[PATCH 0/2] Two small closures patches

[PATCH 0/2] Two small closures patches

[Kent Overstreet]

I'll be sending these to Linus in my next pull request, sending them to
linux-bcache so Coly gets a chance to see them.

Kent Overstreet (2):
  closures: Better memory barriers
  closures: Fix race in closure_sync()

 fs/bcachefs/fs-io-direct.c |  1 +
 include/linux/closure.h    | 12 +++++++++---
 lib/closure.c              |  9 +++++++--
 3 files changed, 17 insertions(+), 5 deletions(-)

-- 
2.42.0

[PATCH 1/2] closures: Better memory barriers

[Kent Overstreet]

atomic_(dec|sub)_return_release() are a thing now - use them.

Also, delete the useless barrier in set_closure_fn(): it's redundant
with the memory barrier in closure_put(0.

Since closure_put() would now otherwise just have a release barrier, we
also need a new barrier when the ref hits 0 -
smp_acquire__after_ctrl_dep().

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 include/linux/closure.h | 2 --
 lib/closure.c           | 6 ++++--
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/include/linux/closure.h b/include/linux/closure.h
index 722a586bb224..bdab17050bc8 100644
--- a/include/linux/closure.h
+++ b/include/linux/closure.h
@@ -233,8 +233,6 @@ static inline void set_closure_fn(struct closure *cl, closure_fn *fn,
 	closure_set_ip(cl);
 	cl->fn = fn;
 	cl->wq = wq;
-	/* between atomic_dec() in closure_put() */
-	smp_mb__before_atomic();
 }
 
 static inline void closure_queue(struct closure *cl)
diff --git a/lib/closure.c b/lib/closure.c
index 0855e698ced1..501dfa277b59 100644
--- a/lib/closure.c
+++ b/lib/closure.c
@@ -21,6 +21,8 @@ static inline void closure_put_after_sub(struct closure *cl, int flags)
 	BUG_ON(!r && (flags & ~CLOSURE_DESTRUCTOR));
 
 	if (!r) {
+		smp_acquire__after_ctrl_dep();
+
 		if (cl->fn && !(flags & CLOSURE_DESTRUCTOR)) {
 			atomic_set(&cl->remaining,
 				   CLOSURE_REMAINING_INITIALIZER);
@@ -43,7 +45,7 @@ static inline void closure_put_after_sub(struct closure *cl, int flags)
 /* For clearing flags with the same atomic op as a put */
 void closure_sub(struct closure *cl, int v)
 {
-	closure_put_after_sub(cl, atomic_sub_return(v, &cl->remaining));
+	closure_put_after_sub(cl, atomic_sub_return_release(v, &cl->remaining));
 }
 EXPORT_SYMBOL(closure_sub);
 
@@ -52,7 +54,7 @@ EXPORT_SYMBOL(closure_sub);
  */
 void closure_put(struct closure *cl)
 {
-	closure_put_after_sub(cl, atomic_dec_return(&cl->remaining));
+	closure_put_after_sub(cl, atomic_dec_return_release(&cl->remaining));
 }
 EXPORT_SYMBOL(closure_put);
 
-- 
2.42.0

[PATCH 2/2] closures: Fix race in closure_sync()

[Kent Overstreet]

As pointed out by Linus, closure_sync() was racy; we could skip blocking
immediately after a get() and a put(), but then that would skip any
barrier corresponding to the other thread's put() barrier.

To fix this, always do the full __closure_sync() sequence whenever any
get() has happened and the closure might have been used by other
threads.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/fs-io-direct.c |  1 +
 include/linux/closure.h    | 10 +++++++++-
 lib/closure.c              |  3 +++
 3 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/fs/bcachefs/fs-io-direct.c b/fs/bcachefs/fs-io-direct.c
index 6a9557e7ecab..5b42a76c4796 100644
--- a/fs/bcachefs/fs-io-direct.c
+++ b/fs/bcachefs/fs-io-direct.c
@@ -113,6 +113,7 @@ static int bch2_direct_IO_read(struct kiocb *req, struct iov_iter *iter)
 	} else {
 		atomic_set(&dio->cl.remaining,
 			   CLOSURE_REMAINING_INITIALIZER + 1);
+		dio->cl.closure_get_happened = true;
 	}
 
 	dio->req	= req;
diff --git a/include/linux/closure.h b/include/linux/closure.h
index bdab17050bc8..de7bb47d8a46 100644
--- a/include/linux/closure.h
+++ b/include/linux/closure.h
@@ -154,6 +154,7 @@ struct closure {
 	struct closure		*parent;
 
 	atomic_t		remaining;
+	bool			closure_get_happened;
 
