Skip to contentSkip to conversation

bpf: expose bpf_{s,g}etsockopt helpers to bind{4,6} hooks

PATCH Applied Stanislav Fomichev 2 participants 5 messages Dec 2, 2020 – Dec 3, 2020

Subsystems (7)
lore
Share & tools
lore
#0 lore raw

cover bpf: expose bpf_{s,g}etsockopt helpers to bind{4,6} hooks

This might be useful for the listener sockets to pre-populate some options. Since those helpers require locked sockets, I'm changing bind hooks to lock/unlock the sockets. This should not cause any performance overhead because at this point there shouldn't be any socket lock contention and the locking/unlocking should be cheap.

Also, as part of the series, I convert test_sock_addr bpf assembly into C (and preserve the narrow load tests) to make it easier to extend with th bpf_setsockopt later on.

v2:
* remove version from bpf programs (Andrii Nakryiko)

Stanislav Fomichev (3):
selftests/bpf: rewrite test_sock_addr bind bpf into C
bpf: allow bpf_{s,g}etsockopt from cgroup bind{4,6} hooks selftests/bpf: extend bind{4,6} programs with a call to bpf_setsockopt

 include/linux/bpf-cgroup.h                    |  12 +-
 net/core/filter.c                             |   4 +
 net/ipv4/af_inet.c                            |   2 +-
 net/ipv6/af_inet6.c                           |   2 +-
 .../testing/selftests/bpf/progs/bind4_prog.c  | 102 +++++++++
 .../testing/selftests/bpf/progs/bind6_prog.c  | 119 +++++++++++
 tools/testing/selftests/bpf/test_sock_addr.c  | 196 ++----------------
 7 files changed, 245 insertions(+), 192 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/progs/bind4_prog.c
 create mode 100644 tools/testing/selftests/bpf/progs/bind6_prog.c
Signature · 1 line
2.29.2.454.gaff20da3a2-goog
#1 lore raw

PATCH 1/3 selftests/bpf: rewrite test_sock_addr bind bpf into C

I'm planning to extend it in the next patches. It's much easier to work with C than BPF assembly.

---
 .../testing/selftests/bpf/progs/bind4_prog.c  |  71 +++++++
 .../testing/selftests/bpf/progs/bind6_prog.c  |  88 ++++++++
 tools/testing/selftests/bpf/test_sock_addr.c  | 196 ++----------------
 3 files changed, 171 insertions(+), 184 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/progs/bind4_prog.c
 create mode 100644 tools/testing/selftests/bpf/progs/bind6_prog.c
tools/testing/selftests/bpf/progs/bind4_prog.c new file +71 -0
new file mode 100644
index 000000000000..0951302a984a
@@ -0,0 +1,71 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <string.h>
+
+#include <linux/stddef.h>
+#include <linux/bpf.h>
+#include <linux/in.h>
+#include <linux/in6.h>
+#include <sys/socket.h>
+#include <netinet/tcp.h>
+#include <linux/if.h>
+#include <errno.h>
+
+#include <bpf/bpf_helpers.h>
+#include <bpf/bpf_endian.h>
+
+#define SERV4_IP		0xc0a801feU /* 192.168.1.254 */
+#define SERV4_PORT		4040
+#define SERV4_REWRITE_IP	0x7f000001U /* 127.0.0.1 */
+#define SERV4_REWRITE_PORT	4444
+
+SEC("cgroup/bind4")
+int bind_v4_prog(struct bpf_sock_addr *ctx)
+{
+	struct bpf_sock *sk;
+	__u32 user_ip4;
+	__u16 user_port;
+
+	sk = ctx->sk;
+	if (!sk)
+		return 0;
+
+	if (sk->family != AF_INET)
+		return 0;
+
+	if (ctx->type != SOCK_STREAM && ctx->type != SOCK_DGRAM)
+		return 0;
+
+	if (ctx->user_ip4 != bpf_htonl(SERV4_IP) ||
+	    ctx->user_port != bpf_htons(SERV4_PORT))
+		return 0;
+
+	// u8 narrow loads:
+	user_ip4 = 0;
+	user_ip4 |= ((volatile __u8 *)&ctx->user_ip4)[0] << 0;
+	user_ip4 |= ((volatile __u8 *)&ctx->user_ip4)[1] << 8;
+	user_ip4 |= ((volatile __u8 *)&ctx->user_ip4)[2] << 16;
+	user_ip4 |= ((volatile __u8 *)&ctx->user_ip4)[3] << 24;
+	if (ctx->user_ip4 != user_ip4)
+		return 0;
+
+	user_port = 0;
+	user_port |= ((volatile __u8 *)&ctx->user_port)[0] << 0;
+	user_port |= ((volatile __u8 *)&ctx->user_port)[1] << 8;
+	if (ctx->user_port != user_port)
+		return 0;
+
+	// u16 narrow loads:
+	user_ip4 = 0;
+	user_ip4 |= ((volatile __u16 *)&ctx->user_ip4)[0] << 0;
+	user_ip4 |= ((volatile __u16 *)&ctx->user_ip4)[1] << 16;
+	if (ctx->user_ip4 != user_ip4)
+		return 0;
+
+	ctx->user_ip4 = bpf_htonl(SERV4_REWRITE_IP);
+	ctx->user_port = bpf_htons(SERV4_REWRITE_PORT);
+
+	return 1;
+}
+
+char _license[] SEC("license") = "GPL";
tools/testing/selftests/bpf/progs/bind6_prog.c new file +88 -0

