Skip to contentSkip to conversation

Add support to set window_clamp from bpf setsockops

PATCH Superseded Prankur gupta 2 participants 4 messages Dec 2, 2020

Subsystems (7)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v3 cover Add support to set window_clamp from bpf setsockops

This patch contains support to set tcp window_field field from bpf setsockops.

v2: Used TCP_WINDOW_CLAMP setsockopt logic for bpf_setsockopt (review comment addressed)

v3: Created a common function for duplicated code (review comment addressed)

Prankur gupta (2):
bpf: Adds support for setting window clamp selftests/bpf: Add Userspace tests for TCP_WINDOW_CLAMP

 include/net/tcp.h                             |  1 +
 net/core/filter.c                             |  3 ++
 net/ipv4/tcp.c                                | 23 ++++++++-----
 tools/testing/selftests/bpf/bpf_tcp_helpers.h |  1 +
 .../selftests/bpf/prog_tests/tcpbpf_user.c    |  4 +++
 .../selftests/bpf/progs/test_tcpbpf_kern.c    | 33 +++++++++++++++++++
 tools/testing/selftests/bpf/test_tcpbpf.h     |  2 ++
 7 files changed, 58 insertions(+), 9 deletions(-)
Signature · 1 line
2.24.1
#1 lore raw

PATCH v3 1/2 bpf: Adds support for setting window clamp

Adds a new bpf_setsockopt for TCP sockets, TCP_BPF_WINDOW_CLAMP, which sets the maximum receiver window size. It will be useful for limiting receiver window based on RTT.

---
 include/net/tcp.h |  1 +
 net/core/filter.c |  3 +++
 net/ipv4/tcp.c    | 23 ++++++++++++++---------
 3 files changed, 18 insertions(+), 9 deletions(-)
include/net/tcp.h +1 -0
index 4aba0f069b05..39ced5882fe3 100644
@@ -406,6 +406,7 @@ void tcp_syn_ack_timeout(const struct request_sock *req);
 int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
 		int flags, int *addr_len);
 int tcp_set_rcvlowat(struct sock *sk, int val);
+int tcp_set_window_clamp(struct sock *sk, struct tcp_sock *tp, int val);
 void tcp_data_ready(struct sock *sk);
 #ifdef CONFIG_MMU
 int tcp_mmap(struct file *file, struct socket *sock,
net/core/filter.c +3 -0
index 2ca5eecebacf..d6225842cfb1 100644
@@ -4910,6 +4910,9 @@ static int _bpf_setsockopt(struct sock *sk, int level, int optname,
 				tp->notsent_lowat = val;
 				sk->sk_write_space(sk);
 				break;
+			case TCP_WINDOW_CLAMP:
+				ret = tcp_set_window_clamp(sk, tp, val);
+				break;
 			default:
 				ret = -EINVAL;
 			}
net/ipv4/tcp.c +14 -9
index b2bc3d7fe9e8..312feb8fcae5 100644
@@ -3022,6 +3022,19 @@ int tcp_sock_set_keepcnt(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepcnt);
 
+int tcp_set_window_clamp(struct sock *sk, struct tcp_sock *tp, int val)
+{
+	if (!val) {
+		if (sk->sk_state != TCP_CLOSE)
+			return -EINVAL;
+		tp->window_clamp = 0;
+	} else {
+		tp->window_clamp = val < SOCK_MIN_RCVBUF / 2 ?
+			SOCK_MIN_RCVBUF / 2 : val;
+	}
+	return 0;
+}
+
 /*
  *	Socket option code for TCP.
  */
@@ -3235,15 +3248,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level, int optname,
 		break;
 
 	case TCP_WINDOW_CLAMP:
-		if (!val) {
-			if (sk->sk_state != TCP_CLOSE) {
-				err = -EINVAL;
-				break;
-			}
-			tp->window_clamp = 0;
-		} else
-			tp->window_clamp = val < SOCK_MIN_RCVBUF / 2 ?
-						SOCK_MIN_RCVBUF / 2 : val;
+		err = tcp_set_window_clamp(sk, tp, val);
 		break;
 
 	case TCP_QUICKACK:
Signature · 1 line
2.24.1
#2 lore raw

Re: [PATCH v3 bpf-next 1/2] bpf: Adds support for setting window clamp

Quoted 42 lines · Prankur gupta

Adds a new bpf_setsockopt for TCP sockets, TCP_BPF_WINDOW_CLAMP, which sets the maximum receiver window size. It will be useful for limiting receiver window based on RTT.

