From: "Christian König" <christian.koenig@amd.com> To: Daniel Stone <daniel@fooishbar.org> Cc: Lucas Stach <l.stach@pengutronix.de>, Pekka Paalanen <ppaalanen@gmail.com>, "Sharma, Shashank" <Shashank.Sharma@amd.com>, lkml <linux-kernel@vger.kernel.org>, dri-devel <dri-devel@lists.freedesktop.org>, Nicolas Dufresne <nicolas@ndufresne.ca>, linaro-mm-sig@lists.linaro.org, Sumit Semwal <sumit.semwal@linaro.org>, linux-media <linux-media@vger.kernel.org> Subject: Re: DMA-buf and uncached system memory Date: Thu, 23 Jun 2022 13:32:18 +0200 [thread overview] Message-ID: <578953dd-6298-2bfe-a8fb-52004b84fd17@amd.com> (raw) In-Reply-To: <CAPj87rOykZv7bjNhHPT4StrsPz8Y_DWqab4Ryq=Qqh77LS2e=Q@mail.gmail.com> Am 23.06.22 um 13:27 schrieb Daniel Stone: > Hi Christian, > > On Thu, 23 Jun 2022 at 12:11, Christian König <christian.koenig@amd.com> wrote: >>> In fact DMA-buf sharing works fine on most of those SoCs because >>> everyone just assumes that all the accelerators don't snoop, so the >>> memory shared via DMA-buf is mostly CPU uncached. It only falls apart >>> for uses like the UVC cameras, where the shared buffer ends up being >>> CPU cached. >> Well then the existing DMA-buf framework is not what you want to use for >> this. >> >>> Non-coherent without explicit domain transfer points is just not going >>> to work. So why can't we solve the issue for DMA-buf in the same way as >>> the DMA API already solved it years ago: by adding the equivalent of >>> the dma_sync calls that do cache maintenance when necessary? On x86 (or >>> any system where things are mostly coherent) you could still no-op them >>> for the common case and only trigger cache cleaning if the importer >>> explicitly says that is going to do a non-snooping access. >> Because DMA-buf is a framework for buffer sharing between cache coherent >> devices which don't signal transitions. >> >> We intentionally didn't implemented any of the dma_sync_* functions >> because that would break the intended use case. >> >> You can of course use DMA-buf in an incoherent environment, but then you >> can't expect that this works all the time. >> >> This is documented behavior and so far we have bluntly rejected any of >> the complains that it doesn't work on most ARM SoCs and I don't really >> see a way to do this differently. > For some strange reason, 'let's do buffer sharing but make sure it > doesn't work on Arm' wasn't exactly the intention of the groups who > came together under the Linaro umbrella to create dmabuf. > > If it's really your belief that dmabuf requires universal snooping, I > recommend you send the patch to update the documentation, as well as > to remove DRIVER_PRIME from, realistically, most non-PCIE drivers. Well, to be honest I think that would indeed be necessary. What we have created are essentially two different worlds, one for PCI devices and one for the rest. This was indeed not the intention, but it's a fact that basically all DMA-buf based PCI drivers assume coherent access. Regards, Christian. > > Cheers, > Daniel
WARNING: multiple messages have this Message-ID (diff)
From: "Christian König" <christian.koenig@amd.com> To: Daniel Stone <daniel@fooishbar.org> Cc: "Sharma, Shashank" <Shashank.Sharma@amd.com>, lkml <linux-kernel@vger.kernel.org>, dri-devel <dri-devel@lists.freedesktop.org>, Nicolas Dufresne <nicolas@ndufresne.ca>, linaro-mm-sig@lists.linaro.org, Pekka Paalanen <ppaalanen@gmail.com>, linux-media <linux-media@vger.kernel.org>, Sumit Semwal <sumit.semwal@linaro.org> Subject: Re: DMA-buf and uncached system memory Date: Thu, 23 Jun 2022 13:32:18 +0200 [thread overview] Message-ID: <578953dd-6298-2bfe-a8fb-52004b84fd17@amd.com> (raw) In-Reply-To: <CAPj87rOykZv7bjNhHPT4StrsPz8Y_DWqab4Ryq=Qqh77LS2e=Q@mail.gmail.com> Am 23.06.22 um 13:27 schrieb Daniel Stone: > Hi Christian, > > On Thu, 23 Jun 2022 at 12:11, Christian König <christian.koenig@amd.com> wrote: >>> In fact DMA-buf sharing works fine on most of those SoCs because >>> everyone just assumes that all the accelerators don't snoop, so the >>> memory shared via DMA-buf is mostly CPU uncached. It only falls apart >>> for uses like the UVC cameras, where the shared buffer ends up being >>> CPU cached. >> Well then the existing DMA-buf framework is not what you want to use for >> this. >> >>> Non-coherent without explicit domain transfer points is just not going >>> to work. So why can't we solve the issue for DMA-buf in the same way as >>> the DMA API already solved it years ago: by adding the equivalent of >>> the dma_sync calls that do cache maintenance when necessary? On x86 (or >>> any system where things are mostly coherent) you could still no-op them >>> for the common case and only trigger cache cleaning if the importer >>> explicitly says that is going to do a non-snooping