Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754301AbYHQM4o (ORCPT ); Sun, 17 Aug 2008 08:56:44 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752758AbYHQM4g (ORCPT ); Sun, 17 Aug 2008 08:56:36 -0400 Received: from mx3.mail.elte.hu ([157.181.1.138]:37943 "EHLO mx3.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752688AbYHQM4f (ORCPT ); Sun, 17 Aug 2008 08:56:35 -0400 Date: Sun, 17 Aug 2008 14:56:14 +0200 From: Ingo Molnar To: FUJITA Tomonori Cc: prarit@redhat.com, jbarnes@virtuousgeek.org, joro@8bytes.org, linux-kernel@vger.kernel.org, linux-pci@vger.kernel.org Subject: Re: [PATCH]: PCI: GART iommu alignment fixes [v2] Message-ID: <20080817125614.GA31875@elte.hu> References: <20080809125116H.fujita.tomonori@lab.ntt.co.jp> <20080815161629.GL27955@elte.hu> <20080815180048.GA26844@elte.hu> <20080816101601K.fujita.tomonori@lab.ntt.co.jp> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20080816101601K.fujita.tomonori@lab.ntt.co.jp> User-Agent: Mutt/1.5.18 (2008-05-17) X-ELTE-VirusStatus: clean X-ELTE-SpamScore: -1.5 X-ELTE-SpamLevel: X-ELTE-SpamCheck: no X-ELTE-SpamVersion: ELTE 2.0 X-ELTE-SpamCheck-Details: score=-1.5 required=5.9 tests=BAYES_00 autolearn=no SpamAssassin version=3.2.3 -1.5 BAYES_00 BODY: Bayesian spam probability is 0 to 1% [score: 0.0000] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6136 Lines: 166 * FUJITA Tomonori wrote: > > Config is: > > > > http://redhat.com/~mingo/misc/config-Fri_Aug_15_18_30_56_CEST_2008.bad > > > > Any idea why that is so? Apparently the alignment change wasnt as benign > > as assumed. > > Ah, sorry, > > @@ -262,7 +264,11 @@ static dma_addr_t dma_map_area(struct device *dev, dma_addr_t phys_mem, > static dma_addr_t > gart_map_simple(struct device *dev, phys_addr_t paddr, size_t size, int dir) > { > - dma_addr_t map = dma_map_area(dev, paddr, size, dir); > + dma_addr_t map; > + unsigned long align_mask; > + > + align_mask = (roundup_pow_of_two(size) >> PAGE_SHIFT) - 1; > + map = dma_map_area(dev, paddr, size, dir, align_mask); > > This code doesn't work with the case size < PAGE_SIZE. > > I think that dmam_alloc_consistent in libata-core fails due to this > bug. > > Can you try this? ok - merged the commit below in tip/x86/iommu. > BTW, I think that this is not urgent stuff at all. ok - i've queued it up for v2.6.28. Ingo --------------> >From 4ae29849888f85a0cee12553995d47ec527ad049 Mon Sep 17 00:00:00 2001 From: Prarit Bhargava Date: Sat, 16 Aug 2008 10:15:32 +0900 Subject: [PATCH] x86 gart: allocate size-aligned address for alloc_coherent, v2 pci_alloc_consistent/dma_alloc_coherent does not return size aligned addresses. >From Documentation/DMA-mapping.txt: "pci_alloc_consistent returns two values: the virtual address which you can use to access it from the CPU and dma_handle which you pass to the card. 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." 1. Modify alloc_iommu to allow for an alignment mask 2. Modify pci_gart_simple to return size-aligned values. 