From 514d6c1959f9b396f1b51850925900adedffb951 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Mon, 23 Apr 2018 12:11:02 -0700 Subject: bpf: Document sockmap '-target bpf' requirement for PROG_TYPE_SK_MSG BPF_PROG_TYPE_SK_MSG programs use a 'void *' for both data and the data_end pointers. Additionally, the verifier ensures that every accesses into the values is a __u64 read. This correctly maps on to the BPF 64-bit architecture. However, to ensure that when building on 32bit architectures that clang uses correct types the '-target bpf' option _must_ be specified. To make this clear add a note to the Documentation. Signed-off-by: John Fastabend Acked-by: Alexei Starovoitov Signed-off-by: Daniel Borkmann --- Documentation/bpf/bpf_devel_QA.txt | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'Documentation') diff --git a/Documentation/bpf/bpf_devel_QA.txt b/Documentation/bpf/bpf_devel_QA.txt index 1a0b704e1a38..da57601153a0 100644 --- a/Documentation/bpf/bpf_devel_QA.txt +++ b/Documentation/bpf/bpf_devel_QA.txt @@ -557,6 +557,14 @@ A: Although LLVM IR generation and optimization try to stay architecture pulls in some header files containing file scope host assembly codes. - You can add "-fno-jump-tables" to work around the switch table issue. - Otherwise, you can use bpf target. + Otherwise, you can use bpf target. Additionally, you _must_ use bpf target + when: + + - Your program uses data structures with pointer or long / unsigned long + types that interface with BPF helpers or context data structures. Access + into these structures is verified by the BPF verifier and may result + in verification failures if the native architecture is not aligned with + the BPF architecture, e.g. 64-bit. An example of this is + BPF_PROG_TYPE_SK_MSG require '-target bpf' Happy BPF hacking! -- cgit v1.2.3-59-g8ed1b