net: cleanup datagram receive helpers

So the scm_stat_{add,del} helper can be invoked with no additional lock held.

This clean-up the code a bit and will make the next patch easier.

Signed-off-by Paolo Abeni <[email protected]>

---
 include/net/af_unix.h |  2 +-
 net/unix/af_unix.c    | 21 ++++++---------------
 2 files changed, 7 insertions(+), 16 deletions(-)

include/net/af_unix.h +1 -1

index 17e10fba2152..5cb65227b7a9 100644
@@ -42,7 +42,7 @@ struct unix_skb_parms {
 } __randomize_layout;
 
 struct scm_stat {
-	u32 nr_fds;
+	atomic_t nr_fds;
 };
 
 #define UNIXCB(skb)	(*(struct unix_skb_parms *)&((skb)->cb))

net/unix/af_unix.c +6 -15

index cbd7dc01e147..145a3965341e 100644
@@ -689,7 +689,8 @@ static void unix_show_fdinfo(struct seq_file *m, struct socket *sock)
 
 	if (sk) {
 		u = unix_sk(sock->sk);
-		seq_printf(m, "scm_fds: %u\n", READ_ONCE(u->scm_stat.nr_fds));
+		seq_printf(m, "scm_fds: %u\n",
+			   atomic_read(&u->scm_stat.nr_fds));
 	}
 }
 

@@ -1598,10 +1599,8 @@ static void scm_stat_add(struct sock *sk, struct sk_buff *skb)
 	struct scm_fp_list *fp = UNIXCB(skb).fp;
 	struct unix_sock *u = unix_sk(sk);
 
-	lockdep_assert_held(&sk->sk_receive_queue.lock);
-
 	if (unlikely(fp && fp->count))
-		u->scm_stat.nr_fds += fp->count;
+		atomic_add(fp->count, &u->scm_stat.nr_fds);
 }
 
 static void scm_stat_del(struct sock *sk, struct sk_buff *skb)

@@ -1609,10 +1608,8 @@ static void scm_stat_del(struct sock *sk, struct sk_buff *skb)
 	struct scm_fp_list *fp = UNIXCB(skb).fp;
 	struct unix_sock *u = unix_sk(sk);
 
-	lockdep_assert_held(&sk->sk_receive_queue.lock);
-
 	if (unlikely(fp && fp->count))
-		u->scm_stat.nr_fds -= fp->count;
+		atomic_sub(fp->count, &u->scm_stat.nr_fds);
 }
 
 /*

@@ -1801,10 +1798,8 @@ static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
 	if (sock_flag(other, SOCK_RCVTSTAMP))
 		__net_timestamp(skb);
 	maybe_add_creds(skb, sock, other);
-	spin_lock(&other->sk_receive_queue.lock);
 	scm_stat_add(other, skb);
-	__skb_queue_tail(&other->sk_receive_queue, skb);
-	spin_unlock(&other->sk_receive_queue.lock);
+	skb_queue_tail(&other->sk_receive_queue, skb);
 	unix_state_unlock(other);
 	other->sk_data_ready(other);
 	sock_put(other);

@@ -1906,10 +1901,8 @@ static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
 			goto pipe_err_free;
 
 		maybe_add_creds(skb, sock, other);
-		spin_lock(&other->sk_receive_queue.lock);
 		scm_stat_add(other, skb);
-		__skb_queue_tail(&other->sk_receive_queue, skb);
-		spin_unlock(&other->sk_receive_queue.lock);
+		skb_queue_tail(&other->sk_receive_queue, skb);
 		unix_state_unlock(other);
 		other->sk_data_ready(other);
 		sent += size;

@@ -2405,9 +2398,7 @@ static int unix_stream_read_generic(struct unix_stream_read_state *state,
 			sk_peek_offset_bwd(sk, chunk);
 
 			if (UNIXCB(skb).fp) {
-				spin_lock(&sk->sk_receive_queue.lock);
 				scm_stat_del(sk, skb);
-				spin_unlock(&sk->sk_receive_queue.lock);
 				unix_detach_fds(&scm, skb);
 			}
 

The only users for such argument are the UDP protocol and the UNIX socket family. We can safely reclaim the accounted memory directly from the UDP code and, after the previous patch, we can do scm stats accounting outside the datagram helpers.

