[PATCH bpf-next v3 2/3] bpf, riscv: add fsession support for trampolines
Björn Töpel
bjorn at kernel.org
Fri Feb 6 05:42:22 PST 2026
Menglong!
On Fri, 6 Feb 2026 at 13:51, <bot+bpf-ci at kernel.org> wrote:
>
> > diff --git a/arch/riscv/net/bpf_jit_comp64.c b/arch/riscv/net/bpf_jit_comp64.c
> > --- a/arch/riscv/net/bpf_jit_comp64.c
> > +++ b/arch/riscv/net/bpf_jit_comp64.c
>
> [ ... ]
>
> > +static int invoke_bpf(struct bpf_tramp_links *tl, int args_off, int retval_off,
> > + int run_ctx_off, int func_meta_off, bool save_ret, u64 func_meta,
> > + int cookie_off, struct rv_jit_context *ctx)
> > +{
> > + int i, cur_cookie = (cookie_off - args_off) / 8;
> > +
> > + for (i = 0; i < tl->nr_links; i++) {
> > + int err;
> > +
> > + if (bpf_prog_calls_session_cookie(&tl->links[i])) {
> ^^^^^^^^^^^^^^
>
> Should this be tl->links[i] instead of &tl->links[i]?
>
> The function bpf_prog_calls_session_cookie() expects a struct bpf_tramp_link
> pointer, but tl->links is an array of pointers, so tl->links[i] is already
> a struct bpf_tramp_link pointer. The extra & passes a pointer-to-pointer.
>
> For comparison, arm64 uses:
>
> if (bpf_prog_calls_session_cookie(fentry->links[i]))
>
> And include/linux/bpf.h uses:
>
> if (bpf_prog_calls_session_cookie(fentries.links[i]))
>
> Both without the &.
>
> > + u64 meta = func_meta | ((u64)cur_cookie << BPF_TRAMP_COOKIE_INDEX_SHIFT);
> > +
> > + emit_store_stack_imm64(RV_REG_T1, -func_meta_off, meta, ctx);
> > + cur_cookie--;
> > + }
> > + err = invoke_bpf_prog(tl->links[i], args_off, retval_off, run_ctx_off,
> > + save_ret, ctx);
>
> Note that invoke_bpf_prog() correctly uses tl->links[i] without the &.
Menglong, the v3 doesn't build. Please refrain from just stressing
versions out. For v4, make sure you properly build/test. On that note;
on what hardware/simulator did you perform the tests on?
More information about the linux-riscv
mailing list