[PATCH bpf-next v2 1/2] bpf: allow to retrieve sol_socket opts from sock_addr progs

[PATCH bpf-next v2 1/2] bpf: allow to retrieve sol_socket opts from sock_addr progs

[Daniel Borkmann]

The _bpf_setsockopt() is able to set some of the SOL_SOCKET level options,
however, _bpf_getsockopt() has little support to actually retrieve them.
This small patch adds few misc options such as SO_MARK, SO_PRIORITY and
SO_BINDTOIFINDEX. For the latter getter and setter are added. The mark and
priority in particular allow to retrieve the options from BPF cgroup hooks
to then implement custom behavior / settings on the syscall hooks compared
to other sockets that stick to the defaults, for example.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
---
 net/core/filter.c | 25 +++++++++++++++++++++++--
 1 file changed, 23 insertions(+), 2 deletions(-)

diff --git a/net/core/filter.c b/net/core/filter.c
index 255aeee72402..9ab94e90d660 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -4770,6 +4770,10 @@ static int _bpf_setsockopt(struct sock *sk, int level, int optname,
 				ifindex = dev->ifindex;
 				dev_put(dev);
 			}
+			fallthrough;
+		case SO_BINDTOIFINDEX:
+			if (optname == SO_BINDTOIFINDEX)
+				ifindex = val;
 			ret = sock_bindtoindex(sk, ifindex, false);
 			break;
 		case SO_KEEPALIVE:
@@ -4932,8 +4936,25 @@ static int _bpf_getsockopt(struct sock *sk, int level, int optname,
 
 	sock_owned_by_me(sk);
 
+	if (level == SOL_SOCKET) {
+		if (optlen != sizeof(int))
+			goto err_clear;
+
+		switch (optname) {
+		case SO_MARK:
+			*((int *)optval) = sk->sk_mark;
+			break;
+		case SO_PRIORITY:
+			*((int *)optval) = sk->sk_priority;
+			break;
+		case SO_BINDTOIFINDEX:
+			*((int *)optval) = sk->sk_bound_dev_if;
+			break;
+		default:
+			goto err_clear;
+		}
 #ifdef CONFIG_INET
-	if (level == SOL_TCP && sk->sk_prot->getsockopt == tcp_getsockopt) {
+	} else if (level == SOL_TCP && sk->sk_prot->getsockopt == tcp_getsockopt) {
 		struct inet_connection_sock *icsk;
 		struct tcp_sock *tp;
 
@@ -4986,12 +5007,12 @@ static int _bpf_getsockopt(struct sock *sk, int level, int optname,
 		default:
 			goto err_clear;
 		}
+#endif
 #endif
 	} else {
 		goto err_clear;
 	}
 	return 0;
-#endif
 err_clear:
 	memset(optval, 0, optlen);
 	return -EINVAL;
-- 
2.21.0

[PATCH bpf-next v2 2/2] bpf: extend bind v4/v6 selftests for mark/prio/bindtoifindex

[Daniel Borkmann]

Extend existing cgroup bind4/bind6 tests to add coverage for setting and
retrieving SO_MARK, SO_PRIORITY and SO_BINDTOIFINDEX at the bind hook.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
---
 v1 -> v2: - Add comment on old vs new optval (Yonghong), rest as-is

 .../testing/selftests/bpf/progs/bind4_prog.c  | 42 +++++++++++++++++--
 .../testing/selftests/bpf/progs/bind6_prog.c  | 42 +++++++++++++++++--
 2 files changed, 76 insertions(+), 8 deletions(-)

diff --git a/tools/testing/selftests/bpf/progs/bind4_prog.c b/tools/testing/selftests/bpf/progs/bind4_prog.c
index c6520f21f5f5..115a3b0ad984 100644
--- a/tools/testing/selftests/bpf/progs/bind4_prog.c
+++ b/tools/testing/selftests/bpf/progs/bind4_prog.c
@@ -29,18 +29,48 @@ static __inline int bind_to_device(struct bpf_sock_addr *ctx)
 	char veth2[IFNAMSIZ] = "test_sock_addr2";
 	char missing[IFNAMSIZ] = "nonexistent_dev";
 	char del_bind[IFNAMSIZ] = "";
+	int veth1_idx, veth2_idx;
 
 	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
-				&veth1, sizeof(veth1)))
+			   &veth1, sizeof(veth1)))
+		return 1;
+	if (bpf_getsockopt(ctx, SOL_SOCKET, SO_BINDTOIFINDEX,
+			   &veth1_idx, sizeof(veth1_idx)) || !veth1_idx)
 		return 1;
 	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
-				&veth2, sizeof(veth2)))
+			   &veth2, sizeof(veth2)))
+		return 1;
+	if (bpf_getsockopt(ctx, SOL_SOCKET, SO_BINDTOIFINDEX,
+			   &veth2_idx, sizeof(veth2_idx)) || !veth2_idx ||
+	    veth1_idx == veth2_idx)
 		return 1;
 	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
-				&missing, sizeof(missing)) != -ENODEV)
+			   &missing, sizeof(missing)) != -ENODEV)
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTOIFINDEX,
+			   &veth1_idx, sizeof(veth1_idx)))
 		return 1;
 	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
-				&del_bind, sizeof(del_bind)))
+			   &del_bind, sizeof(del_bind)))
+		return 1;
+
+	return 0;
+}
+
+static __inline int misc_opts(struct bpf_sock_addr *ctx, int opt)
+{
+	int old, tmp, new = 0xeb9f;
+
+	/* Socket in test case has guarantee that old never equals to new. */
+	if (bpf_getsockopt(ctx, SOL_SOCKET, opt, &old, sizeof(old)) ||
+	    old == new)
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, opt, &new, sizeof(new)))
+		return 1;
+	if (bpf_getsockopt(ctx, SOL_SOCKET, opt, &tmp, sizeof(tmp)) ||
+	    tmp != new)
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, opt, &old, sizeof(old)))
 		return 1;
 
