xfs: convert busy extent tracking to the generic group structure

Split busy extent tracking from struct xfs_perag into its own private
structure, which can be pointed to by the generic group structure.

Note that this structure is now dynamically allocated instead of embedded
as the upcoming zone XFS code doesn't need it and will also have an
unusually high number of groups due to hardware constraints.  Dynamically
allocating the structure this is a big memory saver for this case.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>

7 files changed, 40 insertions, 26 deletions

diff --git a/fs/xfs/libxfs/xfs_ag.c b/fs/xfs/libxfs/xfs_ag.c
index e60469fee875..47e90dbb852b 100644
--- a/fs/xfs/libxfs/xfs_ag.c
+++ b/fs/xfs/libxfs/xfs_ag.c
@@ -230,11 +230,8 @@ xfs_perag_alloc(
 #ifdef __KERNEL__
 	/* Place kernel structure only init below this point. */
 	spin_lock_init(&pag->pag_ici_lock);
-	spin_lock_init(&pag->pagb_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_tree = RB_ROOT;
 #endif /* __KERNEL__ */
 
 	error = xfs_buf_cache_init(&pag->pag_bcache);
diff --git a/fs/xfs/libxfs/xfs_ag.h b/fs/xfs/libxfs/xfs_ag.h
index 042ee0913fb9..7290148fa6e6 100644
--- a/fs/xfs/libxfs/xfs_ag.h
+++ b/fs/xfs/libxfs/xfs_ag.h
@@ -80,11 +80,6 @@ struct xfs_perag {
 	uint8_t		pagf_repair_rmap_level;
 #endif
 
-	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 */
diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
index bfe7b4321c47..4ddd05c97a29 100644
--- a/fs/xfs/libxfs/xfs_alloc.c
+++ b/fs/xfs/libxfs/xfs_alloc.c
@@ -331,7 +331,8 @@ xfs_alloc_compute_aligned(
 	bool		busy;
 
 	/* Trim busy sections out of found extent */
-	busy = xfs_extent_busy_trim(args, &bno, &len, busy_gen);
+	busy = xfs_extent_busy_trim(pag_group(args->pag), args->minlen,
+			args->maxlen, &bno, &len, busy_gen);
 
 	/*
 	 * If we have a largish extent that happens to start before min_agbno,
@@ -1251,7 +1252,7 @@ xfs_alloc_ag_vextent_small(
 	if (fbno == NULLAGBLOCK)
 		goto out;
 
-	xfs_extent_busy_reuse(args->pag, fbno, 1,
+	xfs_extent_busy_reuse(pag_group(args->pag), fbno, 1,
 			      (args->datatype & XFS_ALLOC_NOBUSY));
 
 	if (args->datatype & XFS_ALLOC_USERDATA) {
@@ -1364,7 +1365,8 @@ xfs_alloc_ag_vextent_exact(
 	 */
 	tbno = fbno;
 	tlen = flen;
-	xfs_extent_busy_trim(args, &tbno, &tlen, &busy_gen);
+	xfs_extent_busy_trim(pag_group(args->pag), args->minlen, args->maxlen,
+			&tbno, &tlen, &busy_gen);
 
 	/*
 	 * Give up if the start of the extent is busy, or the freespace isn't
@@ -1757,8 +1759,9 @@ restart:
 			 * the allocation can be retried.
 			 */
 			trace_xfs_alloc_near_busy(args);
-			error = xfs_extent_busy_flush(args->tp, args->pag,
-					acur.busy_gen, alloc_flags);
+			error = xfs_extent_busy_flush(args->tp,
+					pag_group(args->pag), acur.busy_gen,
+					alloc_flags);
 			if (error)
 				goto out;
 
@@ -1873,8 +1876,9 @@ restart:
 			 * the allocation can be retried.
 			 */
 			trace_xfs_alloc_size_busy(args);
-			error = xfs_extent_busy_flush(args->tp, args->pag,
-					busy_gen, alloc_flags);
+			error = xfs_extent_busy_flush(args->tp,
+					pag_group(args->pag), busy_gen,
+					alloc_flags);
 			if (error)
 				goto error0;
 
@@ -1972,8 +1976,9 @@ restart:
 			 * the allocation can be retried.
 			 */
 			trace_xfs_alloc_size_busy(args);
-			error = xfs_extent_busy_flush(args->tp, args->pag,
-					busy_gen, alloc_flags);
+			error = xfs_extent_busy_flush(args->tp,
+					pag_group(args->pag), busy_gen,
+					alloc_flags);
 			if (error)
 				goto error0;
 
@@ -3615,8 +3620,8 @@ xfs_alloc_vextent_finish(
 		if (error)
 			goto out_drop_perag;
 
