drm/i915/kvmgt: Support setting dma map for huge pages

To support huge gtt, we need to support huge pages in kvmgt first.
This patch adds a 'size' param to the intel_gvt_mpt::dma_map_guest_page
API and implements it in kvmgt.

v2: rebase.

Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c
index 54c221dedfe8c1105ee374de6f8b2466899130ae..e26c01da2bd6924842cc78475b0d62b8b1bb89e7 100644 (file)
--- a/drivers/gpu/drm/i915/gvt/gtt.c
+++ b/drivers/gpu/drm/i915/gvt/gtt.c
@@ -1106,7 +1106,7 @@ static int split_64KB_gtt_entry(struct intel_vgpu *vgpu,
 
        for (i = 0; i < GTT_64K_PTE_STRIDE; i++) {
                ret = intel_gvt_hypervisor_dma_map_guest_page(vgpu,
-                                               start_gfn + i, &dma_addr);
+                                       start_gfn + i, PAGE_SIZE, &dma_addr);
                if (ret)
                        return ret;
 
@@ -1152,7 +1152,7 @@ static int ppgtt_populate_shadow_entry(struct intel_vgpu *vgpu,
        };
 
        /* direct shadow */
-       ret = intel_gvt_hypervisor_dma_map_guest_page(vgpu, gfn, &dma_addr);
+       ret = intel_gvt_hypervisor_dma_map_guest_page(vgpu, gfn, PAGE_SIZE, &dma_addr);
        if (ret)
                return -ENXIO;
 
@@ -2080,7 +2080,7 @@ static int emulate_ggtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
                }
 
                ret = intel_gvt_hypervisor_dma_map_guest_page(vgpu, gfn,
-                                                             &dma_addr);
+                                                       PAGE_SIZE, &dma_addr);
                if (ret) {
                        gvt_vgpu_err("fail to populate guest ggtt entry\n");
                        /* guest driver may read/write the entry when partial

diff --git a/drivers/gpu/drm/i915/gvt/hypercall.h b/drivers/gpu/drm/i915/gvt/hypercall.h
index f6dd9f71788834eb388f2703425999c6acedbb04..5af11cf1b48235c46079f376686112c5234abbfd 100644 (file)
--- a/drivers/gpu/drm/i915/gvt/hypercall.h
+++ b/drivers/gpu/drm/i915/gvt/hypercall.h
@@ -53,7 +53,7 @@ struct intel_gvt_mpt {
        unsigned long (*gfn_to_mfn)(unsigned long handle, unsigned long gfn);
 
        int (*dma_map_guest_page)(unsigned long handle, unsigned long gfn,
-                                 dma_addr_t *dma_addr);
+                                 unsigned long size, dma_addr_t *dma_addr);
        void (*dma_unmap_guest_page)(unsigned long handle, dma_addr_t dma_addr);
 
        int (*map_gfn_to_mfn)(unsigned long handle, unsigned long gfn,

diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
index 1466d8769ec9facfc73c9ae4db922d0e09944840..685cb3de6daba0350011fa0982a5e3c4f7f83296 100644 (file)
--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -94,6 +94,7 @@ struct gvt_dma {
        struct rb_node dma_addr_node;
        gfn_t gfn;
        dma_addr_t dma_addr;
+       unsigned long size;
        struct kref ref;
 };
 
@@ -106,45 +107,103 @@ static int kvmgt_guest_init(struct mdev_device *mdev);
 static void intel_vgpu_release_work(struct work_struct *work);
 static bool kvmgt_guest_exit(struct kvmgt_guest_info *info);
 
