	return cur;
@@ -361,11 +363,11 @@ struct xfs_btree_cur *
 xfs_refcountbt_stage_cursor(
 	struct xfs_mount	*mp,
 	struct xbtree_afakeroot	*afake,
-	xfs_agnumber_t		agno)
+	struct xfs_perag	*pag)
 {
 	struct xfs_btree_cur	*cur;
 
-	cur = xfs_refcountbt_init_common(mp, NULL, agno);
+	cur = xfs_refcountbt_init_common(mp, NULL, pag);
 	xfs_btree_stage_afakeroot(cur, afake);
 	return cur;
 }
@@ -450,7 +452,7 @@ int
 xfs_refcountbt_calc_reserves(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_extlen_t		*ask,
 	xfs_extlen_t		*used)
 {
@@ -463,8 +465,7 @@ xfs_refcountbt_calc_reserves(
 	if (!xfs_sb_version_hasreflink(&mp->m_sb))
 		return 0;
 
-
-	error = xfs_alloc_read_agf(mp, tp, agno, 0, &agbp);
+	error = xfs_alloc_read_agf(mp, tp, pag->pag_agno, 0, &agbp);
 	if (error)
 		return error;
 
@@ -479,7 +480,7 @@ xfs_refcountbt_calc_reserves(
 	 * expansion.  We therefore can pretend the space isn't there.
 	 */
 	if (mp->m_sb.sb_logstart &&
-	    XFS_FSB_TO_AGNO(mp, mp->m_sb.sb_logstart) == agno)
+	    XFS_FSB_TO_AGNO(mp, mp->m_sb.sb_logstart) == pag->pag_agno)
 		agblocks -= mp->m_sb.sb_logblocks;
 
 	*ask += xfs_refcountbt_max_size(mp, agblocks);
diff --git a/fs/xfs/libxfs/xfs_refcount_btree.h b/fs/xfs/libxfs/xfs_refcount_btree.h
index 69dc515db671..bd9ed9e1e41f 100644
--- a/fs/xfs/libxfs/xfs_refcount_btree.h
+++ b/fs/xfs/libxfs/xfs_refcount_btree.h
@@ -13,6 +13,7 @@
 struct xfs_buf;
 struct xfs_btree_cur;
 struct xfs_mount;
+struct xfs_perag;
 struct xbtree_afakeroot;
 
 /*
@@ -46,9 +47,9 @@ struct xbtree_afakeroot;
 
 extern struct xfs_btree_cur *xfs_refcountbt_init_cursor(struct xfs_mount *mp,
 		struct xfs_trans *tp, struct xfs_buf *agbp,
-		xfs_agnumber_t agno);
+		struct xfs_perag *pag);
 struct xfs_btree_cur *xfs_refcountbt_stage_cursor(struct xfs_mount *mp,
-		struct xbtree_afakeroot *afake, xfs_agnumber_t agno);
+		struct xbtree_afakeroot *afake, struct xfs_perag *pag);
 extern int xfs_refcountbt_maxrecs(int blocklen, bool leaf);
 extern void xfs_refcountbt_compute_maxlevels(struct xfs_mount *mp);
 
@@ -58,7 +59,7 @@ extern xfs_extlen_t xfs_refcountbt_max_size(struct xfs_mount *mp,
 		xfs_agblock_t agblocks);
 
 extern int xfs_refcountbt_calc_reserves(struct xfs_mount *mp,
-		struct xfs_trans *tp, xfs_agnumber_t agno, xfs_extlen_t *ask,
+		struct xfs_trans *tp, struct xfs_perag *pag, xfs_extlen_t *ask,
 		xfs_extlen_t *used);
 
 void xfs_refcountbt_commit_staged_btree(struct xfs_btree_cur *cur,
diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c
index 10e0cf9949a2..d1dfad0204e3 100644
--- a/fs/xfs/libxfs/xfs_rmap.c
+++ b/fs/xfs/libxfs/xfs_rmap.c
@@ -11,6 +11,7 @@
 #include "xfs_trans_resv.h"
 #include "xfs_bit.h"
 #include "xfs_mount.h"
+#include "xfs_sb.h"
 #include "xfs_defer.h"
 #include "xfs_btree.h"
 #include "xfs_trans.h"
@@ -21,6 +22,7 @@
 #include "xfs_errortag.h"
 #include "xfs_error.h"
 #include "xfs_inode.h"
+#include "xfs_ag.h"
 
 /*
  * Lookup the first record less than or equal to [bno, len, owner, offset]
@@ -79,7 +81,7 @@ xfs_rmap_update(
 	union xfs_btree_rec	rec;
 	int			error;
 
-	trace_xfs_rmap_update(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_rmap_update(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 			irec->rm_startblock, irec->rm_blockcount,
 			irec->rm_owner, irec->rm_offset, irec->rm_flags);
 
@@ -91,7 +93,7 @@ xfs_rmap_update(
 	error = xfs_btree_update(cur, &rec);
 	if (error)
 		trace_xfs_rmap_update_error(cur->bc_mp,
-				cur->bc_ag.agno, error, _RET_IP_);
+				cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -107,7 +109,7 @@ xfs_rmap_insert(
 	int			i;
 	int			error;
 
-	trace_xfs_rmap_insert(rcur->bc_mp, rcur->bc_ag.agno, agbno,
+	trace_xfs_rmap_insert(rcur->bc_mp, rcur->bc_ag.pag->pag_agno, agbno,
 			len, owner, offset, flags);
 
 	error = xfs_rmap_lookup_eq(rcur, agbno, len, owner, offset, flags, &i);
@@ -133,7 +135,7 @@ xfs_rmap_insert(
 done:
 	if (error)
 		trace_xfs_rmap_insert_error(rcur->bc_mp,
-				rcur->bc_ag.agno, error, _RET_IP_);
+				rcur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -149,7 +151,7 @@ xfs_rmap_delete(
 	int			i;
 	int			error;
 
-	trace_xfs_rmap_delete(rcur->bc_mp, rcur->bc_ag.agno, agbno,
+	trace_xfs_rmap_delete(rcur->bc_mp, rcur->bc_ag.pag->pag_agno, agbno,
 			len, owner, offset, flags);
 
 	error = xfs_rmap_lookup_eq(rcur, agbno, len, owner, offset, flags, &i);
@@ -170,7 +172,7 @@ xfs_rmap_delete(
 done:
 	if (error)
 		trace_xfs_rmap_delete_error(rcur->bc_mp,
-				rcur->bc_ag.agno, error, _RET_IP_);
+				rcur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -197,7 +199,7 @@ xfs_rmap_get_rec(
 	int			*stat)
 {
 	struct xfs_mount	*mp = cur->bc_mp;
-	xfs_agnumber_t		agno = cur->bc_ag.agno;
+	xfs_agnumber_t		agno = cur->bc_ag.pag->pag_agno;
 	union xfs_btree_rec	*rec;
 	int			error;
 
@@ -260,7 +262,7 @@ xfs_rmap_find_left_neighbor_helper(
 	struct xfs_find_left_neighbor_info	*info = priv;
 
 	trace_xfs_rmap_find_left_neighbor_candidate(cur->bc_mp,
-			cur->bc_ag.agno, rec->rm_startblock,
+			cur->bc_ag.pag->pag_agno, rec->rm_startblock,
 			rec->rm_blockcount, rec->rm_owner, rec->rm_offset,
 			rec->rm_flags);
 
@@ -312,7 +314,7 @@ xfs_rmap_find_left_neighbor(
 	info.stat = stat;
 
 	trace_xfs_rmap_find_left_neighbor_query(cur->bc_mp,
-			cur->bc_ag.agno, bno, 0, owner, offset, flags);
+			cur->bc_ag.pag->pag_agno, bno, 0, owner, offset, flags);
 
 	error = xfs_rmap_query_range(cur, &info.high, &info.high,
 			xfs_rmap_find_left_neighbor_helper, &info);
@@ -320,7 +322,7 @@ xfs_rmap_find_left_neighbor(
 		error = 0;
 	if (*stat)
 		trace_xfs_rmap_find_left_neighbor_result(cur->bc_mp,
-				cur->bc_ag.agno, irec->rm_startblock,
+				cur->bc_ag.pag->pag_agno, irec->rm_startblock,
 				irec->rm_blockcount, irec->rm_owner,
 				irec->rm_offset, irec->rm_flags);
 	return error;
@@ -336,7 +338,7 @@ xfs_rmap_lookup_le_range_helper(
 	struct xfs_find_left_neighbor_info	*info = priv;
 
 	trace_xfs_rmap_lookup_le_range_candidate(cur->bc_mp,
-			cur->bc_ag.agno, rec->rm_startblock,
+			cur->bc_ag.pag->pag_agno, rec->rm_startblock,
 			rec->rm_blockcount, rec->rm_owner, rec->rm_offset,
 			rec->rm_flags);
 
@@ -385,14 +387,14 @@ xfs_rmap_lookup_le_range(
 	info.stat = stat;
 
 	trace_xfs_rmap_lookup_le_range(cur->bc_mp,
-			cur->bc_ag.agno, bno, 0, owner, offset, flags);
+			cur->bc_ag.pag->pag_agno, bno, 0, owner, offset, flags);
 	error = xfs_rmap_query_range(cur, &info.high, &info.high,
 			xfs_rmap_lookup_le_range_helper, &info);
 	if (error == -ECANCELED)
 		error = 0;
 	if (*stat)
 		trace_xfs_rmap_lookup_le_range_result(cur->bc_mp,
-				cur->bc_ag.agno, irec->rm_startblock,
+				cur->bc_ag.pag->pag_agno, irec->rm_startblock,
 				irec->rm_blockcount, irec->rm_owner,
 				irec->rm_offset, irec->rm_flags);
 	return error;
@@ -498,7 +500,7 @@ xfs_rmap_unmap(
 			(flags & XFS_RMAP_BMBT_BLOCK);
 	if (unwritten)
 		flags |= XFS_RMAP_UNWRITTEN;
-	trace_xfs_rmap_unmap(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_unmap(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 
 	/*
@@ -522,7 +524,7 @@ xfs_rmap_unmap(
 		goto out_error;
 	}
 	trace_xfs_rmap_lookup_le_range_result(cur->bc_mp,
-			cur->bc_ag.agno, ltrec.rm_startblock,
+			cur->bc_ag.pag->pag_agno, ltrec.rm_startblock,
 			ltrec.rm_blockcount, ltrec.rm_owner,
 			ltrec.rm_offset, ltrec.rm_flags);
 	ltoff = ltrec.rm_offset;
@@ -588,7 +590,7 @@ xfs_rmap_unmap(
 
 	if (ltrec.rm_startblock == bno && ltrec.rm_blockcount == len) {
 		/* exact match, simply remove the record from rmap tree */
-		trace_xfs_rmap_delete(mp, cur->bc_ag.agno,
+		trace_xfs_rmap_delete(mp, cur->bc_ag.pag->pag_agno,
 				ltrec.rm_startblock, ltrec.rm_blockcount,
 				ltrec.rm_owner, ltrec.rm_offset,
 				ltrec.rm_flags);
@@ -666,7 +668,7 @@ xfs_rmap_unmap(
 		else
 			cur->bc_rec.r.rm_offset = offset + len;
 		cur->bc_rec.r.rm_flags = flags;
-		trace_xfs_rmap_insert(mp, cur->bc_ag.agno,
+		trace_xfs_rmap_insert(mp, cur->bc_ag.pag->pag_agno,
 				cur->bc_rec.r.rm_startblock,
 				cur->bc_rec.r.rm_blockcount,
 				cur->bc_rec.r.rm_owner,
@@ -678,11 +680,11 @@ xfs_rmap_unmap(
 	}
 
 out_done:
-	trace_xfs_rmap_unmap_done(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_unmap_done(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 out_error:
 	if (error)
-		trace_xfs_rmap_unmap_error(mp, cur->bc_ag.agno,
+		trace_xfs_rmap_unmap_error(mp, cur->bc_ag.pag->pag_agno,
 				error, _RET_IP_);
 	return error;
 }
@@ -694,7 +696,7 @@ int
 xfs_rmap_free(
 	struct xfs_trans		*tp,
 	struct xfs_buf			*agbp,
-	xfs_agnumber_t			agno,
+	struct xfs_perag		*pag,
 	xfs_agblock_t			bno,
 	xfs_extlen_t			len,
 	const struct xfs_owner_info	*oinfo)
@@ -706,7 +708,7 @@ xfs_rmap_free(
 	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
 		return 0;
 
-	cur = xfs_rmapbt_init_cursor(mp, tp, agbp, agno);
+	cur = xfs_rmapbt_init_cursor(mp, tp, agbp, pag);
 
 	error = xfs_rmap_unmap(cur, bno, len, false, oinfo);
 
@@ -773,7 +775,7 @@ xfs_rmap_map(
 			(flags & XFS_RMAP_BMBT_BLOCK);
 	if (unwritten)
 		flags |= XFS_RMAP_UNWRITTEN;
-	trace_xfs_rmap_map(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_map(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 	ASSERT(!xfs_rmap_should_skip_owner_update(oinfo));
 
@@ -795,7 +797,7 @@ xfs_rmap_map(
 			goto out_error;
 		}
 		trace_xfs_rmap_lookup_le_range_result(cur->bc_mp,
-				cur->bc_ag.agno, ltrec.rm_startblock,
+				cur->bc_ag.pag->pag_agno, ltrec.rm_startblock,
 				ltrec.rm_blockcount, ltrec.rm_owner,
 				ltrec.rm_offset, ltrec.rm_flags);
 
@@ -831,7 +833,7 @@ xfs_rmap_map(
 			goto out_error;
 		}
 		trace_xfs_rmap_find_right_neighbor_result(cur->bc_mp,
-			cur->bc_ag.agno, gtrec.rm_startblock,
+			cur->bc_ag.pag->pag_agno, gtrec.rm_startblock,
 			gtrec.rm_blockcount, gtrec.rm_owner,
 			gtrec.rm_offset, gtrec.rm_flags);
 		if (!xfs_rmap_is_mergeable(&gtrec, owner, flags))
@@ -870,7 +872,7 @@ xfs_rmap_map(
 			 * result: |rrrrrrrrrrrrrrrrrrrrrrrrrrrrr|
 			 */
 			ltrec.rm_blockcount += gtrec.rm_blockcount;
-			trace_xfs_rmap_delete(mp, cur->bc_ag.agno,
+			trace_xfs_rmap_delete(mp, cur->bc_ag.pag->pag_agno,
 					gtrec.rm_startblock,
 					gtrec.rm_blockcount,
 					gtrec.rm_owner,
@@ -921,7 +923,7 @@ xfs_rmap_map(
 		cur->bc_rec.r.rm_owner = owner;
 		cur->bc_rec.r.rm_offset = offset;
 		cur->bc_rec.r.rm_flags = flags;
-		trace_xfs_rmap_insert(mp, cur->bc_ag.agno, bno, len,
+		trace_xfs_rmap_insert(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			owner, offset, flags);
 		error = xfs_btree_insert(cur, &i);
 		if (error)
@@ -932,11 +934,11 @@ xfs_rmap_map(
 		}
 	}
 
-	trace_xfs_rmap_map_done(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_map_done(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 out_error:
 	if (error)
-		trace_xfs_rmap_map_error(mp, cur->bc_ag.agno,
+		trace_xfs_rmap_map_error(mp, cur->bc_ag.pag->pag_agno,
 				error, _RET_IP_);
 	return error;
 }
@@ -948,7 +950,7 @@ int
 xfs_rmap_alloc(
 	struct xfs_trans		*tp,
 	struct xfs_buf			*agbp,
-	xfs_agnumber_t			agno,
+	struct xfs_perag		*pag,
 	xfs_agblock_t			bno,
 	xfs_extlen_t			len,
 	const struct xfs_owner_info	*oinfo)
@@ -960,7 +962,7 @@ xfs_rmap_alloc(
 	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
 		return 0;
 
-	cur = xfs_rmapbt_init_cursor(mp, tp, agbp, agno);
+	cur = xfs_rmapbt_init_cursor(mp, tp, agbp, pag);
 	error = xfs_rmap_map(cur, bno, len, false, oinfo);
 
 	xfs_btree_del_cursor(cur, error);
@@ -1010,7 +1012,7 @@ xfs_rmap_convert(
 			(flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK))));
 	oldext = unwritten ? XFS_RMAP_UNWRITTEN : 0;
 	new_endoff = offset + len;
-	trace_xfs_rmap_convert(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_convert(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 
 	/*
@@ -1034,7 +1036,7 @@ xfs_rmap_convert(
 		goto done;
 	}
 	trace_xfs_rmap_lookup_le_range_result(cur->bc_mp,
-			cur->bc_ag.agno, PREV.rm_startblock,
+			cur->bc_ag.pag->pag_agno, PREV.rm_startblock,
 			PREV.rm_blockcount, PREV.rm_owner,
 			PREV.rm_offset, PREV.rm_flags);
 
@@ -1076,7 +1078,7 @@ xfs_rmap_convert(
 			goto done;
 		}
 		trace_xfs_rmap_find_left_neighbor_result(cur->bc_mp,
-				cur->bc_ag.agno, LEFT.rm_startblock,
+				cur->bc_ag.pag->pag_agno, LEFT.rm_startblock,
 				LEFT.rm_blockcount, LEFT.rm_owner,
 				LEFT.rm_offset, LEFT.rm_flags);
 		if (LEFT.rm_startblock + LEFT.rm_blockcount == bno &&
@@ -1114,7 +1116,7 @@ xfs_rmap_convert(
 			goto done;
 		}
 		trace_xfs_rmap_find_right_neighbor_result(cur->bc_mp,
-				cur->bc_ag.agno, RIGHT.rm_startblock,
+				cur->bc_ag.pag->pag_agno, RIGHT.rm_startblock,
 				RIGHT.rm_blockcount, RIGHT.rm_owner,
 				RIGHT.rm_offset, RIGHT.rm_flags);
 		if (bno + len == RIGHT.rm_startblock &&
@@ -1132,7 +1134,7 @@ xfs_rmap_convert(
 	     RIGHT.rm_blockcount > XFS_RMAP_LEN_MAX)
 		state &= ~RMAP_RIGHT_CONTIG;
 
-	trace_xfs_rmap_convert_state(mp, cur->bc_ag.agno, state,
+	trace_xfs_rmap_convert_state(mp, cur->bc_ag.pag->pag_agno, state,
 			_RET_IP_);
 
