bpf: Emit struct bpf_tcp_sock type in vmlinux BTF
authorYonghong Song <yhs@fb.com>
Tue, 2 May 2023 18:05:43 +0000 (11:05 -0700)
committerAndrii Nakryiko <andrii@kernel.org>
Tue, 2 May 2023 23:19:06 +0000 (16:19 -0700)
commitbf6882aebd0ea67558764d44ddeffc7d253c6eb8
treee843ea0a3518abe191b3ddf3325f57b0f4c1ffe6
parent577c34b3be6e55ade4e0391022b9a539b20d77c5
bpf: Emit struct bpf_tcp_sock type in vmlinux BTF

In one of our internal testing, we found a case where
  - uapi struct bpf_tcp_sock is in vmlinux.h where vmlinux.h is not
    generated from the testing kernel
  - struct bpf_tcp_sock is not in vmlinux BTF

The above combination caused bpf load failure as the following
memory access
  struct bpf_tcp_sock *tcp_sock = ...;
  ... tcp_sock->snd_cwnd ...
needs CORE relocation but the relocation cannot be resolved since
the kernel BTF does not have corresponding type.

Similar to other previous cases (nf_conn___init, tcp6_sock, mctcp_sock, etc.),
add the type to vmlinux BTF with BTF_EMIT_TYPE macro.

Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20230502180543.1832140-1-yhs@fb.com
net/core/filter.c