Re: [PATCH rdma-next v9 1/5] RDMA/uverbs: Support QP creation with user allocated memory

[Jiri Pirko <jiri@resnulli.us>]

Tue, Jan 27, 2026 at 02:04:27PM CET, sriharsha.basavapatna@broadcom.com wrote:
>On Tue, Jan 27, 2026 at 5:42 PM Jiri Pirko <jiri@resnulli.us> wrote:
>>
>> Tue, Jan 27, 2026 at 11:31:05AM +0100, sriharsha.basavapatna@broadcom.com wrote:
>> >From: Jiri Pirko <jiri@resnulli.us>
>> >
>> >This patch supports creation of QPs with user allocated memory (umem).
>> >This is similar to the existing CQ umem support. This enables userspace
>> >applications to provide pre-allocated buffers for QP send and receive
>> >queues.
>> >
>> >- Add create_qp_umem device operation to the RDMA device ops.
>> >- Implement get_qp_buffer_umem() helper function to handle both VA-based
>> >  and dmabuf-based umem allocation.
>> >- Extend QP creation handler to support umem attributes for SQ and RQ.
>> >- Add new uAPI attributes to specify umem buffers (VA/length or
>> >  FD/offset combinations).
>> >
>> >Signed-off-by: Jiri Pirko <jiri@resnulli.us>
>> >Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
>>
>> When you send patch in my name, and you do some changes (patch
>> description was not present in my draft at least), I would expect some
>> off-list handshake. Is it too much to ask? Looks like you are in a big
>> hurry, that never brought anything good :/
>
>I didn't know you expected an offline handshake, since I was just
>updating the commit message (and that one other line I mentioned
>earlier). But please feel free to suggest/revert changes or if you
>want to push an updated revision yourself, I'm ok either way.

On which planet you find okay to take someones draft (untested) patch
from discussion, take it, add description and send it on his behalf, all
without consulting him?

I'm in a process of testing the patch. Give me day or two, I will send
it myself.

Thanks!


>Thanks,
>-Harsha