From 75f1cdf1dda92cae037ec848ae63690d91913eac Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Tue, 10 Nov 2009 19:46:20 +0900 Subject: x86: Handle HW IOMMU initialization failure gracefully If HW IOMMU initialization fails (Intel VT-d often does this, typically due to BIOS bugs), we fall back to nommu. It doesn't work for the majority since nowadays we have more than 4GB memory so we must use swiotlb instead of nommu. The problem is that it's too late to initialize swiotlb when HW IOMMU initialization fails. We need to allocate swiotlb memory earlier from bootmem allocator. Chris explained the issue in detail: http://marc.info/?l=linux-kernel&m=125657444317079&w=2 The current x86 IOMMU initialization sequence is too complicated and handling the above issue makes it more hacky. This patch changes x86 IOMMU initialization sequence to handle the above issue cleanly. The new x86 IOMMU initialization sequence are: 1. we initialize the swiotlb (and setting swiotlb to 1) in the case of (max_pfn > MAX_DMA32_PFN && !no_iommu). dma_ops is set to swiotlb_dma_ops or nommu_dma_ops. if swiotlb usage is forced by the boot option, we finish here. 2. we call the detection functions of all the IOMMUs 3. the detection function sets x86_init.iommu.iommu_init to the IOMMU initialization function (so we can avoid calling the initialization functions of all the IOMMUs needlessly). 4. if the IOMMU initialization function doesn't need to swiotlb then sets swiotlb to zero (e.g. the initialization is sucessful). 5. if we find that swiotlb is set to zero, we free swiotlb resource. Signed-off-by: FUJITA Tomonori Cc: chrisw@sous-sol.org Cc: dwmw2@infradead.org Cc: joerg.roedel@amd.com Cc: muli@il.ibm.com LKML-Reference: <1257849980-22640-10-git-send-email-fujita.tomonori@lab.ntt.co.jp> Signed-off-by: Ingo Molnar --- arch/x86/kernel/pci-nommu.c | 9 --------- 1 file changed, 9 deletions(-) (limited to 'arch/x86/kernel/pci-nommu.c') diff --git a/arch/x86/kernel/pci-nommu.c b/arch/x86/kernel/pci-nommu.c index a3933d4330c..875e3822ae6 100644 --- a/arch/x86/kernel/pci-nommu.c +++ b/arch/x86/kernel/pci-nommu.c @@ -103,12 +103,3 @@ struct dma_map_ops nommu_dma_ops = { .sync_sg_for_device = nommu_sync_sg_for_device, .is_phys = 1, }; - -void __init no_iommu_init(void) -{ - if (dma_ops) - return; - - force_iommu = 0; /* no HW IOMMU */ - dma_ops = &nommu_dma_ops; -} -- cgit v1.2.3-70-g09d2 From 8fd524b355daef0945692227e726fb444cebcd4f Mon Sep 17 00:00:00 2001 From: FUJITA Tomonori Date: Sun, 15 Nov 2009 21:19:53 +0900 Subject: x86: Kill bad_dma_address variable This kills bad_dma_address variable, the old mechanism to enable IOMMU drivers to make dma_mapping_error() work in IOMMU's specific way. bad_dma_address variable was introduced to enable IOMMU drivers to make dma_mapping_error() work in IOMMU's specific way. However, it can't handle systems that use both swiotlb and HW IOMMU. SO we introduced dma_map_ops->mapping_error to solve that case. Intel VT-d, GART, and swiotlb already use dma_map_ops->mapping_error. Calgary, AMD IOMMU, and nommu use zero for an error dma address. This adds DMA_ERROR_CODE and converts them to use it (as SPARC and POWER does). Signed-off-by: FUJITA Tomonori Acked-by: Jesse Barnes Cc: muli@il.ibm.com Cc: joerg.roedel@amd.com LKML-Reference: <1258287594-8777-3-git-send-email-fujita.tomonori@lab.ntt.co.jp> Signed-off-by: Ingo Molnar --- arch/x86/include/asm/dma-mapping.h | 5 +++-- arch/x86/kernel/amd_iommu.c | 21 ++++++++++----------- arch/x86/kernel/pci-calgary_64.c | 22 ++++++++++------------ arch/x86/kernel/pci-dma.c | 3 --- arch/x86/kernel/pci-nommu.c | 2 +- 5 files changed, 24 insertions(+), 29 deletions(-) (limited to 'arch/x86/kernel/pci-nommu.c') diff --git a/arch/x86/include/asm/dma-mapping.h