access. >> Because DMA-buf is a framework for buffer sharing between cache coherent >> devices which don't signal transitions. >> >> We intentionally didn't implemented any of the dma_sync_* functions >> because that would break the intended use case. >> >> You can of course use DMA-buf in an incoherent environment, but then you >> can't expect that this works all the time. >> >> This is documented behavior and so far we have bluntly rejected any of >> the complains that it doesn't work on most ARM SoCs and I don't really >> see a way to do this differently. > For some strange reason, 'let's do buffer sharing but make sure it > doesn't work on Arm' wasn't exactly the intention of the groups who > came together under the Linaro umbrella to create dmabuf. > > If it's really your belief that dmabuf requires universal snooping, I > recommend you send the patch to update the documentation, as well as > to remove DRIVER_PRIME from, realistically, most non-PCIE drivers. Well, to be honest I think that would indeed be necessary. What we have created are essentially two different worlds, one for PCI devices and one for the rest. This was indeed not the intention, but it's a fact that basically all DMA-buf based PCI drivers assume coherent access. Regards, Christian. > > Cheers, > Daniel
next prev parent reply other threads:[~2022-06-23 11:32 UTC|newest] Thread overview: 80+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-02-15 8:58 DMA-buf and uncached system memory Christian König 2021-02-15 8:58 ` Christian König 2021-02-15 9:06 ` Simon Ser 2021-02-15 9:06 ` Simon Ser 2021-02-15 9:34 ` Christian König 2021-02-15 9:34 ` Christian König 2021-02-15 11:53 ` Lucas Stach 2021-02-15 11:53 ` Lucas Stach 2021-02-15 12:04 ` Christian König 2021-02-15 12:04 ` Christian König 2021-02-15 12:16 ` Lucas Stach 2021-02-15 12:16 ` Lucas Stach 2021-02-15 12:25 ` Christian König 2021-02-15 12:25 ` Christian König 2021-02-15 14:41 ` David Laight 2021-02-15 14:41 ` David Laight 2021-02-15 14:54 ` [Linaro-mm-sig] " Christian König 2021-02-15 14:54 ` Christian König 2021-02-15 9:49 ` Thomas Zimmermann 2021-02-15 9:49 ` Thomas Zimmermann 2021-02-15 12:00 ` Thomas Zimmermann 2021-02-15 12:00 ` Thomas Zimmermann 2021-02-15 12:10 ` Christian König 2021-02-15 12:10 ` Christian König 2021-02-15 20:46 ` Nicolas Dufresne 2021-02-15 20:46 ` Nicolas Dufresne 2021-02-15 20:39 ` Nicolas Dufresne 2021-02-15 20:39 ` Nicolas Dufresne 2022-06-21 10:17 ` Andy.Hsieh 2022-06-21 10:34 ` Christian König 2022-06-21 15:42 ` Nicolas Dufresne 2022-06-21 15:42 ` Nicolas Dufresne 2022-06-22 9:05 ` [Linaro-mm-sig] " Christian König 2022-06-22 9:05 ` Christian König 2021-02-16 9:25 ` Daniel Vetter 2021-02-16 9:25 ` Daniel Vetter 2022-06-22 19:39 ` Nicolas Dufresne 2022-06-22 19:39 ` Nicolas Dufresne 2022-06-22 23:34 ` Daniel Stone 2022-06-22 23:34 ` Daniel Stone 2022-06-23 6:59 ` Christian König 2022-06-23 6:59 ` Christian König 2022-06-23 7:13 ` Pekka Paalanen 2022-06-23 7:13 ` Pekka Paalanen 2022-06-23 7:26 ` Christian König 2022-06-23 7:26 ` Christian König 2022-06-23 8:04 ` Lucas Stach 2022-06-23 8:14 ` Christian König 2022-06-23 8:58 ` Lucas Stach 2022-06-23 9:09 ` Christian König 2022-06-23 9:33 ` Lucas Stach 2022-06-23 9:46 ` Christian König 2022-06-23 10:13 ` Lucas Stach 2022-06-23 11:10 ` Christian König 2022-06-23 11:27 ` Daniel Stone 2022-06-23 11:27 ` Daniel Stone 2022-06-23 11:32 ` Christian König [this message] 2022-06-23 11:32 ` Christian König 2022-06-24 22:02 ` [Linaro-mm-sig] " Daniel Vetter 2022-06-24 22:02 ` Daniel Vetter 2022-07-04 13:48 ` Christian König 2022-08-09 14:46 ` Daniel Vetter 2022-08-09 14:46 ` Daniel Vetter 2022-08-10 5:55 ` Christian König 2022-06-23 11:29 ` Lucas Stach 2022-06-23 11:54 ` Christian König 2022-06-23 12:14 ` Lucas Stach 2022-06-23 12:52 ` Christian König 2022-06-23 15:26 ` Lucas Stach 2022-06-24 6:54 ` Christian König 2022-06-24 8:10 ` Lucas Stach 2022-06-27 13:54 ` Nicolas Dufresne 2022-06-27 14:06 ` Lucas Stach 2022-06-27 14:30 ` Nicolas Dufresne 2022-06-27 13:51 ` Nicolas Dufresne 2022-06-23 8:13 ` Thomas Zimmermann 2022-06-23 8:26 ` Christian König 2022-06-23 8:42 ` Thomas Zimmermann 2022-08-09 15:01 ` Rob Clark 2022-08-09 15:01 ` Rob Clark
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=578953dd-6298-2bfe-a8fb-52004b84fd17@amd.com \ --to=christian.koenig@amd.com \ --cc=Shashank.Sharma@amd.com \ --cc=daniel@fooishbar.org \ --cc=dri-devel@lists.freedesktop.org \ --cc=l.stach@pengutronix.de \ --cc=linaro-mm-sig@lists.linaro.org \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-media@vger.kernel.org \ --cc=nicolas@ndufresne.ca \ --cc=ppaalanen@gmail.com \ --cc=sumit.semwal@linaro.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: linkBe 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.