Date:   Fri, 11 Sep 2020 21:59:30 -0700
From:   Martin KaFai Lau <kafai@...com>
To:     <bpf@...r.kernel.org>
CC:     Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>, <kernel-team@...com>,
        Lorenz Bauer <lmb@...udflare.com>, <netdev@...r.kernel.org>
Subject: [PATCH RFC bpf-next 2/2] bpf: Enable bpf_skc_to_* sock casting helper to networking prog type

There is a constant need to add more fields into the bpf_tcp_sock
for the bpf programs running at tc, sock_ops...etc.

A current workaround could be to use bpf_probe_read_kernel().  However,
other than making another helper call for reading each field and missing
CO-RE, it is also not as intuitive to use as directly reading
"tp->lsndtime" for example.  While already having perfmon cap to do
bpf_probe_read_kernel(), it will be much easier if the bpf prog can
directly read from the tcp_sock.

This patch tries to do that by using the existing casting-helpers
bpf_skc_to_*() whose func_proto returns a btf_id.  For example, the
func_proto of bpf_skc_to_tcp_sock returns the btf_id of the
kernel "struct tcp_sock".

[ One approach is to make a separate copy of the bpf_skc_to_*
  func_proto and use ARG_PTR_TO_SOCK_COMMON instead of ARG_PTR_TO_BTF_ID.
  More on this later (1). ]

This patch modifies the existing bpf_skc_to_* func_proto to take
ARG_PTR_TO_SOCK_COMMON instead of taking
"ARG_PTR_TO_BTF_ID + &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON]".
That will allow tc, sock_ops,...etc to call these casting helpers
because they already hold the PTR_TO_SOCK_COMMON (or its
equivalent).  For example:

	sk = sock_ops->sk;
	if (!sk)
		return;
	tp = bpf_skc_to_tcp_sock(sk);
	if (!tp)
		return;
	/* Read tp as a PTR_TO_BTF_ID */
	lsndtime = tp->lsndtime;

To ensure the current bpf prog passing a PTR_TO_BTF_ID to
bpf_skc_to_*() still works as is, the verifier is modified such that
ARG_PTR_TO_SOCK_COMMON can accept a reg with reg->type == PTR_TO_BTF_ID
and reg->btf_id is btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON]

To do that, an idea is borrowed from one of the Lorenz's patch:
https://lore.kernel.org/bpf/20200904112401.667645-12-lmb@cloudflare.com/ .
It adds PTR_TO_BTF_ID as one of the acceptable reg->type for
ARG_PTR_TO_SOCK_COMMON and also specifies what btf_id it can take.
By doing this, the bpf_skc_to_* will work as before and can still
take PTR_TO_BTF_ID as the arg.  e.g. The bpf tcp iter will work
as is.

This will also make other existing helper taking ARG_PTR_TO_SOCK_COMMON
works with the pointer obtained from bpf_skc_to_*(). For example:

 	sk = bpf_skc_lookup_tcp(skb, tuple, tuplen, BPF_F_CURRENT_NETNS, 0);
	if (!sk)
		return;
	tp = bpf_skc_to_tcp_sock(sk);
	if (!tp) {
		bpf_sk_release(sk);
		return;
	}
	lsndtime = tp->lsndtime;
	/* Pass tp to bpf_sk_release() will also work */
	bpf_sk_release(tp);

[ (1) Since we need to make the existing helpers taking
  ARG_PTR_TO_SOCK_COMMON to work with PTR_TO_BTF_ID, the
  existing bpf_skc_to_*() func_proto are modified instead
  of making a separate copy. ]

A similar change has not been done to the ARG_PTR_TO_SOCKET yet.  Thus,
ARG_PTR_TO_SOCKET cannot take PTR_TO_BTF_ID for now.