+static void gvt_unpin_guest_page(struct intel_vgpu *vgpu, unsigned long gfn,
+               unsigned long size)
+{
+       int total_pages;
+       int npage;
+       int ret;
+
+       total_pages = roundup(size, PAGE_SIZE) / PAGE_SIZE;
+
+       for (npage = 0; npage < total_pages; npage++) {
+               unsigned long cur_gfn = gfn + npage;
+
+               ret = vfio_unpin_pages(mdev_dev(vgpu->vdev.mdev), &cur_gfn, 1);
+               WARN_ON(ret != 1);
+       }
+}
+
+/* Pin a normal or compound guest page for dma. */
+static int gvt_pin_guest_page(struct intel_vgpu *vgpu, unsigned long gfn,
+               unsigned long size, struct page **page)
+{
+       unsigned long base_pfn = 0;
+       int total_pages;
+       int npage;
+       int ret;
+
+       total_pages = roundup(size, PAGE_SIZE) / PAGE_SIZE;
+       /*
+        * We pin the pages one-by-one to avoid allocating a big arrary
+        * on stack to hold pfns.
+        */
+       for (npage = 0; npage < total_pages; npage++) {
+               unsigned long cur_gfn = gfn + npage;
+               unsigned long pfn;
+
+               ret = vfio_pin_pages(mdev_dev(vgpu->vdev.mdev), &cur_gfn, 1,
+                                    IOMMU_READ | IOMMU_WRITE, &pfn);
+               if (ret != 1) {
+                       gvt_vgpu_err("vfio_pin_pages failed for gfn 0x%lx, ret %d\n",
+                                    cur_gfn, ret);
+                       goto err;
+               }
+
+               if (!pfn_valid(pfn)) {
+                       gvt_vgpu_err("pfn 0x%lx is not mem backed\n", pfn);
+                       npage++;
+                       ret = -EFAULT;
+                       goto err;
+               }
+
+               if (npage == 0)
+                       base_pfn = pfn;
+               else if (base_pfn + npage != pfn) {
+                       gvt_vgpu_err("The pages are not continuous\n");
+                       ret = -EINVAL;
+                       npage++;
+                       goto err;
+               }
+       }
+
+       *page = pfn_to_page(base_pfn);
+       return 0;
+err:
+       gvt_unpin_guest_page(vgpu, gfn, npage * PAGE_SIZE);
+       return ret;
+}
+
 static int gvt_dma_map_page(struct intel_vgpu *vgpu, unsigned long gfn,
-               dma_addr_t *dma_addr)
+               dma_addr_t *dma_addr, unsigned long size)
 {
        struct device *dev = &vgpu->gvt->dev_priv->drm.pdev->dev;
-       struct page *page;
-       unsigned long pfn;
+       struct page *page = NULL;
        int ret;
 
-       /* Pin the page first. */
-       ret = vfio_pin_pages(mdev_dev(vgpu->vdev.mdev), &gfn, 1,
-                            IOMMU_READ | IOMMU_WRITE, &pfn);
-       if (ret != 1) {
-               gvt_vgpu_err("vfio_pin_pages failed for gfn 0x%lx: %d\n",
-                            gfn, ret);
-               return -EINVAL;
-       }
+       ret = gvt_pin_guest_page(vgpu, gfn, size, &page);
+       if (ret)
+               return ret;
 
        /* Setup DMA mapping. */
-       page = pfn_to_page(pfn);
-       *dma_addr = dma_map_page(dev, page, 0, PAGE_SIZE,
-                                PCI_DMA_BIDIRECTIONAL);
-       if (dma_mapping_error(dev, *dma_addr)) {
-               gvt_vgpu_err("DMA mapping failed for gfn 0x%lx\n", gfn);
-               vfio_unpin_pages(mdev_dev(vgpu->vdev.mdev), &gfn, 1);
-               return -ENOMEM;
+       *dma_addr = dma_map_page(dev, page, 0, size, PCI_DMA_BIDIRECTIONAL);
+       ret = dma_mapping_error(dev, *dma_addr);
+       if (ret) {
+               gvt_vgpu_err("DMA mapping failed for pfn 0x%lx, ret %d\n",
+                            page_to_pfn(page), ret);
+               gvt_unpin_guest_page(vgpu, gfn, size);
        }
 
-       return 0;
+       return ret;
 }
 
 static void gvt_dma_unmap_page(struct intel_vgpu *vgpu, unsigned long gfn,
-               dma_addr_t dma_addr)
+               dma_addr_t dma_addr, unsigned long size)
 {
        struct device *dev = &vgpu->gvt->dev_priv->drm.pdev->dev;
-       int ret;
 
-       dma_unmap_page(dev, dma_addr, PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
-       ret = vfio_unpin_pages(mdev_dev(vgpu->vdev.mdev), &gfn, 1);
-       WARN_ON(ret != 1);
+       dma_unmap_page(dev, dma_addr, size, PCI_DMA_BIDIRECTIONAL);
+       gvt_unpin_guest_page(vgpu, gfn, size);
 }
 
 static struct gvt_dma *__gvt_cache_find_dma_addr(struct intel_vgpu *vgpu,
@@ -185,7 +244,7 @@ static struct gvt_dma *__gvt_cache_find_gfn(struct intel_vgpu *vgpu, gfn_t gfn)
 }
 
 static int __gvt_cache_add(struct intel_vgpu *vgpu, gfn_t gfn,
-               dma_addr_t dma_addr)
+               dma_addr_t dma_addr, unsigned long size)
 {
        struct gvt_dma *new, *itr;
        struct rb_node **link, *parent = NULL;
@@ -197,6 +256,7 @@ static int __gvt_cache_add(struct intel_vgpu *vgpu, gfn_t gfn,
        new->vgpu = vgpu;
        new->gfn = gfn;
        new->dma_addr = dma_addr;
+       new->size = size;
        kref_init(&new->ref);
 
