Skip to contentSkip to conversation

net: various KCSAN inspired fixes

PATCH In review Eric Dumazet 2 participants 8 messages Nov 5, 2019 – Nov 7, 2019

Subsystems (16)
lore
Share & tools
lore
#0 lore raw

cover net: various KCSAN inspired fixes

This is a series of minor fixes, mostly dealing with lockless accesses to socket 'sk_ack_backlog', 'sk_max_ack_backlog' ane neighbour 'confirmed' fields.

Eric Dumazet (6):
net: neigh: use long type to store jiffies delta
inet_diag: use jiffies_delta_to_msecs()
net: avoid potential false sharing in neighbor related code
net: use helpers to change sk_ack_backlog
net: annotate lockless accesses to sk->sk_ack_backlog
net: annotate lockless accesses to sk->sk_max_ack_backlog

 include/net/arp.h                       |  4 ++--
 include/net/ndisc.h                     |  8 ++++----
 include/net/sock.h                      | 18 +++++++++---------
 net/atm/signaling.c                     |  2 +-
 net/atm/svc.c                           |  2 +-
 net/ax25/af_ax25.c                      |  2 +-
 net/ax25/ax25_in.c                      |  2 +-
 net/bluetooth/af_bluetooth.c            |  4 ++--
 net/core/neighbour.c                    |  4 ++--
 net/dccp/proto.c                        |  2 +-
 net/decnet/af_decnet.c                  |  2 +-
 net/decnet/dn_nsp_in.c                  |  2 +-
 net/ipv4/af_inet.c                      |  2 +-
 net/ipv4/inet_connection_sock.c         |  2 +-
 net/ipv4/inet_diag.c                    | 15 ++++++---------
 net/ipv4/tcp.c                          |  4 ++--
 net/ipv4/tcp_diag.c                     |  4 ++--
 net/ipv4/tcp_ipv4.c                     |  2 +-
 net/ipv6/tcp_ipv6.c                     |  2 +-
 net/llc/af_llc.c                        |  2 +-
 net/rose/af_rose.c                      |  4 ++--
 net/sched/em_meta.c                     |  4 ++--
 net/sctp/associola.c                    |  4 ++--
 net/sctp/diag.c                         |  4 ++--
 net/sctp/endpointola.c                  |  2 +-
 net/sctp/socket.c                       |  4 ++--
 net/vmw_vsock/af_vsock.c                |  4 ++--
 net/vmw_vsock/hyperv_transport.c        |  2 +-
 net/vmw_vsock/virtio_transport_common.c |  2 +-
 net/vmw_vsock/vmci_transport.c          |  2 +-
 net/x25/af_x25.c                        |  4 ++--
 31 files changed, 59 insertions(+), 62 deletions(-)
Signature · 1 line
2.24.0.rc1.363.gb1bccd3e3d-goog
#1 lore raw

PATCH 1/6 net: neigh: use long type to store jiffies delta

A difference of two unsigned long needs long storage.

Fixes c7fb64db001f ("[NETLINK]: Neighbour table configuration and statistics via rtnetlink")
---
 net/core/neighbour.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
net/core/neighbour.c +2 -2
index 5480edff0c8687a0b3a0e2dca90a83dc7ca52fe8..8c82e95f75394cba3688bc2a960804b1f967f508 100644
@@ -2052,8 +2052,8 @@ static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
 		goto nla_put_failure;
 	{
 		unsigned long now = jiffies;
-		unsigned int flush_delta = now - tbl->last_flush;
-		unsigned int rand_delta = now - tbl->last_rand;
+		long flush_delta = now - tbl->last_flush;
+		long rand_delta = now - tbl->last_rand;
 		struct neigh_hash_table *nht;
 		struct ndt_config ndc = {
 			.ndtc_key_len		= tbl->key_len,
Signature · 1 line
2.24.0.rc1.363.gb1bccd3e3d-goog
#2 lore raw

PATCH 2/6 inet_diag: use jiffies_delta_to_msecs()

In reply to Eric Dumazet

Use jiffies_delta_to_msecs() to avoid reporting 'infinite' timeouts and to cleanup code.

