summarylogtreecommitdiffstats
path: root/0005-btrfs-shrink-delalloc-pages-instead-of-full-inodes.patch
blob: e06f23cc27edae742069aa6a978ef6fd1261c507 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
From 0397aa4e6205543d9b3da11794037e94f8735867 Mon Sep 17 00:00:00 2001
From: Josef Bacik <josef@toxicpanda.com>
Date: Mon, 4 Jan 2021 15:24:11 -0500
Subject: [PATCH 5/6] btrfs: shrink delalloc pages instead of full inodes
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Commit 38d715f494f2 ("btrfs: use btrfs_start_delalloc_roots in
shrink_delalloc") cleaned up how we do delalloc shrinking by utilizing
some infrastructure we have in place to flush inodes that we use for
device replace and snapshot.  However this introduced a pretty serious
performance regression.  The root cause is because before we would
generally use the normal writeback path to reclaim delalloc space, and
for this we would provide it with the number of pages we wanted to
flush.  The referenced commit changed this to flush that many inodes,
which drastically increased the amount of space we were flushing in
certain cases, which severely affected performance.

We cannot revert this patch unfortunately, because Filipe has another
fix that requires the ability to skip flushing inodes that are being
cloned in certain scenarios, which means we need to keep using our
flushing infrastructure or risk re-introducing the deadlock.

Instead to fix this problem we can go back to providing
btrfs_start_delalloc_roots with a number of pages to flush, and then set
up a writeback_control and utilize sync_inode() to handle the flushing
for us.  This gives us the same behavior we had prior to the fix, while
still allowing us to avoid the deadlock that was fixed by Filipe.  The
user reported reproducer was a simple untarring of a large tarball of
the source code for Firefox.  The results are as follows

5.9	0m54.258s
5.10	1m26.212s
Patch	0m38.800s

We are significantly faster because of the work I did around improving
ENOSPC flushing in 5.10 and 5.11, so reverting to the previous write out
behavior gave us a pretty big boost.

CC: stable@vger.kernel.org # 5.10
Reported-by: René Rebe <rene@exactcode.de>
Fixes: 38d715f494f2 ("btrfs: use btrfs_start_delalloc_roots in shrink_delalloc")
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
---
 fs/btrfs/inode.c      | 60 +++++++++++++++++++++++++++++++------------
 fs/btrfs/space-info.c |  4 ++-
 2 files changed, 46 insertions(+), 18 deletions(-)

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 07479250221d..acc47e2ffb46 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -9389,7 +9389,8 @@ static struct btrfs_delalloc_work *btrfs_alloc_delalloc_work(struct inode *inode
  * some fairly slow code that needs optimization. This walks the list
  * of all the inodes with pending delalloc and forces them to disk.
  */
-static int start_delalloc_inodes(struct btrfs_root *root, u64 *nr, bool snapshot,
+static int start_delalloc_inodes(struct btrfs_root *root,
+				 struct writeback_control *wbc, bool snapshot,
 				 bool in_reclaim_context)
 {
 	struct btrfs_inode *binode;
@@ -9398,6 +9399,7 @@ static int start_delalloc_inodes(struct btrfs_root *root, u64 *nr, bool snapshot
 	struct list_head works;
 	struct list_head splice;
 	int ret = 0;
+	bool full_flush = wbc->nr_to_write == LONG_MAX;
 
