From: Breno Leitao <leitao@debian.org>
To: dccp@vger.kernel.org
Subject: Re: [PATCH RFC] io_uring: Pass whole sqe to commands
Date: Thu, 13 Apr 2023 16:47:56 +0000 [thread overview]
Message-ID: <ZDgyPL6UrX/MaBR4@gmail.com> (raw)
In-Reply-To: <20230406165705.3161734-1-leitao@debian.org>
Hello Ming,
On Thu, Apr 13, 2023 at 10:56:49AM +0800, Ming Lei wrote:
> On Thu, Apr 06, 2023 at 09:57:05AM -0700, Breno Leitao wrote:
> > Currently uring CMD operation relies on having large SQEs, but future
> > operations might want to use normal SQE.
> >
> > The io_uring_cmd currently only saves the payload (cmd) part of the SQE,
> > but, for commands that use normal SQE size, it might be necessary to
> > access the initial SQE fields outside of the payload/cmd block. So,
> > saves the whole SQE other than just the pdu.
> >
> > This changes slighlty how the io_uring_cmd works, since the cmd
> > structures and callbacks are not opaque to io_uring anymore. I.e, the
> > callbacks can look at the SQE entries, not only, in the cmd structure.
> >
> > The main advantage is that we don't need to create custom structures for
> > simple commands.
> >
> > Suggested-by: Pavel Begunkov <asml.silence@gmail.com>
> > Signed-off-by: Breno Leitao <leitao@debian.org>
> > ---
>
> ...
>
> > diff --git a/io_uring/uring_cmd.c b/io_uring/uring_cmd.c
> > index 2e4c483075d3..9648134ccae1 100644
> > --- a/io_uring/uring_cmd.c
> > +++ b/io_uring/uring_cmd.c
> > @@ -63,14 +63,15 @@ EXPORT_SYMBOL_GPL(io_uring_cmd_done);
> > int io_uring_cmd_prep_async(struct io_kiocb *req)
> > {
> > struct io_uring_cmd *ioucmd = io_kiocb_to_cmd(req, struct io_uring_cmd);
> > - size_t cmd_size;
> > + size_t size = sizeof(struct io_uring_sqe);
> >
> > BUILD_BUG_ON(uring_cmd_pdu_size(0) != 16);
> > BUILD_BUG_ON(uring_cmd_pdu_size(1) != 80);
> >
> > - cmd_size = uring_cmd_pdu_size(req->ctx->flags & IORING_SETUP_SQE128);
> > + if (req->ctx->flags & IORING_SETUP_SQE128)
> > + size <<= 1;
> >
> > - memcpy(req->async_data, ioucmd->cmd, cmd_size);
> > + memcpy(req->async_data, ioucmd->sqe, size);
>
> The copy will make some fields of sqe become READ TWICE, and driver may see
> different sqe field value compared with the one observed in io_init_req().
This copy only happens if the operation goes to the async path
(calling io_uring_cmd_prep_async()). This only happens if
f_op->uring_cmd() returns -EAGAIN.
ret = file->f_op->uring_cmd(ioucmd, issue_flags);
if (ret = -EAGAIN) {
if (!req_has_async_data(req)) {
if (io_alloc_async_data(req))
return -ENOMEM;
io_uring_cmd_prep_async(req);
}
return -EAGAIN;
}
Are you saying that after this copy, the operation is still reading from
sqe instead of req->async_data?
If you have an example of the two copes flow, that would be great.
Thanks for the review,
Breno
WARNING: multiple messages have this Message-ID (diff)
From: Breno Leitao <leitao@debian.org>
To: Ming Lei <ming.lei@redhat.com>
Cc: asml.silence@gmail.com, axboe@kernel.dk, davem@davemloft.net,
dccp@vger.kernel.org, dsahern@kernel.org, edumazet@google.com,
io-uring@vger.kernel.org, kuba@kernel.org, leit@fb.com,
linux-kernel@vger.kernel.org, marcelo.leitner@gmail.com,
matthieu.baerts@tessares.net, mptcp@lists.linux.dev,
netdev@vger.kernel.org, pabeni@redhat.com,
willemdebruijn.kernel@gmail.com
Subject: Re: [PATCH RFC] io_uring: Pass whole sqe to commands
Date: Thu, 13 Apr 2023 09:47:56 -0700 [thread overview]
Message-ID: <ZDgyPL6UrX/MaBR4@gmail.com> (raw)
In-Reply-To: <ZDdvcSKLa6ZEAhRW@ovpn-8-18.pek2.redhat.com>
Hello Ming,
On Thu, Apr 13, 2023 at 10:56:49AM +0800, Ming Lei wrote:
> On Thu, Apr 06, 2023 at 09:57:05AM -0700, Breno Leitao wrote:
> > Currently uring CMD operation relies on having large SQEs, but future
> > operations might want to use normal SQE.
