Merge tag 'xfs-alloc-perag-conversion' of...

Merge tag 'xfs-alloc-perag-conversion' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs into xfs-6.3-merge-A

xfs: per-ag centric allocation alogrithms

This series continues the work towards making shrinking a filesystem
possible.  We need to be able to stop operations from taking place
on AGs that need to be removed by a shrink, so before shrink can be
implemented we need to have the infrastructure in place to prevent
incursion into AGs that are going to be, or are in the process, of
being removed from active duty.

The focus of this is making operations that depend on access to AGs
use the perag to access and pin the AG in active use, thereby
creating a barrier we can use to delay shrink until all active uses
of an AG have been drained and new uses are prevented.

This series starts by fixing some existing issues that are exposed
by changes later in the series. They stand alone, so can be picked
up independently of the rest of this patchset.

The most complex of these fixes is cleaning up the mess that is the
AGF deadlock avoidance algorithm. This algorithm stores the first
block that is allocated in a transaction in tp->t_firstblock, then
uses this to try to limit future allocations within the transaction
to AGs at or higher than the filesystem block stored in
tp->t_firstblock. This depends on one of the initial bug fixes in
the series to move the deadlock avoidance checks to
xfs_alloc_vextent(), and then builds on it to relax the constraints
of the avoidance algorithm to only be active when a deadlock is
possible.

We also update the algorithm to record allocations from higher AGs
that are allocated from, because we when we need to lock more than
two AGs we still have to ensure lock order is correct. Therefore we
can't lock AGs in the order 1, 3, 2, even though tp->t_firstblock
indicates that we've allocated from AG 1 and so AG is valid to lock.
It's not valid, because we already hold AG 3 locked, and so
tp->t-first_block should actually point at AG 3, not AG 1 in this
situation.

It should now be obvious that the deadlock avoidance algorithm
should record AGs, not filesystem blocks. So the series then changes
the transaction to store the highest AG we've allocated in rather
than a filesystem block we allocated.  This makes it obvious what
the constraints are, and trivial to update as we lock and allocate
from various AGs.

With all the bug fixes out of the way, the series then starts
converting the code to use active references. Active reference
counts are used by high level code that needs to prevent the AG from
being taken out from under it by a shrink operation. The high level
code needs to be able to handle not getting an active reference
gracefully, and the shrink code will need to wait for active
references to drain before continuing.

Active references are implemented just as reference counts right now
- an active reference is taken at perag init during mount, and all
other active references are dependent on the active reference count
being greater than zero. This gives us an initial method of stopping
new active references without needing other infrastructure; just
drop the reference taken at filesystem mount time and when the
refcount then falls to zero no new references can be taken.

In future, this will need to take into account AG control state
(e.g. offline, no alloc, etc) as well as the reference count, but
right now we can implement a basic barrier for shrink with just
reference count manipulations. As such, patches to convert the perag
state to atomic opstate fields similar to the xfs_mount and xlog
opstate fields follow the initial active perag reference counting
patches.

The first target for active reference conversion is the
for_each_perag*() iterators. This captures a lot of high level code
that should skip offline AGs, and introduces the ability to
differentiate between a lookup that didn't have an online AG and the
end of the AG iteration range.

From there, the inode allocation AG selection is converted to active
references, and the perag is driven deeper into the inode allocation
and btree code to replace the xfs_mount. Most of the inode
allocation code operates on a single AG once it is selected, hence
it should pass the perag as the primary referenced object around for
allocation, not the xfs_mount. There is a bit of churn here, but it
emphasises that inode allocation is inherently an allocation group
based operation.

Next the bmap/alloc interface undergoes a major untangling,
reworking xfs_bmap_btalloc() into separate allocation operations for
different contexts and failure handling behaviours. This then allows
us to completely remove the xfs_alloc_vextent() layer via
restructuring the xfs_alloc_vextent/xfs_alloc_ag_vextent() into a
set of realtively simple helper function that describe the
allocation that they are doing. e.g.  xfs_alloc_vextent_exact_bno().

This allows the requirements for accessing AGs to be allocation
context dependent. The allocations that require operation on a
single AG generally can't tolerate failure after the allocation
method and AG has been decided on, and hence the caller needs to
manage the active references to ensure the allocation does not race
with shrink removing the selected AG for the duration of the
operation that requires access to that allocation group.

Other allocations iterate AGs and so the first AG is just a hint -
these do not need to pin a perag first as they can tolerate not
being able to access an AG by simply skipping over it. These require
new perag iteration functions that can start at arbitrary AGs and
wrap around at arbitrary AGs, hence a new set for
for_each_perag_wrap*() helpers to do this.

Next is the rework of the filestreams allocator. This doesn't change
any functionality, but gets rid of the unnecessary multi-pass
selection algorithm when the selected AG is not available. It
currently does a lookup pass which might iterate all AGs to select
an AG, then checks if the AG is acceptible and if not does a "new
AG" pass that is essentially identical to the lookup pass. Both of
these scans also do the same "longest extent in AG" check before
selecting an AG as is done after the AG is selected.

IOWs, the filestreams algorithm can be greatly simplified into a
single new AG selection pass if the there is no current association
or the currently associated AG doesn't have enough contiguous free
space for the allocation to proceed.  With this simplification of
the filestreams allocator, it's then trivial to convert it to use
for_each_perag_wrap() for the AG scan algorithm.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>

* tag 'xfs-alloc-perag-conversion' of git://git.kernel.org/pub/scm/linux/kernel/git/dgc/linux-xfs: (42 commits)
  xfs: refactor the filestreams allocator pick functions
  xfs: return a referenced perag from filestreams allocator
  xfs: pass perag to filestreams tracing
  xfs: use for_each_perag_wrap in xfs_filestream_pick_ag
  xfs: track an active perag reference in filestreams
  xfs: factor out MRU hit case in xfs_filestream_select_ag
  xfs: remove xfs_filestream_select_ag() longest extent check
  xfs: merge new filestream AG selection into xfs_filestream_select_ag()
  xfs: merge filestream AG lookup into xfs_filestream_select_ag()
  xfs: move xfs_bmap_btalloc_filestreams() to xfs_filestreams.c
  xfs: use xfs_bmap_longest_free_extent() in filestreams
  xfs: get rid of notinit from xfs_bmap_longest_free_extent
  xfs: factor out filestreams from xfs_bmap_btalloc_nullfb
  xfs: convert trim to use for_each_perag_range
  xfs: convert xfs_alloc_vextent_iterate_ags() to use perag walker
  xfs: move the minimum agno checks into xfs_alloc_vextent_check_args
  xfs: fold xfs_alloc_ag_vextent() into callers
  xfs: move allocation accounting to xfs_alloc_vextent_set_fsbno()
  xfs: introduce xfs_alloc_vextent_prepare()
  xfs: introduce xfs_alloc_vextent_exact_bno()
  ...

fs/xfs/libxfs/xfs_ag.c

@@ -44,16 +44,15 @@ xfs_perag_get(
 	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) {
+		trace_xfs_perag_get(pag, _RET_IP_);
 		ASSERT(atomic_read(&pag->pag_ref) >= 0);
-		ref = atomic_inc_return(&pag->pag_ref);
+		atomic_inc(&pag->pag_ref);
 	}
 	rcu_read_unlock();
-	trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
 	return pag;
 }
 
@@ -68,7 +67,6 @@ xfs_perag_get_tag(
 {
 	struct xfs_perag	*pag;
 	int			found;
-	int			ref;
 
 	rcu_read_lock();
 	found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
@@ -77,9 +75,9 @@ xfs_perag_get_tag(
 		rcu_read_unlock();
 		return NULL;
 	}
-	ref = atomic_inc_return(&pag->pag_ref);
+	trace_xfs_perag_get_tag(pag, _RET_IP_);
+	atomic_inc(&pag->pag_ref);
 	rcu_read_unlock();
-	trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
 	return pag;
 }
 
@@ -87,11 +85,68 @@ void
 xfs_perag_put(
 	struct xfs_perag	*pag)
 {
-	int	ref;
-
+	trace_xfs_perag_put(pag, _RET_IP_);
 	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_);
+	atomic_dec(&pag->pag_ref);
+}
+
+/*
+ * Active references for perag structures. This is for short term access to the
+ * per ag structures for walking trees or accessing state. If an AG is being
+ * shrunk or is offline, then this will fail to find that AG and return NULL
+ * instead.
+ */
+struct xfs_perag *
+xfs_perag_grab(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno)
+{
+	struct xfs_perag	*pag;
+
+	rcu_read_lock();
+	pag = radix_tree_lookup(&mp->m_perag_tree, agno);
+	if (pag) {
+		trace_xfs_perag_grab(pag, _RET_IP_);
+		if (!atomic_inc_not_zero(&pag->pag_active_ref))
+			pag = NULL;
+	}
+	rcu_read_unlock();
+	return pag;
+}
+
+/*
+ * search from @first to find the next perag with the given tag set.
+ */
+struct xfs_perag *
+xfs_perag_grab_tag(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		first,
+	int			tag)
+{
+	struct xfs_perag	*pag;
+	int			found;
+
+	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;
+	}
+	trace_xfs_perag_grab_tag(pag, _RET_IP_);
+	if (!atomic_inc_not_zero(&pag->pag_active_ref))
+		pag = NULL;
+	rcu_read_unlock();
+	return pag;
+}
+
+void
+xfs_perag_rele(
+	struct xfs_perag	*pag)
+{
+	trace_xfs_perag_rele(pag, _RET_IP_);
+	if (atomic_dec_and_test(&pag->pag_active_ref))
+		wake_up(&pag->pag_active_wq);
 }
 
 /*
@@ -196,6 +251,10 @@ xfs_free_perag(
 		cancel_delayed_work_sync(&pag->pag_blockgc_work);
 		xfs_buf_hash_destroy(pag);
 
+		/* drop the mount's active reference */
+		xfs_perag_rele(pag);
+		XFS_IS_CORRUPT(pag->pag_mount,
+				atomic_read(&pag->pag_active_ref) != 0);
 		call_rcu(&pag->rcu_head, __xfs_free_perag);
 	}
 }
@@ -314,6 +373,7 @@ xfs_initialize_perag(
 		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);
+		init_waitqueue_head(&pag->pag_active_wq);
 		pag->pagb_count = 0;
 		pag->pagb_tree = RB_ROOT;
 #endif /* __KERNEL__ */
@@ -322,6 +382,9 @@ xfs_initialize_perag(
 		if (error)
 			goto out_remove_pag;
 
