Re: [PATCH bpf-next v5 1/3] bpf: Show precise link_type for {uprobe,kprobe}_multi fdinfo

[Tao Chen <chen.dylane@linux.dev>]

在 2025/6/24 23:46, Andrii Nakryiko 写道:
> On Tue, Jun 24, 2025 at 1:41 AM Tao Chen <chen.dylane@linux.dev> wrote:
>>
>> 在 2025/6/24 16:16, Jiri Olsa 写道:
>>> On Mon, Jun 23, 2025 at 01:59:18PM -0700, Andrii Nakryiko wrote:
>>>> On Mon, Jun 23, 2025 at 10:56 AM Alexei Starovoitov
>>>> <alexei.starovoitov@gmail.com> wrote:
>>>>>
>>>>> On Mon, Jun 23, 2025 at 6:44 AM Tao Chen <chen.dylane@linux.dev> wrote:
>>>>>>
>>>>>> Alexei suggested, 'link_type' can be more precise and differentiate
>>>>>> for human in fdinfo. In fact BPF_LINK_TYPE_KPROBE_MULTI includes
>>>>>> kretprobe_multi type, the same as BPF_LINK_TYPE_UPROBE_MULTI, so we
>>>>>> can show it more concretely.
>>>>>>
>>>>>> link_type:      kprobe_multi
>>>>>> link_id:        1
>>>>>> prog_tag:       d2b307e915f0dd37
>>>>>> ...
>>>>>> link_type:      kretprobe_multi
>>>>>> link_id:        2
>>>>>> prog_tag:       ab9ea0545870781d
>>>>>> ...
>>>>>> link_type:      uprobe_multi
>>>>>> link_id:        9
>>>>>> prog_tag:       e729f789e34a8eca
>>>>>> ...
>>>>>> link_type:      uretprobe_multi
>>>>>> link_id:        10
>>>>>> prog_tag:       7db356c03e61a4d4
>>>>>>
>>>>>> Signed-off-by: Tao Chen <chen.dylane@linux.dev>
>>>>>> ---
>>>>>>    include/linux/trace_events.h | 10 ++++++++++
>>>>>>    kernel/bpf/syscall.c         |  9 ++++++++-
>>>>>>    kernel/trace/bpf_trace.c     | 28 ++++++++++++++++++++++++++++
>>>>>>    3 files changed, 46 insertions(+), 1 deletion(-)
>>>>>>
>>>>>> Change list:
>>>>>>     v4 -> v5:
>>>>>>       - Add patch1 to show precise link_type for
>>>>>>         {uprobe,kprobe}_multi.(Alexei)
>>>>>>       - patch2,3 just remove type field, which will be showed in
>>>>>>         link_type
>>>>>>     v4:
>>>>>>     https://lore.kernel.org/bpf/20250619034257.70520-1-chen.dylane@linux.dev
>>>>>>
>>>>>>     v3 -> v4:
>>>>>>       - use %pS to print func info.(Alexei)
>>>>>>     v3:
>>>>>>     https://lore.kernel.org/bpf/20250616130233.451439-1-chen.dylane@linux.dev
>>>>>>
>>>>>>     v2 -> v3:
>>>>>>       - show info in one line for multi events.(Jiri)
>>>>>>     v2:
>>>>>>     https://lore.kernel.org/bpf/20250615150514.418581-1-chen.dylane@linux.dev
>>>>>>
>>>>>>     v1 -> v2:
>>>>>>       - replace 'func_cnt' with 'uprobe_cnt'.(Andrii)
>>>>>>       - print func name is more readable and security for kprobe_multi.(Alexei)
>>>>>>     v1:
>>>>>>     https://lore.kernel.org/bpf/20250612115556.295103-1-chen.dylane@linux.dev
>>>>>>
>>>>>> diff --git a/include/linux/trace_events.h b/include/linux/trace_events.h
>>>>>> index fa9cf4292df..951c91babbc 100644
>>>>>> --- a/include/linux/trace_events.h
>>>>>> +++ b/include/linux/trace_events.h
>>>>>> @@ -780,6 +780,8 @@ int bpf_get_perf_event_info(const struct perf_event *event, u32 *prog_id,
>>>>>>                               unsigned long *missed);
>>>>>>    int bpf_kprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
>>>>>>    int bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
>>>>>> +void bpf_kprobe_multi_link_type_show(const struct bpf_link *link, char *link_type, int len);
>>>>>> +void bpf_uprobe_multi_link_type_show(const struct bpf_link *link, char *link_type, int len);
>>>>>>    #else
>>>>>>    static inline unsigned int trace_call_bpf(struct trace_event_call *call, void *ctx)
>>>>>>    {
>>>>>> @@ -832,6 +834,14 @@ bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *prog)
>>>>>>    {
>>>>>>           return -EOPNOTSUPP;
>>>>>>    }
>>>>>> +static inline void
>>>>>> +bpf_kprobe_multi_link_type_show(const struct bpf_link *link, char *link_type, int len)
>>>>>> +{
>>>>>> +}
>>>>>> +static inline void
>>>>>> +bpf_uprobe_multi_link_type_show(const struct bpf_link *link, char *link_type, int len)