The BPF_FUNC_skc_to_* func_id is added to is_ptr_cast_function().
It ensures the returning reg (BPF_REF_0) which is a PTR_TO_BTF_ID_OR_NULL
also carries the ref_obj_id.  That will keep the ref-tracking works
properly.  It also allow bpf_sk_release() to be called on the ptr
returned by bpf_skc_to_*(), which is shown in the previous
example.

The bpf_skc_to_* helpers are made available to most of the bpf prog
types in filter.c. They are limited by perfmon cap.

Signed-off-by: Martin KaFai Lau <kafai@...com>
---
 kernel/bpf/verifier.c | 22 ++++++++++++--
 net/core/filter.c     | 69 ++++++++++++++++++++++++++++++-------------
 2 files changed, 69 insertions(+), 22 deletions(-)

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 3a5932bd7c22..9eace789e88c 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -486,7 +486,12 @@ static bool is_acquire_function(enum bpf_func_id func_id,
 static bool is_ptr_cast_function(enum bpf_func_id func_id)
 {
 	return func_id == BPF_FUNC_tcp_sock ||
-		func_id == BPF_FUNC_sk_fullsock;
+		func_id == BPF_FUNC_sk_fullsock ||
+		func_id == BPF_FUNC_skc_to_tcp_sock ||
+		func_id == BPF_FUNC_skc_to_tcp6_sock ||
+		func_id == BPF_FUNC_skc_to_udp6_sock ||
+		func_id == BPF_FUNC_skc_to_tcp_timewait_sock ||
+		func_id == BPF_FUNC_skc_to_tcp_request_sock;
 }
 
 /* string representation of 'enum bpf_reg_type' */
@@ -3906,6 +3911,7 @@ static int resolve_map_arg_type(struct bpf_verifier_env *env,
 
 struct bpf_reg_types {
 	const enum bpf_reg_type types[10];
+	u32 *btf_id;
 };
 
 static const struct bpf_reg_types map_key_value_types = {
@@ -3923,7 +3929,9 @@ static const struct bpf_reg_types sock_types = {
 		PTR_TO_SOCKET,
 		PTR_TO_TCP_SOCK,
 		PTR_TO_XDP_SOCK,
+		PTR_TO_BTF_ID,
 	},
+	.btf_id = &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
 };
 
 static const struct bpf_reg_types mem_types = {
@@ -4014,7 +4022,17 @@ static int check_reg_type(struct bpf_verifier_env *env, u32 arg,
 
 found:
 	if (type == PTR_TO_BTF_ID) {
-		u32 *expected_btf_id = fn->arg_btf_id[arg];
+		u32 *expected_btf_id;
+
+		if (arg_type == ARG_PTR_TO_BTF_ID) {
+			expected_btf_id = fn->arg_btf_id[arg];
+		} else {
+			expected_btf_id = compatible->btf_id;
+			if (!expected_btf_id) {
+				 verbose(env, "verifier internal error: missing arg compatible BTF ID\n");
+				 return -EFAULT;
+			}
+		}
 
 		if (!btf_struct_ids_match(&env->log, reg->off, reg->btf_id,
 					  *expected_btf_id)) {
diff --git a/net/core/filter.c b/net/core/filter.c
index 6014e5f40c58..115cde5aaf90 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -77,6 +77,9 @@
 #include <net/transp_v6.h>
 #include <linux/btf_ids.h>
 