+		/* Active ref owned by mount indicates AG is online. */
+		atomic_set(&pag->pag_active_ref, 1);
+
 		/* first new pag is fully initialized */
 		if (first_initialised == NULLAGNUMBER)
 			first_initialised = index;
@@ -824,7 +887,7 @@ xfs_ag_shrink_space(
 	struct xfs_alloc_arg	args = {
 		.tp	= *tpp,
 		.mp	= mp,
-		.type	= XFS_ALLOCTYPE_THIS_BNO,
+		.pag	= pag,
 		.minlen = delta,
 		.maxlen = delta,
 		.oinfo	= XFS_RMAP_OINFO_SKIP_UPDATE,
@@ -856,14 +919,11 @@ xfs_ag_shrink_space(
 	if (delta >= aglen)
 		return -EINVAL;
 
-	args.fsbno = XFS_AGB_TO_FSB(mp, pag->pag_agno, aglen - delta);
-
 	/*
 	 * Make sure that the last inode cluster cannot overlap with the new
 	 * end of the AG, even if it's sparse.
 	 */
-	error = xfs_ialloc_check_shrink(*tpp, pag->pag_agno, agibp,
-			aglen - delta);
+	error = xfs_ialloc_check_shrink(pag, *tpp, agibp, aglen - delta);
 	if (error)
 		return error;
 
@@ -876,7 +936,8 @@ xfs_ag_shrink_space(
 		return error;
 
 	/* internal log shouldn't also show up in the free space btrees */
-	error = xfs_alloc_vextent(&args);
+	error = xfs_alloc_vextent_exact_bno(&args,
+			XFS_AGB_TO_FSB(mp, pag->pag_agno, aglen - delta));
 	if (!error && args.agbno == NULLAGBLOCK)
 		error = -ENOSPC;
 

fs/xfs/libxfs/xfs_ag.h

@@ -32,14 +32,12 @@ struct xfs_ag_resv {
 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 */
+	atomic_t	pag_ref;	/* passive reference count */
+	atomic_t	pag_active_ref;	/* active reference count */
+	wait_queue_head_t pag_active_wq;/* woken active_ref falls to zero */
+	unsigned long	pag_opstate;
 	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 */
@@ -106,16 +104,44 @@ struct xfs_perag {
 #endif /* __KERNEL__ */
 };
 
+/*
+ * Per-AG operational state. These are atomic flag bits.
+ */
+#define XFS_AGSTATE_AGF_INIT		0
+#define XFS_AGSTATE_AGI_INIT		1
+#define XFS_AGSTATE_PREFERS_METADATA	2
+#define XFS_AGSTATE_ALLOWS_INODES	3
+#define XFS_AGSTATE_AGFL_NEEDS_RESET	4
+
+#define __XFS_AG_OPSTATE(name, NAME) \
+static inline bool xfs_perag_ ## name (struct xfs_perag *pag) \
+{ \
+	return test_bit(XFS_AGSTATE_ ## NAME, &pag->pag_opstate); \
+}
+
+__XFS_AG_OPSTATE(initialised_agf, AGF_INIT)
+__XFS_AG_OPSTATE(initialised_agi, AGI_INIT)
+__XFS_AG_OPSTATE(prefers_metadata, PREFERS_METADATA)
+__XFS_AG_OPSTATE(allows_inodes, ALLOWS_INODES)
+__XFS_AG_OPSTATE(agfl_needs_reset, AGFL_NEEDS_RESET)
+
 int xfs_initialize_perag(struct xfs_mount *mp, xfs_agnumber_t agcount,
 			xfs_rfsblock_t dcount, xfs_agnumber_t *maxagi);
 int xfs_initialize_perag_data(struct xfs_mount *mp, xfs_agnumber_t agno);
 void xfs_free_perag(struct xfs_mount *mp);
 
+/* Passive AG references */
 struct xfs_perag *xfs_perag_get(struct xfs_mount *mp, xfs_agnumber_t agno);
 struct xfs_perag *xfs_perag_get_tag(struct xfs_mount *mp, xfs_agnumber_t agno,
 		unsigned int tag);
 void xfs_perag_put(struct xfs_perag *pag);
 
+/* Active AG references */
+struct xfs_perag *xfs_perag_grab(struct xfs_mount *, xfs_agnumber_t);
+struct xfs_perag *xfs_perag_grab_tag(struct xfs_mount *, xfs_agnumber_t,
+				   int tag);
+void xfs_perag_rele(struct xfs_perag *pag);
+
 /*
  * Per-ag geometry infomation and validation
  */
@@ -193,31 +219,86 @@ xfs_perag_next(
 	struct xfs_mount	*mp = pag->pag_mount;
 
 	*agno = pag->pag_agno + 1;
-	xfs_perag_put(pag);
-	if (*agno > end_agno)
-		return NULL;
-	return xfs_perag_get(mp, *agno);
+	xfs_perag_rele(pag);
+	while (*agno <= end_agno) {
+		pag = xfs_perag_grab(mp, *agno);
+		if (pag)
+			return pag;
+		(*agno)++;
+	}
+	return NULL;
 }
 
 #define for_each_perag_range(mp, agno, end_agno, pag) \
-	for ((pag) = xfs_perag_get((mp), (agno)); \
+	for ((pag) = xfs_perag_grab((mp), (agno)); \
 		(pag) != NULL; \
 		(pag) = xfs_perag_next((pag), &(agno), (end_agno)))
 
 #define for_each_perag_from(mp, agno, pag) \
 	for_each_perag_range((mp), (agno), (mp)->m_sb.sb_agcount - 1, (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)); \
+	for ((agno) = 0, (pag) = xfs_perag_grab_tag((mp), 0, (tag)); \
 		(pag) != NULL; \
 		(agno) = (pag)->pag_agno + 1, \
-		xfs_perag_put(pag), \
-		(pag) = xfs_perag_get_tag((mp), (agno), (tag)))
+		xfs_perag_rele(pag), \
+		(pag) = xfs_perag_grab_tag((mp), (agno), (tag)))
+
+static inline struct xfs_perag *
+xfs_perag_next_wrap(
+	struct xfs_perag	*pag,
+	xfs_agnumber_t		*agno,
+	xfs_agnumber_t		stop_agno,
+	xfs_agnumber_t		restart_agno,
+	xfs_agnumber_t		wrap_agno)
+{
+	struct xfs_mount	*mp = pag->pag_mount;
+
+	*agno = pag->pag_agno + 1;
+	xfs_perag_rele(pag);
+	while (*agno != stop_agno) {
+		if (*agno >= wrap_agno) {
+			if (restart_agno >= stop_agno)
+				break;
+			*agno = restart_agno;
+		}
+
+		pag = xfs_perag_grab(mp, *agno);
+		if (pag)
+			return pag;
+		(*agno)++;
+	}
+	return NULL;
+}
+
+/*
+ * Iterate all AGs from start_agno through wrap_agno, then restart_agno through
+ * (start_agno - 1).
+ */
+#define for_each_perag_wrap_range(mp, start_agno, restart_agno, wrap_agno, agno, pag) \
+	for ((agno) = (start_agno), (pag) = xfs_perag_grab((mp), (agno)); \
+		(pag) != NULL; \
+		(pag) = xfs_perag_next_wrap((pag), &(agno), (start_agno), \
+				(restart_agno), (wrap_agno)))
+/*
+ * Iterate all AGs from start_agno through wrap_agno, then 0 through
+ * (start_agno - 1).
+ */
+#define for_each_perag_wrap_at(mp, start_agno, wrap_agno, agno, pag) \
+	for_each_perag_wrap_range((mp), (start_agno), 0, (wrap_agno), (agno), (pag))
+
+/*
+ * Iterate all AGs from start_agno through to the end of the filesystem, then 0
+ * through (start_agno - 1).
+ */
+#define for_each_perag_wrap(mp, start_agno, agno, pag) \
+	for_each_perag_wrap_at((mp), (start_agno), (mp)->m_sb.sb_agcount, \
+				(agno), (pag))
+
 
 struct aghdr_init_data {
 	/* per ag data */

fs/xfs/libxfs/xfs_ag_resv.c

@@ -264,7 +264,7 @@ xfs_ag_resv_init(
 		if (error)
 			goto out;
 
-		error = xfs_finobt_calc_reserves(mp, tp, pag, &ask, &used);
+		error = xfs_finobt_calc_reserves(pag, tp, &ask, &used);
 		if (error)
 			goto out;
 

fs/xfs/libxfs/xfs_alloc.h

@@ -16,25 +16,6 @@ extern struct workqueue_struct *xfs_alloc_wq;
 
 unsigned int xfs_agfl_size(struct xfs_mount *mp);
 
-/*
- * Freespace allocation types.  Argument to xfs_alloc_[v]extent.
- */
-#define XFS_ALLOCTYPE_FIRST_AG	0x02	/* ... start at ag 0 */
-#define XFS_ALLOCTYPE_THIS_AG	0x08	/* anywhere in this a.g. */
-#define XFS_ALLOCTYPE_START_BNO	0x10	/* near this block else anywhere */
-#define XFS_ALLOCTYPE_NEAR_BNO	0x20	/* in this a.g. and near this block */
-#define XFS_ALLOCTYPE_THIS_BNO	0x40	/* at exactly this block */
-
-/* this should become an enum again when the tracing code is fixed */
-typedef unsigned int xfs_alloctype_t;
-
-#define XFS_ALLOC_TYPES \
-	{ XFS_ALLOCTYPE_FIRST_AG,	"FIRST_AG" }, \
-	{ XFS_ALLOCTYPE_THIS_AG,	"THIS_AG" }, \
-	{ XFS_ALLOCTYPE_START_BNO,	"START_BNO" }, \
-	{ XFS_ALLOCTYPE_NEAR_BNO,	"NEAR_BNO" }, \
-	{ XFS_ALLOCTYPE_THIS_BNO,	"THIS_BNO" }
-
 /*
  * Flags for xfs_alloc_fix_freelist.
  */
@@ -68,8 +49,6 @@ typedef struct xfs_alloc_arg {
 	xfs_agblock_t	min_agbno;	/* set an agbno range for NEAR allocs */
 	xfs_agblock_t	max_agbno;	/* ... */
 	xfs_extlen_t	len;		/* output: actual size of extent */
-	xfs_alloctype_t	type;		/* allocation type XFS_ALLOCTYPE_... */
-	xfs_alloctype_t	otype;		/* original allocation type */
 	int		datatype;	/* mask defining data type treatment */
 	char		wasdel;		/* set if allocation was prev delayed */
 	char		wasfromfl;	/* set if allocation is from freelist */
@@ -118,11 +97,43 @@ xfs_alloc_log_agf(
 	uint32_t	fields);/* mask of fields to be logged (XFS_AGF_...) */
 
