bpf: Add struct bpf_tcp_sock and BPF_FUNC_tcp_sock

This patch adds a helper function BPF_FUNC_tcp_sock and it
is currently available for cg_skb and sched_(cls|act):

struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk);

int cg_skb_foo(struct __sk_buff *skb) {
	struct bpf_tcp_sock *tp;
	struct bpf_sock *sk;
	__u32 snd_cwnd;

	sk = skb->sk;
	if (!sk)
		return 1;

	tp = bpf_tcp_sock(sk);
	if (!tp)
		return 1;

	snd_cwnd = tp->snd_cwnd;
	/* ... */

	return 1;
}

A 'struct bpf_tcp_sock' is also added to the uapi bpf.h to provide
read-only access.  bpf_tcp_sock has all the existing tcp_sock's fields
that has already been exposed by the bpf_sock_ops.
i.e. no new tcp_sock's fields are exposed in bpf.h.

This helper returns a pointer to the tcp_sock.  If it is not a tcp_sock
or it cannot be traced back to a tcp_sock by sk_to_full_sk(), it
returns NULL.  Hence, the caller needs to check for NULL before
accessing it.

The current use case is to expose members from tcp_sock
to allow a cg_skb_bpf_prog to provide per cgroup traffic
policing/shaping.
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

include/linux/bpf.h

@@ -204,6 +204,7 @@ enum bpf_return_type {
 	RET_PTR_TO_MAP_VALUE,		/* returns a pointer to map elem value */
 	RET_PTR_TO_MAP_VALUE_OR_NULL,	/* returns a pointer to map elem value or NULL */
 	RET_PTR_TO_SOCKET_OR_NULL,	/* returns a pointer to a socket or NULL */
+	RET_PTR_TO_TCP_SOCK_OR_NULL,	/* returns a pointer to a tcp_sock or NULL */
 };
 
 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
@@ -259,6 +260,8 @@ enum bpf_reg_type {
 	PTR_TO_SOCKET_OR_NULL,	 /* reg points to struct bpf_sock or NULL */
 	PTR_TO_SOCK_COMMON,	 /* reg points to sock_common */
 	PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */
+	PTR_TO_TCP_SOCK,	 /* reg points to struct tcp_sock */
+	PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */
 };
 
 /* The information passed from prog-specific *_is_valid_access
@@ -956,4 +959,31 @@ static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
 }
 #endif
 
+#ifdef CONFIG_INET
+bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
+				  struct bpf_insn_access_aux *info);
+
+u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
+				    const struct bpf_insn *si,
+				    struct bpf_insn *insn_buf,
+				    struct bpf_prog *prog,
+				    u32 *target_size);
+#else
+static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
+						enum bpf_access_type type,
+						struct bpf_insn_access_aux *info)
+{
+	return false;
+}
+
+static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
+						  const struct bpf_insn *si,
+						  struct bpf_insn *insn_buf,
+						  struct bpf_prog *prog,
+						  u32 *target_size)
+{
+	return 0;
+}
+#endif /* CONFIG_INET */
+
 #endif /* _LINUX_BPF_H */

include/uapi/linux/bpf.h

@@ -2337,6 +2337,15 @@ union bpf_attr {
  *	Return
  *		A **struct bpf_sock** pointer on success, or NULL in
  *		case of failure.
+ *
+ * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk)
+ *	Description
+ *		This helper gets a **struct bpf_tcp_sock** pointer from a
+ *		**struct bpf_sock** pointer.
+ *
+ *	Return
+ *		A **struct bpf_tcp_sock** pointer on success, or NULL in
+ *		case of failure.
  */
 #define __BPF_FUNC_MAPPER(FN)		\
 	FN(unspec),			\
@@ -2434,7 +2443,8 @@ union bpf_attr {
 	FN(rc_pointer_rel),		\
 	FN(spin_lock),			\
 	FN(spin_unlock),		\
-	FN(sk_fullsock),
+	FN(sk_fullsock),		\
+	FN(tcp_sock),
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  * function eBPF program intends to call
@@ -2616,6 +2626,45 @@ struct bpf_sock {
 	__u32 state;
 };
 
