Re: [PATCH bpf-next v4 0/4] bpf: Implement two type cast kfuncs

[Alexei Starovoitov <ast@meta.com>]

On 11/22/22 7:18 PM, John Fastabend wrote:
> Alexei Starovoitov wrote:
>> On 11/21/22 5:48 PM, John Fastabend wrote:
>>> Yonghong Song wrote:
>>>>
>>>>
>>>> On 11/20/22 6:10 PM, John Fastabend wrote:
>>>>> Yonghong Song wrote:
>>>>>> Currenty, a non-tracing bpf program typically has a single 'context' argument
>>>>>> with predefined uapi struct type. Following these uapi struct, user is able
>>>>>> to access other fields defined in uapi header. Inside the kernel, the
>>>>>> user-seen 'context' argument is replaced with 'kernel context' (or 'kctx'
>>>>>> in short) which can access more information than what uapi header provides.
>>>>>> To access other info not in uapi header, people typically do two things:
>>>>>>      (1). extend uapi to access more fields rooted from 'context'.
>>>>>>      (2). use bpf_probe_read_kernl() helper to read particular field based on
>>>>>>        kctx.
>>>
>>> [...]
>>>
>>>>>    From myside this allows us to pull in the dev info and from that get
>>>>> netns so fixes a gap we had to split into a kprobe + xdp.
>>>>>
>>>>> If we can get a pointer to the recv queue then with a few reads we
>>>>> get the hash, vlan, etc. (see timestapm thread)
>>>>
>>>> Thanks, John. Glad to see it is useful.
>>>>
>>>>>
>>>>> And then last bit is if we can get a ptr to the net ns list, plus
>>>>
>>>> Unfortunately, currently vmlinux btf does not have non-percpu global
>>>> variables, so net_namespace_list is not available to bpf programs.
>>>> But I think we could do the following with a little bit user space
>>>> initial involvement as a workaround.
>>>
>>> What would you think of another kfunc, bpf_get_global_var() to fetch
>>> the global reference and cast it with a type? I think even if you
>>> had it in BTF you would still need some sort of helper otherwise
>>> how would you know what scope of the var should be and get it
>>> correct in type checker as a TRUSTED arg? I think for my use case
>>> UNTRUSTED is find, seeing we do it with probe_reads already, but
>>> getting a TRUSTED arg seems nicer given it can be known correct
>>> from kernel side.
>>>
>>> I was thinking something like,
>>>
>>>     struct net *head = bpf_get_global_var(net_namespace_list,
>>> 				bpf_core_type_id_kernel(struct *net));
>>
>> We cannot do this as ptr_trusted, since it's an unknown cast.
> 
> I think you _could_ do it if the kfunc new to check the case type
> and knew that net_namespace_list should return that specific global.
> The verifier would special code that var and type.

Hard code it in the verifier just for one or two variables? Ouch.
Let's see whether all export_symbol_gpl can work.

>> The verifier cannot trust bpf prog to do the right thing.
>> But we can enable this buy adding export_symbol_gpl global vars to BTF.
>> Then they will be trusted and their types correct.
>> Pretty much like per-cpu variables.
>>
> 
> Yep this is the more generic way and sounds better to me. Anyone
> working on adding the global var to BTF now?

Alan Maguire looked at it. cc-ing.