 /*
- * Allocate an extent (variable-size).
+ * Allocate an extent anywhere in the specific AG given. If there is no
+ * space matching the requirements in that AG, then the allocation will fail.
  */
-int				/* error */
-xfs_alloc_vextent(
-	xfs_alloc_arg_t	*args);	/* allocation argument structure */
+int xfs_alloc_vextent_this_ag(struct xfs_alloc_arg *args, xfs_agnumber_t agno);
+
+/*
+ * Allocate an extent as close to the target as possible. If there are not
+ * viable candidates in the AG, then fail the allocation.
+ */
+int xfs_alloc_vextent_near_bno(struct xfs_alloc_arg *args,
+		xfs_fsblock_t target);
+
+/*
+ * Allocate an extent exactly at the target given. If this is not possible
+ * then the allocation fails.
+ */
+int xfs_alloc_vextent_exact_bno(struct xfs_alloc_arg *args,
+		xfs_fsblock_t target);
+
+/*
+ * Best effort full filesystem allocation scan.
+ *
+ * Locality aware allocation will be attempted in the initial AG, but on failure
+ * non-localised attempts will be made. The AGs are constrained by previous
+ * allocations in the current transaction. Two passes will be made - the first
+ * non-blocking, the second blocking.
+ */
+int xfs_alloc_vextent_start_ag(struct xfs_alloc_arg *args,
+		xfs_fsblock_t target);
+
+/*
+ * Iterate from the AG indicated from args->fsbno through to the end of the
+ * filesystem attempting blocking allocation. This is for use in last
+ * resort allocation attempts when everything else has failed.
+ */
+int xfs_alloc_vextent_first_ag(struct xfs_alloc_arg *args,
+		xfs_fsblock_t target);
 
 /*
  * Free an extent.

fs/xfs/libxfs/xfs_alloc_btree.c

@@ -315,7 +315,7 @@ xfs_allocbt_verify(
 	level = be16_to_cpu(block->bb_level);
 	if (bp->b_ops->magic[0] == cpu_to_be32(XFS_ABTC_MAGIC))
 		btnum = XFS_BTNUM_CNTi;
-	if (pag && pag->pagf_init) {
+	if (pag && xfs_perag_initialised_agf(pag)) {
 		if (level >= pag->pagf_levels[btnum])
 			return __this_address;
 	} else if (level >= mp->m_alloc_maxlevels)

fs/xfs/libxfs/xfs_bmap.h

@@ -12,6 +12,7 @@ struct xfs_ifork;
 struct xfs_inode;
 struct xfs_mount;
 struct xfs_trans;
+struct xfs_alloc_arg;
 
 /*
  * Argument structure for xfs_bmap_alloc.
@@ -168,6 +169,8 @@ static inline bool xfs_bmap_is_written_extent(struct xfs_bmbt_irec *irec)
 #define xfs_valid_startblock(ip, startblock) \
 	((startblock) != 0 || XFS_IS_REALTIME_INODE(ip))
 
+int	xfs_bmap_longest_free_extent(struct xfs_perag *pag,
+		struct xfs_trans *tp, xfs_extlen_t *blen);
 void	xfs_trim_extent(struct xfs_bmbt_irec *irec, xfs_fileoff_t bno,
 		xfs_filblks_t len);
 unsigned int xfs_bmap_compute_attr_offset(struct xfs_mount *mp);
@@ -220,6 +223,10 @@ int	xfs_bmap_add_extent_unwritten_real(struct xfs_trans *tp,
 		struct xfs_inode *ip, int whichfork,
 		struct xfs_iext_cursor *icur, struct xfs_btree_cur **curp,
 		struct xfs_bmbt_irec *new, int *logflagsp);
+xfs_extlen_t xfs_bmapi_minleft(struct xfs_trans *tp, struct xfs_inode *ip,
+		int fork);
+int	xfs_bmap_btalloc_low_space(struct xfs_bmalloca *ap,
+		struct xfs_alloc_arg *args);
 
 enum xfs_bmap_intent_type {
 	XFS_BMAP_MAP = 1,

fs/xfs/libxfs/xfs_bmap_btree.c

@@ -21,6 +21,7 @@
 #include "xfs_quota.h"
 #include "xfs_trace.h"
 #include "xfs_rmap.h"
+#include "xfs_ag.h"
 
 static struct kmem_cache	*xfs_bmbt_cur_cache;
 
@@ -184,11 +185,11 @@ xfs_bmbt_update_cursor(
 	struct xfs_btree_cur	*src,
 	struct xfs_btree_cur	*dst)
 {
-	ASSERT((dst->bc_tp->t_firstblock != NULLFSBLOCK) ||
+	ASSERT((dst->bc_tp->t_highest_agno != NULLAGNUMBER) ||
 	       (dst->bc_ino.ip->i_diflags & XFS_DIFLAG_REALTIME));
 
 	dst->bc_ino.allocated += src->bc_ino.allocated;
-	dst->bc_tp->t_firstblock = src->bc_tp->t_firstblock;
+	dst->bc_tp->t_highest_agno = src->bc_tp->t_highest_agno;
 
 	src->bc_ino.allocated = 0;
 }
@@ -200,46 +201,32 @@ xfs_bmbt_alloc_block(
 	union xfs_btree_ptr		*new,
 	int				*stat)
 {
-	xfs_alloc_arg_t		args;		/* block allocation args */
-	int			error;		/* error return value */
+	struct xfs_alloc_arg	args;
+	int			error;
 
 	memset(&args, 0, sizeof(args));
 	args.tp = cur->bc_tp;
 	args.mp = cur->bc_mp;
-	args.fsbno = cur->bc_tp->t_firstblock;
 	xfs_rmap_ino_bmbt_owner(&args.oinfo, cur->bc_ino.ip->i_ino,
 			cur->bc_ino.whichfork);
-
-	if (args.fsbno == NULLFSBLOCK) {
-		args.fsbno = be64_to_cpu(start->l);
-		args.type = XFS_ALLOCTYPE_START_BNO;
-		/*
-		 * Make sure there is sufficient room left in the AG to
-		 * complete a full tree split for an extent insert.  If
-		 * we are converting the middle part of an extent then
-		 * we may need space for two tree splits.
-		 *
-		 * We are relying on the caller to make the correct block
-		 * reservation for this operation to succeed.  If the
-		 * reservation amount is insufficient then we may fail a
-		 * block allocation here and corrupt the filesystem.
-		 */
-		args.minleft = args.tp->t_blk_res;
-	} else if (cur->bc_tp->t_flags & XFS_TRANS_LOWMODE) {
-		args.type = XFS_ALLOCTYPE_START_BNO;
-	} else {
-		args.type = XFS_ALLOCTYPE_NEAR_BNO;
-	}
-
 	args.minlen = args.maxlen = args.prod = 1;
 	args.wasdel = cur->bc_ino.flags & XFS_BTCUR_BMBT_WASDEL;
-	if (!args.wasdel && args.tp->t_blk_res == 0) {
-		error = -ENOSPC;
-		goto error0;
-	}
-	error = xfs_alloc_vextent(&args);
+	if (!args.wasdel && args.tp->t_blk_res == 0)
+		return -ENOSPC;
+
+	/*
+	 * If we are coming here from something like unwritten extent
+	 * conversion, there has been no data extent allocation already done, so
+	 * we have to ensure that we attempt to locate the entire set of bmbt
+	 * allocations in the same AG, as xfs_bmapi_write() would have reserved.
+	 */
+	if (cur->bc_tp->t_highest_agno == NULLAGNUMBER)
+		args.minleft = xfs_bmapi_minleft(cur->bc_tp, cur->bc_ino.ip,
+					cur->bc_ino.whichfork);
+
+	error = xfs_alloc_vextent_start_ag(&args, be64_to_cpu(start->l));
 	if (error)
-		goto error0;
+		return error;
 
 	if (args.fsbno == NULLFSBLOCK && args.minleft) {
 		/*
@@ -247,11 +234,10 @@ xfs_bmbt_alloc_block(
 		 * a full btree split.  Try again and if
 		 * successful activate the lowspace algorithm.
 		 */
-		args.fsbno = 0;
-		args.type = XFS_ALLOCTYPE_FIRST_AG;
-		error = xfs_alloc_vextent(&args);
+		args.minleft = 0;
+		error = xfs_alloc_vextent_start_ag(&args, 0);
 		if (error)
-			goto error0;
+			return error;
 		cur->bc_tp->t_flags |= XFS_TRANS_LOWMODE;
 	}
 	if (WARN_ON_ONCE(args.fsbno == NULLFSBLOCK)) {
@@ -260,7 +246,6 @@ xfs_bmbt_alloc_block(
 	}
 
 	ASSERT(args.len == 1);
-	cur->bc_tp->t_firstblock = args.fsbno;
 	cur->bc_ino.allocated++;
 	cur->bc_ino.ip->i_nblocks++;
 	xfs_trans_log_inode(args.tp, cur->bc_ino.ip, XFS_ILOG_CORE);
@@ -271,9 +256,6 @@ xfs_bmbt_alloc_block(
 
 	*stat = 1;
 	return 0;
-
- error0:
-	return error;
 }
 
 STATIC int

fs/xfs/libxfs/xfs_btree.c

@@ -2943,7 +2943,7 @@ xfs_btree_split(
 	DECLARE_COMPLETION_ONSTACK(done);
 
 	if (cur->bc_btnum != XFS_BTNUM_BMAP ||
-	    cur->bc_tp->t_firstblock == NULLFSBLOCK)
+	    cur->bc_tp->t_highest_agno == NULLAGNUMBER)
 		return __xfs_btree_split(cur, level, ptrp, key, curp, stat);
 
 	args.cur = cur;

fs/xfs/libxfs/xfs_ialloc.h

@@ -12,6 +12,7 @@ struct xfs_imap;
 struct xfs_mount;
 struct xfs_trans;
 struct xfs_btree_cur;
+struct xfs_perag;
 