---
 net/ipv4/inet_diag.c | 15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)
net/ipv4/inet_diag.c +6 -9
index 7dc79b973e6edcc64e668e14c71c732ca1187e8f..af154977904c0c249e77e425990a09c62cca4251 100644
@@ -226,17 +226,17 @@ int inet_sk_diag_fill(struct sock *sk, struct inet_connection_sock *icsk,
 		r->idiag_timer = 1;
 		r->idiag_retrans = icsk->icsk_retransmits;
 		r->idiag_expires =
-			jiffies_to_msecs(icsk->icsk_timeout - jiffies);
+			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
 	} else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
 		r->idiag_timer = 4;
 		r->idiag_retrans = icsk->icsk_probes_out;
 		r->idiag_expires =
-			jiffies_to_msecs(icsk->icsk_timeout - jiffies);
+			jiffies_delta_to_msecs(icsk->icsk_timeout - jiffies);
 	} else if (timer_pending(&sk->sk_timer)) {
 		r->idiag_timer = 2;
 		r->idiag_retrans = icsk->icsk_probes_out;
 		r->idiag_expires =
-			jiffies_to_msecs(sk->sk_timer.expires - jiffies);
+			jiffies_delta_to_msecs(sk->sk_timer.expires - jiffies);
 	} else {
 		r->idiag_timer = 0;
 		r->idiag_expires = 0;
@@ -342,16 +342,13 @@ static int inet_twsk_diag_fill(struct sock *sk,
 	r = nlmsg_data(nlh);
 	BUG_ON(tw->tw_state != TCP_TIME_WAIT);
 
-	tmo = tw->tw_timer.expires - jiffies;
-	if (tmo < 0)
-		tmo = 0;
-
 	inet_diag_msg_common_fill(r, sk);
 	r->idiag_retrans      = 0;
 
 	r->idiag_state	      = tw->tw_substate;
 	r->idiag_timer	      = 3;
-	r->idiag_expires      = jiffies_to_msecs(tmo);
+	tmo = tw->tw_timer.expires - jiffies;
+	r->idiag_expires      = jiffies_delta_to_msecs(tmo);
 	r->idiag_rqueue	      = 0;
 	r->idiag_wqueue	      = 0;
 	r->idiag_uid	      = 0;
@@ -385,7 +382,7 @@ static int inet_req_diag_fill(struct sock *sk, struct sk_buff *skb,
 		     offsetof(struct sock, sk_cookie));
 
 	tmo = inet_reqsk(sk)->rsk_timer.expires - jiffies;
-	r->idiag_expires = (tmo >= 0) ? jiffies_to_msecs(tmo) : 0;
+	r->idiag_expires = jiffies_delta_to_msecs(tmo);
 	r->idiag_rqueue	= 0;
 	r->idiag_wqueue	= 0;
 	r->idiag_uid	= 0;
Signature · 1 line
2.24.0.rc1.363.gb1bccd3e3d-goog
#3 lore raw

PATCH 3/6 net: avoid potential false sharing in neighbor related code

In reply to Eric Dumazet

There are common instances of the following construct :

	if (n->confirmed != now)
		n->confirmed = now;

A C compiler could legally remove the conditional.

Use READ_ONCE()/WRITE_ONCE() to avoid this problem.

---
 include/net/arp.h   |  4 ++--
 include/net/ndisc.h |  8 ++++----
 include/net/sock.h  | 12 ++++++------
 3 files changed, 12 insertions(+), 12 deletions(-)
include/net/arp.h +2 -2
index c8f580a0e6b1f5c0853cda1605336fa8eb90917c..4950191f6b2bf424519c7ecf3483336739fea143 100644
@@ -57,8 +57,8 @@ static inline void __ipv4_confirm_neigh(struct net_device *dev, u32 key)
 		unsigned long now = jiffies;
 
 		/* avoid dirtying neighbour */
-		if (n->confirmed != now)
-			n->confirmed = now;
+		if (READ_ONCE(n->confirmed) != now)
+			WRITE_ONCE(n->confirmed, now);
 	}
 	rcu_read_unlock_bh();
 }
include/net/ndisc.h +4 -4
index b2f715ca056726d75033083d1bccee693cea9672..b5ebeb3b0de0e9de3fa495148ab19830d9767e89 100644
@@ -414,8 +414,8 @@ static inline void __ipv6_confirm_neigh(struct net_device *dev,
 		unsigned long now = jiffies;
 
 		/* avoid dirtying neighbour */
-		if (n->confirmed != now)
-			n->confirmed = now;
+		if (READ_ONCE(n->confirmed) != now)
+			WRITE_ONCE(n->confirmed, now);
 	}
 	rcu_read_unlock_bh();
 }
@@ -431,8 +431,8 @@ static inline void __ipv6_confirm_neigh_stub(struct net_device *dev,
 		unsigned long now = jiffies;
 
