[PATCH 11/15] RDMA/hbl: add habanalabs RDMA driver

[Omer Shpigelman]

On 7/17/24 15:33, Leon Romanovsky wrote:
> On Wed, Jul 17, 2024 at 10:51:03AM +0000, Omer Shpigelman wrote:
>> On 7/17/24 10:36, Leon Romanovsky wrote:
>>> On Wed, Jul 17, 2024 at 07:08:59AM +0000, Omer Shpigelman wrote:
>>>> On 7/16/24 16:40, Jason Gunthorpe wrote:
>>>>> On Sun, Jul 14, 2024 at 10:18:12AM +0000, Omer Shpigelman wrote:
>>>>>> On 7/12/24 16:08, Jason Gunthorpe wrote:
>>>>>>> [You don't often get email from jgg at ziepe.ca. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
>>>>>>>
>>>>>>> On Fri, Jun 28, 2024 at 10:24:32AM +0000, Omer Shpigelman wrote:
>>>>>>>
>>>>>>>> We need the core driver to access the IB driver (and to the ETH driver as
>>>>>>>> well). As you wrote, we can't use exported symbols from our IB driver nor
>>>>>>>> rely on function pointers, but what about providing the core driver an ops
>>>>>>>> structure? meaning exporting a register function from the core driver that
>>>>>>>> should be called by the IB driver during auxiliary device probe.
>>>>>>>> Something like:
>>>>>>>>
>>>>>>>> int hbl_cn_register_ib_aux_dev(struct auxiliary_device *adev,
>>>>>>>>                              struct hbl_ib_ops *ops)
>>>>>>>> {
>>>>>>>> ...
>>>>>>>> }
>>>>>>>> EXPORT_SYMBOL(hbl_cn_register_ib_aux_dev);
>>>>>>>
>>>>>>> Definately do not do some kind of double-register like this.
>>>>>>>
>>>>>>> The auxiliary_device scheme can already be extended to provide ops for
>>>>>>> each sub device.
>>>>>>>
>>>>>>> Like
>>>>>>>
>>>>>>> struct habana_driver {
>>>>>>>    struct auxiliary_driver base;
>>>>>>>    const struct habana_ops *ops;
>>>>>>> };
>>>>>>>
>>>>>>> If the ops are justified or not is a different question.
>>>>>>>
>>>>>>
>>>>>> Well, I suggested this double-register option because I got a comment that
>>>>>> the design pattern of embedded ops structure shouldn't be used.
>>>>>> So I'm confused now...
>>>>>
>>>>> Yeah, don't stick ops in random places, but the device_driver is the
>>>>> right place.
>>>>>
>>>>
>>>> Sorry, let me explain again. My original code has an ops structure
>>>> exactly like you are suggesting now (see struct hbl_aux_dev in the first
>>>> patch of the series). But I was instructed not to use this ops structure
>>>> and to rely on exported symbols for inter-driver communication.
>>>> I'll be happy to use this ops structure like in your example rather than
>>>> converting my code to use exported symbols.
>>>> Leon - am I missing anything? what's the verdict here?
>>>
>>> You are missing the main sentence from Jason's response:  "don't stick ops in random places".
>>>
>>> It is fine to have ops in device driver, so the core driver can call them. However, in your
>>> original code, you added ops everywhere. It caused to the need to implement module reference
>>> counting and crazy stuff like calls to lock and unlock functions from the aux driver to the core.
>>>
>>> Verdict is still the same. Core driver should provide EXPORT_SYMBOLs, so the aux driver can call
>>> them directly and enjoy from proper module loading and unloading.
>>>
>>> The aux driver can have ops in the device driver, so the core driver can call them to perform something
>>> specific for that aux driver.
>>>
>>> Calls between aux drivers should be done via the core driver.
>>>
>>> Thanks
>>
>> The only place we have an ops structure is in the device driver,
>> similarly to Jason's example. In our code it is struct hbl_aux_dev. What
>> other random places did you see?
> 
> This is exactly random place.
> 
> I suggest you to take time, learn how existing drivers in netdev and
> RDMA uses auxbus infrastructure and follow the same pattern. There are
> many examples already in the kernel.
> 
> And no, if you do everything right, you won't need custom module
> reference counting and other hacks. There is nothing special in your
> device/driver which requires special treatment.
> 
> Thanks

How come it is a random place?
Look at irdma/i40e - they have an ops struct (struct i40e_ops) embedded
in their shared aux struct (struct i40e_auxiliary_device) which is the
wrapper of the base aux struct (struct auxiliary_device).
This is very similar to what we have - a pointer to an ops struct
(void *aux_ops) embedded in our shared aux struct (struct hbl_aux_dev)
which is the wrapper of the base struct (struct auxiliary_device).

The only difference is that they put their ops struct inside some info
struct (struct i40e_info) and we have a separate pointer for that info
(void *aux_data).

In addition, they use the ops struct for accessing the net driver from the
RDMA driver, meaning son-to-parent communication instead of having an
exported symbol e.g. i40e_client_request_reset().
They have only a single operation as EXPORT_SYMBOL function for
(un)registering the son - i40e_client_device_register() and
i40e_client_device_unregister().

So what is the problem with how we implemented it?