[PATCH v2 0/3] nvmet-rdma: SRQ per completion vector

Max Gurtovoy maxg at mellanox.com
Sat Nov 18 05:57:15 PST 2017



On 11/18/2017 2:52 PM, Leon Romanovsky wrote:
> On Fri, Nov 17, 2017 at 09:32:42PM +0200, Max Gurtovoy wrote:
>>
>>
>> On 11/16/2017 8:36 PM, Sagi Grimberg wrote:
>>>
>>>> Since there is an active discussion regarding the CQ pool
>>>> architecture, I decided to push
>>>> this feature (maybe it can be pushed before CQ pool).
>>>>
>>>> This is a new feature for NVMEoF RDMA target,
>>>
>>> Any chance having this for the rest? isert, srpt, svcrdma?
>>>
>>
>> We can implement it for isert, but I think it's better to see how the CQ
>> pool will be defined first.
>> It can bring a big benefit and improvement for ib_srpt (similar to NVMEoF
>> target) but I'm not sure if I can commit for that one soon..
> 
> Too bad, but I don't see inclusion of generic SRQ pool code in RDMA
> subsystem without actual conversion of existing ULP clients.
> 
> Thanks
> 

This patchset adds this feature to NVMEoF target so actually there are 
ULPs that use it. Same issue we have with mr_pool that only RDMA rw.c 
use it (Now we're adding it to NVMEoF initiators too - in review).
I can add srq_pool to iSER target code but I don't want to re-write it 
again in few weeks, when the CQ pool will be added.
Regarding other ULPs, we don't have a testing environment for them so I 
prefer not to commit on their implementation in the near future.

I don't know why we can't add this feature "as is".
Other ULPs maintainers might use it once it will be pushed.



More information about the Linux-nvme mailing list