All of lore.kernel.org
 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
Subject: [PATCH 02/15] fast_dput(): handle underflows gracefully
Date: Wed,  1 Nov 2023 06:20:51 +0000	[thread overview]
Message-ID: <20231101062104.2104951-2-viro@zeniv.linux.org.uk> (raw)
In-Reply-To: <20231101062104.2104951-1-viro@zeniv.linux.org.uk>

If refcount is less than 1, we should just warn, unlock dentry and
return true, so that the caller doesn't try to do anything else.

Taking care of that leaves the rest of "lockref_put_return() has
failed" case equivalent to "decrement refcount and rejoin the
normal slow path after the point where we grab ->d_lock".

NOTE: lockref_put_return() is strictly a fastpath thing - unlike
the rest of lockref primitives, it does not contain a fallback.
Caller (and it looks like fast_dput() is the only legitimate one
in the entire kernel) has to do that itself.  Reasons for
lockref_put_return() failures:
	* ->d_lock held by somebody
	* refcount <= 0
	* ... or an architecture not supporting lockref use of
cmpxchg - sparc, anything non-SMP, config with spinlock debugging...

We could add a fallback, but it would be a clumsy API - we'd have
to distinguish between:
	(1) refcount > 1 - decremented, lock not held on return
	(2) refcount < 1 - left alone, probably no sense to hold the lock
	(3) refcount is 1, no cmphxcg - decremented, lock held on return
	(4) refcount is 1, cmphxcg supported - decremented, lock *NOT* held
	    on return.
We want to return with no lock held in case (4); that's the whole point of that
thing.  We very much do not want to have the fallback in case (3) return without
a lock, since the caller might have to retake it in that case.
So it wouldn't be more convenient than doing the fallback in the caller and
it would be very easy to screw up, especially since the test coverage would
suck - no way to test (3) and (4) on the same kernel build.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
---
 fs/dcache.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/fs/dcache.c b/fs/dcache.c
index 5ec14df04f11..ddc534b39c22 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -779,12 +779,12 @@ static inline bool fast_dput(struct dentry *dentry)
 	 */
 	if (unlikely(ret < 0)) {
 		spin_lock(&dentry->d_lock);
-		if (dentry->d_lockref.count > 1) {
-			dentry->d_lockref.count--;
+		if (WARN_ON_ONCE(dentry->d_lockref.count <= 0)) {
 			spin_unlock(&dentry->d_lock);
 			return true;
 		}
-		return false;
+		dentry->d_lockref.count--;
+		goto locked;
 	}
 
 	/*
@@ -842,6 +842,7 @@ static inline bool fast_dput(struct dentry *dentry)
 	 * else could have killed it and marked it dead. Either way, we
 	 * don't need to do anything else.
 	 */
+locked:
 	if (dentry->d_lockref.count) {
 		spin_unlock(&dentry->d_lock);
 		return true;
-- 
2.39.2


  reply	other threads:[~2023-11-01  6:21 UTC|newest]

