Re: [PATCH v6 bpf-next 04/17] bpf, x86: add new map type: instructions array

[Anton Protopopov <a.s.protopopov@gmail.com>]

On 25/10/21 10:49AM, Alexei Starovoitov wrote:
> On Sun, Oct 19, 2025 at 1:15 PM Anton Protopopov
> <a.s.protopopov@gmail.com> wrote:
> > diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
> > index d4c93d9e73e4..c8e628410d2c 100644
> > --- a/arch/x86/net/bpf_jit_comp.c
> > +++ b/arch/x86/net/bpf_jit_comp.c
> > @@ -1691,6 +1691,7 @@ static int do_jit(struct bpf_prog *bpf_prog, int *addrs, u8 *image, u8 *rw_image
> >         prog = temp;
> >
> >         for (i = 1; i <= insn_cnt; i++, insn++) {
> > +               u32 abs_xlated_off = bpf_prog->aux->subprog_start + i - 1;
> >                 const s32 imm32 = insn->imm;
> >                 u32 dst_reg = insn->dst_reg;
> >                 u32 src_reg = insn->src_reg;
> > @@ -2751,6 +2752,13 @@ st:                      if (is_imm8(insn->off))
> >                                 return -EFAULT;
> >                         }
> >                         memcpy(rw_image + proglen, temp, ilen);
> > +
> > +                       /*
> > +                        * Instruction arrays need to know how xlated code
> > +                        * maps to jitted code
> > +                        */
> > +                       bpf_prog_update_insn_ptr(bpf_prog, abs_xlated_off, proglen,
> > +                                                image + proglen);
> 
> ...
> 
> > +void bpf_prog_update_insn_ptr(struct bpf_prog *prog,
> > +                             u32 xlated_off,
> > +                             u32 jitted_off,
> > +                             void *jitted_ip)
> > +{
> > +       struct bpf_insn_array *insn_array;
> > +       struct bpf_map *map;
> > +       int i, j;
> > +
> > +       for (i = 0; i < prog->aux->used_map_cnt; i++) {
> > +               map = prog->aux->used_maps[i];
> > +               if (!is_insn_array(map))
> > +                       continue;
> > +
> > +               insn_array = cast_insn_array(map);
> > +               for (j = 0; j < map->max_entries; j++) {
> > +                       if (insn_array->ptrs[j].user_value.xlated_off == xlated_off) {
> > +                               insn_array->ips[j] = (long)jitted_ip;
> > +                               insn_array->ptrs[j].jitted_ip = jitted_ip;
> > +                               insn_array->ptrs[j].user_value.jitted_off = jitted_off;
> > +                       }
> > +               }
> > +       }
> > +}
> 
> This algorithm doesn't scale.
> You're calling bpf_prog_update_insn_ptr() for every insn
> and doing it as many times as they're JIT passes.
> There could be up to 20 passes and hundreds of thousands of insns.
> x86 JIT already keeps the mapping from insn to IP in jit_datat->addrs[].
> Use it. Roughly like:
> insn_array = cast_insn_array(map);
> for (j = 0; j < map->max_entries; j++) {
>    ip = addrs[insn_array->ptrs[j].user_value.xlated_off -
> subprog_start] + image;
> And this is done once per insn_array after JIT is done.

Thanks! Will do.