 	/* reset the cursor back to PREV */
@@ -1162,7 +1164,7 @@ xfs_rmap_convert(
 			error = -EFSCORRUPTED;
 			goto done;
 		}
-		trace_xfs_rmap_delete(mp, cur->bc_ag.agno,
+		trace_xfs_rmap_delete(mp, cur->bc_ag.pag->pag_agno,
 				RIGHT.rm_startblock, RIGHT.rm_blockcount,
 				RIGHT.rm_owner, RIGHT.rm_offset,
 				RIGHT.rm_flags);
@@ -1180,7 +1182,7 @@ xfs_rmap_convert(
 			error = -EFSCORRUPTED;
 			goto done;
 		}
-		trace_xfs_rmap_delete(mp, cur->bc_ag.agno,
+		trace_xfs_rmap_delete(mp, cur->bc_ag.pag->pag_agno,
 				PREV.rm_startblock, PREV.rm_blockcount,
 				PREV.rm_owner, PREV.rm_offset,
 				PREV.rm_flags);
@@ -1210,7 +1212,7 @@ xfs_rmap_convert(
 		 * Setting all of a previous oldext extent to newext.
 		 * The left neighbor is contiguous, the right is not.
 		 */
-		trace_xfs_rmap_delete(mp, cur->bc_ag.agno,
+		trace_xfs_rmap_delete(mp, cur->bc_ag.pag->pag_agno,
 				PREV.rm_startblock, PREV.rm_blockcount,
 				PREV.rm_owner, PREV.rm_offset,
 				PREV.rm_flags);
@@ -1247,7 +1249,7 @@ xfs_rmap_convert(
 			error = -EFSCORRUPTED;
 			goto done;
 		}
-		trace_xfs_rmap_delete(mp, cur->bc_ag.agno,
+		trace_xfs_rmap_delete(mp, cur->bc_ag.pag->pag_agno,
 				RIGHT.rm_startblock, RIGHT.rm_blockcount,
 				RIGHT.rm_owner, RIGHT.rm_offset,
 				RIGHT.rm_flags);
@@ -1326,7 +1328,7 @@ xfs_rmap_convert(
 		NEW.rm_blockcount = len;
 		NEW.rm_flags = newext;
 		cur->bc_rec.r = NEW;
-		trace_xfs_rmap_insert(mp, cur->bc_ag.agno, bno,
+		trace_xfs_rmap_insert(mp, cur->bc_ag.pag->pag_agno, bno,
 				len, owner, offset, newext);
 		error = xfs_btree_insert(cur, &i);
 		if (error)
@@ -1383,7 +1385,7 @@ xfs_rmap_convert(
 		NEW.rm_blockcount = len;
 		NEW.rm_flags = newext;
 		cur->bc_rec.r = NEW;
-		trace_xfs_rmap_insert(mp, cur->bc_ag.agno, bno,
+		trace_xfs_rmap_insert(mp, cur->bc_ag.pag->pag_agno, bno,
 				len, owner, offset, newext);
 		error = xfs_btree_insert(cur, &i);
 		if (error)
@@ -1414,7 +1416,7 @@ xfs_rmap_convert(
 		NEW = PREV;
 		NEW.rm_blockcount = offset - PREV.rm_offset;
 		cur->bc_rec.r = NEW;
-		trace_xfs_rmap_insert(mp, cur->bc_ag.agno,
+		trace_xfs_rmap_insert(mp, cur->bc_ag.pag->pag_agno,
 				NEW.rm_startblock, NEW.rm_blockcount,
 				NEW.rm_owner, NEW.rm_offset,
 				NEW.rm_flags);
@@ -1441,7 +1443,7 @@ xfs_rmap_convert(
 		/* new middle extent - newext */
 		cur->bc_rec.r.rm_flags &= ~XFS_RMAP_UNWRITTEN;
 		cur->bc_rec.r.rm_flags |= newext;
-		trace_xfs_rmap_insert(mp, cur->bc_ag.agno, bno, len,
+		trace_xfs_rmap_insert(mp, cur->bc_ag.pag->pag_agno, bno, len,
 				owner, offset, newext);
 		error = xfs_btree_insert(cur, &i);
 		if (error)
@@ -1465,12 +1467,12 @@ xfs_rmap_convert(
 		ASSERT(0);
 	}
 
-	trace_xfs_rmap_convert_done(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_convert_done(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 done:
 	if (error)
 		trace_xfs_rmap_convert_error(cur->bc_mp,
-				cur->bc_ag.agno, error, _RET_IP_);
+				cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -1506,7 +1508,7 @@ xfs_rmap_convert_shared(
 			(flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK))));
 	oldext = unwritten ? XFS_RMAP_UNWRITTEN : 0;
 	new_endoff = offset + len;
-	trace_xfs_rmap_convert(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_convert(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 
 	/*
@@ -1573,7 +1575,7 @@ xfs_rmap_convert_shared(
 			goto done;
 		}
 		trace_xfs_rmap_find_right_neighbor_result(cur->bc_mp,
-				cur->bc_ag.agno, RIGHT.rm_startblock,
+				cur->bc_ag.pag->pag_agno, RIGHT.rm_startblock,
 				RIGHT.rm_blockcount, RIGHT.rm_owner,
 				RIGHT.rm_offset, RIGHT.rm_flags);
 		if (xfs_rmap_is_mergeable(&RIGHT, owner, newext))
@@ -1589,7 +1591,7 @@ xfs_rmap_convert_shared(
 	     RIGHT.rm_blockcount > XFS_RMAP_LEN_MAX)
 		state &= ~RMAP_RIGHT_CONTIG;
 
-	trace_xfs_rmap_convert_state(mp, cur->bc_ag.agno, state,
+	trace_xfs_rmap_convert_state(mp, cur->bc_ag.pag->pag_agno, state,
 			_RET_IP_);
 	/*
 	 * Switch out based on the FILLING and CONTIG state bits.
@@ -1880,12 +1882,12 @@ xfs_rmap_convert_shared(
 		ASSERT(0);
 	}
 
-	trace_xfs_rmap_convert_done(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_convert_done(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 done:
 	if (error)
 		trace_xfs_rmap_convert_error(cur->bc_mp,
-				cur->bc_ag.agno, error, _RET_IP_);
+				cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -1923,7 +1925,7 @@ xfs_rmap_unmap_shared(
 	xfs_owner_info_unpack(oinfo, &owner, &offset, &flags);
 	if (unwritten)
 		flags |= XFS_RMAP_UNWRITTEN;
-	trace_xfs_rmap_unmap(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_unmap(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 
 	/*
@@ -2072,12 +2074,12 @@ xfs_rmap_unmap_shared(
 			goto out_error;
 	}
 
-	trace_xfs_rmap_unmap_done(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_unmap_done(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 out_error:
 	if (error)
 		trace_xfs_rmap_unmap_error(cur->bc_mp,
-				cur->bc_ag.agno, error, _RET_IP_);
+				cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -2112,7 +2114,7 @@ xfs_rmap_map_shared(
 	xfs_owner_info_unpack(oinfo, &owner, &offset, &flags);
 	if (unwritten)
 		flags |= XFS_RMAP_UNWRITTEN;
-	trace_xfs_rmap_map(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_map(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 
 	/* Is there a left record that abuts our range? */
@@ -2138,7 +2140,7 @@ xfs_rmap_map_shared(
 			goto out_error;
 		}
 		trace_xfs_rmap_find_right_neighbor_result(cur->bc_mp,
-			cur->bc_ag.agno, gtrec.rm_startblock,
+			cur->bc_ag.pag->pag_agno, gtrec.rm_startblock,
 			gtrec.rm_blockcount, gtrec.rm_owner,
 			gtrec.rm_offset, gtrec.rm_flags);
 
@@ -2231,12 +2233,12 @@ xfs_rmap_map_shared(
 			goto out_error;
 	}
 
-	trace_xfs_rmap_map_done(mp, cur->bc_ag.agno, bno, len,
+	trace_xfs_rmap_map_done(mp, cur->bc_ag.pag->pag_agno, bno, len,
 			unwritten, oinfo);
 out_error:
 	if (error)
 		trace_xfs_rmap_map_error(cur->bc_mp,
-				cur->bc_ag.agno, error, _RET_IP_);
+				cur->bc_ag.pag->pag_agno, error, _RET_IP_);
 	return error;
 }
 
@@ -2362,31 +2364,32 @@ xfs_rmap_finish_one(
 	struct xfs_btree_cur		**pcur)
 {
 	struct xfs_mount		*mp = tp->t_mountp;
+	struct xfs_perag		*pag;
 	struct xfs_btree_cur		*rcur;
 	struct xfs_buf			*agbp = NULL;
 	int				error = 0;
-	xfs_agnumber_t			agno;
 	struct xfs_owner_info		oinfo;
 	xfs_agblock_t			bno;
 	bool				unwritten;
 
-	agno = XFS_FSB_TO_AGNO(mp, startblock);
-	ASSERT(agno != NULLAGNUMBER);
+	pag = xfs_perag_get(mp, XFS_FSB_TO_AGNO(mp, startblock));
 	bno = XFS_FSB_TO_AGBNO(mp, startblock);
 
-	trace_xfs_rmap_deferred(mp, agno, type, bno, owner, whichfork,
+	trace_xfs_rmap_deferred(mp, pag->pag_agno, type, bno, owner, whichfork,
 			startoff, blockcount, state);
 
-	if (XFS_TEST_ERROR(false, mp,
-			XFS_ERRTAG_RMAP_FINISH_ONE))
-		return -EIO;
+	if (XFS_TEST_ERROR(false, mp, XFS_ERRTAG_RMAP_FINISH_ONE)) {
+		error = -EIO;
+		goto out_drop;
+	}
+
 
 	/*
 	 * If we haven't gotten a cursor or the cursor AG doesn't match
 	 * the startblock, get one now.
 	 */
 	rcur = *pcur;
-	if (rcur != NULL && rcur->bc_ag.agno != agno) {
+	if (rcur != NULL && rcur->bc_ag.pag != pag) {
 		xfs_rmap_finish_one_cleanup(tp, rcur, 0);
 		rcur = NULL;
 		*pcur = NULL;
@@ -2397,13 +2400,15 @@ xfs_rmap_finish_one(
 		 * rmapbt, because a shape change could cause us to
 		 * allocate blocks.
 		 */
-		error = xfs_free_extent_fix_freelist(tp, agno, &agbp);
+		error = xfs_free_extent_fix_freelist(tp, pag, &agbp);
 		if (error)
-			return error;
-		if (XFS_IS_CORRUPT(tp->t_mountp, !agbp))
-			return -EFSCORRUPTED;
+			goto out_drop;
+		if (XFS_IS_CORRUPT(tp->t_mountp, !agbp)) {
+			error = -EFSCORRUPTED;
+			goto out_drop;
+		}
 
-		rcur = xfs_rmapbt_init_cursor(mp, tp, agbp, agno);
+		rcur = xfs_rmapbt_init_cursor(mp, tp, agbp, pag);
 	}
 	*pcur = rcur;
 
@@ -2441,6 +2446,8 @@ xfs_rmap_finish_one(
 		ASSERT(0);
 		error = -EFSCORRUPTED;
 	}
+out_drop:
+	xfs_perag_put(pag);
 	return error;
 }
 
diff --git a/fs/xfs/libxfs/xfs_rmap.h b/fs/xfs/libxfs/xfs_rmap.h
index abe633403fd1..f2423cf7f1e2 100644
--- a/fs/xfs/libxfs/xfs_rmap.h
+++ b/fs/xfs/libxfs/xfs_rmap.h
@@ -6,6 +6,8 @@
 #ifndef __XFS_RMAP_H__
 #define __XFS_RMAP_H__
 
+struct xfs_perag;
+
 static inline void
 xfs_rmap_ino_bmbt_owner(
 	struct xfs_owner_info	*oi,
@@ -113,10 +115,10 @@ xfs_owner_info_pack(
 }
 
 int xfs_rmap_alloc(struct xfs_trans *tp, struct xfs_buf *agbp,
-		   xfs_agnumber_t agno, xfs_agblock_t bno, xfs_extlen_t len,
+		   struct xfs_perag *pag, xfs_agblock_t bno, xfs_extlen_t len,
 		   const struct xfs_owner_info *oinfo);
 int xfs_rmap_free(struct xfs_trans *tp, struct xfs_buf *agbp,
-		  xfs_agnumber_t agno, xfs_agblock_t bno, xfs_extlen_t len,
+		  struct xfs_perag *pag, xfs_agblock_t bno, xfs_extlen_t len,
 		  const struct xfs_owner_info *oinfo);
 
 int xfs_rmap_lookup_le(struct xfs_btree_cur *cur, xfs_agblock_t bno,
diff --git a/fs/xfs/libxfs/xfs_rmap_btree.c b/fs/xfs/libxfs/xfs_rmap_btree.c
index 9f5bcbd834c3..f29bc71b9950 100644
--- a/fs/xfs/libxfs/xfs_rmap_btree.c
+++ b/fs/xfs/libxfs/xfs_rmap_btree.c
@@ -9,7 +9,6 @@
 #include "xfs_format.h"
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_trans.h"
 #include "xfs_alloc.h"
@@ -20,6 +19,7 @@
 #include "xfs_trace.h"
 #include "xfs_error.h"
 #include "xfs_extent_busy.h"
+#include "xfs_ag.h"
 #include "xfs_ag_resv.h"
 
 /*
@@ -52,7 +52,7 @@ xfs_rmapbt_dup_cursor(
 	struct xfs_btree_cur	*cur)
 {
 	return xfs_rmapbt_init_cursor(cur->bc_mp, cur->bc_tp,
-			cur->bc_ag.agbp, cur->bc_ag.agno);
+				cur->bc_ag.agbp, cur->bc_ag.pag);
 }
 
 STATIC void
@@ -64,13 +64,12 @@ xfs_rmapbt_set_root(
 	struct xfs_buf		*agbp = cur->bc_ag.agbp;
 	struct xfs_agf		*agf = agbp->b_addr;
 	int			btnum = cur->bc_btnum;
-	struct xfs_perag	*pag = agbp->b_pag;
 
 	ASSERT(ptr->s != 0);
 
 	agf->agf_roots[btnum] = ptr->s;
 	be32_add_cpu(&agf->agf_levels[btnum], inc);
-	pag->pagf_levels[btnum] += inc;
+	cur->bc_ag.pag->pagf_levels[btnum] += inc;
 
 	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
 }
@@ -84,6 +83,7 @@ xfs_rmapbt_alloc_block(
 {
 	struct xfs_buf		*agbp = cur->bc_ag.agbp;
 	struct xfs_agf		*agf = agbp->b_addr;
+	struct xfs_perag	*pag = cur->bc_ag.pag;
 	int			error;
 	xfs_agblock_t		bno;
 
@@ -93,21 +93,19 @@ xfs_rmapbt_alloc_block(
 	if (error)
 		return error;
 
-	trace_xfs_rmapbt_alloc_block(cur->bc_mp, cur->bc_ag.agno,
-			bno, 1);
+	trace_xfs_rmapbt_alloc_block(cur->bc_mp, pag->pag_agno, bno, 1);
 	if (bno == NULLAGBLOCK) {
 		*stat = 0;
 		return 0;
 	}
 
-	xfs_extent_busy_reuse(cur->bc_mp, cur->bc_ag.agno, bno, 1,
-			false);
+	xfs_extent_busy_reuse(cur->bc_mp, pag, bno, 1, false);
 
 	new->s = cpu_to_be32(bno);
 	be32_add_cpu(&agf->agf_rmap_blocks, 1);
 	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_RMAP_BLOCKS);
 
-	xfs_ag_resv_rmapbt_alloc(cur->bc_mp, cur->bc_ag.agno);
+	xfs_ag_resv_rmapbt_alloc(cur->bc_mp, pag->pag_agno);
 
 	*stat = 1;
 	return 0;
@@ -120,12 +118,12 @@ xfs_rmapbt_free_block(
 {
 	struct xfs_buf		*agbp = cur->bc_ag.agbp;
 	struct xfs_agf		*agf = agbp->b_addr;
-	struct xfs_perag	*pag;
+	struct xfs_perag	*pag = cur->bc_ag.pag;
 	xfs_agblock_t		bno;
 	int			error;
 
 	bno = xfs_daddr_to_agbno(cur->bc_mp, XFS_BUF_ADDR(bp));
-	trace_xfs_rmapbt_free_block(cur->bc_mp, cur->bc_ag.agno,
+	trace_xfs_rmapbt_free_block(cur->bc_mp, pag->pag_agno,
 			bno, 1);
 	be32_add_cpu(&agf->agf_rmap_blocks, -1);
 	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_RMAP_BLOCKS);
@@ -133,10 +131,9 @@ xfs_rmapbt_free_block(
 	if (error)
 		return error;
 
-	xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1,
+	xfs_extent_busy_insert(cur->bc_tp, pag, bno, 1,
 			      XFS_EXTENT_BUSY_SKIP_DISCARD);
 
-	pag = cur->bc_ag.agbp->b_pag;
 	xfs_ag_resv_free_extent(pag, XFS_AG_RESV_RMAPBT, NULL, 1);
 	return 0;
 }
@@ -215,7 +212,7 @@ xfs_rmapbt_init_ptr_from_cur(
 {
 	struct xfs_agf		*agf = cur->bc_ag.agbp->b_addr;
 
-	ASSERT(cur->bc_ag.agno == be32_to_cpu(agf->agf_seqno));
+	ASSERT(cur->bc_ag.pag->pag_agno == be32_to_cpu(agf->agf_seqno));
 
 	ptr->s = agf->agf_roots[cur->bc_btnum];
 }
@@ -450,7 +447,7 @@ static struct xfs_btree_cur *
 xfs_rmapbt_init_common(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno)
+	struct xfs_perag	*pag)
 {
 	struct xfs_btree_cur	*cur;
 
@@ -462,9 +459,12 @@ xfs_rmapbt_init_common(
 	cur->bc_flags = XFS_BTREE_CRC_BLOCKS | XFS_BTREE_OVERLAPPING;
 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
 	cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_rmap_2);
-	cur->bc_ag.agno = agno;
 	cur->bc_ops = &xfs_rmapbt_ops;
 
+	/* take a reference for the cursor */
+	atomic_inc(&pag->pag_ref);
+	cur->bc_ag.pag = pag;
+
 	return cur;
 }
 
@@ -474,12 +474,12 @@ xfs_rmapbt_init_cursor(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
 	struct xfs_buf		*agbp,
-	xfs_agnumber_t		agno)
+	struct xfs_perag	*pag)
 {
 	struct xfs_agf		*agf = agbp->b_addr;
 	struct xfs_btree_cur	*cur;
 
-	cur = xfs_rmapbt_init_common(mp, tp, agno);
+	cur = xfs_rmapbt_init_common(mp, tp, pag);
 	cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]);
 	cur->bc_ag.agbp = agbp;
 	return cur;
@@ -490,11 +490,11 @@ struct xfs_btree_cur *
 xfs_rmapbt_stage_cursor(
 	struct xfs_mount	*mp,
 	struct xbtree_afakeroot	*afake,
-	xfs_agnumber_t		agno)
+	struct xfs_perag	*pag)
 {
 	struct xfs_btree_cur	*cur;
 
-	cur = xfs_rmapbt_init_common(mp, NULL, agno);
+	cur = xfs_rmapbt_init_common(mp, NULL, pag);
 	xfs_btree_stage_afakeroot(cur, afake);
 	return cur;
 }
@@ -596,7 +596,7 @@ int
 xfs_rmapbt_calc_reserves(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_extlen_t		*ask,
 	xfs_extlen_t		*used)
 {
@@ -609,7 +609,7 @@ xfs_rmapbt_calc_reserves(
 	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
 		return 0;
 
-	error = xfs_alloc_read_agf(mp, tp, agno, 0, &agbp);
+	error = xfs_alloc_read_agf(mp, tp, pag->pag_agno, 0, &agbp);
 	if (error)
 		return error;
 