 /* Move inodes in clusters of this size */
 #define	XFS_INODE_BIG_CLUSTER_SIZE	8192
@@ -47,7 +48,7 @@ int xfs_difree(struct xfs_trans *tp, struct xfs_perag *pag,
  */
 int
 xfs_imap(
-	struct xfs_mount *mp,		/* file system mount structure */
+	struct xfs_perag *pag,
 	struct xfs_trans *tp,		/* transaction pointer */
 	xfs_ino_t	ino,		/* inode to locate */
 	struct xfs_imap	*imap,		/* location map structure */
@@ -106,7 +107,7 @@ int xfs_ialloc_cluster_alignment(struct xfs_mount *mp);
 void xfs_ialloc_setup_geometry(struct xfs_mount *mp);
 xfs_ino_t xfs_ialloc_calc_rootino(struct xfs_mount *mp, int sunit);
 
-int xfs_ialloc_check_shrink(struct xfs_trans *tp, xfs_agnumber_t agno,
+int xfs_ialloc_check_shrink(struct xfs_perag *pag, struct xfs_trans *tp,
 		struct xfs_buf *agibp, xfs_agblock_t new_length);
 
 #endif	/* __XFS_IALLOC_H__ */

fs/xfs/libxfs/xfs_ialloc_btree.c

@@ -36,8 +36,8 @@ STATIC struct xfs_btree_cur *
 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.pag, cur->bc_btnum);
+	return xfs_inobt_init_cursor(cur->bc_ag.pag, cur->bc_tp,
+			cur->bc_ag.agbp, cur->bc_btnum);
 }
 
 STATIC void
@@ -103,15 +103,15 @@ __xfs_inobt_alloc_block(
 	memset(&args, 0, sizeof(args));
 	args.tp = cur->bc_tp;
 	args.mp = cur->bc_mp;
+	args.pag = cur->bc_ag.pag;
 	args.oinfo = XFS_RMAP_OINFO_INOBT;
-	args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_ag.pag->pag_agno, sbno);
 	args.minlen = 1;
 	args.maxlen = 1;
 	args.prod = 1;
-	args.type = XFS_ALLOCTYPE_NEAR_BNO;
 	args.resv = resv;
 
-	error = xfs_alloc_vextent(&args);
+	error = xfs_alloc_vextent_near_bno(&args,
+			XFS_AGB_TO_FSB(args.mp, args.pag->pag_agno, sbno));
 	if (error)
 		return error;
 
@@ -291,8 +291,8 @@ xfs_inobt_verify(
 	 * Similarly, during log recovery we will have a perag structure
 	 * attached, but the agi information will not yet have been initialised
 	 * from the on disk AGI. We don't currently use any of this information,
-	 * but beware of the landmine (i.e. need to check pag->pagi_init) if we
-	 * ever do.
+	 * but beware of the landmine (i.e. need to check
+	 * xfs_perag_initialised_agi(pag)) if we ever do.
 	 */
 	if (xfs_has_crc(mp)) {
 		fa = xfs_btree_sblock_v5hdr_verify(bp);
@@ -427,11 +427,11 @@ static const struct xfs_btree_ops xfs_finobt_ops = {
  */
 static struct xfs_btree_cur *
 xfs_inobt_init_common(
-	struct xfs_mount	*mp,		/* file system mount point */
-	struct xfs_trans	*tp,		/* transaction pointer */
 	struct xfs_perag	*pag,
+	struct xfs_trans	*tp,		/* transaction pointer */
 	xfs_btnum_t		btnum)		/* ialloc or free ino btree */
 {
+	struct xfs_mount	*mp = pag->pag_mount;
 	struct xfs_btree_cur	*cur;
 
 	cur = xfs_btree_alloc_cursor(mp, tp, btnum,
@@ -456,16 +456,15 @@ xfs_inobt_init_common(
 /* Create an inode btree cursor. */
 struct xfs_btree_cur *
 xfs_inobt_init_cursor(
-	struct xfs_mount	*mp,
+	struct xfs_perag	*pag,
 	struct xfs_trans	*tp,
 	struct xfs_buf		*agbp,
-	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, pag, btnum);
+	cur = xfs_inobt_init_common(pag, tp, btnum);
 	if (btnum == XFS_BTNUM_INO)
 		cur->bc_nlevels = be32_to_cpu(agi->agi_level);
 	else
@@ -477,14 +476,13 @@ xfs_inobt_init_cursor(
 /* Create an inode btree cursor with a fake root for staging. */
 struct xfs_btree_cur *
 xfs_inobt_stage_cursor(
-	struct xfs_mount	*mp,
-	struct xbtree_afakeroot	*afake,
 	struct xfs_perag	*pag,
+	struct xbtree_afakeroot	*afake,
 	xfs_btnum_t		btnum)
 {
 	struct xfs_btree_cur	*cur;
 
-	cur = xfs_inobt_init_common(mp, NULL, pag, btnum);
+	cur = xfs_inobt_init_common(pag, NULL, btnum);
 	xfs_btree_stage_afakeroot(cur, afake);
 	return cur;
 }
@@ -708,9 +706,8 @@ xfs_inobt_max_size(
 /* Read AGI and create inobt cursor. */
 int
 xfs_inobt_cur(
-	struct xfs_mount	*mp,
-	struct xfs_trans	*tp,
 	struct xfs_perag	*pag,
+	struct xfs_trans	*tp,
 	xfs_btnum_t		which,
 	struct xfs_btree_cur	**curpp,
 	struct xfs_buf		**agi_bpp)
@@ -725,16 +722,15 @@ xfs_inobt_cur(
 	if (error)
 		return error;
 
-	cur = xfs_inobt_init_cursor(mp, tp, *agi_bpp, pag, which);
+	cur = xfs_inobt_init_cursor(pag, tp, *agi_bpp, which);
 	*curpp = cur;
 	return 0;
 }
 
 static int
 xfs_inobt_count_blocks(
-	struct xfs_mount	*mp,
-	struct xfs_trans	*tp,
 	struct xfs_perag	*pag,
+	struct xfs_trans	*tp,
 	xfs_btnum_t		btnum,
 	xfs_extlen_t		*tree_blocks)
 {
@@ -742,7 +738,7 @@ xfs_inobt_count_blocks(
 	struct xfs_btree_cur	*cur = NULL;
 	int			error;
 
-	error = xfs_inobt_cur(mp, tp, pag, btnum, &cur, &agbp);
+	error = xfs_inobt_cur(pag, tp, btnum, &cur, &agbp);
 	if (error)
 		return error;
 
@@ -779,22 +775,21 @@ xfs_finobt_read_blocks(
  */
 int
 xfs_finobt_calc_reserves(
-	struct xfs_mount	*mp,
-	struct xfs_trans	*tp,
 	struct xfs_perag	*pag,
+	struct xfs_trans	*tp,
 	xfs_extlen_t		*ask,
 	xfs_extlen_t		*used)
 {
 	xfs_extlen_t		tree_len = 0;
 	int			error;
 
-	if (!xfs_has_finobt(mp))
+	if (!xfs_has_finobt(pag->pag_mount))
 		return 0;
 
-	if (xfs_has_inobtcounts(mp))
+	if (xfs_has_inobtcounts(pag->pag_mount))
 		error = xfs_finobt_read_blocks(pag, tp, &tree_len);
 	else
-		error = xfs_inobt_count_blocks(mp, tp, pag, XFS_BTNUM_FINO,
+		error = xfs_inobt_count_blocks(pag, tp, XFS_BTNUM_FINO,
 				&tree_len);
 	if (error)
 		return error;

fs/xfs/libxfs/xfs_ialloc_btree.h

@@ -46,12 +46,10 @@ struct xfs_perag;
 		 (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 *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, struct xfs_perag *pag,
-		xfs_btnum_t btnum);
+extern struct xfs_btree_cur *xfs_inobt_init_cursor(struct xfs_perag *pag,
+		struct xfs_trans *tp, struct xfs_buf *agbp, xfs_btnum_t btnum);
+struct xfs_btree_cur *xfs_inobt_stage_cursor(struct xfs_perag *pag,
+		struct xbtree_afakeroot *afake, xfs_btnum_t btnum);
 extern int xfs_inobt_maxrecs(struct xfs_mount *, int, int);
 
 /* ir_holemask to inode allocation bitmap conversion */
@@ -64,13 +62,13 @@ int xfs_inobt_rec_check_count(struct xfs_mount *,
 #define xfs_inobt_rec_check_count(mp, rec)	0
 #endif	/* DEBUG */
 
-int xfs_finobt_calc_reserves(struct xfs_mount *mp, struct xfs_trans *tp,
-		struct xfs_perag *pag, xfs_extlen_t *ask, xfs_extlen_t *used);
+int xfs_finobt_calc_reserves(struct xfs_perag *perag, struct xfs_trans *tp,
+		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,
-		struct xfs_perag *pag, xfs_btnum_t btnum,
-		struct xfs_btree_cur **curpp, struct xfs_buf **agi_bpp);
+int xfs_inobt_cur(struct xfs_perag *pag, struct xfs_trans *tp,
+		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,
 		struct xfs_trans *tp, struct xfs_buf *agbp);

fs/xfs/libxfs/xfs_refcount_btree.c

@@ -67,14 +67,14 @@ xfs_refcountbt_alloc_block(
 	memset(&args, 0, sizeof(args));
 	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.pag->pag_agno,
-			xfs_refc_block(args.mp));
+	args.pag = cur->bc_ag.pag;
 	args.oinfo = XFS_RMAP_OINFO_REFC;
 	args.minlen = args.maxlen = args.prod = 1;
 	args.resv = XFS_AG_RESV_METADATA;
 
