Re: [PATCH bpf-next 4/9] samples: bpf: refactor task_fd_query program with libbpf

["Daniel T. Lee" <danieltimlee@gmail.com>]

On Wed, Nov 18, 2020 at 11:58 AM Andrii Nakryiko
<andrii.nakryiko@gmail.com> wrote:
>
> On Tue, Nov 17, 2020 at 6:57 AM Daniel T. Lee <danieltimlee@gmail.com> wrote:
> >
> > This commit refactors the existing kprobe program with libbpf bpf
> > loader. To attach bpf program, this uses generic bpf_program__attach()
> > approach rather than using bpf_load's load_bpf_file().
> >
> > To attach bpf to perf_event, instead of using previous ioctl method,
> > this commit uses bpf_program__attach_perf_event since it manages the
> > enable of perf_event and attach of BPF programs to it, which is much
> > more intuitive way to achieve.
> >
> > Also, explicit close(fd) has been removed since event will be closed
> > inside bpf_link__destroy() automatically.
> >
> > DEBUGFS macro from trace_helpers has been used to control uprobe events.
> > Furthermore, to prevent conflict of same named uprobe events, O_TRUNC
> > flag has been used to clear 'uprobe_events' interface.
> >
> > Signed-off-by: Daniel T. Lee <danieltimlee@gmail.com>
> > ---
> >  samples/bpf/Makefile             |   2 +-
> >  samples/bpf/task_fd_query_user.c | 101 ++++++++++++++++++++++---------
> >  2 files changed, 74 insertions(+), 29 deletions(-)
> >
>
> [...]
>
> >  static int test_debug_fs_uprobe(char *binary_path, long offset, bool is_return)
> >  {
> > +       char buf[256], event_alias[sizeof("test_1234567890")];
> >         const char *event_type = "uprobe";
> >         struct perf_event_attr attr = {};
> > -       char buf[256], event_alias[sizeof("test_1234567890")];
> >         __u64 probe_offset, probe_addr;
> >         __u32 len, prog_id, fd_type;
> > -       int err, res, kfd, efd;
> > +       int err = -1, res, kfd, efd;
> > +       struct bpf_link *link;
> >         ssize_t bytes;
> >
> > -       snprintf(buf, sizeof(buf), "/sys/kernel/debug/tracing/%s_events",
> > -                event_type);
> > -       kfd = open(buf, O_WRONLY | O_APPEND, 0);
> > +       snprintf(buf, sizeof(buf), DEBUGFS "%s_events", event_type);
> > +       kfd = open(buf, O_WRONLY | O_TRUNC, 0);
>
> O_TRUNC will also remove other events, created by users. Not a great
> experience. Let's leave the old behavior?
>

The reason why I used O_TRUNC is, it gets conflict error during tests.
I'm not sure if it is a bug of ftrace uprobes_events or not, but seems adding
same name of uprobe_events with another type seems not working.
(adding uretprobes after uprobes returns an error)

    samples/bpf # echo 'p:uprobes/test_500836 ./task_fd_query:0x3d80'
>> /sys/kernel/debug/tracing/uprobe_events
    samples/bpf # cat /sys/kernel/debug/tracing/uprobe_events
     p:uprobes/test_500836 ./task_fd_query:0x0000000000003d80
    samples/bpf# echo 'r:uprobes/test_500836 ./task_fd_query:0x3d80'
>> /sys/kernel/debug/tracing/uprobe_events
     bash: echo: write error: File exists

Since this gets error, I've just truncated on every open of this interface.

> >         CHECK_PERROR_RET(kfd < 0);
> >
> >         res = snprintf(event_alias, sizeof(event_alias), "test_%d", getpid());
> > @@ -240,8 +252,8 @@ static int test_debug_fs_uprobe(char *binary_path, long offset, bool is_return)
> >         close(kfd);
> >         kfd = -1;
> >
> > -       snprintf(buf, sizeof(buf), "/sys/kernel/debug/tracing/events/%ss/%s/id",
> > -                event_type, event_alias);
> > +       snprintf(buf, sizeof(buf), DEBUGFS "events/%ss/%s/id", event_type,
>
> I'd leave the string verbatim here (and above), I think it's better
> that way and easier to figure out what's written where. And then no
> need to expose DEBUGFS.
>

Sounds great. I'll keep the string path as it was.

> > +                event_alias);
> >         efd = open(buf, O_RDONLY, 0);
> >         CHECK_PERROR_RET(efd < 0);
> >
>
> [...]



-- 
Best,
Daniel T. Lee