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 <[email protected]>
Reviewed-by: Darrick J. Wong <[email protected]>
Signed-off-by: Darrick J. Wong <[email protected]>

Author: Christoph Hellwig

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);

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 */

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 @@ xfs_alloc_ag_vextent_near(
 			 * 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 @@ xfs_alloc_ag_vextent_size(
 			 * 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 @@ xfs_alloc_ag_vextent_size(
 			 * 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:

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;
 }

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;
 }

fs/xfs/libxfs/xfs_group.h

@@ -15,6 +15,11 @@ struct xfs_group {
 #ifdef __KERNEL__
 	/* -- 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.

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);

fs/xfs/scrub/alloc_repair.c

@@ -132,13 +132,16 @@ int
 xrep_setup_ag_allocbt(
 	struct xfs_scrub	*sc)
 {
+	struct xfs_group	*xg = pag_group(sc->sa.pag);
+	unsigned int		busy_gen;
+
 	/*
 	 * Make sure the busy extent list is clear because we can't put extents
 	 * on there twice.
 	 */
-	if (xfs_extent_busy_list_empty(sc->sa.pag, &busy_gen))
+	if (xfs_extent_busy_list_empty(xg, &busy_gen))
 		return 0;
-	return xfs_extent_busy_flush(sc->tp, sc->sa.pag, busy_gen, 0);
+	return xfs_extent_busy_flush(sc->tp, xg, busy_gen, 0);
 }
 
 /* Check for any obvious conflicts in the free extent. */
@@ -866,7 +869,7 @@ xrep_allocbt(
 	 * on there twice.  In theory we cleared this before we started, but
 	 * let's not risk the filesystem.
 	 */
-	if (!xfs_extent_busy_list_empty(sc->sa.pag, &busy_gen)) {
+	if (!xfs_extent_busy_list_empty(pag_group(sc->sa.pag), &busy_gen)) {
 		error = -EDEADLOCK;
 		goto out_ra;
 	}

fs/xfs/scrub/reap.c

@@ -137,7 +137,7 @@ xreap_put_freelist(
 			agfl_bp, agbno, 0);
 	if (error)
 		return error;
-	xfs_extent_busy_insert(sc->tp, sc->sa.pag, agbno, 1,
+	xfs_extent_busy_insert(sc->tp, pag_group(sc->sa.pag), agbno, 1,
 			XFS_EXTENT_BUSY_SKIP_DISCARD);
 
 	return 0;

fs/xfs/xfs_discard.c

@@ -117,10 +117,12 @@ xfs_discard_extents(
 
 	blk_start_plug(&plug);
 	list_for_each_entry(busyp, &extents->extent_list, list) {
-		trace_xfs_discard_extent(busyp->pag, busyp->bno, busyp->length);
+		struct xfs_perag	*pag = to_perag(busyp->group);
+
+		trace_xfs_discard_extent(pag, busyp->bno, busyp->length);
 
 		error = __blkdev_issue_discard(mp->m_ddev_targp->bt_bdev,
-				xfs_agbno_to_daddr(busyp->pag, busyp->bno),
+				xfs_agbno_to_daddr(pag, busyp->bno),
 				XFS_FSB_TO_BB(mp, busyp->length),
 				GFP_KERNEL, &bio);
 		if (error && error != -EOPNOTSUPP) {
@@ -271,12 +273,12 @@ xfs_trim_gather_extents(
 		 * If any blocks in the range are still busy, skip the
 		 * discard and try again the next time.
 		 */
-		if (xfs_extent_busy_search(pag, fbno, flen)) {
+		if (xfs_extent_busy_search(pag_group(pag), fbno, flen)) {
 			trace_xfs_discard_busy(pag, fbno, flen);
 			goto next_extent;
 		}
 
-		xfs_extent_busy_insert_discard(pag, fbno, flen,
+		xfs_extent_busy_insert_discard(pag_group(pag), fbno, flen,
 				&extents->extent_list);
 next_extent:
 		if (tcur->by_bno)