Thread overview: 119+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-10-30  0:37 [RFC] simplifying fast_dput(), dentry_kill() et.al Al Viro
2023-10-30 21:53 ` Al Viro
2023-10-30 22:18   ` Linus Torvalds
2023-10-31  0:18     ` Al Viro
2023-10-31  1:53       ` Al Viro
2023-10-31  6:12         ` Al Viro
2023-11-01  6:18           ` Al Viro
2023-11-01  6:20           ` [PATCH 01/15] fast_dput(): having ->d_delete() is not reason to delay refcount decrement Al Viro
2023-11-01  6:20             ` Al Viro [this message]
2023-11-01  6:20             ` [PATCH 03/15] fast_dput(): new rules for refcount Al Viro
2023-11-01  6:20             ` [PATCH 04/15] __dput_to_list(): do decrement of refcount in the caller Al Viro
2023-11-01  6:20             ` [PATCH 05/15] retain_dentry(): lift decrement of ->d_count into callers Al Viro
2023-11-01  6:20             ` [PATCH 06/15] __dentry_kill(): get consistent rules for ->d_count Al Viro
2023-11-01  6:20             ` [PATCH 07/15] dentry_kill(): don't bother with retain_dentry() on slow path Al Viro
2023-11-01  6:20             ` [PATCH 08/15] Call retain_dentry() with refcount 0 Al Viro
2023-11-01  6:20             ` [PATCH 09/15] fold the call of retain_dentry() into fast_dput() Al Viro
2023-11-01  8:45               ` Al Viro
2023-11-01 17:30                 ` Linus Torvalds
2023-11-01 18:19                   ` Al Viro
2023-11-10  4:20                     ` lockless case of retain_dentry() (was Re: [PATCH 09/15] fold the call of retain_dentry() into fast_dput()) Al Viro
2023-11-10  5:57                       ` Linus Torvalds
2023-11-10  6:22                         ` Linus Torvalds
2023-11-22  6:29                           ` Guo Ren
2023-11-10  8:19                         ` Al Viro
2023-11-22  7:19                         ` Guo Ren
2023-11-22 17:20                           ` Linus Torvalds
2023-11-22 17:52                             ` Linus Torvalds
2023-11-22 18:05                               ` Linus Torvalds
2023-11-22 19:11                               ` Linus Torvalds
2023-11-29  7:14                                 ` Guo Ren
2023-11-29 12:25                                 ` Guo Ren
2023-11-29 14:42                                   ` Linus Torvalds
2023-11-26 16:39                             ` Guo Ren
2023-11-26 16:51                               ` Linus Torvalds
2023-11-30 10:00                                 ` Guo Ren
2023-12-01  1:09                                   ` Linus Torvalds
2023-12-01  3:36                                     ` Guo Ren
2023-12-01  5:15                                       ` Linus Torvalds
2023-12-01  7:31                                         ` Guo Ren
2023-11-26 16:51                               ` Guo Ren
2023-11-26 17:06                               ` Linus Torvalds
2023-11-26 17:59                                 ` Linus Torvalds
2023-11-29  9:52                                 ` Guo Ren
2023-11-01  6:20             ` [PATCH 10/15] don't try to cut corners in shrink_lock_dentry() Al Viro
2023-11-01  6:21             ` [PATCH 11/15] fold dentry_kill() into dput() Al Viro
2023-11-01  6:21             ` [PATCH 12/15] get rid of __dget() Al Viro
2023-11-01  6:21             ` [PATCH 13/15] shrink_dentry_list(): no need to check that dentry refcount is marked dead Al Viro
2023-11-01  6:21             ` [PATCH 14/15] to_shrink_list(): call only if refcount is 0 Al Viro
2023-11-01  6:21             ` [PATCH 15/15] switch select_collect{,2}() to use of to_shrink_list() Al Viro
2023-11-01  2:22       ` [RFC] simplifying fast_dput(), dentry_kill() et.al Al Viro
2023-11-01 14:29         ` Benjamin Coddington
2023-11-05 19:54       ` Al Viro
2023-11-05 21:59         ` Al Viro
2023-11-06  5:53         ` Al Viro
2023-11-07  2:08           ` Al Viro
2023-11-09  6:19             ` [RFC][PATCHSET v2] " Al Viro
2023-11-09  6:20               ` [PATCH 01/22] struct dentry: get rid of randomize_layout idiocy Al Viro
2023-11-09  6:20                 ` [PATCH 02/22] switch nfsd_client_rmdir() to use of simple_recursive_removal() Al Viro
2023-11-09 13:42                   ` Christian Brauner
2023-11-09 14:01                   ` Chuck Lever
2023-11-09 18:47                     ` Al Viro
2023-11-09 18:50                       ` Chuck Lever III