 		/* avoid dirtying neighbour */
-		if (n->confirmed != now)
-			n->confirmed = now;
+		if (READ_ONCE(n->confirmed) != now)
+			WRITE_ONCE(n->confirmed, now);
 	}
 	rcu_read_unlock_bh();
 }
include/net/sock.h +6 -6
index ac6042d0af328a7f6e18f8882d2ee0679a0e59c5..f2f853439b6576925e39f6db010964762e39ccf2 100644
@@ -1939,8 +1939,8 @@ struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
 
 static inline void sk_dst_confirm(struct sock *sk)
 {
-	if (!sk->sk_dst_pending_confirm)
-		sk->sk_dst_pending_confirm = 1;
+	if (!READ_ONCE(sk->sk_dst_pending_confirm))
+		WRITE_ONCE(sk->sk_dst_pending_confirm, 1);
 }
 
 static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
@@ -1950,10 +1950,10 @@ static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
 		unsigned long now = jiffies;
 
 		/* avoid dirtying neighbour */
-		if (n->confirmed != now)
-			n->confirmed = now;
-		if (sk && sk->sk_dst_pending_confirm)
-			sk->sk_dst_pending_confirm = 0;
+		if (READ_ONCE(n->confirmed) != now)
+			WRITE_ONCE(n->confirmed, now);
+		if (sk && READ_ONCE(sk->sk_dst_pending_confirm))
+			WRITE_ONCE(sk->sk_dst_pending_confirm, 0);
 	}
 }
 
Signature · 1 line
2.24.0.rc1.363.gb1bccd3e3d-goog
#4 lore raw

PATCH 4/6 net: use helpers to change sk_ack_backlog

In reply to Eric Dumazet

Writers are holding a lock, but many readers do not.

Following patch will add appropriate barriers in
sk_acceptq_removed() and sk_acceptq_added().

---
 net/atm/signaling.c                     | 2 +-
 net/atm/svc.c                           | 2 +-
 net/ax25/af_ax25.c                      | 2 +-
 net/ax25/ax25_in.c                      | 2 +-
 net/bluetooth/af_bluetooth.c            | 4 ++--
 net/decnet/af_decnet.c                  | 2 +-
 net/decnet/dn_nsp_in.c                  | 2 +-
 net/llc/af_llc.c                        | 2 +-
 net/rose/af_rose.c                      | 4 ++--
 net/sctp/associola.c                    | 4 ++--
 net/sctp/endpointola.c                  | 2 +-
 net/vmw_vsock/af_vsock.c                | 4 ++--
 net/vmw_vsock/hyperv_transport.c        | 2 +-
 net/vmw_vsock/virtio_transport_common.c | 2 +-
 net/vmw_vsock/vmci_transport.c          | 2 +-
 net/x25/af_x25.c                        | 4 ++--
 16 files changed, 21 insertions(+), 21 deletions(-)
net/atm/signaling.c +1 -1
index 6c11cdf4dd4ce4737dd38bcceccc43dfd272c338..fbd0c5e7b299993bd3a4ae4a6c5f543cb30c5091 100644
@@ -109,7 +109,7 @@ static int sigd_send(struct atm_vcc *vcc, struct sk_buff *skb)
 			dev_kfree_skb(skb);
 			goto as_indicate_complete;
 		}
-		sk->sk_ack_backlog++;
+		sk_acceptq_added(sk);
 		skb_queue_tail(&sk->sk_receive_queue, skb);
 		pr_debug("waking sk_sleep(sk) 0x%p\n", sk_sleep(sk));
 		sk->sk_state_change(sk);
net/atm/svc.c +1 -1
index 908cbb8654f532548b20f166492a2e02de6324d6..ba144d035e3d41e8ba8115b9b4aa54762fca0d01 100644
@@ -381,7 +381,7 @@ static int svc_accept(struct socket *sock, struct socket *newsock, int flags,
 				    msg->pvc.sap_addr.vpi,
 				    msg->pvc.sap_addr.vci);
 		dev_kfree_skb(skb);
-		sk->sk_ack_backlog--;
+		sk_acceptq_removed(sk);
 		if (error) {
 			sigd_enq2(NULL, as_reject, old_vcc, NULL, NULL,
 				  &old_vcc->qos, error);
net/ax25/af_ax25.c +1 -1
index bb222b882b6776481a17c9a4cc9ed6d97150c986..324306d6fde02cb67e13073986bde1c4ed7eae5f 100644
@@ -1384,7 +1384,7 @@ static int ax25_accept(struct socket *sock, struct socket *newsock, int flags,
 
