xfs: garbage collect old cowextsz reservations

Trim CoW reservations made on behalf of a cowextsz hint if they get too
old or we run low on quota, so long as we don't have dirty data awaiting
writeback or directio operations in progress.

Garbage collection of the cowextsize extents are kept separate from
prealloc extent reaping because setting the CoW prealloc lifetime to a
(much) higher value than the regular prealloc extent lifetime has been
useful for combatting CoW fragmentation on VM hosts where the VMs
experience bursty write behaviors and we can keep the utilization ratios
low enough that we don't start to run out of space.  IOWs, it benefits
us to keep the CoW fork reservations around for as long as we can unless
we run out of blocks or hit inode reclaim.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

fs/xfs/xfs_bmap_util.c

@@ -1891,6 +1891,8 @@ xfs_swap_extents(
 		cowfp = ip->i_cowfp;
 		ip->i_cowfp = tip->i_cowfp;
 		tip->i_cowfp = cowfp;
+		xfs_inode_set_cowblocks_tag(ip);
+		xfs_inode_set_cowblocks_tag(tip);
 	}
 
 	xfs_trans_log_inode(tp, ip,  src_log_flags);

fs/xfs/xfs_file.c

@@ -782,6 +782,9 @@ xfs_file_buffered_aio_write(
 		enospc = xfs_inode_free_quota_eofblocks(ip);
 		if (enospc)
 			goto write_retry;
+		enospc = xfs_inode_free_quota_cowblocks(ip);
+		if (enospc)
+			goto write_retry;
 	} else if (ret == -ENOSPC && !enospc) {
 		struct xfs_eofblocks eofb = {0};
 

fs/xfs/xfs_globals.c

@@ -21,8 +21,8 @@
 /*
  * Tunable XFS parameters.  xfs_params is required even when CONFIG_SYSCTL=n,
  * other XFS code uses these values.  Times are measured in centisecs (i.e.
- * 100ths of a second) with the exception of eofb_timer, which is measured in
- * seconds.
+ * 100ths of a second) with the exception of eofb_timer and cowb_timer, which
+ * are measured in seconds.
  */
 xfs_param_t xfs_params = {
 			  /*	MIN		DFLT		MAX	*/
@@ -42,6 +42,7 @@ xfs_param_t xfs_params = {
 	.inherit_nodfrg	= {	0,		1,		1	},
 	.fstrm_timer	= {	1,		30*100,		3600*100},
 	.eofb_timer	= {	1,		300,		3600*24},
+	.cowb_timer	= {	1,		1800,		3600*24},
 };
 
 struct xfs_globals xfs_globals = {

fs/xfs/xfs_icache.c

@@ -33,6 +33,7 @@
 #include "xfs_bmap_util.h"
 #include "xfs_dquot_item.h"
 #include "xfs_dquot.h"
+#include "xfs_reflink.h"
 
 #include <linux/kthread.h>
 #include <linux/freezer.h>
@@ -792,6 +793,33 @@ xfs_eofblocks_worker(
 	xfs_queue_eofblocks(mp);
 }
 
+/*
+ * Background scanning to trim preallocated CoW space. This is queued
+ * based on the 'speculative_cow_prealloc_lifetime' tunable (5m by default).
+ * (We'll just piggyback on the post-EOF prealloc space workqueue.)
+ */
+STATIC void
+xfs_queue_cowblocks(
+	struct xfs_mount *mp)
+{
+	rcu_read_lock();
+	if (radix_tree_tagged(&mp->m_perag_tree, XFS_ICI_COWBLOCKS_TAG))
+		queue_delayed_work(mp->m_eofblocks_workqueue,
+				   &mp->m_cowblocks_work,
+				   msecs_to_jiffies(xfs_cowb_secs * 1000));
+	rcu_read_unlock();
+}
+
+void
+xfs_cowblocks_worker(
+	struct work_struct *work)
+{
+	struct xfs_mount *mp = container_of(to_delayed_work(work),
+				struct xfs_mount, m_cowblocks_work);
+	xfs_icache_free_cowblocks(mp, NULL);
+	xfs_queue_cowblocks(mp);
+}
+
 int
 xfs_inode_ag_iterator(
 	struct xfs_mount	*mp,
@@ -1348,18 +1376,30 @@ xfs_inode_free_eofblocks(
 	return ret;
 }
 
