From: John Fastabend <john.fastabend@gmail.com>
Date: Fri, 27 Oct 2017 09:45:34 -0700
Subject: bpf: bpf_compute_data uses incorrect cb structure
Patch-mainline: v4.14-rc7
Git-commit: 8108a77515126f6db4374e8593956e20430307c0
References: bsc#1109837

SK_SKB program types use bpf_compute_data to store the end of the
packet data. However, bpf_compute_data assumes the cb is stored in the
qdisc layer format. But, for SK_SKB this is the wrong layer of the
stack for this type.

It happens to work (sort of!) because in most cases nothing happens
to be overwritten today. This is very fragile and error prone.
Fortunately, we have another hole in tcp_skb_cb we can use so lets
put the data_end value there.

Note, SK_SKB program types do not use data_meta, they are failed by
sk_skb_is_valid_access().

Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
 include/net/tcp.h    |    1 +
 kernel/bpf/sockmap.c |   12 ++++++++++--
 net/core/filter.c    |   27 ++++++++++++++++++++++++++-
 3 files changed, 37 insertions(+), 3 deletions(-)

--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -835,6 +835,7 @@ struct tcp_skb_cb {
 			__u32 key;
 			__u32 flags;
 			struct bpf_map *map;
+			void *data_end;
 		} bpf;
 	};
 };
--- a/kernel/bpf/sockmap.c
+++ b/kernel/bpf/sockmap.c
@@ -93,6 +93,14 @@ static inline struct smap_psock *smap_ps
 	return rcu_dereference_sk_user_data(sk);
 }
 
+/* compute the linear packet data range [data, data_end) for skb when
+ * sk_skb type programs are in use.
+ */
+static inline void bpf_compute_data_end_sk_skb(struct sk_buff *skb)
+{
+	TCP_SKB_CB(skb)->bpf.data_end = skb->data + skb_headlen(skb);
+}
+
 static int smap_verdict_func(struct smap_psock *psock, struct sk_buff *skb)
 {
 	struct bpf_prog *prog = READ_ONCE(psock->bpf_verdict);
@@ -108,7 +116,7 @@ static int smap_verdict_func(struct smap
 	 */
 	TCP_SKB_CB(skb)->bpf.map = NULL;
 	skb->sk = psock->sock;
-	bpf_compute_data_end(skb);
+	bpf_compute_data_end_sk_skb(skb);
 	preempt_disable();
 	rc = (*prog->bpf_func)(skb, prog->insnsi);
 	preempt_enable();
@@ -368,7 +376,7 @@ static int smap_parse_func_strparser(str
 	 * any socket yet.
 	 */
 	skb->sk = psock->sock;
-	bpf_compute_data_end(skb);
+	bpf_compute_data_end_sk_skb(skb);
 	rc = (*prog->bpf_func)(skb, prog->insnsi);
 	skb->sk = NULL;
 	rcu_read_unlock();
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -4242,6 +4242,31 @@ static u32 sock_ops_convert_ctx_access(e
 	return insn - insn_buf;
 }
 
+static u32 sk_skb_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;
+	int off;
+
+	switch (si->off) {
+	case offsetof(struct __sk_buff, data_end):
+		off  = si->off;
+		off -= offsetof(struct __sk_buff, data_end);
+		off += offsetof(struct sk_buff, cb);
+		off += offsetof(struct tcp_skb_cb, bpf.data_end);
+		*insn++ = BPF_LDX_MEM(BPF_SIZEOF(void *), si->dst_reg,
+				      si->src_reg, off);
+		break;
+	default:
+		return bpf_convert_ctx_access(type, si, insn_buf, prog,
+					      target_size);
+	}
+
+	return insn - insn_buf;
+}
+
 const struct bpf_verifier_ops sk_filter_prog_ops = {
 	.get_func_proto		= sk_filter_func_proto,
 	.is_valid_access	= sk_filter_is_valid_access,
@@ -4300,7 +4325,7 @@ const struct bpf_verifier_ops sock_ops_p
 const struct bpf_verifier_ops sk_skb_prog_ops = {
 	.get_func_proto		= sk_skb_func_proto,
 	.is_valid_access	= sk_skb_is_valid_access,
-	.convert_ctx_access	= bpf_convert_ctx_access,
+	.convert_ctx_access	= sk_skb_convert_ctx_access,
 	.gen_prologue		= sk_skb_prologue,
 };