[PATCH bpf-next] libbpf: Add support for dynamic program attach target

[PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Eelco Chaudron]

Currently when you want to attach a trace program to a bpf program
the section name needs to match the tracepoint/function semantics.

However the addition of the bpf_program__set_attach_target() API
allows you to specify the tracepoint/function dynamically.

The call flow would look something like this:

  xdp_fd = bpf_prog_get_fd_by_id(id);
  trace_obj = bpf_object__open_file("func.o", NULL);
  prog = bpf_object__find_program_by_title(trace_obj,
                                           "fentry/myfunc");
  bpf_program__set_attach_target(prog, xdp_fd,
                                 "fentry/xdpfilt_blk_all");
  bpf_object__load(trace_obj)

Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
---
 tools/lib/bpf/libbpf.c   |   41 +++++++++++++++++++++++++++++++++++------
 tools/lib/bpf/libbpf.h   |    4 ++++
 tools/lib/bpf/libbpf.map |    1 +
 3 files changed, 40 insertions(+), 6 deletions(-)

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 514b1a524abb..2ce879c301bb 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -4933,15 +4933,16 @@ load_program(struct bpf_program *prog, struct bpf_insn *insns, int insns_cnt,
 	return ret;
 }
 
-static int libbpf_find_attach_btf_id(struct bpf_program *prog);
+static int libbpf_find_attach_btf_id(struct bpf_program *prog,
+				     const char *name);
 
 int bpf_program__load(struct bpf_program *prog, char *license, __u32 kern_ver)
 {
 	int err = 0, fd, i, btf_id;
 
-	if (prog->type == BPF_PROG_TYPE_TRACING ||
-	    prog->type == BPF_PROG_TYPE_EXT) {
-		btf_id = libbpf_find_attach_btf_id(prog);
+	if ((prog->type == BPF_PROG_TYPE_TRACING ||
+	     prog->type == BPF_PROG_TYPE_EXT) && !prog->attach_btf_id) {
+		btf_id = libbpf_find_attach_btf_id(prog, NULL);
 		if (btf_id <= 0)
 			return btf_id;
 		prog->attach_btf_id = btf_id;
@@ -6202,6 +6203,31 @@ void bpf_program__set_expected_attach_type(struct bpf_program *prog,
 	prog->expected_attach_type = type;
 }
 
+int bpf_program__set_attach_target(struct bpf_program *prog,
+				   int attach_prog_fd,
+				   const char *attach_func_name)
+{
+	__u32 org_attach_prog_fd;
+	int btf_id;
+
+	if (!prog || attach_prog_fd < 0 || !attach_func_name)
+		return -EINVAL;
+
+	org_attach_prog_fd = prog->attach_prog_fd;
+	prog->attach_prog_fd = attach_prog_fd;
+
+	btf_id = libbpf_find_attach_btf_id(prog,
+					   attach_func_name);
+
+	if (btf_id < 0) {
+		prog->attach_prog_fd = org_attach_prog_fd;
+		return btf_id;
+	}
+
+	prog->attach_btf_id = btf_id;
+	return 0;
+}
+
 #define BPF_PROG_SEC_IMPL(string, ptype, eatype, is_attachable, btf, atype) \
 	{ string, sizeof(string) - 1, ptype, eatype, is_attachable, btf, atype }
 
@@ -6633,13 +6659,16 @@ static int libbpf_find_prog_btf_id(const char *name, __u32 attach_prog_fd)
 	return err;
 }
 
-static int libbpf_find_attach_btf_id(struct bpf_program *prog)
+static int libbpf_find_attach_btf_id(struct bpf_program *prog,
+				     const char *name)
 {
 	enum bpf_attach_type attach_type = prog->expected_attach_type;
 	__u32 attach_prog_fd = prog->attach_prog_fd;
-	const char *name = prog->section_name;
 	int i, err;
 
+	if (!name)
+		name = prog->section_name;
+
 	if (!name)
 		return -EINVAL;
 
diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
index 3fe12c9d1f92..02fc58a21a7f 100644
--- a/tools/lib/bpf/libbpf.h
+++ b/tools/lib/bpf/libbpf.h
@@ -334,6 +334,10 @@ LIBBPF_API void
 bpf_program__set_expected_attach_type(struct bpf_program *prog,
 				      enum bpf_attach_type type);
 