-int
-xfs_icache_free_eofblocks(
+static int
+__xfs_icache_free_eofblocks(
 	struct xfs_mount	*mp,
-	struct xfs_eofblocks	*eofb)
+	struct xfs_eofblocks	*eofb,
+	int			(*execute)(struct xfs_inode *ip, int flags,
+					   void *args),
+	int			tag)
 {
 	int flags = SYNC_TRYLOCK;
 
 	if (eofb && (eofb->eof_flags & XFS_EOF_FLAGS_SYNC))
 		flags = SYNC_WAIT;
 
-	return xfs_inode_ag_iterator_tag(mp, xfs_inode_free_eofblocks, flags,
-					 eofb, XFS_ICI_EOFBLOCKS_TAG);
+	return xfs_inode_ag_iterator_tag(mp, execute, flags,
+					 eofb, tag);
+}
+
+int
+xfs_icache_free_eofblocks(
+	struct xfs_mount	*mp,
+	struct xfs_eofblocks	*eofb)
+{
+	return __xfs_icache_free_eofblocks(mp, eofb, xfs_inode_free_eofblocks,
+			XFS_ICI_EOFBLOCKS_TAG);
 }
 
 /*
@@ -1368,9 +1408,11 @@ xfs_icache_free_eofblocks(
  * failure. We make a best effort by including each quota under low free space
  * conditions (less than 1% free space) in the scan.
  */
-int
-xfs_inode_free_quota_eofblocks(
-	struct xfs_inode *ip)
+static int
+__xfs_inode_free_quota_eofblocks(
+	struct xfs_inode	*ip,
+	int			(*execute)(struct xfs_mount *mp,
+					   struct xfs_eofblocks	*eofb))
 {
 	int scan = 0;
 	struct xfs_eofblocks eofb = {0};
@@ -1406,14 +1448,25 @@ xfs_inode_free_quota_eofblocks(
 	}
 
 	if (scan)
-		xfs_icache_free_eofblocks(ip->i_mount, &eofb);
+		execute(ip->i_mount, &eofb);
 
 	return scan;
 }
 
-void
-xfs_inode_set_eofblocks_tag(
-	xfs_inode_t	*ip)
+int
+xfs_inode_free_quota_eofblocks(
+	struct xfs_inode *ip)
+{
+	return __xfs_inode_free_quota_eofblocks(ip, xfs_icache_free_eofblocks);
+}
+
+static void
+__xfs_inode_set_eofblocks_tag(
+	xfs_inode_t	*ip,
+	void		(*execute)(struct xfs_mount *mp),
+	void		(*set_tp)(struct xfs_mount *mp, xfs_agnumber_t agno,
+				  int error, unsigned long caller_ip),
+	int		tag)
 {
 	struct xfs_mount *mp = ip->i_mount;
 	struct xfs_perag *pag;
@@ -1431,35 +1484,44 @@ xfs_inode_set_eofblocks_tag(
 
 	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
 	spin_lock(&pag->pag_ici_lock);
-	trace_xfs_inode_set_eofblocks_tag(ip);
 
-	tagged = radix_tree_tagged(&pag->pag_ici_root,
-				   XFS_ICI_EOFBLOCKS_TAG);
+	tagged = radix_tree_tagged(&pag->pag_ici_root, tag);
 	radix_tree_tag_set(&pag->pag_ici_root,
-			   XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino),
-			   XFS_ICI_EOFBLOCKS_TAG);
+			   XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino), tag);
 	if (!tagged) {
 		/* propagate the eofblocks tag up into the perag radix tree */
 		spin_lock(&ip->i_mount->m_perag_lock);
 		radix_tree_tag_set(&ip->i_mount->m_perag_tree,
 				   XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino),
-				   XFS_ICI_EOFBLOCKS_TAG);
+				   tag);
 		spin_unlock(&ip->i_mount->m_perag_lock);
 
 		/* kick off background trimming */
-		xfs_queue_eofblocks(ip->i_mount);
+		execute(ip->i_mount);
 