 #ifdef CONFIG_DEBUG_CLOSURES
 #define CLOSURE_MAGIC_DEAD	0xc054dead
@@ -185,7 +186,11 @@ static inline unsigned closure_nr_remaining(struct closure *cl)
  */
 static inline void closure_sync(struct closure *cl)
 {
-	if (closure_nr_remaining(cl) != 1)
+#ifdef CONFIG_DEBUG_CLOSURES
+	BUG_ON(closure_nr_remaining(cl) != 1 && !cl->closure_get_happened);
+#endif
+
+	if (cl->closure_get_happened)
 		__closure_sync(cl);
 }
 
@@ -257,6 +262,8 @@ static inline void closure_queue(struct closure *cl)
  */
 static inline void closure_get(struct closure *cl)
 {
+	cl->closure_get_happened = true;
+
 #ifdef CONFIG_DEBUG_CLOSURES
 	BUG_ON((atomic_inc_return(&cl->remaining) &
 		CLOSURE_REMAINING_MASK) <= 1);
@@ -279,6 +286,7 @@ static inline void closure_init(struct closure *cl, struct closure *parent)
 		closure_get(parent);
 
 	atomic_set(&cl->remaining, CLOSURE_REMAINING_INITIALIZER);
+	cl->closure_get_happened = false;
 
 	closure_debug_create(cl);
 	closure_set_ip(cl);
diff --git a/lib/closure.c b/lib/closure.c
index 501dfa277b59..f86c9eeafb35 100644
--- a/lib/closure.c
+++ b/lib/closure.c
@@ -23,6 +23,8 @@ static inline void closure_put_after_sub(struct closure *cl, int flags)
 	if (!r) {
 		smp_acquire__after_ctrl_dep();
 
+		cl->closure_get_happened = false;
+
 		if (cl->fn && !(flags & CLOSURE_DESTRUCTOR)) {
 			atomic_set(&cl->remaining,
 				   CLOSURE_REMAINING_INITIALIZER);
@@ -92,6 +94,7 @@ bool closure_wait(struct closure_waitlist *waitlist, struct closure *cl)
 	if (atomic_read(&cl->remaining) & CLOSURE_WAITING)
 		return false;
 
+	cl->closure_get_happened = true;
 	closure_set_waiting(cl, _RET_IP_);
 	atomic_add(CLOSURE_WAITING + 1, &cl->remaining);
 	llist_add(&cl->list, &waitlist->list);
-- 
2.42.0

Re: [PATCH 0/2] Two small closures patches

[Coly Li]

> 2023年11月1日 00:24，Kent Overstreet <kent.overstreet@linux.dev> 写道：
> 
> I'll be sending these to Linus in my next pull request, sending them to
> linux-bcache so Coly gets a chance to see them.
> 
> Kent Overstreet (2):
>  closures: Better memory barriers
>  closures: Fix race in closure_sync()
> 
> fs/bcachefs/fs-io-direct.c |  1 +
> include/linux/closure.h    | 12 +++++++++---
> lib/closure.c              |  9 +++++++--
> 3 files changed, 17 insertions(+), 5 deletions(-)

The whole series got tested with current bcache code, it works fine.

Acked-by: Coly Li <colyli@suse.de <mailto:colyli@suse.de>>

Thanks.

Coly Li

Re: [PATCH 0/2] Two small closures patches

[Coly Li]

> 2023年11月19日 00:43，Coly Li <colyli@suse.de> 写道：
> 
> 
> 
>> 2023年11月1日 00:24，Kent Overstreet <kent.overstreet@linux.dev> 写道：
>> 
>> I'll be sending these to Linus in my next pull request, sending them to
>> linux-bcache so Coly gets a chance to see them.
>> 
>> Kent Overstreet (2):
>> closures: Better memory barriers
>> closures: Fix race in closure_sync()
>> 
>> fs/bcachefs/fs-io-direct.c |  1 +
>> include/linux/closure.h    | 12 +++++++++---
>> lib/closure.c              |  9 +++++++--
>> 3 files changed, 17 insertions(+), 5 deletions(-)
> 
> The whole series got tested with current bcache code, it works fine.
> 
> Acked-by: Coly Li <colyli@suse.de <mailto:colyli@suse.de>>

The above line was automatically modified by the MacOS Mail client, sigh…
Please use the following line,

Acked-by: Coly Li <colyli@suse.de>


Thanks.

Coly Li