 	return 0;
@@ -93,6 +123,10 @@ int bind_v4_prog(struct bpf_sock_addr *ctx)
 	if (bind_to_device(ctx))
 		return 0;
 
+	/* Test for misc socket options. */
+	if (misc_opts(ctx, SO_MARK) || misc_opts(ctx, SO_PRIORITY))
+		return 0;
+
 	ctx->user_ip4 = bpf_htonl(SERV4_REWRITE_IP);
 	ctx->user_port = bpf_htons(SERV4_REWRITE_PORT);
 
diff --git a/tools/testing/selftests/bpf/progs/bind6_prog.c b/tools/testing/selftests/bpf/progs/bind6_prog.c
index 4358e44dcf47..4c0d348034b9 100644
--- a/tools/testing/selftests/bpf/progs/bind6_prog.c
+++ b/tools/testing/selftests/bpf/progs/bind6_prog.c
@@ -35,18 +35,48 @@ static __inline int bind_to_device(struct bpf_sock_addr *ctx)
 	char veth2[IFNAMSIZ] = "test_sock_addr2";
 	char missing[IFNAMSIZ] = "nonexistent_dev";
 	char del_bind[IFNAMSIZ] = "";
+	int veth1_idx, veth2_idx;
 
 	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
-				&veth1, sizeof(veth1)))
+			   &veth1, sizeof(veth1)))
+		return 1;
+	if (bpf_getsockopt(ctx, SOL_SOCKET, SO_BINDTOIFINDEX,
+			   &veth1_idx, sizeof(veth1_idx)) || !veth1_idx)
 		return 1;
 	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
-				&veth2, sizeof(veth2)))
+			   &veth2, sizeof(veth2)))
+		return 1;
+	if (bpf_getsockopt(ctx, SOL_SOCKET, SO_BINDTOIFINDEX,
+			   &veth2_idx, sizeof(veth2_idx)) || !veth2_idx ||
+	    veth1_idx == veth2_idx)
 		return 1;
 	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
-				&missing, sizeof(missing)) != -ENODEV)
+			   &missing, sizeof(missing)) != -ENODEV)
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTOIFINDEX,
+			   &veth1_idx, sizeof(veth1_idx)))
 		return 1;
 	if (bpf_setsockopt(ctx, SOL_SOCKET, SO_BINDTODEVICE,
-				&del_bind, sizeof(del_bind)))
+			   &del_bind, sizeof(del_bind)))
+		return 1;
+
+	return 0;
+}
+
+static __inline int misc_opts(struct bpf_sock_addr *ctx, int opt)
+{
+	int old, tmp, new = 0xeb9f;
+
+	/* Socket in test case has guarantee that old never equals to new. */
+	if (bpf_getsockopt(ctx, SOL_SOCKET, opt, &old, sizeof(old)) ||
+	    old == new)
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, opt, &new, sizeof(new)))
+		return 1;
+	if (bpf_getsockopt(ctx, SOL_SOCKET, opt, &tmp, sizeof(tmp)) ||
+	    tmp != new)
+		return 1;
+	if (bpf_setsockopt(ctx, SOL_SOCKET, opt, &old, sizeof(old)))
 		return 1;
 
 	return 0;
@@ -107,6 +137,10 @@ int bind_v6_prog(struct bpf_sock_addr *ctx)
 	if (bind_to_device(ctx))
 		return 0;
 
+	/* Test for misc socket options. */
+	if (misc_opts(ctx, SO_MARK) || misc_opts(ctx, SO_PRIORITY))
+		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);
-- 
2.21.0

Re: [PATCH bpf-next v2 1/2] bpf: allow to retrieve sol_socket opts from sock_addr progs

[Alexei Starovoitov]

On Mon, Jan 11, 2021 at 3:09 PM Daniel Borkmann <daniel@iogearbox.net> wrote:
>
> The _bpf_setsockopt() is able to set some of the SOL_SOCKET level options,
> however, _bpf_getsockopt() has little support to actually retrieve them.
> This small patch adds few misc options such as SO_MARK, SO_PRIORITY and
> SO_BINDTOIFINDEX. For the latter getter and setter are added. The mark and
> priority in particular allow to retrieve the options from BPF cgroup hooks
> to then implement custom behavior / settings on the syscall hooks compared
> to other sockets that stick to the defaults, for example.
>
> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
> Acked-by: Yonghong Song <yhs@fb.com>

Applied.