MIME-Version: 1.0
References: <20230203031742.1730761-1-imagedong@tencent.com> <20230203031742.1730761-2-imagedong@tencent.com>
In-Reply-To: <20230203031742.1730761-2-imagedong@tencent.com>
From:   Andrii Nakryiko <andrii.nakryiko@gmail.com>
Date:   Mon, 6 Feb 2023 11:58:51 -0800
Message-ID: <CAEf4Bzbig9DmCEJd0i64gA=jzczK7n=joqiD0MAa6sFSgd=WAw@mail.gmail.com>
Subject: Re: [PATCH bpf-next 1/2] libbpf: add support to set kprobe/uprobe
 attach mode
To:     menglong8.dong@gmail.com
Cc:     alan.maguire@oracle.com, ast@kernel.org, daniel@iogearbox.net,
        andrii@kernel.org, martin.lau@linux.dev, song@kernel.org,
        yhs@fb.com, john.fastabend@gmail.com, kpsingh@kernel.org,
        sdf@google.com, haoluo@google.com, jolsa@kernel.org,
        bpf@vger.kernel.org, linux-kernel@vger.kernel.org,
        Menglong Dong <imagedong@tencent.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Feb 2, 2023 at 7:18 PM <menglong8.dong@gmail.com> wrote:
>
> From: Menglong Dong <imagedong@tencent.com>
>
> By default, libbpf will attach the kprobe/uprobe eBPF program in the
> latest mode that supported by kernel. In this patch, we add the support
> to let users manually attach kprobe/uprobe in legacy or perf mode.
>
> There are 3 mode that supported by the kernel to attach kprobe/uprobe:
>
>   LEGACY: create perf event in legacy way and don't use bpf_link
>   PERF: create perf event with perf_event_open() and don't use bpf_link
>   LINK: create perf event with perf_event_open() and use bpf_link
>
> Users now can manually choose the mode with
> bpf_program__attach_uprobe_opts()/bpf_program__attach_kprobe_opts().
>
> Link: https://lore.kernel.org/bpf/20230113093427.1666466-1-imagedong@tencent.com/
> Signed-off-by: Menglong Dong <imagedong@tencent.com>
> ---
>  tools/lib/bpf/libbpf.c | 26 +++++++++++++++++++++++++-
>  tools/lib/bpf/libbpf.h | 19 ++++++++++++++++---
>  2 files changed, 41 insertions(+), 4 deletions(-)
>
> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index eed5cec6f510..0d20bf1ee301 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c
> @@ -9784,7 +9784,7 @@ struct bpf_link *bpf_program__attach_perf_event_opts(const struct bpf_program *p
>         link->link.dealloc = &bpf_link_perf_dealloc;
>         link->perf_event_fd = pfd;
>
> -       if (kernel_supports(prog->obj, FEAT_PERF_LINK)) {
> +       if (kernel_supports(prog->obj, FEAT_PERF_LINK) && !opts->no_link) {
>                 DECLARE_LIBBPF_OPTS(bpf_link_create_opts, link_opts,
>                         .perf_event.bpf_cookie = OPTS_GET(opts, bpf_cookie, 0));
>
> @@ -10148,16 +10148,28 @@ bpf_program__attach_kprobe_opts(const struct bpf_program *prog,
>         struct bpf_link *link;
>         size_t offset;
>         bool retprobe, legacy;
> +       enum probe_mode mode;
>         int pfd, err;
>
>         if (!OPTS_VALID(opts, bpf_kprobe_opts))
>                 return libbpf_err_ptr(-EINVAL);
>
> +       mode = OPTS_GET(opts, mode, PROBE_MODE_DEFAULT);
>         retprobe = OPTS_GET(opts, retprobe, false);
>         offset = OPTS_GET(opts, offset, 0);
>         pe_opts.bpf_cookie = OPTS_GET(opts, bpf_cookie, 0);
>
>         legacy = determine_kprobe_perf_type() < 0;
> +       switch (mode) {
> +       case PROBE_MODE_LEGACY:
> +               legacy = true;
> +       case PROBE_MODE_PERF:
> +               pe_opts.no_link = true;
> +               break;
> +       default:
> +               break;
> +       }
> +
>         if (!legacy) {
>                 pfd = perf_event_open_probe(false /* uprobe */, retprobe,
>                                             func_name, offset,
> @@ -10817,10 +10829,12 @@ bpf_program__attach_uprobe_opts(const struct bpf_program *prog, pid_t pid,
>         int pfd, err;
>         bool retprobe, legacy;
>         const char *func_name;
> +       enum probe_mode mode;
>
>         if (!OPTS_VALID(opts, bpf_uprobe_opts))
>                 return libbpf_err_ptr(-EINVAL);
>
> +       mode = OPTS_GET(opts, mode, PROBE_MODE_DEFAULT);
>         retprobe = OPTS_GET(opts, retprobe, false);
>         ref_ctr_off = OPTS_GET(opts, ref_ctr_offset, 0);
>         pe_opts.bpf_cookie = OPTS_GET(opts, bpf_cookie, 0);
> @@ -10849,6 +10863,16 @@ bpf_program__attach_uprobe_opts(const struct bpf_program *prog, pid_t pid,
>         }
>
>         legacy = determine_uprobe_perf_type() < 0;
> +       switch (mode) {
> +       case PROBE_MODE_LEGACY:
> +               legacy = true;
> +       case PROBE_MODE_PERF:
> +               pe_opts.no_link = true;
> +               break;
> +       default:
> +               break;
> +       }
> +