-		trace_xfs_perag_set_eofblocks(ip->i_mount, pag->pag_agno,
-					      -1, _RET_IP_);
+		set_tp(ip->i_mount, pag->pag_agno, -1, _RET_IP_);
 	}
 
 	spin_unlock(&pag->pag_ici_lock);
 	xfs_perag_put(pag);
 }
 
 void
-xfs_inode_clear_eofblocks_tag(
+xfs_inode_set_eofblocks_tag(
 	xfs_inode_t	*ip)
+{
+	trace_xfs_inode_set_eofblocks_tag(ip);
+	return __xfs_inode_set_eofblocks_tag(ip, xfs_queue_eofblocks,
+			trace_xfs_perag_set_eofblocks,
+			XFS_ICI_EOFBLOCKS_TAG);
+}
+
+static void
+__xfs_inode_clear_eofblocks_tag(
+	xfs_inode_t	*ip,
+	void		(*clear_tp)(struct xfs_mount *mp, xfs_agnumber_t agno,
+				    int error, unsigned long caller_ip),
+	int		tag)
 {
 	struct xfs_mount *mp = ip->i_mount;
 	struct xfs_perag *pag;
@@ -1470,23 +1532,141 @@ xfs_inode_clear_eofblocks_tag(
 
 	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
 	spin_lock(&pag->pag_ici_lock);
-	trace_xfs_inode_clear_eofblocks_tag(ip);
 
 	radix_tree_tag_clear(&pag->pag_ici_root,
-			     XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino),
-			     XFS_ICI_EOFBLOCKS_TAG);
-	if (!radix_tree_tagged(&pag->pag_ici_root, XFS_ICI_EOFBLOCKS_TAG)) {
+			     XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino), tag);
+	if (!radix_tree_tagged(&pag->pag_ici_root, tag)) {
 		/* clear the eofblocks tag from the perag radix tree */
 		spin_lock(&ip->i_mount->m_perag_lock);
 		radix_tree_tag_clear(&ip->i_mount->m_perag_tree,
 				     XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino),
-				     XFS_ICI_EOFBLOCKS_TAG);
+				     tag);
 		spin_unlock(&ip->i_mount->m_perag_lock);
-		trace_xfs_perag_clear_eofblocks(ip->i_mount, pag->pag_agno,
-					       -1, _RET_IP_);
+		clear_tp(ip->i_mount, pag->pag_agno, -1, _RET_IP_);
 	}
 
