Re: [PATCH for-next v4] RDMA/efa: Add EFA query MR support

["Margolin, Michael" <mrgolin@amazon.com>]

On 1/9/2024 8:05 PM, Leon Romanovsky wrote:
> CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you can confirm the sender and know the content is safe.
>
>
>
> On Mon, Jan 08, 2024 at 09:40:32PM +0200, Leon Romanovsky wrote:
>>
>> On Mon, Jan 8, 2024, at 20:06, Jason Gunthorpe wrote:
>>> On Mon, Jan 08, 2024 at 08:01:40PM +0200, Leon Romanovsky wrote:
>>>>> I was saying in the rdma-core PR that this field shouldn't even
>>>>> exist..
>>>> Something like that?
>>> Yeah, like that. However it is difficult to get the out valid uattr
>>> back in the rdma-core side.
>>>
>>> This is best if the ID's can have well defined not-valid values such
>>> as 0 or -1.
>> Michael tried something like that in previous versions by defining 0xffff as not valid.
>>
>> I didn't like it because there's no promise from PCI core that it is invalid value.
> Michael,
> What do you think?
>
> Thanks

In this case I prefer to keep the explicit validity for consistency with
the device originated values.

In general I think it will be useful to have some convenient way to get
attr validity from the ioctl mechanism in rdma-core.

Michael