@@ -624,7 +624,7 @@ xfs_rmapbt_calc_reserves(
 	 * expansion.  We therefore can pretend the space isn't there.
 	 */
 	if (mp->m_sb.sb_logstart &&
-	    XFS_FSB_TO_AGNO(mp, mp->m_sb.sb_logstart) == agno)
+	    XFS_FSB_TO_AGNO(mp, mp->m_sb.sb_logstart) == pag->pag_agno)
 		agblocks -= mp->m_sb.sb_logblocks;
 
 	/* Reserve 1% of the AG or enough for 1 block per record. */
diff --git a/fs/xfs/libxfs/xfs_rmap_btree.h b/fs/xfs/libxfs/xfs_rmap_btree.h
index 115c3455a734..88d8d18788a2 100644
--- a/fs/xfs/libxfs/xfs_rmap_btree.h
+++ b/fs/xfs/libxfs/xfs_rmap_btree.h
@@ -43,9 +43,9 @@ struct xbtree_afakeroot;
 
 struct xfs_btree_cur *xfs_rmapbt_init_cursor(struct xfs_mount *mp,
 				struct xfs_trans *tp, struct xfs_buf *bp,
-				xfs_agnumber_t agno);
+				struct xfs_perag *pag);
 struct xfs_btree_cur *xfs_rmapbt_stage_cursor(struct xfs_mount *mp,
-		struct xbtree_afakeroot *afake, xfs_agnumber_t agno);
+		struct xbtree_afakeroot *afake, struct xfs_perag *pag);
 void xfs_rmapbt_commit_staged_btree(struct xfs_btree_cur *cur,
 		struct xfs_trans *tp, struct xfs_buf *agbp);
 int xfs_rmapbt_maxrecs(int blocklen, int leaf);
@@ -57,6 +57,6 @@ extern xfs_extlen_t xfs_rmapbt_max_size(struct xfs_mount *mp,
 		xfs_agblock_t agblocks);
 
 extern int xfs_rmapbt_calc_reserves(struct xfs_mount *mp, struct xfs_trans *tp,
-		xfs_agnumber_t agno, xfs_extlen_t *ask, xfs_extlen_t *used);
+		struct xfs_perag *pag, xfs_extlen_t *ask, xfs_extlen_t *used);
 
 #endif	/* __XFS_RMAP_BTREE_H__ */
diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c
index dfbbcbd448c1..04f5386446db 100644
--- a/fs/xfs/libxfs/xfs_sb.c
+++ b/fs/xfs/libxfs/xfs_sb.c
@@ -15,7 +15,6 @@
 #include "xfs_ialloc.h"
 #include "xfs_alloc.h"
 #include "xfs_error.h"
-#include "xfs_trace.h"
 #include "xfs_trans.h"
 #include "xfs_buf_item.h"
 #include "xfs_bmap_btree.h"
@@ -25,72 +24,12 @@
 #include "xfs_refcount_btree.h"
 #include "xfs_da_format.h"
 #include "xfs_health.h"
+#include "xfs_ag.h"
 
 /*
  * Physical superblock buffer manipulations. Shared with libxfs in userspace.
  */
 
-/*
- * Reference counting access wrappers to the perag structures.
- * Because we never free per-ag structures, the only thing we
- * have to protect against changes is the tree structure itself.
- */
-struct xfs_perag *
-xfs_perag_get(
-	struct xfs_mount	*mp,
-	xfs_agnumber_t		agno)
-{
-	struct xfs_perag	*pag;
-	int			ref = 0;
-
-	rcu_read_lock();
-	pag = radix_tree_lookup(&mp->m_perag_tree, agno);
-	if (pag) {
-		ASSERT(atomic_read(&pag->pag_ref) >= 0);
-		ref = atomic_inc_return(&pag->pag_ref);
-	}
-	rcu_read_unlock();
-	trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
-	return pag;
-}
-
-/*
- * search from @first to find the next perag with the given tag set.
- */
-struct xfs_perag *
-xfs_perag_get_tag(
-	struct xfs_mount	*mp,
-	xfs_agnumber_t		first,
-	int			tag)
-{
-	struct xfs_perag	*pag;
-	int			found;
-	int			ref;
-
-	rcu_read_lock();
-	found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
-					(void **)&pag, first, 1, tag);
-	if (found <= 0) {
-		rcu_read_unlock();
-		return NULL;
-	}
-	ref = atomic_inc_return(&pag->pag_ref);
-	rcu_read_unlock();
-	trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
-	return pag;
-}
-
-void
-xfs_perag_put(
-	struct xfs_perag	*pag)
-{
-	int	ref;
-
-	ASSERT(atomic_read(&pag->pag_ref) > 0);
-	ref = atomic_dec_return(&pag->pag_ref);
-	trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
-}
-
 /* Check all the superblock fields we care about when reading one in. */
 STATIC int
 xfs_validate_sb_read(
@@ -842,78 +781,6 @@ xfs_sb_mount_common(
 }
 
 /*
- * xfs_initialize_perag_data
- *
- * Read in each per-ag structure so we can count up the number of
- * allocated inodes, free inodes and used filesystem blocks as this
- * information is no longer persistent in the superblock. Once we have
- * this information, write it into the in-core superblock structure.
- */
-int
-xfs_initialize_perag_data(
-	struct xfs_mount *mp,
-	xfs_agnumber_t	agcount)
-{
-	xfs_agnumber_t	index;
-	xfs_perag_t	*pag;
-	xfs_sb_t	*sbp = &mp->m_sb;
-	uint64_t	ifree = 0;
-	uint64_t	ialloc = 0;
-	uint64_t	bfree = 0;
-	uint64_t	bfreelst = 0;
-	uint64_t	btree = 0;
-	uint64_t	fdblocks;
-	int		error = 0;
-
-	for (index = 0; index < agcount; index++) {
-		/*
-		 * read the agf, then the agi. This gets us
-		 * all the information we need and populates the
-		 * per-ag structures for us.
-		 */
-		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
-		if (error)
-			return error;
-
-		error = xfs_ialloc_pagi_init(mp, NULL, index);
-		if (error)
-			return error;
-		pag = xfs_perag_get(mp, index);
-		ifree += pag->pagi_freecount;
-		ialloc += pag->pagi_count;
-		bfree += pag->pagf_freeblks;
-		bfreelst += pag->pagf_flcount;
-		btree += pag->pagf_btreeblks;
-		xfs_perag_put(pag);
-	}
-	fdblocks = bfree + bfreelst + btree;
-
-	/*
-	 * If the new summary counts are obviously incorrect, fail the
-	 * mount operation because that implies the AGFs are also corrupt.
-	 * Clear FS_COUNTERS so that we don't unmount with a dirty log, which
-	 * will prevent xfs_repair from fixing anything.
-	 */
-	if (fdblocks > sbp->sb_dblocks || ifree > ialloc) {
-		xfs_alert(mp, "AGF corruption. Please run xfs_repair.");
-		error = -EFSCORRUPTED;
-		goto out;
-	}
-
-	/* Overwrite incore superblock counters with just-read data */
-	spin_lock(&mp->m_sb_lock);
-	sbp->sb_ifree = ifree;
-	sbp->sb_icount = ialloc;
-	sbp->sb_fdblocks = fdblocks;
-	spin_unlock(&mp->m_sb_lock);
-
-	xfs_reinit_percpu_counters(mp);
-out:
-	xfs_fs_mark_healthy(mp, XFS_SICK_FS_COUNTERS);
-	return error;
-}
-
-/*
  * xfs_log_sb() can be used to copy arbitrary changes to the in-core superblock
  * into the superblock buffer to be logged.  It does not provide the higher
  * level of locking that is needed to protect the in-core superblock from
@@ -989,17 +856,18 @@ int
 xfs_update_secondary_sbs(
 	struct xfs_mount	*mp)
 {
-	xfs_agnumber_t		agno;
+	struct xfs_perag	*pag;
+	xfs_agnumber_t		agno = 1;
 	int			saved_error = 0;
 	int			error = 0;
 	LIST_HEAD		(buffer_list);
 
 	/* update secondary superblocks. */
-	for (agno = 1; agno < mp->m_sb.sb_agcount; agno++) {
+	for_each_perag_from(mp, agno, pag) {
 		struct xfs_buf		*bp;
 
 		error = xfs_buf_get(mp->m_ddev_targp,
-				 XFS_AG_DADDR(mp, agno, XFS_SB_DADDR),
+				 XFS_AG_DADDR(mp, pag->pag_agno, XFS_SB_DADDR),
 				 XFS_FSS_TO_BB(mp, 1), &bp);
 		/*
 		 * If we get an error reading or writing alternate superblocks,
@@ -1011,7 +879,7 @@ xfs_update_secondary_sbs(
 		if (error) {
 			xfs_warn(mp,
 		"error allocating secondary superblock for ag %d",
-				agno);
+				pag->pag_agno);
 			if (!saved_error)
 				saved_error = error;
 			continue;
@@ -1032,7 +900,7 @@ xfs_update_secondary_sbs(
 		if (error) {
 			xfs_warn(mp,
 		"write error %d updating a secondary superblock near ag %d",
-				error, agno);
+				error, pag->pag_agno);
 			if (!saved_error)
 				saved_error = error;
 			continue;
diff --git a/fs/xfs/libxfs/xfs_sb.h b/fs/xfs/libxfs/xfs_sb.h
index f79f9dc632b6..0c1602d9b53d 100644
--- a/fs/xfs/libxfs/xfs_sb.h
+++ b/fs/xfs/libxfs/xfs_sb.h
@@ -13,15 +13,6 @@ struct xfs_trans;
 struct xfs_fsop_geom;
 struct xfs_perag;
 
-/*
- * perag get/put wrappers for ref counting
- */
-extern struct xfs_perag *xfs_perag_get(struct xfs_mount *, xfs_agnumber_t);
-extern struct xfs_perag *xfs_perag_get_tag(struct xfs_mount *, xfs_agnumber_t,
-					   int tag);
-extern void	xfs_perag_put(struct xfs_perag *pag);
-extern int	xfs_initialize_perag_data(struct xfs_mount *, xfs_agnumber_t);
-
 extern void	xfs_log_sb(struct xfs_trans *tp);
 extern int	xfs_sync_sb(struct xfs_mount *mp, bool wait);
 extern int	xfs_sync_sb_buf(struct xfs_mount *mp);
diff --git a/fs/xfs/libxfs/xfs_types.c b/fs/xfs/libxfs/xfs_types.c
index 04801362e1a7..e8f4abee7892 100644
--- a/fs/xfs/libxfs/xfs_types.c
+++ b/fs/xfs/libxfs/xfs_types.c
@@ -11,6 +11,7 @@
 #include "xfs_trans_resv.h"
 #include "xfs_bit.h"
 #include "xfs_mount.h"
+#include "xfs_ag.h"
 
 /* Find the size of the AG, in blocks. */
 inline xfs_agblock_t
@@ -222,12 +223,13 @@ xfs_icount_range(
 	unsigned long long	*max)
 {
 	unsigned long long	nr_inos = 0;
+	struct xfs_perag	*pag;
 	xfs_agnumber_t		agno;
 
 	/* root, rtbitmap, rtsum all live in the first chunk */
 	*min = XFS_INODES_PER_CHUNK;
 
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
+	for_each_perag(mp, agno, pag) {
 		xfs_agino_t	first, last;
 
 		xfs_agino_range(mp, agno, &first, &last);
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index 7a2f9b5f2db5..64a7a30f4ac0 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -14,6 +14,7 @@
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
 #include "xfs_rmap.h"
+#include "xfs_ag.h"
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 
diff --git a/fs/xfs/scrub/agheader_repair.c b/fs/xfs/scrub/agheader_repair.c
index 23690f824ffa..e95f8c98f0f7 100644
--- a/fs/xfs/scrub/agheader_repair.c
+++ b/fs/xfs/scrub/agheader_repair.c
@@ -20,6 +20,7 @@
 #include "xfs_rmap.h"
 #include "xfs_rmap_btree.h"
 #include "xfs_refcount_btree.h"
+#include "xfs_ag.h"
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/trace.h"
@@ -245,8 +246,8 @@ xrep_agf_calc_from_btrees(
 	int			error;
 
 	/* Update the AGF counters from the bnobt. */
-	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno,
-			XFS_BTNUM_BNO);
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp,
+			sc->sa.pag, XFS_BTNUM_BNO);
 	error = xfs_alloc_query_all(cur, xrep_agf_walk_allocbt, &raa);
 	if (error)
 		goto err;
@@ -259,8 +260,8 @@ xrep_agf_calc_from_btrees(
 	agf->agf_longest = cpu_to_be32(raa.longest);
 
 	/* Update the AGF counters from the cntbt. */
-	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno,
-			XFS_BTNUM_CNT);
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp,
+			sc->sa.pag, XFS_BTNUM_CNT);
 	error = xfs_btree_count_blocks(cur, &blocks);
 	if (error)
 		goto err;
@@ -268,7 +269,7 @@ xrep_agf_calc_from_btrees(
 	btreeblks += blocks - 1;
 
 	/* Update the AGF counters from the rmapbt. */
-	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno);
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.pag);
 	error = xfs_btree_count_blocks(cur, &blocks);
 	if (error)
 		goto err;
@@ -281,7 +282,7 @@ xrep_agf_calc_from_btrees(
 	/* Update the AGF counters from the refcountbt. */
 	if (xfs_sb_version_hasreflink(&mp->m_sb)) {
 		cur = xfs_refcountbt_init_cursor(mp, sc->tp, agf_bp,
-				sc->sa.agno);
+				sc->sa.pag);
 		error = xfs_btree_count_blocks(cur, &blocks);
 		if (error)
 			goto err;
@@ -453,7 +454,7 @@ xrep_agfl_walk_rmap(
 
 	/* Record all the OWN_AG blocks. */
 	if (rec->rm_owner == XFS_RMAP_OWN_AG) {
-		fsb = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_ag.agno,
+		fsb = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_ag.pag->pag_agno,
 				rec->rm_startblock);
 		error = xbitmap_set(ra->freesp, fsb, rec->rm_blockcount);
 		if (error)
@@ -489,23 +490,23 @@ xrep_agfl_collect_blocks(
 	xbitmap_init(&ra.agmetablocks);
 
 	/* Find all space used by the free space btrees & rmapbt. */
-	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno);
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.pag);
 	error = xfs_rmap_query_all(cur, xrep_agfl_walk_rmap, &ra);
 	if (error)
 		goto err;
 	xfs_btree_del_cursor(cur, error);
 
 	/* Find all blocks currently being used by the bnobt. */
-	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno,
-			XFS_BTNUM_BNO);
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp,
+			sc->sa.pag, XFS_BTNUM_BNO);
 	error = xbitmap_set_btblocks(&ra.agmetablocks, cur);
 	if (error)
 		goto err;
 	xfs_btree_del_cursor(cur, error);
 
 	/* Find all blocks currently being used by the cntbt. */
-	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno,
-			XFS_BTNUM_CNT);
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp,
+			sc->sa.pag, XFS_BTNUM_CNT);
 	error = xbitmap_set_btblocks(&ra.agmetablocks, cur);
 	if (error)
 		goto err;
@@ -805,8 +806,8 @@ xrep_agi_calc_from_btrees(
 	xfs_agino_t		freecount;
 	int			error;
 
-	cur = xfs_inobt_init_cursor(mp, sc->tp, agi_bp, sc->sa.agno,
-			XFS_BTNUM_INO);
+	cur = xfs_inobt_init_cursor(mp, sc->tp, agi_bp,
+			sc->sa.pag, XFS_BTNUM_INO);
 	error = xfs_ialloc_count_inodes(cur, &count, &freecount);
 	if (error)
 		goto err;
@@ -827,8 +828,8 @@ xrep_agi_calc_from_btrees(
 	    xfs_sb_version_hasinobtcounts(&mp->m_sb)) {
 		xfs_agblock_t	blocks;
 
-		cur = xfs_inobt_init_cursor(mp, sc->tp, agi_bp, sc->sa.agno,
-				XFS_BTNUM_FINO);
+		cur = xfs_inobt_init_cursor(mp, sc->tp, agi_bp,
+				sc->sa.pag, XFS_BTNUM_FINO);
 		error = xfs_btree_count_blocks(cur, &blocks);
 		if (error)
 			goto err;
diff --git a/fs/xfs/scrub/alloc.c b/fs/xfs/scrub/alloc.c
index 2720bd7fe53b..d5741980094a 100644
--- a/fs/xfs/scrub/alloc.c
+++ b/fs/xfs/scrub/alloc.c
@@ -15,6 +15,7 @@
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
+#include "xfs_ag.h"
 
 /*
  * Set us up to scrub free space btrees.
@@ -93,7 +94,7 @@ xchk_allocbt_rec(
 	union xfs_btree_rec	*rec)
 {
 	struct xfs_mount	*mp = bs->cur->bc_mp;
-	xfs_agnumber_t		agno = bs->cur->bc_ag.agno;
+	xfs_agnumber_t		agno = bs->cur->bc_ag.pag->pag_agno;
 	xfs_agblock_t		bno;
 	xfs_extlen_t		len;
 
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
index b5ebf1d1b4db..0f125583189f 100644
--- a/fs/xfs/scrub/bmap.c
+++ b/fs/xfs/scrub/bmap.c
@@ -22,6 +22,7 @@
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
+#include "xfs_ag.h"
 
 /* Set us up with an inode's bmap. */
 int
@@ -514,7 +515,7 @@ xchk_bmap_check_rmap(
 			xchk_fblock_set_corrupt(sc, sbcri->whichfork,
 					rec->rm_offset);
 		if (irec.br_startblock != XFS_AGB_TO_FSB(sc->mp,
-				cur->bc_ag.agno, rec->rm_startblock))
+				cur->bc_ag.pag->pag_agno, rec->rm_startblock))
 			xchk_fblock_set_corrupt(sc, sbcri->whichfork,
 					rec->rm_offset);
 		if (irec.br_blockcount > rec->rm_blockcount)
@@ -544,18 +545,18 @@ STATIC int
 xchk_bmap_check_ag_rmaps(
 	struct xfs_scrub		*sc,
 	int				whichfork,
-	xfs_agnumber_t			agno)
+	struct xfs_perag		*pag)
 {
 	struct xchk_bmap_check_rmap_info	sbcri;
 	struct xfs_btree_cur		*cur;
 	struct xfs_buf			*agf;
 	int				error;
 
-	error = xfs_alloc_read_agf(sc->mp, sc->tp, agno, 0, &agf);
+	error = xfs_alloc_read_agf(sc->mp, sc->tp, pag->pag_agno, 0, &agf);
 	if (error)
 		return error;
 
-	cur = xfs_rmapbt_init_cursor(sc->mp, sc->tp, agf, agno);
+	cur = xfs_rmapbt_init_cursor(sc->mp, sc->tp, agf, pag);
 
 	sbcri.sc = sc;
 	sbcri.whichfork = whichfork;
@@ -575,6 +576,7 @@ xchk_bmap_check_rmaps(
 	int			whichfork)
 {
 	struct xfs_ifork	*ifp = XFS_IFORK_PTR(sc->ip, whichfork);
+	struct xfs_perag	*pag;
 	xfs_agnumber_t		agno;
 	bool			zero_size;
 	int			error;
@@ -607,15 +609,16 @@ xchk_bmap_check_rmaps(
 	    (zero_size || ifp->if_nextents > 0))
 		return 0;
 