b/arch/x86/include/asm/dma-mapping.h index 6a25d5d4283..0f6c02f3b7d 100644 --- a/arch/x86/include/asm/dma-mapping.h +++ b/arch/x86/include/asm/dma-mapping.h @@ -20,7 +20,8 @@ # define ISA_DMA_BIT_MASK DMA_BIT_MASK(32) #endif -extern dma_addr_t bad_dma_address; +#define DMA_ERROR_CODE 0 + extern int iommu_merge; extern struct device x86_dma_fallback_dev; extern int panic_on_overflow; @@ -48,7 +49,7 @@ static inline int dma_mapping_error(struct device *dev, dma_addr_t dma_addr) if (ops->mapping_error) return ops->mapping_error(dev, dma_addr); - return (dma_addr == bad_dma_address); + return (dma_addr == DMA_ERROR_CODE); } #define dma_alloc_noncoherent(d, s, h, f) dma_alloc_coherent(d, s, h, f) diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c index 66237fde758..093bd526c94 100644 --- a/arch/x86/kernel/amd_iommu.c +++ b/arch/x86/kernel/amd_iommu.c @@ -928,7 +928,7 @@ static unsigned long dma_ops_alloc_addresses(struct device *dev, } if (unlikely(address == -1)) - address = bad_dma_address; + address = DMA_ERROR_CODE; WARN_ON((address + (PAGE_SIZE*pages)) > dom->aperture_size); @@ -1544,7 +1544,7 @@ static dma_addr_t dma_ops_domain_map(struct amd_iommu *iommu, pte = dma_ops_get_pte(dom, address); if (!pte) - return bad_dma_address; + return DMA_ERROR_CODE; __pte = paddr | IOMMU_PTE_P | IOMMU_PTE_FC; @@ -1625,7 +1625,7 @@ static dma_addr_t __map_single(struct device *dev, retry: address = dma_ops_alloc_addresses(dev, dma_dom, pages, align_mask, dma_mask); - if (unlikely(address == bad_dma_address)) { + if (unlikely(address == DMA_ERROR_CODE)) { /* * setting next_address here will let the address * allocator only scan the new allocated range in the @@ -1646,7 +1646,7 @@ retry: start = address; for (i = 0; i < pages; ++i) { ret = dma_ops_domain_map(iommu, dma_dom, start, paddr, dir); - if (ret == bad_dma_address) + if (ret == DMA_ERROR_CODE) goto out_unmap; paddr += PAGE_SIZE; @@ -1674,7 +1674,7 @@ out_unmap: dma_ops_free_addresses(dma_dom, address, pages); - return bad_dma_address; + return DMA_ERROR_CODE; } /* @@ -1690,7 +1690,7 @@ static void __unmap_single(struct amd_iommu *iommu, dma_addr_t i, start; unsigned int pages; - if ((dma_addr == bad_dma_address) || + if ((dma_addr == DMA_ERROR_CODE) || (dma_addr + size > dma_dom->aperture_size)) return; @@ -1732,7 +1732,7 @@ static dma_addr_t map_page(struct device *dev, struct page *page, INC_STATS_COUNTER(cnt_map_single); if (!check_device(dev)) - return bad_dma_address; + return DMA_ERROR_CODE; dma_mask = *dev->dma_mask; @@ -1743,12 +1743,12 @@ static dma_addr_t map_page(struct device *dev, struct page *page, return (dma_addr_t)paddr; if (!dma_ops_domain(domain)) - return bad_dma_address; + return DMA_ERROR_CODE; spin_lock_irqsave(&domain->lock, flags); addr = __map_single(dev, iommu, domain->priv, paddr, size, dir, false, dma_mask); - if (addr == bad_dma_address) + if (addr == DMA_ERROR_CODE) goto out; iommu_completion_wait(iommu); @@ -1957,7 +1957,7 @@ static void *alloc_coherent(struct device *dev, size_t size, *dma_addr = __map_single(dev, iommu, domain->priv, paddr, size, DMA_BIDIRECTIONAL, true, dma_mask); - if (*dma_addr == bad_dma_address) { + if (*dma_addr == DMA_ERROR_CODE) { spin_unlock_irqrestore(&domain->lock, flags); goto out_free; } @@ -2110,7 +2110,6 @@ int __init amd_iommu_init_dma_ops(void) prealloc_protection_domains(); iommu_detected = 1; - bad_dma_address = 0; swiotlb = 0; #ifdef CONFIG_GART_IOMMU gart_iommu_aperture_disabled = 1; diff --git a/arch/x86/kernel/pci-calgary_64.c b/arch/x86/kernel/pci-calgary_64.c index c84ad037f58..af9f436096a 100644 --- a/arch/x86/kernel/pci-calgary_64.c +++ b/arch/x86/kernel/pci-calgary_64.c @@ -245,7 +245,7 @@ static unsigned long iommu_range_alloc(struct device *dev, if (panic_on_overflow) panic("Calgary: fix the allocator.