AMD IOMMU: make dma_ops_free_pagetable generic

[firefly-linux-kernel-4.4.55.git] / arch / x86 / kernel / amd_iommu.c

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index e4899e0e878740726bfa7ea56e655c53a6b88f37..8a0fd3d099735a072c7c936d60502af39d4c6d8a 100644 (file)
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -24,6 +24,7 @@
 #include <linux/iommu-helper.h>
 #include <asm/proto.h>
 #include <asm/iommu.h>
+#include <asm/gart.h>
 #include <asm/amd_iommu_types.h>
 #include <asm/amd_iommu.h>
 
@@ -187,6 +188,8 @@ static int iommu_queue_command(struct amd_iommu *iommu, struct iommu_cmd *cmd)
 
        spin_lock_irqsave(&iommu->lock, flags);
        ret = __iommu_queue_command(iommu, cmd);
+       if (!ret)
+               iommu->need_sync = 1;
        spin_unlock_irqrestore(&iommu->lock, flags);
 
        return ret;
@@ -210,10 +213,13 @@ static int iommu_completion_wait(struct amd_iommu *iommu)
        cmd.data[0] = CMD_COMPL_WAIT_INT_MASK;
        CMD_SET_TYPE(&cmd, CMD_COMPL_WAIT);
 
-       iommu->need_sync = 0;
-
        spin_lock_irqsave(&iommu->lock, flags);
 
+       if (!iommu->need_sync)
+               goto out;
+
+       iommu->need_sync = 0;
+
        ret = __iommu_queue_command(iommu, &cmd);
 
        if (ret)
@@ -230,8 +236,9 @@ static int iommu_completion_wait(struct amd_iommu *iommu)
        status &= ~MMIO_STATUS_COM_WAIT_INT_MASK;
        writel(status, iommu->mmio_base + MMIO_STATUS_OFFSET);
 
-       if (unlikely((i == EXIT_LOOP_COUNT) && printk_ratelimit()))
-               printk(KERN_WARNING "AMD IOMMU: Completion wait loop failed\n");
+       if (unlikely(i == EXIT_LOOP_COUNT))
+               panic("AMD IOMMU: Completion wait loop failed\n");
+
 out:
        spin_unlock_irqrestore(&iommu->lock, flags);
 
@@ -254,8 +261,6 @@ static int iommu_queue_inv_dev_entry(struct amd_iommu *iommu, u16 devid)
 
        ret = iommu_queue_command(iommu, &cmd);
 
-       iommu->need_sync = 1;
-
        return ret;
 }
 
@@ -281,8 +286,6 @@ static int iommu_queue_inv_iommu_pages(struct amd_iommu *iommu,
 
        ret = iommu_queue_command(iommu, &cmd);
 
-       iommu->need_sync = 1;
-
        return ret;
 }
 
@@ -335,15 +338,15 @@ static void iommu_flush_tlb(struct amd_iommu *iommu, u16 domid)
  * supporting all features of AMD IOMMU page tables like level skipping
  * and full 64 bit address spaces.
  */
-static int iommu_map(struct protection_domain *dom,
-                    unsigned long bus_addr,
-                    unsigned long phys_addr,
-                    int prot)
+static int iommu_map_page(struct protection_domain *dom,
+                         unsigned long bus_addr,
+                         unsigned long phys_addr,
+                         int prot)
 {
        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))
@@ -437,7 +440,7 @@ static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
 
        for (addr = e->address_start; addr < e->address_end;
             addr += PAGE_SIZE) {
-               ret = iommu_map(&dma_dom->domain, addr, addr, e->prot);
+               ret = iommu_map_page(&dma_dom->domain, addr, addr, e->prot);
                if (ret)
                        return ret;
                /*
@@ -584,12 +587,12 @@ static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
        iommu_area_reserve(dom->bitmap, start_page, pages);
 }
 
-static void dma_ops_free_pagetable(struct dma_ops_domain *dma_dom)
+static void free_pagetable(struct protection_domain *domain)
 {
        int i, j;
        u64 *p1, *p2, *p3;
 
-       p1 = dma_dom->domain.pt_root;
+       p1 = domain->pt_root;
 
        if (!p1)
                return;
@@ -599,7 +602,7 @@ 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]);
@@ -610,6 +613,8 @@ static void dma_ops_free_pagetable(struct dma_ops_domain *dma_dom)
        }
 
        free_page((unsigned long)p1);
+
+       domain->pt_root = NULL;
 }
 
 /*
@@ -621,7 +626,7 @@ static void dma_ops_domain_free(struct dma_ops_domain *dom)
        if (!dom)
                return;
 
-       dma_ops_free_pagetable(dom);
+       free_pagetable(&dom->domain);
 
        kfree(dom->pte_pages);
 
@@ -762,8 +767,6 @@ static void set_device_domain(struct amd_iommu *iommu,
        write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
 
        iommu_queue_inv_dev_entry(iommu, devid);
-
-       iommu->need_sync = 1;
 }
 
 /*****************************************************************************
@@ -858,6 +861,9 @@ static int get_device_resources(struct device *dev,
                print_devid(_bdf, 1);
        }
 
+       if (domain_for_device(_bdf) == NULL)
+               set_device_domain(*iommu, *domain, _bdf);
+
        return 1;
 }
 
@@ -908,7 +914,7 @@ 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);
@@ -920,8 +926,8 @@ 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,
@@ -981,7 +987,8 @@ 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);
@@ -1031,8 +1038,7 @@ static dma_addr_t map_single(struct device *dev, phys_addr_t paddr,
        if (addr == bad_dma_address)
                goto out;
 
-       if (unlikely(iommu->need_sync))
-               iommu_completion_wait(iommu);
+       iommu_completion_wait(iommu);
 
 out:
        spin_unlock_irqrestore(&domain->lock, flags);
@@ -1060,8 +1066,7 @@ static void unmap_single(struct device *dev, dma_addr_t dma_addr,
 
        __unmap_single(iommu, domain->priv, dma_addr, size, dir);
 
-       if (unlikely(iommu->need_sync))
-               iommu_completion_wait(iommu);
+       iommu_completion_wait(iommu);
 
        spin_unlock_irqrestore(&domain->lock, flags);
 }
@@ -1127,8 +1132,7 @@ static int map_sg(struct device *dev, struct scatterlist *sglist,
                        goto unmap;
        }
 
-       if (unlikely(iommu->need_sync))
-               iommu_completion_wait(iommu);
+       iommu_completion_wait(iommu);
 
 out:
        spin_unlock_irqrestore(&domain->lock, flags);
@@ -1173,8 +1177,7 @@ static void unmap_sg(struct device *dev, struct scatterlist *sglist,
                s->dma_address = s->dma_length = 0;
        }
 
-       if (unlikely(iommu->need_sync))
-               iommu_completion_wait(iommu);
+       iommu_completion_wait(iommu);
 
        spin_unlock_irqrestore(&domain->lock, flags);
 }
@@ -1225,8 +1228,7 @@ static void *alloc_coherent(struct device *dev, size_t size,
                goto out;
        }
 
-       if (unlikely(iommu->need_sync))
-               iommu_completion_wait(iommu);
+       iommu_completion_wait(iommu);
 
 out:
        spin_unlock_irqrestore(&domain->lock, flags);
@@ -1257,8 +1259,7 @@ static void free_coherent(struct device *dev, size_t size,
 
        __unmap_single(iommu, domain->priv, dma_addr, size, DMA_BIDIRECTIONAL);
 
-       if (unlikely(iommu->need_sync))
-               iommu_completion_wait(iommu);
+       iommu_completion_wait(iommu);
 
        spin_unlock_irqrestore(&domain->lock, flags);