Re: [RFC PATCH net-next v5 04/11] net/smc: implement some unsupported operations of loopback-ism

[Alexandra Winter]

On 09.04.24 03:44, Wen Gu wrote:
> 
> 
> On 2024/4/4 23:15, Niklas Schnelle wrote:
>> On Thu, 2024-04-04 at 21:12 +0800, Wen Gu wrote:
>>>
>>> On 2024/4/4 19:42, Niklas Schnelle wrote:
>>>> On Thu, 2024-04-04 at 17:32 +0800, Wen Gu wrote:
>>>>>
>>>>> On 2024/4/4 00:25, Gerd Bayer wrote:
>>>>>> On Sun, 2024-03-24 at 21:55 +0800, Wen Gu wrote:
>>>>>>> This implements some operations that loopback-ism does not support
>>>>>>> currently:
>>>>>>>     - vlan operations, since there is no strong use-case for it.
>>>>>>>     - signal_event operations, since there is no event to be processed
>>>>>>> by the loopback-ism device.
>>>>>>
>>>>>> Hi Wen,
>>>>>>
>>>>>> I wonder if the these operations that are not supported by loopback-ism
>>>>>> should rather be marked "optional" in the struct smcd_ops, and the
>>>>>> calling code should call these only when they are implemented.
>>>>>>
>>>>>> Of course this would mean more changes to net/smc/smc_core.c - but
>>>>>> loopback-ism could omit these "boiler-plate" functions.
>>>>>>
>>>>>
>>>>> Hi Gerd.
>>>>>
>>>>> Thank you for the thoughts! I agree that checks like 'if(smcd->ops->xxx)'
>>>>> can avoid the device driver from implementing unsupported operations. But I
>>>>> am afraid that which operations need to be defined as 'optional' may differ
>>>>> from different device perspectives (e.g. for loopback-ism they are vlan-related
>>>>> opts and signal_event). So I perfer to simply let the smc protocol assume
>>>>> that all operations have been implemented, and let drivers to decide which
>>>>> ones are unsupported in implementation. What do you think?
>>>>>
>>>>> Thanks!
>>>>>
>>>>
>>>> I agree with Gerd, in my opinion it is better to document ops as
>>>> optional and then allow their function pointers to be NULL and check
>>>> for that. Acting like they are supported and then they turn out to be
>>>> nops to me seems to contradict the principle of least surprises. I also
>>>> think we can find a subset of mandatory ops without which SMC-D is
>>>> impossible and then everything else should be optional.
>>>
>>> I see. If we all agree to classify smcd_ops into mandatory and optional ones,
>>> I'll add a patch to mark the optional ops and check if they are implemented.
>>
>> Keep in mind I don't speak for the SMC maintainers but that does sound
>> reasonable to me.
>>
> 
> Hi Wenjia and Jan, do you have any comments on this and [1]? Thanks!
> 
> [1] https://lore.kernel.org/netdev/60b4aec0b4bf4474d651b653c86c280dafc4518a.camel@xxxxxxxxxxxxx/
> 
>>>
>>>>
>>>> As a first guess I think the following options may be mandatory:
>>>>
>>>> * query_remote_gid()
>>>> * register_dmb()/unregister_dmb()
>>>> * move_data()
>>>>     For this one could argue that either move_data() or
>>>>     attach_dmb()/detach_dmb() is required though personally I would
>>>>     prefer to always have move_data() as a fallback and simple API
>>>> * supports_v2()
>>>> * get_local_gid()
>>>> * get_chid()
>>>> * get_dev()
>>> I agree with this classification. Just one point, maybe we can take
>>> supports_v2() as an optional ops, like support_dmb_nocopy()? e.g. if
>>> it is not implemented, we treat it as an ISMv1.
>>>
>>> Thanks!
>>
>> Interpreting a NULL supports_v2() as not supporting v2 sounds
>> reasonable to me.
> 

Let me add my thoughts to the discussion:
For the vlan operations and signal_event operations that loopback-ism does
not support:
I like the idea to set the ops to NULL and make sure the caller checks that
and can live with it. That is readable and efficient.

I don't think there is a need to discuss a strategy now, which ops could be
optional in the future. This is all inside the kernel. loopback-ism is even 
inside the smc module. Such comments in the code get outdated very easily.

I would propose to mark those as optional struct smcd_ops, where all callers can 
handle a NULL pointer and still be productive.
Future support of other devices for SMC-D can update that.