Received-SPF: pass (google.com: domain of linux-kernel+bounces-138843-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Precedence: bulk
MIME-Version: 1.0
References: <20211113062222.3743909-1-jay.xu@rock-chips.com>
 <1da5cdf0-ccb8-3740-cf96-794c4d5b2eb4@amd.com> <3175d41a-fc44-4741-91ac-005c8f21abb8@vivo.com>
 <9e6f1f52-db49-43bb-a0c2-b0ad12c28aa1@amd.com> <5c7ac24c-79fa-45fc-a4fd-5b8fc77a741b@vivo.com>
 <CABdmKX1OZ9yT3YQA9e7JzKND9wfiL-hnf87Q6v9pwtnAeLHrdA@mail.gmail.com> <0df41277-ded5-4a42-9df5-864d2b6646f5@amd.com>
In-Reply-To: <0df41277-ded5-4a42-9df5-864d2b6646f5@amd.com>
From: "T.J. Mercier" <tjmercier@google.com>
Date: Wed, 10 Apr 2024 08:07:15 -0700
Message-ID: <CABdmKX03Of7OE_9PfAy5DWcCwwvQvJGXDjzAE8c4WRrz_0Z_eg@mail.gmail.com>
Subject: Re: [PATCH] dma-buf: add DMA_BUF_IOCTL_SYNC_PARTIAL support
To: =?UTF-8?Q?Christian_K=C3=B6nig?= <christian.koenig@amd.com>
Cc: Rong Qianfeng <11065417@vivo.com>, Rong Qianfeng <rongqianfeng@vivo.com>, 
	Jianqun Xu <jay.xu@rock-chips.com>, sumit.semwal@linaro.org, 
	pekka.paalanen@collabora.com, daniel.vetter@ffwll.ch, jason@jlekstrand.net, 
	linux-media@vger.kernel.org, dri-devel@lists.freedesktop.org, 
	linaro-mm-sig@lists.linaro.org, linux-kernel@vger.kernel.org, 
	linux-rockchip@lists.infradead.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Wed, Apr 10, 2024 at 7:22=E2=80=AFAM Christian K=C3=B6nig
<christian.koenig@amd.com> wrote:
>
> Am 09.04.24 um 18:37 schrieb T.J. Mercier:
> > On Tue, Apr 9, 2024 at 12:34=E2=80=AFAM Rong Qianfeng <11065417@vivo.co=
m> wrote:
> >>
> >> =E5=9C=A8 2024/4/8 15:58, Christian K=C3=B6nig =E5=86=99=E9=81=93:
> >>> Am 07.04.24 um 09:50 schrieb Rong Qianfeng:
> >>>> [SNIP]
> >>>>> Am 13.11.21 um 07:22 schrieb Jianqun Xu:
> >>>>>> Add DMA_BUF_IOCTL_SYNC_PARTIAL support for user to sync dma-buf wi=
th
> >>>>>> offset and len.
> >>>>> You have not given an use case for this so it is a bit hard to
> >>>>> review. And from the existing use cases I don't see why this should
> >>>>> be necessary.
> >>>>>
> >>>>> Even worse from the existing backend implementation I don't even se=
e
> >>>>> how drivers should be able to fulfill this semantics.
> >>>>>
> >>>>> Please explain further,
> >>>>> Christian.
> >>>> Here is a practical case:
> >>>> The user space can allocate a large chunk of dma-buf for
> >>>> self-management, used as a shared memory pool.
> >>>> Small dma-buf can be allocated from this shared memory pool and
> >>>> released back to it after use, thus improving the speed of dma-buf
> >>>> allocation and release.
> >>>> Additionally, custom functionalities such as memory statistics and
> >>>> boundary checking can be implemented in the user space.
> >>>> Of course, the above-mentioned functionalities require the
> >>>> implementation of a partial cache sync interface.
> >>> Well that is obvious, but where is the code doing that?
> >>>
> >>> You can't send out code without an actual user of it. That will
> >>> obviously be rejected.
> >>>
> >>> Regards,
> >>> Christian.
> >> In fact, we have already used the user-level dma-buf memory pool in th=
e
> >> camera shooting scenario on the phone.
> >>
> >>   From the test results, The execution time of the photo shooting
> >> algorithm has been reduced from 3.8s to 3s.
> >>
> > For phones, the (out of tree) Android version of the system heap has a
> > page pool connected to a shrinker.
>
> Well, it should be obvious but I'm going to repeat it here: Submitting
> kernel patches for our of tree code is a rather *extreme* no-go.
>
Sorry I think my comment led to some confusion. I wasn't suggesting
you should take the patch; it's clearly incomplete. I was pointing out
another option to Rong. It appears Rong is creating a single oversized
dma-buf, and subdividing it in userspace to avoid multiple allocations
for multiple buffers. That would lead to a need for a partial sync of
the buffer from userspace. Instead of that, using a heap with a page
pool gets you kind of the same thing with a lot less headache in
userspace, and no need for partial sync. So I wanted to share that
option, and that can go in just Android if necessary without this
patch.

> That in kernel code *must* have an in kernel user is a number one rule.
>
> When somebody violates this rule he pretty much disqualifying himself as
> reliable source of patches since maintainers then have to assume that
> this person tries to submit code which doesn't have a justification to
> be upstream.
>
> So while this actually looks useful from the technical side as long as
> nobody does an implementation based on an upstream driver I absolutely
> have to reject it from the organizational side.
>
> Regards,
> Christian.
>
> >   That allows you to skip page
> > allocation without fully pinning the memory like you get when
> > allocating a dma-buf that's way larger than necessary. If it's for a
> > camera maybe you need physically contiguous memory, but it's also
> > possible to set that up.
> >
> > https://android.googlesource.com/kernel/common/+/refs/heads/android14-6=
1/drivers/dma-buf/heaps/system_heap.c#377
> >
> >
> >> To be honest, I didn't understand your concern "...how drivers should =
be
> >> able to fulfill this semantics." Can you please help explain it in mor=
e
> >> detail?
> >>
> >> Thanks,
> >>
> >> Rong Qianfeng.
> >>
> >>>> Thanks
> >>>> Rong Qianfeng.
>