Linux-XFS Archive on lore.kernel.org
 help / color / Atom feed
From: "Darrick J. Wong" <djwong@kernel.org>
To: Christoph Hellwig <hch@lst.de>,
	linux-xfs@vger.kernel.org, hch@infradead.org,
	david@fromorbit.com
Subject: [PATCH v4.1 9/9] xfs: parallelize block preallocation garbage collection
Date: Mon, 25 Jan 2021 21:12:26 -0800
Message-ID: <20210126051226.GU7698@magnolia> (raw)
In-Reply-To: <161142805211.2173480.18150564632716126076.stgit@magnolia>

From: Darrick J. Wong <djwong@kernel.org>

Split the block preallocation garbage collection work into per-AG work
items so that we can take advantage of parallelization.

Note that sysadmins /can/ tweak the max concurrency level of the blockgc
workqueue via /sys/bus/workqueue/devices/xfs-conv!${device}/max_active.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
v4.1: rebase due to earlier WQ_* changes
---
 fs/xfs/xfs_icache.c |   42 ++++++++++++++++++++++++++++++------------
 fs/xfs/xfs_mount.c  |    3 +++
 fs/xfs/xfs_mount.h  |    5 +++--
 fs/xfs/xfs_super.c  |    4 ++--
 4 files changed, 38 insertions(+), 16 deletions(-)

diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
index fbd59c702f23..d1d9cd4c8ad4 100644
--- a/fs/xfs/xfs_icache.c
+++ b/fs/xfs/xfs_icache.c
@@ -1333,12 +1333,12 @@ xfs_inode_free_eofblocks(
  */
 static inline void
 xfs_blockgc_queue(
-	struct xfs_mount	*mp)
+	struct xfs_perag	*pag)
 {
 	rcu_read_lock();
-	if (radix_tree_tagged(&mp->m_perag_tree, XFS_ICI_BLOCKGC_TAG))
-		queue_delayed_work(mp->m_blockgc_workqueue,
-				   &mp->m_blockgc_work,
+	if (radix_tree_tagged(&pag->pag_ici_root, XFS_ICI_BLOCKGC_TAG))
+		queue_delayed_work(pag->pag_mount->m_blockgc_workqueue,
+				   &pag->pag_blockgc_work,
 				   msecs_to_jiffies(xfs_blockgc_secs * 1000));
 	rcu_read_unlock();
 }
@@ -1380,7 +1380,7 @@ xfs_blockgc_set_iflag(
 		spin_unlock(&ip->i_mount->m_perag_lock);
 
 		/* kick off background trimming */
-		xfs_blockgc_queue(ip->i_mount);
+		xfs_blockgc_queue(pag);
 
 		trace_xfs_perag_set_blockgc(ip->i_mount, pag->pag_agno, -1,
 				_RET_IP_);
@@ -1555,12 +1555,24 @@ xfs_inode_clear_cowblocks_tag(
 	return xfs_blockgc_clear_iflag(ip, XFS_ICOWBLOCKS);
 }
 
+#define for_each_perag_tag(mp, next_agno, pag, tag) \
+	for ((next_agno) = 0, (pag) = xfs_perag_get_tag((mp), 0, (tag)); \
+		(pag) != NULL; \
+		(next_agno) = (pag)->pag_agno + 1, \
+		xfs_perag_put(pag), \
+		(pag) = xfs_perag_get_tag((mp), (next_agno), (tag)))
+
+
 /* Disable post-EOF and CoW block auto-reclamation. */
 void
 xfs_blockgc_stop(
 	struct xfs_mount	*mp)
 {
-	cancel_delayed_work_sync(&mp->m_blockgc_work);
+	struct xfs_perag	*pag;
+	xfs_agnumber_t		agno;
+
+	for_each_perag_tag(mp, agno, pag, XFS_ICI_BLOCKGC_TAG)
+		cancel_delayed_work_sync(&pag->pag_blockgc_work);
 }
 
 /* Enable post-EOF and CoW block auto-reclamation. */
@@ -1568,7 +1580,11 @@ void
 xfs_blockgc_start(
 	struct xfs_mount	*mp)
 {
-	xfs_blockgc_queue(mp);
+	struct xfs_perag	*pag;
+	xfs_agnumber_t		agno;
+
+	for_each_perag_tag(mp, agno, pag, XFS_ICI_BLOCKGC_TAG)
+		xfs_blockgc_queue(pag);
 }
 
 /* Scan one incore inode for block preallocations that we can remove. */
@@ -1595,18 +1611,20 @@ void
 xfs_blockgc_worker(
 	struct work_struct	*work)
 {
-	struct xfs_mount	*mp = container_of(to_delayed_work(work),
-					struct xfs_mount, m_blockgc_work);
+	struct xfs_perag	*pag = container_of(to_delayed_work(work),
+					struct xfs_perag, pag_blockgc_work);
+	struct xfs_mount	*mp = pag->pag_mount;
 	int			error;
 
 	if (!sb_start_write_trylock(mp->m_super))
 		return;
-	error = xfs_inode_walk(mp, 0, xfs_blockgc_scan_inode, NULL,
+	error = xfs_inode_walk_ag(pag, 0, xfs_blockgc_scan_inode, NULL,
 			XFS_ICI_BLOCKGC_TAG);
 	if (error)
-		xfs_info(mp, "preallocation gc worker failed, err=%d", error);
+		xfs_info(mp, "AG %u preallocation gc worker failed, err=%d",
+				pag->pag_agno, error);
 	sb_end_write(mp->m_super);
-	xfs_blockgc_queue(mp);
+	xfs_blockgc_queue(pag);
 }
 