-	for (agno = 0; agno < sc->mp->m_sb.sb_agcount; agno++) {
-		error = xchk_bmap_check_ag_rmaps(sc, whichfork, agno);
+	for_each_perag(sc->mp, agno, pag) {
+		error = xchk_bmap_check_ag_rmaps(sc, whichfork, pag);
 		if (error)
-			return error;
+			break;
 		if (sc->sm->sm_flags & XFS_SCRUB_OFLAG_CORRUPT)
 			break;
 	}
-
-	return 0;
+	if (pag)
+		xfs_perag_put(pag);
+	return error;
 }
 
 /*
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index be38c960da85..cadfd5799909 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -12,7 +12,6 @@
 #include "xfs_btree.h"
 #include "xfs_log_format.h"
 #include "xfs_trans.h"
-#include "xfs_sb.h"
 #include "xfs_inode.h"
 #include "xfs_icache.h"
 #include "xfs_alloc.h"
@@ -26,6 +25,7 @@
 #include "xfs_trans_priv.h"
 #include "xfs_attr.h"
 #include "xfs_reflink.h"
+#include "xfs_ag.h"
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/trace.h"
@@ -460,49 +460,48 @@ xchk_ag_btcur_init(
 	struct xchk_ag		*sa)
 {
 	struct xfs_mount	*mp = sc->mp;
-	xfs_agnumber_t		agno = sa->agno;
 
 	xchk_perag_get(sc->mp, sa);
 	if (sa->agf_bp &&
 	    xchk_ag_btree_healthy_enough(sc, sa->pag, XFS_BTNUM_BNO)) {
 		/* Set up a bnobt cursor for cross-referencing. */
 		sa->bno_cur = xfs_allocbt_init_cursor(mp, sc->tp, sa->agf_bp,
-				agno, XFS_BTNUM_BNO);
+				sa->pag, XFS_BTNUM_BNO);
 	}
 
 	if (sa->agf_bp &&
 	    xchk_ag_btree_healthy_enough(sc, sa->pag, XFS_BTNUM_CNT)) {
 		/* Set up a cntbt cursor for cross-referencing. */
 		sa->cnt_cur = xfs_allocbt_init_cursor(mp, sc->tp, sa->agf_bp,
-				agno, XFS_BTNUM_CNT);
+				sa->pag, XFS_BTNUM_CNT);
 	}
 
 	/* Set up a inobt cursor for cross-referencing. */
 	if (sa->agi_bp &&
 	    xchk_ag_btree_healthy_enough(sc, sa->pag, XFS_BTNUM_INO)) {
 		sa->ino_cur = xfs_inobt_init_cursor(mp, sc->tp, sa->agi_bp,
-					agno, XFS_BTNUM_INO);
+				sa->pag, XFS_BTNUM_INO);
 	}
 
 	/* Set up a finobt cursor for cross-referencing. */
 	if (sa->agi_bp && xfs_sb_version_hasfinobt(&mp->m_sb) &&
 	    xchk_ag_btree_healthy_enough(sc, sa->pag, XFS_BTNUM_FINO)) {
 		sa->fino_cur = xfs_inobt_init_cursor(mp, sc->tp, sa->agi_bp,
-				agno, XFS_BTNUM_FINO);
+				sa->pag, XFS_BTNUM_FINO);
 	}
 
 	/* Set up a rmapbt cursor for cross-referencing. */
 	if (sa->agf_bp && xfs_sb_version_hasrmapbt(&mp->m_sb) &&
 	    xchk_ag_btree_healthy_enough(sc, sa->pag, XFS_BTNUM_RMAP)) {
 		sa->rmap_cur = xfs_rmapbt_init_cursor(mp, sc->tp, sa->agf_bp,
-				agno);
+				sa->pag);
 	}
 
 	/* Set up a refcountbt cursor for cross-referencing. */
 	if (sa->agf_bp && xfs_sb_version_hasreflink(&mp->m_sb) &&
 	    xchk_ag_btree_healthy_enough(sc, sa->pag, XFS_BTNUM_REFC)) {
 		sa->refc_cur = xfs_refcountbt_init_cursor(mp, sc->tp,
-				sa->agf_bp, agno);
+				sa->agf_bp, sa->pag);
 	}
 }
 
diff --git a/fs/xfs/scrub/fscounters.c b/fs/xfs/scrub/fscounters.c
index f1d1a8c58853..fd7941e04ae1 100644
--- a/fs/xfs/scrub/fscounters.c
+++ b/fs/xfs/scrub/fscounters.c
@@ -9,11 +9,11 @@
 #include "xfs_format.h"
 #include "xfs_trans_resv.h"
 #include "xfs_mount.h"
-#include "xfs_sb.h"
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
 #include "xfs_health.h"
 #include "xfs_btree.h"
+#include "xfs_ag.h"
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/trace.h"
@@ -71,11 +71,11 @@ xchk_fscount_warmup(
 	xfs_agnumber_t		agno;
 	int			error = 0;
 
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		pag = xfs_perag_get(mp, agno);
-
+	for_each_perag(mp, agno, pag) {
+		if (xchk_should_terminate(sc, &error))
+			break;
 		if (pag->pagi_init && pag->pagf_init)
-			goto next_loop_perag;
+			continue;
 
 		/* Lock both AG headers. */
 		error = xfs_ialloc_read_agi(mp, sc->tp, agno, &agi_bp);
@@ -89,21 +89,15 @@ xchk_fscount_warmup(
 		 * These are supposed to be initialized by the header read
 		 * function.
 		 */
-		error = -EFSCORRUPTED;
-		if (!pag->pagi_init || !pag->pagf_init)
+		if (!pag->pagi_init || !pag->pagf_init) {
+			error = -EFSCORRUPTED;
 			break;
+		}
 
 		xfs_buf_relse(agf_bp);
 		agf_bp = NULL;
 		xfs_buf_relse(agi_bp);
 		agi_bp = NULL;
-next_loop_perag:
-		xfs_perag_put(pag);
-		pag = NULL;
-		error = 0;
-
-		if (xchk_should_terminate(sc, &error))
-			break;
 	}
 
 	if (agf_bp)
@@ -196,13 +190,14 @@ retry:
 	fsc->ifree = 0;
 	fsc->fdblocks = 0;
 
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		pag = xfs_perag_get(mp, agno);
+	for_each_perag(mp, agno, pag) {
+		if (xchk_should_terminate(sc, &error))
+			break;
 
 		/* This somehow got unset since the warmup? */
 		if (!pag->pagi_init || !pag->pagf_init) {
-			xfs_perag_put(pag);
-			return -EFSCORRUPTED;
+			error = -EFSCORRUPTED;
+			break;
 		}
 
 		/* Count all the inodes */
@@ -216,10 +211,8 @@ retry:
 			fsc->fdblocks += pag->pagf_btreeblks;
 		} else {
 			error = xchk_fscount_btreeblks(sc, fsc, agno);
-			if (error) {
-				xfs_perag_put(pag);
+			if (error)
 				break;
-			}
 		}
 
 		/*
@@ -229,12 +222,9 @@ retry:
 		fsc->fdblocks -= pag->pag_meta_resv.ar_reserved;
 		fsc->fdblocks -= pag->pag_rmapbt_resv.ar_orig_reserved;
 
-		xfs_perag_put(pag);
-
-		if (xchk_should_terminate(sc, &error))
-			break;
 	}
-
+	if (pag)
+		xfs_perag_put(pag);
 	if (error)
 		return error;
 
diff --git a/fs/xfs/scrub/health.c b/fs/xfs/scrub/health.c
index 3de59b5c2ce6..2e61df3bca83 100644
--- a/fs/xfs/scrub/health.c
+++ b/fs/xfs/scrub/health.c
@@ -8,7 +8,7 @@
 #include "xfs_shared.h"
 #include "xfs_format.h"
 #include "xfs_btree.h"
-#include "xfs_sb.h"
+#include "xfs_ag.h"
 #include "xfs_health.h"
 #include "scrub/scrub.h"
 #include "scrub/health.h"
diff --git a/fs/xfs/scrub/ialloc.c b/fs/xfs/scrub/ialloc.c
index 8d9f3fb0cd22..30e568596b79 100644
--- a/fs/xfs/scrub/ialloc.c
+++ b/fs/xfs/scrub/ialloc.c
@@ -21,6 +21,7 @@
 #include "scrub/common.h"
 #include "scrub/btree.h"
 #include "scrub/trace.h"
+#include "xfs_ag.h"
 
 /*
  * Set us up to scrub inode btrees.
@@ -103,7 +104,7 @@ xchk_iallocbt_chunk(
 	xfs_extlen_t			len)
 {
 	struct xfs_mount		*mp = bs->cur->bc_mp;
-	xfs_agnumber_t			agno = bs->cur->bc_ag.agno;
+	xfs_agnumber_t			agno = bs->cur->bc_ag.pag->pag_agno;
 	xfs_agblock_t			bno;
 
 	bno = XFS_AGINO_TO_AGBNO(mp, agino);
@@ -163,7 +164,7 @@ xchk_iallocbt_check_cluster_ifree(
 	 * the record, compute which fs inode we're talking about.
 	 */
 	agino = irec->ir_startino + irec_ino;
-	fsino = XFS_AGINO_TO_INO(mp, bs->cur->bc_ag.agno, agino);
+	fsino = XFS_AGINO_TO_INO(mp, bs->cur->bc_ag.pag->pag_agno, agino);
 	irec_free = (irec->ir_free & XFS_INOBT_MASK(irec_ino));
 
 	if (be16_to_cpu(dip->di_magic) != XFS_DINODE_MAGIC ||
@@ -213,7 +214,7 @@ xchk_iallocbt_check_cluster(
 	struct xfs_mount		*mp = bs->cur->bc_mp;
 	struct xfs_buf			*cluster_bp;
 	unsigned int			nr_inodes;
-	xfs_agnumber_t			agno = bs->cur->bc_ag.agno;
+	xfs_agnumber_t			agno = bs->cur->bc_ag.pag->pag_agno;
 	xfs_agblock_t			agbno;
 	unsigned int			cluster_index;
 	uint16_t			cluster_mask = 0;
@@ -423,7 +424,7 @@ xchk_iallocbt_rec(
 	struct xchk_iallocbt		*iabt = bs->private;
 	struct xfs_inobt_rec_incore	irec;
 	uint64_t			holes;
-	xfs_agnumber_t			agno = bs->cur->bc_ag.agno;
+	xfs_agnumber_t			agno = bs->cur->bc_ag.pag->pag_agno;
 	xfs_agino_t			agino;
 	xfs_extlen_t			len;
 	int				holecount;
diff --git a/fs/xfs/scrub/refcount.c b/fs/xfs/scrub/refcount.c
index 744530a66c0c..7014b7408bad 100644
--- a/fs/xfs/scrub/refcount.c
+++ b/fs/xfs/scrub/refcount.c
@@ -13,6 +13,7 @@
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
+#include "xfs_ag.h"
 
 /*
  * Set us up to scrub reference count btrees.
@@ -333,7 +334,7 @@ xchk_refcountbt_rec(
 {
 	struct xfs_mount	*mp = bs->cur->bc_mp;
 	xfs_agblock_t		*cow_blocks = bs->private;
-	xfs_agnumber_t		agno = bs->cur->bc_ag.agno;
+	xfs_agnumber_t		agno = bs->cur->bc_ag.pag->pag_agno;
 	xfs_agblock_t		bno;
 	xfs_extlen_t		len;
 	xfs_nlink_t		refcount;
diff --git a/fs/xfs/scrub/repair.c b/fs/xfs/scrub/repair.c
index c2857d854c83..5cf1c3707b6a 100644
--- a/fs/xfs/scrub/repair.c
+++ b/fs/xfs/scrub/repair.c
@@ -22,6 +22,7 @@
 #include "xfs_rmap_btree.h"
 #include "xfs_refcount_btree.h"
 #include "xfs_extent_busy.h"
+#include "xfs_ag.h"
 #include "xfs_ag_resv.h"
 #include "xfs_quota.h"
 #include "scrub/scrub.h"
@@ -303,7 +304,7 @@ xrep_alloc_ag_block(
 			return error;
 		if (bno == NULLAGBLOCK)
 			return -ENOSPC;
-		xfs_extent_busy_reuse(sc->mp, sc->sa.agno, bno,
+		xfs_extent_busy_reuse(sc->mp, sc->sa.pag, bno,
 				1, false);
 		*fsbno = XFS_AGB_TO_FSB(sc->mp, sc->sa.agno, bno);
 		if (resv == XFS_AG_RESV_RMAPBT)
@@ -508,7 +509,7 @@ xrep_put_freelist(
 	 * create an rmap for the block prior to merging it or else other
 	 * parts will break.
 	 */
-	error = xfs_rmap_alloc(sc->tp, sc->sa.agf_bp, sc->sa.agno, agbno, 1,
+	error = xfs_rmap_alloc(sc->tp, sc->sa.agf_bp, sc->sa.pag, agbno, 1,
 			&XFS_RMAP_OINFO_AG);
 	if (error)
 		return error;
@@ -518,7 +519,7 @@ xrep_put_freelist(
 			agbno, 0);
 	if (error)
 		return error;
-	xfs_extent_busy_insert(sc->tp, sc->sa.agno, agbno, 1,
+	xfs_extent_busy_insert(sc->tp, sc->sa.pag, agbno, 1,
 			XFS_EXTENT_BUSY_SKIP_DISCARD);
 
 	return 0;
@@ -554,7 +555,7 @@ xrep_reap_block(
 	} else {
 		agf_bp = sc->sa.agf_bp;
 	}
-	cur = xfs_rmapbt_init_cursor(sc->mp, sc->tp, agf_bp, agno);
+	cur = xfs_rmapbt_init_cursor(sc->mp, sc->tp, agf_bp, sc->sa.pag);
 
 	/* Can we find any other rmappings? */
 	error = xfs_rmap_has_other_keys(cur, agbno, 1, oinfo, &has_other_rmap);
@@ -576,7 +577,8 @@ xrep_reap_block(
 	 * to run xfs_repair.
 	 */
 	if (has_other_rmap)
-		error = xfs_rmap_free(sc->tp, agf_bp, agno, agbno, 1, oinfo);
+		error = xfs_rmap_free(sc->tp, agf_bp, sc->sa.pag, agbno,
+					1, oinfo);
 	else if (resv == XFS_AG_RESV_AGFL)
 		error = xrep_put_freelist(sc, agbno);
 	else
@@ -891,7 +893,7 @@ xrep_find_ag_btree_roots(
 		fab->height = 0;
 	}
 
-	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno);
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.pag);
 	error = xfs_rmap_query_all(cur, xrep_findroot_rmap, &ri);
 	xfs_btree_del_cursor(cur, error);
 
diff --git a/fs/xfs/scrub/rmap.c b/fs/xfs/scrub/rmap.c
index a4f17477c5d1..fc306573f0ac 100644
--- a/fs/xfs/scrub/rmap.c
+++ b/fs/xfs/scrub/rmap.c
@@ -15,6 +15,7 @@
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
+#include "xfs_ag.h"
 
 /*
  * Set us up to scrub reverse mapping btrees.
@@ -91,7 +92,7 @@ xchk_rmapbt_rec(
 {
 	struct xfs_mount	*mp = bs->cur->bc_mp;
 	struct xfs_rmap_irec	irec;
-	xfs_agnumber_t		agno = bs->cur->bc_ag.agno;
+	xfs_agnumber_t		agno = bs->cur->bc_ag.pag->pag_agno;
 	bool			non_inode;
 	bool			is_unwritten;
 	bool			is_bmbt;
diff --git a/fs/xfs/scrub/trace.c b/fs/xfs/scrub/trace.c
index 2c6c248be823..03882a605a3c 100644
--- a/fs/xfs/scrub/trace.c
+++ b/fs/xfs/scrub/trace.c
@@ -13,6 +13,7 @@
 #include "xfs_inode.h"
 #include "xfs_btree.h"
 #include "scrub/scrub.h"
+#include "xfs_ag.h"
 
 /* Figure out which block the btree cursor was pointing to. */
 static inline xfs_fsblock_t
@@ -26,7 +27,7 @@ xchk_btree_cur_fsbno(
 		 cur->bc_flags & XFS_BTREE_LONG_PTRS)
 		return XFS_INO_TO_FSB(cur->bc_mp, cur->bc_ino.ip->i_ino);
 	else if (!(cur->bc_flags & XFS_BTREE_LONG_PTRS))
-		return XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_ag.agno, 0);
+		return XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_ag.pag->pag_agno, 0);
 	return NULLFSBLOCK;
 }
 
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index a55471612150..b4ee9d3532f0 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -10,7 +10,6 @@
 #include "xfs_format.h"
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_trace.h"
 #include "xfs_log.h"
@@ -19,6 +18,7 @@
 #include "xfs_buf_item.h"
 #include "xfs_errortag.h"
 #include "xfs_error.h"
+#include "xfs_ag.h"
 
 static kmem_zone_t *xfs_buf_zone;
 
diff --git a/fs/xfs/xfs_discard.c b/fs/xfs/xfs_discard.c
index f979d0d7e6cd..736df5660f1f 100644
--- a/fs/xfs/xfs_discard.c
+++ b/fs/xfs/xfs_discard.c
@@ -8,7 +8,6 @@
 #include "xfs_format.h"
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_btree.h"
 #include "xfs_alloc_btree.h"
@@ -18,6 +17,7 @@
 #include "xfs_extent_busy.h"
 #include "xfs_trace.h"
 #include "xfs_log.h"
+#include "xfs_ag.h"
 
 STATIC int
 xfs_trim_extents(
@@ -50,7 +50,7 @@ xfs_trim_extents(
 		goto out_put_perag;
 	agf = agbp->b_addr;
 
-	cur = xfs_allocbt_init_cursor(mp, NULL, agbp, agno, XFS_BTNUM_CNT);
+	cur = xfs_allocbt_init_cursor(mp, NULL, agbp, pag, XFS_BTNUM_CNT);
 
 	/*
 	 * Look up the longest btree in the AGF and start with it.
@@ -108,7 +108,7 @@ xfs_trim_extents(
 		 * If any blocks in the range are still busy, skip the
 		 * discard and try again the next time.
 		 */
-		if (xfs_extent_busy_search(mp, agno, fbno, flen)) {
+		if (xfs_extent_busy_search(mp, pag, fbno, flen)) {
 			trace_xfs_discard_busy(mp, agno, fbno, flen);
 			goto next_extent;
 		}
diff --git a/fs/xfs/xfs_extent_busy.c b/fs/xfs/xfs_extent_busy.c
index a4075685d9eb..ad22a003f959 100644
--- a/fs/xfs/xfs_extent_busy.c
+++ b/fs/xfs/xfs_extent_busy.c
@@ -11,39 +11,37 @@
 #include "xfs_log_format.h"
 #include "xfs_shared.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_alloc.h"
 #include "xfs_extent_busy.h"
 #include "xfs_trace.h"
 #include "xfs_trans.h"
 #include "xfs_log.h"
+#include "xfs_ag.h"
 
 void
 xfs_extent_busy_insert(
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_agblock_t		bno,
 	xfs_extlen_t		len,
 	unsigned int		flags)
 {
 	struct xfs_extent_busy	*new;
 	struct xfs_extent_busy	*busyp;
-	struct xfs_perag	*pag;
 	struct rb_node		**rbp;
 	struct rb_node		*parent = NULL;
 
 	new = kmem_zalloc(sizeof(struct xfs_extent_busy), 0);
-	new->agno = agno;
+	new->agno = pag->pag_agno;
 	new->bno = bno;
 	new->length = len;
 	INIT_LIST_HEAD(&new->list);
 	new->flags = flags;
 