 	spin_unlock(&pag->pag_ici_lock);
 	xfs_perag_put(pag);
 }
 
+void
+xfs_inode_clear_eofblocks_tag(
+	xfs_inode_t	*ip)
+{
+	trace_xfs_inode_clear_eofblocks_tag(ip);
+	return __xfs_inode_clear_eofblocks_tag(ip,
+			trace_xfs_perag_clear_eofblocks, XFS_ICI_EOFBLOCKS_TAG);
+}
+
+/*
+ * Automatic CoW Reservation Freeing
+ *
+ * These functions automatically garbage collect leftover CoW reservations
+ * that were made on behalf of a cowextsize hint when we start to run out
+ * of quota or when the reservations sit around for too long.  If the file
+ * has dirty pages or is undergoing writeback, its CoW reservations will
+ * be retained.
+ *
+ * The actual garbage collection piggybacks off the same code that runs
+ * the speculative EOF preallocation garbage collector.
+ */
+STATIC int
+xfs_inode_free_cowblocks(
+	struct xfs_inode	*ip,
+	int			flags,
+	void			*args)
+{
+	int ret;
+	struct xfs_eofblocks *eofb = args;
+	bool need_iolock = true;
+	int match;
+
+	ASSERT(!eofb || (eofb && eofb->eof_scan_owner != 0));
+
+	if (!xfs_reflink_has_real_cow_blocks(ip)) {
+		trace_xfs_inode_free_cowblocks_invalid(ip);
+		xfs_inode_clear_cowblocks_tag(ip);
+		return 0;
+	}
+
+	/*
+	 * If the mapping is dirty or under writeback we cannot touch the
+	 * CoW fork.  Leave it alone if we're in the midst of a directio.
+	 */
+	if (mapping_tagged(VFS_I(ip)->i_mapping, PAGECACHE_TAG_DIRTY) ||
+	    mapping_tagged(VFS_I(ip)->i_mapping, PAGECACHE_TAG_WRITEBACK) ||
+	    atomic_read(&VFS_I(ip)->i_dio_count))
+		return 0;
+
+	if (eofb) {
+		if (eofb->eof_flags & XFS_EOF_FLAGS_UNION)
+			match = xfs_inode_match_id_union(ip, eofb);
+		else
+			match = xfs_inode_match_id(ip, eofb);
+		if (!match)
+			return 0;
+
+		/* skip the inode if the file size is too small */
+		if (eofb->eof_flags & XFS_EOF_FLAGS_MINFILESIZE &&
+		    XFS_ISIZE(ip) < eofb->eof_min_file_size)
+			return 0;
+
+		/*
+		 * A scan owner implies we already hold the iolock. Skip it in
+		 * xfs_free_eofblocks() to avoid deadlock. This also eliminates
+		 * the possibility of EAGAIN being returned.
+		 */
+		if (eofb->eof_scan_owner == ip->i_ino)
+			need_iolock = false;
+	}
+
+	/* Free the CoW blocks */
+	if (need_iolock) {
+		xfs_ilock(ip, XFS_IOLOCK_EXCL);
+		xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
+	}
+
+	ret = xfs_reflink_cancel_cow_range(ip, 0, NULLFILEOFF);
+
+	if (need_iolock) {
+		xfs_iunlock(ip, XFS_MMAPLOCK_EXCL);
+		xfs_iunlock(ip, XFS_IOLOCK_EXCL);
+	}
+
+	return ret;
+}
+
+int
+xfs_icache_free_cowblocks(
+	struct xfs_mount	*mp,
+	struct xfs_eofblocks	*eofb)
+{
+	return __xfs_icache_free_eofblocks(mp, eofb, xfs_inode_free_cowblocks,
+			XFS_ICI_COWBLOCKS_TAG);
+}
+
+int
+xfs_inode_free_quota_cowblocks(
+	struct xfs_inode *ip)
+{
+	return __xfs_inode_free_quota_eofblocks(ip, xfs_icache_free_cowblocks);
+}
+
+void
+xfs_inode_set_cowblocks_tag(
+	xfs_inode_t	*ip)
+{
+	trace_xfs_inode_set_eofblocks_tag(ip);
+	return __xfs_inode_set_eofblocks_tag(ip, xfs_queue_cowblocks,
+			trace_xfs_perag_set_eofblocks,
+			XFS_ICI_COWBLOCKS_TAG);
+}
+
+void
+xfs_inode_clear_cowblocks_tag(
+	xfs_inode_t	*ip)
+{
+	trace_xfs_inode_clear_eofblocks_tag(ip);
+	return __xfs_inode_clear_eofblocks_tag(ip,
+			trace_xfs_perag_clear_eofblocks, XFS_ICI_COWBLOCKS_TAG);
+}

fs/xfs/xfs_icache.h

@@ -40,6 +40,7 @@ struct xfs_eofblocks {
 					   in xfs_inode_ag_iterator */
 #define XFS_ICI_RECLAIM_TAG	0	/* inode is to be reclaimed */
 #define XFS_ICI_EOFBLOCKS_TAG	1	/* inode has blocks beyond EOF */
+#define XFS_ICI_COWBLOCKS_TAG	2	/* inode can have cow blocks to gc */
 
 /*
  * Flags for xfs_iget()
@@ -70,6 +71,12 @@ int xfs_inode_free_quota_eofblocks(struct xfs_inode *ip);
 void xfs_eofblocks_worker(struct work_struct *);
 void xfs_queue_eofblocks(struct xfs_mount *);
 
+void xfs_inode_set_cowblocks_tag(struct xfs_inode *ip);
+void xfs_inode_clear_cowblocks_tag(struct xfs_inode *ip);
+int xfs_icache_free_cowblocks(struct xfs_mount *, struct xfs_eofblocks *);
+int xfs_inode_free_quota_cowblocks(struct xfs_inode *ip);
+void xfs_cowblocks_worker(struct work_struct *);
+
 int xfs_inode_ag_iterator(struct xfs_mount *mp,
 	int (*execute)(struct xfs_inode *ip, int flags, void *args),
 	int flags, void *args);

fs/xfs/xfs_inode.c

@@ -1629,8 +1629,10 @@ xfs_itruncate_extents(
 	/*
 	 * Clear the reflink flag if we truncated everything.
 	 */
-	if (ip->i_d.di_nblocks == 0 && xfs_is_reflink_inode(ip))
+	if (ip->i_d.di_nblocks == 0 && xfs_is_reflink_inode(ip)) {
 		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
+		xfs_inode_clear_cowblocks_tag(ip);
+	}
 
 	/*
 	 * Always re-log the inode so that our permanent transaction can keep