public inbox for linux-fsdevel@vger.kernel.org
 help / color / mirror / Atom feed
From: Al Viro <viro@zeniv.linux.org.uk>
To: linux-fsdevel@vger.kernel.org
Cc: agruenba@redhat.com, amir73il@gmail.com, brauner@kernel.org,
	ceph-devel@vger.kernel.org, dhowells@redhat.com,
	hubcap@omnibond.com, jack@suse.cz, krisman@kernel.org,
	linux-nfs@vger.kernel.org, miklos@szeredi.hu,
	torvalds@linux-foundation.org
Subject: [PATCH 03/20] make take_dentry_name_snapshot() lockless
Date: Fri, 10 Jan 2025 02:42:46 +0000	[thread overview]
Message-ID: <20250110024303.4157645-3-viro@zeniv.linux.org.uk> (raw)
In-Reply-To: <20250110024303.4157645-1-viro@zeniv.linux.org.uk>

Use ->d_seq instead of grabbing ->d_lock; in case of shortname dentries
that avoids any stores to shared data objects and in case of long names
we are down to (unavoidable) atomic_inc on the external_name refcount.

Makes the thing safer as well - the areas where ->d_seq is held odd are
all nested inside the areas where ->d_lock is held, and the latter are
much more numerous.

NOTE: now that there is a lockless path where we might try to grab
a reference to an already doomed external_name instance, it is no
longer possible for external_name.u.count and external_name.u.head
to share space (kudos to Linus for spotting that).

To reduce the noice this commit just turns external_name.u into
a struct (instead of union); the next commit will dissolve it.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
---
 fs/dcache.c | 35 +++++++++++++++++++++++++----------
 1 file changed, 25 insertions(+), 10 deletions(-)

diff --git a/fs/dcache.c b/fs/dcache.c
index 52662a5d08e4..f387dc97df86 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -296,9 +296,9 @@ static inline int dentry_cmp(const struct dentry *dentry, const unsigned char *c
 }
 
 struct external_name {
-	union {
-		atomic_t count;
-		struct rcu_head head;
+	struct {
+		atomic_t count;		// ->count and ->head can't be combined
+		struct rcu_head head;	// see take_dentry_name_snapshot()
 	} u;
 	unsigned char name[];
 };
@@ -329,15 +329,30 @@ static inline int dname_external(const struct dentry *dentry)
 
 void take_dentry_name_snapshot(struct name_snapshot *name, struct dentry *dentry)
 {
-	spin_lock(&dentry->d_lock);
-	name->name = dentry->d_name;
-	if (unlikely(dname_external(dentry))) {
-		atomic_inc(&external_name(dentry)->u.count);
-	} else {
+	unsigned seq;
+	const unsigned char *s;
+
+	rcu_read_lock();
+retry:
+	seq = read_seqcount_begin(&dentry->d_seq);
+	s = READ_ONCE(dentry->d_name.name);
+	name->name.hash_len = dentry->d_name.hash_len;
+	name->name.name = name->inline_name.string;
+	if (likely(s == dentry->d_shortname.string)) {
 		name->inline_name = dentry->d_shortname;
-		name->name.name = name->inline_name.string;
+	} else {
+		struct external_name *p;
+		p = container_of(s, struct external_name, name[0]);
+		// get a valid reference
+		if (unlikely(!atomic_inc_not_zero(&p->u.count)))
+			goto retry;
+		name->name.name = s;
 	}
-	spin_unlock(&dentry->d_lock);
+	if (read_seqcount_retry(&dentry->d_seq, seq)) {
+		release_dentry_name_snapshot(name);
+		goto retry;
+	}
+	rcu_read_unlock();
 }
 EXPORT_SYMBOL(take_dentry_name_snapshot);
 
-- 
2.39.5


  parent reply	other threads:[~2025-01-10  2:43 UTC|newest]

