Linux kernel -stable discussions
 help / color / mirror / Atom feed
From: Oleg Nesterov <oleg@redhat.com>
To: Linus Torvalds <torvalds@linux-foundation.org>,
	Mani Tofigh via B4 Relay <devnull+mt4022.columbia.edu@kernel.org>
Cc: Breno Leitao <leitao@debian.org>,
	Christian Brauner <brauner@kernel.org>,
	Mateusz Guzik <mjguzik@gmail.com>,
	Alexander Viro <viro@zeniv.linux.org.uk>, Jan Kara <jack@suse.cz>,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	tz2294@columbia.edu, stable@vger.kernel.org,
	Mani Tofigh <mt4022@columbia.edu>
Subject: Re: [PATCH] pipe: keep per-write wakeups for EPOLLEXCLUSIVE consumers
Date: Sun, 22 Nov 2026 18:09:24 +0100	[thread overview]
Message-ID: <awMhxEuMHULvk8SA@redhat.com> (raw)
In-Reply-To: <av3SmRMGDLVAwMzK@redhat.com>

On 11/18, Oleg Nesterov wrote:
>
> On 11/18, Linus Torvalds wrote:
> >
> > Some duplication - with explicit naming of the locking status for the
> > special cases that are called with the lock already held - is much
> > preferable to dynamically doing the locking
>
> OK, this was my plan B ;) I'll try to make a patch this week.

Damn, I am stipid.

So I wrote the patches, but just before sending them, I realized I was wrong.
Just for the record, see the diff I was going to send at the end.

Yes, __wait_event_interruptible_locked() is buggy, but only in that do_wait_intr*()
use list_add_tail() regardless of WQ_FLAG_EXCLUSIVE.

I missed that fact that do_wait_intr*() always return 0 if they drop wq->lock,
and in this case __wait_event_interruptible_locked() will always recheck the
condition with this lock held. So an exclusive waiter can't miss miss a wakeup.

So we only need to change do_wait_intr*() to check WQ_FLAG_EXCLUSIVE and use
__add_wait_queue() if it is set. It doesn't make sense to refactor this code.

Mani, can you resend your patch? perhaps it also makes sense to turn the
"exclusive" arg into wq_flags like the Linus's commit cea5f91daa3ac does.

Oleg.
---

diff --git a/include/linux/wait.h b/include/linux/wait.h
index b3e94c71cd3e..c30e6b2ee231 100644
--- a/include/linux/wait.h
+++ b/include/linux/wait.h
@@ -298,7 +298,6 @@ extern void init_wait_entry(struct wait_queue_entry *wq_entry, int flags);
  * on purpose; we use long where we can return timeout values and int
  * otherwise.
  */
-
 #define ___wait_event(wq_head, condition, state, wq_flag, ret, cmd)		\
 ({										\
 	__label__ __out;							\
@@ -326,6 +325,35 @@ extern void init_wait_entry(struct wait_queue_entry *wq_entry, int flags);
 __out:	__ret;									\
 })
 
+/*
+ * Same as ___wait_event(), but called with wq_head->lock held.
+ * @cmd must release and reacquire this lock around schedule().
+ */
+#define ___wait_event_locked(wq_head, condition, state, wq_flag, ret, cmd)	\
+({										\
+	__label__ __out;							\
+	struct wait_queue_entry __wq_entry;					\
+	long __ret = ret;	/* explicit shadow */				\
+										\
+	init_wait_entry(&__wq_entry, wq_flag);					\
+	for (;;) {								\
+		long __int = prepare_to_wait_event_locked(&wq_head,		\
+							  &__wq_entry, state);	\
+		if (condition)							\
+			break;							\
+										\
+		if (___wait_is_interruptible(state) && __int) {			\
+			__ret = __int;						\
+			goto __out;						\
+		}								\
+										\
+		cmd;								\
+	}									\
+	__set_current_state(TASK_RUNNING);					\
+	__remove_wait_queue(&wq_head, &__wq_entry);				\
+__out:	__ret;									\
+})
+
 #define __wait_event(wq_head, condition)					\
 	(void)__wait_event_state(wq_head, condition, TASK_UNINTERRUPTIBLE)
 
@@ -767,26 +795,16 @@ do {										\
 	__ret;									\
 })
 
-extern int do_wait_intr(wait_queue_head_t *, wait_queue_entry_t *);
-extern int do_wait_intr_irq(wait_queue_head_t *, wait_queue_entry_t *);
-
-#define __wait_event_interruptible_locked(wq, condition, exclusive, fn)		\
-({										\
-	int __ret;								\
-	DEFINE_WAIT(__wait);							\
-	if (exclusive)								\
-		__wait.flags |= WQ_FLAG_EXCLUSIVE;				\
-	do {									\
-		__ret = fn(&(wq), &__wait);					\
-		if (__ret)							\
-			break;							\
-	} while (!(condition));							\
-	__remove_wait_queue(&(wq), &__wait);					\
-	__set_current_state(TASK_RUNNING);					\
-	__ret;									\
-})
+extern void do_wait_intr(wait_queue_head_t *);
+extern void do_wait_intr_irq(wait_queue_head_t *);
 