 	/* Now attach up the new socket */
 	kfree_skb(skb);
-	sk->sk_ack_backlog--;
+	sk_acceptq_removed(sk);
 	newsock->state = SS_CONNECTED;
 
 out:
net/ax25/ax25_in.c +1 -1
index dcdbaeeb2358a45d18ca0987ae5969b8f9a76cad..cd6afe895db9910fb7f8f1abad9016307c660850 100644
@@ -356,7 +356,7 @@ static int ax25_rcv(struct sk_buff *skb, struct net_device *dev,
 
 		make->sk_state = TCP_ESTABLISHED;
 
-		sk->sk_ack_backlog++;
+		sk_acceptq_added(sk);
 		bh_unlock_sock(sk);
 	} else {
 		if (!mine)
net/bluetooth/af_bluetooth.c +2 -2
index 5f508c50649d0abc317ed83cc0768f6b7a64de96..3fd124927d4d176864004aecdd206407a1f6b836 100644
@@ -173,7 +173,7 @@ void bt_accept_enqueue(struct sock *parent, struct sock *sk, bool bh)
 	else
 		release_sock(sk);
 
-	parent->sk_ack_backlog++;
+	sk_acceptq_added(parent);
 }
 EXPORT_SYMBOL(bt_accept_enqueue);
 
@@ -185,7 +185,7 @@ void bt_accept_unlink(struct sock *sk)
 	BT_DBG("sk %p state %d", sk, sk->sk_state);
 
 	list_del_init(&bt_sk(sk)->accept_q);
-	bt_sk(sk)->parent->sk_ack_backlog--;
+	sk_acceptq_removed(bt_sk(sk)->parent);
 	bt_sk(sk)->parent = NULL;
 	sock_put(sk);
 }
net/decnet/af_decnet.c +1 -1
index 3349ea81f9016fb785ec888fadf86faa4d859ed7..e19a92a62e142611f78d942ba22d2b1908430660 100644
@@ -1091,7 +1091,7 @@ static int dn_accept(struct socket *sock, struct socket *newsock, int flags,
 	}
 
 	cb = DN_SKB_CB(skb);
-	sk->sk_ack_backlog--;
+	sk_acceptq_removed(sk);
 	newsk = dn_alloc_sock(sock_net(sk), newsock, sk->sk_allocation, kern);
 	if (newsk == NULL) {
 		release_sock(sk);
net/decnet/dn_nsp_in.c +1 -1
index e4161e0c86aa1190bdba483f857922ef12ce0bb6..c68503a180259467505bd1346995ea1bc00df755 100644
@@ -328,7 +328,7 @@ static void dn_nsp_conn_init(struct sock *sk, struct sk_buff *skb)
 		return;
 	}
 
-	sk->sk_ack_backlog++;
+	sk_acceptq_added(sk);
 	skb_queue_tail(&sk->sk_receive_queue, skb);
 	sk->sk_state_change(sk);
 }
net/llc/af_llc.c +1 -1
index c74f44dfaa22a5020880ea218c6607ace8fd5e22..50d2c9749db36da84f0e84c254771ee5e6c9cef9 100644
@@ -705,7 +705,7 @@ static int llc_ui_accept(struct socket *sock, struct socket *newsock, int flags,
 
 	/* put original socket back into a clean listen state. */
 	sk->sk_state = TCP_LISTEN;
-	sk->sk_ack_backlog--;
+	sk_acceptq_removed(sk);
 	dprintk("%s: ok success on %02X, client on %02X\n", __func__,
 		llc_sk(sk)->addr.sllc_sap, newllc->daddr.lsap);
 frees:
net/rose/af_rose.c +2 -2
index 6a0df7c8a939e4976aa2815127885e127dd864fc..46b8ff24020d7bb6788e985686bef12b4fbc83d0 100644
@@ -906,7 +906,7 @@ static int rose_accept(struct socket *sock, struct socket *newsock, int flags,
 	/* Now attach up the new socket */
 	skb->sk = NULL;
 	kfree_skb(skb);
-	sk->sk_ack_backlog--;
+	sk_acceptq_removed(sk);
 
 out_release:
 	release_sock(sk);
@@ -1011,7 +1011,7 @@ int rose_rx_call_request(struct sk_buff *skb, struct net_device *dev, struct ros
 	make_rose->va        = 0;
 	make_rose->vr        = 0;
 	make_rose->vl        = 0;
-	sk->sk_ack_backlog++;
+	sk_acceptq_added(sk);
 
 	rose_insert_socket(make);
 
net/sctp/associola.c +2 -2
index 1ba893b85dad79786ec3fd55c2870072c6c4efa9..1b9809ad772528b3596efc9ae96780dbc70f7cc2 100644
@@ -324,7 +324,7 @@ void sctp_association_free(struct sctp_association *asoc)
 		 * socket.
 		 */
 		if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
-			sk->sk_ack_backlog--;
+			sk_acceptq_removed(sk);
 	}
 
