closures: Better memory barriers

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

@@ -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)

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);