+LIBBPF_API int
+bpf_program__set_attach_target(struct bpf_program *prog, int attach_prog_fd,
+			       const char *attach_func_name);
+
 LIBBPF_API bool bpf_program__is_socket_filter(const struct bpf_program *prog);
 LIBBPF_API bool bpf_program__is_tracepoint(const struct bpf_program *prog);
 LIBBPF_API bool bpf_program__is_raw_tracepoint(const struct bpf_program *prog);
diff --git a/tools/lib/bpf/libbpf.map b/tools/lib/bpf/libbpf.map
index b035122142bb..8aba5438a3f0 100644
--- a/tools/lib/bpf/libbpf.map
+++ b/tools/lib/bpf/libbpf.map
@@ -230,6 +230,7 @@ LIBBPF_0.0.7 {
 		bpf_program__name;
 		bpf_program__is_extension;
 		bpf_program__is_struct_ops;
+		bpf_program__set_attach_target;
 		bpf_program__set_extension;
 		bpf_program__set_struct_ops;
 		btf__align_of;

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Toke Høiland-Jørgensen]

Eelco Chaudron <echaudro@redhat.com> writes:

> Currently when you want to attach a trace program to a bpf program
> the section name needs to match the tracepoint/function semantics.
>
> However the addition of the bpf_program__set_attach_target() API
> allows you to specify the tracepoint/function dynamically.
>
> The call flow would look something like this:
>
>   xdp_fd = bpf_prog_get_fd_by_id(id);
>   trace_obj = bpf_object__open_file("func.o", NULL);
>   prog = bpf_object__find_program_by_title(trace_obj,
>                                            "fentry/myfunc");
>   bpf_program__set_attach_target(prog, xdp_fd,
>                                  "fentry/xdpfilt_blk_all");

I think it would be better to have the attach type as a separate arg
instead of encoding it in the function name. I.e., rather:

   bpf_program__set_attach_target(prog, xdp_fd,
                                  "xdpfilt_blk_all", BPF_TRACE_FENTRY);

-Toke

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Andrii Nakryiko]

On Wed, Feb 12, 2020 at 4:32 AM Eelco Chaudron <echaudro@redhat.com> wrote:
>
> Currently when you want to attach a trace program to a bpf program
> the section name needs to match the tracepoint/function semantics.
>
> However the addition of the bpf_program__set_attach_target() API
> allows you to specify the tracepoint/function dynamically.
>
> The call flow would look something like this:
>
>   xdp_fd = bpf_prog_get_fd_by_id(id);
>   trace_obj = bpf_object__open_file("func.o", NULL);
>   prog = bpf_object__find_program_by_title(trace_obj,
>                                            "fentry/myfunc");
>   bpf_program__set_attach_target(prog, xdp_fd,
>                                  "fentry/xdpfilt_blk_all");
>   bpf_object__load(trace_obj)
>
> Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
> ---
>  tools/lib/bpf/libbpf.c   |   41 +++++++++++++++++++++++++++++++++++------
>  tools/lib/bpf/libbpf.h   |    4 ++++
>  tools/lib/bpf/libbpf.map |    1 +
>  3 files changed, 40 insertions(+), 6 deletions(-)
>
> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index 514b1a524abb..2ce879c301bb 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c
> @@ -4933,15 +4933,16 @@ load_program(struct bpf_program *prog, struct bpf_insn *insns, int insns_cnt,
>         return ret;
>  }
>
> -static int libbpf_find_attach_btf_id(struct bpf_program *prog);
> +static int libbpf_find_attach_btf_id(struct bpf_program *prog,
> +                                    const char *name);
>
>  int bpf_program__load(struct bpf_program *prog, char *license, __u32 kern_ver)
>  {
>         int err = 0, fd, i, btf_id;
>
> -       if (prog->type == BPF_PROG_TYPE_TRACING ||
> -           prog->type == BPF_PROG_TYPE_EXT) {
> -               btf_id = libbpf_find_attach_btf_id(prog);
> +       if ((prog->type == BPF_PROG_TYPE_TRACING ||
> +            prog->type == BPF_PROG_TYPE_EXT) && !prog->attach_btf_id) {
> +               btf_id = libbpf_find_attach_btf_id(prog, NULL);
>                 if (btf_id <= 0)
>                         return btf_id;
>                 prog->attach_btf_id = btf_id;
> @@ -6202,6 +6203,31 @@ void bpf_program__set_expected_attach_type(struct bpf_program *prog,
>         prog->expected_attach_type = type;
>  }
>
> +int bpf_program__set_attach_target(struct bpf_program *prog,
> +                                  int attach_prog_fd,
> +                                  const char *attach_func_name)
> +{
> +       __u32 org_attach_prog_fd;
> +       int btf_id;
> +
> +       if (!prog || attach_prog_fd < 0 || !attach_func_name)
> +               return -EINVAL;
> +
> +       org_attach_prog_fd = prog->attach_prog_fd;
> +       prog->attach_prog_fd = attach_prog_fd;
> +
> +       btf_id = libbpf_find_attach_btf_id(prog,
> +                                          attach_func_name);
> +
> +       if (btf_id < 0) {
> +               prog->attach_prog_fd = org_attach_prog_fd;

