Linux filesystem development
 help / color / mirror / Atom feed
From: Al Viro <viro@zeniv.linux.org.uk>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: linux-fsdevel@vger.kernel.org,
	Christian Brauner <brauner@kernel.org>, Jan Kara <jack@suse.cz>,
	NeilBrown <neil@brown.name>
Subject: [RFC PATCH 21/25] shrinking rcu_read_lock() scope in d_alloc_parallel()
Date: Tue,  5 May 2026 06:54:08 +0100	[thread overview]
Message-ID: <20260505055412.1261144-22-viro@zeniv.linux.org.uk> (raw)
In-Reply-To: <20260505055412.1261144-1-viro@zeniv.linux.org.uk>

	The current use of rcu_read_lock() uses in d_alloc_parallel()
is fairly opaque - the single large scope serves two purposes.

	We start with lookup in normal hash, and there rcu_read_lock()
scope puts __d_lookup_rcu() and subsequent lockref_get_not_dead() into
the same RCU read-side critical area.

	If no match is found, we proceed to lock the hash chain of
in-lookup hash and scan that for a match.  If we find a match, we want
to grab it and wait for lookup in progress to finish.  Since the bitlock
we use for these hash chains has to nest inside ->d_lock, we need to
unlock the chain first and use lockref_get_not_dead() on the match.
That has to be done without breaking the RCU read-side critical area,
and we use the same rcu_read_lock() scope to bridge over.

	The thing is, after having grabbed the reference (and it is
very unlikely to fail) we proceed to grab ->d_lock - d_wait_lookup()
and __d_lookup_unhash()/__d_wake_in_lookup_waiters() are using that for
serialization.	That makes lockref_get_not_dead() pointless - trying
to avoid grabbing ->d_lock for refcount increment, only to grab it
anyway immediately after that.	If we grab ->d_lock first and replace
lockref_get_not_dead() with direct check for sign and increment if
non-negative we can move rcu_read_unlock() to immediately after grabbing
->d_lock.  Moreover, we don't need the RCU read-side critical area to
be contiguous since before earlier __d_lookup_rcu() - we can just as
well terminate the earlier one ASAP and call rcu_read_lock() again only
after having found a match (if any) in the in-lookup hash chain.

	That makes the entire thing easier to follow and the purpose
of those rcu_read_lock() calls easier to describe - the first scope is
for __d_lookup_rcu() + lockref_get_not_dead(), the second one bridges
over from the bitlock scope to the ->d_lock scope on the match found in
in-lookup hash.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
---
 fs/dcache.c | 29 ++++++++++++-----------------
 1 file changed, 12 insertions(+), 17 deletions(-)

diff --git a/fs/dcache.c b/fs/dcache.c
index 2a342d618303..f3c8d46867a9 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -2729,38 +2729,33 @@ struct dentry *d_alloc_parallel(struct dentry *parent,
 	spin_unlock(&parent->d_lock);
 
 retry:
-	rcu_read_lock();
 	seq = smp_load_acquire(&parent->d_inode->i_dir_seq);
 	r_seq = read_seqbegin(&rename_lock);
+	rcu_read_lock();
 	dentry = __d_lookup_rcu(parent, name, &d_seq);
 	if (unlikely(dentry)) {
 		if (!lockref_get_not_dead(&dentry->d_lockref)) {
 			rcu_read_unlock();
 			goto retry;
 		}
+		rcu_read_unlock();
 		if (read_seqcount_retry(&dentry->d_seq, d_seq)) {
-			rcu_read_unlock();
 			dput(dentry);
 			goto retry;
 		}
-		rcu_read_unlock();
 		dput(new);
 		return dentry;
 	}
-	if (unlikely(read_seqretry(&rename_lock, r_seq))) {
-		rcu_read_unlock();
+	rcu_read_unlock();
+	if (unlikely(read_seqretry(&rename_lock, r_seq)))
 		goto retry;
-	}
 
-	if (unlikely(seq & 1)) {
-		rcu_read_unlock();
+	if (unlikely(seq & 1))
 		goto retry;
-	}
 
 	hlist_bl_lock(b);
 	if (unlikely(READ_ONCE(parent->d_inode->i_dir_seq) != seq)) {
 		hlist_bl_unlock(b);
-		rcu_read_unlock();
 		goto retry;
 	}
 	/*
@@ -2777,19 +2772,20 @@ struct dentry *d_alloc_parallel(struct dentry *parent,
 			continue;
 		if (!d_same_name(dentry, parent, name))
 			continue;
+		rcu_read_lock();
 		hlist_bl_unlock(b);
+		spin_lock(&dentry->d_lock);
+		rcu_read_unlock();
 		/* now we can try to grab a reference */
-		if (!lockref_get_not_dead(&dentry->d_lockref)) {
-			rcu_read_unlock();
+		if (unlikely(dentry->d_lockref.count < 0)) {
+			spin_unlock(&dentry->d_lock);
 			goto retry;
 		}
-
-		rcu_read_unlock();
 		/*
 		 * somebody is likely to be still doing lookup for it;
-		 * wait for them to finish
+		 * pin it and wait for them to finish
 		 */
-		spin_lock(&dentry->d_lock);
+		dget_dlock(dentry);
 		d_wait_lookup(dentry);
 		/*
 		 * it's not in-lookup anymore; in principle we should repeat
@@ -2810,7 +2806,6 @@ struct dentry *d_alloc_parallel(struct dentry *parent,
 		dput(new);
 		return dentry;
 	}
-	rcu_read_unlock();
 	hlist_bl_add_head(&new->d_in_lookup_hash, b);
 	hlist_bl_unlock(b);
 	return new;
-- 
2.47.3


  parent reply	other threads:[~2026-05-05  5:54 UTC|newest]

