vhost-iova-tree: Implement an IOVA-only tree

Creates and supports an IOVA-only tree to support a SVQ IOVA->HVA and
GPA->IOVA tree for host-only and guest-backed memory, respectively, in
the next patch.

The IOVA allocator still allocates an IOVA range but now adds this range
to the IOVA-only tree as well as to the full IOVA->HVA tree.

In the next patch, the full IOVA->HVA tree will be split into a partial
SVQ IOVA->HVA tree and a GPA->IOVA tree. The motivation behind having an
IOVA-only tree was to have a single tree that would keep track of all
allocated IOVA ranges between the partial SVQ IOVA->HVA and GPA->IOVA
trees.

Signed-off-by: Jonah Palmer <jonah.palmer@oracle.com>
Acked-by: Eugenio Pérez <eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Message-Id: <20250217144936.3589907-2-jonah.palmer@oracle.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

hw/virtio/vhost-iova-tree.c

@@ -28,6 +28,9 @@ struct VhostIOVATree {
 
     /* IOVA address to qemu memory maps. */
     IOVATree *iova_taddr_map;
+
+    /* Allocated IOVA addresses */
+    IOVATree *iova_map;
 };
 
 /**
@@ -44,6 +47,7 @@ VhostIOVATree *vhost_iova_tree_new(hwaddr iova_first, hwaddr iova_last)
     tree->iova_last = iova_last;
 
     tree->iova_taddr_map = iova_tree_new();
+    tree->iova_map = iova_tree_new();
     return tree;
 }
 
@@ -53,6 +57,7 @@ VhostIOVATree *vhost_iova_tree_new(hwaddr iova_first, hwaddr iova_last)
 void vhost_iova_tree_delete(VhostIOVATree *iova_tree)
 {
     iova_tree_destroy(iova_tree->iova_taddr_map);
+    iova_tree_destroy(iova_tree->iova_map);
     g_free(iova_tree);
 }
 
@@ -75,6 +80,7 @@ const DMAMap *vhost_iova_tree_find_iova(const VhostIOVATree *tree,
  *
  * @tree: The iova tree
  * @map: The iova map
+ * @taddr: The translated address (HVA)
  *
  * Returns:
  * - IOVA_OK if the map fits in the container
@@ -83,19 +89,26 @@ const DMAMap *vhost_iova_tree_find_iova(const VhostIOVATree *tree,
  *
  * It returns assignated iova in map->iova if return value is VHOST_DMA_MAP_OK.
  */
-int vhost_iova_tree_map_alloc(VhostIOVATree *tree, DMAMap *map)
+int vhost_iova_tree_map_alloc(VhostIOVATree *tree, DMAMap *map, hwaddr taddr)
 {
+    int ret;
+
     /* Some vhost devices do not like addr 0. Skip first page */
     hwaddr iova_first = tree->iova_first ?: qemu_real_host_page_size();
 
-    if (map->translated_addr + map->size < map->translated_addr ||
+    if (taddr + map->size < taddr || map->perm == IOMMU_NONE) {
-        map->perm == IOMMU_NONE) {
         return IOVA_ERR_INVALID;
     }
 
-    /* Allocate a node in IOVA address */
+    /* Allocate a node in the IOVA-only tree */
-    return iova_tree_alloc_map(tree->iova_taddr_map, map, iova_first,
+    ret = iova_tree_alloc_map(tree->iova_map, map, iova_first, tree->iova_last);
-                               tree->iova_last);
+    if (unlikely(ret != IOVA_OK)) {
+        return ret;
+    }
+
+    /* Insert a node in the IOVA->HVA tree */
+    map->translated_addr = taddr;
+    return iova_tree_insert(tree->iova_taddr_map, map);
 }
 
 /**
@@ -107,4 +120,5 @@ int vhost_iova_tree_map_alloc(VhostIOVATree *tree, DMAMap *map)
 void vhost_iova_tree_remove(VhostIOVATree *iova_tree, DMAMap map)
 {
     iova_tree_remove(iova_tree->iova_taddr_map, map);
+    iova_tree_remove(iova_tree->iova_map, map);
 }

hw/virtio/vhost-iova-tree.h

@@ -21,7 +21,8 @@ G_DEFINE_AUTOPTR_CLEANUP_FUNC(VhostIOVATree, vhost_iova_tree_delete);
 
 const DMAMap *vhost_iova_tree_find_iova(const VhostIOVATree *iova_tree,
                                         const DMAMap *map);
-int vhost_iova_tree_map_alloc(VhostIOVATree *iova_tree, DMAMap *map);
+int vhost_iova_tree_map_alloc(VhostIOVATree *iova_tree, DMAMap *map,
+                              hwaddr taddr);
 void vhost_iova_tree_remove(VhostIOVATree *iova_tree, DMAMap map);
 
