Merge branch 'for-io_uring-add-napi-busy-polling-support'

Merge netdev bits of io_uring busy polling support.

Jens Axboe says:

====================
io_uring: add napi busy polling support

I finally got around to testing this patchset in its current form, and
results look fine to me. It Works. Using the basic ping/pong test that's
part of the liburing addition, without enabling NAPI I get:

Stock settings, no NAPI, 100k packets:

 rtt(us) min/avg/max/mdev = 31.730/37.006/87.960/0.497

 and with -t10 -b enabled:

 rtt(us) min/avg/max/mdev = 23.250/29.795/63.511/1.203

In short, this patchset enables per io_uring NAPI enablement, rather
than need to enable that globally. This allows targeted NAPI usage with
io_uring.

Here's Stefan's v15 posting, which predates this one:

https://lore.kernel.org/io-uring/20230608163839.2891748-1-shr@devkernel.io/
====================

Link: https://lore.kernel.org/r/20240206163422.646218-1-axboe@kernel.dkSigned-off-by: Jakub Kicinski <kuba@kernel.org>

include/net/busy_poll.h

@@ -48,6 +48,10 @@ void napi_busy_loop(unsigned int napi_id,
 		    bool (*loop_end)(void *, unsigned long),
 		    void *loop_end_arg, bool prefer_busy_poll, u16 budget);
 
+void napi_busy_loop_rcu(unsigned int napi_id,
+			bool (*loop_end)(void *, unsigned long),
+			void *loop_end_arg, bool prefer_busy_poll, u16 budget);
+
 #else /* CONFIG_NET_RX_BUSY_POLL */
 static inline unsigned long net_busy_loop_on(void)
 {

net/core/dev.c

@@ -6185,8 +6185,13 @@ static void __busy_poll_stop(struct napi_struct *napi, bool skip_schedule)
 	clear_bit(NAPI_STATE_SCHED, &napi->state);
 }
 
-static void busy_poll_stop(struct napi_struct *napi, void *have_poll_lock, bool prefer_busy_poll,
-			   u16 budget)
+enum {
+	NAPI_F_PREFER_BUSY_POLL	= 1,
+	NAPI_F_END_ON_RESCHED	= 2,
+};
+
+static void busy_poll_stop(struct napi_struct *napi, void *have_poll_lock,
+			   unsigned flags, u16 budget)
 {
 	bool skip_schedule = false;
 	unsigned long timeout;
@@ -6206,7 +6211,7 @@ static void busy_poll_stop(struct napi_struct *napi, void *have_poll_lock, bool
 
 	local_bh_disable();
 
-	if (prefer_busy_poll) {
+	if (flags & NAPI_F_PREFER_BUSY_POLL) {
 		napi->defer_hard_irqs_count = READ_ONCE(napi->dev->napi_defer_hard_irqs);
 		timeout = READ_ONCE(napi->dev->gro_flush_timeout);
 		if (napi->defer_hard_irqs_count && timeout) {
@@ -6230,23 +6235,23 @@ static void busy_poll_stop(struct napi_struct *napi, void *have_poll_lock, bool
 	local_bh_enable();
 }
 
-void napi_busy_loop(unsigned int napi_id,
-		    bool (*loop_end)(void *, unsigned long),
-		    void *loop_end_arg, bool prefer_busy_poll, u16 budget)
+static void __napi_busy_loop(unsigned int napi_id,
+		      bool (*loop_end)(void *, unsigned long),
+		      void *loop_end_arg, unsigned flags, u16 budget)
 {
 	unsigned long start_time = loop_end ? busy_loop_current_time() : 0;
 	int (*napi_poll)(struct napi_struct *napi, int budget);
 	void *have_poll_lock = NULL;
 	struct napi_struct *napi;
 
+	WARN_ON_ONCE(!rcu_read_lock_held());
+
 restart:
 	napi_poll = NULL;
 
-	rcu_read_lock();
-
 	napi = napi_by_id(napi_id);
 	if (!napi)
-		goto out;
+		return;
 
 	if (!IS_ENABLED(CONFIG_PREEMPT_RT))
 		preempt_disable();
@@ -6262,14 +6267,14 @@ void napi_busy_loop(unsigned int napi_id,
 			 */
 			if (val & (NAPIF_STATE_DISABLE | NAPIF_STATE_SCHED |
 				   NAPIF_STATE_IN_BUSY_POLL)) {
-				if (prefer_busy_poll)
+				if (flags & NAPI_F_PREFER_BUSY_POLL)
 					set_bit(NAPI_STATE_PREFER_BUSY_POLL, &napi->state);
 				goto count;
 			}
 			if (cmpxchg(&napi->state, val,
 				    val | NAPIF_STATE_IN_BUSY_POLL |
 					  NAPIF_STATE_SCHED) != val) {
-				if (prefer_busy_poll)
+				if (flags & NAPI_F_PREFER_BUSY_POLL)
 					set_bit(NAPI_STATE_PREFER_BUSY_POLL, &napi->state);
 				goto count;
 			}
@@ -6289,12 +6294,15 @@ void napi_busy_loop(unsigned int napi_id,
 			break;
 
 		if (unlikely(need_resched())) {
+			if (flags & NAPI_F_END_ON_RESCHED)
+				break;
 			if (napi_poll)
-				busy_poll_stop(napi, have_poll_lock, prefer_busy_poll, budget);
+				busy_poll_stop(napi, have_poll_lock, flags, budget);
 			if (!IS_ENABLED(CONFIG_PREEMPT_RT))
 				preempt_enable();
 			rcu_read_unlock();
 			cond_resched();
+			rcu_read_lock();
 			if (loop_end(loop_end_arg, start_time))
 				return;
 			goto restart;
@@ -6302,10 +6310,31 @@ void napi_busy_loop(unsigned int napi_id,
 		cpu_relax();
 	}
 	if (napi_poll)
-		busy_poll_stop(napi, have_poll_lock, prefer_busy_poll, budget);
+		busy_poll_stop(napi, have_poll_lock, flags, budget);
 	if (!IS_ENABLED(CONFIG_PREEMPT_RT))
 		preempt_enable();
-out:
+}
+
+void napi_busy_loop_rcu(unsigned int napi_id,
+			bool (*loop_end)(void *, unsigned long),
+			void *loop_end_arg, bool prefer_busy_poll, u16 budget)
+{
+	unsigned flags = NAPI_F_END_ON_RESCHED;
+
+	if (prefer_busy_poll)
+		flags |= NAPI_F_PREFER_BUSY_POLL;
+
+	__napi_busy_loop(napi_id, loop_end, loop_end_arg, flags, budget);
+}
+
+void napi_busy_loop(unsigned int napi_id,
+		    bool (*loop_end)(void *, unsigned long),
+		    void *loop_end_arg, bool prefer_busy_poll, u16 budget)
+{
+	unsigned flags = prefer_busy_poll ? NAPI_F_PREFER_BUSY_POLL : 0;
+
+	rcu_read_lock();
+	__napi_busy_loop(napi_id, loop_end, loop_end_arg, flags, budget);
 	rcu_read_unlock();
 }
 EXPORT_SYMBOL(napi_busy_loop);