Thread (17 messages) 17 messages, 3 authors, 2026-01-24

Re: [PATCH bpf-next v12 04/13] bpf: support fsession for bpf_session_is_return

From: bot+bpf-ci@kernel.org
Date: 2026-01-24 04:03:32
Also in: bpf, lkml

quoted hunk ↗ jump to hunk
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 13f0a2de33b7..f7baeb8278ca 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
[ ... ]
quoted hunk ↗ jump to hunk
@@ -3341,34 +3349,39 @@ __bpf_kfunc __u64 *bpf_session_cookie(void *ctx)

 __bpf_kfunc_end_defs();

-BTF_KFUNCS_START(kprobe_multi_kfunc_set_ids)
+BTF_KFUNCS_START(session_kfunc_set_ids)
 BTF_ID_FLAGS(func, bpf_session_is_return)
 BTF_ID_FLAGS(func, bpf_session_cookie)
-BTF_KFUNCS_END(kprobe_multi_kfunc_set_ids)
+BTF_KFUNCS_END(session_kfunc_set_ids)

-static int bpf_kprobe_multi_filter(const struct bpf_prog *prog, u32 kfunc_id)
+static int bpf_session_filter(const struct bpf_prog *prog, u32 kfunc_id)
 {
-	if (!btf_id_set8_contains(&kprobe_multi_kfunc_set_ids, kfunc_id))
+	if (!btf_id_set8_contains(&session_kfunc_set_ids, kfunc_id))
 		return 0;

-	if (!is_kprobe_session(prog) && !is_uprobe_session(prog))
+	if (!is_kprobe_session(prog) && !is_uprobe_session(prog) && !is_trace_fsession(prog))
 		return -EACCES;
The filter now allows fsession programs to call both bpf_session_is_return()
and bpf_session_cookie(). The verifier inline handles bpf_session_is_return()
for fsession, but what happens when an fsession program calls
bpf_session_cookie()?

Looking at bpf_session_cookie() in the same file:

    __bpf_kfunc __u64 *bpf_session_cookie(void *ctx)
    {
        struct bpf_session_run_ctx *session_ctx;
        session_ctx = container_of(current->bpf_ctx, struct bpf_session_run_ctx, run_ctx);
        return session_ctx->data;
    }

This assumes current->bpf_ctx points to a struct bpf_session_run_ctx, but
fsession programs use struct bpf_tramp_run_ctx (set up in __bpf_prog_enter*).
The container_of() will return an incorrect pointer, and session_ctx->data
will read from the wrong memory location (the saved_run_ctx field of
bpf_tramp_run_ctx instead of the session cookie).

Is this intentional to defer bpf_session_cookie() support to a later commit?
If so, should the filter perhaps only allow bpf_session_is_return() for
fsession at this point?

Note: I see that commit b8f4610f8738 ("bpf: support fsession for
bpf_session_cookie") in the same series adds the verifier inline for
bpf_session_cookie() with fsession, which would resolve this. However, at
this commit, calling bpf_session_cookie() from an fsession program would
produce incorrect results.
 	return 0;
 }
[ ... ]
+static int __init bpf_trace_kfuncs_init(void)
 {
-	return register_btf_kfunc_id_set(BPF_PROG_TYPE_KPROBE, &bpf_kprobe_multi_kfunc_set);
+	int err = 0;
+
+	err = err ?: register_btf_kfunc_id_set(BPF_PROG_TYPE_KPROBE, &bpf_session_kfunc_set);
+	err = err ?: register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, &bpf_session_kfunc_set);
+
+	return err;
 }

---
AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md

CI run summary: https://github.com/kernel-patches/bpf/actions/runs/21308666922
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help