Merge tag 'xfs-perag-conv-tag' of...

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 ...

Merge tag 'xfs-perag-conv-tag' of...
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 ...
c3eabd36 · Darrick J. Wong · ebf2e337 · 50920116 · c3eabd36 · c3eabd36
Commit c3eabd36 authored Jun 08, 2021 by Darrick J. Wong
60 changed files
--- 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)

--- 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 */

--- 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;


--- 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

--- 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 @@ xfs_alloc_ag_vextent_size(
 	 * 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 @@ xfs_alloc_ag_vextent_size(
 	 * 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 @@ xfs_alloc_fix_freelist(
 * 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 @@ xfs_alloc_vextent(
 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;
 }


--- 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);

--- 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;
 }

--- 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,

--- 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"


 /*

--- 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"

--- 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,

--- 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;

--- a/fs/xfs/libxfs/xfs_ialloc.c
+++ b/fs/xfs/libxfs/xfs_ialloc.c
--- 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.

--- 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;
 }

--- 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,

--- a/fs/xfs/libxfs/xfs_refcount.c
+++ b/fs/xfs/libxfs/xfs_refcount.c
--- 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

--- 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);

--- 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,

--- a/fs/xfs/libxfs/xfs_rmap.c
+++ b/fs/xfs/libxfs/xfs_rmap.c
--- 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,

--- 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. */

--- 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__ */
--- 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(
@@ -841,78 +780,6 @@ xfs_sb_mount_common(
 	mp->m_ag_max_usable = xfs_alloc_ag_max_usable(mp);
 }

-/*
- * 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
@@ -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;

--- 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);

--- 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);

--- 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"


--- 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;

--- 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;


--- 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;
 }

 /*

--- 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);
 	}
 }


--- 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 @@ xchk_fscount_aggregate_agcounts(
 	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 @@ xchk_fscount_aggregate_agcounts(
 			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 @@ xchk_fscount_aggregate_agcounts(
 		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;


--- 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"

--- 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;

--- 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;

--- 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);


--- 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;

--- 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;
 }


--- 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;


--- 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;
 		}

--- 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 @@ xfs_extent_busy_update_extent(
 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 @@ xfs_extent_busy_reuse(
 			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);
 	}
 }


--- 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

--- 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"

--- 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 @@ __xfs_getfsmap_datadev(
 		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;

--- 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;
 	}

--- 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. */

--- 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(

--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
--- a/fs/xfs/xfs_iwalk.c
+++ b/fs/xfs/xfs_iwalk.c
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
--- a/fs/xfs/xfs_qm.c
+++ b/fs/xfs/xfs_qm.c
--- a/fs/xfs/xfs_reflink.c
+++ b/fs/xfs/xfs_reflink.c
--- 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>

--- a/fs/xfs/xfs_symlink.c
+++ b/fs/xfs/xfs_symlink.c
--- a/fs/xfs/xfs_trace.c
+++ b/fs/xfs/xfs_trace.c
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h