AMD IOMMU updates for 2.6.28-rc8
Hi Ingo,
The following changes since commit 218d11a8b071b23b76c484fd5f72a4fe3306801e:
Jonathan Corbet (1):
Fix a race condition in FASYNC handling
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux-2.6-iommu.git iommu-fixes-2.6.28
Joerg Roedel (5):
AMD IOMMU: fix iommu_map_page function
AMD IOMMU: fix loop counter in free_pagetable function AMD IOMMU: fix typo in comment
AMD IOMMU: fix WARN_ON in dma_ops unmap path
AMD IOMMU: __unmap_single: check for bad_dma_address instead of 0
arch/x86/kernel/amd_iommu.c | 13 +++++++------ 1 files changed, 7 insertions(+), 6 deletions(-)
This are some small bugfixes which were found during the implementation of the KVM support and further code review. Please pull.
Joerg
arch/x86/kernel/amd_iommu.c +7 -6
@@ -344,7 +344,7 @@ static int iommu_map(struct protection_domain *dom,u64 __pte, *pte, *page; bus_addr =PAGE_ALIGN (bus_addr); - phys_addr =PAGE_ALIGN (bus_addr); + phys_addr =PAGE_ALIGN (phys_addr);/* only support 512GB address spaces for now */ if (bus_addr >IOMMU_MAP_SIZE_L3 || !(prot &IOMMU_PROT_MASK ))
@@ -600,7 +600,7 @@ static void dma_ops_free_pagetable(struct dma_ops_domain *dma_dom)continue ; p2 =IOMMU_PTE_PAGE (p1[i]); -for (j =0 ; j <512 ; ++i) { +for (j =0 ; j <512 ; ++j) {if (!IOMMU_PTE_PRESENT (p2[j]))continue ; p3 =IOMMU_PTE_PAGE (p2[j]);
@@ -910,7 +910,7 @@ static void dma_ops_domain_unmap(struct amd_iommu *iommu,if (address >= dom->aperture_size)return ; -WARN_ON (address &0xfffULL || address > dom->aperture_size); +WARN_ON (address & ~PAGE_MASK || address >= dom->aperture_size); pte = dom->pte_pages[IOMMU_PTE_L1_INDEX (address)]; pte +=IOMMU_PTE_L0_INDEX (address);
@@ -922,8 +922,8 @@ static void dma_ops_domain_unmap(struct amd_iommu *iommu,/* * This function contains common code for mapping of a physically -* contiguous memory region into DMA address space. It is uses by all -* mapping functions provided by this IOMMU driver. +* contiguous memory region into DMA address space. It is used by all +* mapping functions provided with this IOMMU driver. * Must be called with the domain lock held. */ static dma_addr_t __map_single(struct device *dev,
@@ -983,7 +983,8 @@ static void __unmap_single(struct amd_iommu *iommu,dma_addr_t i, start;unsigned int pages; -if ((dma_addr ==0 ) || (dma_addr + size > dma_dom->aperture_size)) +if ((dma_addr == bad_dma_address) || + (dma_addr + size > dma_dom->aperture_size))return ; pages =iommu_num_pages (dma_addr, size,PAGE_SIZE );
Signature · 6 lines
| AMD Saxony Limited Liability Company & Co. KG
Operating | Wilschdorfer Landstr. 101, 01109 Dresden, Germany
System | Register Court Dresden: HRA 4896
Research | General Partner authorized to represent:
Center | AMD Saxony LLC (Wilmington, Delaware, US)
| General Manager of AMD Saxony LLC: Dr. Hans-R. Deppe, Thomas McCoyQuoted 17 lines · Joerg Roedel
Hi Ingo,
The following changes since commit
218d11a8b071b23b76c484fd5f72a4fe3306801e:
Jonathan Corbet (1):
Fix a race condition in FASYNC handlingare available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux-2.6-iommu.gitiommu-fixes-2.6.28Joerg Roedel (5):
AMD IOMMU: fixiommu_map_pagefunction
AMD IOMMU: fix loop counter infree_pagetablefunction AMD IOMMU: fix typo in comment
AMD IOMMU: fixWARN_ONindma_opsunmap path
AMD IOMMU:__unmap_single: check forbad_dma_addressinstead of 0
Quoted 5 lines
arch/x86/kernel/amd_iommu.c | 13 +++++++------ 1 files changed, 7 insertions(+), 6 deletions(-)This are some small bugfixes which were found during the implementation of the KVM support and further code review. Please pull.
pulled into tip/x86/urgent, thanks Joerg!
Ingo