bpf: Make dst_port field in struct bpf_sock 16-bit wide

Menglong Dong reports that the documentation for the dst_port field in
struct bpf_sock is inaccurate and confusing. From the BPF program PoV, the
field is a zero-padded 16-bit integer in network byte order. The value
appears to the BPF user as if laid out in memory as so:

  offsetof(struct bpf_sock, dst_port) + 0  <port MSB>
                                      + 8  <port LSB>
                                      +16  0x00
                                      +24  0x00

32-, 16-, and 8-bit wide loads from the field are all allowed, but only if
the offset into the field is 0.

32-bit wide loads from dst_port are especially confusing. The loaded value,
after converting to host byte order with bpf_ntohl(dst_port), contains the
port number in the upper 16-bits.

Remove the confusion by splitting the field into two 16-bit fields. For
backward compatibility, allow 32-bit wide loads from offsetof(struct
bpf_sock, dst_port).

While at it, allow loads 8-bit loads at offset [0] and [1] from dst_port.

Reported-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
Link: https://lore.kernel.org/r/20220130115518.213259-2-jakub@cloudflare.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

1 files changed, 2 insertions, 1 deletions

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 4a2f7041ebae..a7f0ddedac1f 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -5574,7 +5574,8 @@ struct bpf_sock {
 	__u32 src_ip4;
 	__u32 src_ip6[4];
 	__u32 src_port;		/* host byte order */
-	__u32 dst_port;		/* network byte order */
+	__be16 dst_port;	/* network byte order */
+	__u16 :16;		/* zero padding */
 	__u32 dst_ip4;
 	__u32 dst_ip6[4];
 	__u32 state;