2023-11-09  6:20                 ` [PATCH 03/22] coda_flag_children(): cope with dentries turning negative Al Viro
2023-11-09 13:43                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 04/22] dentry: switch the lists of children to hlist Al Viro
2023-11-09 13:48                   ` Christian Brauner
2023-11-09 19:32                     ` Al Viro
2023-11-09  6:20                 ` [PATCH 05/22] centralize killing dentry from shrink list Al Viro
2023-11-09 13:49                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 06/22] get rid of __dget() Al Viro
2023-11-09 13:50                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 07/22] shrink_dentry_list(): no need to check that dentry refcount is marked dead Al Viro
2023-11-09 13:53                   ` Christian Brauner
2023-11-09 20:28                     ` Al Viro
2023-11-09  6:20                 ` [PATCH 08/22] fast_dput(): having ->d_delete() is not reason to delay refcount decrement Al Viro
2023-11-09 13:58                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 09/22] fast_dput(): handle underflows gracefully Al Viro
2023-11-09 14:46                   ` Christian Brauner
2023-11-09 20:39                     ` Al Viro
2023-11-09  6:20                 ` [PATCH 10/22] fast_dput(): new rules for refcount Al Viro
2023-11-09 14:54                   ` Christian Brauner
2023-11-09 20:52                     ` Al Viro
2023-11-09  6:20                 ` [PATCH 11/22] __dput_to_list(): do decrement of refcount in the callers Al Viro
2023-11-09 15:21                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 12/22] Make retain_dentry() neutral with respect to refcounting Al Viro
2023-11-09 15:22                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 13/22] __dentry_kill(): get consistent rules for victim's refcount Al Viro
2023-11-09 15:27                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 14/22] dentry_kill(): don't bother with retain_dentry() on slow path Al Viro
2023-11-09 15:53                   ` Christian Brauner
2023-11-09 21:29                     ` Al Viro
2023-11-09  6:20                 ` [PATCH 15/22] Call retain_dentry() with refcount 0 Al Viro
2023-11-09 16:09                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 16/22] fold the call of retain_dentry() into fast_dput() Al Viro
2023-11-09 16:17                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 17/22] don't try to cut corners in shrink_lock_dentry() Al Viro
2023-11-09 17:20                   ` Christian Brauner
2023-11-09 21:45                     ` Al Viro
2023-11-10  9:07                       ` Christian Brauner
2023-11-09 17:39                   ` Linus Torvalds
2023-11-09 18:11                     ` Linus Torvalds
2023-11-09 18:20                     ` Al Viro
2023-11-09  6:20                 ` [PATCH 18/22] fold dentry_kill() into dput() Al Viro
2023-11-09 17:22                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 19/22] to_shrink_list(): call only if refcount is 0 Al Viro
2023-11-09 17:29                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 20/22] switch select_collect{,2}() to use of to_shrink_list() Al Viro
2023-11-09 17:31                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 21/22] d_prune_aliases(): use a shrink list Al Viro
2023-11-09 17:33                   ` Christian Brauner
2023-11-09  6:20                 ` [PATCH 22/22] __dentry_kill(): new locking scheme Al Viro
2023-11-10 13:34                   ` Christian Brauner
2023-11-09 13:33                 ` [PATCH 01/22] struct dentry: get rid of randomize_layout idiocy Christian Brauner
2023-10-31  2:25     ` [RFC] simplifying fast_dput(), dentry_kill() et.al Gao Xiang
2023-10-31  2:29       ` Gao Xiang
2023-10-31  3:02       ` Linus Torvalds
2023-10-31  3:13         ` Gao Xiang
2023-10-31  3:26         ` Al Viro
2023-10-31  3:41           ` 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=20231101062104.2104951-2-viro@zeniv.linux.org.uk \
    --to=viro@zeniv.linux.org.uk \
    --cc=linux-fsdevel@vger.kernel.org \
    --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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.