 	INIT_LIST_HEAD(&works);
 	INIT_LIST_HEAD(&splice);
@@ -9426,18 +9428,24 @@ static int start_delalloc_inodes(struct btrfs_root *root, u64 *nr, bool snapshot
 		if (snapshot)
 			set_bit(BTRFS_INODE_SNAPSHOT_FLUSH,
 				&binode->runtime_flags);
-		work = btrfs_alloc_delalloc_work(inode);
-		if (!work) {
-			iput(inode);
-			ret = -ENOMEM;
-			goto out;
-		}
-		list_add_tail(&work->list, &works);
-		btrfs_queue_work(root->fs_info->flush_workers,
-				 &work->work);
-		if (*nr != U64_MAX) {
-			(*nr)--;
-			if (*nr == 0)
+		if (full_flush) {
+			work = btrfs_alloc_delalloc_work(inode);
+			if (!work) {
+				iput(inode);
+				ret = -ENOMEM;
+				goto out;
+			}
+			list_add_tail(&work->list, &works);
+			btrfs_queue_work(root->fs_info->flush_workers,
+					 &work->work);
+		} else {
+			ret = sync_inode(inode, wbc);
+			if (!ret &&
+			    test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
+				     &BTRFS_I(inode)->runtime_flags))
+				ret = sync_inode(inode, wbc);
+			btrfs_add_delayed_iput(inode);
+			if (ret || wbc->nr_to_write <= 0)
 				goto out;
 		}
 		cond_resched();
@@ -9463,18 +9471,29 @@ static int start_delalloc_inodes(struct btrfs_root *root, u64 *nr, bool snapshot
 
 int btrfs_start_delalloc_snapshot(struct btrfs_root *root)
 {
+	struct writeback_control wbc = {
+		.nr_to_write = LONG_MAX,
+		.sync_mode = WB_SYNC_NONE,
+		.range_start = 0,
+		.range_end = LLONG_MAX,
+	};
 	struct btrfs_fs_info *fs_info = root->fs_info;
-	u64 nr = U64_MAX;
 
 	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
 		return -EROFS;
 
-	return start_delalloc_inodes(root, &nr, true, false);
+	return start_delalloc_inodes(root, &wbc, true, false);
 }
 
 int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, u64 nr,
 			       bool in_reclaim_context)
 {
+	struct writeback_control wbc = {
+		.nr_to_write = (nr == U64_MAX) ? LONG_MAX : (unsigned long)nr,
+		.sync_mode = WB_SYNC_NONE,
+		.range_start = 0,
+		.range_end = LLONG_MAX,
+	};
 	struct btrfs_root *root;
 	struct list_head splice;
 	int ret;
@@ -9488,6 +9507,13 @@ int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, u64 nr,
 	spin_lock(&fs_info->delalloc_root_lock);
 	list_splice_init(&fs_info->delalloc_roots, &splice);
 	while (!list_empty(&splice) && nr) {
+		/*
+		 * Reset nr_to_write here so we know that we're doing a full
+		 * flush.
+		 */
+		if (nr == U64_MAX)
+			wbc.nr_to_write = LONG_MAX;
+
 		root = list_first_entry(&splice, struct btrfs_root,
 					delalloc_root);
 		root = btrfs_grab_root(root);
@@ -9496,9 +9522,9 @@ int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, u64 nr,
 			       &fs_info->delalloc_roots);
 		spin_unlock(&fs_info->delalloc_root_lock);
 
-		ret = start_delalloc_inodes(root, &nr, false, in_reclaim_context);
+		ret = start_delalloc_inodes(root, &wbc, false, in_reclaim_context);
 		btrfs_put_root(root);
-		if (ret < 0)
+		if (ret < 0 || wbc.nr_to_write <= 0)
 			goto out;
 		spin_lock(&fs_info->delalloc_root_lock);
 	}
diff --git a/fs/btrfs/space-info.c b/fs/btrfs/space-info.c
index 67e55c5479b8..e8347461c8dd 100644
--- a/fs/btrfs/space-info.c
+++ b/fs/btrfs/space-info.c
@@ -532,7 +532,9 @@ static void shrink_delalloc(struct btrfs_fs_info *fs_info,
 
 	loops = 0;
 	while ((delalloc_bytes || dio_bytes) && loops < 3) {
-		btrfs_start_delalloc_roots(fs_info, items, true);
+		u64 nr_pages = min(delalloc_bytes, to_reclaim) >> PAGE_SHIFT;
+
+		btrfs_start_delalloc_roots(fs_info, nr_pages, true);
 
 		loops++;
 		if (wait_ordered && !trans) {
-- 
2.30.0