diff mbox series

[PROTOTYPE,6/6] vfio: Disable only RAM_BLOCK_DISCARD_T_UNCOORDINATED discards

Message ID 20200924160423.106747-7-david@redhat.com
State New
Headers show
Series virtio-mem: vfio support | expand

Commit Message

David Hildenbrand Sept. 24, 2020, 4:04 p.m. UTC 
This unlocks virtio-mem with vfio. A virtio-mem device properly notifies
about all accessible/mapped blocks inside a managed memory region -
whenever blocks become accessible and whenever blocks become inaccessible.

Note: The block size of a virtio-mem device has to be set to sane sizes,
depending on the maximum hotplug size - to not run out of vfio mappings.
The default virtio-mem block size is usually in the range of a couple of
MBs. Linux kernels (x86-64) don't support block sizes > 128MB
with an initial memory size of < 64 MB - and above that only in some
cases 2GB. The larger the blocks, the less likely that a lot of
memory can get unplugged again. The smaller the blocks, the slower
memory hot(un)plug will be.

Assume you want to hotplug 256GB - the block size would have to be at
least 8 MB (resulting in 32768 distinct mappings).

It's expected that the block size will be comparatively large when
virtio-mem is used with vfio in the future (e.g., 128MB, 1G, 2G) -
something Linux guests will have to be optimized for.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Wei Yang <richardw.yang@linux.intel.com>
Cc: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
---
 hw/vfio/common.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
diff mbox series

Patch

diff --git a/hw/vfio/common.c b/hw/vfio/common.c
index a3aaf70dd8..4d82296967 100644
--- a/hw/vfio/common.c
+++ b/hw/vfio/common.c
@@ -1392,8 +1392,12 @@  static int vfio_connect_container(VFIOGroup *group, AddressSpace *as,
      * new memory, it will not yet set ram_block_discard_set_required() and
      * therefore, neither stops us here or deals with the sudden memory
      * consumption of inflated memory.
+     *
+     * We do support discarding for memory regions where accessible pieces
+     * are coordinated via the SparseRAMNotifier.
      */
-    ret = ram_block_discard_disable(true);
+    ret = ram_block_discard_type_disable(RAM_BLOCK_DISCARD_T_UNCOORDINATED,
+                                         true);
     if (ret) {
         error_setg_errno(errp, -ret, "Cannot set discarding of RAM broken");
         return ret;
@@ -1564,7 +1568,7 @@  close_fd_exit:
     close(fd);
 
 put_space_exit:
-    ram_block_discard_disable(false);
+    ram_block_discard_type_disable(RAM_BLOCK_DISCARD_T_UNCOORDINATED, false);
     vfio_put_address_space(space);
 
     return ret;
@@ -1686,7 +1690,8 @@  void vfio_put_group(VFIOGroup *group)
     }
 
     if (!group->ram_block_discard_allowed) {
-        ram_block_discard_disable(false);
+        ram_block_discard_type_disable(RAM_BLOCK_DISCARD_T_UNCOORDINATED,
+                                       false);
     }
     vfio_kvm_device_del_group(group);
     vfio_disconnect_container(group);
@@ -1740,7 +1745,8 @@  int vfio_get_device(VFIOGroup *group, const char *name,
 
         if (!group->ram_block_discard_allowed) {
             group->ram_block_discard_allowed = true;
-            ram_block_discard_disable(false);
+            ram_block_discard_type_disable(RAM_BLOCK_DISCARD_T_UNCOORDINATED,
+                                           false);
         }
     }