Re: bpf: btf: Change how section is supported in btf_header

Re: bpf: btf: Change how section is supported in btf_header

[Dmitry Vyukov]

Hi Martin,

I am looking at the subj commit:

 static int btf_add_type(struct btf_verifier_env *env, struct btf_type *t)
@@ -1754,9 +1756,9 @@ static int btf_check_all_metas(struct
btf_verifier_env *env)
        struct btf_header *hdr;
        void *cur, *end;

-       hdr = btf->hdr;
+       hdr = &btf->hdr;
        cur = btf->nohdr_data + hdr->type_off;
-       end = btf->nohdr_data + hdr->str_off;
+       end = btf->nohdr_data + hdr->type_len;

Shouldn't this be:

+       end = cur + hdr->type_len;

? Or otherwise I am having trouble understanding meaning of fields.

On a related note, what's between header and type_off? Is type_off
supposed to be 0 always?

Thanks

Re: bpf: btf: Change how section is supported in btf_header

[Martin KaFai Lau]

On Tue, Sep 11, 2018 at 06:40:05PM +0200, Dmitry Vyukov wrote:
> Hi Martin,
> 
> I am looking at the subj commit:
> 
>  static int btf_add_type(struct btf_verifier_env *env, struct btf_type *t)
> @@ -1754,9 +1756,9 @@ static int btf_check_all_metas(struct
> btf_verifier_env *env)
>         struct btf_header *hdr;
>         void *cur, *end;
> 
> -       hdr = btf->hdr;
> +       hdr = &btf->hdr;
>         cur = btf->nohdr_data + hdr->type_off;
> -       end = btf->nohdr_data + hdr->str_off;
> +       end = btf->nohdr_data + hdr->type_len;
> 
> Shouldn't this be:
> 
> +       end = cur + hdr->type_len;
> 
> ? Or otherwise I am having trouble understanding meaning of fields.
You are correct.  Thanks for pointing this out.
Do you want to post an offical patch for the bpf branch?

> 
> On a related note, what's between header and type_off? Is type_off
> supposed to be 0 always?
type section is always the first section for now (i.e. immediately after
the header).  Some other sections could be introduced later and it could
be located before the type section such that the type_off will not be 0.

Re: bpf: btf: Change how section is supported in btf_header

[Dmitry Vyukov]

On Tue, Sep 11, 2018 at 8:10 PM, Martin KaFai Lau <kafai@fb.com> wrote:
> On Tue, Sep 11, 2018 at 06:40:05PM +0200, Dmitry Vyukov wrote:
>> Hi Martin,
>>
>> I am looking at the subj commit:
>>
>>  static int btf_add_type(struct btf_verifier_env *env, struct btf_type *t)
>> @@ -1754,9 +1756,9 @@ static int btf_check_all_metas(struct
>> btf_verifier_env *env)
>>         struct btf_header *hdr;
>>         void *cur, *end;
>>
>> -       hdr = btf->hdr;
>> +       hdr = &btf->hdr;
>>         cur = btf->nohdr_data + hdr->type_off;
>> -       end = btf->nohdr_data + hdr->str_off;
>> +       end = btf->nohdr_data + hdr->type_len;
>>
>> Shouldn't this be:
>>
>> +       end = cur + hdr->type_len;
>>
>> ? Or otherwise I am having trouble understanding meaning of fields.
> You are correct.  Thanks for pointing this out.
> Do you want to post an offical patch for the bpf branch?

no :)

Only if you want to send a syzkaller patch with btf descriptions ;)

>> On a related note, what's between header and type_off? Is type_off
>> supposed to be 0 always?
> type section is always the first section for now (i.e. immediately after
> the header).  Some other sections could be introduced later and it could
> be located before the type section such that the type_off will not be 0.

I see. Thanks.