[PATCH rfc 03/12] bpf: treat fmodret tracing program's arguments as trusted

[Roman Gushchin <roman.gushchin@linux.dev>]

*** DO NOT MERGE! ***

This is a temporarily workaround, which will be fixed/replaced
in the next version.

--

Bpf oom handler hook has to:
1) have a trusted pointer to the oom_control structure,
2) return a value,
3) be sleepable to use cgroup iterator functions.

fmodret tracing programs fulfill 2) and 3).
This patch enables 1), however this change contradicts
the commit c6b0337f0120 ("bpf: Don't mark arguments to fentry/fexit
programs as trusted.").

Signed-off-by: Roman Gushchin <roman.gushchin@linux.dev>
---
 kernel/bpf/btf.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index a91822bae043..aa86c4eabfa0 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -6424,7 +6424,14 @@ static bool prog_args_trusted(const struct bpf_prog *prog)
 
 	switch (prog->type) {
 	case BPF_PROG_TYPE_TRACING:
-		return atype == BPF_TRACE_RAW_TP || atype == BPF_TRACE_ITER;
+		switch (atype) {
+		case BPF_TRACE_RAW_TP:
+		case BPF_TRACE_ITER:
+		case BPF_MODIFY_RETURN:
+			return true;
+		default:
+			return false;
+		}
 	case BPF_PROG_TYPE_LSM:
 		return bpf_lsm_is_trusted(prog);
 	case BPF_PROG_TYPE_STRUCT_OPS:
-- 
2.49.0.901.g37484f566f-goog