> >
> > The io_uring_cmd currently only saves the payload (cmd) part of the SQE,
> > but, for commands that use normal SQE size, it might be necessary to
> > access the initial SQE fields outside of the payload/cmd block. So,
> > saves the whole SQE other than just the pdu.
> >
> > This changes slighlty how the io_uring_cmd works, since the cmd
> > structures and callbacks are not opaque to io_uring anymore. I.e, the
> > callbacks can look at the SQE entries, not only, in the cmd structure.
> >
> > The main advantage is that we don't need to create custom structures for
> > simple commands.
> >
> > Suggested-by: Pavel Begunkov <asml.silence@gmail.com>
> > Signed-off-by: Breno Leitao <leitao@debian.org>
> > ---
>
> ...
>
> > diff --git a/io_uring/uring_cmd.c b/io_uring/uring_cmd.c
> > index 2e4c483075d3..9648134ccae1 100644
> > --- a/io_uring/uring_cmd.c
> > +++ b/io_uring/uring_cmd.c
> > @@ -63,14 +63,15 @@ EXPORT_SYMBOL_GPL(io_uring_cmd_done);
> > int io_uring_cmd_prep_async(struct io_kiocb *req)
> > {
> > struct io_uring_cmd *ioucmd = io_kiocb_to_cmd(req, struct io_uring_cmd);
> > - size_t cmd_size;
> > + size_t size = sizeof(struct io_uring_sqe);
> >
> > BUILD_BUG_ON(uring_cmd_pdu_size(0) != 16);
> > BUILD_BUG_ON(uring_cmd_pdu_size(1) != 80);
> >
> > - cmd_size = uring_cmd_pdu_size(req->ctx->flags & IORING_SETUP_SQE128);
> > + if (req->ctx->flags & IORING_SETUP_SQE128)
> > + size <<= 1;
> >
> > - memcpy(req->async_data, ioucmd->cmd, cmd_size);
> > + memcpy(req->async_data, ioucmd->sqe, size);
>
> The copy will make some fields of sqe become READ TWICE, and driver may see
> different sqe field value compared with the one observed in io_init_req().
This copy only happens if the operation goes to the async path
(calling io_uring_cmd_prep_async()). This only happens if
f_op->uring_cmd() returns -EAGAIN.
ret = file->f_op->uring_cmd(ioucmd, issue_flags);
if (ret == -EAGAIN) {
if (!req_has_async_data(req)) {
if (io_alloc_async_data(req))
return -ENOMEM;
io_uring_cmd_prep_async(req);
}
return -EAGAIN;
}
Are you saying that after this copy, the operation is still reading from
sqe instead of req->async_data?
If you have an example of the two copes flow, that would be great.
Thanks for the review,
Breno
next prev parent reply other threads:[~2023-04-13 16:47 UTC|newest]
Thread overview: 108+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-04-06 16:57 [PATCH RFC] io_uring: Pass whole sqe to commands Breno Leitao
2023-04-06 16:57 ` Breno Leitao
2023-04-06 17:50 ` io_uring: Pass whole sqe to commands: Tests Results MPTCP CI
2023-04-07 18:51 ` [PATCH RFC] io_uring: Pass whole sqe to commands Keith Busch
2023-04-07 18:51 ` Keith Busch
2023-04-07 19:43 ` io_uring: Pass whole sqe to commands: Tests Results MPTCP CI
2023-04-11 12:22 ` [PATCH RFC] io_uring: Pass whole sqe to commands Breno Leitao
2023-04-11 12:22 ` Breno Leitao
2023-04-11 12:39 ` Pavel Begunkov
2023-04-11 12:39 ` Pavel Begunkov
2023-04-13 2:56 ` Ming Lei
2023-04-13 2:56 ` Ming Lei
2023-04-13 16:47 ` Breno Leitao [this message]
2023-04-13 16:47 ` Breno Leitao
2023-04-14 2:12 ` Ming Lei
2023-04-14 2:12 ` Ming Lei
2023-04-14 13:12 ` Pavel Begunkov
2023-04-14 13:12 ` Pavel Begunkov
2023-04-14 13:59 ` Ming Lei
2023-04-14 13:59 ` Ming Lei
2023-04-14 14:56 ` Pavel Begunkov
2023-04-14 14:56 ` Pavel Begunkov