+#define __wait_event_interruptible_locked(wq_head, condition, cmd)		\
+	___wait_event_locked(wq_head, condition, TASK_INTERRUPTIBLE,		\
+			     0, 0, cmd(&wq_head))
 
+#define __wait_event_interruptible_exclusive_locked(wq_head, condition, cmd)	\
+	___wait_event_locked(wq_head, condition, TASK_INTERRUPTIBLE,		\
+			     WQ_FLAG_EXCLUSIVE, 0, cmd(&wq_head))
 /**
  * wait_event_interruptible_locked - sleep until a condition gets true
  * @wq: the waitqueue to wait on
@@ -810,9 +828,14 @@ extern int do_wait_intr_irq(wait_queue_head_t *, wait_queue_entry_t *);
  * The function will return -ERESTARTSYS if it was interrupted by a
  * signal and 0 if @condition evaluated to true.
  */
-#define wait_event_interruptible_locked(wq, condition)				\
-	((condition)								\
-	 ? 0 : __wait_event_interruptible_locked(wq, condition, 0, do_wait_intr))
+#define wait_event_interruptible_locked(wq_head, condition)			\
+({										\
+	int __ret = 0;								\
+	if (!(condition))							\
+		__ret = __wait_event_interruptible_locked(wq_head, condition,	\
+					     		  do_wait_intr);	\
+	__ret;									\
+})
 
 /**
  * wait_event_interruptible_locked_irq - sleep until a condition gets true
@@ -837,9 +860,14 @@ extern int do_wait_intr_irq(wait_queue_head_t *, wait_queue_entry_t *);
  * The function will return -ERESTARTSYS if it was interrupted by a
  * signal and 0 if @condition evaluated to true.
  */
-#define wait_event_interruptible_locked_irq(wq, condition)			\
-	((condition)								\
-	 ? 0 : __wait_event_interruptible_locked(wq, condition, 0, do_wait_intr_irq))
+#define wait_event_interruptible_locked_irq(wq_head, condition)			\
+({										\
+	int __ret = 0;								\
+	if (!(condition))							\
+		__ret = __wait_event_interruptible_locked(wq_head, condition,	\
+					     		  do_wait_intr_irq);	\
+	__ret;									\
+})
 
 /**
  * wait_event_interruptible_exclusive_locked - sleep exclusively until a condition gets true
@@ -868,9 +896,14 @@ extern int do_wait_intr_irq(wait_queue_head_t *, wait_queue_entry_t *);
  * The function will return -ERESTARTSYS if it was interrupted by a
  * signal and 0 if @condition evaluated to true.
  */
-#define wait_event_interruptible_exclusive_locked(wq, condition)		\
-	((condition)								\
-	 ? 0 : __wait_event_interruptible_locked(wq, condition, 1, do_wait_intr))
+#define wait_event_interruptible_exclusive_locked(wq_head, condition)			\
+({											\
+	int __ret = 0;									\
+	if (!(condition))								\
+		__ret = __wait_event_interruptible_exclusive_locked(wq_head, condition,	\
+					     			    do_wait_intr);	\
+	__ret;										\
+})
 
 /**
  * wait_event_interruptible_exclusive_locked_irq - sleep until a condition gets true
@@ -899,10 +932,14 @@ extern int do_wait_intr_irq(wait_queue_head_t *, wait_queue_entry_t *);
  * The function will return -ERESTARTSYS if it was interrupted by a
  * signal and 0 if @condition evaluated to true.
  */
-#define wait_event_interruptible_exclusive_locked_irq(wq, condition)		\
-	((condition)								\
-	 ? 0 : __wait_event_interruptible_locked(wq, condition, 1, do_wait_intr_irq))
-
+#define wait_event_interruptible_exclusive_locked_irq(wq_head, condition)		\
+({											\
+	int __ret = 0;									\
+	if (!(condition))								\
+		__ret = __wait_event_interruptible_exclusive_locked(wq_head, condition,	\
+					     			    do_wait_intr_irq);	\
+	__ret;										\
+})
 
 #define __wait_event_killable(wq, condition)					\
 	__wait_event_state(wq, condition, TASK_KILLABLE)
@@ -1218,6 +1255,7 @@ do {										\
  */
 void prepare_to_wait(struct wait_queue_head *wq_head, struct wait_queue_entry *wq_entry, int state);
 bool prepare_to_wait_exclusive(struct wait_queue_head *wq_head, struct wait_queue_entry *wq_entry, int state);
+long prepare_to_wait_event_locked(struct wait_queue_head *wq_head, struct wait_queue_entry *wq_entry, int state);
 long prepare_to_wait_event(struct wait_queue_head *wq_head, struct wait_queue_entry *wq_entry, int state);
 void finish_wait(struct wait_queue_head *wq_head, struct wait_queue_entry *wq_entry);
 long wait_woken(struct wait_queue_entry *wq_entry, unsigned mode, long timeout);
diff --git a/kernel/sched/wait.c b/kernel/sched/wait.c
index d0521b54eeb3..b7d03ceb80cd 100644
--- a/kernel/sched/wait.c
+++ b/kernel/sched/wait.c
@@ -269,12 +269,13 @@ void init_wait_entry(struct wait_queue_entry *wq_entry, int flags)
 }
 EXPORT_SYMBOL(init_wait_entry);
 