Overall this cleans up a bit some datagram-related helpers, and avoids an indirect call per packet in the UDP receive path.

v1 -> v2:
- call scm_stat_del() only when not peeking - Kirill
- fix build issue with CONFIG_INET_ESPINTCP

Signed-off-by Paolo Abeni <[email protected]>

---
 include/linux/skbuff.h | 12 ++----------
 net/core/datagram.c    | 25 +++++++------------------
 net/ipv4/udp.c         | 14 ++++++++------
 net/unix/af_unix.c     |  7 +++++--
 net/xfrm/espintcp.c    |  2 +-
 5 files changed, 23 insertions(+), 37 deletions(-)

include/linux/skbuff.h +2 -10

index 5b50278c4bc8..21749b2cdc9b 100644
@@ -3514,23 +3514,15 @@ int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
 					  struct sk_buff_head *queue,
 					  unsigned int flags,
-					  void (*destructor)(struct sock *sk,
-							   struct sk_buff *skb),
 					  int *off, int *err,
 					  struct sk_buff **last);
 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
 					struct sk_buff_head *queue,
-					unsigned int flags,
-					void (*destructor)(struct sock *sk,
-							   struct sk_buff *skb),
-					int *off, int *err,
+					unsigned int flags, int *off, int *err,
 					struct sk_buff **last);
 struct sk_buff *__skb_recv_datagram(struct sock *sk,
 				    struct sk_buff_head *sk_queue,
-				    unsigned int flags,
-				    void (*destructor)(struct sock *sk,
-						       struct sk_buff *skb),
-				    int *off, int *err);
+				    unsigned int flags, int *off, int *err);
 struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
 				  int *err);
 __poll_t datagram_poll(struct file *file, struct socket *sock,

net/core/datagram.c +7 -18

index a78e7f864c1e..4213081c6ed3 100644
@@ -166,8 +166,6 @@ static struct sk_buff *skb_set_peeked(struct sk_buff *skb)
 struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
 					  struct sk_buff_head *queue,
 					  unsigned int flags,
-					  void (*destructor)(struct sock *sk,
-							   struct sk_buff *skb),
 					  int *off, int *err,
 					  struct sk_buff **last)
 {

@@ -198,8 +196,6 @@ struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
 			refcount_inc(&skb->users);
 		} else {
 			__skb_unlink(skb, queue);
-			if (destructor)
-				destructor(sk, skb);
 		}
 		*off = _off;
 		return skb;

@@ -212,7 +208,6 @@ struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
  *	@sk: socket
  *	@queue: socket queue from which to receive
  *	@flags: MSG\_ flags
- *	@destructor: invoked under the receive lock on successful dequeue
  *	@off: an offset in bytes to peek skb from. Returns an offset
  *	      within an skb where data actually starts
  *	@err: error code returned

@@ -245,10 +240,7 @@ struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
  */
 struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
 					struct sk_buff_head *queue,
-					unsigned int flags,
-					void (*destructor)(struct sock *sk,
-							   struct sk_buff *skb),
-					int *off, int *err,
+					unsigned int flags, int *off, int *err,
 					struct sk_buff **last)
 {
 	struct sk_buff *skb;

@@ -269,8 +261,8 @@ struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
 		 * However, this function was correct in any case. 8)
 		 */
 		spin_lock_irqsave(&queue->lock, cpu_flags);
-		skb = __skb_try_recv_from_queue(sk, queue, flags, destructor,
-						off, &error, last);
+		skb = __skb_try_recv_from_queue(sk, queue, flags, off, &error,
+						last);
 		spin_unlock_irqrestore(&queue->lock, cpu_flags);
 		if (error)
 			goto no_packet;

@@ -293,10 +285,7 @@ EXPORT_SYMBOL(__skb_try_recv_datagram);
 
 struct sk_buff *__skb_recv_datagram(struct sock *sk,
 				    struct sk_buff_head *sk_queue,
-				    unsigned int flags,
-				    void (*destructor)(struct sock *sk,
-						       struct sk_buff *skb),
-				    int *off, int *err)
+				    unsigned int flags, int *off, int *err)
 {
 	struct sk_buff *skb, *last;
 	long timeo;