 	/* trace before insert to be able to see failed inserts */
-	trace_xfs_extent_busy(tp->t_mountp, agno, bno, len);
+	trace_xfs_extent_busy(tp->t_mountp, pag->pag_agno, bno, len);
 
-	pag = xfs_perag_get(tp->t_mountp, new->agno);
 	spin_lock(&pag->pagb_lock);
 	rbp = &pag->pagb_tree.rb_node;
 	while (*rbp) {
@@ -66,7 +64,6 @@ xfs_extent_busy_insert(
 
 	list_add(&new->list, &tp->t_busy);
 	spin_unlock(&pag->pagb_lock);
-	xfs_perag_put(pag);
 }
 
 /*
@@ -81,21 +78,17 @@ xfs_extent_busy_insert(
 int
 xfs_extent_busy_search(
 	struct xfs_mount	*mp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_agblock_t		bno,
 	xfs_extlen_t		len)
 {
-	struct xfs_perag	*pag;
 	struct rb_node		*rbp;
 	struct xfs_extent_busy	*busyp;
 	int			match = 0;
 
-	pag = xfs_perag_get(mp, agno);
+	/* find closest start bno overlap */
 	spin_lock(&pag->pagb_lock);
-
 	rbp = pag->pagb_tree.rb_node;
-
-	/* find closest start bno overlap */
 	while (rbp) {
 		busyp = rb_entry(rbp, struct xfs_extent_busy, rb_node);
 		if (bno < busyp->bno) {
@@ -115,7 +108,6 @@ xfs_extent_busy_search(
 		}
 	}
 	spin_unlock(&pag->pagb_lock);
-	xfs_perag_put(pag);
 	return match;
 }
 
@@ -281,17 +273,14 @@ out_force_log:
 void
 xfs_extent_busy_reuse(
 	struct xfs_mount	*mp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_agblock_t		fbno,
 	xfs_extlen_t		flen,
 	bool			userdata)
 {
-	struct xfs_perag	*pag;
 	struct rb_node		*rbp;
 
 	ASSERT(flen > 0);
-
-	pag = xfs_perag_get(mp, agno);
 	spin_lock(&pag->pagb_lock);
 restart:
 	rbp = pag->pagb_tree.rb_node;
@@ -314,7 +303,6 @@ restart:
 			goto restart;
 	}
 	spin_unlock(&pag->pagb_lock);
-	xfs_perag_put(pag);
 }
 
 /*
@@ -605,12 +593,11 @@ void
 xfs_extent_busy_wait_all(
 	struct xfs_mount	*mp)
 {
+	struct xfs_perag	*pag;
 	DEFINE_WAIT		(wait);
 	xfs_agnumber_t		agno;
 
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		struct xfs_perag *pag = xfs_perag_get(mp, agno);
-
+	for_each_perag(mp, agno, pag) {
 		do {
 			prepare_to_wait(&pag->pagb_wait, &wait, TASK_KILLABLE);
 			if  (RB_EMPTY_ROOT(&pag->pagb_tree))
@@ -618,8 +605,6 @@ xfs_extent_busy_wait_all(
 			schedule();
 		} while (1);
 		finish_wait(&pag->pagb_wait, &wait);
-
-		xfs_perag_put(pag);
 	}
 }
 
diff --git a/fs/xfs/xfs_extent_busy.h b/fs/xfs/xfs_extent_busy.h
index 8aea07100092..4a118131059f 100644
--- a/fs/xfs/xfs_extent_busy.h
+++ b/fs/xfs/xfs_extent_busy.h
@@ -9,6 +9,7 @@
 #define	__XFS_EXTENT_BUSY_H__
 
 struct xfs_mount;
+struct xfs_perag;
 struct xfs_trans;
 struct xfs_alloc_arg;
 
@@ -31,7 +32,7 @@ struct xfs_extent_busy {
 };
 
 void
-xfs_extent_busy_insert(struct xfs_trans *tp, xfs_agnumber_t agno,
+xfs_extent_busy_insert(struct xfs_trans *tp, struct xfs_perag *pag,
 	xfs_agblock_t bno, xfs_extlen_t len, unsigned int flags);
 
 void
@@ -39,11 +40,11 @@ xfs_extent_busy_clear(struct xfs_mount *mp, struct list_head *list,
 	bool do_discard);
 
 int
-xfs_extent_busy_search(struct xfs_mount *mp, xfs_agnumber_t agno,
+xfs_extent_busy_search(struct xfs_mount *mp, struct xfs_perag *pag,
 	xfs_agblock_t bno, xfs_extlen_t len);
 
 void
-xfs_extent_busy_reuse(struct xfs_mount *mp, xfs_agnumber_t agno,
+xfs_extent_busy_reuse(struct xfs_mount *mp, struct xfs_perag *pag,
 	xfs_agblock_t fbno, xfs_extlen_t flen, bool userdata);
 
 bool
diff --git a/fs/xfs/xfs_filestream.c b/fs/xfs/xfs_filestream.c
index db23e455eb91..eed6ca5f8f91 100644
--- a/fs/xfs/xfs_filestream.c
+++ b/fs/xfs/xfs_filestream.c
@@ -9,13 +9,13 @@
 #include "xfs_format.h"
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_inode.h"
 #include "xfs_bmap.h"
 #include "xfs_alloc.h"
 #include "xfs_mru_cache.h"
 #include "xfs_trace.h"
+#include "xfs_ag.h"
 #include "xfs_ag_resv.h"
 #include "xfs_trans.h"
 #include "xfs_filestream.h"
diff --git a/fs/xfs/xfs_fsmap.c b/fs/xfs/xfs_fsmap.c
index 34f2b971ce43..7d0b09c1366e 100644
--- a/fs/xfs/xfs_fsmap.c
+++ b/fs/xfs/xfs_fsmap.c
@@ -24,6 +24,7 @@
 #include "xfs_refcount_btree.h"
 #include "xfs_alloc_btree.h"
 #include "xfs_rtalloc.h"
+#include "xfs_ag.h"
 
 /* Convert an xfs_fsmap to an fsmap. */
 static void
@@ -157,10 +158,10 @@ struct xfs_getfsmap_info {
 	struct xfs_fsmap_head	*head;
 	struct fsmap		*fsmap_recs;	/* mapping records */
 	struct xfs_buf		*agf_bp;	/* AGF, for refcount queries */
+	struct xfs_perag	*pag;		/* AG info, if applicable */
 	xfs_daddr_t		next_daddr;	/* next daddr we expect */
 	u64			missing_owner;	/* owner of holes */
 	u32			dev;		/* device id */
-	xfs_agnumber_t		agno;		/* AG number, if applicable */
 	struct xfs_rmap_irec	low;		/* low rmap key */
 	struct xfs_rmap_irec	high;		/* high rmap key */
 	bool			last;		/* last extent? */
@@ -203,14 +204,13 @@ xfs_getfsmap_is_shared(
 	*stat = false;
 	if (!xfs_sb_version_hasreflink(&mp->m_sb))
 		return 0;
-	/* rt files will have agno set to NULLAGNUMBER */
-	if (info->agno == NULLAGNUMBER)
+	/* rt files will have no perag structure */
+	if (!info->pag)
 		return 0;
 
 	/* Are there any shared blocks here? */
 	flen = 0;
-	cur = xfs_refcountbt_init_cursor(mp, tp, info->agf_bp,
-			info->agno);
+	cur = xfs_refcountbt_init_cursor(mp, tp, info->agf_bp, info->pag);
 
 	error = xfs_refcount_find_shared(cur, rec->rm_startblock,
 			rec->rm_blockcount, &fbno, &flen, false);
@@ -311,7 +311,8 @@ xfs_getfsmap_helper(
 	if (info->head->fmh_entries >= info->head->fmh_count)
 		return -ECANCELED;
 
-	trace_xfs_fsmap_mapping(mp, info->dev, info->agno, rec);
+	trace_xfs_fsmap_mapping(mp, info->dev,
+			info->pag ? info->pag->pag_agno : NULLAGNUMBER, rec);
 
 	fmr.fmr_device = info->dev;
 	fmr.fmr_physical = rec_daddr;
@@ -354,7 +355,7 @@ xfs_getfsmap_datadev_helper(
 	xfs_fsblock_t			fsb;
 	xfs_daddr_t			rec_daddr;
 
-	fsb = XFS_AGB_TO_FSB(mp, cur->bc_ag.agno, rec->rm_startblock);
+	fsb = XFS_AGB_TO_FSB(mp, cur->bc_ag.pag->pag_agno, rec->rm_startblock);
 	rec_daddr = XFS_FSB_TO_DADDR(mp, fsb);
 
 	return xfs_getfsmap_helper(cur->bc_tp, info, rec, rec_daddr);
@@ -372,7 +373,7 @@ xfs_getfsmap_datadev_bnobt_helper(
 	struct xfs_rmap_irec		irec;
 	xfs_daddr_t			rec_daddr;
 
-	rec_daddr = XFS_AGB_TO_DADDR(mp, cur->bc_ag.agno,
+	rec_daddr = XFS_AGB_TO_DADDR(mp, cur->bc_ag.pag->pag_agno,
 			rec->ar_startblock);
 
 	irec.rm_startblock = rec->ar_startblock;
@@ -429,8 +430,8 @@ xfs_getfsmap_logdev(
 	info->high.rm_flags = XFS_RMAP_KEY_FLAGS | XFS_RMAP_REC_FLAGS;
 	info->missing_owner = XFS_FMR_OWN_FREE;
 
-	trace_xfs_fsmap_low_key(mp, info->dev, info->agno, &info->low);
-	trace_xfs_fsmap_high_key(mp, info->dev, info->agno, &info->high);
+	trace_xfs_fsmap_low_key(mp, info->dev, NULLAGNUMBER, &info->low);
+	trace_xfs_fsmap_high_key(mp, info->dev, NULLAGNUMBER, &info->high);
 
 	if (keys[0].fmr_physical > 0)
 		return 0;
@@ -508,8 +509,8 @@ __xfs_getfsmap_rtdev(
 	info->high.rm_blockcount = 0;
 	xfs_getfsmap_set_irec_flags(&info->high, &keys[1]);
 
-	trace_xfs_fsmap_low_key(mp, info->dev, info->agno, &info->low);
-	trace_xfs_fsmap_high_key(mp, info->dev, info->agno, &info->high);
+	trace_xfs_fsmap_low_key(mp, info->dev, NULLAGNUMBER, &info->low);
+	trace_xfs_fsmap_high_key(mp, info->dev, NULLAGNUMBER, &info->high);
 
 	return query_fn(tp, info);
 }
@@ -572,6 +573,7 @@ __xfs_getfsmap_datadev(
 	void				*priv)
 {
 	struct xfs_mount		*mp = tp->t_mountp;
+	struct xfs_perag		*pag;
 	struct xfs_btree_cur		*bt_cur = NULL;
 	xfs_fsblock_t			start_fsb;
 	xfs_fsblock_t			end_fsb;
@@ -610,20 +612,20 @@ __xfs_getfsmap_datadev(
 	start_ag = XFS_FSB_TO_AGNO(mp, start_fsb);
 	end_ag = XFS_FSB_TO_AGNO(mp, end_fsb);
 
-	/* Query each AG */
-	for (info->agno = start_ag; info->agno <= end_ag; info->agno++) {
+	for_each_perag_range(mp, start_ag, end_ag, pag) {
 		/*
 		 * Set the AG high key from the fsmap high key if this
 		 * is the last AG that we're querying.
 		 */
-		if (info->agno == end_ag) {
+		info->pag = pag;
+		if (pag->pag_agno == end_ag) {
 			info->high.rm_startblock = XFS_FSB_TO_AGBNO(mp,
 					end_fsb);
 			info->high.rm_offset = XFS_BB_TO_FSBT(mp,
 					keys[1].fmr_offset);
 			error = xfs_fsmap_owner_to_rmap(&info->high, &keys[1]);
 			if (error)
-				goto err;
+				break;
 			xfs_getfsmap_set_irec_flags(&info->high, &keys[1]);
 		}
 
@@ -634,38 +636,45 @@ __xfs_getfsmap_datadev(
 			info->agf_bp = NULL;
 		}
 
-		error = xfs_alloc_read_agf(mp, tp, info->agno, 0,
+		error = xfs_alloc_read_agf(mp, tp, pag->pag_agno, 0,
 				&info->agf_bp);
 		if (error)
-			goto err;
+			break;
 
-		trace_xfs_fsmap_low_key(mp, info->dev, info->agno, &info->low);
-		trace_xfs_fsmap_high_key(mp, info->dev, info->agno,
+		trace_xfs_fsmap_low_key(mp, info->dev, pag->pag_agno,
+				&info->low);
+		trace_xfs_fsmap_high_key(mp, info->dev, pag->pag_agno,
 				&info->high);
 
 		error = query_fn(tp, info, &bt_cur, priv);
 		if (error)
-			goto err;
+			break;
 
 		/*
 		 * Set the AG low key to the start of the AG prior to
 		 * moving on to the next AG.
 		 */
-		if (info->agno == start_ag) {
+		if (pag->pag_agno == start_ag) {
 			info->low.rm_startblock = 0;
 			info->low.rm_owner = 0;
 			info->low.rm_offset = 0;
 			info->low.rm_flags = 0;
 		}
-	}
 
-	/* Report any gap at the end of the AG */
-	info->last = true;
-	error = query_fn(tp, info, &bt_cur, priv);
-	if (error)
-		goto err;
+		/*
+		 * If this is the last AG, report any gap at the end of it
+		 * before we drop the reference to the perag when the loop
+		 * terminates.
+		 */
+		if (pag->pag_agno == end_ag) {
+			info->last = true;
+			error = query_fn(tp, info, &bt_cur, priv);
+			if (error)
+				break;
+		}
+		info->pag = NULL;
+	}
 
-err:
 	if (bt_cur)
 		xfs_btree_del_cursor(bt_cur, error < 0 ? XFS_BTREE_ERROR :
 							 XFS_BTREE_NOERROR);
@@ -673,6 +682,13 @@ err:
 		xfs_trans_brelse(tp, info->agf_bp);
 		info->agf_bp = NULL;
 	}
+	if (info->pag) {
+		xfs_perag_put(info->pag);
+		info->pag = NULL;
+	} else if (pag) {
+		/* loop termination case */
+		xfs_perag_put(pag);
+	}
 
 	return error;
 }
@@ -691,7 +707,7 @@ xfs_getfsmap_datadev_rmapbt_query(
 
 	/* Allocate cursor for this AG and query_range it. */
 	*curpp = xfs_rmapbt_init_cursor(tp->t_mountp, tp, info->agf_bp,
-			info->agno);
+			info->pag);
 	return xfs_rmap_query_range(*curpp, &info->low, &info->high,
 			xfs_getfsmap_datadev_helper, info);
 }
@@ -724,7 +740,7 @@ xfs_getfsmap_datadev_bnobt_query(
 
 	/* Allocate cursor for this AG and query_range it. */
 	*curpp = xfs_allocbt_init_cursor(tp->t_mountp, tp, info->agf_bp,
-			info->agno, XFS_BTNUM_BNO);
+			info->pag, XFS_BTNUM_BNO);
 	key->ar_startblock = info->low.rm_startblock;
 	key[1].ar_startblock = info->high.rm_startblock;
 	return xfs_alloc_query_range(*curpp, key, &key[1],
@@ -937,7 +953,7 @@ xfs_getfsmap(
 
 		info.dev = handlers[i].dev;
 		info.last = false;
-		info.agno = NULLAGNUMBER;
+		info.pag = NULL;
 		error = handlers[i].fn(tp, dkeys, &info);
 		if (error)
 			break;
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index be9cf88d2ad7..07c745cd483e 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -576,10 +576,8 @@ xfs_fs_reserve_ag_blocks(
 	int			err2;
 
 	mp->m_finobt_nores = false;
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		pag = xfs_perag_get(mp, agno);
+	for_each_perag(mp, agno, pag) {
 		err2 = xfs_ag_resv_init(pag, NULL);
-		xfs_perag_put(pag);
 		if (err2 && !error)
 			error = err2;
 	}
@@ -605,10 +603,8 @@ xfs_fs_unreserve_ag_blocks(
 	int			error = 0;
 	int			err2;
 
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		pag = xfs_perag_get(mp, agno);
+	for_each_perag(mp, agno, pag) {
 		err2 = xfs_ag_resv_free(pag);
-		xfs_perag_put(pag);
 		if (err2 && !error)
 			error = err2;
 	}
diff --git a/fs/xfs/xfs_health.c b/fs/xfs/xfs_health.c
index 8e0cb05a7142..5de3195f6cb2 100644
--- a/fs/xfs/xfs_health.c
+++ b/fs/xfs/xfs_health.c
@@ -9,11 +9,11 @@
 #include "xfs_format.h"
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_inode.h"
 #include "xfs_trace.h"
 #include "xfs_health.h"
+#include "xfs_ag.h"
 
 /*
  * Warn about metadata corruption that we detected but haven't fixed, and
@@ -34,14 +34,12 @@ xfs_health_unmount(
 		return;
 
 	/* Measure AG corruption levels. */
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		pag = xfs_perag_get(mp, agno);
+	for_each_perag(mp, agno, pag) {
 		xfs_ag_measure_sickness(pag, &sick, &checked);
 		if (sick) {
 			trace_xfs_ag_unfixed_corruption(mp, agno, sick);
 			warn = true;
 		}
-		xfs_perag_put(pag);
 	}
 
 	/* Measure realtime volume corruption levels. */
diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
index 3c81daca0e9a..7dad83a6f586 100644
--- a/fs/xfs/xfs_icache.c
+++ b/fs/xfs/xfs_icache.c
@@ -9,7 +9,6 @@
 #include "xfs_format.h"
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_inode.h"
 #include "xfs_trans.h"
@@ -23,6 +22,7 @@
 #include "xfs_dquot.h"
 #include "xfs_reflink.h"
 #include "xfs_ialloc.h"
+#include "xfs_ag.h"
 
 #include <linux/iversion.h>
 
@@ -1061,15 +1061,13 @@ xfs_reclaim_inodes_ag(
 	int			*nr_to_scan)
 {
 	struct xfs_perag	*pag;
-	xfs_agnumber_t		ag = 0;
+	xfs_agnumber_t		agno;
 
-	while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
+	for_each_perag_tag(mp, agno, pag, XFS_ICI_RECLAIM_TAG) {
 		unsigned long	first_index = 0;
 		int		done = 0;
 		int		nr_found = 0;
 
-		ag = pag->pag_agno + 1;
-
 		first_index = READ_ONCE(pag->pag_ici_reclaim_cursor);
 		do {
 			struct xfs_inode *batch[XFS_LOOKUP_BATCH];
@@ -1134,7 +1132,6 @@ xfs_reclaim_inodes_ag(
 		if (done)
 			first_index = 0;
 		WRITE_ONCE(pag->pag_ici_reclaim_cursor, first_index);
-		xfs_perag_put(pag);
 	}
 }
 