-	error = xfs_alloc_vextent(&args);
+	error = xfs_alloc_vextent_near_bno(&args,
+			XFS_AGB_TO_FSB(args.mp, args.pag->pag_agno,
+					xfs_refc_block(args.mp)));
 	if (error)
 		goto out_error;
 	trace_xfs_refcountbt_alloc_block(cur->bc_mp, cur->bc_ag.pag->pag_agno,
@@ -227,7 +227,7 @@ xfs_refcountbt_verify(
 		return fa;
 
 	level = be16_to_cpu(block->bb_level);
-	if (pag && pag->pagf_init) {
+	if (pag && xfs_perag_initialised_agf(pag)) {
 		if (level >= pag->pagf_refcount_level)
 			return __this_address;
 	} else if (level >= mp->m_refc_maxlevels)

fs/xfs/libxfs/xfs_rmap_btree.c

@@ -313,7 +313,7 @@ xfs_rmapbt_verify(
 		return fa;
 
 	level = be16_to_cpu(block->bb_level);
-	if (pag && pag->pagf_init) {
+	if (pag && xfs_perag_initialised_agf(pag)) {
 		if (level >= pag->pagf_levels[XFS_BTNUM_RMAPi])
 			return __this_address;
 	} else if (level >= mp->m_rmap_maxlevels)

fs/xfs/libxfs/xfs_sb.c

@@ -909,7 +909,8 @@ xfs_sb_mount_common(
 	struct xfs_mount	*mp,
 	struct xfs_sb		*sbp)
 {
-	mp->m_agfrotor = mp->m_agirotor = 0;
+	mp->m_agfrotor = 0;
+	atomic_set(&mp->m_agirotor, 0);
 	mp->m_maxagi = mp->m_sb.sb_agcount;
 	mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
 	mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;

fs/xfs/scrub/agheader_repair.c

@@ -191,14 +191,15 @@ xrep_agf_init_header(
 	struct xfs_agf		*old_agf)
 {
 	struct xfs_mount	*mp = sc->mp;
+	struct xfs_perag	*pag = sc->sa.pag;
 	struct xfs_agf		*agf = agf_bp->b_addr;
 
 	memcpy(old_agf, agf, sizeof(*old_agf));
 	memset(agf, 0, BBTOB(agf_bp->b_length));
 	agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
 	agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
-	agf->agf_seqno = cpu_to_be32(sc->sa.pag->pag_agno);
-	agf->agf_length = cpu_to_be32(sc->sa.pag->block_count);
+	agf->agf_seqno = cpu_to_be32(pag->pag_agno);
+	agf->agf_length = cpu_to_be32(pag->block_count);
 	agf->agf_flfirst = old_agf->agf_flfirst;
 	agf->agf_fllast = old_agf->agf_fllast;
 	agf->agf_flcount = old_agf->agf_flcount;
@@ -206,8 +207,8 @@ xrep_agf_init_header(
 		uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_meta_uuid);
 
 	/* Mark the incore AGF data stale until we're done fixing things. */
-	ASSERT(sc->sa.pag->pagf_init);
-	sc->sa.pag->pagf_init = 0;
+	ASSERT(xfs_perag_initialised_agf(pag));
+	clear_bit(XFS_AGSTATE_AGF_INIT, &pag->pag_opstate);
 }
 
 /* Set btree root information in an AGF. */
@@ -333,7 +334,7 @@ xrep_agf_commit_new(
 	pag->pagf_levels[XFS_BTNUM_RMAPi] =
 			be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAPi]);
 	pag->pagf_refcount_level = be32_to_cpu(agf->agf_refcount_level);
-	pag->pagf_init = 1;
+	set_bit(XFS_AGSTATE_AGF_INIT, &pag->pag_opstate);
 
 	return 0;
 }
@@ -434,7 +435,7 @@ xrep_agf(
 
 out_revert:
 	/* Mark the incore AGF state stale and revert the AGF. */
-	sc->sa.pag->pagf_init = 0;
+	clear_bit(XFS_AGSTATE_AGF_INIT, &sc->sa.pag->pag_opstate);
 	memcpy(agf, &old_agf, sizeof(old_agf));
 	return error;
 }
@@ -618,7 +619,7 @@ xrep_agfl_update_agf(
 	xfs_force_summary_recalc(sc->mp);
 
 	/* Update the AGF counters. */
-	if (sc->sa.pag->pagf_init)
+	if (xfs_perag_initialised_agf(sc->sa.pag))
 		sc->sa.pag->pagf_flcount = flcount;
 	agf->agf_flfirst = cpu_to_be32(0);
 	agf->agf_flcount = cpu_to_be32(flcount);
@@ -822,14 +823,15 @@ xrep_agi_init_header(
 	struct xfs_agi		*old_agi)
 {
 	struct xfs_agi		*agi = agi_bp->b_addr;
+	struct xfs_perag	*pag = sc->sa.pag;
 	struct xfs_mount	*mp = sc->mp;
 
 	memcpy(old_agi, agi, sizeof(*old_agi));
 	memset(agi, 0, BBTOB(agi_bp->b_length));
 	agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
 	agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
-	agi->agi_seqno = cpu_to_be32(sc->sa.pag->pag_agno);
-	agi->agi_length = cpu_to_be32(sc->sa.pag->block_count);
+	agi->agi_seqno = cpu_to_be32(pag->pag_agno);
+	agi->agi_length = cpu_to_be32(pag->block_count);
 	agi->agi_newino = cpu_to_be32(NULLAGINO);
 	agi->agi_dirino = cpu_to_be32(NULLAGINO);
 	if (xfs_has_crc(mp))
@@ -840,8 +842,8 @@ xrep_agi_init_header(
 			sizeof(agi->agi_unlinked));
 
 	/* Mark the incore AGF data stale until we're done fixing things. */
-	ASSERT(sc->sa.pag->pagi_init);
-	sc->sa.pag->pagi_init = 0;
+	ASSERT(xfs_perag_initialised_agi(pag));
+	clear_bit(XFS_AGSTATE_AGI_INIT, &pag->pag_opstate);
 }
 
 /* Set btree root information in an AGI. */
@@ -873,8 +875,7 @@ xrep_agi_calc_from_btrees(
 	xfs_agino_t		freecount;
 	int			error;
 
-	cur = xfs_inobt_init_cursor(mp, sc->tp, agi_bp,
-			sc->sa.pag, XFS_BTNUM_INO);
+	cur = xfs_inobt_init_cursor(sc->sa.pag, sc->tp, agi_bp, XFS_BTNUM_INO);
 	error = xfs_ialloc_count_inodes(cur, &count, &freecount);
 	if (error)
 		goto err;
@@ -894,8 +895,8 @@ xrep_agi_calc_from_btrees(
 	if (xfs_has_finobt(mp) && xfs_has_inobtcounts(mp)) {
 		xfs_agblock_t	blocks;
 
-		cur = xfs_inobt_init_cursor(mp, sc->tp, agi_bp,
-				sc->sa.pag, XFS_BTNUM_FINO);
+		cur = xfs_inobt_init_cursor(sc->sa.pag, sc->tp, agi_bp,
+				XFS_BTNUM_FINO);
 		error = xfs_btree_count_blocks(cur, &blocks);
 		if (error)
 			goto err;
@@ -929,7 +930,7 @@ xrep_agi_commit_new(
 	pag = sc->sa.pag;
 	pag->pagi_count = be32_to_cpu(agi->agi_count);
 	pag->pagi_freecount = be32_to_cpu(agi->agi_freecount);
-	pag->pagi_init = 1;
+	set_bit(XFS_AGSTATE_AGI_INIT, &pag->pag_opstate);
 
 	return 0;
 }
@@ -994,7 +995,7 @@ xrep_agi(
 
 out_revert:
 	/* Mark the incore AGI state stale and revert the AGI. */
-	sc->sa.pag->pagi_init = 0;
+	clear_bit(XFS_AGSTATE_AGI_INIT, &sc->sa.pag->pag_opstate);
 	memcpy(agi, &old_agi, sizeof(old_agi));
 	return error;
 }

fs/xfs/scrub/bmap.c

@@ -662,7 +662,7 @@ xchk_bmap_check_rmaps(
 		error = xchk_bmap_check_ag_rmaps(sc, whichfork, pag);
 		if (error ||
 		    (sc->sm->sm_flags & XFS_SCRUB_OFLAG_CORRUPT)) {
-			xfs_perag_put(pag);
+			xfs_perag_rele(pag);
 			return error;
 		}
 	}

fs/xfs/scrub/common.c

@@ -478,15 +478,15 @@ xchk_ag_btcur_init(
 	/* 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,
-				sa->pag, XFS_BTNUM_INO);
+		sa->ino_cur = xfs_inobt_init_cursor(sa->pag, sc->tp, sa->agi_bp,
+				XFS_BTNUM_INO);
 	}
 
 	/* Set up a finobt cursor for cross-referencing. */
 	if (sa->agi_bp && xfs_has_finobt(mp) &&
 	    xchk_ag_btree_healthy_enough(sc, sa->pag, XFS_BTNUM_FINO)) {
-		sa->fino_cur = xfs_inobt_init_cursor(mp, sc->tp, sa->agi_bp,
-				sa->pag, XFS_BTNUM_FINO);
+		sa->fino_cur = xfs_inobt_init_cursor(sa->pag, sc->tp, sa->agi_bp,
+				XFS_BTNUM_FINO);
 	}
 
 	/* Set up a rmapbt cursor for cross-referencing. */
@@ -636,6 +636,7 @@ xchk_get_inode(
 {
 	struct xfs_imap		imap;
 	struct xfs_mount	*mp = sc->mp;
+	struct xfs_perag	*pag;
 	struct xfs_inode	*ip_in = XFS_I(file_inode(sc->file));
 	struct xfs_inode	*ip = NULL;
 	int			error;
@@ -671,10 +672,14 @@ xchk_get_inode(
 		 * Otherwise, we really couldn't find it so tell userspace
 		 * that it no longer exists.
 		 */
-		error = xfs_imap(sc->mp, sc->tp, sc->sm->sm_ino, &imap,
-				XFS_IGET_UNTRUSTED | XFS_IGET_DONTCACHE);
-		if (error)
-			return -ENOENT;
+		pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, sc->sm->sm_ino));
+		if (pag) {
+			error = xfs_imap(pag, sc->tp, sc->sm->sm_ino, &imap,
+					XFS_IGET_UNTRUSTED | XFS_IGET_DONTCACHE);
+			xfs_perag_put(pag);
+			if (error)
+				return -ENOENT;
+		}
 		error = -EFSCORRUPTED;
 		fallthrough;
 	default:

fs/xfs/scrub/fscounters.c

@@ -86,7 +86,8 @@ xchk_fscount_warmup(
 	for_each_perag(mp, agno, pag) {
 		if (xchk_should_terminate(sc, &error))
 			break;
-		if (pag->pagi_init && pag->pagf_init)
+		if (xfs_perag_initialised_agi(pag) &&
+		    xfs_perag_initialised_agf(pag))
 			continue;
 
