Re: [PATCH -tip 1/3] IA64: fix swiotlb alloc_coherent for non DMA_64BIT_MASK devices

[Yasunori Goto <y-goto@jp.fujitsu.com>]

> On Fri, 30 Jan 2009 07:54:07 +0900
> Yasunori Goto <y-goto@jp.fujitsu.com> wrote:
> 
> > > On Thu, 29 Jan 2009 20:41:07 +0900
> > > Yasunori Goto <y-goto@jp.fujitsu.com> wrote:
> > > 
> > > > > Before the dma ops unification, IA64 always uses GFP_DMA for
> > > > > dma_alloc_coherent like:
> > > > > 
> > > > > #define dma_alloc_coherent(dev, size, handle, gfp)	\
> > > > > 	platform_dma_alloc_coherent(dev, size, handle, (gfp) | GFP_DMA)
> > > > > 
> > > > > This GFP_DMA enforcement doesn't make sense for IOMMUs since they can
> > > > > do address translation to give addresses that devices can access
> > > > > to. The IOMMU drivers ignore the zone flag. However, this is still
> > > > > necessary for swiotlb since it can't do address translation.
> > > > > 
> > > > > We don't always need to use GFP_DMA for swiotlb. We need GFP_DMA for
> > > > > devices incapable of 64bit DMA.
> > > > > 
> > > > > This patch is sorta updated version of:
> > > > > 
> > > > > http://marc.info/?l=linux-kernel&m=122638215612705&w=2
> > > > > 
> > > > > Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
> > > > > Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
> > > > > ---
> > > > >  arch/ia64/kernel/pci-swiotlb.c |   10 +++++++++-
> > > > >  1 files changed, 9 insertions(+), 1 deletions(-)
> > > > > 
> > > > > diff --git a/arch/ia64/kernel/pci-swiotlb.c b/arch/ia64/kernel/pci-swiotlb.c
> > > > > index 717ad4f..573f02c 100644
> > > > > --- a/arch/ia64/kernel/pci-swiotlb.c
> > > > > +++ b/arch/ia64/kernel/pci-swiotlb.c
> > > > > @@ -13,8 +13,16 @@
> > > > >  int swiotlb __read_mostly;
> > > > >  EXPORT_SYMBOL(swiotlb);
> > > > >  
> > > > > +static void *ia64_swiotlb_alloc_coherent(struct device *dev, size_t size,
> > > > > +					 dma_addr_t *dma_handle, gfp_t gfp)
> > > > > +{
> > > > > +	if (dev->coherent_dma_mask != DMA_64BIT_MASK)
> > > > > +		gfp |= GFP_DMA;
> > > > > +	return swiotlb_alloc_coherent(dev, size, dma_handle, gfp);
> > > > > +}
> > > > > +
> > > > >  struct dma_map_ops swiotlb_dma_ops = {
> > > > > -	.alloc_coherent = swiotlb_alloc_coherent,
> > > > > +	.alloc_coherent = ia64_swiotlb_alloc_coherent,
> > > > >  	.free_coherent = swiotlb_free_coherent,
> > > > >  	.map_page = swiotlb_map_page,
> > > > >  	.unmap_page = swiotlb_unmap_page,
> > > > > -- 
> > > > > 1.6.0.6
> > > > 
> > > > 
> > > > Though I may be misunderstanding something, this initialization seems to be
> > > > called only when CONFIG_IA64_DIG_VTD is on.
> > > 
> > > Hmm, what's 'this initialization'?
> > > 
> > > This patch always sets swiotlb_dma_ops.alloc_coherent to
> > > ia64_swiotlb_alloc_coherent.
> > 
> > I tested your patch, but, ia64_swiotlb_alloc_coherent() was not called with
> > CONFIG_GENERIC.
> 
> You use the generic kernel on a DIG box, right? If so, yeah, there is
> something wrong.
> 
> 
> > swiotlb_dma_ops is set by pci_swiotlb_init().
> > However, it seems to be not called when CONFIG_IA64_DIG/Generic configuration.
> 
> You are talking about setting dma_ops to swiotlb_dma_ops?
> 
> I think that swiotlb_dma_init() also sets dma_ops to swiotlb_dma_ops.

Certainly.
I overlooked the 7th patch of your unifying dma_ops patch set.
ia64_swiotlb_alloc_coherent was called when I tested again.
I'm sorry.

However, GFP_DMA is still always on in ia64_swiotlb_alloc_coherent()
regardless of the value of dev->coherent_dma_mask.

After following patch applied, GFP_DMA is set correctly by your patch.
Am I still overlooking other your patch?

Bye.



---
 arch/ia64/include/asm/dma-mapping.h |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: dma_fix_test/arch/ia64/include/asm/dma-mapping.h
===================================================================
--- dma_fix_test.orig/arch/ia64/include/asm/dma-mapping.h       2009-01-30 19:36:33.000000000 +0900
+++ dma_fix_test/arch/ia64/include/asm/dma-mapping.h    2009-01-30 20:20:06.000000000 +0900
@@ -71,7 +71,7 @@ static inline void *dma_alloc_coherent(s
                                       dma_addr_t *daddr, gfp_t gfp)
 {
        struct dma_mapping_ops *ops = platform_dma_get_ops(dev);
-       return ops->alloc_coherent(dev, size, daddr, gfp | GFP_DMA);
+       return ops->alloc_coherent(dev, size, daddr, gfp);
 }

 static inline void dma_free_coherent(struct device *dev, size_t size,


-- 
Yasunori Goto