@@ -1554,14 +1551,6 @@ xfs_inode_clear_cowblocks_tag(
 	return xfs_blockgc_clear_iflag(ip, XFS_ICOWBLOCKS);
 }
 
-#define for_each_perag_tag(mp, next_agno, pag, tag) \
-	for ((next_agno) = 0, (pag) = xfs_perag_get_tag((mp), 0, (tag)); \
-		(pag) != NULL; \
-		(next_agno) = (pag)->pag_agno + 1, \
-		xfs_perag_put(pag), \
-		(pag) = xfs_perag_get_tag((mp), (next_agno), (tag)))
-
-
 /* Disable post-EOF and CoW block auto-reclamation. */
 void
 xfs_blockgc_stop(
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index e4c2da4566f1..3cdcfa67b290 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -11,7 +11,6 @@
 #include "xfs_format.h"
 #include "xfs_log_format.h"
 #include "xfs_trans_resv.h"
-#include "xfs_sb.h"
 #include "xfs_mount.h"
 #include "xfs_defer.h"
 #include "xfs_inode.h"
@@ -35,6 +34,7 @@
 #include "xfs_log.h"
 #include "xfs_bmap_btree.h"
 #include "xfs_reflink.h"
+#include "xfs_ag.h"
 
 kmem_zone_t *xfs_inode_zone;
 
@@ -45,7 +45,8 @@ kmem_zone_t *xfs_inode_zone;
 #define	XFS_ITRUNC_MAX_EXTENTS	2
 
 STATIC int xfs_iunlink(struct xfs_trans *, struct xfs_inode *);
-STATIC int xfs_iunlink_remove(struct xfs_trans *, struct xfs_inode *);
+STATIC int xfs_iunlink_remove(struct xfs_trans *tp, struct xfs_perag *pag,
+	struct xfs_inode *);
 
 /*
  * helper function to extract extent size hint from inode
@@ -778,7 +779,7 @@ xfs_inode_inherit_flags2(
  * Initialise a newly allocated inode and return the in-core inode to the
  * caller locked exclusively.
  */
-static int
+int
 xfs_init_new_inode(
 	struct user_namespace	*mnt_userns,
 	struct xfs_trans	*tp,
@@ -915,57 +916,6 @@ xfs_init_new_inode(
 }
 
 /*
- * Allocates a new inode from disk and return a pointer to the incore copy. This
- * routine will internally commit the current transaction and allocate a new one
- * if we needed to allocate more on-disk free inodes to perform the requested
- * operation.
- *
- * If we are allocating quota inodes, we do not have a parent inode to attach to
- * or associate with (i.e. dp == NULL) because they are not linked into the
- * directory structure - they are attached directly to the superblock - and so
- * have no parent.
- */
-int
-xfs_dir_ialloc(
-	struct user_namespace	*mnt_userns,
-	struct xfs_trans	**tpp,
-	struct xfs_inode	*dp,
-	umode_t			mode,
-	xfs_nlink_t		nlink,
-	dev_t			rdev,
-	prid_t			prid,
-	bool			init_xattrs,
-	struct xfs_inode	**ipp)
-{
-	struct xfs_buf		*agibp;
-	xfs_ino_t		parent_ino = dp ? dp->i_ino : 0;
-	xfs_ino_t		ino;
-	int			error;
-
-	ASSERT((*tpp)->t_flags & XFS_TRANS_PERM_LOG_RES);
-
-	/*
-	 * Call the space management code to pick the on-disk inode to be
-	 * allocated.
-	 */
-	error = xfs_dialloc_select_ag(tpp, parent_ino, mode, &agibp);
-	if (error)
-		return error;
-
-	if (!agibp)
-		return -ENOSPC;
-
-	/* Allocate an inode from the selected AG */
-	error = xfs_dialloc_ag(*tpp, agibp, parent_ino, &ino);
-	if (error)
-		return error;
-	ASSERT(ino != NULLFSINO);
-
-	return xfs_init_new_inode(mnt_userns, *tpp, dp, ino, mode, nlink, rdev,
-				  prid, init_xattrs, ipp);
-}
-
-/*
  * Decrement the link count on an inode & log the change.  If this causes the
  * link count to go to zero, move the inode to AGI unlinked list so that it can
  * be freed when the last active reference goes away via xfs_inactive().
@@ -1022,6 +972,7 @@ xfs_create(
 	struct xfs_dquot	*pdqp = NULL;
 	struct xfs_trans_res	*tres;
 	uint			resblks;
+	xfs_ino_t		ino;
 
 	trace_xfs_create(dp, name);
 
@@ -1078,14 +1029,16 @@ xfs_create(
 	 * entry pointing to them, but a directory also the "." entry
 	 * pointing to itself.
 	 */
-	error = xfs_dir_ialloc(mnt_userns, &tp, dp, mode, is_dir ? 2 : 1, rdev,
-			       prid, init_xattrs, &ip);
+	error = xfs_dialloc(&tp, dp->i_ino, mode, &ino);
+	if (!error)
+		error = xfs_init_new_inode(mnt_userns, tp, dp, ino, mode,
+				is_dir ? 2 : 1, rdev, prid, init_xattrs, &ip);
 	if (error)
 		goto out_trans_cancel;
 
 	/*
 	 * Now we join the directory inode to the transaction.  We do not do it
-	 * earlier because xfs_dir_ialloc might commit the previous transaction
+	 * earlier because xfs_dialloc might commit the previous transaction
 	 * (and release all the locks).  An error from here on will result in
 	 * the transaction cancel unlocking dp so don't do it explicitly in the
 	 * error path.
@@ -1175,6 +1128,7 @@ xfs_create_tmpfile(
 	struct xfs_dquot	*pdqp = NULL;
 	struct xfs_trans_res	*tres;
 	uint			resblks;
+	xfs_ino_t		ino;
 
 	if (XFS_FORCED_SHUTDOWN(mp))
 		return -EIO;
@@ -1199,8 +1153,10 @@ xfs_create_tmpfile(
 	if (error)
 		goto out_release_dquots;
 
-	error = xfs_dir_ialloc(mnt_userns, &tp, dp, mode, 0, 0, prid,
-				false, &ip);
+	error = xfs_dialloc(&tp, dp->i_ino, mode, &ino);
+	if (!error)
+		error = xfs_init_new_inode(mnt_userns, tp, dp, ino, mode,
+				0, 0, prid, false, &ip);
 	if (error)
 		goto out_trans_cancel;
 
@@ -1315,7 +1271,11 @@ xfs_link(
 	 * Handle initial link state of O_TMPFILE inode
 	 */
 	if (VFS_I(sip)->i_nlink == 0) {
-		error = xfs_iunlink_remove(tp, sip);
+		struct xfs_perag	*pag;
+
+		pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, sip->i_ino));
+		error = xfs_iunlink_remove(tp, pag, sip);
+		xfs_perag_put(pag);
 		if (error)
 			goto error_return;
 	}
@@ -2008,7 +1968,7 @@ xfs_iunlink_destroy(
 STATIC int
 xfs_iunlink_update_bucket(
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	struct xfs_buf		*agibp,
 	unsigned int		bucket_index,
 	xfs_agino_t		new_agino)
@@ -2017,10 +1977,10 @@ xfs_iunlink_update_bucket(
 	xfs_agino_t		old_value;
 	int			offset;
 
-	ASSERT(xfs_verify_agino_or_null(tp->t_mountp, agno, new_agino));
+	ASSERT(xfs_verify_agino_or_null(tp->t_mountp, pag->pag_agno, new_agino));
 
 	old_value = be32_to_cpu(agi->agi_unlinked[bucket_index]);
-	trace_xfs_iunlink_update_bucket(tp->t_mountp, agno, bucket_index,
+	trace_xfs_iunlink_update_bucket(tp->t_mountp, pag->pag_agno, bucket_index,
 			old_value, new_agino);
 
 	/*
@@ -2044,7 +2004,7 @@ xfs_iunlink_update_bucket(
 STATIC void
 xfs_iunlink_update_dinode(
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_agino_t		agino,
 	struct xfs_buf		*ibp,
 	struct xfs_dinode	*dip,
@@ -2054,9 +2014,9 @@ xfs_iunlink_update_dinode(
 	struct xfs_mount	*mp = tp->t_mountp;
 	int			offset;
 
-	ASSERT(xfs_verify_agino_or_null(mp, agno, next_agino));
+	ASSERT(xfs_verify_agino_or_null(mp, pag->pag_agno, next_agino));
 
-	trace_xfs_iunlink_update_dinode(mp, agno, agino,
+	trace_xfs_iunlink_update_dinode(mp, pag->pag_agno, agino,
 			be32_to_cpu(dip->di_next_unlinked), next_agino);
 
 	dip->di_next_unlinked = cpu_to_be32(next_agino);
@@ -2074,7 +2034,7 @@ STATIC int
 xfs_iunlink_update_inode(
 	struct xfs_trans	*tp,
 	struct xfs_inode	*ip,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_agino_t		next_agino,
 	xfs_agino_t		*old_next_agino)
 {
@@ -2084,7 +2044,7 @@ xfs_iunlink_update_inode(
 	xfs_agino_t		old_value;
 	int			error;
 
-	ASSERT(xfs_verify_agino_or_null(mp, agno, next_agino));
+	ASSERT(xfs_verify_agino_or_null(mp, pag->pag_agno, next_agino));
 
 	error = xfs_imap_to_bp(mp, tp, &ip->i_imap, &ibp);
 	if (error)
@@ -2093,7 +2053,7 @@ xfs_iunlink_update_inode(
 
 	/* Make sure the old pointer isn't garbage. */
 	old_value = be32_to_cpu(dip->di_next_unlinked);
-	if (!xfs_verify_agino_or_null(mp, agno, old_value)) {
+	if (!xfs_verify_agino_or_null(mp, pag->pag_agno, old_value)) {
 		xfs_inode_verifier_error(ip, -EFSCORRUPTED, __func__, dip,
 				sizeof(*dip), __this_address);
 		error = -EFSCORRUPTED;
@@ -2116,7 +2076,7 @@ xfs_iunlink_update_inode(
 	}
 
 	/* Ok, update the new pointer. */
-	xfs_iunlink_update_dinode(tp, agno, XFS_INO_TO_AGINO(mp, ip->i_ino),
+	xfs_iunlink_update_dinode(tp, pag, XFS_INO_TO_AGINO(mp, ip->i_ino),
 			ibp, dip, &ip->i_imap, next_agino);
 	return 0;
 out:
@@ -2137,10 +2097,10 @@ xfs_iunlink(
 	struct xfs_inode	*ip)
 {
 	struct xfs_mount	*mp = tp->t_mountp;
+	struct xfs_perag	*pag;
 	struct xfs_agi		*agi;
 	struct xfs_buf		*agibp;
 	xfs_agino_t		next_agino;
-	xfs_agnumber_t		agno = XFS_INO_TO_AGNO(mp, ip->i_ino);
 	xfs_agino_t		agino = XFS_INO_TO_AGINO(mp, ip->i_ino);
 	short			bucket_index = agino % XFS_AGI_UNLINKED_BUCKETS;
 	int			error;
@@ -2149,10 +2109,12 @@ xfs_iunlink(
 	ASSERT(VFS_I(ip)->i_mode != 0);
 	trace_xfs_iunlink(ip);
 
+	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
+
 	/* Get the agi buffer first.  It ensures lock ordering on the list. */
-	error = xfs_read_agi(mp, tp, agno, &agibp);
+	error = xfs_read_agi(mp, tp, pag->pag_agno, &agibp);
 	if (error)
-		return error;
+		goto out;
 	agi = agibp->b_addr;
 
 	/*
@@ -2162,9 +2124,10 @@ xfs_iunlink(
 	 */
 	next_agino = be32_to_cpu(agi->agi_unlinked[bucket_index]);
 	if (next_agino == agino ||
-	    !xfs_verify_agino_or_null(mp, agno, next_agino)) {
+	    !xfs_verify_agino_or_null(mp, pag->pag_agno, next_agino)) {
 		xfs_buf_mark_corrupt(agibp);
-		return -EFSCORRUPTED;
+		error = -EFSCORRUPTED;
+		goto out;
 	}
 
 	if (next_agino != NULLAGINO) {
@@ -2174,23 +2137,26 @@ xfs_iunlink(
 		 * There is already another inode in the bucket, so point this
 		 * inode to the current head of the list.
 		 */
-		error = xfs_iunlink_update_inode(tp, ip, agno, next_agino,
+		error = xfs_iunlink_update_inode(tp, ip, pag, next_agino,
 				&old_agino);
 		if (error)
-			return error;
+			goto out;
 		ASSERT(old_agino == NULLAGINO);
 
 		/*
 		 * agino has been unlinked, add a backref from the next inode
 		 * back to agino.
 		 */
-		error = xfs_iunlink_add_backref(agibp->b_pag, agino, next_agino);
+		error = xfs_iunlink_add_backref(pag, agino, next_agino);
 		if (error)
-			return error;
+			goto out;
 	}
 
 	/* Point the head of the list to point to this inode. */
-	return xfs_iunlink_update_bucket(tp, agno, agibp, bucket_index, agino);
+	error = xfs_iunlink_update_bucket(tp, pag, agibp, bucket_index, agino);
+out:
+	xfs_perag_put(pag);
+	return error;
 }
 
 /* Return the imap, dinode pointer, and buffer for an inode. */
@@ -2238,14 +2204,13 @@ xfs_iunlink_map_ino(
 STATIC int
 xfs_iunlink_map_prev(
 	struct xfs_trans	*tp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_agino_t		head_agino,
 	xfs_agino_t		target_agino,
 	xfs_agino_t		*agino,
 	struct xfs_imap		*imap,
 	struct xfs_dinode	**dipp,
-	struct xfs_buf		**bpp,
-	struct xfs_perag	*pag)
+	struct xfs_buf		**bpp)
 {
 	struct xfs_mount	*mp = tp->t_mountp;
 	xfs_agino_t		next_agino;
@@ -2257,7 +2222,8 @@ xfs_iunlink_map_prev(
 	/* See if our backref cache can find it faster. */
 	*agino = xfs_iunlink_lookup_backref(pag, target_agino);
 	if (*agino != NULLAGINO) {
-		error = xfs_iunlink_map_ino(tp, agno, *agino, imap, dipp, bpp);
+		error = xfs_iunlink_map_ino(tp, pag->pag_agno, *agino, imap,
+				dipp, bpp);
 		if (error)
 			return error;
 
@@ -2273,7 +2239,7 @@ xfs_iunlink_map_prev(
 		WARN_ON_ONCE(1);
 	}
 
-	trace_xfs_iunlink_map_prev_fallback(mp, agno);
+	trace_xfs_iunlink_map_prev_fallback(mp, pag->pag_agno);
 
 	/* Otherwise, walk the entire bucket until we find it. */
 	next_agino = head_agino;
@@ -2284,8 +2250,8 @@ xfs_iunlink_map_prev(
 			xfs_trans_brelse(tp, *bpp);
 
 		*agino = next_agino;
-		error = xfs_iunlink_map_ino(tp, agno, next_agino, imap, dipp,
-				bpp);
+		error = xfs_iunlink_map_ino(tp, pag->pag_agno, next_agino, imap,
+				dipp, bpp);
 		if (error)
 			return error;
 
@@ -2294,7 +2260,7 @@ xfs_iunlink_map_prev(
 		 * Make sure this pointer is valid and isn't an obvious
 		 * infinite loop.
 		 */
-		if (!xfs_verify_agino(mp, agno, unlinked_agino) ||
+		if (!xfs_verify_agino(mp, pag->pag_agno, unlinked_agino) ||
 		    next_agino == unlinked_agino) {
 			XFS_CORRUPTION_ERROR(__func__,
 					XFS_ERRLEVEL_LOW, mp,
@@ -2314,6 +2280,7 @@ xfs_iunlink_map_prev(
 STATIC int
 xfs_iunlink_remove(
 	struct xfs_trans	*tp,
+	struct xfs_perag	*pag,
 	struct xfs_inode	*ip)
 {
 	struct xfs_mount	*mp = tp->t_mountp;
@@ -2321,7 +2288,6 @@ xfs_iunlink_remove(
 	struct xfs_buf		*agibp;
 	struct xfs_buf		*last_ibp;
 	struct xfs_dinode	*last_dip = NULL;
-	xfs_agnumber_t		agno = XFS_INO_TO_AGNO(mp, ip->i_ino);
 	xfs_agino_t		agino = XFS_INO_TO_AGINO(mp, ip->i_ino);
 	xfs_agino_t		next_agino;
 	xfs_agino_t		head_agino;
@@ -2331,7 +2297,7 @@ xfs_iunlink_remove(
 	trace_xfs_iunlink_remove(ip);
 
 	/* Get the agi buffer first.  It ensures lock ordering on the list. */
-	error = xfs_read_agi(mp, tp, agno, &agibp);
+	error = xfs_read_agi(mp, tp, pag->pag_agno, &agibp);
 	if (error)
 		return error;
 	agi = agibp->b_addr;
@@ -2341,7 +2307,7 @@ xfs_iunlink_remove(
 	 * go on.  Make sure the head pointer isn't garbage.
 	 */
 	head_agino = be32_to_cpu(agi->agi_unlinked[bucket_index]);
-	if (!xfs_verify_agino(mp, agno, head_agino)) {
+	if (!xfs_verify_agino(mp, pag->pag_agno, head_agino)) {
 		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
 				agi, sizeof(*agi));
 		return -EFSCORRUPTED;
@@ -2352,7 +2318,7 @@ xfs_iunlink_remove(
 	 * the old pointer value so that we can update whatever was previous
 	 * to us in the list to point to whatever was next in the list.
 	 */
-	error = xfs_iunlink_update_inode(tp, ip, agno, NULLAGINO, &next_agino);
+	error = xfs_iunlink_update_inode(tp, ip, pag, NULLAGINO, &next_agino);
 	if (error)
 		return error;
 
@@ -2364,8 +2330,7 @@ xfs_iunlink_remove(
 	 * this inode's backref to point from the next inode.
 	 */
 	if (next_agino != NULLAGINO) {
-		error = xfs_iunlink_change_backref(agibp->b_pag, next_agino,
-				NULLAGINO);
+		error = xfs_iunlink_change_backref(pag, next_agino, NULLAGINO);
 		if (error)
 			return error;
 	}
@@ -2375,14 +2340,13 @@ xfs_iunlink_remove(
 		xfs_agino_t	prev_agino;
 
 		/* We need to search the list for the inode being freed. */
-		error = xfs_iunlink_map_prev(tp, agno, head_agino, agino,
-				&prev_agino, &imap, &last_dip, &last_ibp,
-				agibp->b_pag);
+		error = xfs_iunlink_map_prev(tp, pag, head_agino, agino,
+				&prev_agino, &imap, &last_dip, &last_ibp);
 		if (error)
 			return error;
 
 		/* Point the previous inode on the list to the next inode. */
-		xfs_iunlink_update_dinode(tp, agno, prev_agino, last_ibp,
+		xfs_iunlink_update_dinode(tp, pag, prev_agino, last_ibp,
 				last_dip, &imap, next_agino);
 
 		/*
@@ -2398,7 +2362,7 @@ xfs_iunlink_remove(
 	}
 
 	/* Point the head of the list to the next unlinked inode. */
-	return xfs_iunlink_update_bucket(tp, agno, agibp, bucket_index,
+	return xfs_iunlink_update_bucket(tp, pag, agibp, bucket_index,
 			next_agino);
 }
 