 /*
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index be9ce114527f..52370d0a3f43 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -126,6 +126,7 @@ __xfs_free_perag(
 {
 	struct xfs_perag *pag = container_of(head, struct xfs_perag, rcu_head);
 
+	ASSERT(!delayed_work_pending(&pag->pag_blockgc_work));
 	ASSERT(atomic_read(&pag->pag_ref) == 0);
 	kmem_free(pag);
 }
@@ -146,6 +147,7 @@ xfs_free_perag(
 		spin_unlock(&mp->m_perag_lock);
 		ASSERT(pag);
 		ASSERT(atomic_read(&pag->pag_ref) == 0);
+		cancel_delayed_work_sync(&pag->pag_blockgc_work);
 		xfs_iunlink_destroy(pag);
 		xfs_buf_hash_destroy(pag);
 		call_rcu(&pag->rcu_head, __xfs_free_perag);
@@ -201,6 +203,7 @@ xfs_initialize_perag(
 		pag->pag_agno = index;
 		pag->pag_mount = mp;
 		spin_lock_init(&pag->pag_ici_lock);
+		INIT_DELAYED_WORK(&pag->pag_blockgc_work, xfs_blockgc_worker);
 		INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC);
 
 		error = xfs_buf_hash_init(pag);
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 316e0d79cc40..659ad95fe3e0 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -177,8 +177,6 @@ typedef struct xfs_mount {
 	uint64_t		m_resblks_avail;/* available reserved blocks */
 	uint64_t		m_resblks_save;	/* reserved blks @ remount,ro */
 	struct delayed_work	m_reclaim_work;	/* background inode reclaim */
-	struct delayed_work	m_blockgc_work; /* background prealloc blocks
-						     trimming */
 	struct xfs_kobj		m_kobj;
 	struct xfs_kobj		m_error_kobj;
 	struct xfs_kobj		m_error_meta_kobj;
@@ -367,6 +365,9 @@ typedef struct xfs_perag {
 	/* Blocks reserved for the reverse mapping btree. */
 	struct xfs_ag_resv	pag_rmapbt_resv;
 
+	/* background prealloc block trimming */
+	struct delayed_work	pag_blockgc_work;
+
 	/* reference count */
 	uint8_t			pagf_refcount_level;
 
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index ea942089d074..2b04818627e9 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -35,6 +35,7 @@
 #include "xfs_refcount_item.h"
 #include "xfs_bmap_item.h"
 #include "xfs_reflink.h"
+#include "xfs_pwork.h"
 
 #include <linux/magic.h>
 #include <linux/fs_context.h>
@@ -519,7 +520,7 @@ xfs_init_mount_workqueues(
 		goto out_destroy_cil;
 
 	mp->m_blockgc_workqueue = alloc_workqueue("xfs-blockgc/%s",
-			XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM),
+			XFS_WQFLAGS(WQ_UNBOUND | WQ_FREEZABLE | WQ_MEM_RECLAIM),
 			0, mp->m_super->s_id);
 	if (!mp->m_blockgc_workqueue)
 		goto out_destroy_reclaim;
@@ -1842,7 +1843,6 @@ static int xfs_init_fs_context(
 	mutex_init(&mp->m_growlock);
 	INIT_WORK(&mp->m_flush_inodes_work, xfs_flush_inodes_worker);
 	INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
-	INIT_DELAYED_WORK(&mp->m_blockgc_work, xfs_blockgc_worker);
 	mp->m_kobj.kobject.kset = xfs_kset;
 	/*
 	 * We don't create the finobt per-ag space reservation until after log

  reply index

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-23 18:53 [PATCHSET v4 0/9] xfs: consolidate posteof and cowblocks cleanup Darrick J. Wong
2021-01-23 18:53 ` [PATCH 1/9] xfs: relocate the eofb/cowb workqueue functions Darrick J. Wong
2021-01-23 18:53 ` [PATCH 2/9] xfs: hide xfs_icache_free_eofblocks Darrick J. Wong
2021-01-23 18:53 ` [PATCH 3/9] xfs: hide xfs_icache_free_cowblocks Darrick J. Wong
2021-01-23 18:53 ` [PATCH 4/9] xfs: remove trivial eof/cowblocks functions Darrick J. Wong
2021-01-23 18:53 ` [PATCH 5/9] xfs: consolidate incore inode radix tree posteof/cowblocks tags Darrick J. Wong
2021-01-23 18:53 ` [PATCH 6/9] xfs: consolidate the eofblocks and cowblocks workers Darrick J. Wong
2021-01-23 18:54 ` [PATCH 7/9] xfs: only walk the incore inode tree once per blockgc scan Darrick J. Wong
2021-01-23 18:54 ` [PATCH 8/9] xfs: rename block gc start and stop functions Darrick J. Wong
2021-01-23 18:54 ` [PATCH 9/9] xfs: parallelize block preallocation garbage collection Darrick J. Wong
2021-01-26  5:12   ` Darrick J. Wong [this message]
2021-01-26  5:13 ` [PATCH 10/9] xfs: expose the blockgc workqueue knobs publicly Darrick J. Wong
2021-01-27 17:03   ` Christoph Hellwig

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210126051226.GU7698@magnolia \
    --to=djwong@kernel.org \
    --cc=david@fromorbit.com \
    --cc=hch@infradead.org \
    --cc=hch@lst.de \
    --cc=linux-xfs@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-XFS Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-xfs/0 linux-xfs/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-xfs linux-xfs/ https://lore.kernel.org/linux-xfs \
		linux-xfs@vger.kernel.org
	public-inbox-index linux-xfs

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-xfs


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git