I don't think there is a need to restore original attach_prog_fd (most
probably it's going to be invalid either way). If explicit
set_attach_target fails, user application will have to fail or do some
other set_attach_target call.

> +               return btf_id;
> +       }
> +
> +       prog->attach_btf_id = btf_id;
> +       return 0;
> +}
> +
>  #define BPF_PROG_SEC_IMPL(string, ptype, eatype, is_attachable, btf, atype) \
>         { string, sizeof(string) - 1, ptype, eatype, is_attachable, btf, atype }
>
> @@ -6633,13 +6659,16 @@ static int libbpf_find_prog_btf_id(const char *name, __u32 attach_prog_fd)
>         return err;
>  }
>
> -static int libbpf_find_attach_btf_id(struct bpf_program *prog)
> +static int libbpf_find_attach_btf_id(struct bpf_program *prog,
> +                                    const char *name)
>  {
>         enum bpf_attach_type attach_type = prog->expected_attach_type;
>         __u32 attach_prog_fd = prog->attach_prog_fd;
> -       const char *name = prog->section_name;
>         int i, err;
>
> +       if (!name)
> +               name = prog->section_name;
> +

I second Toke, name should be just a function name, not including
"fentry/" (and others) part. If user want to programmatically
set/override attach type, we already have
bpf_program__set_expected_attach_type() API for that. So this
function's logic should do prefix/name extraction from
prog->section_name only if name is not explicitly specified.

>         if (!name)
>                 return -EINVAL;
>
> diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
> index 3fe12c9d1f92..02fc58a21a7f 100644
> --- a/tools/lib/bpf/libbpf.h
> +++ b/tools/lib/bpf/libbpf.h
> @@ -334,6 +334,10 @@ LIBBPF_API void
>  bpf_program__set_expected_attach_type(struct bpf_program *prog,
>                                       enum bpf_attach_type type);
>
> +LIBBPF_API int
> +bpf_program__set_attach_target(struct bpf_program *prog, int attach_prog_fd,
> +                              const char *attach_func_name);
> +
>  LIBBPF_API bool bpf_program__is_socket_filter(const struct bpf_program *prog);
>  LIBBPF_API bool bpf_program__is_tracepoint(const struct bpf_program *prog);
>  LIBBPF_API bool bpf_program__is_raw_tracepoint(const struct bpf_program *prog);
> diff --git a/tools/lib/bpf/libbpf.map b/tools/lib/bpf/libbpf.map
> index b035122142bb..8aba5438a3f0 100644
> --- a/tools/lib/bpf/libbpf.map
> +++ b/tools/lib/bpf/libbpf.map
> @@ -230,6 +230,7 @@ LIBBPF_0.0.7 {
>                 bpf_program__name;
>                 bpf_program__is_extension;
>                 bpf_program__is_struct_ops;
> +               bpf_program__set_attach_target;
>                 bpf_program__set_extension;
>                 bpf_program__set_struct_ops;
>                 btf__align_of;
>

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Andrii Nakryiko]

On Wed, Feb 12, 2020 at 5:05 AM Toke Høiland-Jørgensen <toke@redhat.com> wrote:
>
> Eelco Chaudron <echaudro@redhat.com> writes:
>
> > Currently when you want to attach a trace program to a bpf program
> > the section name needs to match the tracepoint/function semantics.
> >
> > However the addition of the bpf_program__set_attach_target() API
> > allows you to specify the tracepoint/function dynamically.
> >
> > The call flow would look something like this:
> >
> >   xdp_fd = bpf_prog_get_fd_by_id(id);
> >   trace_obj = bpf_object__open_file("func.o", NULL);
> >   prog = bpf_object__find_program_by_title(trace_obj,
> >                                            "fentry/myfunc");
> >   bpf_program__set_attach_target(prog, xdp_fd,
> >                                  "fentry/xdpfilt_blk_all");
>
> I think it would be better to have the attach type as a separate arg
> instead of encoding it in the function name. I.e., rather:
>
>    bpf_program__set_attach_target(prog, xdp_fd,
>                                   "xdpfilt_blk_all", BPF_TRACE_FENTRY);

I agree about not specifying section name prefix (e.g., fentry/). But
disagree that expected attach type (BPF_TRACE_FENTRY) should be part
of this API. We already have bpf_program__set_expected_attach_type()
API, no need to duplicate it here.