 	/* Mark as dead, so other users can know this structure is
@@ -1073,7 +1073,7 @@ void sctp_assoc_migrate(struct sctp_association *assoc, struct sock *newsk)
 
 	/* Decrement the backlog value for a TCP-style socket. */
 	if (sctp_style(oldsk, TCP))
-		oldsk->sk_ack_backlog--;
+		sk_acceptq_removed(oldsk);
 
 	/* Release references to the old endpoint and the sock.  */
 	sctp_endpoint_put(assoc->ep);
net/sctp/endpointola.c +1 -1
index ea53049d1db663e63def7f5e7e22e788c7456d22..9d05b2e7bce24cad0633efd54ca7223b0426de9a 100644
@@ -164,7 +164,7 @@ void sctp_endpoint_add_asoc(struct sctp_endpoint *ep,
 
 	/* Increment the backlog value for a TCP-style listening socket. */
 	if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
-		sk->sk_ack_backlog++;
+		sk_acceptq_added(sk);
 }
 
 /* Free the endpoint structure.  Delay cleanup until
net/vmw_vsock/af_vsock.c +2 -2
index 582a3e4dfce295fa67100468335bc8dbf8dc1095..e4da879161b74366dfccfec296f3f68a3626f35c 100644
@@ -439,7 +439,7 @@ static void vsock_pending_work(struct work_struct *work)
 	if (vsock_is_pending(sk)) {
 		vsock_remove_pending(listener, sk);
 
-		listener->sk_ack_backlog--;
+		sk_acceptq_removed(listener);
 	} else if (!vsk->rejected) {
 		/* We are not on the pending list and accept() did not reject
 		 * us, so we must have been accepted by our user process.  We
@@ -1301,7 +1301,7 @@ static int vsock_accept(struct socket *sock, struct socket *newsock, int flags,
 		err = -listener->sk_err;
 
 	if (connected) {
-		listener->sk_ack_backlog--;
+		sk_acceptq_removed(listener);
 
 		lock_sock_nested(connected, SINGLE_DEPTH_NESTING);
 		vconnected = vsock_sk(connected);
net/vmw_vsock/hyperv_transport.c +1 -1
index bef8772116ec82dc43df0120b3eb6987d111b858..7fa09c5e4625ca801ef3812d8c8cc836f38b08e9 100644
@@ -428,7 +428,7 @@ static void hvs_open_connection(struct vmbus_channel *chan)
 
 	if (conn_from_host) {
 		new->sk_state = TCP_ESTABLISHED;
-		sk->sk_ack_backlog++;
+		sk_acceptq_added(sk);
 
 		hvs_addr_init(&vnew->local_addr, if_type);
 		hvs_remote_addr_init(&vnew->remote_addr, &vnew->local_addr);
net/vmw_vsock/virtio_transport_common.c +1 -1
index d02c9b41a768e6a9a968e1b685a8ed2f390212e6..193f959e51efee2fb9ae8d8ed0a7c8d3f894a58d 100644
@@ -1066,7 +1066,7 @@ virtio_transport_recv_listen(struct sock *sk, struct virtio_vsock_pkt *pkt)
 		return -ENOMEM;
 	}
 
-	sk->sk_ack_backlog++;
+	sk_acceptq_added(sk);
 
 	lock_sock_nested(child, SINGLE_DEPTH_NESTING);
 
net/vmw_vsock/vmci_transport.c +1 -1
index 8c9c4ed90fa707b93208742a808253b81f67f489..6ba98a1efe2e08de7a99e92f3ee6a9a39e3f9086 100644
@@ -1098,7 +1098,7 @@ static int vmci_transport_recv_listen(struct sock *sk,
 	}
 
 	vsock_add_pending(sk, pending);
-	sk->sk_ack_backlog++;
+	sk_acceptq_added(sk);
 
 	pending->sk_state = TCP_SYN_SENT;
 	vmci_trans(vpending)->produce_size =
net/x25/af_x25.c +2 -2
index 6aee9f5e8e7155461eac93074ba975bac3af08c7..c34f7d0776046f2c15b668b66f020be8008ea731 100644
@@ -891,7 +891,7 @@ static int x25_accept(struct socket *sock, struct socket *newsock, int flags,
 	/* Now attach up the new socket */
 	skb->sk = NULL;
 	kfree_skb(skb);
-	sk->sk_ack_backlog--;
+	sk_acceptq_removed(sk);
 	newsock->state = SS_CONNECTED;
 	rc = 0;
 out2:
@@ -1062,7 +1062,7 @@ int x25_rx_call_request(struct sk_buff *skb, struct x25_neigh *nb,
 	skb_copy_from_linear_data(skb, makex25->calluserdata.cuddata, skb->len);
 	makex25->calluserdata.cudlength = skb->len;
 
