io_uring/napi: use ktime in busy polling

It's more natural to use ktime/ns instead of keeping around usec,
especially since we're comparing it against user provided timers,
so convert napi busy poll internal handling to ktime. It's also nicer
since the type (ktime_t vs unsigned long) now tells the unit of measure.

Keep everything as ktime, which we convert to/from micro seconds for
IORING_[UN]REGISTER_NAPI. The net/ busy polling works seems to work with
usec, however it's not real usec as shift by 10 is used to get it from
nsecs, see busy_loop_current_time(), so it's easy to get truncated nsec
back and we get back better precision.

Note, we can further improve it later by removing the truncation and
maybe convincing net/ to use ktime/ns instead.

Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/95e7ec8d095069a3ed5d40a4bc6f8b586698bc7e.1722003776.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>

Author: isilence

include/linux/io_uring_types.h

@@ -404,7 +404,7 @@ struct io_ring_ctx {
 	spinlock_t		napi_lock;	/* napi_list lock */
 
 	/* napi busy poll default timeout */
-	unsigned int		napi_busy_poll_to;
+	ktime_t			napi_busy_poll_dt;
 	bool			napi_prefer_busy_poll;
 	bool			napi_enabled;
 

io_uring/io_uring.h

@@ -43,7 +43,7 @@ struct io_wait_queue {
 	ktime_t timeout;
 
 #ifdef CONFIG_NET_RX_BUSY_POLL
-	unsigned int napi_busy_poll_to;
+	ktime_t napi_busy_poll_dt;
 	bool napi_prefer_busy_poll;
 #endif
 };

io_uring/napi.c

@@ -33,6 +33,12 @@ static struct io_napi_entry *io_napi_hash_find(struct hlist_head *hash_list,
 	return NULL;
 }
 
+static inline ktime_t net_to_ktime(unsigned long t)
+{
+	/* napi approximating usecs, reverse busy_loop_current_time */
+	return ns_to_ktime(t << 10);
+}
+
 void __io_napi_add(struct io_ring_ctx *ctx, struct socket *sock)
 {
 	struct hlist_head *hash_list;
@@ -102,14 +108,14 @@ static inline void io_napi_remove_stale(struct io_ring_ctx *ctx, bool is_stale)
 		__io_napi_remove_stale(ctx);
 }
 
-static inline bool io_napi_busy_loop_timeout(unsigned long start_time,
-					     unsigned long bp_usec)
+static inline bool io_napi_busy_loop_timeout(ktime_t start_time,
+					     ktime_t bp)
 {
-	if (bp_usec) {
-		unsigned long end_time = start_time + bp_usec;
-		unsigned long now = busy_loop_current_time();
+	if (bp) {
+		ktime_t end_time = ktime_add(start_time, bp);
+		ktime_t now = net_to_ktime(busy_loop_current_time());
 
-		return time_after(now, end_time);
+		return ktime_after(now, end_time);
 	}
 
 	return true;
@@ -124,7 +130,8 @@ static bool io_napi_busy_loop_should_end(void *data,
 		return true;
 	if (io_should_wake(iowq) || io_has_work(iowq->ctx))
 		return true;
-	if (io_napi_busy_loop_timeout(start_time, iowq->napi_busy_poll_to))
+	if (io_napi_busy_loop_timeout(net_to_ktime(start_time),
+				      iowq->napi_busy_poll_dt))
 		return true;
 
 	return false;
@@ -181,10 +188,12 @@ static void io_napi_blocking_busy_loop(struct io_ring_ctx *ctx,
  */
 void io_napi_init(struct io_ring_ctx *ctx)
 {
+	u64 sys_dt = READ_ONCE(sysctl_net_busy_poll) * NSEC_PER_USEC;
+
 	INIT_LIST_HEAD(&ctx->napi_list);
 	spin_lock_init(&ctx->napi_lock);
 	ctx->napi_prefer_busy_poll = false;
-	ctx->napi_busy_poll_to = READ_ONCE(sysctl_net_busy_poll);
+	ctx->napi_busy_poll_dt = ns_to_ktime(sys_dt);
 }
 
