vhost: Rework memslot filtering and fix "used_memslot" tracking

Having multiple vhost devices, some filtering out fd-less memslots and
some not, can mess up the "used_memslot" accounting. Consequently our
"free memslot" checks become unreliable and we might run out of free
memslots at runtime later.

An example sequence which can trigger a potential issue that involves
different vhost backends (vhost-kernel and vhost-user) and hotplugged
memory devices can be found at [1].

Let's make the filtering mechanism less generic and distinguish between
backends that support private memslots (without a fd) and ones that only
support shared memslots (with a fd). Track the used_memslots for both
cases separately and use the corresponding value when required.

Note: Most probably we should filter out MAP_PRIVATE fd-based RAM regions
(for example, via memory-backend-memfd,...,shared=off or as default with
 memory-backend-file) as well. When not using MAP_SHARED, it might not work
as expected. Add a TODO for now.

[1] https://lkml.kernel.org/r/fad9136f-08d3-3fd9-71a1-502069c000cf@redhat.com

Message-ID: <20230926185738.277351-2-david@redhat.com>
Fixes: 988a27754bbb ("vhost: allow backends to filter memory sections")
Cc: Tiwei Bie <tiwei.bie@intel.com>
Acked-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>

1 files changed, 47 insertions, 9 deletions

diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
index 9cfac40fde..0edd31a7e8 100644
--- a/hw/virtio/vhost.c
+++ b/hw/virtio/vhost.c
@@ -45,20 +45,33 @@
 static struct vhost_log *vhost_log;
 static struct vhost_log *vhost_log_shm;
 
+/* Memslots used by backends that support private memslots (without an fd). */
 static unsigned int used_memslots;
+
+/* Memslots used by backends that only support shared memslots (with an fd). */
+static unsigned int used_shared_memslots;
+
 static QLIST_HEAD(, vhost_dev) vhost_devices =
     QLIST_HEAD_INITIALIZER(vhost_devices);
 
 bool vhost_has_free_slot(void)
 {
-    unsigned int slots_limit = ~0U;
+    unsigned int free = UINT_MAX;
     struct vhost_dev *hdev;
 
     QLIST_FOREACH(hdev, &vhost_devices, entry) {
         unsigned int r = hdev->vhost_ops->vhost_backend_memslots_limit(hdev);
-        slots_limit = MIN(slots_limit, r);
+        unsigned int cur_free;
+
+        if (hdev->vhost_ops->vhost_backend_no_private_memslots &&
+            hdev->vhost_ops->vhost_backend_no_private_memslots(hdev)) {
+            cur_free = r - used_shared_memslots;
+        } else {
+            cur_free = r - used_memslots;
+        }
+        free = MIN(free, cur_free);
     }
-    return slots_limit > used_memslots;
+    return free > 0;
 }
 
 static void vhost_dev_sync_region(struct vhost_dev *dev,
@@ -474,8 +487,7 @@ static int vhost_verify_ring_mappings(struct vhost_dev *dev,
  * vhost_section: identify sections needed for vhost access
  *
  * We only care about RAM sections here (where virtqueue and guest
- * internals accessed by virtio might live). If we find one we still
- * allow the backend to potentially filter it out of our list.
+ * internals accessed by virtio might live).
  */
 static bool vhost_section(struct vhost_dev *dev, MemoryRegionSection *section)
 {
@@ -502,8 +514,16 @@ static bool vhost_section(struct vhost_dev *dev, MemoryRegionSection *section)
             return false;
         }
 
-        if (dev->vhost_ops->vhost_backend_mem_section_filter &&
-            !dev->vhost_ops->vhost_backend_mem_section_filter(dev, section)) {
+        /*
+         * Some backends (like vhost-user) can only handle memory regions
+         * that have an fd (can be mapped into a different process). Filter
+         * the ones without an fd out, if requested.
+         *
+         * TODO: we might have to limit to MAP_SHARED as well.
+         */
+        if (memory_region_get_fd(section->mr) < 0 &&
+            dev->vhost_ops->vhost_backend_no_private_memslots &&
+            dev->vhost_ops->vhost_backend_no_private_memslots(dev)) {
             trace_vhost_reject_section(mr->name, 2);
             return false;
         }
@@ -568,7 +588,14 @@ static void vhost_commit(MemoryListener *listener)
                        dev->n_mem_sections * sizeof dev->mem->regions[0];
     dev->mem = g_realloc(dev->mem, regions_size);
     dev->mem->nregions = dev->n_mem_sections;
-    used_memslots = dev->mem->nregions;
+
+    if (dev->vhost_ops->vhost_backend_no_private_memslots &&
+        dev->vhost_ops->vhost_backend_no_private_memslots(dev)) {
+        used_shared_memslots = dev->mem->nregions;
+    } else {
+        used_memslots = dev->mem->nregions;
+    }
+
     for (i = 0; i < dev->n_mem_sections; i++) {
         struct vhost_memory_region *cur_vmr = dev->mem->regions + i;
         struct MemoryRegionSection *mrs = dev->mem_sections + i;
@@ -1400,6 +1427,7 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,
                    VhostBackendType backend_type, uint32_t busyloop_timeout,
                    Error **errp)
 {
+    unsigned int used;
     uint64_t features;
     int i, r, n_initialized_vqs = 0;
 
@@ -1495,7 +1523,17 @@ int vhost_dev_init(struct vhost_dev *hdev, void *opaque,
     memory_listener_register(&hdev->memory_listener, &address_space_memory);
     QLIST_INSERT_HEAD(&vhost_devices, hdev, entry);
 
-    if (used_memslots > hdev->vhost_ops->vhost_backend_memslots_limit(hdev)) {
+    /*
+     * The listener we registered properly updated the corresponding counter.
+     * So we can trust that these values are accurate.
+     */
+    if (hdev->vhost_ops->vhost_backend_no_private_memslots &&
+        hdev->vhost_ops->vhost_backend_no_private_memslots(hdev)) {
+        used = used_shared_memslots;
+    } else {
+        used = used_memslots;
+    }
+    if (used > hdev->vhost_ops->vhost_backend_memslots_limit(hdev)) {
         error_setg(errp, "vhost backend memory slots limit is less"
                    " than current number of present memory slots");
         r = -EINVAL;