-	sk->sk_ack_backlog++;
+	sk_acceptq_added(sk);
 
 	x25_insert_socket(make);
 
Signature · 1 line
2.24.0.rc1.363.gb1bccd3e3d-goog
#5 lore raw

PATCH 5/6 net: annotate lockless accesses to sk->sk_ack_backlog

In reply to Eric Dumazet

sk->sk_ack_backlog can be read without any lock being held. We need to use READ_ONCE()/WRITE_ONCE() to avoid load/store tearing and/or potential KCSAN warnings.

---
 include/net/sock.h  | 6 +++---
 net/ipv4/tcp.c      | 2 +-
 net/ipv4/tcp_diag.c | 2 +-
 net/ipv4/tcp_ipv4.c | 2 +-
 net/ipv6/tcp_ipv6.c | 2 +-
 net/sched/em_meta.c | 2 +-
 net/sctp/diag.c     | 2 +-
 7 files changed, 9 insertions(+), 9 deletions(-)
include/net/sock.h +3 -3
index f2f853439b6576925e39f6db010964762e39ccf2..a126784aa7d9b6f59c8937c8c94d5bd7843988a4 100644
@@ -859,17 +859,17 @@ static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
 
 static inline void sk_acceptq_removed(struct sock *sk)
 {
-	sk->sk_ack_backlog--;
+	WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog - 1);
 }
 
 static inline void sk_acceptq_added(struct sock *sk)
 {
-	sk->sk_ack_backlog++;
+	WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog + 1);
 }
 
 static inline bool sk_acceptq_is_full(const struct sock *sk)
 {
-	return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
+	return READ_ONCE(sk->sk_ack_backlog) > sk->sk_max_ack_backlog;
 }
 
 /*
net/ipv4/tcp.c +1 -1
index 1dd25189d83f2c7404336f8378be23c4beaa7ed7..68375f7ffdce1fbbb4cf443660703c98b61fd9e3 100644
@@ -3225,7 +3225,7 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
 		 * tcpi_unacked -> Number of children ready for accept()
 		 * tcpi_sacked  -> max backlog
 		 */
-		info->tcpi_unacked = sk->sk_ack_backlog;
+		info->tcpi_unacked = READ_ONCE(sk->sk_ack_backlog);
 		info->tcpi_sacked = sk->sk_max_ack_backlog;
 		return;
 	}
net/ipv4/tcp_diag.c +1 -1
index 549506162ddeca22f6dd87dfe1c5c13cea6e2b69..edfbab54c46f4cac1b0a7960718d0b6308978957 100644
@@ -21,7 +21,7 @@ static void tcp_diag_get_info(struct sock *sk, struct inet_diag_msg *r,
 	struct tcp_info *info = _info;
 
 	if (inet_sk_state_load(sk) == TCP_LISTEN) {
-		r->idiag_rqueue = sk->sk_ack_backlog;
+		r->idiag_rqueue = READ_ONCE(sk->sk_ack_backlog);
 		r->idiag_wqueue = sk->sk_max_ack_backlog;
 	} else if (sk->sk_type == SOCK_STREAM) {
 		const struct tcp_sock *tp = tcp_sk(sk);
net/ipv4/tcp_ipv4.c +1 -1
index 899e100a68e6ab8fcf7b2c4d2a9d179745a782b5..92282f98dc82290bfaf53acc050182e4cc3be1eb 100644
@@ -2451,7 +2451,7 @@ static void get_tcp4_sock(struct sock *sk, struct seq_file *f, int i)
 
 	state = inet_sk_state_load(sk);
 	if (state == TCP_LISTEN)
-		rx_queue = sk->sk_ack_backlog;
+		rx_queue = READ_ONCE(sk->sk_ack_backlog);
 	else
 		/* Because we don't lock the socket,
 		 * we might find a transient negative value.
net/ipv6/tcp_ipv6.c +1 -1
index 4804b6dc5e6519a457e631bc1438a14f85477567..81f51335e326fad57d3e0e1ce23926b276e95e92 100644
@@ -1891,7 +1891,7 @@ static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
 
 	state = inet_sk_state_load(sp);
 	if (state == TCP_LISTEN)
-		rx_queue = sp->sk_ack_backlog;
+		rx_queue = READ_ONCE(sp->sk_ack_backlog);
 	else
 		/* Because we don't lock the socket,
 		 * we might find a transient negative value.
net/sched/em_meta.c +1 -1
index 3177dcb173161629a801278db38fabeb6fcdbdd9..ebb6e2430861d23a42431e4143f229395d9321c5 100644
@@ -521,7 +521,7 @@ META_COLLECTOR(int_sk_ack_bl)
 		*err = -1;
 		return;
 	}
-	dst->value = sk->sk_ack_backlog;
+	dst->value = READ_ONCE(sk->sk_ack_backlog);
 }
 