 /*
@@ -217,7 +226,7 @@ void io_napi_free(struct io_ring_ctx *ctx)
 int io_register_napi(struct io_ring_ctx *ctx, void __user *arg)
 {
 	const struct io_uring_napi curr = {
-		.busy_poll_to 	  = ctx->napi_busy_poll_to,
+		.busy_poll_to 	  = ktime_to_us(ctx->napi_busy_poll_dt),
 		.prefer_busy_poll = ctx->napi_prefer_busy_poll
 	};
 	struct io_uring_napi napi;
@@ -232,7 +241,7 @@ int io_register_napi(struct io_ring_ctx *ctx, void __user *arg)
 	if (copy_to_user(arg, &curr, sizeof(curr)))
 		return -EFAULT;
 
-	WRITE_ONCE(ctx->napi_busy_poll_to, napi.busy_poll_to);
+	WRITE_ONCE(ctx->napi_busy_poll_dt, napi.busy_poll_to * NSEC_PER_USEC);
 	WRITE_ONCE(ctx->napi_prefer_busy_poll, !!napi.prefer_busy_poll);
 	WRITE_ONCE(ctx->napi_enabled, true);
 	return 0;
@@ -249,14 +258,14 @@ int io_register_napi(struct io_ring_ctx *ctx, void __user *arg)
 int io_unregister_napi(struct io_ring_ctx *ctx, void __user *arg)
 {
 	const struct io_uring_napi curr = {
-		.busy_poll_to 	  = ctx->napi_busy_poll_to,
+		.busy_poll_to 	  = ktime_to_us(ctx->napi_busy_poll_dt),
 		.prefer_busy_poll = ctx->napi_prefer_busy_poll
 	};
 
 	if (arg && copy_to_user(arg, &curr, sizeof(curr)))
 		return -EFAULT;
 
-	WRITE_ONCE(ctx->napi_busy_poll_to, 0);
+	WRITE_ONCE(ctx->napi_busy_poll_dt, 0);
 	WRITE_ONCE(ctx->napi_prefer_busy_poll, false);
 	WRITE_ONCE(ctx->napi_enabled, false);
 	return 0;
@@ -275,23 +284,20 @@ int io_unregister_napi(struct io_ring_ctx *ctx, void __user *arg)
 void __io_napi_adjust_timeout(struct io_ring_ctx *ctx, struct io_wait_queue *iowq,
 			      struct timespec64 *ts)
 {
-	unsigned int poll_to = READ_ONCE(ctx->napi_busy_poll_to);
+	ktime_t poll_dt = READ_ONCE(ctx->napi_busy_poll_dt);
 
 	if (ts) {
 		struct timespec64 poll_to_ts;
 
-		poll_to_ts = ns_to_timespec64(1000 * (s64)poll_to);
+		poll_to_ts = ns_to_timespec64(ktime_to_ns(poll_dt));
 		if (timespec64_compare(ts, &poll_to_ts) < 0) {
 			s64 poll_to_ns = timespec64_to_ns(ts);
-			if (poll_to_ns > 0) {
-				u64 val = poll_to_ns + 999;
-				do_div(val, 1000);
-				poll_to = val;
-			}
+			if (poll_to_ns > 0)
+				poll_dt = ns_to_ktime(poll_to_ns);
 		}
 	}
 
-	iowq->napi_busy_poll_to = poll_to;
+	iowq->napi_busy_poll_dt = poll_dt;
 }
 
 /*
@@ -320,7 +326,7 @@ int io_napi_sqpoll_busy_poll(struct io_ring_ctx *ctx)
 	LIST_HEAD(napi_list);
 	bool is_stale = false;
 
-	if (!READ_ONCE(ctx->napi_busy_poll_to))
+	if (!READ_ONCE(ctx->napi_busy_poll_dt))
 		return 0;
 	if (list_empty_careful(&ctx->napi_list))
 		return 0;

io_uring/napi.h

@@ -55,7 +55,7 @@ static inline void io_napi_add(struct io_kiocb *req)
 	struct io_ring_ctx *ctx = req->ctx;
 	struct socket *sock;
 
-	if (!READ_ONCE(ctx->napi_busy_poll_to))
+	if (!READ_ONCE(ctx->napi_busy_poll_dt))
 		return;
 
 	sock = sock_from_file(req->file);