>
> -Toke
>

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Song Liu]

> On Feb 12, 2020, at 9:34 AM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> 
> On Wed, Feb 12, 2020 at 4:32 AM Eelco Chaudron <echaudro@redhat.com> wrote:
>> 
>> Currently when you want to attach a trace program to a bpf program
>> the section name needs to match the tracepoint/function semantics.
>> 
>> However the addition of the bpf_program__set_attach_target() API
>> allows you to specify the tracepoint/function dynamically.
>> 
>> The call flow would look something like this:
>> 
>>  xdp_fd = bpf_prog_get_fd_by_id(id);
>>  trace_obj = bpf_object__open_file("func.o", NULL);
>>  prog = bpf_object__find_program_by_title(trace_obj,
>>                                           "fentry/myfunc");
>>  bpf_program__set_attach_target(prog, xdp_fd,
>>                                 "fentry/xdpfilt_blk_all");
>>  bpf_object__load(trace_obj)
>> 
>> Signed-off-by: Eelco Chaudron <echaudro@redhat.com>


I am trying to solve the same problem with slightly different approach. 

It works as the following (with skeleton):

	obj = myobject_bpf__open_opts(&opts);
	bpf_object__for_each_program(prog, obj->obj)
		bpf_program__overwrite_section_name(prog, new_names[id++]);
	err = myobject_bpf__load(obj);

I don't have very strong preference. But I think my approach is simpler? 

Thanks,
Song


diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 514b1a524abb..4c29a7181d57 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -238,6 +238,8 @@ struct bpf_program {
        __u32 line_info_rec_size;
        __u32 line_info_cnt;
        __u32 prog_flags;
+
+       char *overwritten_section_name;
 };

 struct bpf_struct_ops {
@@ -442,6 +444,7 @@ static void bpf_program__exit(struct bpf_program *prog)
        zfree(&prog->pin_name);
        zfree(&prog->insns);
        zfree(&prog->reloc_desc);
+       zfree(&prog->overwritten_section_name);

        prog->nr_reloc = 0;
        prog->insns_cnt = 0;
@@ -6637,7 +6640,7 @@ static int libbpf_find_attach_btf_id(struct bpf_program *prog)
 {
        enum bpf_attach_type attach_type = prog->expected_attach_type;
        __u32 attach_prog_fd = prog->attach_prog_fd;
-       const char *name = prog->section_name;
+       const char *name = prog->overwritten_section_name ? : prog->section_name;
        int i, err;

        if (!name)
@@ -8396,3 +8399,11 @@ void bpf_object__destroy_skeleton(struct bpf_object_skeleton *s)
        free(s->progs);
        free(s);
 }
+
+char *bpf_program__overwrite_section_name(struct bpf_program *prog,
+                                         const char *sec_name)
+{
+       prog->overwritten_section_name = strdup(sec_name);
+
+       return prog->overwritten_section_name;
+}
diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
index 3fe12c9d1f92..02f0d8b57cc4 100644
--- a/tools/lib/bpf/libbpf.h
+++ b/tools/lib/bpf/libbpf.h
@@ -595,6 +595,10 @@ bpf_program__bpil_addr_to_offs(struct bpf_prog_info_linear *info_linear);
 LIBBPF_API void
 bpf_program__bpil_offs_to_addr(struct bpf_prog_info_linear *info_linear);

+LIBBPF_API char *
+bpf_program__overwrite_section_name(struct bpf_program *prog,
+                                   const char *sec_name);
+
 /*
  * A helper function to get the number of possible CPUs before looking up
  * per-CPU maps. Negative errno is returned on failure.
diff --git a/tools/lib/bpf/libbpf.map b/tools/lib/bpf/libbpf.map
index b035122142bb..ed26c20729db 100644
--- a/tools/lib/bpf/libbpf.map
+++ b/tools/lib/bpf/libbpf.map
@@ -235,3 +235,8 @@ LIBBPF_0.0.7 {
                btf__align_of;
                libbpf_find_kernel_btf;
 } LIBBPF_0.0.6;
+
+LIBBPF_0.0.8 {
+       global:
+               bpf_program__overwrite_section_name;
+} LIBBPF_0.0.7;