---
 include/net/tcp.h |  1 +
 net/core/filter.c |  3 +++
 net/ipv4/tcp.c    | 23 ++++++++++++++---------
 3 files changed, 18 insertions(+), 9 deletions(-)
include/net/tcp.h +1 -0
index 4aba0f069b05..39ced5882fe3 100644
@@ -406,6 +406,7 @@ void tcp_syn_ack_timeout(const struct request_sock *req);
 int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
 		int flags, int *addr_len);
 int tcp_set_rcvlowat(struct sock *sk, int val);
+int tcp_set_window_clamp(struct sock *sk, struct tcp_sock *tp, int val);
 void tcp_data_ready(struct sock *sk);
 #ifdef CONFIG_MMU
 int tcp_mmap(struct file *file, struct socket *sock,
net/core/filter.c +3 -0
index 2ca5eecebacf..d6225842cfb1 100644
@@ -4910,6 +4910,9 @@ static int _bpf_setsockopt(struct sock *sk, int level, int optname,
 				tp->notsent_lowat = val;
 				sk->sk_write_space(sk);
 				break;
+			case TCP_WINDOW_CLAMP:
+				ret = tcp_set_window_clamp(sk, tp, val);
+				break;
 			default:
 				ret = -EINVAL;
 			}
net/ipv4/tcp.c +0 -0
index b2bc3d7fe9e8..312feb8fcae5 100644

net/ipv4/tcp.c int tcp_sock_set_keepcnt(struct sock *sk, int val) · line 3025

Quoted 4 lines
net/ipv4/tcp.c +1 -0
@@ -3022,6 +3022,19 @@ int tcp_sock_set_keepcnt(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepcnt);
 
+int tcp_set_window_clamp(struct sock *sk, struct tcp_sock *tp, int val)

No TCP function pass both sk and tp (which are identical values)

Prefer :

int tcp_set_window_clamp(struct sock *sk, int val)

{
        struct tcp_sock *tp = tcp_sk(sk);
        ...

This will allow optimal code generation.

Quoted 33 lines
net/ipv4/tcp.c +13 -9
+{
+	if (!val) {
+		if (sk->sk_state != TCP_CLOSE)
+			return -EINVAL;
+		tp->window_clamp = 0;
+	} else {
+		tp->window_clamp = val < SOCK_MIN_RCVBUF / 2 ?
+			SOCK_MIN_RCVBUF / 2 : val;
+	}
+	return 0;
+}
+
 /*
  *	Socket option code for TCP.
  */
@@ -3235,15 +3248,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level, int optname,
 		break;
 
 	case TCP_WINDOW_CLAMP:
-		if (!val) {
-			if (sk->sk_state != TCP_CLOSE) {
-				err = -EINVAL;
-				break;
-			}
-			tp->window_clamp = 0;
-		} else
-			tp->window_clamp = val < SOCK_MIN_RCVBUF / 2 ?
-						SOCK_MIN_RCVBUF / 2 : val;
+		err = tcp_set_window_clamp(sk, tp, val);
 		break;
 
 	case TCP_QUICKACK:
#3 lore raw

PATCH v3 2/2 selftests/bpf: Add Userspace tests for TCP_WINDOW_CLAMP

In reply to Prankur gupta

Adding selftests for new added functionality to set TCP_WINDOW_CLAMP from bpf setsockopt.

