Re: [RFC PATCH bpf-next 1/2] libbpf: Add libbpf_probe_bpf_kfunc API

[Tao Chen <chen.dylane@gmail.com>]

在 2025/1/23 06:22, Andrii Nakryiko 写道:
> On Wed, Jan 22, 2025 at 9:14 AM Tao Chen <chen.dylane@gmail.com> wrote:
>>
>> Similarly to libbpf_probe_bpf_helper, the libbpf_probe_bpf_kfunc
>> used to test the availability of the different eBPF kfuncs on the
>> current system.
>>
>> Signed-off-by: Tao Chen <chen.dylane@gmail.com>
>> ---
>>   tools/lib/bpf/libbpf.h        | 16 +++++++++++++++-
>>   tools/lib/bpf/libbpf.map      |  1 +
>>   tools/lib/bpf/libbpf_probes.c | 36 +++++++++++++++++++++++++++++++++++
>>   3 files changed, 52 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
>> index 3020ee45303a..3b6d33578a16 100644
>> --- a/tools/lib/bpf/libbpf.h
>> +++ b/tools/lib/bpf/libbpf.h
>> @@ -1680,7 +1680,21 @@ LIBBPF_API int libbpf_probe_bpf_map_type(enum bpf_map_type map_type, const void
>>    */
>>   LIBBPF_API int libbpf_probe_bpf_helper(enum bpf_prog_type prog_type,
>>                                         enum bpf_func_id helper_id, const void *opts);
>> -
>> +/**
>> + * @brief **libbpf_probe_bpf_kfunc()** detects if host kernel supports the
>> + * use of a given BPF kfunc from specified BPF program type.
>> + * @param prog_type BPF program type used to check the support of BPF kfunc
>> + * @param kfunc_id The btf ID of BPF kfunc to check support for
>> + * @param opts reserved for future extensibility, should be NULL
>> + * @return 1, if given combination of program type and kfunc is supported; 0,
>> + * if the combination is not supported; negative error code if feature
>> + * detection for provided input arguments failed or can't be performed
>> + *
>> + * Make sure the process has required set of CAP_* permissions (or runs as
>> + * root) when performing feature checking.
>> + */
>> +LIBBPF_API int libbpf_probe_bpf_kfunc(enum bpf_prog_type prog_type,
>> +                                     int kfunc_id, const void *opts);
>>   /**
>>    * @brief **libbpf_num_possible_cpus()** is a helper function to get the
>>    * number of possible CPUs that the host kernel supports and expects.
>> diff --git a/tools/lib/bpf/libbpf.map b/tools/lib/bpf/libbpf.map
>> index a8b2936a1646..e93fae101efd 100644
>> --- a/tools/lib/bpf/libbpf.map
>> +++ b/tools/lib/bpf/libbpf.map
>> @@ -436,4 +436,5 @@ LIBBPF_1.6.0 {
>>                  bpf_linker__add_buf;
>>                  bpf_linker__add_fd;
>>                  bpf_linker__new_fd;
>> +               libbpf_probe_bpf_kfunc;
>>   } LIBBPF_1.5.0;
>> diff --git a/tools/lib/bpf/libbpf_probes.c b/tools/lib/bpf/libbpf_probes.c
>> index 9dfbe7750f56..bc1cf2afbe87 100644
>> --- a/tools/lib/bpf/libbpf_probes.c
>> +++ b/tools/lib/bpf/libbpf_probes.c
>> @@ -413,6 +413,42 @@ int libbpf_probe_bpf_map_type(enum bpf_map_type map_type, const void *opts)
>>          return libbpf_err(ret);
>>   }
>>
>> +int libbpf_probe_bpf_kfunc(enum bpf_prog_type prog_type, int kfunc_id,
>> +                          const void *opts)
>> +{
>> +       struct bpf_insn insns[] = {
>> +               BPF_EXIT_INSN(),
>> +               BPF_EXIT_INSN(),
>> +       };
>> +       const size_t insn_cnt = ARRAY_SIZE(insns);
>> +       int err;
>> +       char buf[4096];
>> +
>> +       if (opts)
>> +               return libbpf_err(-EINVAL);
> 
> note how libbpf_probe_bpf_helper() rejects some program types because
> they can't be really loaded. Let's keep it consistent?
> 

Hi andrii, thank you for your guidance, i will add it later.

> pw-bot: cr
> 
>> +
>> +       insns[0].code = BPF_JMP | BPF_CALL;
>> +       insns[0].src_reg = BPF_PSEUDO_KFUNC_CALL;
>> +       insns[0].imm = kfunc_id;
>> +
>> +       /* Now only support kfunc from vmlinux */
>> +       insns[0].off = 0;
> 
> why not support modules from the very beginning?
> 

So can we add a new parameter named like "off"? If it's a module, pass 
the BTF offset to insns[0].off. If it's vmlinux, pass 0.

>> +
>> +       buf[0] = '\0';
>> +       err = probe_prog_load(prog_type, insns, insn_cnt, buf, sizeof(buf));
>> +       if (err < 0)
>> +               return libbpf_err(err);
>> +
>> +       /* If BPF verifier recognizes BPF kfunc but it's not supported for
>> +        * given BPF program type, it will emit "calling kernel function
>> +        * bpf_cpumask_create is not allowed"
>> +        */
>> +       if (err == 0 && strstr(buf, "not allowed"))
> 
> Looking at kernel code, if kfunc ID is not recognized, it seems like
> the verifier won't print anything, is that right? If that's the case,
> then this API will behave differently from libbpf_probe_bpf_helper(),
> which isn't great.
> 

You mean kfunc id is invalid? i try set kfunc id with -1
ret = libbpf_probe_bpf_kfunc(BPF_PROG_TYPE_SYSCALL, -1, NULL);
And the verifier will print like:
"kernel btf_id 4294967295 is not a function"

So "not a function" may also be checked, i will add it in v2.

>> +               return 0;
>> +
>> +       return 1; /* assume supported */
>> +}
>> +
>>   int libbpf_probe_bpf_helper(enum bpf_prog_type prog_type, enum bpf_func_id helper_id,
>>                              const void *opts)
>>   {
>> --
>> 2.43.0
>>


-- 
Best Regards
Dylane Chen