@@ -304,8 +293,8 @@ struct sk_buff *__skb_recv_datagram(struct sock *sk,
 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
 
 	do {
-		skb = __skb_try_recv_datagram(sk, sk_queue, flags, destructor,
-					      off, err, &last);
+		skb = __skb_try_recv_datagram(sk, sk_queue, flags, off, err,
+					      &last);
 		if (skb)
 			return skb;
 

@@ -326,7 +315,7 @@ struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags,
 
 	return __skb_recv_datagram(sk, &sk->sk_receive_queue,
 				   flags | (noblock ? MSG_DONTWAIT : 0),
-				   NULL, &off, err);
+				   &off, err);
 }
 EXPORT_SYMBOL(skb_recv_datagram);
 

net/ipv4/udp.c +8 -6

index 08a41f1e1cd2..a68e2ac37f26 100644
@@ -1671,10 +1671,11 @@ struct sk_buff *__skb_recv_udp(struct sock *sk, unsigned int flags,
 		error = -EAGAIN;
 		do {
 			spin_lock_bh(&queue->lock);
-			skb = __skb_try_recv_from_queue(sk, queue, flags,
-							udp_skb_destructor,
-							off, err, &last);
+			skb = __skb_try_recv_from_queue(sk, queue, flags, off,
+							err, &last);
 			if (skb) {
+				if (!(flags & MSG_PEEK))
+					udp_skb_destructor(sk, skb);
 				spin_unlock_bh(&queue->lock);
 				return skb;
 			}

@@ -1692,9 +1693,10 @@ struct sk_buff *__skb_recv_udp(struct sock *sk, unsigned int flags,
 			spin_lock(&sk_queue->lock);
 			skb_queue_splice_tail_init(sk_queue, queue);
 
-			skb = __skb_try_recv_from_queue(sk, queue, flags,
-							udp_skb_dtor_locked,
-							off, err, &last);
+			skb = __skb_try_recv_from_queue(sk, queue, flags, off,
+							err, &last);
+			if (skb && !(flags & MSG_PEEK))
+				udp_skb_dtor_locked(sk, skb);
 			spin_unlock(&sk_queue->lock);
 			spin_unlock_bh(&queue->lock);
 			if (skb)

net/unix/af_unix.c +5 -2

index 145a3965341e..103bf94d2ab5 100644
@@ -2102,9 +2102,12 @@ static int unix_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
 
 		skip = sk_peek_offset(sk, flags);
 		skb = __skb_try_recv_datagram(sk, &sk->sk_receive_queue, flags,
-					      scm_stat_del, &skip, &err, &last);
-		if (skb)
+					      &skip, &err, &last);
+		if (skb) {
+			if (!(flags & MSG_PEEK))
+				scm_stat_del(sk, skb);
 			break;
+		}
 
 		mutex_unlock(&u->iolock);
 

net/xfrm/espintcp.c +1 -1

index f15d6a564b0e..037ea156d2f9 100644
@@ -100,7 +100,7 @@ static int espintcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 
 	flags |= nonblock ? MSG_DONTWAIT : 0;
 
-	skb = __skb_recv_datagram(sk, &ctx->ike_queue, flags, NULL, &off, &err);
+	skb = __skb_recv_datagram(sk, &ctx->ike_queue, flags, &off, &err);
 	if (!skb)
 		return err;
 

Quoted 8 lines · Paolo Abeni

The only users for such argument are the UDP protocol and the UNIX socket family. We can safely reclaim the accounted memory directly from the UDP code and, after the previous patch, we can do scm stats accounting outside the datagram helpers.

Overall this cleans up a bit some datagram-related helpers, and avoids an indirect call per packet in the UDP receive path.

Quoted 5 lines

v1 -> v2:
- call scm_stat_del() only when not peeking - Kirill
- fix build issue with CONFIG_INET_ESPINTCP

Signed-off-by Paolo Abeni <[email protected]>

Reviewed-by Kirill Tkhai <[email protected]>

Several receive helpers have an optional destructor argument, which uglify the code a bit and is taxed by retpoline overhead.

This series refactor the code so that we can drop such optional argument, cleaning the helpers a bit and avoiding an indirect call in fast path.

The first patch refactor a bit the caller, so that the second patch actually dropping the argument is more straight-forward

v1 -> v2:
- call scm_stat_del() only when not peeking - Kirill
- fix build issue with CONFIG_INET_ESPINTCP