-		ASSERT(!xfs_extent_busy_search(args->pag, args->agbno,
-				args->len));
+		ASSERT(!xfs_extent_busy_search(pag_group(args->pag),
+				args->agbno, args->len));
 	}
 
 	xfs_ag_resv_alloc_extent(args->pag, args->resv, args);
@@ -4014,7 +4019,7 @@ __xfs_free_extent(
 
 	if (skip_discard)
 		busy_flags |= XFS_EXTENT_BUSY_SKIP_DISCARD;
-	xfs_extent_busy_insert(tp, pag, agbno, len, busy_flags);
+	xfs_extent_busy_insert(tp, pag_group(pag), agbno, len, busy_flags);
 	return 0;
 
 err_release:
diff --git a/fs/xfs/libxfs/xfs_alloc_btree.c b/fs/xfs/libxfs/xfs_alloc_btree.c
index 88e1545ed4c9..e69a1bb13f7f 100644
--- a/fs/xfs/libxfs/xfs_alloc_btree.c
+++ b/fs/xfs/libxfs/xfs_alloc_btree.c
@@ -86,7 +86,7 @@ xfs_allocbt_alloc_block(
 	}
 
 	atomic64_inc(&cur->bc_mp->m_allocbt_blks);
-	xfs_extent_busy_reuse(cur->bc_ag.pag, bno, 1, false);
+	xfs_extent_busy_reuse(pag_group(cur->bc_ag.pag), bno, 1, false);
 
 	new->s = cpu_to_be32(bno);
 
@@ -110,7 +110,7 @@ xfs_allocbt_free_block(
 		return error;
 
 	atomic64_dec(&cur->bc_mp->m_allocbt_blks);
-	xfs_extent_busy_insert(cur->bc_tp, agbp->b_pag, bno, 1,
+	xfs_extent_busy_insert(cur->bc_tp, pag_group(agbp->b_pag), bno, 1,
 			      XFS_EXTENT_BUSY_SKIP_DISCARD);
 	return 0;
 }
diff --git a/fs/xfs/libxfs/xfs_group.c b/fs/xfs/libxfs/xfs_group.c
index 8532dc2f8628..5c6fa5d76a91 100644
--- a/fs/xfs/libxfs/xfs_group.c
+++ b/fs/xfs/libxfs/xfs_group.c
@@ -10,6 +10,7 @@
 #include "xfs_mount.h"
 #include "xfs_error.h"
 #include "xfs_trace.h"
+#include "xfs_extent_busy.h"
 #include "xfs_group.h"
 
 /*
@@ -161,6 +162,9 @@ xfs_group_free(
 	XFS_IS_CORRUPT(mp, atomic_read(&xg->xg_ref) != 0);
 
 	xfs_defer_drain_free(&xg->xg_intents_drain);
+#ifdef __KERNEL__
+	kfree(xg->xg_busy_extents);
+#endif
 
 	if (uninit)
 		uninit(xg);
@@ -185,6 +189,9 @@ xfs_group_insert(
 	xg->xg_type = type;
 
 #ifdef __KERNEL__
+	xg->xg_busy_extents = xfs_extent_busy_alloc();
+	if (!xg->xg_busy_extents)
+		return -ENOMEM;
 	spin_lock_init(&xg->xg_state_lock);
 	xfs_hooks_init(&xg->xg_rmap_update_hooks);
 #endif
@@ -196,9 +203,14 @@ xfs_group_insert(
 	error = xa_insert(&mp->m_groups[type].xa, index, xg, GFP_KERNEL);
 	if (error) {
 		WARN_ON_ONCE(error == -EBUSY);
-		xfs_defer_drain_free(&xg->xg_intents_drain);
-		return error;
+		goto out_drain;
 	}
 
 	return 0;
+out_drain:
+	xfs_defer_drain_free(&xg->xg_intents_drain);
+#ifdef __KERNEL__
+	kfree(xg->xg_busy_extents);
+#endif
+	return error;
 }
diff --git a/fs/xfs/libxfs/xfs_group.h b/fs/xfs/libxfs/xfs_group.h
index a87b9b80ef75..0ff6e1d5635c 100644
--- a/fs/xfs/libxfs/xfs_group.h
+++ b/fs/xfs/libxfs/xfs_group.h
@@ -16,6 +16,11 @@ struct xfs_group {
 	/* -- kernel only structures below this line -- */
 
 	/*
+	 * Track freed but not yet committed extents.
+	 */
+	struct xfs_extent_busy_tree *xg_busy_extents;
+
+	/*
 	 * Bitsets of per-ag metadata that have been checked and/or are sick.
 	 * Callers should hold xg_state_lock before accessing this field.
 	 */
diff --git a/fs/xfs/libxfs/xfs_rmap_btree.c b/fs/xfs/libxfs/xfs_rmap_btree.c
index 6fd460fc7c9c..b37eaf37c7fd 100644
--- a/fs/xfs/libxfs/xfs_rmap_btree.c
+++ b/fs/xfs/libxfs/xfs_rmap_btree.c
@@ -102,7 +102,7 @@ xfs_rmapbt_alloc_block(
 		return 0;
 	}
 
-	xfs_extent_busy_reuse(pag, bno, 1, false);
+	xfs_extent_busy_reuse(pag_group(pag), bno, 1, false);
 
 	new->s = cpu_to_be32(bno);
 	be32_add_cpu(&agf->agf_rmap_blocks, 1);
@@ -136,7 +136,7 @@ xfs_rmapbt_free_block(
 	if (error)
 		return error;
 
-	xfs_extent_busy_insert(cur->bc_tp, pag, bno, 1,
+	xfs_extent_busy_insert(cur->bc_tp, pag_group(pag), bno, 1,
 			      XFS_EXTENT_BUSY_SKIP_DISCARD);
 
 	xfs_ag_resv_free_extent(pag, XFS_AG_RESV_RMAPBT, NULL, 1);