+static const struct bpf_func_proto *
+bpf_sk_base_func_proto(enum bpf_func_id func_id);
+
 int copy_bpf_fprog_from_user(struct sock_fprog *dst, sockptr_t src, int len)
 {
 	if (in_compat_syscall()) {
@@ -6619,7 +6622,7 @@ sock_addr_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 			return NULL;
 		}
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -6638,7 +6641,7 @@ sk_filter_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	case BPF_FUNC_perf_event_output:
 		return &bpf_skb_event_output_proto;
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -6799,7 +6802,7 @@ tc_cls_act_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_sk_assign_proto;
 #endif
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -6840,7 +6843,7 @@ xdp_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_tcp_gen_syncookie_proto;
 #endif
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -6882,7 +6885,7 @@ sock_ops_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_tcp_sock_proto;
 #endif /* CONFIG_INET */
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -6928,7 +6931,7 @@ sk_msg_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_get_cgroup_classid_curr_proto;
 #endif
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -6970,7 +6973,7 @@ sk_skb_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_skc_lookup_tcp_proto;
 #endif
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -6981,7 +6984,7 @@ flow_dissector_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	case BPF_FUNC_skb_load_bytes:
 		return &bpf_flow_dissector_load_bytes_proto;
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -7008,7 +7011,7 @@ lwt_out_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	case BPF_FUNC_skb_under_cgroup:
 		return &bpf_skb_under_cgroup_proto;
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -9745,7 +9748,7 @@ sk_lookup_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 	case BPF_FUNC_sk_release:
 		return &bpf_sk_release_proto;
 	default:
-		return bpf_base_func_proto(func_id);
+		return bpf_sk_base_func_proto(func_id);
 	}
 }
 
@@ -9912,8 +9915,7 @@ const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto = {
 	.func			= bpf_skc_to_tcp6_sock,
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
-	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
+	.arg1_type		= ARG_PTR_TO_SOCK_COMMON,
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_TCP6],
 };
 
@@ -9929,8 +9931,7 @@ const struct bpf_func_proto bpf_skc_to_tcp_sock_proto = {
 	.func			= bpf_skc_to_tcp_sock,
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
-	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
+	.arg1_type		= ARG_PTR_TO_SOCK_COMMON,
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_TCP],
 };
 
@@ -9953,8 +9954,7 @@ const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto = {
 	.func			= bpf_skc_to_tcp_timewait_sock,
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
-	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
+	.arg1_type		= ARG_PTR_TO_SOCK_COMMON,
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_TCP_TW],
 };
 
@@ -9977,8 +9977,7 @@ const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto = {
 	.func			= bpf_skc_to_tcp_request_sock,
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
-	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
+	.arg1_type		= ARG_PTR_TO_SOCK_COMMON,
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_TCP_REQ],
 };
 
@@ -9999,7 +9998,37 @@ const struct bpf_func_proto bpf_skc_to_udp6_sock_proto = {
 	.func			= bpf_skc_to_udp6_sock,
 	.gpl_only		= false,
 	.ret_type		= RET_PTR_TO_BTF_ID_OR_NULL,
-	.arg1_type		= ARG_PTR_TO_BTF_ID,
-	.arg1_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_SOCK_COMMON],
+	.arg1_type		= ARG_PTR_TO_SOCK_COMMON,
 	.ret_btf_id		= &btf_sock_ids[BTF_SOCK_TYPE_UDP6],
 };
+
+static const struct bpf_func_proto *
+bpf_sk_base_func_proto(enum bpf_func_id func_id)
+{
+       const struct bpf_func_proto *func;
+
+       switch (func_id) {
+       case BPF_FUNC_skc_to_tcp6_sock:
+	       func = &bpf_skc_to_tcp6_sock_proto;
+	       break;
+       case BPF_FUNC_skc_to_tcp_sock:
+	       func = &bpf_skc_to_tcp_sock_proto;
+	       break;
+       case BPF_FUNC_skc_to_tcp_timewait_sock:
+	       func = &bpf_skc_to_tcp_timewait_sock_proto;
+	       break;
+       case BPF_FUNC_skc_to_tcp_request_sock:
+	       func = &bpf_skc_to_tcp_request_sock_proto;
+	       break;
+       case BPF_FUNC_skc_to_udp6_sock:
+	       func = &bpf_skc_to_udp6_sock_proto;
+	       break;
+       default:
+	       return bpf_base_func_proto(func_id);
+       }
+
+       if (!perfmon_capable())
+	       return NULL;
+
+       return func;
+}
-- 
2.24.1

Powered by blists - more mailing lists