88 lines not on this page · raw message

tools/testing/selftests/bpf/test_sock_addr.c +12 -184

214 lines not on this page · raw message

Signature · 1 line
2.29.2.454.gaff20da3a2-goog
#2 lore raw

PATCH 2/3 bpf: allow bpf_{s,g}etsockopt from cgroup bind{4,6} hooks

In reply to Stanislav Fomichev

I have to now lock/unlock socket for the bind hook execution. That shouldn't cause any overhead because the socket is unbound and shouldn't receive any traffic.

---
 include/linux/bpf-cgroup.h | 12 ++++++------
 net/core/filter.c          |  4 ++++
 net/ipv4/af_inet.c         |  2 +-
 net/ipv6/af_inet6.c        |  2 +-
 4 files changed, 12 insertions(+), 8 deletions(-)
include/linux/bpf-cgroup.h +6 -6
index ed71bd1a0825..72e69a0e1e8c 100644
@@ -246,11 +246,11 @@ int bpf_percpu_cgroup_storage_update(struct bpf_map *map, void *key,
 	__ret;								       \
 })
 
-#define BPF_CGROUP_RUN_PROG_INET4_BIND(sk, uaddr)			       \
-	BPF_CGROUP_RUN_SA_PROG(sk, uaddr, BPF_CGROUP_INET4_BIND)
+#define BPF_CGROUP_RUN_PROG_INET4_BIND_LOCK(sk, uaddr)			       \
+	BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_INET4_BIND, NULL)
 
-#define BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr)			       \
-	BPF_CGROUP_RUN_SA_PROG(sk, uaddr, BPF_CGROUP_INET6_BIND)
+#define BPF_CGROUP_RUN_PROG_INET6_BIND_LOCK(sk, uaddr)			       \
+	BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_INET6_BIND, NULL)
 
 #define BPF_CGROUP_PRE_CONNECT_ENABLED(sk) (cgroup_bpf_enabled && \
 					    sk->sk_prot->pre_connect)
@@ -434,8 +434,8 @@ static inline int bpf_percpu_cgroup_storage_update(struct bpf_map *map,
 #define BPF_CGROUP_RUN_PROG_INET_EGRESS(sk,skb) ({ 0; })
 #define BPF_CGROUP_RUN_PROG_INET_SOCK(sk) ({ 0; })
 #define BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk) ({ 0; })
-#define BPF_CGROUP_RUN_PROG_INET4_BIND(sk, uaddr) ({ 0; })
-#define BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr) ({ 0; })
+#define BPF_CGROUP_RUN_PROG_INET4_BIND_LOCK(sk, uaddr) ({ 0; })
+#define BPF_CGROUP_RUN_PROG_INET6_BIND_LOCK(sk, uaddr) ({ 0; })
 #define BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk) ({ 0; })
 #define BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk) ({ 0; })
 #define BPF_CGROUP_RUN_PROG_INET4_CONNECT(sk, uaddr) ({ 0; })
net/core/filter.c +4 -0
index 2ca5eecebacf..21d91dcf0260 100644
@@ -6995,6 +6995,8 @@ sock_addr_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_sk_storage_delete_proto;
 	case BPF_FUNC_setsockopt:
 		switch (prog->expected_attach_type) {
+		case BPF_CGROUP_INET4_BIND:
+		case BPF_CGROUP_INET6_BIND:
 		case BPF_CGROUP_INET4_CONNECT:
 		case BPF_CGROUP_INET6_CONNECT:
 			return &bpf_sock_addr_setsockopt_proto;
@@ -7003,6 +7005,8 @@ sock_addr_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		}
 	case BPF_FUNC_getsockopt:
 		switch (prog->expected_attach_type) {
+		case BPF_CGROUP_INET4_BIND:
+		case BPF_CGROUP_INET6_BIND:
 		case BPF_CGROUP_INET4_CONNECT:
 		case BPF_CGROUP_INET6_CONNECT:
 			return &bpf_sock_addr_getsockopt_proto;
net/ipv4/af_inet.c +1 -1
index b7260c8cef2e..b94fa8eb831b 100644
@@ -450,7 +450,7 @@ int inet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 	/* BPF prog is run before any checks are done so that if the prog
 	 * changes context in a wrong way it will be caught.
 	 */
-	err = BPF_CGROUP_RUN_PROG_INET4_BIND(sk, uaddr);
+	err = BPF_CGROUP_RUN_PROG_INET4_BIND_LOCK(sk, uaddr);
 	if (err)
 		return err;
 
net/ipv6/af_inet6.c +1 -1
index e648fbebb167..a7e3d170af51 100644
@@ -451,7 +451,7 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 	/* BPF prog is run before any checks are done so that if the prog
 	 * changes context in a wrong way it will be caught.
 	 */
-	err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
+	err = BPF_CGROUP_RUN_PROG_INET6_BIND_LOCK(sk, uaddr);
 	if (err)
 		return err;
 
Signature · 1 line
2.29.2.454.gaff20da3a2-goog
#3 lore raw

PATCH 3/3 selftests/bpf: extend bind{4,6} programs with a call to bpf_setsockopt

In reply to Stanislav Fomichev

To make sure it doesn't trigger sock_owned_by_me splat.