 		/* Lock both AG headers. */
@@ -101,7 +102,8 @@ xchk_fscount_warmup(
 		 * These are supposed to be initialized by the header read
 		 * function.
 		 */
-		if (!pag->pagi_init || !pag->pagf_init) {
+		if (!xfs_perag_initialised_agi(pag) ||
+		    !xfs_perag_initialised_agf(pag)) {
 			error = -EFSCORRUPTED;
 			break;
 		}
@@ -117,7 +119,7 @@ xchk_fscount_warmup(
 	if (agi_bp)
 		xfs_buf_relse(agi_bp);
 	if (pag)
-		xfs_perag_put(pag);
+		xfs_perag_rele(pag);
 	return error;
 }
 
@@ -220,7 +222,8 @@ xchk_fscount_aggregate_agcounts(
 			break;
 
 		/* This somehow got unset since the warmup? */
-		if (!pag->pagi_init || !pag->pagf_init) {
+		if (!xfs_perag_initialised_agi(pag) ||
+		    !xfs_perag_initialised_agf(pag)) {
 			error = -EFSCORRUPTED;
 			break;
 		}
@@ -249,7 +252,7 @@ xchk_fscount_aggregate_agcounts(
 
 	}
 	if (pag)
-		xfs_perag_put(pag);
+		xfs_perag_rele(pag);
 	if (error) {
 		xchk_set_incomplete(sc);
 		return error;

fs/xfs/scrub/repair.c

@@ -206,7 +206,7 @@ xrep_calc_ag_resblks(
 		return 0;
 
 	pag = xfs_perag_get(mp, sm->sm_agno);
-	if (pag->pagi_init) {
+	if (xfs_perag_initialised_agi(pag)) {
 		/* Use in-core icount if possible. */
 		icount = pag->pagi_count;
 	} else {
@@ -326,15 +326,14 @@ xrep_alloc_ag_block(
 
 	args.tp = sc->tp;
 	args.mp = sc->mp;
+	args.pag = sc->sa.pag;
 	args.oinfo = *oinfo;
-	args.fsbno = XFS_AGB_TO_FSB(args.mp, sc->sa.pag->pag_agno, 0);
 	args.minlen = 1;
 	args.maxlen = 1;
 	args.prod = 1;
-	args.type = XFS_ALLOCTYPE_THIS_AG;
 	args.resv = resv;
 
-	error = xfs_alloc_vextent(&args);
+	error = xfs_alloc_vextent_this_ag(&args, sc->sa.pag->pag_agno);
 	if (error)
 		return error;
 	if (args.fsbno == NULLFSBLOCK)

fs/xfs/xfs_bmap_util.c

@@ -1410,7 +1410,7 @@ xfs_swap_extent_rmap(
 
 		/* Unmap the old blocks in the source file. */
 		while (tirec.br_blockcount) {
-			ASSERT(tp->t_firstblock == NULLFSBLOCK);
+			ASSERT(tp->t_highest_agno == NULLAGNUMBER);
 			trace_xfs_swap_extent_rmap_remap_piece(tip, &tirec);
 
 			/* Read extent from the source file */

fs/xfs/xfs_discard.c

@@ -21,23 +21,20 @@
 
 STATIC int
 xfs_trim_extents(
-	struct xfs_mount	*mp,
-	xfs_agnumber_t		agno,
+	struct xfs_perag	*pag,
 	xfs_daddr_t		start,
 	xfs_daddr_t		end,
 	xfs_daddr_t		minlen,
 	uint64_t		*blocks_trimmed)
 {
+	struct xfs_mount	*mp = pag->pag_mount;
 	struct block_device	*bdev = mp->m_ddev_targp->bt_bdev;
 	struct xfs_btree_cur	*cur;
 	struct xfs_buf		*agbp;
 	struct xfs_agf		*agf;
-	struct xfs_perag	*pag;
 	int			error;
 	int			i;
 
-	pag = xfs_perag_get(mp, agno);
-
 	/*
 	 * Force out the log.  This means any transactions that might have freed
 	 * space before we take the AGF buffer lock are now on disk, and the
@@ -47,7 +44,7 @@ xfs_trim_extents(
 
 	error = xfs_alloc_read_agf(pag, NULL, 0, &agbp);
 	if (error)
-		goto out_put_perag;
+		return error;
 	agf = agbp->b_addr;
 
 	cur = xfs_allocbt_init_cursor(mp, NULL, agbp, pag, XFS_BTNUM_CNT);
@@ -71,10 +68,10 @@ xfs_trim_extents(
 
 		error = xfs_alloc_get_rec(cur, &fbno, &flen, &i);
 		if (error)
-			goto out_del_cursor;
+			break;
 		if (XFS_IS_CORRUPT(mp, i != 1)) {
 			error = -EFSCORRUPTED;
-			goto out_del_cursor;
+			break;
 		}
 		ASSERT(flen <= be32_to_cpu(agf->agf_longest));
 
@@ -83,15 +80,15 @@ xfs_trim_extents(
 		 * the format the range/len variables are supplied in by
 		 * userspace.
 		 */
-		dbno = XFS_AGB_TO_DADDR(mp, agno, fbno);
+		dbno = XFS_AGB_TO_DADDR(mp, pag->pag_agno, fbno);
 		dlen = XFS_FSB_TO_BB(mp, flen);
 
 		/*
 		 * Too small?  Give up.
 		 */
 		if (dlen < minlen) {
-			trace_xfs_discard_toosmall(mp, agno, fbno, flen);
-			goto out_del_cursor;
+			trace_xfs_discard_toosmall(mp, pag->pag_agno, fbno, flen);
+			break;
 		}
 
 		/*
@@ -100,7 +97,7 @@ xfs_trim_extents(
 		 * down partially overlapping ranges for now.
 		 */
 		if (dbno + dlen < start || dbno > end) {
-			trace_xfs_discard_exclude(mp, agno, fbno, flen);
+			trace_xfs_discard_exclude(mp, pag->pag_agno, fbno, flen);
 			goto next_extent;
 		}
 
@@ -109,32 +106,30 @@ xfs_trim_extents(
 		 * discard and try again the next time.
 		 */
 		if (xfs_extent_busy_search(mp, pag, fbno, flen)) {
-			trace_xfs_discard_busy(mp, agno, fbno, flen);
+			trace_xfs_discard_busy(mp, pag->pag_agno, fbno, flen);
 			goto next_extent;
 		}
 
-		trace_xfs_discard_extent(mp, agno, fbno, flen);
+		trace_xfs_discard_extent(mp, pag->pag_agno, fbno, flen);
 		error = blkdev_issue_discard(bdev, dbno, dlen, GFP_NOFS);
 		if (error)
-			goto out_del_cursor;
+			break;
 		*blocks_trimmed += flen;
 
 next_extent:
 		error = xfs_btree_decrement(cur, 0, &i);
 		if (error)
-			goto out_del_cursor;
+			break;
 
 		if (fatal_signal_pending(current)) {
 			error = -ERESTARTSYS;
-			goto out_del_cursor;
+			break;
 		}
 	}
 
 out_del_cursor:
 	xfs_btree_del_cursor(cur, error);
 	xfs_buf_relse(agbp);
-out_put_perag:
-	xfs_perag_put(pag);
 	return error;
 }
 
@@ -152,11 +147,12 @@ xfs_ioc_trim(
 	struct xfs_mount		*mp,
 	struct fstrim_range __user	*urange)
 {
+	struct xfs_perag	*pag;
 	unsigned int		granularity =
 		bdev_discard_granularity(mp->m_ddev_targp->bt_bdev);
 	struct fstrim_range	range;
 	xfs_daddr_t		start, end, minlen;
-	xfs_agnumber_t		start_agno, end_agno, agno;
+	xfs_agnumber_t		agno;
 	uint64_t		blocks_trimmed = 0;
 	int			error, last_error = 0;
 
@@ -193,18 +189,18 @@ xfs_ioc_trim(
 	end = start + BTOBBT(range.len) - 1;
 
 	if (end > XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) - 1)
-		end = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)- 1;
-
-	start_agno = xfs_daddr_to_agno(mp, start);
-	end_agno = xfs_daddr_to_agno(mp, end);
+		end = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) - 1;
 
-	for (agno = start_agno; agno <= end_agno; agno++) {
-		error = xfs_trim_extents(mp, agno, start, end, minlen,
+	agno = xfs_daddr_to_agno(mp, start);
+	for_each_perag_range(mp, agno, xfs_daddr_to_agno(mp, end), pag) {
+		error = xfs_trim_extents(pag, start, end, minlen,
 					  &blocks_trimmed);
 		if (error) {
 			last_error = error;
-			if (error == -ERESTARTSYS)
+			if (error == -ERESTARTSYS) {
+				xfs_perag_rele(pag);
 				break;
+			}
 		}
 	}
 

fs/xfs/xfs_filestream.h

@@ -9,13 +9,13 @@
 struct xfs_mount;
 struct xfs_inode;
 struct xfs_bmalloca;
+struct xfs_alloc_arg;
 
 int xfs_filestream_mount(struct xfs_mount *mp);
 void xfs_filestream_unmount(struct xfs_mount *mp);
 void xfs_filestream_deassociate(struct xfs_inode *ip);
-xfs_agnumber_t xfs_filestream_lookup_ag(struct xfs_inode *ip);
-int xfs_filestream_new_ag(struct xfs_bmalloca *ap, xfs_agnumber_t *agp);
-int xfs_filestream_peek_ag(struct xfs_mount *mp, xfs_agnumber_t agno);
+int xfs_filestream_select_ag(struct xfs_bmalloca *ap,
+		struct xfs_alloc_arg *args, xfs_extlen_t *blen);
 
 static inline int
 xfs_inode_is_filestream(

fs/xfs/xfs_fsmap.c

@@ -688,11 +688,11 @@ __xfs_getfsmap_datadev(
 		info->agf_bp = NULL;
 	}
 	if (info->pag) {
-		xfs_perag_put(info->pag);
+		xfs_perag_rele(info->pag);
 		info->pag = NULL;
 	} else if (pag) {
 		/* loop termination case */
-		xfs_perag_put(pag);
+		xfs_perag_rele(pag);
 	}
 
 	return error;

fs/xfs/xfs_icache.c

@@ -255,7 +255,7 @@ xfs_perag_set_inode_tag(
 		break;
 	}
 
-	trace_xfs_perag_set_inode_tag(mp, pag->pag_agno, tag, _RET_IP_);
+	trace_xfs_perag_set_inode_tag(pag, _RET_IP_);
 }
 