>>>>>> +{
>>>>>> +}
>>>>>>    #endif
>>>>>>
>>>>>>    enum {
>>>>>> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>>>>>> index 51ba1a7aa43..43b821b37bc 100644
>>>>>> --- a/kernel/bpf/syscall.c
>>>>>> +++ b/kernel/bpf/syscall.c
>>>>>> @@ -3226,9 +3226,16 @@ static void bpf_link_show_fdinfo(struct seq_file *m, struct file *filp)
>>>>>>           const struct bpf_prog *prog = link->prog;
>>>>>>           enum bpf_link_type type = link->type;
>>>>>>           char prog_tag[sizeof(prog->tag) * 2 + 1] = { };
>>>>>> +       char link_type[64] = {};
>>>>>>
>>>>>>           if (type < ARRAY_SIZE(bpf_link_type_strs) && bpf_link_type_strs[type]) {
>>>>>> -               seq_printf(m, "link_type:\t%s\n", bpf_link_type_strs[type]);
>>>>>> +               if (link->type == BPF_LINK_TYPE_KPROBE_MULTI)
>>>>>> +                       bpf_kprobe_multi_link_type_show(link, link_type, sizeof(link_type));
>>>>>> +               else if (link->type == BPF_LINK_TYPE_UPROBE_MULTI)
>>>>>> +                       bpf_uprobe_multi_link_type_show(link, link_type, sizeof(link_type));
>>>>>> +               else
>>>>>> +                       strscpy(link_type, bpf_link_type_strs[type], sizeof(link_type));
>>>>>> +               seq_printf(m, "link_type:\t%s\n", link_type);
>>>>>
>>>>> New callbacks just to print a string?
>>>>> Let's find a different way.
>>>>>
>>>>> How about moving 'flags' from bpf_[ku]probe_multi_link into bpf_link ?
>>>>> (There is a 7 byte hole there anyway)
>>>>> and checking flags inline.
>>>>>
>>>>> Jiri, Andrii,
>>>>>
>>>>> better ideas?
>>>>
>>>> We can just remember original attr->link_create.attach_type in
>>>> bpf_link itself, and then have a small helper that will accept link
>>>> type and attach type, and fill out link type representation based on
>>>> those two. Internally we can do the special-casing of  uprobe vs
>>>> uretprobe and kprobe vs kretprobe transparently to all the other code.
>>>> And use that here in show_fdinfo
>>>
>>> but you'd still need the flags, no? to find out if it's return probe
>>>
>>> I tried what Alexei suggested and it seems ok and simple enough
>>>
>>> jirka
>>>
>>>
>>> ---
>>> diff --git a/include/linux/bpf.h b/include/linux/bpf.h
>>> index 5dd556e89cce..287c956cdbd2 100644
>>> --- a/include/linux/bpf.h
>>> +++ b/include/linux/bpf.h
>>> @@ -1702,6 +1702,7 @@ struct bpf_link {
>>>         * link's semantics is determined by target attach hook
>>>         */
>>>        bool sleepable;
>>> +     u32 flags;
>>>        /* rcu is used before freeing, work can be used to schedule that
>>>         * RCU-based freeing before that, so they never overlap
>>>         */
>>> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>>> index 56500381c28a..f1d9ee9717a1 100644
>>> --- a/kernel/bpf/syscall.c
>>> +++ b/kernel/bpf/syscall.c
>>> @@ -3228,7 +3228,14 @@ static void bpf_link_show_fdinfo(struct seq_file *m, struct file *filp)
>>>        char prog_tag[sizeof(prog->tag) * 2 + 1] = { };
>>>
>>>        if (type < ARRAY_SIZE(bpf_link_type_strs) && bpf_link_type_strs[type]) {
>>> -             seq_printf(m, "link_type:\t%s\n", bpf_link_type_strs[type]);
>>> +             if (link->type == BPF_LINK_TYPE_KPROBE_MULTI)
>>> +                     seq_printf(m, "link_type:\t%s\n", link->flags == BPF_F_KPROBE_MULTI_RETURN ?
>>> +                                "kretprobe_multi" : "kprobe_multi");
>>> +             else if (link->type == BPF_LINK_TYPE_UPROBE_MULTI)
>>> +                     seq_printf(m, "link_type:\t%s\n", link->flags == BPF_F_UPROBE_MULTI_RETURN ?
>>> +                                "uretprobe_multi" : "uprobe_multi");
>>> +             else
>>> +                     seq_printf(m, "link_type:\t%s\n", bpf_link_type_strs[type]);
>>>        } else {
