From mboxrd@z Thu Jan 1 00:00:00 1970 From: Marek Szyprowski Subject: RE: [PATCH 8/8 RESEND] ARM: dma-mapping: add support for IOMMU mapper Date: Thu, 26 Jan 2012 09:09:26 +0100 Message-ID: <007a01ccdc01$d2028400$76078c00$%szyprowski@samsung.com> References: <1323448798-18184-9-git-send-email-m.szyprowski@samsung.com> <1326124161-2220-1-git-send-email-m.szyprowski@samsung.com> <20120125125916.GE1068@n2100.arm.linux.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7BIT Return-path: Received: from mailout3.w1.samsung.com ([210.118.77.13]:48558 "EHLO mailout3.w1.samsung.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751382Ab2AZIJc (ORCPT ); Thu, 26 Jan 2012 03:09:32 -0500 In-reply-to: <20120125125916.GE1068@n2100.arm.linux.org.uk> Content-language: pl Sender: linux-arch-owner@vger.kernel.org List-ID: To: 'Russell King - ARM Linux' Cc: linux-arm-kernel@lists.infradead.org, linaro-mm-sig@lists.linaro.org, linux-mm@kvack.org, linux-arch@vger.kernel.org, linux-samsung-soc@vger.kernel.org, iommu@lists.linux-foundation.org, 'Kyungmin Park' , 'Arnd Bergmann' , 'Joerg Roedel' , 'Shariq Hasnain' , 'Chunsang Jeong' , 'Krishna Reddy' , 'KyongHo Cho' , Andrzej Pietrasiewicz , 'Benjamin Herrenschmidt' Hello, On Wednesday, January 25, 2012 1:59 PM Russell King - ARM Linux wrote: > On Mon, Jan 09, 2012 at 04:49:21PM +0100, Marek Szyprowski wrote: > > This patch add a complete implementation of DMA-mapping API for > > devices that have IOMMU support. All DMA-mapping calls are supported. > > > > This patch contains some of the code kindly provided by Krishna Reddy > > and Andrzej Pietrasiewicz > > > > Signed-off-by: Marek Szyprowski > > Signed-off-by: Kyungmin Park > > > > --- > > > > Hello, > > > > This is the corrected version of the previous patch from the "[PATCH 0/8 > > v4] ARM: DMA-mapping framework redesign" thread which can be found here: > > http://www.spinics.net/lists/linux-mm/msg27382.html > > > > Previous version had very nasty bug which causes memory trashing if > > DMA-mapping managed to allocate pages larger than 4KiB. The problem was > > in __iommu_alloc_buffer() function which did not check how many pages > > has been left to allocate. > > This patch seems to be incomplete. > > If the standard DMA API is used (the one which exists in current kernels) > and NEED_SG_DMA_LENGTH is enabled, then where do we set the DMA length > in the scatterlist? Standard DMA API is also updated to work correctly with NEED_SG_DMA_LENGTH, please notice the following chunk: ----- @@ -644,6 +659,9 @@ int arm_dma_map_sg(struct device *dev, struct scatterlist *sg, int nents, int i, j; for_each_sg(sg, s, nents, i) { +#ifdef CONFIG_NEED_SG_DMA_LENGTH + s->dma_length = s->length; +#endif s->dma_address = ops->map_page(dev, sg_page(s), s->offset, s->length, dir, attrs); if (dma_mapping_error(dev, s->dma_address)) ----- (http://www.spinics.net/lists/arm-kernel/msg154889.html for the reference) > > diff --git a/arch/arm/include/asm/dma-iommu.h b/arch/arm/include/asm/dma-iommu.h > > new file mode 100644 > > index 0000000..6668b41 > > --- /dev/null > > +++ b/arch/arm/include/asm/dma-iommu.h > > @@ -0,0 +1,36 @@ > > +#ifndef ASMARM_DMA_IOMMU_H > > +#define ASMARM_DMA_IOMMU_H > > + > > +#ifdef __KERNEL__ > > + > > +#include > > +#include > > +#include > > +#include > > + > > +#include > > I can't see anything in here which needs asm/memory.h - if files which > include this need it, please include it in there so we can see why it's > needed. Ok, I will fix this issue. > > > + > > +struct dma_iommu_mapping { > > + /* iommu specific data */ > > + struct iommu_domain *domain; > > + > > + void *bitmap; > > + size_t bits; > > + unsigned int order; > > + dma_addr_t base; > > + > > + spinlock_t lock; > > + struct kref kref; > > +}; > > + > > +struct dma_iommu_mapping * > > +arm_iommu_create_mapping(struct bus_type *bus, dma_addr_t base, size_t size, > > + int order); > > + > > +void arm_iommu_release_mapping(struct dma_iommu_mapping *mapping); > > + > > +int arm_iommu_attach_device(struct device *dev, > > + struct dma_iommu_mapping *mapping); > > + > > +#endif /* __KERNEL__ */ > > +#endif > > diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c > > index 4845c09..2287b01 100644 > > --- a/arch/arm/mm/dma-mapping.c > > +++ b/arch/arm/mm/dma-mapping.c > > @@ -27,6 +27,9 @@ > > #include > > #include > > > > +#include > > linux/ includes should be grouped together. Ok, I will fix this. > > > diff --git a/arch/arm/mm/vmregion.h b/arch/arm/mm/vmregion.h > > index 15e9f04..6bbc402 100644 > > --- a/arch/arm/mm/vmregion.h > > +++ b/arch/arm/mm/vmregion.h > > @@ -17,7 +17,7 @@ struct arm_vmregion { > > struct list_head vm_list; > > unsigned long vm_start; > > unsigned long vm_end; > > - struct page *vm_pages; > > + void *priv; > > I want to think about that - I may wish to export the vm_pages via > the new dma-mappings file to provide additional information. For IOMMU case I need to store a page array for each allocated buffer. I haven't analyzed it yet, but maybe it would be possible to use standard vmalloc style entries and avoid creating separate arm_vmregion for coherent allocations? Best regards -- Marek Szyprowski Samsung Poland R&D Center