Thread overview: 53+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-05  5:53 [RFC PATCH 00/25] assorted dcache cleanups and fixes Al Viro
2026-05-05  5:53 ` [RFC PATCH 01/25] VFS: use wait_var_event for waiting in d_alloc_parallel() Al Viro
2026-05-05  5:53 ` [RFC PATCH 02/25] alloc_path_pseudo(): make sure we don't end up with NORCU dentries for directories Al Viro
2026-05-05  8:21   ` NeilBrown
2026-05-05 17:48     ` Al Viro
2026-05-05  5:53 ` [RFC PATCH 03/25] fix a race between d_find_any_alias() and final dput() of NORCU dentries Al Viro
2026-05-05 17:06   ` Linus Torvalds
2026-05-05 20:29     ` Al Viro
2026-05-05  5:53 ` [RFC PATCH 04/25] find_acceptable_alias(): skip NORCU aliases with zero refcount Al Viro
2026-05-05  5:53 ` [RFC PATCH 05/25] select_collect(): ignore dentries on shrink lists if they have positive refcounts Al Viro
2026-05-05  5:53 ` [RFC PATCH 06/25] make to_shrink_list() return whether it has moved dentry to list Al Viro
2026-05-05  5:53 ` [RFC PATCH 07/25] kill d_dispose_if_unused() Al Viro
2026-05-05  5:53 ` [RFC PATCH 08/25] d_prune_aliases(): make sure to skip NORCU aliases Al Viro
2026-05-05  5:53 ` [RFC PATCH 09/25] shrink_dentry_list(): start with removing from shrink list Al Viro
2026-05-07 20:39   ` Al Viro
2026-05-05  5:53 ` [RFC PATCH 10/25] fold lock_for_kill() into shrink_kill() Al Viro
2026-05-05  5:53 ` [RFC PATCH 11/25] fold lock_for_kill() and __dentry_kill() into common helper Al Viro
2026-05-05  5:53 ` [RFC PATCH 12/25] reducing rcu_read_lock() scopes in dput and friends, step 1 Al Viro
2026-05-05  8:55   ` NeilBrown
2026-05-05 14:22     ` Al Viro
2026-05-05 21:58       ` NeilBrown
2026-05-05 16:47   ` Linus Torvalds
2026-05-05 22:42     ` Al Viro
2026-05-07  7:35       ` Al Viro
2026-05-07 15:32         ` Linus Torvalds
2026-05-05  5:54 ` [RFC PATCH 13/25] reducing rcu_read_lock() scopes in dput and friends, step 2 Al Viro
2026-05-05  5:54 ` [RFC PATCH 14/25] reducing rcu_read_lock() scopes in dput and friends, step 3 Al Viro
2026-05-05  5:54 ` [RFC PATCH 15/25] reducing rcu_read_lock() scopes in dput and friends, step 4 Al Viro
2026-05-05  5:54 ` [RFC PATCH 16/25] reducing rcu_read_lock() scopes in dput and friends, step 5 Al Viro
2026-05-05  5:54 ` [RFC PATCH 17/25] reducing rcu_read_lock() scopes in dput and friends, step 6 Al Viro
2026-05-05  5:54 ` [RFC PATCH 18/25] adjust calling conventions of lock_for_kill(), fold __dentry_kill() into dentry_kill() Al Viro
2026-05-05  5:54 ` [RFC PATCH 19/25] document dentry_kill() Al Viro
2026-05-05  5:54 ` [RFC PATCH 20/25] d_walk(): shrink rcu_read_lock() scope Al Viro
2026-05-05 17:01   ` Linus Torvalds
2026-05-05 20:05     ` Al Viro
2026-05-05 21:40       ` Frederic Weisbecker
2026-05-05 22:50         ` Al Viro
2026-05-06  3:49           ` Paul E. McKenney
2026-05-07 22:39             ` NeilBrown
2026-05-07 23:21               ` Paul E. McKenney
2026-05-08 14:47                 ` Al Viro
2026-05-08 22:03                   ` Paul E. McKenney
2026-05-08 23:03                     ` Al Viro
2026-05-08  3:01         ` Al Viro
2026-05-05  5:54 ` Al Viro [this message]
2026-05-07 21:52   ` [RFC PATCH 21/25] shrinking rcu_read_lock() scope in d_alloc_parallel() Jori Koolstra
2026-05-08  3:12     ` Al Viro
2026-05-08  9:28       ` Jori Koolstra
2026-05-05  5:54 ` [RFC PATCH 22/25] shrink_dentry_tree(): unify the calls of shrink_dentry_list() Al Viro
2026-05-05  5:54 ` [RFC PATCH 23/25] wind ->s_roots via ->d_sib instead of ->d_hash Al Viro
2026-05-05  5:54 ` [RFC PATCH 24/25] nfs: get rid of fake root dentries Al Viro
2026-05-05  5:54 ` [RFC PATCH 25/25] make cursors NORCU Al Viro
2026-05-05 17:09 ` [RFC PATCH 00/25] assorted dcache cleanups and fixes Linus Torvalds

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=20260505055412.1261144-22-viro@zeniv.linux.org.uk \
    --to=viro@zeniv.linux.org.uk \
    --cc=brauner@kernel.org \
    --cc=jack@suse.cz \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=neil@brown.name \
    --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