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Andrii Nakryiko]

On Wed, Feb 12, 2020 at 10:07 AM Song Liu <songliubraving@fb.com> wrote:
>
>
>
> > On Feb 12, 2020, at 9:34 AM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> >
> > On Wed, Feb 12, 2020 at 4:32 AM Eelco Chaudron <echaudro@redhat.com> wrote:
> >>
> >> Currently when you want to attach a trace program to a bpf program
> >> the section name needs to match the tracepoint/function semantics.
> >>
> >> However the addition of the bpf_program__set_attach_target() API
> >> allows you to specify the tracepoint/function dynamically.
> >>
> >> The call flow would look something like this:
> >>
> >>  xdp_fd = bpf_prog_get_fd_by_id(id);
> >>  trace_obj = bpf_object__open_file("func.o", NULL);
> >>  prog = bpf_object__find_program_by_title(trace_obj,
> >>                                           "fentry/myfunc");
> >>  bpf_program__set_attach_target(prog, xdp_fd,
> >>                                 "fentry/xdpfilt_blk_all");
> >>  bpf_object__load(trace_obj)
> >>
> >> Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
>
>
> I am trying to solve the same problem with slightly different approach.
>
> It works as the following (with skeleton):
>
>         obj = myobject_bpf__open_opts(&opts);
>         bpf_object__for_each_program(prog, obj->obj)
>                 bpf_program__overwrite_section_name(prog, new_names[id++]);
>         err = myobject_bpf__load(obj);
>
> I don't have very strong preference. But I think my approach is simpler?

I prefer bpf_program__set_attach_target() approach. Section name is a
program identifier and a *hint* for libbpf to determine program type,
attach type, and whatever else makes sense. But there still should be
an API to set all that manually at runtime, thus
bpf_program__set_attach_target(). Doing same by overriding section
name feels like a hack, plus it doesn't handle overriding
attach_program_fd at all.

>
> Thanks,
> Song
>
>
> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index 514b1a524abb..4c29a7181d57 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c
> @@ -238,6 +238,8 @@ struct bpf_program {
>         __u32 line_info_rec_size;
>         __u32 line_info_cnt;
>         __u32 prog_flags;
> +
> +       char *overwritten_section_name;
>  };
>
>  struct bpf_struct_ops {
> @@ -442,6 +444,7 @@ static void bpf_program__exit(struct bpf_program *prog)
>         zfree(&prog->pin_name);
>         zfree(&prog->insns);
>         zfree(&prog->reloc_desc);
> +       zfree(&prog->overwritten_section_name);
>
>         prog->nr_reloc = 0;
>         prog->insns_cnt = 0;
> @@ -6637,7 +6640,7 @@ static int libbpf_find_attach_btf_id(struct bpf_program *prog)
>  {
>         enum bpf_attach_type attach_type = prog->expected_attach_type;
>         __u32 attach_prog_fd = prog->attach_prog_fd;
> -       const char *name = prog->section_name;
> +       const char *name = prog->overwritten_section_name ? : prog->section_name;
>         int i, err;
>
>         if (!name)
> @@ -8396,3 +8399,11 @@ void bpf_object__destroy_skeleton(struct bpf_object_skeleton *s)
>         free(s->progs);
>         free(s);
>  }
> +
> +char *bpf_program__overwrite_section_name(struct bpf_program *prog,
> +                                         const char *sec_name)
> +{
> +       prog->overwritten_section_name = strdup(sec_name);
> +
> +       return prog->overwritten_section_name;
> +}
> diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
> index 3fe12c9d1f92..02f0d8b57cc4 100644
> --- a/tools/lib/bpf/libbpf.h
> +++ b/tools/lib/bpf/libbpf.h
> @@ -595,6 +595,10 @@ bpf_program__bpil_addr_to_offs(struct bpf_prog_info_linear *info_linear);
>  LIBBPF_API void
>  bpf_program__bpil_offs_to_addr(struct bpf_prog_info_linear *info_linear);
>
> +LIBBPF_API char *
> +bpf_program__overwrite_section_name(struct bpf_program *prog,
> +                                   const char *sec_name);
> +
>  /*
>   * A helper function to get the number of possible CPUs before looking up
>   * per-CPU maps. Negative errno is returned on failure.
> diff --git a/tools/lib/bpf/libbpf.map b/tools/lib/bpf/libbpf.map
> index b035122142bb..ed26c20729db 100644
> --- a/tools/lib/bpf/libbpf.map
> +++ b/tools/lib/bpf/libbpf.map
> @@ -235,3 +235,8 @@ LIBBPF_0.0.7 {
>                 btf__align_of;
>                 libbpf_find_kernel_btf;
>  } LIBBPF_0.0.6;
> +
> +LIBBPF_0.0.8 {
> +       global:
> +               bpf_program__overwrite_section_name;
> +} LIBBPF_0.0.7;