3. Fixup the alignment calculation in the iommu-helper code. 4. Fix possible overflow in alloc_iommu's boundary_size calculation. (It is possible that alloc_iommu()'s boundary_size overflows as dma_get_seg_boundary can return 0xffffffff. In that case, further usage of boundary_size triggers a BUG_ON() in the iommu code.) End result: When allocating from IOMMU, pci_alloc_consistent/dma_alloc_coherent will now return a size aligned value. Signed-off-by: Prarit Bhargava [ FUJITA Tomonori : boot hang fix. ] Signed-off-by: Ingo Molnar --- arch/x86/kernel/pci-gart_64.c | 25 ++++++++++++++++--------- 1 files changed, 16 insertions(+), 9 deletions(-) diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c index cdab678..4d8efb0 100644 --- a/arch/x86/kernel/pci-gart_64.c +++ b/arch/x86/kernel/pci-gart_64.c @@ -82,7 +82,8 @@ AGPEXTERN __u32 *agp_gatt_table; static unsigned long next_bit; /* protected by iommu_bitmap_lock */ static int need_flush; /* global flush state. set for each gart wrap */ -static unsigned long alloc_iommu(struct device *dev, int size) +static unsigned long alloc_iommu(struct device *dev, int size, + unsigned long align_mask) { unsigned long offset, flags; unsigned long boundary_size; @@ -95,11 +96,12 @@ static unsigned long alloc_iommu(struct device *dev, int size) spin_lock_irqsave(&iommu_bitmap_lock, flags); offset = iommu_area_alloc(iommu_gart_bitmap, iommu_pages, next_bit, - size, base_index, boundary_size, 0); + size, base_index, boundary_size, align_mask); if (offset == -1) { need_flush = 1; offset = iommu_area_alloc(iommu_gart_bitmap, iommu_pages, 0, - size, base_index, boundary_size, 0); + size, base_index, boundary_size, + align_mask); } if (offset != -1) { next_bit = offset+size; @@ -236,10 +238,10 @@ nonforced_iommu(struct device *dev, unsigned long addr, size_t size) * Caller needs to check if the iommu is needed and flush. */ static dma_addr_t dma_map_area(struct device *dev, dma_addr_t phys_mem, - size_t size, int dir) + size_t size, int dir, unsigned long align_mask) { unsigned long npages = iommu_num_pages(phys_mem, size); - unsigned long iommu_page = alloc_iommu(dev, npages); + unsigned long iommu_page = alloc_iommu(dev, npages, align_mask); int i; if (iommu_page == -1) { @@ -262,7 +264,11 @@ static dma_addr_t dma_map_area(struct device *dev, dma_addr_t phys_mem, static dma_addr_t gart_map_simple(struct device *dev, phys_addr_t paddr, size_t size, int dir) { - dma_addr_t map = dma_map_area(dev, paddr, size, dir); + dma_addr_t map; + unsigned long align_mask; + + align_mask = (1UL << get_order(size)) - 1; + map = dma_map_area(dev, paddr, size, dir, align_mask); flush_gart(); @@ -281,7 +287,8 @@ gart_map_single(struct device *dev, phys_addr_t paddr, size_t size, int dir) if (!need_iommu(dev, paddr, size)) return paddr; - bus = gart_map_simple(dev, paddr, size, dir); + bus = dma_map_area(dev, paddr, size, dir, 0); + flush_gart(); return bus; } @@ -340,7 +347,7 @@ static int dma_map_sg_nonforce(struct device *dev, struct scatterlist *sg, unsigned long addr = sg_phys(s); if (nonforced_iommu(dev, addr, s->length)) { - addr = dma_map_area(dev, addr, s->length, dir); + addr = dma_map_area(dev, addr, s->length, dir, 0); if (addr == bad_dma_address) { if (i > 0) gart_unmap_sg(dev, sg, i, dir); @@ -362,7 +369,7 @@ static int __dma_map_cont(struct device *dev, struct scatterlist *start, int nelems, struct scatterlist *sout, unsigned long pages) { - unsigned long iommu_start = alloc_iommu(dev, pages); + unsigned long iommu_start = alloc_iommu(dev, pages, 0); unsigned long iommu_page = iommu_start; struct scatterlist *s; int i; -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/