 META_COLLECTOR(int_sk_max_ack_bl)
net/sctp/diag.c +1 -1
index 0851166b917597b08becf9bf9d5873287b375828..f873f15407de4e7d9a246d41e07602f33da8064d 100644
@@ -425,7 +425,7 @@ static void sctp_diag_get_info(struct sock *sk, struct inet_diag_msg *r,
 		r->idiag_rqueue = atomic_read(&infox->asoc->rmem_alloc);
 		r->idiag_wqueue = infox->asoc->sndbuf_used;
 	} else {
-		r->idiag_rqueue = sk->sk_ack_backlog;
+		r->idiag_rqueue = READ_ONCE(sk->sk_ack_backlog);
 		r->idiag_wqueue = sk->sk_max_ack_backlog;
 	}
 	if (infox->sctpinfo)
Signature · 1 line
2.24.0.rc1.363.gb1bccd3e3d-goog
#6 lore raw

PATCH 6/6 net: annotate lockless accesses to sk->sk_max_ack_backlog

In reply to Eric Dumazet

sk->sk_max_ack_backlog can be read without any lock being held at least in TCP/DCCP cases.

We need to use READ_ONCE()/WRITE_ONCE() to avoid load/store tearing and/or potential KCSAN warnings.

---
 include/net/sock.h              | 2 +-
 net/dccp/proto.c                | 2 +-
 net/ipv4/af_inet.c              | 2 +-
 net/ipv4/inet_connection_sock.c | 2 +-
 net/ipv4/tcp.c                  | 2 +-
 net/ipv4/tcp_diag.c             | 2 +-
 net/sched/em_meta.c             | 2 +-
 net/sctp/diag.c                 | 2 +-
 net/sctp/socket.c               | 4 ++--
 9 files changed, 10 insertions(+), 10 deletions(-)
include/net/sock.h +1 -1
index a126784aa7d9b6f59c8937c8c94d5bd7843988a4..d4d3ef5ba0490366e1e25884a5edf54186c940d8 100644
@@ -869,7 +869,7 @@ static inline void sk_acceptq_added(struct sock *sk)
 
 static inline bool sk_acceptq_is_full(const struct sock *sk)
 {
-	return READ_ONCE(sk->sk_ack_backlog) > sk->sk_max_ack_backlog;
+	return READ_ONCE(sk->sk_ack_backlog) > READ_ONCE(sk->sk_max_ack_backlog);
 }
 
 /*
net/dccp/proto.c +1 -1
index 5bad08dc431611d5387c8d3c1858ee2c43cb9b68..a52e8ba1ced046b178fa069b1e0d690c537c6bc0 100644
@@ -944,7 +944,7 @@ int inet_dccp_listen(struct socket *sock, int backlog)
 	if (!((1 << old_state) & (DCCPF_CLOSED | DCCPF_LISTEN)))
 		goto out;
 
-	sk->sk_max_ack_backlog = backlog;
+	WRITE_ONCE(sk->sk_max_ack_backlog, backlog);
 	/* Really, if the socket is already in listen state
 	 * we can only allow the backlog to be adjusted.
 	 */
net/ipv4/af_inet.c +1 -1
index 70f92aaca4110b3ecd691949203f28978597e9c9..53de8e00990e276448df1c60e47620be3b58f517 100644
@@ -208,7 +208,7 @@ int inet_listen(struct socket *sock, int backlog)
 	if (!((1 << old_state) & (TCPF_CLOSE | TCPF_LISTEN)))
 		goto out;
 