---
 .../testing/selftests/bpf/progs/bind4_prog.c  | 31 +++++++++++++++++++
 .../testing/selftests/bpf/progs/bind6_prog.c  | 31 +++++++++++++++++++
 2 files changed, 62 insertions(+)
tools/testing/selftests/bpf/progs/bind4_prog.c +31 -0
index 0951302a984a..c6520f21f5f5 100644
@@ -19,6 +19,33 @@
 #define SERV4_REWRITE_IP	0x7f000001U /* 127.0.0.1 */
 #define SERV4_REWRITE_PORT	4444
 
+#ifndef IFNAMSIZ
+#define IFNAMSIZ 16
+#endif
+
+static __inline int bind_to_device(struct bpf_sock_addr *ctx)
+{
+	char veth1[IFNAMSIZ] = "test_sock_addr1";
+	char veth2[IFNAMSIZ] = "test_sock_addr2";
+	char missing[IFNAMSIZ] = "nonexistent_dev";
+	char del_bind[IFNAMSIZ] = "";
+
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
+				&veth1, sizeof(veth1)))
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
+				&veth2, sizeof(veth2)))
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
+				&missing, sizeof(missing)) != -ENODEV)
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
+				&del_bind, sizeof(del_bind)))
+		return 1;
+
+	return 0;
+}
+
 SEC("cgroup/bind4")
 int bind_v4_prog(struct bpf_sock_addr *ctx)
 {
@@ -62,6 +89,10 @@ int bind_v4_prog(struct bpf_sock_addr *ctx)
 	if (ctx->user_ip4 != user_ip4)
 		return 0;
 
+	/* Bind to device and unbind it. */
+	if (bind_to_device(ctx))
+		return 0;
+
 	ctx->user_ip4 = bpf_htonl(SERV4_REWRITE_IP);
 	ctx->user_port = bpf_htons(SERV4_REWRITE_PORT);
 
tools/testing/selftests/bpf/progs/bind6_prog.c +31 -0
index 16da1cf85418..4358e44dcf47 100644
@@ -25,6 +25,33 @@
 #define SERV6_REWRITE_IP_3	0x00000001
 #define SERV6_REWRITE_PORT	6666
 
+#ifndef IFNAMSIZ
+#define IFNAMSIZ 16
+#endif
+
+static __inline int bind_to_device(struct bpf_sock_addr *ctx)
+{
+	char veth1[IFNAMSIZ] = "test_sock_addr1";
+	char veth2[IFNAMSIZ] = "test_sock_addr2";
+	char missing[IFNAMSIZ] = "nonexistent_dev";
+	char del_bind[IFNAMSIZ] = "";
+
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
+				&veth1, sizeof(veth1)))
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
+				&veth2, sizeof(veth2)))
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
+				&missing, sizeof(missing)) != -ENODEV)
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
+				&del_bind, sizeof(del_bind)))
+		return 1;
+
+	return 0;
+}
+
 SEC("cgroup/bind6")
 int bind_v6_prog(struct bpf_sock_addr *ctx)
 {
@@ -76,6 +103,10 @@ int bind_v6_prog(struct bpf_sock_addr *ctx)
 			return 0;
 	}
 
+	/* Bind to device and unbind it. */
+	if (bind_to_device(ctx))
+		return 0;
+
 	ctx->user_ip6[0] = bpf_htonl(SERV6_REWRITE_IP_0);
 	ctx->user_ip6[1] = bpf_htonl(SERV6_REWRITE_IP_1);
 	ctx->user_ip6[2] = bpf_htonl(SERV6_REWRITE_IP_2);
Signature · 1 line
2.29.2.454.gaff20da3a2-goog
#4 lore raw
In reply to Stanislav Fomichev
Quoted 7 lines · Stanislav Fomichev

This might be useful for the listener sockets to pre-populate some options. Since those helpers require locked sockets, I'm changing bind hooks to lock/unlock the sockets. This should not cause any performance overhead because at this point there shouldn't be any socket lock contention and the locking/unlocking should be cheap.

Quoted 6 lines

Also, as part of the series, I convert test_sock_addr bpf assembly into C (and preserve the narrow load tests) to make it easier to extend with th bpf_setsockopt later on.

v2:
* remove version from bpf programs (Andrii Nakryiko)

Applied, Thanks