 /* Clear a tag on both the AG incore inode tree and the AG radix tree. */
@@ -289,7 +289,7 @@ xfs_perag_clear_inode_tag(
 	radix_tree_tag_clear(&mp->m_perag_tree, pag->pag_agno, tag);
 	spin_unlock(&mp->m_perag_lock);
 
-	trace_xfs_perag_clear_inode_tag(mp, pag->pag_agno, tag, _RET_IP_);
+	trace_xfs_perag_clear_inode_tag(pag, _RET_IP_);
 }
 
 /*
@@ -586,7 +586,7 @@ xfs_iget_cache_miss(
 	if (!ip)
 		return -ENOMEM;
 
-	error = xfs_imap(mp, tp, ip->i_ino, &ip->i_imap, flags);
+	error = xfs_imap(pag, tp, ip->i_ino, &ip->i_imap, flags);
 	if (error)
 		goto out_destroy;
 
@@ -1767,7 +1767,7 @@ xfs_icwalk(
 		if (error) {
 			last_error = error;
 			if (error == -EFSCORRUPTED) {
-				xfs_perag_put(pag);
+				xfs_perag_rele(pag);
 				break;
 			}
 		}

fs/xfs/xfs_inode.c

@@ -1367,7 +1367,7 @@ xfs_itruncate_extents_flags(
 
 	unmap_len = XFS_MAX_FILEOFF - first_unmap_block + 1;
 	while (unmap_len > 0) {
-		ASSERT(tp->t_firstblock == NULLFSBLOCK);
+		ASSERT(tp->t_highest_agno == NULLAGNUMBER);
 		error = __xfs_bunmapi(tp, ip, first_unmap_block, &unmap_len,
 				flags, XFS_ITRUNC_MAX_EXTENTS);
 		if (error)

fs/xfs/xfs_iwalk.c

@@ -275,7 +275,7 @@ xfs_iwalk_ag_start(
 
 	/* Set up a fresh cursor and empty the inobt cache. */
 	iwag->nr_recs = 0;
-	error = xfs_inobt_cur(mp, tp, pag, XFS_BTNUM_INO, curpp, agi_bpp);
+	error = xfs_inobt_cur(pag, tp, XFS_BTNUM_INO, curpp, agi_bpp);
 	if (error)
 		return error;
 
@@ -390,7 +390,7 @@ xfs_iwalk_run_callbacks(
 	}
 
 	/* ...and recreate the cursor just past where we left off. */
-	error = xfs_inobt_cur(mp, iwag->tp, iwag->pag, XFS_BTNUM_INO, curpp,
+	error = xfs_inobt_cur(iwag->pag, iwag->tp, XFS_BTNUM_INO, curpp,
 			agi_bpp);
 	if (error)
 		return error;
@@ -591,7 +591,7 @@ xfs_iwalk(
 	}
 
 	if (iwag.pag)
-		xfs_perag_put(pag);
+		xfs_perag_rele(pag);
 	xfs_iwalk_free(&iwag);
 	return error;
 }
@@ -683,7 +683,7 @@ xfs_iwalk_threaded(
 			break;
 	}
 	if (pag)
-		xfs_perag_put(pag);
+		xfs_perag_rele(pag);
 	if (polled)
 		xfs_pwork_poll(&pctl);
 	return xfs_pwork_destroy(&pctl);
@@ -776,7 +776,7 @@ xfs_inobt_walk(
 	}
 
 	if (iwag.pag)
-		xfs_perag_put(pag);
+		xfs_perag_rele(pag);
 	xfs_iwalk_free(&iwag);
 	return error;
 }

fs/xfs/xfs_mount.h

@@ -210,8 +210,7 @@ typedef struct xfs_mount {
 	struct xfs_error_cfg	m_error_cfg[XFS_ERR_CLASS_MAX][XFS_ERR_ERRNO_MAX];
 	struct xstats		m_stats;	/* per-fs stats */
 	xfs_agnumber_t		m_agfrotor;	/* last ag where space found */
-	xfs_agnumber_t		m_agirotor;	/* last ag dir inode alloced */
-	spinlock_t		m_agirotor_lock;/* .. and lock protecting it */
+	atomic_t		m_agirotor;	/* last ag dir inode alloced */
 
 	/* Memory shrinker to throttle and reprioritize inodegc */
 	struct shrinker		m_inodegc_shrinker;

fs/xfs/xfs_reflink.c

@@ -610,7 +610,7 @@ xfs_reflink_cancel_cow_blocks(
 			if (error)
 				break;
 		} else if (del.br_state == XFS_EXT_UNWRITTEN || cancel_real) {
-			ASSERT((*tpp)->t_firstblock == NULLFSBLOCK);
+			ASSERT((*tpp)->t_highest_agno == NULLAGNUMBER);
 
 			/* Free the CoW orphan record. */
 			xfs_refcount_free_cow_extent(*tpp, del.br_startblock,
@@ -927,7 +927,7 @@ xfs_reflink_recover_cow(
 	for_each_perag(mp, agno, pag) {
 		error = xfs_refcount_recover_cow_leftovers(mp, pag);
 		if (error) {
-			xfs_perag_put(pag);
+			xfs_perag_rele(pag);
 			break;
 		}
 	}

fs/xfs/xfs_super.c

@@ -247,6 +247,32 @@ xfs_fs_show_options(
 	return 0;
 }
 
+static bool
+xfs_set_inode_alloc_perag(
+	struct xfs_perag	*pag,
+	xfs_ino_t		ino,
+	xfs_agnumber_t		max_metadata)
+{
+	if (!xfs_is_inode32(pag->pag_mount)) {
+		set_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate);
+		clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate);
+		return false;
+	}
+
+	if (ino > XFS_MAXINUMBER_32) {
+		clear_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate);
+		clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate);
+		return false;
+	}
+
+	set_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate);
+	if (pag->pag_agno < max_metadata)
+		set_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate);
+	else
+		clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate);
+	return true;
+}
+
 /*
  * Set parameters for inode allocation heuristics, taking into account
  * filesystem size and inode32/inode64 mount options; i.e. specifically
@@ -310,24 +336,8 @@ xfs_set_inode_alloc(
 		ino = XFS_AGINO_TO_INO(mp, index, agino);
 
 		pag = xfs_perag_get(mp, index);
-
-		if (xfs_is_inode32(mp)) {
-			if (ino > XFS_MAXINUMBER_32) {
-				pag->pagi_inodeok = 0;
-				pag->pagf_metadata = 0;
-			} else {
-				pag->pagi_inodeok = 1;
-				maxagi++;
-				if (index < max_metadata)
-					pag->pagf_metadata = 1;
-				else
-					pag->pagf_metadata = 0;
-			}
-		} else {
-			pag->pagi_inodeok = 1;
-			pag->pagf_metadata = 0;
-		}
-
+		if (xfs_set_inode_alloc_perag(pag, ino, max_metadata))
+			maxagi++;
 		xfs_perag_put(pag);
 	}
 
@@ -1922,7 +1932,6 @@ static int xfs_init_fs_context(
 		return -ENOMEM;
 
 	spin_lock_init(&mp->m_sb_lock);
-	spin_lock_init(&mp->m_agirotor_lock);
 	INIT_RADIX_TREE(&mp->m_perag_tree, GFP_ATOMIC);
 	spin_lock_init(&mp->m_perag_lock);
 	mutex_init(&mp->m_growlock);

fs/xfs/xfs_trace.h

@@ -74,6 +74,7 @@ struct xfs_inobt_rec_incore;
 union xfs_btree_ptr;
 struct xfs_dqtrx;
 struct xfs_icwalk;
+struct xfs_perag;
 
 #define XFS_ATTR_FILTER_FLAGS \
 	{ XFS_ATTR_ROOT,	"ROOT" }, \
@@ -159,36 +160,40 @@ TRACE_EVENT(xlog_intent_recovery_failed,
 );
 
 DECLARE_EVENT_CLASS(xfs_perag_class,
-	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, int refcount,
-		 unsigned long caller_ip),
-	TP_ARGS(mp, agno, refcount, caller_ip),
+	TP_PROTO(struct xfs_perag *pag, unsigned long caller_ip),
+	TP_ARGS(pag, caller_ip),
 	TP_STRUCT__entry(
 		__field(dev_t, dev)
 		__field(xfs_agnumber_t, agno)
 		__field(int, refcount)
+		__field(int, active_refcount)
 		__field(unsigned long, caller_ip)
 	),
 	TP_fast_assign(
-		__entry->dev = mp->m_super->s_dev;
-		__entry->agno = agno;
-		__entry->refcount = refcount;
+		__entry->dev = pag->pag_mount->m_super->s_dev;
+		__entry->agno = pag->pag_agno;
+		__entry->refcount = atomic_read(&pag->pag_ref);
+		__entry->active_refcount = atomic_read(&pag->pag_active_ref);
 		__entry->caller_ip = caller_ip;
 	),
-	TP_printk("dev %d:%d agno 0x%x refcount %d caller %pS",
+	TP_printk("dev %d:%d agno 0x%x passive refs %d active refs %d caller %pS",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  __entry->agno,
 		  __entry->refcount,
+		  __entry->active_refcount,
 		  (char *)__entry->caller_ip)
 );
 
