2 files changed, 336 insertions, 302 deletions

diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
index 4fb7a8f83c8a..8c7bb9befdab 100644
--- a/drivers/vfio/vfio_iommu_type1.c
+++ b/drivers/vfio/vfio_iommu_type1.c
@@ -30,7 +30,6 @@
 #include <linux/iommu.h>
 #include <linux/module.h>
 #include <linux/mm.h>
-#include <linux/pci.h>          /* pci_bus_type */
 #include <linux/rbtree.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
@@ -55,11 +54,17 @@ MODULE_PARM_DESC(disable_hugepages,
                  "Disable VFIO IOMMU support for IOMMU hugepages.");
 
 struct vfio_iommu {
-        struct iommu_domain     *domain;
+        struct list_head        domain_list;
         struct mutex            lock;
         struct rb_root          dma_list;
+        bool v2;
+};
+
+struct vfio_domain {
+        struct iommu_domain     *domain;
+        struct list_head        next;
         struct list_head        group_list;
-        bool                    cache;
+        int                     prot;           /* IOMMU_CACHE */
 };
 
 struct vfio_dma {
@@ -99,7 +104,7 @@ static struct vfio_dma *vfio_find_dma(struct vfio_iommu *iommu,
         return NULL;
 }
 
-static void vfio_insert_dma(struct vfio_iommu *iommu, struct vfio_dma *new)
+static void vfio_link_dma(struct vfio_iommu *iommu, struct vfio_dma *new)
 {
         struct rb_node **link = &iommu->dma_list.rb_node, *parent = NULL;
         struct vfio_dma *dma;
@@ -118,7 +123,7 @@ static void vfio_insert_dma(struct vfio_iommu *iommu, struct vfio_dma *new)
         rb_insert_color(&new->node, &iommu->dma_list);
 }
 
-static void vfio_remove_dma(struct vfio_iommu *iommu, struct vfio_dma *old)
+static void vfio_unlink_dma(struct vfio_iommu *iommu, struct vfio_dma *old)
 {
         rb_erase(&old->node, &iommu->dma_list);
 }
@@ -322,32 +327,39 @@ static long vfio_unpin_pages(unsigned long pfn, long npage,
         return unlocked;
 }
 
-static int vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma,
-                            dma_addr_t iova, size_t *size)
+static void vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma)
 {
-        dma_addr_t start = iova, end = iova + *size;
+        dma_addr_t iova = dma->iova, end = dma->iova + dma->size;
+        struct vfio_domain *domain, *d;
         long unlocked = 0;
 
+        if (!dma->size)
+                return;
+        /*
+         * We use the IOMMU to track the physical addresses, otherwise we'd
+         * need a much more complicated tracking system.  Unfortunately that
+         * means we need to use one of the iommu domains to figure out the
+         * pfns to unpin.  The rest need to be unmapped in advance so we have
+         * no iommu translations remaining when the pages are unpinned.
+         */
+        domain = d = list_first_entry(&iommu->domain_list,
+                                      struct vfio_domain, next);
+
+        list_for_each_entry_continue(d, &iommu->domain_list, next)
+                iommu_unmap(d->domain, dma->iova, dma->size);
+
         while (iova < end) {
                 size_t unmapped;
                 phys_addr_t phys;
 
-                /*
-                 * We use the IOMMU to track the physical address.  This
-                 * saves us from having a lot more entries in our mapping
-                 * tree.  The downside is that we don't track the size
-                 * used to do the mapping.  We request unmap of a single
-                 * page, but expect IOMMUs that support large pages to
-                 * unmap a larger chunk.
-                 */
-                phys = iommu_iova_to_phys(iommu->domain, iova);
+                phys = iommu_iova_to_phys(domain->domain, iova);
                 if (WARN_ON(!phys)) {
                         iova += PAGE_SIZE;
                         continue;
                 }
 
-                unmapped = iommu_unmap(iommu->domain, iova, PAGE_SIZE);
-                if (!unmapped)
+                unmapped = iommu_unmap(domain->domain, iova, PAGE_SIZE);
+                if (WARN_ON(!unmapped))
                         break;
 
                 unlocked += vfio_unpin_pages(phys >> PAGE_SHIFT,
@@ -357,119 +369,26 @@ static int vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma,
         }
 
         vfio_lock_acct(-unlocked);
-
-        *size = iova - start;
-
-        return 0;
 }
 
-static int vfio_remove_dma_overlap(struct vfio_iommu *iommu, dma_addr_t start,
-                                   size_t *size, struct vfio_dma *dma)
+static void vfio_remove_dma(struct vfio_iommu *iommu, struct vfio_dma *dma)
 {
-        size_t offset, overlap, tmp;
-        struct vfio_dma *split;
-        int ret;
-
-        if (!*size)
-                return 0;
-
-        /*
-         * Existing dma region is completely covered, unmap all.  This is
-         * the likely case since userspace tends to map and unmap buffers
-         * in one shot rather than multiple mappings within a buffer.
-         */
-        if (likely(start <= dma->iova &&
-                   start + *size >= dma->iova + dma->size)) {
-                *size = dma->size;
-                ret = vfio_unmap_unpin(iommu, dma, dma->iova, size);
-                if (ret)
-                        return ret;
-
-                /*
-                 * Did we remove more than we have?  Should never happen
-                 * since a vfio_dma is contiguous in iova and vaddr.
-                 */
-                WARN_ON(*size != dma->size);
-
-                vfio_remove_dma(iommu, dma);
-                kfree(dma);
-                return 0;
-        }
-
-        /* Overlap low address of existing range */
-        if (start <= dma->iova) {
-                overlap = start + *size - dma->iova;
-                ret = vfio_unmap_unpin(iommu, dma, dma->iova, &overlap);
-                if (ret)
-                        return ret;
-
-                vfio_remove_dma(iommu, dma);
-
-                /*
-                 * Check, we may have removed to whole vfio_dma.  If not
-                 * fixup and re-insert.
-                 */
-                if (overlap < dma->size) {
-                        dma->iova += overlap;
-                        dma->vaddr += overlap;
-                        dma->size -= overlap;
-                        vfio_insert_dma(iommu, dma);
-                } else
-                        kfree(dma);
-
-                *size = overlap;
-                return 0;
-        }
-
+        vfio_unmap_unpin(iommu, dma);
+        vfio_unlink_dma(iommu, dma);
+        kfree(dma);
+}