2023-04-16 9:51 ` Ming Lei
2023-04-16 9:51 ` Ming Lei
-- strict thread matches above, loose matches on Subject: below --
2023-04-06 14:43 [RFC PATCH 4/4] net: add uring_cmd callback to raw "protocol" Breno Leitao
2023-04-06 14:43 ` Breno Leitao
2023-04-06 14:43 [RFC PATCH 3/4] net: add uring_cmd callback to TCP Breno Leitao
2023-04-06 14:43 ` Breno Leitao
2023-04-06 20:35 ` kernel test robot
2023-04-06 14:43 [RFC PATCH 2/4] net: add uring_cmd callback to UDP Breno Leitao
2023-04-06 14:43 ` Breno Leitao
2023-04-06 19:03 ` kernel test robot
2023-04-11 12:54 ` Pavel Begunkov
2023-04-11 12:54 ` Pavel Begunkov
2023-04-06 14:43 [RFC PATCH 1/4] net: wire up support for file_operations->uring_cmd() Breno Leitao
2023-04-06 14:43 ` Breno Leitao
2023-04-06 14:43 [PATCH 0/5] add initial io_uring_cmd support for sockets Breno Leitao
2023-04-06 14:43 ` Breno Leitao
2023-04-06 15:34 ` Willem de Bruijn
2023-04-06 15:34 ` Willem de Bruijn
2023-04-06 15:59 ` Breno Leitao
2023-04-06 15:59 ` Breno Leitao
2023-04-06 16:41 ` Keith Busch
2023-04-06 16:41 ` Keith Busch
2023-04-06 16:49 ` Jens Axboe
2023-04-06 16:49 ` Jens Axboe
2023-04-06 16:58 ` Breno Leitao
2023-04-06 16:58 ` Breno Leitao
2023-04-06 18:16 ` Willem de Bruijn
2023-04-06 18:16 ` Willem de Bruijn
2023-04-07 2:46 ` David Ahern
2023-04-07 2:46 ` David Ahern
2023-04-11 11:59 ` Breno Leitao
2023-04-11 12:00 ` Breno Leitao
2023-04-11 14:36 ` David Ahern
2023-04-11 14:36 ` David Ahern
2023-04-11 14:41 ` Jens Axboe
2023-04-11 14:41 ` Jens Axboe
2023-04-11 14:51 ` Willem de Bruijn
2023-04-11 14:51 ` Willem de Bruijn
2023-04-11 14:54 ` Jens Axboe
2023-04-11 14:54 ` Jens Axboe
2023-04-11 15:00 ` Willem de Bruijn
2023-04-11 15:00 ` Willem de Bruijn
2023-04-11 15:06 ` Jens Axboe
2023-04-11 15:06 ` Jens Axboe
2023-04-11 15:10 ` David Ahern
2023-04-11 15:10 ` David Ahern
2023-04-11 15:17 ` Jens Axboe
2023-04-11 15:17 ` Jens Axboe
2023-04-11 15:24 ` Willem de Bruijn
2023-04-11 15:24 ` Willem de Bruijn
2023-04-11 15:27 ` David Ahern
2023-04-11 15:27 ` David Ahern
2023-04-11 15:28 ` Jens Axboe
2023-04-11 15:28 ` Jens Axboe
2023-04-11 15:29 ` Jens Axboe
2023-04-11 15:29 ` Jens Axboe
2023-04-12 7:39 ` David Laight
2023-04-12 7:39 ` David Laight
2023-04-12 13:53 ` Breno Leitao
2023-04-12 13:53 ` Breno Leitao
2023-04-12 14:28 ` Willem de Bruijn
2023-04-12 14:28 ` Willem de Bruijn
2023-04-13 0:02 ` Breno Leitao
2023-04-13 0:02 ` Breno Leitao
2023-04-13 14:24 ` Willem de Bruijn
2023-04-13 14:24 ` Willem de Bruijn
2023-04-13 14:45 ` Jakub Kicinski
2023-04-13 14:45 ` Jakub Kicinski
2023-04-13 14:57 ` David Laight
2023-04-13 14:57 ` David Laight
2023-04-18 13:23 ` Breno Leitao
2023-04-18 13:23 ` Breno Leitao
2023-04-18 19:41 ` Willem de Bruijn
2023-04-18 19:41 ` Willem de Bruijn
2023-04-20 14:43 ` Breno Leitao
2023-04-20 14:43 ` Breno Leitao
2023-04-20 16:48 ` Willem de Bruijn
2023-04-20 16:48 ` Willem de Bruijn
2023-05-02 9:21 ` Adrien Delorme
2023-05-02 9:21 ` Adrien Delorme
2023-05-02 13:03 ` Pavel Begunkov
2023-05-02 13:03 ` Pavel Begunkov
2023-05-03 13:11 ` Adrien Delorme
2023-05-03 13:11 ` Adrien Delorme
2023-05-03 13:27 ` David Laight
2023-05-03 13:27 ` David Laight
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=ZDgyPL6UrX/MaBR4@gmail.com \
--to=leitao@debian.org \
--cc=dccp@vger.kernel.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.