Re: [PATCH bpf-next 1/7] bpf: Refactor trampoline update code

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Mar 3, 2020 at 2:24 PM KP Singh <kpsingh@xxxxxxxxxxxx> wrote:
>
> On 03-Mär 14:12, Andrii Nakryiko wrote:
> > On Tue, Mar 3, 2020 at 6:13 AM KP Singh <kpsingh@xxxxxxxxxxxx> wrote:
> > >
> > > From: KP Singh <kpsingh@xxxxxxxxxx>
> > >
> > > As we need to introduce a third type of attachment for trampolines, the
> > > flattened signature of arch_prepare_bpf_trampoline gets even more
> > > complicated.
> > >
> > > Refactor the prog and count argument to arch_prepare_bpf_trampoline to
> > > use bpf_tramp_progs to simplify the addition and accounting for new
> > > attachment types.
> > >
> > > Signed-off-by: KP Singh <kpsingh@xxxxxxxxxx>
> > > ---
> > >  arch/x86/net/bpf_jit_comp.c | 31 +++++++++---------
> > >  include/linux/bpf.h         | 13 ++++++--
> > >  kernel/bpf/bpf_struct_ops.c | 13 +++++++-
> > >  kernel/bpf/trampoline.c     | 63 +++++++++++++++++++++----------------
> > >  4 files changed, 75 insertions(+), 45 deletions(-)
> > >
> > > diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
> > > index 9ba08e9abc09..15c7d28bc05c 100644
> > > --- a/arch/x86/net/bpf_jit_comp.c
> > > +++ b/arch/x86/net/bpf_jit_comp.c
> > > @@ -1362,12 +1362,12 @@ static void restore_regs(const struct btf_func_model *m, u8 **prog, int nr_args,
> > >  }
> > >
> > >  static int invoke_bpf(const struct btf_func_model *m, u8 **pprog,
> > > -                     struct bpf_prog **progs, int prog_cnt, int stack_size)
> > > +                     struct bpf_tramp_progs *tp, int stack_size)
> >
> > nit: it's `tp` here, but `tprogs` in arch_prepare_bpf_trampoline. It's
> > minor, but would be nice to stick to consistent naming.
>
> I did this to ~distinguish~ that rather than being an array of
> tprogs it's a pointer to one of its members e.g.
> &tprogs[BPF_TRAMP_FEXIT]).
>
> I change it if you feel this is not a valuable disntinction.

I think it's important distinction, but naming doesn't really help
with it... Not sure how you can make it more clear, though.
>
> >
> > >  {
> > >         u8 *prog = *pprog;
> > >         int cnt = 0, i;
> > >
> > > -       for (i = 0; i < prog_cnt; i++) {
> > > +       for (i = 0; i < tp->nr_progs; i++) {
> > >                 if (emit_call(&prog, __bpf_prog_enter, prog))
> > >                         return -EINVAL;
> > >                 /* remember prog start time returned by __bpf_prog_enter */
> > > @@ -1376,17 +1376,17 @@ static int invoke_bpf(const struct btf_func_model *m, u8 **pprog,
> > >                 /* arg1: lea rdi, [rbp - stack_size] */
> > >                 EMIT4(0x48, 0x8D, 0x7D, -stack_size);
> > >                 /* arg2: progs[i]->insnsi for interpreter */
> > > -               if (!progs[i]->jited)
> > > +               if (!tp->progs[i]->jited)
> > >                         emit_mov_imm64(&prog, BPF_REG_2,
> > > -                                      (long) progs[i]->insnsi >> 32,
> > > -                                      (u32) (long) progs[i]->insnsi);
> > > +                                      (long) tp->progs[i]->insnsi >> 32,
> > > +                                      (u32) (long) tp->progs[i]->insnsi);
> > >                 /* call JITed bpf program or interpreter */
> > > -               if (emit_call(&prog, progs[i]->bpf_func, prog))
> > > +               if (emit_call(&prog, tp->progs[i]->bpf_func, prog))
> > >                         return -EINVAL;
> > >
> > >                 /* arg1: mov rdi, progs[i] */
> > > -               emit_mov_imm64(&prog, BPF_REG_1, (long) progs[i] >> 32,
> > > -                              (u32) (long) progs[i]);
> > > +               emit_mov_imm64(&prog, BPF_REG_1, (long) tp->progs[i] >> 32,
> > > +                              (u32) (long) tp->progs[i]);
> > >                 /* arg2: mov rsi, rbx <- start time in nsec */
> > >                 emit_mov_reg(&prog, true, BPF_REG_2, BPF_REG_6);
> > >                 if (emit_call(&prog, __bpf_prog_exit, prog))
> > > @@ -1458,12 +1458,13 @@ static int invoke_bpf(const struct btf_func_model *m, u8 **pprog,
> > >   */
> > >  int arch_prepare_bpf_trampoline(void *image, void *image_end,
> > >                                 const struct btf_func_model *m, u32 flags,
> > > -                               struct bpf_prog **fentry_progs, int fentry_cnt,
> > > -                               struct bpf_prog **fexit_progs, int fexit_cnt,
> > > +                               struct bpf_tramp_progs *tprogs,
> > >                                 void *orig_call)
> > >  {
> > >         int cnt = 0, nr_args = m->nr_args;
> > >         int stack_size = nr_args * 8;
> > > +       struct bpf_tramp_progs *fentry = &tprogs[BPF_TRAMP_FENTRY];
> > > +       struct bpf_tramp_progs *fexit = &tprogs[BPF_TRAMP_FEXIT];
> > >         u8 *prog;
> > >
> > >         /* x86-64 supports up to 6 arguments. 7+ can be added in the future */
> >
> > [...]
> >
> > > diff --git a/kernel/bpf/bpf_struct_ops.c b/kernel/bpf/bpf_struct_ops.c
> > > index c498f0fffb40..a011a77b21fa 100644
> > > --- a/kernel/bpf/bpf_struct_ops.c
> > > +++ b/kernel/bpf/bpf_struct_ops.c
> > > @@ -320,6 +320,7 @@ static int bpf_struct_ops_map_update_elem(struct bpf_map *map, void *key,
> > >         struct bpf_struct_ops_value *uvalue, *kvalue;
> > >         const struct btf_member *member;
> > >         const struct btf_type *t = st_ops->type;
> > > +       struct bpf_tramp_progs *tprogs = NULL;
> > >         void *udata, *kdata;
> > >         int prog_fd, err = 0;
> > >         void *image;
> > > @@ -425,10 +426,19 @@ static int bpf_struct_ops_map_update_elem(struct bpf_map *map, void *key,
> > >                         goto reset_unlock;
> > >                 }
> > >
> > > +               tprogs = kcalloc(BPF_TRAMP_MAX, sizeof(struct bpf_tramp_progs),
> >
> > nit: sizeof(*tprogs) ?
>
> Sure. I can fix it.
>
> >
> > > +                                GFP_KERNEL);
> > > +               if (!tprogs) {
> > > +                       err = -ENOMEM;
> > > +                       goto reset_unlock;
> > > +               }
> > > +
> > > +               *tprogs[BPF_TRAMP_FENTRY].progs = prog;
> >
> > I'm very confused what's going on here, why * at the beginning here,
> > but no * below?.. It seems unnecessary.
>
> The progs member of bpf_tramp_progs is
>
>   struct bpf_tramp_progs {
>         struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS];
>         int nr_progs;
>   };
>
> Equivalent to the **progs we had before in the signature of
> arch_prepare_bpf_trampoline.
>
>     *tprogs[BPF_TRAMP_FENTRY].progs = prog;
>
> is setting the program in the **progs. The one below is setting the
> count. Am I missing something :)

