bpf: Compute data_end dynamically with JIT code

Currently, we compute ->data_end with a compile-time constant
offset of skb. But as Jakub pointed out, we can actually compute
it in eBPF JIT code at run-time, so that we can competely get
rid of ->data_end. This is similar to skb_shinfo(skb) computation
in bpf_convert_shinfo_access().

Suggested-by: Jakub Sitnicki <jakub@cloudflare.com>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/bpf/20210223184934.6054-4-xiyou.wangcong@gmail.com

diff --git a/include/net/tcp.h b/include/net/tcp.h
index c00e125..947ef5d 100644 (file)
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -886,18 +886,12 @@ struct tcp_skb_cb {
                struct {
                        __u32 flags;
                        struct sock *sk_redir;
-                       void *data_end;
                } bpf;
        };
 };
 
 #define TCP_SKB_CB(__skb)      ((struct tcp_skb_cb *)&((__skb)->cb[0]))
 
-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 inline bool tcp_skb_bpf_ingress(const struct sk_buff *skb)
 {
        return TCP_SKB_CB(skb)->bpf.flags & BPF_F_INGRESS;

diff --git a/net/core/filter.c b/net/core/filter.c
index adfdad2..13bcf24 100644 (file)
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -1863,10 +1863,7 @@ static const struct bpf_func_proto bpf_sk_fullsock_proto = {
 static inline int sk_skb_try_make_writable(struct sk_buff *skb,
                                           unsigned int write_len)
 {
-       int err = __bpf_try_make_writable(skb, write_len);
-
-       bpf_compute_data_end_sk_skb(skb);
-       return err;
+       return __bpf_try_make_writable(skb, write_len);
 }
 
 BPF_CALL_2(sk_skb_pull_data, struct sk_buff *, skb, u32, len)
@@ -3577,7 +3574,6 @@ BPF_CALL_4(sk_skb_adjust_room, struct sk_buff *, skb, s32, len_diff,
                        return -ENOMEM;
                __skb_pull(skb, len_diff_abs);
        }
-       bpf_compute_data_end_sk_skb(skb);
        if (tls_sw_has_ctx_rx(skb->sk)) {
                struct strp_msg *rxm = strp_msg(skb);
 
@@ -3742,10 +3738,7 @@ static const struct bpf_func_proto bpf_skb_change_tail_proto = {
 BPF_CALL_3(sk_skb_change_tail, struct sk_buff *, skb, u32, new_len,
           u64, flags)
 {
-       int ret = __bpf_skb_change_tail(skb, new_len, flags);
-
-       bpf_compute_data_end_sk_skb(skb);
-       return ret;
+       return __bpf_skb_change_tail(skb, new_len, flags);
 }
 
 static const struct bpf_func_proto sk_skb_change_tail_proto = {
@@ -3808,10 +3801,7 @@ static const struct bpf_func_proto bpf_skb_change_head_proto = {
 BPF_CALL_3(sk_skb_change_head, struct sk_buff *, skb, u32, head_room,
           u64, flags)
 {
-       int ret = __bpf_skb_change_head(skb, head_room, flags);
-
-       bpf_compute_data_end_sk_skb(skb);
-       return ret;
+       return __bpf_skb_change_head(skb, head_room, flags);
 }
 
 static const struct bpf_func_proto sk_skb_change_head_proto = {
@@ -9655,22 +9645,40 @@ static u32 sock_ops_convert_ctx_access(enum bpf_access_type type,
        return insn - insn_buf;
 }
 
+/* data_end = skb->data + skb_headlen() */
+static struct bpf_insn *bpf_convert_data_end_access(const struct bpf_insn *si,
+                                                   struct bpf_insn *insn)
+{
+       /* si->dst_reg = skb->data */
+       *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct sk_buff, data),
+                             si->dst_reg, si->src_reg,
+                             offsetof(struct sk_buff, data));
+       /* AX = skb->len */
+       *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct sk_buff, len),
+                             BPF_REG_AX, si->src_reg,
+                             offsetof(struct sk_buff, len));
+       /* si->dst_reg = skb->data + skb->len */
+       *insn++ = BPF_ALU64_REG(BPF_ADD, si->dst_reg, BPF_REG_AX);
+       /* AX = skb->data_len */
+       *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(struct sk_buff, data_len),
+                             BPF_REG_AX, si->src_reg,
+                             offsetof(struct sk_buff, data_len));
+       /* si->dst_reg = skb->data + skb->len - skb->data_len */
+       *insn++ = BPF_ALU64_REG(BPF_SUB, si->dst_reg, BPF_REG_AX);
+
+       return insn;
+}
+
 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);
+               insn = bpf_convert_data_end_access(si, insn);
                break;
        default:
                return bpf_convert_ctx_access(type, si, insn_buf, prog,

diff --git a/net/core/skmsg.c b/net/core/skmsg.c
index d00c9a4..8822001 100644 (file)
--- a/net/core/skmsg.c
+++ b/net/core/skmsg.c
@@ -746,7 +746,6 @@ EXPORT_SYMBOL_GPL(sk_psock_msg_verdict);
 static int sk_psock_bpf_run(struct sk_psock *psock, struct bpf_prog *prog,
                            struct sk_buff *skb)
 {
-       bpf_compute_data_end_sk_skb(skb);
        return bpf_prog_run_pin_on_cpu(prog, skb);
 }