From: "yangx.jy@fujitsu.com" <yangx.jy@fujitsu.com>
To: Jason Gunthorpe <jgg@nvidia.com>
Cc: Tom Talpey <tom@talpey.com>,
"Gromadzki, Tomasz" <tomasz.gromadzki@intel.com>,
"linux-rdma@vger.kernel.org" <linux-rdma@vger.kernel.org>,
"yanjun.zhu@linux.dev" <yanjun.zhu@linux.dev>,
"rpearsonhpe@gmail.com" <rpearsonhpe@gmail.com>,
"y-goto@fujitsu.com" <y-goto@fujitsu.com>,
"lizhijian@fujitsu.com" <lizhijian@fujitsu.com>
Subject: Re: [RFC PATCH 0/2] RDMA/rxe: Add RDMA Atomic Write operation
Date: Tue, 11 Jan 2022 02:34:51 +0000 [thread overview]
Message-ID: <61DCECAF.10206@fujitsu.com> (raw)
In-Reply-To: <20220110154205.GD2328285@nvidia.com>
On 2022/1/10 23:42, Jason Gunthorpe wrote:
> On Thu, Jan 06, 2022 at 01:54:30AM +0000, yangx.jy@fujitsu.com wrote:
>> On 2022/1/6 8:01, Jason Gunthorpe wrote:
>>> On Wed, Jan 05, 2022 at 01:00:42AM +0000, yangx.jy@fujitsu.com wrote:
>>>
>>>> 1) In kernel, current SoftRoCE copies the content of struct rdma to RETH
>>>> and copies the content of struct atomic to AtomicETH.
>>>> 2) IBTA defines that RDMA Atomic Write uses RETH + payload.
>>>> According to these two reasons, I perfer to tweak the existing struct rdma.
>>> No this is basically meaningless
>>>
>>> The wr struct is designed as a 'tagged union' where the op specified
>>> which union is in effect.
>>>
>>> It turns out that the op generally specifies the network headers as
>>> well, but that is just a side effect.
>>>
>>>>>> How about adding a member in struct rdma? for example:
>>>>>> struct {
>>>>>> uint64_t remote_addr;
>>>>>> uint32_t rkey;
>>>>>> uint64_t wr_value:
>>>>>> } rdma;
>>>>> Yes, that's what Tomasz and I were suggesting - a new template for the
>>>>> ATOMIC_WRITE request payload. The three fields are to be supplied by
>>>>> the verb consumer when posting the work request.
>>>> OK, I will update the patch in this way.
>>> We are not extending the ib_send_wr anymore anyhow.
>>>
>>> You should implement new ops inside struct ibv_qp_ex as function
>>> calls.
>> Hi Jason,
>>
>> For SoftRoCE, do you mean that I only need to extend struct rxe_send_wr
>> and add ibv_wr_rdma_atomic_write() ?
>> struct rxe_send_wr {
>> ...
>> struct {
>> __aligned_u64 remote_addr;
>> + __aligned_u64 atomic_wr;
>> __u32 rkey;
>> __u32 reserved;
>> } rdma;
> You can't make a change like this to anything in
> include/uapi/rdma/rdma_user_rxe.h, it has to remain compatiable.
Hi Jason,
How about adding atomic_wr member at the end of struct rdma? like this:
------------------------------------------------------
struct rxe_send_wr {
...
struct {
__aligned_u64 remote_addr;
__u32 rkey;
__u32 reserved;
+ __aligned_u64 atomic_wr;
} rdma;
...
}
------------------------------------------------------
If it is also wrong, how to extend struct rdma?
>
>> static inline void ibv_wr_rdma_atomic_write(struct ibv_qp_ex *qp,
>> uint32_t rkey,
>> uint64_t remote_addr)
>> {
>> qp->wr_rdma_atomic_write(qp, rkey, remote_addr);
>> }
> Yes, something like that
>
>> Besides, could you tell me why we cannot extend struct ibv_send_wr for
>> ibv_post_send()?
> The ABI is not allowed to change so what is doable with it is very
> limited.
Thanks for your explanation.
Best Regards,
Xiao Yang
> Jason
next prev parent reply other threads:[~2022-01-11 2:34 UTC|newest]
Thread overview: 33+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-12-30 12:14 [RFC PATCH 0/2] RDMA/rxe: Add RDMA Atomic Write operation Xiao Yang
2021-12-30 12:14 ` [RFC PATCH 1/2] RDMA/rxe: Rename send_atomic_ack() and atomic member of struct resp_res Xiao Yang
2021-12-30 12:14 ` [RFC PATCH 2/2] RDMA/rxe: Add RDMA Atomic Write operation Xiao Yang
2021-12-30 21:39 ` Tom Talpey
2021-12-31 8:29 ` yangx.jy
2021-12-31 15:09 ` Tom Talpey
[not found] ` <61D563B4.2070106@fujitsu.com>
2022-01-07 15:50 ` Tom Talpey
2022-01-07 17:11 ` Jason Gunthorpe
2022-01-12 9:24 ` yangx.jy
2022-01-05 23:53 ` Jason Gunthorpe
2022-01-06 10:52 ` yangx.jy
2022-01-06 13:00 ` Jason Gunthorpe
2022-01-07 2:15 ` yangx.jy
2022-01-07 12:22 ` Jason Gunthorpe
2022-01-07 15:38 ` Tom Talpey
2022-01-07 19:28 ` Jason Gunthorpe
2022-01-07 20:11 ` Tom Talpey
2021-12-31 3:01 ` lizhijian
2021-12-31 6:02 ` yangx.jy
2021-12-30 19:21 ` [RFC PATCH 0/2] " Gromadzki, Tomasz
2021-12-30 21:42 ` Tom Talpey
2021-12-31 6:30 ` yangx.jy
2022-01-04 9:28 ` yangx.jy
2022-01-04 15:17 ` Tom Talpey
2022-01-05 1:00 ` yangx.jy
2022-01-06 0:01 ` Jason Gunthorpe
2022-01-06 1:54 ` yangx.jy
2022-01-10 15:42 ` Jason Gunthorpe
2022-01-11 2:34 ` yangx.jy [this message]
2022-01-11 23:29 ` Jason Gunthorpe
2022-02-11 13:18 ` Gromadzki, Tomasz
2022-02-17 3:50 ` yangx.jy
2022-02-19 10:37 ` Leon Romanovsky
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=61DCECAF.10206@fujitsu.com \
--to=yangx.jy@fujitsu.com \
--cc=jgg@nvidia.com \
--cc=linux-rdma@vger.kernel.org \
--cc=lizhijian@fujitsu.com \
--cc=rpearsonhpe@gmail.com \
--cc=tom@talpey.com \
--cc=tomasz.gromadzki@intel.com \
--cc=y-goto@fujitsu.com \
--cc=yanjun.zhu@linux.dev \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).