>>>                WARN_ONCE(1, "missing BPF_LINK_TYPE(...) for link type %u\n", type);
>>>                seq_printf(m, "link_type:\t<%u>\n", type);
>>> diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
>>> index 0a06ea6638fe..81d7a4e5ae15 100644
>>> --- a/kernel/trace/bpf_trace.c
>>> +++ b/kernel/trace/bpf_trace.c
>>> @@ -2466,7 +2466,6 @@ struct bpf_kprobe_multi_link {
>>>        u32 cnt;
>>>        u32 mods_cnt;
>>>        struct module **mods;
>>> -     u32 flags;
>>>    };
>>>
>>>    struct bpf_kprobe_multi_run_ctx {
>>> @@ -2586,7 +2585,7 @@ static int bpf_kprobe_multi_link_fill_link_info(const struct bpf_link *link,
>>>
>>>        kmulti_link = container_of(link, struct bpf_kprobe_multi_link, link);
>>>        info->kprobe_multi.count = kmulti_link->cnt;
>>> -     info->kprobe_multi.flags = kmulti_link->flags;
>>> +     info->kprobe_multi.flags = kmulti_link->link.flags;
>>>        info->kprobe_multi.missed = kmulti_link->fp.nmissed;
>>>
>>>        if (!uaddrs)
>>> @@ -2976,7 +2975,7 @@ int bpf_kprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
>>>        link->addrs = addrs;
>>>        link->cookies = cookies;
>>>        link->cnt = cnt;
>>> -     link->flags = flags;
>>> +     link->link.flags = flags;
>>>
>>>        if (cookies) {
>>>                /*
>>> @@ -3045,7 +3044,6 @@ struct bpf_uprobe_multi_link {
>>>        struct path path;
>>>        struct bpf_link link;
>>>        u32 cnt;
>>> -     u32 flags;
>>>        struct bpf_uprobe *uprobes;
>>>        struct task_struct *task;
>>>    };
>>> @@ -3109,7 +3107,7 @@ static int bpf_uprobe_multi_link_fill_link_info(const struct bpf_link *link,
>>>
>>>        umulti_link = container_of(link, struct bpf_uprobe_multi_link, link);
>>>        info->uprobe_multi.count = umulti_link->cnt;
>>> -     info->uprobe_multi.flags = umulti_link->flags;
>>> +     info->uprobe_multi.flags = umulti_link->link.flags;
>>>        info->uprobe_multi.pid = umulti_link->task ?
>>>                                 task_pid_nr_ns(umulti_link->task, task_active_pid_ns(current)) : 0;
>>>
>>> @@ -3369,7 +3367,7 @@ int bpf_uprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
>>>        link->uprobes = uprobes;
>>>        link->path = path;
>>>        link->task = task;
>>> -     link->flags = flags;
>>> +     link->link.flags = flags;
>>>
>>>        bpf_link_init(&link->link, BPF_LINK_TYPE_UPROBE_MULTI,
>>>                      &bpf_uprobe_multi_link_lops, prog);
>>
>> Hi, Jiri, Andrii,
>>
>> Jiri's patch looks more simple, and i see other struct xx_links wrap
>> bpf_link, which have attach_type field like:
>> struct sockmap_link {
>>           struct bpf_link link;
>>           struct bpf_map *map;
>>           enum bpf_attach_type attach_type;
>> };
>> If we create attach_type filed in bpf_link, maybe these struct xx_link
>> should also be modified. BTW, as Jiri said, we still can not find return
>> probe type from attach_type.
> 
> You are right, I somehow was under impression that ret vs non-retprobe
> comes from attach type as well.
> 
> Ok, moving flags into common bpf_link struct sounds good to me. I'd
> still move attach_type into bpf_link, together with flags, for
> generality (and update all those links that already include
> attach_type as you mentioned). We can make it a single-byte field to
> not increase bpf_link size unnecessarily (by using bitfield size).
> 

Well，can we complete this in two steps?

1. Create a common field in bpf_link used for flags or attach_type, and 
realise the precise link_type feature as Jiri and Alexei said, the 
review of this part has been revised almost completely.

2. Move the attach_type from struct bpf_xx_link into bpf_link, this will 
involve a lot of changes, i will send a separate patchset to finish it.

>>
>> --
>> Best Regards
>> Tao Chen


-- 
Best Regards
Tao Chen