Re: [PATCH v1 net-next 4/7] af_unix: Use cached value for SOCK_STREAM in unix_inq_len().

From
Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Date
2025-07-06 14:13:19
Reply to
Kuniyuki Iwashima, 2025-07-02
Kuniyuki Iwashima wrote:
quoted hunk ↗ jump to hunk
Compared to TCP, ioctl(SIOCINQ) for AF_UNIX SOCK_STREAM socket is more
expensive, as unix_inq_len() requires iterating through the receive queue
and accumulating skb->len.

Let's cache the value for SOCK_STREAM to a new field during sendmsg()
and recvmsg().

The field is protected by the receive queue lock.

Note that ioctl(SIOCINQ) for SOCK_DGRAM returns the length of the first
skb in the queue.

SOCK_SEQPACKET still requires iterating through the queue because we do
not touch functions shared with unix_dgram_ops.  But, if really needed,
we can support it by switching __skb_try_recv_datagram() to a custom
version.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 include/net/af_unix.h |  1 +
 net/unix/af_unix.c    | 38 ++++++++++++++++++++++++++++----------
 2 files changed, 29 insertions(+), 10 deletions(-)
diff --git a/include/net/af_unix.h b/include/net/af_unix.h
index 1af1841b7601..603f8cd026e5 100644
--- a/include/net/af_unix.h
+++ b/include/net/af_unix.h
@@ -47,6 +47,7 @@ struct unix_sock {
 #define peer_wait		peer_wq.wait
 	wait_queue_entry_t	peer_wake;
 	struct scm_stat		scm_stat;
+	int			inq_len;
 #if IS_ENABLED(CONFIG_AF_UNIX_OOB)
 	struct sk_buff		*oob_skb;
 #endif
diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
index fa2081713dad..aade29d65570 100644
--- a/net/unix/af_unix.c
+++ b/net/unix/af_unix.c
@@ -2297,6 +2297,7 @@ static int queue_oob(struct sock *sk, struct msghdr *msg, struct sock *other,
 
 	spin_lock(&other->sk_receive_queue.lock);
 	WRITE_ONCE(ousk->oob_skb, skb);
+	WRITE_ONCE(ousk->inq_len, ousk->inq_len + 1);
 	__skb_queue_tail(&other->sk_receive_queue, skb);
 	spin_unlock(&other->sk_receive_queue.lock);
 
@@ -2319,6 +2320,7 @@ static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
 	struct sock *sk = sock->sk;
 	struct sk_buff *skb = NULL;
 	struct sock *other = NULL;
+	struct unix_sock *otheru;
 	struct scm_cookie scm;
 	bool fds_sent = false;
 	int err, sent = 0;
@@ -2342,14 +2344,16 @@ static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
 	if (msg->msg_namelen) {
 		err = READ_ONCE(sk->sk_state) == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
 		goto out_err;
-	} else {
-		other = unix_peer(sk);
-		if (!other) {
-			err = -ENOTCONN;
-			goto out_err;
-		}
 	}
 
+	other = unix_peer(sk);
+	if (!other) {
+		err = -ENOTCONN;
+		goto out_err;
+	}
+
+	otheru = unix_sk(other);
+
 	if (READ_ONCE(sk->sk_shutdown) & SEND_SHUTDOWN)
 		goto out_pipe;
 
@@ -2418,7 +2422,12 @@ static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
 
 		unix_maybe_add_creds(skb, sk, other);
 		scm_stat_add(other, skb);
-		skb_queue_tail(&other->sk_receive_queue, skb);
+
+		spin_lock(&other->sk_receive_queue.lock);
+		WRITE_ONCE(otheru->inq_len, otheru->inq_len + skb->len);
+		__skb_queue_tail(&other->sk_receive_queue, skb);
+		spin_unlock(&other->sk_receive_queue.lock);
+
The change from spin_lock_irqsave here and below is intentional, I
assume. If respinning, worth stating explicitly.
Thread (15 messages) 15 messages, 3 authors, last reply 2025-07-09
Keyboard shortcuts
hback out one level
jnext message in thread / next series-diff position
kprevious message in thread / previous series-diff position
ldrill in
Escclose help / fold thread tree
?toggle this help