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Song Liu]

> On Feb 12, 2020, at 10:14 AM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> 
> On Wed, Feb 12, 2020 at 10:07 AM Song Liu <songliubraving@fb.com> wrote:
>> 
>> 
>> 
>>> On Feb 12, 2020, at 9:34 AM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
>>> 
>>> On Wed, Feb 12, 2020 at 4:32 AM Eelco Chaudron <echaudro@redhat.com> wrote:
>>>> 
>>>> Currently when you want to attach a trace program to a bpf program
>>>> the section name needs to match the tracepoint/function semantics.
>>>> 
>>>> However the addition of the bpf_program__set_attach_target() API
>>>> allows you to specify the tracepoint/function dynamically.
>>>> 
>>>> The call flow would look something like this:
>>>> 
>>>> xdp_fd = bpf_prog_get_fd_by_id(id);
>>>> trace_obj = bpf_object__open_file("func.o", NULL);
>>>> prog = bpf_object__find_program_by_title(trace_obj,
>>>>                                          "fentry/myfunc");
>>>> bpf_program__set_attach_target(prog, xdp_fd,
>>>>                                "fentry/xdpfilt_blk_all");
>>>> bpf_object__load(trace_obj)
>>>> 
>>>> Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
>> 
>> 
>> I am trying to solve the same problem with slightly different approach.
>> 
>> It works as the following (with skeleton):
>> 
>>        obj = myobject_bpf__open_opts(&opts);
>>        bpf_object__for_each_program(prog, obj->obj)
>>                bpf_program__overwrite_section_name(prog, new_names[id++]);
>>        err = myobject_bpf__load(obj);
>> 
>> I don't have very strong preference. But I think my approach is simpler?
> 
> I prefer bpf_program__set_attach_target() approach. Section name is a
> program identifier and a *hint* for libbpf to determine program type,
> attach type, and whatever else makes sense. But there still should be
> an API to set all that manually at runtime, thus
> bpf_program__set_attach_target(). Doing same by overriding section
> name feels like a hack, plus it doesn't handle overriding
> attach_program_fd at all.

We already have bpf_object_open_opts to handle different attach_program_fd. 
Can we depreciate bpf_object_open_opts.attach_prog_fd with the 
bpf_program__set_attach_target() approach?

Thanks,
Song

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Andrii Nakryiko]

On Wed, Feb 12, 2020 at 10:29 AM Song Liu <songliubraving@fb.com> wrote:
>
>
>
> > On Feb 12, 2020, at 10:14 AM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> >
> > On Wed, Feb 12, 2020 at 10:07 AM Song Liu <songliubraving@fb.com> wrote:
> >>
> >>
> >>
> >>> On Feb 12, 2020, at 9:34 AM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> >>>
> >>> On Wed, Feb 12, 2020 at 4:32 AM Eelco Chaudron <echaudro@redhat.com> wrote:
> >>>>
> >>>> Currently when you want to attach a trace program to a bpf program
> >>>> the section name needs to match the tracepoint/function semantics.
> >>>>
> >>>> However the addition of the bpf_program__set_attach_target() API
> >>>> allows you to specify the tracepoint/function dynamically.
> >>>>
> >>>> The call flow would look something like this:
> >>>>
> >>>> xdp_fd = bpf_prog_get_fd_by_id(id);
> >>>> trace_obj = bpf_object__open_file("func.o", NULL);
> >>>> prog = bpf_object__find_program_by_title(trace_obj,
> >>>>                                          "fentry/myfunc");
> >>>> bpf_program__set_attach_target(prog, xdp_fd,
> >>>>                                "fentry/xdpfilt_blk_all");
> >>>> bpf_object__load(trace_obj)
> >>>>
> >>>> Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
> >>
> >>
> >> I am trying to solve the same problem with slightly different approach.
> >>
> >> It works as the following (with skeleton):
> >>
> >>        obj = myobject_bpf__open_opts(&opts);
> >>        bpf_object__for_each_program(prog, obj->obj)
> >>                bpf_program__overwrite_section_name(prog, new_names[id++]);
> >>        err = myobject_bpf__load(obj);
> >>
> >> I don't have very strong preference. But I think my approach is simpler?
> >
> > I prefer bpf_program__set_attach_target() approach. Section name is a
> > program identifier and a *hint* for libbpf to determine program type,
> > attach type, and whatever else makes sense. But there still should be
> > an API to set all that manually at runtime, thus
> > bpf_program__set_attach_target(). Doing same by overriding section
> > name feels like a hack, plus it doesn't handle overriding
> > attach_program_fd at all.
>
> We already have bpf_object_open_opts to handle different attach_program_fd.

Not really, because open_opts apply to bpf_object and all its
bpf_programs, not to individual bpf_program. So it works only if BPF
application has only one BPF program. If you have many, you can only
set the same attach_program_fd for all of them. Basically, open_opts'
attach_prog_fd should be treated as a default or fallback
attach_prog_fd.

> Can we depreciate bpf_object_open_opts.attach_prog_fd with the
> bpf_program__set_attach_target() approach?

bpf_program__set_attach_target() overrides attach_prog_fd, yes. But we
can't just deprecate that option because it's part of an API already,
even though adding it to open opts was probably a mistake. But for
simple BPF apps with single BPF program it does work fine, so...

>
> Thanks,
> Song
>

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Song Liu]

> On Feb 12, 2020, at 10:34 AM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> 
> On Wed, Feb 12, 2020 at 10:29 AM Song Liu <songliubraving@fb.com> wrote:
>> 
>> 
>> 
>>> On Feb 12, 2020, at 10:14 AM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
>>> 
>>> On Wed, Feb 12, 2020 at 10:07 AM Song Liu <songliubraving@fb.com> wrote:
>>>> 
>>>> 
>>>> 
>>>>> On Feb 12, 2020, at 9:34 AM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
>>>>> 
>>>>> On Wed, Feb 12, 2020 at 4:32 AM Eelco Chaudron <echaudro@redhat.com> wrote:
>>>>>> 
>>>>>> Currently when you want to attach a trace program to a bpf program
>>>>>> the section name needs to match the tracepoint/function semantics.
>>>>>> 
>>>>>> However the addition of the bpf_program__set_attach_target() API
>>>>>> allows you to specify the tracepoint/function dynamically.
>>>>>> 
>>>>>> The call flow would look something like this:
>>>>>> 
>>>>>> xdp_fd = bpf_prog_get_fd_by_id(id);
>>>>>> trace_obj = bpf_object__open_file("func.o", NULL);
>>>>>> prog = bpf_object__find_program_by_title(trace_obj,
>>>>>>                                         "fentry/myfunc");
>>>>>> bpf_program__set_attach_target(prog, xdp_fd,
>>>>>>                               "fentry/xdpfilt_blk_all");
>>>>>> bpf_object__load(trace_obj)
>>>>>> 
>>>>>> Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
>>>> 
>>>> 
>>>> I am trying to solve the same problem with slightly different approach.
>>>> 
>>>> It works as the following (with skeleton):
>>>> 
>>>>       obj = myobject_bpf__open_opts(&opts);
>>>>       bpf_object__for_each_program(prog, obj->obj)
>>>>               bpf_program__overwrite_section_name(prog, new_names[id++]);
>>>>       err = myobject_bpf__load(obj);
>>>> 
>>>> I don't have very strong preference. But I think my approach is simpler?
>>> 
>>> I prefer bpf_program__set_attach_target() approach. Section name is a
>>> program identifier and a *hint* for libbpf to determine program type,
>>> attach type, and whatever else makes sense. But there still should be
>>> an API to set all that manually at runtime, thus
>>> bpf_program__set_attach_target(). Doing same by overriding section
>>> name feels like a hack, plus it doesn't handle overriding
>>> attach_program_fd at all.
>> 
>> We already have bpf_object_open_opts to handle different attach_program_fd.
> 
> Not really, because open_opts apply to bpf_object and all its
> bpf_programs, not to individual bpf_program. So it works only if BPF
> application has only one BPF program. If you have many, you can only
> set the same attach_program_fd for all of them. Basically, open_opts'
> attach_prog_fd should be treated as a default or fallback
> attach_prog_fd.