Ok, so it's setting entry 0 in bpf_tramp_progs->progs array, right?
Wouldn't it be less mind-bending and confusing written this way:

tprogs[BPF_TRAMP_FENTRY].progs[0] = prog;

?

Syntax you used treats fixed-length progs array as a pointer, which is
valid C, but not the best C either.

>
> >
> > > +               tprogs[BPF_TRAMP_FENTRY].nr_progs = 1;
> > >                 err = arch_prepare_bpf_trampoline(image,
> > >                                                   st_map->image + PAGE_SIZE,
> > >                                                   &st_ops->func_models[i], 0,
> > > -                                                 &prog, 1, NULL, 0, NULL);
> > > +                                                 tprogs, NULL);
> > >                 if (err < 0)
> > >                         goto reset_unlock;
> > >
> > > @@ -469,6 +479,7 @@ static int bpf_struct_ops_map_update_elem(struct bpf_map *map, void *key,
> > >         memset(uvalue, 0, map->value_size);
> > >         memset(kvalue, 0, map->value_size);
> > >  unlock:
> > > +       kfree(tprogs);
> > >         mutex_unlock(&st_map->lock);
> > >         return err;
> > >  }
> > > diff --git a/kernel/bpf/trampoline.c b/kernel/bpf/trampoline.c
> > > index 704fa787fec0..9daeb094f054 100644
> > > --- a/kernel/bpf/trampoline.c
> > > +++ b/kernel/bpf/trampoline.c
> > > @@ -190,40 +190,50 @@ static int register_fentry(struct bpf_trampoline *tr, void *new_addr)
> > >         return ret;
> > >  }
> > >
> > > -/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
> > > - * bytes on x86.  Pick a number to fit into BPF_IMAGE_SIZE / 2
> > > - */
> > > -#define BPF_MAX_TRAMP_PROGS 40
> > > +static struct bpf_tramp_progs *
> > > +bpf_trampoline_update_progs(struct bpf_trampoline *tr, int *total)
> > > +{
> > > +       struct bpf_tramp_progs *tprogs;
> > > +       struct bpf_prog **progs;
> > > +       struct bpf_prog_aux *aux;
> > > +       int kind;
> > > +
> > > +       *total = 0;
> > > +       tprogs = kcalloc(BPF_TRAMP_MAX, sizeof(struct bpf_tramp_progs),
> >
> > same nit as above, sizeof(*tprogs) is shorter and less error-prone
>
> Sure, can fix it.
>
> - KP
>
> >
> > > +                        GFP_KERNEL);
> > > +       if (!tprogs)
> > > +               return ERR_PTR(-ENOMEM);
> > > +
> > > +       for (kind = 0; kind < BPF_TRAMP_MAX; kind++) {
> > > +               tprogs[kind].nr_progs = tr->progs_cnt[kind];
> > > +               *total += tr->progs_cnt[kind];
> > > +               progs = tprogs[kind].progs;
> > > +
> > > +               hlist_for_each_entry(aux, &tr->progs_hlist[kind], tramp_hlist)
> > > +                       *progs++ = aux->prog;
> > > +       }
> > > +       return tprogs;
> > > +}
> > >
> >
> > [...]




[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux