Re: [RFC] Host1x/TegraDRM UAPI (drm_tegra_submit_relocation)

From: Dmitry Osipenko <digetx@gmail.com>
To: Mikko Perttunen <cyndis@kapsi.fi>,
	Thierry Reding <thierry.reding@gmail.com>,
	Jon Hunter <jonathanh@nvidia.com>,
	David Airlie <airlied@linux.ie>, Daniel Vetter <daniel@ffwll.ch>,
	sumit.semwal@linaro.org, gustavo@padovan.org
Cc: "linux-tegra@vger.kernel.org" <linux-tegra@vger.kernel.org>,
	talho@nvidia.com, bhuntsman@nvidia.com,
	dri-devel <dri-devel@lists.freedesktop.org>
Subject: Re: [RFC] Host1x/TegraDRM UAPI (drm_tegra_submit_relocation)
Date: Fri, 26 Jun 2020 01:50:51 +0300	[thread overview]
Message-ID: <eb3cac23-cd8c-732b-684c-c2d531de00a2@gmail.com> (raw)
In-Reply-To: <8d60baf4-45e8-296a-279e-dc105966361c@kapsi.fi>

25.06.2020 12:27, Mikko Perttunen пишет:
> On 6/25/20 1:33 AM, Dmitry Osipenko wrote:
>> 23.06.2020 15:09, Mikko Perttunen пишет:
>>> struct drm_tegra_submit_relocation {
>>>          /* [in] Index of GATHER or GATHER_UPTR command in commands. */
>>>          __u32 gather_command_index;
>>>
>>>          /*
>>>           * [in] Mapping handle (obtained through CHANNEL_MAP) of the
>>> memory
>>>           *   the address of which will be patched in.
>>>           */
>>>          __u32 mapping_id;
>>>
>>>          /*
>>>           * [in] Offset in the gather that will be patched.
>>>           */
>>>          __u64 gather_offset;
>>>
>>>          /*
>>>           * [in] Offset in target buffer whose paddr/IOVA will be
>>> written
>>>           *   to the gather.
>>>           */
>>>          __u64 target_offset;
>>>
>>>          /*
>>>           * [in] Number of bits the resulting address will be logically
>>>           *   shifted right before writing to gather.
>>>           */
>>>          __u32 shift;
>>>
>>>          __u32 reserved[1];
>>> };
>>
>> We will also need read/write direction flag here for the
>> DMA-reservations set up, it will be used for the implicit BO fencing by
>> the job's scheduler.
>>
> 
> Ideally on newer chips with context isolation, we no longer know what
> DMA-BUFs are being used by the job at submit time - they would just be
> pointers after being MAPped.

The DMA-BUFs themselves shouldn't be needed, but GEMs should.

> Do you know how other GPUs deal with DMA reservations - I expect
> separate MAP and SUBMIT steps would be pretty common?

I can't instantly recall what other drivers do, could be worthwhile to
take a closer look.

> Do you have to
> pass the DMA-BUF to both steps (i.e. do IOMMU mapping during MAP, and
> manage reservations at SUBMIT)?

Yes, this sounds good to me if DMA-BUF part is replaced with a GEM.

Please see my other reply regarding the MAP IOCTL where I'm suggesting
to back mapping IDs with a GEM.
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel