vmm: device_manager: Tie PCI bus to NUMA node 0

Make sure the unique PCI bus is tied to the default NUMA node 0, and
update the documentation to let the users know about this special case.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>

docs/memory.md

@@ -451,3 +451,10 @@ _Example_
 --numa guest_numa_id=0,memory_zones=mem0:mem2
 --numa guest_numa_id=1,memory_zones=mem1
 ```
+
+### PCI bus
+
+Cloud Hypervisor supports only one PCI bus, which is why it has been tied to
+the NUMA node 0 by default. It is the user responsibility to organize the NUMA
+nodes correctly so that vCPUs and guest RAM which should be located on the same
+NUMA node as the PCI bus end up on the NUMA node 0.

vmm/src/device_manager.rs

@@ -3744,6 +3744,14 @@ impl Aml for DeviceManager {
         let supp = aml::Name::new("SUPP".into(), &aml::ZERO);
         pci_dsdt_inner_data.push(&supp);
 
+        // Since Cloud Hypervisor supports only one PCI bus, it can be tied
+        // to the NUMA node 0. It's up to the user to organize the NUMA nodes
+        // so that the PCI bus relates to the expected vCPUs and guest RAM.
+        let proximity_domain = 0u32;
+        let pxm_return = aml::Return::new(&proximity_domain);
+        let pxm = aml::Method::new("_PXM".into(), 0, false, vec![&pxm_return]);
+        pci_dsdt_inner_data.push(&pxm);
+
         let pci_dsm = PciDsmMethod {};
         pci_dsdt_inner_data.push(&pci_dsm);