Re: [PATCH V2] drm/amdgpu: limit DMA size to PAGE_SIZE for scatter-gather buffers

From: Robin Murphy <robin.murphy@arm.com>
To: "Christian König" <christian.koenig@amd.com>,
	"Christoph Hellwig" <hch@lst.de>
Cc: Sinan Kaya <okaya@codeaurora.org>,
	amd-gfx@lists.freedesktop.org, timur@codeaurora.org,
	sulrich@codeaurora.org, Tom St Denis <tom.stdenis@amd.com>,
	"David (ChunMing) Zhou" <David1.Zhou@amd.com>,
	Emily Deng <Emily.Deng@amd.com>, David Airlie <airlied@linux.ie>,
	linux-arm-msm@vger.kernel.org,
	Felix Kuehling <Felix.Kuehling@amd.com>,
	open list <linux-kernel@vger.kernel.org>,
	"open list:DRM DRIVERS" <dri-devel@lists.freedesktop.org>,
	David Panariti <David.Panariti@amd.com>, Jim Qu <Jim.Qu@amd.com>,
	Huang Rui <ray.huang@amd.com>, Roger He <Hongbo.He@amd.com>,
	Monk Liu <Monk.Liu@amd.com>, Feifei Xu <Feifei.Xu@amd.com>,
	Alex Deucher <alexander.deucher@amd.com>,
	linux-arm-kernel@lists.infradead.org,
	iommu@lists.linux-foundation.org
Subject: Re: [PATCH V2] drm/amdgpu: limit DMA size to PAGE_SIZE for scatter-gather buffers
Date: Thu, 12 Apr 2018 14:51:00 +0100	[thread overview]
Message-ID: <7a3e7459-2965-54c8-ff0c-6edb3a6aecb5@arm.com> (raw)
In-Reply-To: <6802bd0c-7da6-659a-7e34-8f05d7c6422c@amd.com>

On 12/04/18 10:42, Christian König wrote:
> Am 12.04.2018 um 08:26 schrieb Christoph Hellwig:
>> On Wed, Apr 11, 2018 at 01:03:59PM +0100, Robin Murphy wrote:
>>> On 10/04/18 21:59, Sinan Kaya wrote:
>>>> Code is expecing to observe the same number of buffers returned from
>>>> dma_map_sg() function compared to sg_alloc_table_from_pages(). This
>>>> doesn't hold true universally especially for systems with IOMMU.
>>> So why not fix said code? It's clearly not a real hardware 
>>> limitation, and
>>> the map_sg() APIs have potentially returned fewer than nents since 
>>> forever,
>>> so there's really no excuse.
>> Yes, relying on dma_map_sg returning the same number of entries as passed
>> it is completely bogus.
> 
> I agree that the common DRM functions should be able to deal with both, 
> but we should let the driver side decide if it wants multiple addresses 
> combined or not.
> 
>>
>>>> IOMMU driver tries to combine buffers into a single DMA address as much
>>>> as it can. The right thing is to tell the DMA layer how much combining
>>>> IOMMU can do.
>>> Disagree; this is a dodgy hack, since you'll now end up passing
>>> scatterlists into dma_map_sg() which already violate max_seg_size to 
>>> begin
>>> with, and I think a conscientious DMA API implementation would be at 
>>> rights
>>> to fail the mapping for that reason (I know arm64 happens not to, but 
>>> that
>>> was a deliberate design decision to make my life easier at the time).
>> Agreed.
> 
> Sounds like my understanding of max_seg_size is incorrect, what exactly 
> should that describe?

The segment size and boundary mask are there to represent a device's 
hardware scatter-gather capabilities - a lot of things have limits on 
the size and alignment of the data pointed to by a single descriptor 
(e.g. an xHCI TRB, where the data buffer must not cross a 64KB boundary) 
- although they can also be relevant to non-scatter-gather devices if 
they just have limits on how big/aligned a single DMA transfer can be.

Robin.