 #endif

hw/virtio/vhost-vdpa.c

@@ -360,14 +360,20 @@ static void vhost_vdpa_listener_region_add(MemoryListener *listener,
     llsize = int128_sub(llend, int128_make64(iova));
     if (s->shadow_data) {
         int r;
+        hwaddr hw_vaddr = (hwaddr)(uintptr_t)vaddr;
 
-        mem_region.translated_addr = (hwaddr)(uintptr_t)vaddr,
         mem_region.size = int128_get64(llsize) - 1,
         mem_region.perm = IOMMU_ACCESS_FLAG(true, section->readonly),
 
-        r = vhost_iova_tree_map_alloc(s->iova_tree, &mem_region);
+        r = vhost_iova_tree_map_alloc(s->iova_tree, &mem_region, hw_vaddr);
         if (unlikely(r != IOVA_OK)) {
             error_report("Can't allocate a mapping (%d)", r);
+
+            if (mem_region.translated_addr == hw_vaddr) {
+                error_report("Insertion to IOVA->HVA tree failed");
+                /* Remove the mapping from the IOVA-only tree */
+                goto fail_map;
+            }
             goto fail;
         }
 
@@ -1142,16 +1148,23 @@ static void vhost_vdpa_svq_unmap_rings(struct vhost_dev *dev,
  *
  * @v: Vhost-vdpa device
  * @needle: The area to search iova
+ * @taddr: The translated address (HVA)
  * @errorp: Error pointer
  */
 static bool vhost_vdpa_svq_map_ring(struct vhost_vdpa *v, DMAMap *needle,
-                                    Error **errp)
+                                    hwaddr taddr, Error **errp)
 {
     int r;
 
-    r = vhost_iova_tree_map_alloc(v->shared->iova_tree, needle);
+    r = vhost_iova_tree_map_alloc(v->shared->iova_tree, needle, taddr);
     if (unlikely(r != IOVA_OK)) {
         error_setg(errp, "Cannot allocate iova (%d)", r);
+
+        if (needle->translated_addr == taddr) {
+            error_append_hint(errp, "Insertion to IOVA->HVA tree failed");
+            /* Remove the mapping from the IOVA-only tree */
+            vhost_iova_tree_remove(v->shared->iova_tree, *needle);
+        }
         return false;
     }
 
@@ -1192,11 +1205,11 @@ static bool vhost_vdpa_svq_map_rings(struct vhost_dev *dev,
     vhost_svq_get_vring_addr(svq, &svq_addr);
 
     driver_region = (DMAMap) {
-        .translated_addr = svq_addr.desc_user_addr,
         .size = driver_size - 1,
         .perm = IOMMU_RO,
     };
-    ok = vhost_vdpa_svq_map_ring(v, &driver_region, errp);
+    ok = vhost_vdpa_svq_map_ring(v, &driver_region, svq_addr.desc_user_addr,
+                                 errp);
     if (unlikely(!ok)) {
         error_prepend(errp, "Cannot create vq driver region: ");
         return false;
@@ -1206,11 +1219,11 @@ static bool vhost_vdpa_svq_map_rings(struct vhost_dev *dev,
     addr->avail_user_addr = driver_region.iova + avail_offset;
 
     device_region = (DMAMap) {
-        .translated_addr = svq_addr.used_user_addr,
         .size = device_size - 1,
         .perm = IOMMU_RW,
     };
-    ok = vhost_vdpa_svq_map_ring(v, &device_region, errp);
+    ok = vhost_vdpa_svq_map_ring(v, &device_region, svq_addr.used_user_addr,
+                                 errp);
     if (unlikely(!ok)) {
         error_prepend(errp, "Cannot create vq device region: ");
         vhost_vdpa_svq_unmap_ring(v, driver_region.translated_addr);

net/vhost-vdpa.c

@@ -510,14 +510,20 @@ static int vhost_vdpa_cvq_map_buf(struct vhost_vdpa *v, void *buf, size_t size,
                                   bool write)
 {
     DMAMap map = {};
+    hwaddr taddr = (hwaddr)(uintptr_t)buf;
     int r;
 
-    map.translated_addr = (hwaddr)(uintptr_t)buf;
     map.size = size - 1;
     map.perm = write ? IOMMU_RW : IOMMU_RO,
-    r = vhost_iova_tree_map_alloc(v->shared->iova_tree, &map);
+    r = vhost_iova_tree_map_alloc(v->shared->iova_tree, &map, taddr);
     if (unlikely(r != IOVA_OK)) {
         error_report("Cannot map injected element");
+
+        if (map.translated_addr == taddr) {
+            error_report("Insertion to IOVA->HVA tree failed");
+            /* Remove the mapping from the IOVA-only tree */
+            goto dma_map_err;
+        }
         return r;
     }