Merge branch 'bpf: Allow skb dynptr for tp_btf'

Philo Lu says:

====================
This makes bpf_dynptr_from_skb usable for tp_btf, so that we can easily
parse skb in tracepoints. This has been discussed in [0], and Martin
suggested to use dynptr (instead of helpers like bpf_skb_load_bytes).

For safety, skb dynptr shouldn't be used in fentry/fexit. This is achieved
by add KF_TRUSTED_ARGS flag in bpf_dynptr_from_skb defination, because
pointers passed by tracepoint are trusted (PTR_TRUSTED) while those of
fentry/fexit are not.

Another problem raises that NULL pointers could be passed to tracepoint,
such as trace_tcp_send_reset, and we need to recognize them. This is done
by add a "__nullable" suffix in the func_proto of the tracepoint,
discussed in [1].

2 Test cases are added, one for "__nullable" suffix, and the other for
using skb dynptr in tp_btf.

changelog
v2 -> v3 (Andrii Nakryiko):
 Patch 1:
  - Remove prog type check in prog_arg_maybe_null()
  - Add bpf_put_raw_tracepoint() after get()
  - Use kallsyms_lookup() instead of sprintf("%ps")
 Patch 2: Add separate test "tp_btf_nullable", and use full failure msg
v1 -> v2:
 - Add "__nullable" suffix support (Alexei Starovoitov)
 - Replace "struct __sk_buff*" with "void*" in test (Martin KaFai Lau)

[0]
https://lore.kernel.org/all/20240205121038.41344-1-lulie@linux.alibaba.com/T/
[1]
https://lore.kernel.org/all/20240430121805.104618-1-lulie@linux.alibaba.com/T/
====================

Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>

1 files changed, 3 insertions, 0 deletions

diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 520f49f422fe..7e03a98ccb7c 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -6523,6 +6523,9 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 	if (prog_args_trusted(prog))
 		info->reg_type |= PTR_TRUSTED;
 
+	if (btf_param_match_suffix(btf, &args[arg], "__nullable"))
+		info->reg_type |= PTR_MAYBE_NULL;
+
 	if (tgt_prog) {
 		enum bpf_prog_type tgt_type;