@@ -2409,12 +2373,11 @@ xfs_iunlink_remove(
  */
 static void
 xfs_ifree_mark_inode_stale(
-	struct xfs_buf		*bp,
+	struct xfs_perag	*pag,
 	struct xfs_inode	*free_ip,
 	xfs_ino_t		inum)
 {
-	struct xfs_mount	*mp = bp->b_mount;
-	struct xfs_perag	*pag = bp->b_pag;
+	struct xfs_mount	*mp = pag->pag_mount;
 	struct xfs_inode_log_item *iip;
 	struct xfs_inode	*ip;
 
@@ -2504,10 +2467,11 @@ out_iflags_unlock:
  * inodes that are in memory - they all must be marked stale and attached to
  * the cluster buffer.
  */
-STATIC int
+static int
 xfs_ifree_cluster(
-	struct xfs_inode	*free_ip,
 	struct xfs_trans	*tp,
+	struct xfs_perag	*pag,
+	struct xfs_inode	*free_ip,
 	struct xfs_icluster	*xic)
 {
 	struct xfs_mount	*mp = free_ip->i_mount;
@@ -2569,7 +2533,7 @@ xfs_ifree_cluster(
 		 * already marked XFS_ISTALE.
 		 */
 		for (i = 0; i < igeo->inodes_per_cluster; i++)
-			xfs_ifree_mark_inode_stale(bp, free_ip, inum + i);
+			xfs_ifree_mark_inode_stale(pag, free_ip, inum + i);
 
 		xfs_trans_stale_inode_buf(tp, bp);
 		xfs_trans_binval(tp, bp);
@@ -2592,9 +2556,11 @@ xfs_ifree(
 	struct xfs_trans	*tp,
 	struct xfs_inode	*ip)
 {
-	int			error;
+	struct xfs_mount	*mp = ip->i_mount;
+	struct xfs_perag	*pag;
 	struct xfs_icluster	xic = { 0 };
 	struct xfs_inode_log_item *iip = ip->i_itemp;
+	int			error;
 
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
 	ASSERT(VFS_I(ip)->i_nlink == 0);
@@ -2602,16 +2568,18 @@ xfs_ifree(
 	ASSERT(ip->i_disk_size == 0 || !S_ISREG(VFS_I(ip)->i_mode));
 	ASSERT(ip->i_nblocks == 0);
 
+	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
+
 	/*
 	 * Pull the on-disk inode from the AGI unlinked list.
 	 */
-	error = xfs_iunlink_remove(tp, ip);
+	error = xfs_iunlink_remove(tp, pag, ip);
 	if (error)
-		return error;
+		goto out;
 
-	error = xfs_difree(tp, ip->i_ino, &xic);
+	error = xfs_difree(tp, pag, ip->i_ino, &xic);
 	if (error)
-		return error;
+		goto out;
 
 	/*
 	 * Free any local-format data sitting around before we reset the
@@ -2626,7 +2594,7 @@ xfs_ifree(
 
 	VFS_I(ip)->i_mode = 0;		/* mark incore inode as free */
 	ip->i_diflags = 0;
-	ip->i_diflags2 = ip->i_mount->m_ino_geo.new_diflags2;
+	ip->i_diflags2 = mp->m_ino_geo.new_diflags2;
 	ip->i_forkoff = 0;		/* mark the attr fork not in use */
 	ip->i_df.if_format = XFS_DINODE_FMT_EXTENTS;
 	if (xfs_iflags_test(ip, XFS_IPRESERVE_DM_FIELDS))
@@ -2645,8 +2613,9 @@ xfs_ifree(
 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 
 	if (xic.deleted)
-		error = xfs_ifree_cluster(ip, tp, &xic);
-
+		error = xfs_ifree_cluster(tp, pag, ip, &xic);
+out:
+	xfs_perag_put(pag);
 	return error;
 }
 
@@ -3250,8 +3219,13 @@ xfs_rename(
 	 * in future.
 	 */
 	if (wip) {
+		struct xfs_perag	*pag;
+
 		ASSERT(VFS_I(wip)->i_nlink == 0);
-		error = xfs_iunlink_remove(tp, wip);
+
+		pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, wip->i_ino));
+		error = xfs_iunlink_remove(tp, pag, wip);
+		xfs_perag_put(pag);
 		if (error)
 			goto out_trans_cancel;
 
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index ca826cfba91c..4b6703dbffb8 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -431,11 +431,10 @@ void		xfs_lock_two_inodes(struct xfs_inode *ip0, uint ip0_mode,
 xfs_extlen_t	xfs_get_extsz_hint(struct xfs_inode *ip);
 xfs_extlen_t	xfs_get_cowextsz_hint(struct xfs_inode *ip);
 
-int		xfs_dir_ialloc(struct user_namespace *mnt_userns,
-			       struct xfs_trans **tpp, struct xfs_inode *dp,
-			       umode_t mode, xfs_nlink_t nlink, dev_t dev,
-			       prid_t prid, bool need_xattr,
-			       struct xfs_inode **ipp);
+int xfs_init_new_inode(struct user_namespace *mnt_userns, struct xfs_trans *tp,
+		struct xfs_inode *pip, xfs_ino_t ino, umode_t mode,
+		xfs_nlink_t nlink, dev_t rdev, prid_t prid, bool init_xattrs,
+		struct xfs_inode **ipp);
 
 static inline int
 xfs_itruncate_extents(
diff --git a/fs/xfs/xfs_iwalk.c b/fs/xfs/xfs_iwalk.c
index c4a340f1f1e1..917d51eefee3 100644
--- a/fs/xfs/xfs_iwalk.c
+++ b/fs/xfs/xfs_iwalk.c
@@ -21,6 +21,7 @@
 #include "xfs_health.h"
 #include "xfs_trans.h"
 #include "xfs_pwork.h"
+#include "xfs_ag.h"
 
 /*
  * Walking Inodes in the Filesystem
@@ -51,6 +52,7 @@ struct xfs_iwalk_ag {
 
 	struct xfs_mount		*mp;
 	struct xfs_trans		*tp;
+	struct xfs_perag		*pag;
 
 	/* Where do we start the traversal? */
 	xfs_ino_t			startino;
@@ -90,7 +92,7 @@ struct xfs_iwalk_ag {
 STATIC void
 xfs_iwalk_ichunk_ra(
 	struct xfs_mount		*mp,
-	xfs_agnumber_t			agno,
+	struct xfs_perag		*pag,
 	struct xfs_inobt_rec_incore	*irec)
 {
 	struct xfs_ino_geometry		*igeo = M_IGEO(mp);
@@ -106,7 +108,7 @@ xfs_iwalk_ichunk_ra(
 
 		imask = xfs_inobt_maskn(i, igeo->inodes_per_cluster);
 		if (imask & ~irec->ir_free) {
-			xfs_btree_reada_bufs(mp, agno, agbno,
+			xfs_btree_reada_bufs(mp, pag->pag_agno, agbno,
 					igeo->blocks_per_cluster,
 					&xfs_inode_buf_ops);
 		}
@@ -174,26 +176,25 @@ xfs_iwalk_free(
 /* For each inuse inode in each cached inobt record, call our function. */
 STATIC int
 xfs_iwalk_ag_recs(
-	struct xfs_iwalk_ag		*iwag)
+	struct xfs_iwalk_ag	*iwag)
 {
-	struct xfs_mount		*mp = iwag->mp;
-	struct xfs_trans		*tp = iwag->tp;
-	xfs_ino_t			ino;
-	unsigned int			i, j;
-	xfs_agnumber_t			agno;
-	int				error;
+	struct xfs_mount	*mp = iwag->mp;
+	struct xfs_trans	*tp = iwag->tp;
+	struct xfs_perag	*pag = iwag->pag;
+	xfs_ino_t		ino;
+	unsigned int		i, j;
+	int			error;
 
-	agno = XFS_INO_TO_AGNO(mp, iwag->startino);
 	for (i = 0; i < iwag->nr_recs; i++) {
 		struct xfs_inobt_rec_incore	*irec = &iwag->recs[i];
 
-		trace_xfs_iwalk_ag_rec(mp, agno, irec);
+		trace_xfs_iwalk_ag_rec(mp, pag->pag_agno, irec);
 
 		if (xfs_pwork_want_abort(&iwag->pwork))
 			return 0;
 
 		if (iwag->inobt_walk_fn) {
-			error = iwag->inobt_walk_fn(mp, tp, agno, irec,
+			error = iwag->inobt_walk_fn(mp, tp, pag->pag_agno, irec,
 					iwag->data);
 			if (error)
 				return error;
@@ -211,7 +212,8 @@ xfs_iwalk_ag_recs(
 				continue;
 
 			/* Otherwise call our function. */
-			ino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino + j);
+			ino = XFS_AGINO_TO_INO(mp, pag->pag_agno,
+						irec->ir_startino + j);
 			error = iwag->iwalk_fn(mp, tp, ino, iwag->data);
 			if (error)
 				return error;
@@ -257,7 +259,6 @@ xfs_iwalk_del_inobt(
 STATIC int
 xfs_iwalk_ag_start(
 	struct xfs_iwalk_ag	*iwag,
-	xfs_agnumber_t		agno,
 	xfs_agino_t		agino,
 	struct xfs_btree_cur	**curpp,
 	struct xfs_buf		**agi_bpp,
@@ -265,12 +266,13 @@ xfs_iwalk_ag_start(
 {
 	struct xfs_mount	*mp = iwag->mp;
 	struct xfs_trans	*tp = iwag->tp;
+	struct xfs_perag	*pag = iwag->pag;
 	struct xfs_inobt_rec_incore *irec;
 	int			error;
 
 	/* Set up a fresh cursor and empty the inobt cache. */
 	iwag->nr_recs = 0;
-	error = xfs_inobt_cur(mp, tp, agno, XFS_BTNUM_INO, curpp, agi_bpp);
+	error = xfs_inobt_cur(mp, tp, pag, XFS_BTNUM_INO, curpp, agi_bpp);
 	if (error)
 		return error;
 
@@ -304,7 +306,7 @@ xfs_iwalk_ag_start(
 	if (XFS_IS_CORRUPT(mp, *has_more != 1))
 		return -EFSCORRUPTED;
 
-	iwag->lastino = XFS_AGINO_TO_INO(mp, agno,
+	iwag->lastino = XFS_AGINO_TO_INO(mp, pag->pag_agno,
 				irec->ir_startino + XFS_INODES_PER_CHUNK - 1);
 
 	/*
@@ -345,7 +347,6 @@ out_advance:
 STATIC int
 xfs_iwalk_run_callbacks(
 	struct xfs_iwalk_ag		*iwag,
-	xfs_agnumber_t			agno,
 	struct xfs_btree_cur		**curpp,
 	struct xfs_buf			**agi_bpp,
 	int				*has_more)
@@ -376,7 +377,7 @@ xfs_iwalk_run_callbacks(
 		return 0;
 
 	/* ...and recreate the cursor just past where we left off. */
-	error = xfs_inobt_cur(mp, tp, agno, XFS_BTNUM_INO, curpp, agi_bpp);
+	error = xfs_inobt_cur(mp, tp, iwag->pag, XFS_BTNUM_INO, curpp, agi_bpp);
 	if (error)
 		return error;
 
@@ -390,17 +391,17 @@ xfs_iwalk_ag(
 {
 	struct xfs_mount		*mp = iwag->mp;
 	struct xfs_trans		*tp = iwag->tp;
+	struct xfs_perag		*pag = iwag->pag;
 	struct xfs_buf			*agi_bp = NULL;
 	struct xfs_btree_cur		*cur = NULL;
-	xfs_agnumber_t			agno;
 	xfs_agino_t			agino;
 	int				has_more;
 	int				error = 0;
 
 	/* Set up our cursor at the right place in the inode btree. */
-	agno = XFS_INO_TO_AGNO(mp, iwag->startino);
+	ASSERT(pag->pag_agno == XFS_INO_TO_AGNO(mp, iwag->startino));
 	agino = XFS_INO_TO_AGINO(mp, iwag->startino);
-	error = xfs_iwalk_ag_start(iwag, agno, agino, &cur, &agi_bp, &has_more);
+	error = xfs_iwalk_ag_start(iwag, agino, &cur, &agi_bp, &has_more);
 
 	while (!error && has_more) {
 		struct xfs_inobt_rec_incore	*irec;
@@ -417,7 +418,7 @@ xfs_iwalk_ag(
 			break;
 
 		/* Make sure that we always move forward. */
-		rec_fsino = XFS_AGINO_TO_INO(mp, agno, irec->ir_startino);
+		rec_fsino = XFS_AGINO_TO_INO(mp, pag->pag_agno, irec->ir_startino);
 		if (iwag->lastino != NULLFSINO &&
 		    XFS_IS_CORRUPT(mp, iwag->lastino >= rec_fsino)) {
 			error = -EFSCORRUPTED;
@@ -438,7 +439,7 @@ xfs_iwalk_ag(
 		 * walking the inodes.
 		 */
 		if (iwag->iwalk_fn)
-			xfs_iwalk_ichunk_ra(mp, agno, irec);
+			xfs_iwalk_ichunk_ra(mp, pag, irec);
 
 		/*
 		 * If there's space in the buffer for more records, increment
@@ -458,15 +459,14 @@ xfs_iwalk_ag(
 		 * we would be if we had been able to increment like above.
 		 */
 		ASSERT(has_more);
-		error = xfs_iwalk_run_callbacks(iwag, agno, &cur, &agi_bp,
-				&has_more);
+		error = xfs_iwalk_run_callbacks(iwag, &cur, &agi_bp, &has_more);
 	}
 
 	if (iwag->nr_recs == 0 || error)
 		goto out;
 
 	/* Walk the unprocessed records in the cache. */
-	error = xfs_iwalk_run_callbacks(iwag, agno, &cur, &agi_bp, &has_more);
+	error = xfs_iwalk_run_callbacks(iwag, &cur, &agi_bp, &has_more);
 
 out:
 	xfs_iwalk_del_inobt(tp, &cur, &agi_bp, error);
@@ -555,6 +555,7 @@ xfs_iwalk(
 		.pwork		= XFS_PWORK_SINGLE_THREADED,
 		.lastino	= NULLFSINO,
 	};
+	struct xfs_perag	*pag;
 	xfs_agnumber_t		agno = XFS_INO_TO_AGNO(mp, startino);
 	int			error;
 
@@ -565,15 +566,19 @@ xfs_iwalk(
 	if (error)
 		return error;
 
-	for (; agno < mp->m_sb.sb_agcount; agno++) {
+	for_each_perag_from(mp, agno, pag) {
+		iwag.pag = pag;
 		error = xfs_iwalk_ag(&iwag);
 		if (error)
 			break;
 		iwag.startino = XFS_AGINO_TO_INO(mp, agno + 1, 0);
 		if (flags & XFS_INOBT_WALK_SAME_AG)
 			break;
+		iwag.pag = NULL;
 	}
 
+	if (iwag.pag)
+		xfs_perag_put(pag);
 	xfs_iwalk_free(&iwag);
 	return error;
 }
@@ -598,6 +603,7 @@ xfs_iwalk_ag_work(
 	error = xfs_iwalk_ag(iwag);
 	xfs_iwalk_free(iwag);
 out:
+	xfs_perag_put(iwag->pag);
 	kmem_free(iwag);
 	return error;
 }
@@ -617,6 +623,7 @@ xfs_iwalk_threaded(
 	void			*data)
 {
 	struct xfs_pwork_ctl	pctl;
+	struct xfs_perag	*pag;
 	xfs_agnumber_t		agno = XFS_INO_TO_AGNO(mp, startino);
 	int			error;
 
@@ -627,7 +634,7 @@ xfs_iwalk_threaded(
 	if (error)
 		return error;
 
-	for (; agno < mp->m_sb.sb_agcount; agno++) {
+	for_each_perag_from(mp, agno, pag) {
 		struct xfs_iwalk_ag	*iwag;
 
 		if (xfs_pwork_ctl_want_abort(&pctl))
@@ -635,17 +642,25 @@ xfs_iwalk_threaded(
 
 		iwag = kmem_zalloc(sizeof(struct xfs_iwalk_ag), 0);
 		iwag->mp = mp;
+
+		/*
+		 * perag is being handed off to async work, so take another
+		 * reference for the async work to release.
+		 */
+		atomic_inc(&pag->pag_ref);
+		iwag->pag = pag;
 		iwag->iwalk_fn = iwalk_fn;
 		iwag->data = data;
 		iwag->startino = startino;
 		iwag->sz_recs = xfs_iwalk_prefetch(inode_records);
 		iwag->lastino = NULLFSINO;
 		xfs_pwork_queue(&pctl, &iwag->pwork);
-		startino = XFS_AGINO_TO_INO(mp, agno + 1, 0);
+		startino = XFS_AGINO_TO_INO(mp, pag->pag_agno + 1, 0);
 		if (flags & XFS_INOBT_WALK_SAME_AG)
 			break;
 	}
-
+	if (pag)
+		xfs_perag_put(pag);
 	if (polled)
 		xfs_pwork_poll(&pctl);
 	return xfs_pwork_destroy(&pctl);
@@ -715,6 +730,7 @@ xfs_inobt_walk(
 		.pwork		= XFS_PWORK_SINGLE_THREADED,
 		.lastino	= NULLFSINO,
 	};
+	struct xfs_perag	*pag;
 	xfs_agnumber_t		agno = XFS_INO_TO_AGNO(mp, startino);
 	int			error;
 
@@ -725,15 +741,19 @@ xfs_inobt_walk(
 	if (error)
 		return error;
 
-	for (; agno < mp->m_sb.sb_agcount; agno++) {
+	for_each_perag_from(mp, agno, pag) {
+		iwag.pag = pag;
 		error = xfs_iwalk_ag(&iwag);
 		if (error)
 			break;
-		iwag.startino = XFS_AGINO_TO_INO(mp, agno + 1, 0);
+		iwag.startino = XFS_AGINO_TO_INO(mp, pag->pag_agno + 1, 0);
 		if (flags & XFS_INOBT_WALK_SAME_AG)
 			break;
+		iwag.pag = NULL;
 	}
 
+	if (iwag.pag)
+		xfs_perag_put(pag);
 	xfs_iwalk_free(&iwag);
 	return error;
 }
diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index e5dd1c0c2f03..1227503d2246 100644
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -25,6 +25,7 @@
 #include "xfs_icache.h"
 #include "xfs_error.h"
 #include "xfs_buf_item.h"
+#include "xfs_ag.h"
 
 #define BLK_AVG(blk1, blk2)	((blk1+blk2) >> 1)
 
@@ -2741,21 +2742,17 @@ STATIC void
 xlog_recover_process_iunlinks(
 	struct xlog	*log)
 {
-	xfs_mount_t	*mp;
-	xfs_agnumber_t	agno;
-	xfs_agi_t	*agi;
-	struct xfs_buf	*agibp;
-	xfs_agino_t	agino;
-	int		bucket;
-	int		error;
-
-	mp = log->l_mp;
+	struct xfs_mount	*mp = log->l_mp;
+	struct xfs_perag	*pag;
+	xfs_agnumber_t		agno;
+	struct xfs_agi		*agi;
+	struct xfs_buf		*agibp;
+	xfs_agino_t		agino;
+	int			bucket;
+	int			error;
 