-long prepare_to_wait_event(struct wait_queue_head *wq_head, struct wait_queue_entry *wq_entry, int state)
+long prepare_to_wait_event_locked(struct wait_queue_head *wq_head,
+				  struct wait_queue_entry *wq_entry, int state)
 {
-	unsigned long flags;
 	long ret = 0;
 
-	spin_lock_irqsave(&wq_head->lock, flags);
+	lockdep_assert_held(&wq_head->lock);
+
 	if (signal_pending_state(state, current)) {
 		/*
 		 * Exclusive waiter must not fail if it was selected by wakeup,
@@ -299,50 +300,37 @@ long prepare_to_wait_event(struct wait_queue_head *wq_head, struct wait_queue_en
 		}
 		set_current_state(state);
 	}
-	spin_unlock_irqrestore(&wq_head->lock, flags);
 
 	return ret;
 }
+EXPORT_SYMBOL(prepare_to_wait_event_locked);
+
+long prepare_to_wait_event(struct wait_queue_head *wq_head,
+			   struct wait_queue_entry *wq_entry, int state)
+{
+	guard(spinlock_irqsave)(&wq_head->lock);
+
+	return prepare_to_wait_event_locked(wq_head, wq_entry, state);
+}
 EXPORT_SYMBOL(prepare_to_wait_event);
 
 /*
- * Note! These two wait functions are entered with the
- * wait-queue lock held (and interrupts off in the _irq
- * case), so there is no race with testing the wakeup
- * condition in the caller before they add the wait
- * entry to the wake queue.
+ * Note! These two wait functions are entered with the wait-queue
+ * lock held (and interrupts off in the _irq case).
  */
-int do_wait_intr(wait_queue_head_t *wq, wait_queue_entry_t *wait)
+void do_wait_intr(wait_queue_head_t *wq)
 {
-	if (likely(list_empty(&wait->entry)))
-		__add_wait_queue_entry_tail(wq, wait);
-
-	set_current_state(TASK_INTERRUPTIBLE);
-	if (signal_pending(current))
-		return -ERESTARTSYS;
-
 	spin_unlock(&wq->lock);
 	schedule();
 	spin_lock(&wq->lock);
-
-	return 0;
 }
 EXPORT_SYMBOL(do_wait_intr);
 
-int do_wait_intr_irq(wait_queue_head_t *wq, wait_queue_entry_t *wait)
+void do_wait_intr_irq(wait_queue_head_t *wq)
 {
-	if (likely(list_empty(&wait->entry)))
-		__add_wait_queue_entry_tail(wq, wait);
-
-	set_current_state(TASK_INTERRUPTIBLE);
-	if (signal_pending(current))
-		return -ERESTARTSYS;
-
 	spin_unlock_irq(&wq->lock);
 	schedule();
 	spin_lock_irq(&wq->lock);
-
-	return 0;
 }
 EXPORT_SYMBOL(do_wait_intr_irq);
 


      parent reply	other threads:[~2026-11-22 17:09 UTC|newest]

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-11 21:02 [PATCH] pipe: keep per-write wakeups for EPOLLEXCLUSIVE consumers Mani Tofigh via B4 Relay
2026-11-11 23:34 ` Oleg Nesterov
2026-11-12 10:26   ` Oleg Nesterov
2026-11-12 12:56 ` Oleg Nesterov
2026-11-17 13:27 ` Oleg Nesterov
2026-11-17 16:22   ` Linus Torvalds
2026-11-17 17:25     ` Oleg Nesterov
2026-11-17 17:35       ` Linus Torvalds
2026-11-17 17:58         ` Oleg Nesterov
2026-11-17 19:12           ` Linus Torvalds
2026-11-17 19:48             ` Linus Torvalds
2026-11-17 20:41               ` Oleg Nesterov
2026-11-18 13:10               ` Oleg Nesterov
2026-11-18 16:01                 ` Linus Torvalds
2026-11-18 16:30                   ` Oleg Nesterov
2026-11-18 16:39                     ` Linus Torvalds
2026-11-22 17:09                     ` Oleg Nesterov [this message]

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=awMhxEuMHULvk8SA@redhat.com \
    --to=oleg@redhat.com \
    --cc=brauner@kernel.org \
    --cc=devnull+mt4022.columbia.edu@kernel.org \
    --cc=jack@suse.cz \
    --cc=leitao@debian.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mjguzik@gmail.com \
    --cc=mt4022@columbia.edu \
    --cc=stable@vger.kernel.org \
    --cc=torvalds@linux-foundation.org \
    --cc=tz2294@columbia.edu \
    --cc=viro@zeniv.linux.org.uk \
    /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