---
 tools/testing/selftests/bpf/bpf_tcp_helpers.h |  1 +
 .../selftests/bpf/prog_tests/tcpbpf_user.c    |  4 +++
 .../selftests/bpf/progs/test_tcpbpf_kern.c    | 33 +++++++++++++++++++
 tools/testing/selftests/bpf/test_tcpbpf.h     |  2 ++
 4 files changed, 40 insertions(+)
tools/testing/selftests/bpf/bpf_tcp_helpers.h +1 -0
index 2915664c335d..6a9053162cf2 100644
@@ -56,6 +56,7 @@ struct tcp_sock {
 	__u32	rcv_nxt;
 	__u32	snd_nxt;
 	__u32	snd_una;
+	__u32	window_clamp;
 	__u8	ecn_flags;
 	__u32	delivered;
 	__u32	delivered_ce;
tools/testing/selftests/bpf/prog_tests/tcpbpf_user.c +4 -0
index ab5281475f44..87923d2865b7 100644
@@ -42,6 +42,10 @@ static void verify_result(struct tcpbpf_globals *result)
 
 	/* check getsockopt for SAVED_SYN */
 	ASSERT_EQ(result->tcp_saved_syn, 1, "tcp_saved_syn");
+
+	/* check getsockopt for window_clamp */
+	ASSERT_EQ(result->window_clamp_client, 9216, "window_clamp_client");
+	ASSERT_EQ(result->window_clamp_server, 9216, "window_clamp_server");
 }
 
 static void run_test(struct tcpbpf_globals *result)
tools/testing/selftests/bpf/progs/test_tcpbpf_kern.c +33 -0
index e85e49deba70..94f50f7e94d6 100644
@@ -12,17 +12,41 @@
 #include <linux/tcp.h>
 #include <bpf/bpf_helpers.h>
 #include <bpf/bpf_endian.h>
+#include "bpf_tcp_helpers.h"
 #include "test_tcpbpf.h"
 
 struct tcpbpf_globals global = {};
 int _version SEC("version") = 1;
 
+/**
+ * SOL_TCP is defined in <netinet/tcp.h> while
+ * TCP_SAVED_SYN is defined in already included <linux/tcp.h>
+ */
+#ifndef SOL_TCP
+#define SOL_TCP 6
+#endif
+
+static __always_inline int get_tp_window_clamp(struct bpf_sock_ops *skops)
+{
+	struct bpf_sock *sk;
+	struct tcp_sock *tp;
+
+	sk = skops->sk;
+	if (!sk)
+		return -1;
+	tp = bpf_skc_to_tcp_sock(sk);
+	if (!tp)
+		return -1;
+	return tp->window_clamp;
+}
+
 SEC("sockops")
 int bpf_testcb(struct bpf_sock_ops *skops)
 {
 	char header[sizeof(struct ipv6hdr) + sizeof(struct tcphdr)];
 	struct bpf_sock_ops *reuse = skops;
 	struct tcphdr *thdr;
+	int window_clamp = 9216;
 	int good_call_rv = 0;
 	int bad_call_rv = 0;
 	int save_syn = 1;
@@ -75,6 +99,11 @@ int bpf_testcb(struct bpf_sock_ops *skops)
 	global.event_map |= (1 << op);
 
 	switch (op) {
+	case BPF_SOCK_OPS_TCP_CONNECT_CB:
+		rv = bpf_setsockopt(skops, SOL_TCP, TCP_WINDOW_CLAMP,
+				    &window_clamp, sizeof(window_clamp));
+		global.window_clamp_client = get_tp_window_clamp(skops);
+		break;
 	case BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB:
 		/* Test failure to set largest cb flag (assumes not defined) */
 		global.bad_cb_test_rv = bpf_sock_ops_cb_flags_set(skops, 0x80);
@@ -100,6 +129,10 @@ int bpf_testcb(struct bpf_sock_ops *skops)
 				global.tcp_saved_syn = v;
 			}
 		}
+		rv = bpf_setsockopt(skops, SOL_TCP, TCP_WINDOW_CLAMP,
+				    &window_clamp, sizeof(window_clamp));
+
+		global.window_clamp_server = get_tp_window_clamp(skops);
 		break;
 	case BPF_SOCK_OPS_RTO_CB:
 		break;
tools/testing/selftests/bpf/test_tcpbpf.h +2 -0
index 0ed33521cbbb..9dd9b5590f9d 100644
@@ -16,5 +16,7 @@ struct tcpbpf_globals {
 	__u32 num_close_events;
 	__u32 tcp_save_syn;
 	__u32 tcp_saved_syn;
+	__u32 window_clamp_client;
+	__u32 window_clamp_server;
 };
 #endif
Signature · 1 line
2.24.1