+struct bpf_tcp_sock {
+	__u32 snd_cwnd;		/* Sending congestion window		*/
+	__u32 srtt_us;		/* smoothed round trip time << 3 in usecs */
+	__u32 rtt_min;
+	__u32 snd_ssthresh;	/* Slow start size threshold		*/
+	__u32 rcv_nxt;		/* What we want to receive next		*/
+	__u32 snd_nxt;		/* Next sequence we send		*/
+	__u32 snd_una;		/* First byte we want an ack for	*/
+	__u32 mss_cache;	/* Cached effective mss, not including SACKS */
+	__u32 ecn_flags;	/* ECN status bits.			*/
+	__u32 rate_delivered;	/* saved rate sample: packets delivered */
+	__u32 rate_interval_us;	/* saved rate sample: time elapsed */
+	__u32 packets_out;	/* Packets which are "in flight"	*/
+	__u32 retrans_out;	/* Retransmitted packets out		*/
+	__u32 total_retrans;	/* Total retransmits for entire connection */
+	__u32 segs_in;		/* RFC4898 tcpEStatsPerfSegsIn
+				 * total number of segments in.
+				 */
+	__u32 data_segs_in;	/* RFC4898 tcpEStatsPerfDataSegsIn
+				 * total number of data segments in.
+				 */
+	__u32 segs_out;		/* RFC4898 tcpEStatsPerfSegsOut
+				 * The total number of segments sent.
+				 */
+	__u32 data_segs_out;	/* RFC4898 tcpEStatsPerfDataSegsOut
+				 * total number of data segments sent.
+				 */
+	__u32 lost_out;		/* Lost packets			*/
+	__u32 sacked_out;	/* SACK'd packets			*/
+	__u64 bytes_received;	/* RFC4898 tcpEStatsAppHCThruOctetsReceived
+				 * sum(delta(rcv_nxt)), or how many bytes
+				 * were acked.
+				 */
+	__u64 bytes_acked;	/* RFC4898 tcpEStatsAppHCThruOctetsAcked
+				 * sum(delta(snd_una)), or how many bytes
+				 * were acked.
+				 */
+};
+
 struct bpf_sock_tuple {
 	union {
 		struct {

kernel/bpf/verifier.c

@@ -334,14 +334,16 @@ static bool type_is_pkt_pointer(enum bpf_reg_type type)
 static bool type_is_sk_pointer(enum bpf_reg_type type)
 {
 	return type == PTR_TO_SOCKET ||
-		type == PTR_TO_SOCK_COMMON;
+		type == PTR_TO_SOCK_COMMON ||
+		type == PTR_TO_TCP_SOCK;
 }
 
 static bool reg_type_may_be_null(enum bpf_reg_type type)
 {
 	return type == PTR_TO_MAP_VALUE_OR_NULL ||
 	       type == PTR_TO_SOCKET_OR_NULL ||
-	       type == PTR_TO_SOCK_COMMON_OR_NULL;
+	       type == PTR_TO_SOCK_COMMON_OR_NULL ||
+	       type == PTR_TO_TCP_SOCK_OR_NULL;
 }
 
 static bool type_is_refcounted(enum bpf_reg_type type)
@@ -407,6 +409,8 @@ static const char * const reg_type_str[] = {
 	[PTR_TO_SOCKET_OR_NULL] = "sock_or_null",
 	[PTR_TO_SOCK_COMMON]	= "sock_common",
 	[PTR_TO_SOCK_COMMON_OR_NULL] = "sock_common_or_null",
+	[PTR_TO_TCP_SOCK]	= "tcp_sock",
+	[PTR_TO_TCP_SOCK_OR_NULL] = "tcp_sock_or_null",
 };
 