Thread overview: 96+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-01-10  2:38 [PATCHES][RFC][CFT] ->d_revalidate() calling conventions changes (->d_parent/->d_name stability problems) Al Viro
2025-01-10  2:42 ` [PATCH 01/20] make sure that DNAME_INLINE_LEN is a multiple of word size Al Viro
2025-01-10  2:42   ` [PATCH 02/20] dcache: back inline names with a struct-wrapped array of unsigned long Al Viro
2025-01-10  9:35     ` Jan Kara
2025-01-10 16:24       ` Al Viro
2025-01-10  2:42   ` Al Viro [this message]
2025-01-10  9:45     ` [PATCH 03/20] make take_dentry_name_snapshot() lockless Jan Kara
2025-01-10  2:42   ` [PATCH 04/20] dissolve external_name.u into separate members Al Viro
2025-01-10  7:34     ` David Howells
2025-01-10 16:46       ` Al Viro
2025-01-10  2:42   ` [PATCH 05/20] ext4 fast_commit: make use of name_snapshot primitives Al Viro
2025-01-10  9:15     ` Jan Kara
2025-01-10  2:42   ` [PATCH 06/20] generic_ci_d_compare(): use shortname_storage Al Viro
2025-01-10  2:42   ` [PATCH 07/20] Pass parent directory inode and expected name to ->d_revalidate() Al Viro
2025-01-10  2:42   ` [PATCH 08/20] afs_d_revalidate(): use stable name and parent inode passed by caller Al Viro
2025-01-10  2:42   ` [PATCH 09/20] ceph_d_revalidate(): use stable " Al Viro
2025-01-10 19:45     ` Viacheslav Dubeyko
2025-01-10  2:42   ` [PATCH 10/20] ceph_d_revalidate(): propagate stable name down into request enconding Al Viro
2025-01-10  2:42   ` [PATCH 11/20] fscrypt_d_revalidate(): use stable parent inode passed by caller Al Viro
2025-01-10  2:42   ` [PATCH 12/20] exfat_d_revalidate(): " Al Viro
2025-01-10  2:42   ` [PATCH 13/20] vfat_revalidate{,_ci}(): " Al Viro
2025-01-10  2:42   ` [PATCH 14/20] fuse_dentry_revalidate(): use stable parent inode and name " Al Viro
2025-01-10  2:42   ` [PATCH 15/20] gfs2_drevalidate(): " Al Viro
2025-01-10 19:20     ` Andreas Grünbacher
2025-01-10  2:42   ` [PATCH 16/20] nfs{,4}_lookup_validate(): use stable parent inode " Al Viro
2025-01-10  2:43   ` [PATCH 17/20] nfs: fix ->d_revalidate() UAF on ->d_name accesses Al Viro
2025-01-10  2:43   ` [PATCH 18/20] ocfs2_dentry_revalidate(): use stable parent inode and name passed by caller Al Viro
2025-01-10  9:54     ` Jan Kara
2025-01-10  2:43   ` [PATCH 19/20] orangefs_d_revalidate(): " Al Viro
2025-01-10  3:06     ` Linus Torvalds
2025-01-10  2:43   ` [PATCH 20/20] 9p: fix ->rename_sem exclusion Al Viro
2025-01-10  3:11     ` Linus Torvalds
2025-01-10  5:53       ` Al Viro
2025-01-10  9:21   ` [PATCH 01/20] make sure that DNAME_INLINE_LEN is a multiple of word size Jan Kara
2025-01-16  5:21 ` [PATCHES v2][RFC][CFT] ->d_revalidate() calling conventions changes (->d_parent/->d_name stability problems) Al Viro
2025-01-16  5:22   ` [PATCH v2 01/20] make sure that DNAME_INLINE_LEN is a multiple of word size Al Viro
2025-01-16  5:22     ` [PATCH v2 02/20] dcache: back inline names with a struct-wrapped array of unsigned long Al Viro
2025-01-16  5:23     ` [PATCH v2 03/20] make take_dentry_name_snapshot() lockless Al Viro
2025-01-16  5:23     ` [PATCH v2 04/20] dissolve external_name.u into separate members Al Viro
2025-01-16 10:06       ` Jan Kara
2025-01-16  5:23     ` [PATCH v2 05/20] ext4 fast_commit: make use of name_snapshot primitives Al Viro
2025-01-16  5:23     ` [PATCH v2 06/20] generic_ci_d_compare(): use shortname_storage Al Viro
2025-01-16 15:38       ` Gabriel Krisman Bertazi
2025-01-16 15:46         ` Al Viro
2025-01-16 15:53           ` Gabriel Krisman Bertazi
2025-01-16  5:23     ` [PATCH v2 07/20] Pass parent directory inode and expected name to ->d_revalidate() Al Viro
2025-01-16 15:15       ` Gabriel Krisman Bertazi
2025-01-17 18:55       ` Jeff Layton
2025-01-17 19:00         ` Al Viro
2025-01-16  5:23     ` [PATCH v2 08/20] afs_d_revalidate(): use stable name and parent inode passed by caller Al Viro
2025-01-22 20:27       ` David Howells
2025-01-22 21:01         ` Al Viro
2025-01-22 21:24           ` Al Viro
2025-01-22 21:55             ` David Howells
2025-01-16  5:23     ` [PATCH v2 09/20] ceph_d_revalidate(): use stable " Al Viro
2025-01-17 18:35       ` Jeff Layton
