Re: [PATCH]: PCI: GART iommu alignment fixes [v2]

[FUJITA Tomonori]

On Mon, 18 Aug 2008 00:48:05 +0900
FUJITA Tomonori <fujita.tomonori@xxxxxxxxxxxxx> wrote:

> On Sun, 17 Aug 2008 17:42:13 +0200
> Ingo Molnar <mingo@xxxxxxx> wrote:
> 
> > 
> > * FUJITA Tomonori <fujita.tomonori@xxxxxxxxxxxxx> wrote:
> > 
> > > The above commit log of Prarit's patch is completely wrong (so I wrote 
> > > this patch). To avoid misunderstanding, can you apply this patch with 
> > > a proper description like this:
> > 
> > sure, i've created the commit below - is that fine?
> > 
> > 	Ingo
> 
> I think that 'v2' in the subject is unnecessary. The rest looks fine.
> 
> Thanks,
> 
> 
> > --------------->
> > From 3f18931be15856db01a216995670412e4421f3c9 Mon Sep 17 00:00:00 2001
> > From: FUJITA Tomonori <fujita.tomonori@xxxxxxxxxxxxx>
> > Date: Mon, 18 Aug 2008 00:36:18 +0900
> > Subject: [PATCH] x86 gart: allocate size-aligned address for alloc_coherent, v2
> > 
> > This patch changes GART IOMMU to return a size aligned address wrt
> > dma_alloc_coherent, as DMA-mapping.txt defines:
> > 
> > The cpu return address and the DMA bus master address are both
> > guaranteed to be aligned to the smallest PAGE_SIZE order which
> > is greater than or equal to the requested size.  This invariant
> > exists (for example) to guarantee that if you allocate a chunk
> > which is smaller than or equal to 64 kilobytes, the extent of the
> > buffer you receive will not cross a 64K boundary.

Ops, can you remove the following part? I confused
segment_boundary_mask with max_segment_size. A device needs to set
segment_boundary_mask properly if it doesn't want a buffer that
crosses a 64K boundary. We set the default boundary to 4G.


> > But it is very unlikely that this matters. As DMA-mapping.txt
> > explains, This invariant is to avoid the boundary problem (such as
> > 64K). Now the majority of IOMMUs including GART (except for Intel
> > IOMMU) don't allocate a buffer that crosses a 64K boundary wrt all the
> > DMA mapping interfaces (dma_alloc_coherent, dma_map_sg, and
> > dma_map_single) because of segment_boundary_mask in struct
> > device_dma_parameters.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/