Re: [PATCH bpf-next v16 08/14] bpf: pass attached BTF to the bpf_struct_ops subsystem

[Kui-Feng Lee <sinquersw@gmail.com>]

On 1/19/24 10:05, Kui-Feng Lee wrote:
> 
> 
> On 1/18/24 13:56, Martin KaFai Lau wrote:
>> On 1/17/24 5:49 PM, thinker.li@gmail.com wrote:
>>> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
>>> index 0744a1f194fa..ff41f7736618 100644
>>> --- a/kernel/bpf/verifier.c
>>> +++ b/kernel/bpf/verifier.c
>>> @@ -20234,6 +20234,7 @@ static int check_struct_ops_btf_id(struct 
>>> bpf_verifier_env *env)
>>>       const struct btf_member *member;
>>>       struct bpf_prog *prog = env->prog;
>>>       u32 btf_id, member_idx;
>>> +    struct btf *btf;
>>>       const char *mname;
>>>       if (!prog->gpl_compatible) {
>>> @@ -20241,8 +20242,10 @@ static int check_struct_ops_btf_id(struct 
>>> bpf_verifier_env *env)
>>>           return -EINVAL;
>>>       }
>>> +    btf = prog->aux->attach_btf ?: bpf_get_btf_vmlinux();
>>> +
>>
>> just "btf = prog->aux->attach_btf;" which was assigned to 
>> bpf_get_btf_vmlinux() for the non-module case. Take a look at 
>> bpf_prog_load() in syscall.c
> 
> You are right. I have been too protective here.

Just try to remove it. The test case libbpf_probe_prog_types crashes
the kernel for attach_btf is null. It sets neither attach_btf_id nor 
attach_prog_fd.