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		/*
-		 * Find the agi for this ag.
-		 */
-		error = xfs_read_agi(mp, NULL, agno, &agibp);
+	for_each_perag(mp, agno, pag) {
+		error = xfs_read_agi(mp, NULL, pag->pag_agno, &agibp);
 		if (error) {
 			/*
 			 * AGI is b0rked. Don't process it.
@@ -2781,7 +2778,7 @@ xlog_recover_process_iunlinks(
 			agino = be32_to_cpu(agi->agi_unlinked[bucket]);
 			while (agino != NULLAGINO) {
 				agino = xlog_recover_process_one_iunlink(mp,
-							agno, agino, bucket);
+						pag->pag_agno, agino, bucket);
 				cond_resched();
 			}
 		}
@@ -3493,27 +3490,28 @@ xlog_recover_cancel(
  */
 STATIC void
 xlog_recover_check_summary(
-	struct xlog	*log)
+	struct xlog		*log)
 {
-	xfs_mount_t	*mp;
-	struct xfs_buf	*agfbp;
-	struct xfs_buf	*agibp;
-	xfs_agnumber_t	agno;
-	uint64_t	freeblks;
-	uint64_t	itotal;
-	uint64_t	ifree;
-	int		error;
+	struct xfs_mount	*mp = log->l_mp;
+	struct xfs_perag	*pag;
+	struct xfs_buf		*agfbp;
+	struct xfs_buf		*agibp;
+	xfs_agnumber_t		agno;
+	uint64_t		freeblks;
+	uint64_t		itotal;
+	uint64_t		ifree;
+	int			error;
 
 	mp = log->l_mp;
 
 	freeblks = 0LL;
 	itotal = 0LL;
 	ifree = 0LL;
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		error = xfs_read_agf(mp, NULL, agno, 0, &agfbp);
+	for_each_perag(mp, agno, pag) {
+		error = xfs_read_agf(mp, NULL, pag->pag_agno, 0, &agfbp);
 		if (error) {
 			xfs_alert(mp, "%s agf read failed agno %d error %d",
-						__func__, agno, error);
+						__func__, pag->pag_agno, error);
 		} else {
 			struct xfs_agf	*agfp = agfbp->b_addr;
 
@@ -3522,10 +3520,10 @@ xlog_recover_check_summary(
 			xfs_buf_relse(agfbp);
 		}
 
-		error = xfs_read_agi(mp, NULL, agno, &agibp);
+		error = xfs_read_agi(mp, NULL, pag->pag_agno, &agibp);
 		if (error) {
 			xfs_alert(mp, "%s agi read failed agno %d error %d",
-						__func__, agno, error);
+						__func__, pag->pag_agno, error);
 		} else {
 			struct xfs_agi	*agi = agibp->b_addr;
 
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index bdfee1943796..c3a96fb3ad80 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -32,6 +32,7 @@
 #include "xfs_extent_busy.h"
 #include "xfs_health.h"
 #include "xfs_trace.h"
+#include "xfs_ag.h"
 
 static DEFINE_MUTEX(xfs_uuid_table_mutex);
 static int xfs_uuid_table_size;
@@ -119,41 +120,6 @@ xfs_uuid_unmount(
 	mutex_unlock(&xfs_uuid_table_mutex);
 }
 
-
-STATIC void
-__xfs_free_perag(
-	struct rcu_head	*head)
-{
-	struct xfs_perag *pag = container_of(head, struct xfs_perag, rcu_head);
-
-	ASSERT(!delayed_work_pending(&pag->pag_blockgc_work));
-	ASSERT(atomic_read(&pag->pag_ref) == 0);
-	kmem_free(pag);
-}
-
-/*
- * Free up the per-ag resources associated with the mount structure.
- */
-STATIC void
-xfs_free_perag(
-	xfs_mount_t	*mp)
-{
-	xfs_agnumber_t	agno;
-	struct xfs_perag *pag;
-
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		spin_lock(&mp->m_perag_lock);
-		pag = radix_tree_delete(&mp->m_perag_tree, agno);
-		spin_unlock(&mp->m_perag_lock);
-		ASSERT(pag);
-		ASSERT(atomic_read(&pag->pag_ref) == 0);
-		cancel_delayed_work_sync(&pag->pag_blockgc_work);
-		xfs_iunlink_destroy(pag);
-		xfs_buf_hash_destroy(pag);
-		call_rcu(&pag->rcu_head, __xfs_free_perag);
-	}
-}
-
 /*
  * Check size of device based on the (data/realtime) block count.
  * Note: this check is used by the growfs code as well as mount.
@@ -172,96 +138,6 @@ xfs_sb_validate_fsb_count(
 	return 0;
 }
 
-int
-xfs_initialize_perag(
-	xfs_mount_t	*mp,
-	xfs_agnumber_t	agcount,
-	xfs_agnumber_t	*maxagi)
-{
-	xfs_agnumber_t	index;
-	xfs_agnumber_t	first_initialised = NULLAGNUMBER;
-	xfs_perag_t	*pag;
-	int		error = -ENOMEM;
-
-	/*
-	 * Walk the current per-ag tree so we don't try to initialise AGs
-	 * that already exist (growfs case). Allocate and insert all the
-	 * AGs we don't find ready for initialisation.
-	 */
-	for (index = 0; index < agcount; index++) {
-		pag = xfs_perag_get(mp, index);
-		if (pag) {
-			xfs_perag_put(pag);
-			continue;
-		}
-
-		pag = kmem_zalloc(sizeof(*pag), KM_MAYFAIL);
-		if (!pag) {
-			error = -ENOMEM;
-			goto out_unwind_new_pags;
-		}
-		pag->pag_agno = index;
-		pag->pag_mount = mp;
-		spin_lock_init(&pag->pag_ici_lock);
-		INIT_DELAYED_WORK(&pag->pag_blockgc_work, xfs_blockgc_worker);
-		INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC);
-
-		error = xfs_buf_hash_init(pag);
-		if (error)
-			goto out_free_pag;
-		init_waitqueue_head(&pag->pagb_wait);
-		spin_lock_init(&pag->pagb_lock);
-		pag->pagb_count = 0;
-		pag->pagb_tree = RB_ROOT;
-
-		error = radix_tree_preload(GFP_NOFS);
-		if (error)
-			goto out_hash_destroy;
-
-		spin_lock(&mp->m_perag_lock);
-		if (radix_tree_insert(&mp->m_perag_tree, index, pag)) {
-			WARN_ON_ONCE(1);
-			spin_unlock(&mp->m_perag_lock);
-			radix_tree_preload_end();
-			error = -EEXIST;
-			goto out_hash_destroy;
-		}
-		spin_unlock(&mp->m_perag_lock);
-		radix_tree_preload_end();
-		/* first new pag is fully initialized */
-		if (first_initialised == NULLAGNUMBER)
-			first_initialised = index;
-		error = xfs_iunlink_init(pag);
-		if (error)
-			goto out_hash_destroy;
-		spin_lock_init(&pag->pag_state_lock);
-	}
-
-	index = xfs_set_inode_alloc(mp, agcount);
-
-	if (maxagi)
-		*maxagi = index;
-
-	mp->m_ag_prealloc_blocks = xfs_prealloc_blocks(mp);
-	return 0;
-
-out_hash_destroy:
-	xfs_buf_hash_destroy(pag);
-out_free_pag:
-	kmem_free(pag);
-out_unwind_new_pags:
-	/* unwind any prior newly initialized pags */
-	for (index = first_initialised; index < agcount; index++) {
-		pag = radix_tree_delete(&mp->m_perag_tree, index);
-		if (!pag)
-			break;
-		xfs_buf_hash_destroy(pag);
-		xfs_iunlink_destroy(pag);
-		kmem_free(pag);
-	}
-	return error;
-}
-
 /*
  * xfs_readsb
  *
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index bb67274ee23f..c78b63fe779a 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -12,6 +12,7 @@ struct xfs_mru_cache;
 struct xfs_ail;
 struct xfs_quotainfo;
 struct xfs_da_geometry;
+struct xfs_perag;
 
 /* dynamic preallocation free space thresholds, 5% down to 1% */
 enum {
@@ -297,117 +298,12 @@ xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
 	return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
 }
 
-/* per-AG block reservation data structures*/
-struct xfs_ag_resv {
-	/* number of blocks originally reserved here */
-	xfs_extlen_t			ar_orig_reserved;
-	/* number of blocks reserved here */
-	xfs_extlen_t			ar_reserved;
-	/* number of blocks originally asked for */
-	xfs_extlen_t			ar_asked;
-};
-
-/*
- * Per-ag incore structure, copies of information in agf and agi, to improve the
- * performance of allocation group selection.
- */
-typedef struct xfs_perag {
-	struct xfs_mount *pag_mount;	/* owner filesystem */
-	xfs_agnumber_t	pag_agno;	/* AG this structure belongs to */
-	atomic_t	pag_ref;	/* perag reference count */
-	char		pagf_init;	/* this agf's entry is initialized */
-	char		pagi_init;	/* this agi's entry is initialized */
-	char		pagf_metadata;	/* the agf is preferred to be metadata */
-	char		pagi_inodeok;	/* The agi is ok for inodes */
-	uint8_t		pagf_levels[XFS_BTNUM_AGF];
-					/* # of levels in bno & cnt btree */
-	bool		pagf_agflreset; /* agfl requires reset before use */
-	uint32_t	pagf_flcount;	/* count of blocks in freelist */
-	xfs_extlen_t	pagf_freeblks;	/* total free blocks */
-	xfs_extlen_t	pagf_longest;	/* longest free space */
-	uint32_t	pagf_btreeblks;	/* # of blocks held in AGF btrees */
-	xfs_agino_t	pagi_freecount;	/* number of free inodes */
-	xfs_agino_t	pagi_count;	/* number of allocated inodes */
-
-	/*
-	 * Inode allocation search lookup optimisation.
-	 * If the pagino matches, the search for new inodes
-	 * doesn't need to search the near ones again straight away
-	 */
-	xfs_agino_t	pagl_pagino;
-	xfs_agino_t	pagl_leftrec;
-	xfs_agino_t	pagl_rightrec;
-
-	/*
-	 * Bitsets of per-ag metadata that have been checked and/or are sick.
-	 * Callers should hold pag_state_lock before accessing this field.
-	 */
-	uint16_t	pag_checked;
-	uint16_t	pag_sick;
-	spinlock_t	pag_state_lock;
-
-	spinlock_t	pagb_lock;	/* lock for pagb_tree */
-	struct rb_root	pagb_tree;	/* ordered tree of busy extents */
-	unsigned int	pagb_gen;	/* generation count for pagb_tree */
-	wait_queue_head_t pagb_wait;	/* woken when pagb_gen changes */
-
-	atomic_t        pagf_fstrms;    /* # of filestreams active in this AG */
-
-	spinlock_t	pag_ici_lock;	/* incore inode cache lock */
-	struct radix_tree_root pag_ici_root;	/* incore inode cache root */
-	int		pag_ici_reclaimable;	/* reclaimable inodes */
-	unsigned long	pag_ici_reclaim_cursor;	/* reclaim restart point */
-
-	/* buffer cache index */
-	spinlock_t	pag_buf_lock;	/* lock for pag_buf_hash */
-	struct rhashtable pag_buf_hash;
-
-	/* for rcu-safe freeing */
-	struct rcu_head	rcu_head;
-	int		pagb_count;	/* pagb slots in use */
-
-	/* Blocks reserved for all kinds of metadata. */
-	struct xfs_ag_resv	pag_meta_resv;
-	/* Blocks reserved for the reverse mapping btree. */
-	struct xfs_ag_resv	pag_rmapbt_resv;
-
-	/* background prealloc block trimming */
-	struct delayed_work	pag_blockgc_work;
-
-	/* reference count */
-	uint8_t			pagf_refcount_level;
-
-	/*
-	 * Unlinked inode information.  This incore information reflects
-	 * data stored in the AGI, so callers must hold the AGI buffer lock
-	 * or have some other means to control concurrency.
-	 */
-	struct rhashtable	pagi_unlinked_hash;
-} xfs_perag_t;
-
-static inline struct xfs_ag_resv *
-xfs_perag_resv(
-	struct xfs_perag	*pag,
-	enum xfs_ag_resv_type	type)
-{
-	switch (type) {
-	case XFS_AG_RESV_METADATA:
-		return &pag->pag_meta_resv;
-	case XFS_AG_RESV_RMAPBT:
-		return &pag->pag_rmapbt_resv;
-	default:
-		return NULL;
-	}
-}
-
-int xfs_buf_hash_init(xfs_perag_t *pag);
-void xfs_buf_hash_destroy(xfs_perag_t *pag);
+int xfs_buf_hash_init(struct xfs_perag *pag);
+void xfs_buf_hash_destroy(struct xfs_perag *pag);
 
 extern void	xfs_uuid_table_free(void);
 extern uint64_t xfs_default_resblks(xfs_mount_t *mp);
 extern int	xfs_mountfs(xfs_mount_t *mp);
-extern int	xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
-				     xfs_agnumber_t *maxagi);
 extern void	xfs_unmountfs(xfs_mount_t *);
 
 extern int	xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
diff --git a/fs/xfs/xfs_qm.c b/fs/xfs/xfs_qm.c
index 4bf949a89d0d..fe341f3fd419 100644
--- a/fs/xfs/xfs_qm.c
+++ b/fs/xfs/xfs_qm.c
@@ -23,6 +23,8 @@
 #include "xfs_trace.h"
 #include "xfs_icache.h"
 #include "xfs_error.h"
+#include "xfs_ag.h"
+#include "xfs_ialloc.h"
 
 /*
  * The global quota manager. There is only one of these for the entire
@@ -787,8 +789,12 @@ xfs_qm_qino_alloc(
 		return error;
 
 	if (need_alloc) {
-		error = xfs_dir_ialloc(&init_user_ns, &tp, NULL, S_IFREG, 1, 0,
-				       0, false, ipp);
+		xfs_ino_t	ino;
+
+		error = xfs_dialloc(&tp, 0, S_IFREG, &ino);
+		if (!error)
+			error = xfs_init_new_inode(&init_user_ns, tp, NULL, ino,
+					S_IFREG, 1, 0, 0, false, ipp);
 		if (error) {
 			xfs_trans_cancel(tp);
 			return error;
diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c
index 060695d6d56a..c256104772cb 100644
--- a/fs/xfs/xfs_reflink.c
+++ b/fs/xfs/xfs_reflink.c
@@ -27,7 +27,7 @@
 #include "xfs_quota.h"
 #include "xfs_reflink.h"
 #include "xfs_iomap.h"
-#include "xfs_sb.h"
+#include "xfs_ag.h"
 #include "xfs_ag_resv.h"
 
 /*
@@ -144,7 +144,7 @@ xfs_reflink_find_shared(
 	if (error)
 		return error;
 
-	cur = xfs_refcountbt_init_cursor(mp, tp, agbp, agno);
+	cur = xfs_refcountbt_init_cursor(mp, tp, agbp, agbp->b_pag);
 
 	error = xfs_refcount_find_shared(cur, agbno, aglen, fbno, flen,
 			find_end_of_shared);
@@ -755,16 +755,19 @@ int
 xfs_reflink_recover_cow(
 	struct xfs_mount	*mp)
 {
+	struct xfs_perag	*pag;
 	xfs_agnumber_t		agno;
 	int			error = 0;
 
 	if (!xfs_sb_version_hasreflink(&mp->m_sb))
 		return 0;
 
-	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
-		error = xfs_refcount_recover_cow_leftovers(mp, agno);
-		if (error)
+	for_each_perag(mp, agno, pag) {
+		error = xfs_refcount_recover_cow_leftovers(mp, pag);
+		if (error) {
+			xfs_perag_put(pag);
 			break;
+		}
 	}
 
 	return error;
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index a2dab05332ac..688309dbe18b 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -36,6 +36,7 @@
 #include "xfs_bmap_item.h"
 #include "xfs_reflink.h"
 #include "xfs_pwork.h"
+#include "xfs_ag.h"
 
 #include <linux/magic.h>
 #include <linux/fs_context.h>
diff --git a/fs/xfs/xfs_symlink.c b/fs/xfs/xfs_symlink.c
index 99fbec32c10a..1525636f4065 100644
--- a/fs/xfs/xfs_symlink.c
+++ b/fs/xfs/xfs_symlink.c
@@ -21,6 +21,7 @@
 #include "xfs_trans_space.h"
 #include "xfs_trace.h"
 #include "xfs_trans.h"
+#include "xfs_ialloc.h"
 
 /* ----- Kernel only functions below ----- */
 int
@@ -161,6 +162,7 @@ xfs_symlink(
 	struct xfs_dquot	*gdqp = NULL;
 	struct xfs_dquot	*pdqp = NULL;
 	uint			resblks;
+	xfs_ino_t		ino;
 
 	*ipp = NULL;
 
@@ -223,8 +225,11 @@ xfs_symlink(
 	/*
 	 * Allocate an inode for the symlink.
 	 */
-	error = xfs_dir_ialloc(mnt_userns, &tp, dp, S_IFLNK | (mode & ~S_IFMT),
-			       1, 0, prid, false, &ip);
+	error = xfs_dialloc(&tp, dp->i_ino, S_IFLNK, &ino);
+	if (!error)
+		error = xfs_init_new_inode(mnt_userns, tp, dp, ino,
+				S_IFLNK | (mode & ~S_IFMT), 1, 0, prid,
+				false, &ip);
 	if (error)
 		goto out_trans_cancel;
 
diff --git a/fs/xfs/xfs_trace.c b/fs/xfs/xfs_trace.c
index 9b8d703dc9fd..7e01e00550ac 100644
--- a/fs/xfs/xfs_trace.c
+++ b/fs/xfs/xfs_trace.c
@@ -30,6 +30,8 @@
 #include "xfs_fsmap.h"
 #include "xfs_btree_staging.h"
 #include "xfs_icache.h"
+#include "xfs_ag.h"
+#include "xfs_ag_resv.h"
 
 /*
  * We include this last to have the helpers above available for the trace
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 808ae337b222..5ba9c6396dcb 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3730,7 +3730,7 @@ TRACE_EVENT(xfs_btree_commit_afakeroot,
 	TP_fast_assign(
 		__entry->dev = cur->bc_mp->m_super->s_dev;
 		__entry->btnum = cur->bc_btnum;
-		__entry->agno = cur->bc_ag.agno;
+		__entry->agno = cur->bc_ag.pag->pag_agno;
 		__entry->agbno = cur->bc_ag.afake->af_root;
 		__entry->levels = cur->bc_ag.afake->af_levels;
 		__entry->blocks = cur->bc_ag.afake->af_blocks;
@@ -3845,7 +3845,7 @@ TRACE_EVENT(xfs_btree_bload_block,
 			__entry->agno = XFS_FSB_TO_AGNO(cur->bc_mp, fsb);
 			__entry->agbno = XFS_FSB_TO_AGBNO(cur->bc_mp, fsb);
 		} else {
-			__entry->agno = cur->bc_ag.agno;
+			__entry->agno = cur->bc_ag.pag->pag_agno;
 			__entry->agbno = be32_to_cpu(ptr->s);
 		}
 		__entry->nr_records = nr_records;