Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> writes:
> On Wed, Sep 16, 2020 at 2:21 PM Toke Høiland-Jørgensen <toke@xxxxxxxxxx> wrote:
>>
>> Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> writes:
>>
>> > On Wed, Sep 16, 2020 at 1:37 PM Andrii Nakryiko
>> > <andrii.nakryiko@xxxxxxxxx> wrote:
>> >>
>> >> On Tue, Sep 15, 2020 at 5:50 PM Toke Høiland-Jørgensen <toke@xxxxxxxxxx> wrote:
>> >> >
>> >> > From: Toke Høiland-Jørgensen <toke@xxxxxxxxxx>
>> >> >
>> >> > This adds support for supplying a target btf ID for the bpf_link_create()
>> >> > operation, and adds a new bpf_program__attach_freplace() high-level API for
>> >> > attaching freplace functions with a target.
>> >> >
>> >> > Signed-off-by: Toke Høiland-Jørgensen <toke@xxxxxxxxxx>
>> >> > ---
>> >> > tools/lib/bpf/bpf.c | 1 +
>> >> > tools/lib/bpf/bpf.h | 3 ++-
>> >> > tools/lib/bpf/libbpf.c | 24 ++++++++++++++++++------
>> >> > tools/lib/bpf/libbpf.h | 3 +++
>> >> > tools/lib/bpf/libbpf.map | 1 +
>> >> > 5 files changed, 25 insertions(+), 7 deletions(-)
>> >> >
>> >> > diff --git a/tools/lib/bpf/bpf.c b/tools/lib/bpf/bpf.c
>> >> > index 82b983ff6569..e928456c0dd6 100644
>> >> > --- a/tools/lib/bpf/bpf.c
>> >> > +++ b/tools/lib/bpf/bpf.c
>> >> > @@ -599,6 +599,7 @@ int bpf_link_create(int prog_fd, int target_fd,
>> >> > attr.link_create.iter_info =
>> >> > ptr_to_u64(OPTS_GET(opts, iter_info, (void *)0));
>> >> > attr.link_create.iter_info_len = OPTS_GET(opts, iter_info_len, 0);
>> >> > + attr.link_create.target_btf_id = OPTS_GET(opts, target_btf_id, 0);
>> >> >
>> >> > return sys_bpf(BPF_LINK_CREATE, &attr, sizeof(attr));
>> >> > }
>> >> > diff --git a/tools/lib/bpf/bpf.h b/tools/lib/bpf/bpf.h
>> >> > index 015d13f25fcc..f8dbf666b62b 100644
>> >> > --- a/tools/lib/bpf/bpf.h
>> >> > +++ b/tools/lib/bpf/bpf.h
>> >> > @@ -174,8 +174,9 @@ struct bpf_link_create_opts {
>> >> > __u32 flags;
>> >> > union bpf_iter_link_info *iter_info;
>> >> > __u32 iter_info_len;
>> >> > + __u32 target_btf_id;
>> >> > };
>> >> > -#define bpf_link_create_opts__last_field iter_info_len
>> >> > +#define bpf_link_create_opts__last_field target_btf_id
>> >> >
>> >> > LIBBPF_API int bpf_link_create(int prog_fd, int target_fd,
>> >> > enum bpf_attach_type attach_type,
>> >> > diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
>> >> > index 550950eb1860..165131c73f40 100644
>> >> > --- a/tools/lib/bpf/libbpf.c
>> >> > +++ b/tools/lib/bpf/libbpf.c
>> >> > @@ -9322,12 +9322,14 @@ static struct bpf_link *attach_iter(const struct bpf_sec_def *sec,
>> >> >
>> >> > static struct bpf_link *
>> >> > bpf_program__attach_fd(struct bpf_program *prog, int target_fd,
>> >> > - const char *target_name)
>> >> > + int target_btf_id, const char *target_name)
>> >> > {
>> >> > enum bpf_attach_type attach_type;
>> >> > char errmsg[STRERR_BUFSIZE];
>> >> > struct bpf_link *link;
>> >> > int prog_fd, link_fd;
>> >> > + DECLARE_LIBBPF_OPTS(bpf_link_create_opts, opts,
>> >> > + .target_btf_id = target_btf_id);
>> >> >
>> >> > prog_fd = bpf_program__fd(prog);
>> >> > if (prog_fd < 0) {
>> >> > @@ -9340,8 +9342,12 @@ bpf_program__attach_fd(struct bpf_program *prog, int target_fd,
>> >> > return ERR_PTR(-ENOMEM);
>> >> > link->detach = &bpf_link__detach_fd;
>> >> >
>> >> > - attach_type = bpf_program__get_expected_attach_type(prog);
>> >> > - link_fd = bpf_link_create(prog_fd, target_fd, attach_type, NULL);
>> >> > + if (bpf_program__get_type(prog) == BPF_PROG_TYPE_EXT)
>> >> > + attach_type = BPF_TRACE_FREPLACE;
>> >>
>> >> doing this unconditionally will break an old-style freplace without
>> >> target_fd/btf_id on older kernels. Safe and simple way would be to
>> >> continue using raw_tracepoint_open when there is no target_fd/btf_id,
>> >> and use LINK_CREATE for newer stuff. Alternatively, you'd need to do
>> >> feature detection, but it's still would be nice to handle old-style
>> >> attach through raw_tracepoint_open for bpf_program__attach_freplace.
>> >>
>> >> so I suggest leaving bpf_program__attach_fd() as is and to create a
>> >> custom bpf_program__attach_freplace() implementation.
>>
>> Sure, I'll take another pass at this. Not sure how useful feature
>> detection in libbpf is; if the caller passes a target, libbpf can't
>> really do much if the kernel doesn't support it...
>
> I was thinking about bpf_program__attach_freplace(prog, 0, NULL) doing
> bpf_raw_tracepoint_open(). It would be nice to support this, for API
> uniformity, no?
Yeah, sure, that we can do :)
-Toke