\n"); else - return bad_dma_address; + return DMA_ERROR_CODE; } } @@ -261,11 +261,11 @@ static dma_addr_t iommu_alloc(struct device *dev, struct iommu_table *tbl, void *vaddr, unsigned int npages, int direction) { unsigned long entry; - dma_addr_t ret = bad_dma_address; + dma_addr_t ret = DMA_ERROR_CODE; entry = iommu_range_alloc(dev, tbl, npages); - if (unlikely(entry == bad_dma_address)) + if (unlikely(entry == DMA_ERROR_CODE)) goto error; /* set the return dma address */ @@ -280,7 +280,7 @@ static dma_addr_t iommu_alloc(struct device *dev, struct iommu_table *tbl, error: printk(KERN_WARNING "Calgary: failed to allocate %u pages in " "iommu %p\n", npages, tbl); - return bad_dma_address; + return DMA_ERROR_CODE; } static void iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr, @@ -291,8 +291,8 @@ static void iommu_free(struct iommu_table *tbl, dma_addr_t dma_addr, unsigned long flags; /* were we called with bad_dma_address? */ - badend = bad_dma_address + (EMERGENCY_PAGES * PAGE_SIZE); - if (unlikely((dma_addr >= bad_dma_address) && (dma_addr < badend))) { + badend = DMA_ERROR_CODE + (EMERGENCY_PAGES * PAGE_SIZE); + if (unlikely((dma_addr >= DMA_ERROR_CODE) && (dma_addr < badend))) { WARN(1, KERN_ERR "Calgary: driver tried unmapping bad DMA " "address 0x%Lx\n", dma_addr); return; @@ -374,7 +374,7 @@ static int calgary_map_sg(struct device *dev, struct scatterlist *sg, npages = iommu_num_pages(vaddr, s->length, PAGE_SIZE); entry = iommu_range_alloc(dev, tbl, npages); - if (entry == bad_dma_address) { + if (entry == DMA_ERROR_CODE) { /* makes sure unmap knows to stop */ s->dma_length = 0; goto error; @@ -392,7 +392,7 @@ static int calgary_map_sg(struct device *dev, struct scatterlist *sg, error: calgary_unmap_sg(dev, sg, nelems, dir, NULL); for_each_sg(sg, s, nelems, i) { - sg->dma_address = bad_dma_address; + sg->dma_address = DMA_ERROR_CODE; sg->dma_length = 0; } return 0; @@ -447,7 +447,7 @@ static void* calgary_alloc_coherent(struct device *dev, size_t size, /* set up tces to cover the allocated range */ mapping = iommu_alloc(dev, tbl, ret, npages, DMA_BIDIRECTIONAL); - if (mapping == bad_dma_address) + if (mapping == DMA_ERROR_CODE) goto free; *dma_handle = mapping; return ret; @@ -728,7 +728,7 @@ static void __init calgary_reserve_regions(struct pci_dev *dev) struct iommu_table *tbl = pci_iommu(dev->bus); /* reserve EMERGENCY_PAGES from bad_dma_address and up */ - iommu_range_reserve(tbl, bad_dma_address, EMERGENCY_PAGES); + iommu_range_reserve(tbl, DMA_ERROR_CODE, EMERGENCY_PAGES); /* avoid the BIOS/VGA first 640KB-1MB region */ /* for CalIOC2 - avoid the entire first MB */ @@ -1359,8 +1359,6 @@ static int __init calgary_iommu_init(void) return ret; } - bad_dma_address = 0x0; - return 0; } diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c index bf621b9ee26..afcc58b69c7 100644 --- a/arch/x86/kernel/pci-dma.c +++ b/arch/x86/kernel/pci-dma.c @@ -43,9 +43,6 @@ int iommu_detected __read_mostly = 0; */ int iommu_pass_through __read_mostly; -dma_addr_t bad_dma_address __read_mostly = 0; -EXPORT_SYMBOL(bad_dma_address); - /* Dummy device used for NULL arguments (normally ISA). */ struct device x86_dma_fallback_dev = { .init_name = "fallback device", diff --git a/arch/x86/kernel/pci-nommu.c b/arch/x86/kernel/pci-nommu.c index 875e3822ae6..22be12b60a8 100644 --- a/arch/x86/kernel/pci-nommu.c +++ b/arch/x86/kernel/pci-nommu.c @@ -33,7 +33,7 @@ static dma_addr_t nommu_map_page(struct device *dev, struct page *page, dma_addr_t bus = page_to_phys(page) + offset; WARN_ON(size == 0); if (!check_addr("map_single", dev, bus, size)) - return bad_dma_address; + return DMA_ERROR_CODE; flush_write_buffers(); return bus; } -- cgit v1.2.3-70-g09d2