2025-01-16  5:23     ` [PATCH v2 10/20] ceph_d_revalidate(): propagate stable name down into request enconding Al Viro
2025-01-17 18:35       ` Jeff Layton
2025-01-16  5:23     ` [PATCH v2 11/20] fscrypt_d_revalidate(): use stable parent inode passed by caller Al Viro
2025-01-17 15:20       ` Jeff Layton
2025-01-16  5:23     ` [PATCH v2 12/20] exfat_d_revalidate(): " Al Viro
2025-01-16  5:23     ` [PATCH v2 13/20] vfat_revalidate{,_ci}(): " Al Viro
2025-01-17 15:22       ` Jeff Layton
2025-01-16  5:23     ` [PATCH v2 14/20] fuse_dentry_revalidate(): use stable parent inode and name " Al Viro
2025-01-17 15:18       ` Jeff Layton
2025-01-16  5:23     ` [PATCH v2 15/20] gfs2_drevalidate(): " Al Viro
2025-01-16  5:23     ` [PATCH v2 16/20] nfs{,4}_lookup_validate(): use stable parent inode " Al Viro
2025-01-17 14:05       ` Jeff Layton
2025-01-16  5:23     ` [PATCH v2 17/20] nfs: fix ->d_revalidate() UAF on ->d_name accesses Al Viro
2025-01-17 15:12       ` Jeff Layton
2025-01-16  5:23     ` [PATCH v2 18/20] ocfs2_dentry_revalidate(): use stable parent inode and name passed by caller Al Viro
2025-01-16  5:23     ` [PATCH v2 19/20] orangefs_d_revalidate(): " Al Viro
2025-01-16  5:23     ` [PATCH v2 20/20] 9p: fix ->rename_sem exclusion Al Viro
2025-01-23  1:45   ` [PATCHES v3][RFC][CFT] ->d_revalidate() calling conventions changes (->d_parent/->d_name stability problems) Al Viro
2025-01-23  1:46     ` [PATCH v3 01/20] make sure that DNAME_INLINE_LEN is a multiple of word size Al Viro
2025-01-23  1:46       ` [PATCH v3 02/20] dcache: back inline names with a struct-wrapped array of unsigned long Al Viro
2025-01-23  1:46       ` [PATCH v3 03/20] make take_dentry_name_snapshot() lockless Al Viro
2025-01-23  1:46       ` [PATCH v3 04/20] dissolve external_name.u into separate members Al Viro
2025-01-23  1:46       ` [PATCH v3 05/20] ext4 fast_commit: make use of name_snapshot primitives Al Viro
2025-01-23  1:46       ` [PATCH v3 06/20] generic_ci_d_compare(): use shortname_storage Al Viro
2025-01-23  1:46       ` [PATCH v3 07/20] Pass parent directory inode and expected name to ->d_revalidate() Al Viro
2025-01-23  1:46       ` [PATCH v3 08/20] afs_d_revalidate(): use stable name and parent inode passed by caller Al Viro
2025-01-23  1:46       ` [PATCH v3 09/20] ceph_d_revalidate(): use stable " Al Viro
2025-01-23  1:46       ` [PATCH v3 10/20] ceph_d_revalidate(): propagate stable name down into request encoding Al Viro
2025-01-23  1:46       ` [PATCH v3 11/20] fscrypt_d_revalidate(): use stable parent inode passed by caller Al Viro
2025-01-23  1:46       ` [PATCH v3 12/20] exfat_d_revalidate(): " Al Viro
2025-01-23  1:46       ` [PATCH v3 13/20] vfat_revalidate{,_ci}(): " Al Viro
2025-01-23  1:46       ` [PATCH v3 14/20] fuse_dentry_revalidate(): use stable parent inode and name " Al Viro
2025-01-23 10:51         ` Miklos Szeredi
2025-01-23  1:46       ` [PATCH v3 15/20] gfs2_drevalidate(): " Al Viro
2025-01-23  1:46       ` [PATCH v3 16/20] nfs{,4}_lookup_validate(): use stable parent inode " Al Viro
2025-01-23  1:46       ` [PATCH v3 17/20] nfs: fix ->d_revalidate() UAF on ->d_name accesses Al Viro
2025-01-23  1:46       ` [PATCH v3 18/20] ocfs2_dentry_revalidate(): use stable parent inode and name passed by caller Al Viro
2025-01-23  1:46       ` [PATCH v3 19/20] orangefs_d_revalidate(): " Al Viro
2025-01-25 16:25         ` Mike Marshall
2025-01-23  1:46       ` [PATCH v3 20/20] 9p: fix ->rename_sem exclusion Al Viro

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=20250110024303.4157645-3-viro@zeniv.linux.org.uk \
    --to=viro@zeniv.linux.org.uk \
    --cc=agruenba@redhat.com \
    --cc=amir73il@gmail.com \
    --cc=brauner@kernel.org \
    --cc=ceph-devel@vger.kernel.org \
    --cc=dhowells@redhat.com \
    --cc=hubcap@omnibond.com \
    --cc=jack@suse.cz \
    --cc=krisman@kernel.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-nfs@vger.kernel.org \
    --cc=miklos@szeredi.hu \
    --cc=torvalds@linux-foundation.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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox