From 5de6bd4c23babc6a00fcd7dc7c90705c9d0ec69e Mon Sep 17 00:00:00 2001 From: Chuanxiao Dong Date: Thu, 9 Feb 2017 11:37:11 +0800 Subject: [PATCH] drm/i915/gvt: Map shadow page before using it in shadow page table MFN usually refers to "Machine Frame Number" in virtulization world. Currently GVT-g populates the shadow PPGTT/GGTT page table with MFN according to the translation between MFN and Guest PFN. When host IOMMU is enabled, GPU DMA transactions go through the IOMMU, GPU needs an IOVA<->MFN mapping to walk the shadow page table in host main memory. This patch will map a page in IOMMU page table before using it in shadow page table and release the map when a shadow page is going to be freed. Signed-off-by: Zhi Wang Signed-off-by: Chuanxiao Dong Signed-off-by: Zhenyu Wang --- drivers/gpu/drm/i915/gvt/gtt.c | 70 ++++++++++++++++++++++++---------- 1 file changed, 50 insertions(+), 20 deletions(-) diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c index 47dec4acf7ff..28c92346db0e 100644 --- a/drivers/gpu/drm/i915/gvt/gtt.c +++ b/drivers/gpu/drm/i915/gvt/gtt.c @@ -606,21 +606,33 @@ struct intel_vgpu_guest_page *intel_vgpu_find_guest_page( static inline int init_shadow_page(struct intel_vgpu *vgpu, struct intel_vgpu_shadow_page *p, int type) { + struct device *kdev = &vgpu->gvt->dev_priv->drm.pdev->dev; + dma_addr_t daddr; + + daddr = dma_map_page(kdev, p->page, 0, 4096, PCI_DMA_BIDIRECTIONAL); + if (dma_mapping_error(kdev, daddr)) { + gvt_err("fail to map dma addr\n"); + return -EINVAL; + } + p->vaddr = page_address(p->page); p->type = type; INIT_HLIST_NODE(&p->node); - p->mfn = intel_gvt_hypervisor_virt_to_mfn(p->vaddr); - if (p->mfn == INTEL_GVT_INVALID_ADDR) - return -EFAULT; - + p->mfn = daddr >> GTT_PAGE_SHIFT; hash_add(vgpu->gtt.shadow_page_hash_table, &p->node, p->mfn); return 0; } -static inline void clean_shadow_page(struct intel_vgpu_shadow_page *p) +static inline void clean_shadow_page(struct intel_vgpu *vgpu, + struct intel_vgpu_shadow_page *p) { + struct device *kdev = &vgpu->gvt->dev_priv->drm.pdev->dev; + + dma_unmap_page(kdev, p->mfn << GTT_PAGE_SHIFT, 4096, + PCI_DMA_BIDIRECTIONAL); + if (!hlist_unhashed(&p->node)) hash_del(&p->node); } @@ -670,7 +682,7 @@ static void ppgtt_free_shadow_page(struct intel_vgpu_ppgtt_spt *spt) { trace_spt_free(spt->vgpu->id, spt, spt->shadow_page.type); - clean_shadow_page(&spt->shadow_page); + clean_shadow_page(spt->vgpu, &spt->shadow_page); intel_vgpu_clean_guest_page(spt->vgpu, &spt->guest_page); list_del_init(&spt->post_shadow_list); @@ -1875,8 +1887,9 @@ static int alloc_scratch_pages(struct intel_vgpu *vgpu, int page_entry_num = GTT_PAGE_SIZE >> vgpu->gvt->device_info.gtt_entry_size_shift; void *scratch_pt; - unsigned long mfn; int i; + struct device *dev = &vgpu->gvt->dev_priv->drm.pdev->dev; + dma_addr_t daddr; if (WARN_ON(type < GTT_TYPE_PPGTT_PTE_PT || type >= GTT_TYPE_MAX)) return -EINVAL; @@ -1887,16 +1900,18 @@ static int alloc_scratch_pages(struct intel_vgpu *vgpu, return -ENOMEM; } - mfn = intel_gvt_hypervisor_virt_to_mfn(scratch_pt); - if (mfn == INTEL_GVT_INVALID_ADDR) { - gvt_err("fail to translate vaddr:0x%lx\n", (unsigned long)scratch_pt); - free_page((unsigned long)scratch_pt); - return -EFAULT; + daddr = dma_map_page(dev, virt_to_page(scratch_pt), 0, + 4096, PCI_DMA_BIDIRECTIONAL); + if (dma_mapping_error(dev, daddr)) { + gvt_err("fail to dmamap scratch_pt\n"); + __free_page(virt_to_page(scratch_pt)); + return -ENOMEM; } - gtt->scratch_pt[type].page_mfn = mfn; + gtt->scratch_pt[type].page_mfn = + (unsigned long)(daddr >> GTT_PAGE_SHIFT); gtt->scratch_pt[type].page = virt_to_page(scratch_pt); gvt_dbg_mm("vgpu%d create scratch_pt: type %d mfn=0x%lx\n", - vgpu->id, type, mfn); + vgpu->id, type, gtt->scratch_pt[type].page_mfn); /* Build the tree by full filled the scratch pt with the entries which * point to the next level scratch pt or scratch page. The @@ -1930,9 +1945,14 @@ static int alloc_scratch_pages(struct intel_vgpu *vgpu, static int release_scratch_page_tree(struct intel_vgpu *vgpu) { int i; + struct device *dev = &vgpu->gvt->dev_priv->drm.pdev->dev; + dma_addr_t daddr; for (i = GTT_TYPE_PPGTT_PTE_PT; i < GTT_TYPE_MAX; i++) { if (vgpu->gtt.scratch_pt[i].page != NULL) { + daddr = (dma_addr_t)(vgpu->gtt.scratch_pt[i].page_mfn << + GTT_PAGE_SHIFT); + dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL); __free_page(vgpu->gtt.scratch_pt[i].page); vgpu->gtt.scratch_pt[i].page = NULL; vgpu->gtt.scratch_pt[i].page_mfn = 0; @@ -2192,6 +2212,8 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt) { int ret; void *page; + struct device *dev = &gvt->dev_priv->drm.pdev->dev; + dma_addr_t daddr; gvt_dbg_core("init gtt\n"); @@ -2209,14 +2231,16 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt) gvt_err("fail to allocate scratch ggtt page\n"); return -ENOMEM; } - gvt->gtt.scratch_ggtt_page = virt_to_page(page); - gvt->gtt.scratch_ggtt_mfn = intel_gvt_hypervisor_virt_to_mfn(page); - if (gvt->gtt.scratch_ggtt_mfn == INTEL_GVT_INVALID_ADDR) { - gvt_err("fail to translate scratch ggtt page\n"); - __free_page(gvt->gtt.scratch_ggtt_page); - return -EFAULT; + daddr = dma_map_page(dev, virt_to_page(page), 0, + 4096, PCI_DMA_BIDIRECTIONAL); + if (dma_mapping_error(dev, daddr)) { + gvt_err("fail to dmamap scratch ggtt page\n"); + __free_page(virt_to_page(page)); + return -ENOMEM; } + gvt->gtt.scratch_ggtt_page = virt_to_page(page); + gvt->gtt.scratch_ggtt_mfn = (unsigned long)(daddr >> GTT_PAGE_SHIFT); if (enable_out_of_sync) { ret = setup_spt_oos(gvt); @@ -2239,6 +2263,12 @@ int intel_gvt_init_gtt(struct intel_gvt *gvt) */ void intel_gvt_clean_gtt(struct intel_gvt *gvt) { + struct device *dev = &gvt->dev_priv->drm.pdev->dev; + dma_addr_t daddr = (dma_addr_t)(gvt->gtt.scratch_ggtt_mfn << + GTT_PAGE_SHIFT); + + dma_unmap_page(dev, daddr, 4096, PCI_DMA_BIDIRECTIONAL); + __free_page(gvt->gtt.scratch_ggtt_page); if (enable_out_of_sync)