I think this is a good place to also return errors early if, say, user
requested LINK mode, but that mode is not supported by kernel. Instead
of returning some -ENOTSUP generic error, we can error out early?
Similar for PERF mode, if only legacy is supported, and so on. Similar
for attach_uprobe, of course.

>         if (!legacy) {
>                 pfd = perf_event_open_probe(true /* uprobe */, retprobe, binary_path,
>                                             func_offset, pid, ref_ctr_off);
> diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
> index 8777ff21ea1d..7fb474e036a3 100644
> --- a/tools/lib/bpf/libbpf.h
> +++ b/tools/lib/bpf/libbpf.h
> @@ -451,8 +451,10 @@ struct bpf_perf_event_opts {
>         size_t sz;
>         /* custom user-provided value fetchable through bpf_get_attach_cookie() */
>         __u64 bpf_cookie;
> +       /* don't use bpf_link when attach eBPF pprogram */

typo: pprogram

> +       bool no_link;

I've been struggling with this "no_link" name a bit. It is quite
confusing considering that from libbpf's API side we do return `struct
bpf_link`. So I'm thinking that maybe "force_ioctl_attach" would be a
better way to describe it (and will be scary enough for people not
knowing what this is about to not set it to true?)

Also, we'll need size_t: 0 at the end to avoid uninitialized padding
issues (like we have in kprobe_opts and uprobe_opts)

>  };
> -#define bpf_perf_event_opts__last_field bpf_cookie
> +#define bpf_perf_event_opts__last_field no_link
>
>  LIBBPF_API struct bpf_link *
>  bpf_program__attach_perf_event(const struct bpf_program *prog, int pfd);
> @@ -461,6 +463,13 @@ LIBBPF_API struct bpf_link *
>  bpf_program__attach_perf_event_opts(const struct bpf_program *prog, int pfd,
>                                     const struct bpf_perf_event_opts *opts);
>
> +enum probe_mode {

shall we call it probe_attach_mode?

also let's elaborate a bit more in doc comment that specifying mode
will force libbpf to use that, but if kernel doesn't support it --
then we'll error out.

> +       PROBE_MODE_DEFAULT = 0, /* latest supported by kernel */
> +       PROBE_MODE_LEGACY,
> +       PROBE_MODE_PERF,
> +       PROBE_MODE_LINK,
> +};
> +
>  struct bpf_kprobe_opts {
>         /* size of this struct, for forward/backward compatiblity */
>         size_t sz;
> @@ -470,9 +479,11 @@ struct bpf_kprobe_opts {
>         size_t offset;
>         /* kprobe is return probe */
>         bool retprobe;
> +       /* kprobe attach mode */
> +       enum probe_mode mode;

nit: mode -> attach_mode?

>         size_t :0;
>  };
> -#define bpf_kprobe_opts__last_field retprobe
> +#define bpf_kprobe_opts__last_field mode
>
>  LIBBPF_API struct bpf_link *
>  bpf_program__attach_kprobe(const struct bpf_program *prog, bool retprobe,
> @@ -570,9 +581,11 @@ struct bpf_uprobe_opts {
>          * binary_path.
>          */
>         const char *func_name;
> +       /* uprobe attach mode */
> +       enum probe_mode mode;
>         size_t :0;
>  };
> -#define bpf_uprobe_opts__last_field func_name
> +#define bpf_uprobe_opts__last_field mode

ditto

>
>  /**
>   * @brief **bpf_program__attach_uprobe()** attaches a BPF program
> --
> 2.39.0
>