        /* gfn_cache maps gfn to struct gvt_dma. */
@@ -254,7 +314,7 @@ static void gvt_cache_destroy(struct intel_vgpu *vgpu)
                        break;
                }
                dma = rb_entry(node, struct gvt_dma, gfn_node);
-               gvt_dma_unmap_page(vgpu, dma->gfn, dma->dma_addr);
+               gvt_dma_unmap_page(vgpu, dma->gfn, dma->dma_addr, dma->size);
                __gvt_cache_remove_entry(vgpu, dma);
                mutex_unlock(&vgpu->vdev.cache_lock);
        }
@@ -509,7 +569,8 @@ static int intel_vgpu_iommu_notifier(struct notifier_block *nb,
                        if (!entry)
                                continue;
 
-                       gvt_dma_unmap_page(vgpu, entry->gfn, entry->dma_addr);
+                       gvt_dma_unmap_page(vgpu, entry->gfn, entry->dma_addr,
+                                          entry->size);
                        __gvt_cache_remove_entry(vgpu, entry);
                }
                mutex_unlock(&vgpu->vdev.cache_lock);
@@ -1616,7 +1677,7 @@ static unsigned long kvmgt_gfn_to_pfn(unsigned long handle, unsigned long gfn)
 }
 
 int kvmgt_dma_map_guest_page(unsigned long handle, unsigned long gfn,
-               dma_addr_t *dma_addr)
+               unsigned long size, dma_addr_t *dma_addr)
 {
        struct kvmgt_guest_info *info;
        struct intel_vgpu *vgpu;
@@ -1633,11 +1694,11 @@ int kvmgt_dma_map_guest_page(unsigned long handle, unsigned long gfn,
 
        entry = __gvt_cache_find_gfn(info->vgpu, gfn);
        if (!entry) {
-               ret = gvt_dma_map_page(vgpu, gfn, dma_addr);
+               ret = gvt_dma_map_page(vgpu, gfn, dma_addr, size);
                if (ret)
                        goto err_unlock;
 
-               ret = __gvt_cache_add(info->vgpu, gfn, *dma_addr);
+               ret = __gvt_cache_add(info->vgpu, gfn, *dma_addr, size);
                if (ret)
                        goto err_unmap;
        } else {
@@ -1649,7 +1710,7 @@ int kvmgt_dma_map_guest_page(unsigned long handle, unsigned long gfn,
        return 0;
 
 err_unmap:
-       gvt_dma_unmap_page(vgpu, gfn, *dma_addr);
+       gvt_dma_unmap_page(vgpu, gfn, *dma_addr, size);
 err_unlock:
        mutex_unlock(&info->vgpu->vdev.cache_lock);
        return ret;
@@ -1659,7 +1720,8 @@ static void __gvt_dma_release(struct kref *ref)
 {
        struct gvt_dma *entry = container_of(ref, typeof(*entry), ref);
 
-       gvt_dma_unmap_page(entry->vgpu, entry->gfn, entry->dma_addr);
+       gvt_dma_unmap_page(entry->vgpu, entry->gfn, entry->dma_addr,
+                          entry->size);
        __gvt_cache_remove_entry(entry->vgpu, entry);
 }
 

diff --git a/drivers/gpu/drm/i915/gvt/mpt.h b/drivers/gpu/drm/i915/gvt/mpt.h
index 32ffcd566cddf8965a53d0534682a50fab8594d4..67f19992b226f29a13d408be36da8b4820095b13 100644 (file)
--- a/drivers/gpu/drm/i915/gvt/mpt.h
+++ b/drivers/gpu/drm/i915/gvt/mpt.h
@@ -230,17 +230,18 @@ static inline unsigned long intel_gvt_hypervisor_gfn_to_mfn(
 /**
  * intel_gvt_hypervisor_dma_map_guest_page - setup dma map for guest page
  * @vgpu: a vGPU
- * @gpfn: guest pfn
+ * @gfn: guest pfn
+ * @size: page size
  * @dma_addr: retrieve allocated dma addr
  *
  * Returns:
  * 0 on success, negative error code if failed.
  */
 static inline int intel_gvt_hypervisor_dma_map_guest_page(
-               struct intel_vgpu *vgpu, unsigned long gfn,
+               struct intel_vgpu *vgpu, unsigned long gfn, unsigned long size,
                dma_addr_t *dma_addr)
 {
-       return intel_gvt_host.mpt->dma_map_guest_page(vgpu->handle, gfn,
+       return intel_gvt_host.mpt->dma_map_guest_page(vgpu->handle, gfn, size,
                                                      dma_addr);
 }