 #define DEFINE_PERAG_REF_EVENT(name)	\
 DEFINE_EVENT(xfs_perag_class, name,	\
-	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, int refcount,	\
-		 unsigned long caller_ip),					\
-	TP_ARGS(mp, agno, refcount, caller_ip))
+	TP_PROTO(struct xfs_perag *pag, unsigned long caller_ip), \
+	TP_ARGS(pag, caller_ip))
 DEFINE_PERAG_REF_EVENT(xfs_perag_get);
 DEFINE_PERAG_REF_EVENT(xfs_perag_get_tag);
 DEFINE_PERAG_REF_EVENT(xfs_perag_put);
+DEFINE_PERAG_REF_EVENT(xfs_perag_grab);
+DEFINE_PERAG_REF_EVENT(xfs_perag_grab_tag);
+DEFINE_PERAG_REF_EVENT(xfs_perag_rele);
 DEFINE_PERAG_REF_EVENT(xfs_perag_set_inode_tag);
 DEFINE_PERAG_REF_EVENT(xfs_perag_clear_inode_tag);
 
@@ -634,8 +639,8 @@ DEFINE_BUF_ITEM_EVENT(xfs_trans_bhold_release);
 DEFINE_BUF_ITEM_EVENT(xfs_trans_binval);
 
 DECLARE_EVENT_CLASS(xfs_filestream_class,
-	TP_PROTO(struct xfs_mount *mp, xfs_ino_t ino, xfs_agnumber_t agno),
-	TP_ARGS(mp, ino, agno),
+	TP_PROTO(struct xfs_perag *pag, xfs_ino_t ino),
+	TP_ARGS(pag, ino),
 	TP_STRUCT__entry(
 		__field(dev_t, dev)
 		__field(xfs_ino_t, ino)
@@ -643,10 +648,10 @@ DECLARE_EVENT_CLASS(xfs_filestream_class,
 		__field(int, streams)
 	),
 	TP_fast_assign(
-		__entry->dev = mp->m_super->s_dev;
+		__entry->dev = pag->pag_mount->m_super->s_dev;
 		__entry->ino = ino;
-		__entry->agno = agno;
-		__entry->streams = xfs_filestream_peek_ag(mp, agno);
+		__entry->agno = pag->pag_agno;
+		__entry->streams = atomic_read(&pag->pagf_fstrms);
 	),
 	TP_printk("dev %d:%d ino 0x%llx agno 0x%x streams %d",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
@@ -656,39 +661,40 @@ DECLARE_EVENT_CLASS(xfs_filestream_class,
 )
 #define DEFINE_FILESTREAM_EVENT(name) \
 DEFINE_EVENT(xfs_filestream_class, name, \
-	TP_PROTO(struct xfs_mount *mp, xfs_ino_t ino, xfs_agnumber_t agno), \
-	TP_ARGS(mp, ino, agno))
+	TP_PROTO(struct xfs_perag *pag, xfs_ino_t ino), \
+	TP_ARGS(pag, ino))
 DEFINE_FILESTREAM_EVENT(xfs_filestream_free);
 DEFINE_FILESTREAM_EVENT(xfs_filestream_lookup);
 DEFINE_FILESTREAM_EVENT(xfs_filestream_scan);
 
 TRACE_EVENT(xfs_filestream_pick,
-	TP_PROTO(struct xfs_inode *ip, xfs_agnumber_t agno,
-		 xfs_extlen_t free, int nscan),
-	TP_ARGS(ip, agno, free, nscan),
+	TP_PROTO(struct xfs_perag *pag, xfs_ino_t ino, xfs_extlen_t free),
+	TP_ARGS(pag, ino, free),
 	TP_STRUCT__entry(
 		__field(dev_t, dev)
 		__field(xfs_ino_t, ino)
 		__field(xfs_agnumber_t, agno)
 		__field(int, streams)
 		__field(xfs_extlen_t, free)
-		__field(int, nscan)
 	),
 	TP_fast_assign(
-		__entry->dev = VFS_I(ip)->i_sb->s_dev;
-		__entry->ino = ip->i_ino;
-		__entry->agno = agno;
-		__entry->streams = xfs_filestream_peek_ag(ip->i_mount, agno);
+		__entry->dev = pag->pag_mount->m_super->s_dev;
+		__entry->ino = ino;
+		if (pag) {
+			__entry->agno = pag->pag_agno;
+			__entry->streams = atomic_read(&pag->pagf_fstrms);
+		} else {
+			__entry->agno = NULLAGNUMBER;
+			__entry->streams = 0;
+		}
 		__entry->free = free;
-		__entry->nscan = nscan;
 	),
-	TP_printk("dev %d:%d ino 0x%llx agno 0x%x streams %d free %d nscan %d",
+	TP_printk("dev %d:%d ino 0x%llx agno 0x%x streams %d free %d",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  __entry->ino,
 		  __entry->agno,
 		  __entry->streams,
-		  __entry->free,
-		  __entry->nscan)
+		  __entry->free)
 );
 
 DECLARE_EVENT_CLASS(xfs_lock_class,
@@ -1795,13 +1801,11 @@ DECLARE_EVENT_CLASS(xfs_alloc_class,
 		__field(xfs_extlen_t, alignment)
 		__field(xfs_extlen_t, minalignslop)
 		__field(xfs_extlen_t, len)
-		__field(short, type)
-		__field(short, otype)
 		__field(char, wasdel)
 		__field(char, wasfromfl)
 		__field(int, resv)
 		__field(int, datatype)
-		__field(xfs_fsblock_t, firstblock)
+		__field(xfs_agnumber_t, highest_agno)
 	),
 	TP_fast_assign(
 		__entry->dev = args->mp->m_super->s_dev;
@@ -1816,18 +1820,16 @@ DECLARE_EVENT_CLASS(xfs_alloc_class,
 		__entry->alignment = args->alignment;
 		__entry->minalignslop = args->minalignslop;
 		__entry->len = args->len;
-		__entry->type = args->type;
-		__entry->otype = args->otype;
 		__entry->wasdel = args->wasdel;
 		__entry->wasfromfl = args->wasfromfl;
 		__entry->resv = args->resv;
 		__entry->datatype = args->datatype;
-		__entry->firstblock = args->tp->t_firstblock;
+		__entry->highest_agno = args->tp->t_highest_agno;
 	),
 	TP_printk("dev %d:%d agno 0x%x agbno 0x%x minlen %u maxlen %u mod %u "
 		  "prod %u minleft %u total %u alignment %u minalignslop %u "
-		  "len %u type %s otype %s wasdel %d wasfromfl %d resv %d "
-		  "datatype 0x%x firstblock 0x%llx",
+		  "len %u wasdel %d wasfromfl %d resv %d "
+		  "datatype 0x%x highest_agno 0x%x",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  __entry->agno,
 		  __entry->agbno,
@@ -1840,13 +1842,11 @@ DECLARE_EVENT_CLASS(xfs_alloc_class,
 		  __entry->alignment,
 		  __entry->minalignslop,
 		  __entry->len,
-		  __print_symbolic(__entry->type, XFS_ALLOC_TYPES),
-		  __print_symbolic(__entry->otype, XFS_ALLOC_TYPES),
 		  __entry->wasdel,
 		  __entry->wasfromfl,
 		  __entry->resv,
 		  __entry->datatype,
-		  (unsigned long long)__entry->firstblock)
+		  __entry->highest_agno)
 )
 
 #define DEFINE_ALLOC_EVENT(name) \
@@ -1877,6 +1877,7 @@ DEFINE_ALLOC_EVENT(xfs_alloc_small_notenough);
 DEFINE_ALLOC_EVENT(xfs_alloc_small_done);
 DEFINE_ALLOC_EVENT(xfs_alloc_small_error);
 DEFINE_ALLOC_EVENT(xfs_alloc_vextent_badargs);
+DEFINE_ALLOC_EVENT(xfs_alloc_vextent_skip_deadlock);
 DEFINE_ALLOC_EVENT(xfs_alloc_vextent_nofix);
 DEFINE_ALLOC_EVENT(xfs_alloc_vextent_noagbp);
 DEFINE_ALLOC_EVENT(xfs_alloc_vextent_loopfailed);

fs/xfs/xfs_trans.c

@@ -102,7 +102,7 @@ xfs_trans_dup(
 	INIT_LIST_HEAD(&ntp->t_items);
 	INIT_LIST_HEAD(&ntp->t_busy);
 	INIT_LIST_HEAD(&ntp->t_dfops);
-	ntp->t_firstblock = NULLFSBLOCK;
+	ntp->t_highest_agno = NULLAGNUMBER;
 
 	ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
 	ASSERT(tp->t_ticket != NULL);
@@ -278,7 +278,7 @@ xfs_trans_alloc(
 	INIT_LIST_HEAD(&tp->t_items);
 	INIT_LIST_HEAD(&tp->t_busy);
 	INIT_LIST_HEAD(&tp->t_dfops);
-	tp->t_firstblock = NULLFSBLOCK;
+	tp->t_highest_agno = NULLAGNUMBER;
 
 	error = xfs_trans_reserve(tp, resp, blocks, rtextents);
 	if (error == -ENOSPC && want_retry) {
@@ -1078,10 +1078,10 @@ xfs_trans_cancel(
 	/*
 	 * It's never valid to cancel a transaction with deferred ops attached,
 	 * because the transaction is effectively dirty.  Complain about this
-	 * loudly before freeing the in-memory defer items.
+	 * loudly before freeing the in-memory defer items and shutting down the
+	 * filesystem.
 	 */
 	if (!list_empty(&tp->t_dfops)) {
-		ASSERT(xfs_is_shutdown(mp) || list_empty(&tp->t_dfops));
 		ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES);
 		dirty = true;
 		xfs_defer_cancel(tp);

fs/xfs/xfs_trans.h

@@ -132,7 +132,7 @@ typedef struct xfs_trans {
 	unsigned int		t_rtx_res;	/* # of rt extents resvd */
 	unsigned int		t_rtx_res_used;	/* # of resvd rt extents used */
 	unsigned int		t_flags;	/* misc flags */
-	xfs_fsblock_t		t_firstblock;	/* first block allocated */
+	xfs_agnumber_t		t_highest_agno;	/* highest AGF locked */
 	struct xlog_ticket	*t_ticket;	/* log mgr ticket */
 	struct xfs_mount	*t_mountp;	/* ptr to fs mount struct */
 	struct xfs_dquot_acct   *t_dqinfo;	/* acctg info for dquots */