-	sk->sk_max_ack_backlog = backlog;
+	WRITE_ONCE(sk->sk_max_ack_backlog, backlog);
 	/* Really, if the socket is already in listen state
 	 * we can only allow the backlog to be adjusted.
 	 */
net/ipv4/inet_connection_sock.c +1 -1
index eb30fc1770def741950215f59a4e3ab0f91c6293..e4c6e8b4049063f5239a5e99a185016ad3bb5790 100644
@@ -716,7 +716,7 @@ static void reqsk_timer_handler(struct timer_list *t)
 	 * ones are about to clog our table.
 	 */
 	qlen = reqsk_queue_len(queue);
-	if ((qlen << 1) > max(8U, sk_listener->sk_max_ack_backlog)) {
+	if ((qlen << 1) > max(8U, READ_ONCE(sk_listener->sk_max_ack_backlog))) {
 		int young = reqsk_queue_len_young(queue) << 1;
 
 		while (thresh > 2) {
net/ipv4/tcp.c +1 -1
index 68375f7ffdce1fbbb4cf443660703c98b61fd9e3..fb1666440e1064a9ab2f2993b23fdb744e82f5c5 100644
@@ -3226,7 +3226,7 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
 		 * tcpi_sacked  -> max backlog
 		 */
 		info->tcpi_unacked = READ_ONCE(sk->sk_ack_backlog);
-		info->tcpi_sacked = sk->sk_max_ack_backlog;
+		info->tcpi_sacked = READ_ONCE(sk->sk_max_ack_backlog);
 		return;
 	}
 
net/ipv4/tcp_diag.c +1 -1
index edfbab54c46f4cac1b0a7960718d0b6308978957..0d08f9e2d8d0322fcdd3a465a3a9712b36605954 100644
@@ -22,7 +22,7 @@ static void tcp_diag_get_info(struct sock *sk, struct inet_diag_msg *r,
 
 	if (inet_sk_state_load(sk) == TCP_LISTEN) {
 		r->idiag_rqueue = READ_ONCE(sk->sk_ack_backlog);
-		r->idiag_wqueue = sk->sk_max_ack_backlog;
+		r->idiag_wqueue = READ_ONCE(sk->sk_max_ack_backlog);
 	} else if (sk->sk_type == SOCK_STREAM) {
 		const struct tcp_sock *tp = tcp_sk(sk);
 
net/sched/em_meta.c +1 -1
index ebb6e2430861d23a42431e4143f229395d9321c5..d99966a55c84fa0f5142ed72faeceb9baab86f5e 100644
@@ -532,7 +532,7 @@ META_COLLECTOR(int_sk_max_ack_bl)
 		*err = -1;
 		return;
 	}
-	dst->value = sk->sk_max_ack_backlog;
+	dst->value = READ_ONCE(sk->sk_max_ack_backlog);
 }
 
 META_COLLECTOR(int_sk_prio)
net/sctp/diag.c +1 -1
index f873f15407de4e7d9a246d41e07602f33da8064d..8a15146faaebdcb869233a08318e4fb5a1e1129b 100644
@@ -426,7 +426,7 @@ static void sctp_diag_get_info(struct sock *sk, struct inet_diag_msg *r,
 		r->idiag_wqueue = infox->asoc->sndbuf_used;
 	} else {
 		r->idiag_rqueue = READ_ONCE(sk->sk_ack_backlog);
-		r->idiag_wqueue = sk->sk_max_ack_backlog;
+		r->idiag_wqueue = READ_ONCE(sk->sk_max_ack_backlog);
 	}
 	if (infox->sctpinfo)
 		sctp_get_sctp_info(sk, infox->asoc, infox->sctpinfo);
net/sctp/socket.c +2 -2
index ffd3262b7a41eac2e3d825c3f0665066f376ea3c..53abb97e0061c14fd4a9c3090a4a5cbe0af9c5a9 100644
@@ -8376,7 +8376,7 @@ static int sctp_listen_start(struct sock *sk, int backlog)
 		}
 	}
 
-	sk->sk_max_ack_backlog = backlog;
+	WRITE_ONCE(sk->sk_max_ack_backlog, backlog);
 	return sctp_hash_endpoint(ep);
 }
 
@@ -8430,7 +8430,7 @@ int sctp_inet_listen(struct socket *sock, int backlog)
 
 	/* If we are already listening, just update the backlog */
 	if (sctp_sstate(sk, LISTENING))
-		sk->sk_max_ack_backlog = backlog;
+		WRITE_ONCE(sk->sk_max_ack_backlog, backlog);
 	else {
 		err = sctp_listen_start(sk, backlog);
 		if (err)
Signature · 1 line
2.24.0.rc1.363.gb1bccd3e3d-goog
maintainer
#7 lore raw
In reply to Eric Dumazet

From: Eric Dumazet <[email protected]>
Date: Tue, 5 Nov 2019 14:11:48 -0800

Quoted 3 lines

This is a series of minor fixes, mostly dealing with lockless accesses to socket 'sk_ack_backlog', 'sk_max_ack_backlog' ane neighbour 'confirmed' fields.

Series applied, thanks Eric.