linux-kernel - [PATCH 8/8] fs: use unregister_shrinker_delayed_{initiate, finalize} for super

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 31 May 2023 09:57:42 +0000
From:   Qi Zheng <qi.zheng@...ux.dev>
To:     akpm@...ux-foundation.org, tkhai@...ru, roman.gushchin@...ux.dev,
        vbabka@...e.cz, viro@...iv.linux.org.uk, brauner@...nel.org,
        djwong@...nel.org, hughd@...gle.com, paulmck@...nel.org,
        muchun.song@...ux.dev
Cc:     linux-mm@...ck.org, linux-fsdevel@...r.kernel.org,
        linux-xfs@...r.kernel.org, linux-kernel@...r.kernel.org,
        Qi Zheng <zhengqi.arch@...edance.com>
Subject: [PATCH 8/8] fs: use unregister_shrinker_delayed_{initiate, finalize} for super_block shrinker

From: Kirill Tkhai <tkhai@...ru>

Previous patches made all the data, which is touched from
super_cache_count(), destroyed from destroy_super_work():
s_dentry_lru, s_inode_lru and super_block::s_fs_info.

super_cache_scan() can't be called after SB_ACTIVE is cleared
in generic_shutdown_super().

So, it safe to move heavy unregister_shrinker_delayed_finalize()
part to delayed work, i.e. it's safe for parallel do_shrink_slab()
to be executed between unregister_shrinker_delayed_initiate() and
destroy_super_work()->unregister_shrinker_delayed_finalize().

This makes the heavy synchronize_srcu() to do not affect on user-visible
unregistration speed (since now it's executed from workqueue).

All further time-critical for unregistration places may be written
in the same conception.

Signed-off-by: Kirill Tkhai <tkhai@...ru>
Signed-off-by: Qi Zheng <zhengqi.arch@...edance.com>
---
 fs/super.c         | 4 +++-
 include/linux/fs.h | 5 +++++
 2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/fs/super.c b/fs/super.c
index 4e9d08224f86..c61efb74fa7f 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -165,6 +165,8 @@ static void destroy_super_work(struct work_struct *work)
 							destroy_work);
 	int i;
 
+	unregister_shrinker_delayed_finalize(&s->s_shrink);
+
 	WARN_ON(list_lru_count(&s->s_dentry_lru));
 	WARN_ON(list_lru_count(&s->s_inode_lru));
 	list_lru_destroy(&s->s_dentry_lru);
@@ -337,7 +339,7 @@ void deactivate_locked_super(struct super_block *s)
 {
 	struct file_system_type *fs = s->s_type;
 	if (atomic_dec_and_test(&s->s_active)) {
-		unregister_shrinker(&s->s_shrink);
+		unregister_shrinker_delayed_initiate(&s->s_shrink);
 		fs->kill_sb(s);
 
 		put_filesystem(fs);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 30b46d0facfc..869dd8de91a5 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1929,6 +1929,11 @@ struct super_operations {
 	ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
 	struct dquot **(*get_dquots)(struct inode *);
 #endif
+	/*
+	 * Shrinker may call these two function on destructing super_block
+	 * till unregister_shrinker_delayed_finalize() has completed
+	 * in destroy_super_work(), and they must care about that.
+	 */
 	long (*nr_cached_objects)(struct super_block *,
 				  struct shrink_control *);
 	long (*free_cached_objects)(struct super_block *,
-- 
2.30.2