Fair enough. I will use set_attach_target in my code. 

> 
>> Can we depreciate bpf_object_open_opts.attach_prog_fd with the
>> bpf_program__set_attach_target() approach?
> 
> bpf_program__set_attach_target() overrides attach_prog_fd, yes. But we
> can't just deprecate that option because it's part of an API already,
> even though adding it to open opts was probably a mistake. But for
> simple BPF apps with single BPF program it does work fine, so...

Maybe add a warning saying "attach_prog_fd is deprecated, xxx"?

Thanks,
Song

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Toke Høiland-Jørgensen]

Andrii Nakryiko <andrii.nakryiko@gmail.com> writes:

> On Wed, Feb 12, 2020 at 5:05 AM Toke Høiland-Jørgensen <toke@redhat.com> wrote:
>>
>> Eelco Chaudron <echaudro@redhat.com> writes:
>>
>> > Currently when you want to attach a trace program to a bpf program
>> > the section name needs to match the tracepoint/function semantics.
>> >
>> > However the addition of the bpf_program__set_attach_target() API
>> > allows you to specify the tracepoint/function dynamically.
>> >
>> > The call flow would look something like this:
>> >
>> >   xdp_fd = bpf_prog_get_fd_by_id(id);
>> >   trace_obj = bpf_object__open_file("func.o", NULL);
>> >   prog = bpf_object__find_program_by_title(trace_obj,
>> >                                            "fentry/myfunc");
>> >   bpf_program__set_attach_target(prog, xdp_fd,
>> >                                  "fentry/xdpfilt_blk_all");
>>
>> I think it would be better to have the attach type as a separate arg
>> instead of encoding it in the function name. I.e., rather:
>>
>>    bpf_program__set_attach_target(prog, xdp_fd,
>>                                   "xdpfilt_blk_all", BPF_TRACE_FENTRY);
>
> I agree about not specifying section name prefix (e.g., fentry/). But
> disagree that expected attach type (BPF_TRACE_FENTRY) should be part
> of this API. We already have bpf_program__set_expected_attach_type()
> API, no need to duplicate it here.

Ah yes, forgot about that; just keeping that and making this function
name only is fine with me :)

-Toke

Re: [PATCH bpf-next] libbpf: Add support for dynamic program attach target

[Eelco Chaudron]

On 12 Feb 2020, at 22:52, Toke Høiland-Jørgensen wrote:

> Andrii Nakryiko <andrii.nakryiko@gmail.com> writes:
>
>> On Wed, Feb 12, 2020 at 5:05 AM Toke Høiland-Jørgensen 
>> <toke@redhat.com> wrote:
>>>
>>> Eelco Chaudron <echaudro@redhat.com> writes:
>>>
>>>> Currently when you want to attach a trace program to a bpf program
>>>> the section name needs to match the tracepoint/function semantics.
>>>>
>>>> However the addition of the bpf_program__set_attach_target() API
>>>> allows you to specify the tracepoint/function dynamically.
>>>>
>>>> The call flow would look something like this:
>>>>
>>>>   xdp_fd = bpf_prog_get_fd_by_id(id);
>>>>   trace_obj = bpf_object__open_file("func.o", NULL);
>>>>   prog = bpf_object__find_program_by_title(trace_obj,
>>>>                                            "fentry/myfunc");
>>>>   bpf_program__set_attach_target(prog, xdp_fd,
>>>>                                  "fentry/xdpfilt_blk_all");
>>>
>>> I think it would be better to have the attach type as a separate arg
>>> instead of encoding it in the function name. I.e., rather:
>>>
>>>    bpf_program__set_attach_target(prog, xdp_fd,
>>>                                   "xdpfilt_blk_all", 
>>> BPF_TRACE_FENTRY);
>>
>> I agree about not specifying section name prefix (e.g., fentry/). But
>> disagree that expected attach type (BPF_TRACE_FENTRY) should be part
>> of this API. We already have bpf_program__set_expected_attach_type()
>> API, no need to duplicate it here.
>
> Ah yes, forgot about that; just keeping that and making this function
> name only is fine with me :)

Toke/Andrii,

Thanks for the feedback, will send out a v2 soon.

//Eelco