 static char slot_type_char[] = {
@@ -1209,6 +1213,8 @@ static bool is_spillable_regtype(enum bpf_reg_type type)
 	case PTR_TO_SOCKET_OR_NULL:
 	case PTR_TO_SOCK_COMMON:
 	case PTR_TO_SOCK_COMMON_OR_NULL:
+	case PTR_TO_TCP_SOCK:
+	case PTR_TO_TCP_SOCK_OR_NULL:
 		return true;
 	default:
 		return false;
@@ -1662,6 +1668,9 @@ static int check_sock_access(struct bpf_verifier_env *env, int insn_idx,
 	case PTR_TO_SOCKET:
 		valid = bpf_sock_is_valid_access(off, size, t, &info);
 		break;
+	case PTR_TO_TCP_SOCK:
+		valid = bpf_tcp_sock_is_valid_access(off, size, t, &info);
+		break;
 	default:
 		valid = false;
 	}
@@ -1823,6 +1832,9 @@ static int check_ptr_alignment(struct bpf_verifier_env *env,
 	case PTR_TO_SOCK_COMMON:
 		pointer_desc = "sock_common ";
 		break;
+	case PTR_TO_TCP_SOCK:
+		pointer_desc = "tcp_sock ";
+		break;
 	default:
 		break;
 	}
@@ -3148,6 +3160,10 @@ static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn
 			/* For mark_ptr_or_null_reg() */
 			regs[BPF_REG_0].id = ++env->id_gen;
 		}
+	} else if (fn->ret_type == RET_PTR_TO_TCP_SOCK_OR_NULL) {
+		mark_reg_known_zero(env, regs, BPF_REG_0);
+		regs[BPF_REG_0].type = PTR_TO_TCP_SOCK_OR_NULL;
+		regs[BPF_REG_0].id = ++env->id_gen;
 	} else {
 		verbose(env, "unknown return type %d of func %s#%d\n",
 			fn->ret_type, func_id_name(func_id), func_id);
@@ -3409,6 +3425,8 @@ static int adjust_ptr_min_max_vals(struct bpf_verifier_env *env,
 	case PTR_TO_SOCKET_OR_NULL:
 	case PTR_TO_SOCK_COMMON:
 	case PTR_TO_SOCK_COMMON_OR_NULL:
+	case PTR_TO_TCP_SOCK:
+	case PTR_TO_TCP_SOCK_OR_NULL:
 		verbose(env, "R%d pointer arithmetic on %s prohibited\n",
 			dst, reg_type_str[ptr_reg->type]);
 		return -EACCES;
@@ -4644,6 +4662,8 @@ static void mark_ptr_or_null_reg(struct bpf_func_state *state,
 			reg->type = PTR_TO_SOCKET;
 		} else if (reg->type == PTR_TO_SOCK_COMMON_OR_NULL) {
 			reg->type = PTR_TO_SOCK_COMMON;
+		} else if (reg->type == PTR_TO_TCP_SOCK_OR_NULL) {
+			reg->type = PTR_TO_TCP_SOCK;
 		}
 		if (is_null || !(reg_is_refcounted(reg) ||
 				 reg_may_point_to_spin_lock(reg))) {
@@ -5839,6 +5859,8 @@ static bool regsafe(struct bpf_reg_state *rold, struct bpf_reg_state *rcur,
 	case PTR_TO_SOCKET_OR_NULL:
 	case PTR_TO_SOCK_COMMON:
 	case PTR_TO_SOCK_COMMON_OR_NULL:
+	case PTR_TO_TCP_SOCK:
+	case PTR_TO_TCP_SOCK_OR_NULL:
 		/* Only valid matches are exact, which memcmp() above
 		 * would have accepted
 		 */
@@ -6161,6 +6183,8 @@ static bool reg_type_mismatch_ok(enum bpf_reg_type type)
 	case PTR_TO_SOCKET_OR_NULL:
 	case PTR_TO_SOCK_COMMON:
 	case PTR_TO_SOCK_COMMON_OR_NULL:
+	case PTR_TO_TCP_SOCK:
+	case PTR_TO_TCP_SOCK_OR_NULL:
 		return false;
 	default:
 		return true;
@@ -7166,6 +7190,9 @@ static int convert_ctx_accesses(struct bpf_verifier_env *env)
 		case PTR_TO_SOCK_COMMON:
 			convert_ctx_access = bpf_sock_convert_ctx_access;
 			break;
+		case PTR_TO_TCP_SOCK:
+			convert_ctx_access = bpf_tcp_sock_convert_ctx_access;
+			break;
 		default:
 			continue;
 		}

net/core/filter.c

@@ -5315,6 +5315,79 @@ static const struct bpf_func_proto bpf_sock_addr_sk_lookup_udp_proto = {
 	.arg5_type	= ARG_ANYTHING,
 };
 
+bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
+				  struct bpf_insn_access_aux *info)
+{
+	if (off < 0 || off >= offsetofend(struct bpf_tcp_sock, bytes_acked))
+		return false;
+
+	if (off % size != 0)
+		return false;
+
+	switch (off) {
+	case offsetof(struct bpf_tcp_sock, bytes_received):
+	case offsetof(struct bpf_tcp_sock, bytes_acked):
+		return size == sizeof(__u64);
+	default:
+		return size == sizeof(__u32);
+	}
+}
+
+u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
+				    const struct bpf_insn *si,
+				    struct bpf_insn *insn_buf,
+				    struct bpf_prog *prog, u32 *target_size)
+{
+	struct bpf_insn *insn = insn_buf;
+
+#define BPF_TCP_SOCK_GET_COMMON(FIELD)					\
+	do {								\
+		BUILD_BUG_ON(FIELD_SIZEOF(struct tcp_sock, FIELD) >	\
+			     FIELD_SIZEOF(struct bpf_tcp_sock, FIELD));	\
+		*insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct tcp_sock, FIELD),\
+				      si->dst_reg, si->src_reg,		\
+				      offsetof(struct tcp_sock, FIELD)); \
+	} while (0)
+
+	CONVERT_COMMON_TCP_SOCK_FIELDS(struct bpf_tcp_sock,
+				       BPF_TCP_SOCK_GET_COMMON);
+
+	if (insn > insn_buf)
+		return insn - insn_buf;
+
+	switch (si->off) {
+	case offsetof(struct bpf_tcp_sock, rtt_min):
+		BUILD_BUG_ON(FIELD_SIZEOF(struct tcp_sock, rtt_min) !=
+			     sizeof(struct minmax));
+		BUILD_BUG_ON(sizeof(struct minmax) <
+			     sizeof(struct minmax_sample));
+
+		*insn++ = BPF_LDX_MEM(BPF_W, si->dst_reg, si->src_reg,
+				      offsetof(struct tcp_sock, rtt_min) +
+				      offsetof(struct minmax_sample, v));
+		break;
+	}
+
+	return insn - insn_buf;
+}
+
+BPF_CALL_1(bpf_tcp_sock, struct sock *, sk)
+{
+	sk = sk_to_full_sk(sk);
+
+	if (sk_fullsock(sk) && sk->sk_protocol == IPPROTO_TCP)
+		return (unsigned long)sk;
+
+	return (unsigned long)NULL;
+}
+
+static const struct bpf_func_proto bpf_tcp_sock_proto = {
+	.func		= bpf_tcp_sock,
+	.gpl_only	= false,
+	.ret_type	= RET_PTR_TO_TCP_SOCK_OR_NULL,
+	.arg1_type	= ARG_PTR_TO_SOCK_COMMON,
+};
+
 #endif /* CONFIG_INET */
 
 bool bpf_helper_changes_pkt_data(void *func)
@@ -5470,6 +5543,10 @@ cg_skb_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_get_local_storage_proto;
 	case BPF_FUNC_sk_fullsock:
 		return &bpf_sk_fullsock_proto;
+#ifdef CONFIG_INET
+	case BPF_FUNC_tcp_sock:
+		return &bpf_tcp_sock_proto;
+#endif
 	default:
 		return sk_filter_func_proto(func_id, prog);
 	}
@@ -5560,6 +5637,8 @@ tc_cls_act_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_sk_lookup_udp_proto;
 	case BPF_FUNC_sk_release:
 		return &bpf_sk_release_proto;
+	case BPF_FUNC_tcp_sock:
+		return &bpf_tcp_sock_proto;
 #endif
 	default:
 		return bpf_base_func_proto(func_id);