[Qemu-devel] [PATCH v3 3/7] numa: make sure that all cpus have has_node_id set if numa is enabled

[Igor Mammedov <imammedo@redhat.com>]

It fixes/add missing _PXM object for non mapped CPU (x86)
and missing fdt node (virt-arm).

It ensures that possible_cpus contains complete mapping if
numa is enabled by the time machine_init() is executed.

As result non completely mapped CPUs:
 1) appear in ACPI/fdt blobs
 2) QMP query-hotpluggable-cpus command shows bound nodes for such CPUs
 3) allows to drop checks for has_node_id in numa only code,
   reducing number of invariants incomplete mapping could produce
 4) moves fixup/implicit node init from runtime numa_cpu_pre_plug()
   (when CPU object is created) to machine_numa_finish_init() which
   helps to fix [1, 2] and make possible_cpus complete source
   of numa mapping available even before CPUs are created.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
---
 hw/arm/virt-acpi-build.c |  4 +---
 hw/core/machine.c        | 16 ++++++++++------
 hw/i386/acpi-build.c     |  3 +--
 hw/i386/pc.c             |  4 +---
 numa.c                   | 16 +++++-----------
 5 files changed, 18 insertions(+), 25 deletions(-)

diff --git a/hw/arm/virt-acpi-build.c b/hw/arm/virt-acpi-build.c
index e585206..977a794 100644
--- a/hw/arm/virt-acpi-build.c
+++ b/hw/arm/virt-acpi-build.c
@@ -496,12 +496,10 @@ build_srat(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms)
     srat->reserved1 = cpu_to_le32(1);
 
     for (i = 0; i < cpu_list->len; ++i) {
-        int node_id = cpu_list->cpus[i].props.has_node_id ?
-            cpu_list->cpus[i].props.node_id : 0;
         core = acpi_data_push(table_data, sizeof(*core));
         core->type = ACPI_SRAT_PROCESSOR_GICC;
         core->length = sizeof(*core);
-        core->proximity = cpu_to_le32(node_id);
+        core->proximity = cpu_to_le32(cpu_list->cpus[i].props.node_id);
         core->acpi_processor_uid = cpu_to_le32(i);
         core->flags = cpu_to_le32(1);
     }
diff --git a/hw/core/machine.c b/hw/core/machine.c
index aaf3cff..964b75d 100644
--- a/hw/core/machine.c
+++ b/hw/core/machine.c
@@ -720,19 +720,23 @@ static void machine_numa_finish_init(MachineState *machine)
         const CPUArchId *cpu_slot = &possible_cpus->cpus[i];
 
         if (!cpu_slot->props.has_node_id) {
-            if (default_mapping) {
-                /* fetch default mapping from board and enable it */
-                CpuInstanceProperties props = cpu_slot->props;
-                props.has_node_id = true;
-                machine_set_cpu_numa_node(machine, &props, &error_fatal);
-            } else {
+            /* fetch default mapping from board and enable it */
+            CpuInstanceProperties props = cpu_slot->props;
+
+            if (!default_mapping) {
                 /* record slots with not set mapping,
                  * TODO: make it hard error in future */
                 char *cpu_str = cpu_slot_to_string(cpu_slot);
                 g_string_append_printf(s, "%sCPU %d [%s]",
                                        s->len ? ", " : "", i, cpu_str);
                 g_free(cpu_str);
+
+                /* non mapped cpus used to fallback to node 0 */
+                props.node_id = 0;
             }
+
+            props.has_node_id = true;
+            machine_set_cpu_numa_node(machine, &props, &error_fatal);
         }
     }
     if (s->len) {
diff --git a/hw/i386/acpi-build.c b/hw/i386/acpi-build.c
index afcadac..873880d 100644
--- a/hw/i386/acpi-build.c
+++ b/hw/i386/acpi-build.c
@@ -2335,8 +2335,7 @@ build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
     srat->reserved1 = cpu_to_le32(1);
 
     for (i = 0; i < apic_ids->len; i++) {
-        int node_id = apic_ids->cpus[i].props.has_node_id ?
-            apic_ids->cpus[i].props.node_id : 0;
+        int node_id = apic_ids->cpus[i].props.node_id;
         uint32_t apic_id = apic_ids->cpus[i].arch_id;
 
         if (apic_id < 255) {
diff --git a/hw/i386/pc.c b/hw/i386/pc.c
index cf09949..84f0a6f 100644
--- a/hw/i386/pc.c
+++ b/hw/i386/pc.c
@@ -788,9 +788,7 @@ static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
     for (i = 0; i < cpus->len; i++) {
         unsigned int apic_id = cpus->cpus[i].arch_id;
         assert(apic_id < pcms->apic_id_limit);
-        if (cpus->cpus[i].props.has_node_id) {
-            numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
-        }
+        numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
     }
     for (i = 0; i < nb_numa_nodes; i++) {
         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
diff --git a/numa.c b/numa.c
index 45ad70a..abed45c 100644
--- a/numa.c
+++ b/numa.c
@@ -510,22 +510,16 @@ void parse_numa_opts(MachineState *ms)
 
 void numa_cpu_pre_plug(const CPUArchId *slot, DeviceState *dev, Error **errp)
 {
-    int mapped_node_id; /* set by -numa option */
     int node_id = object_property_get_int(OBJECT(dev), "node-id", &error_abort);
 
-    /* by default CPUState::numa_node was 0 if it wasn't set explicitly
-     * TODO: make it error when incomplete numa mapping support is removed
-     */
-    mapped_node_id = slot->props.node_id;
-    if (!slot->props.has_node_id) {
-         mapped_node_id = 0;
-    }
-
     if (node_id == CPU_UNSET_NUMA_NODE_ID) {
         /* due to bug in libvirt, it doesn't pass node-id from props on
          * device_add as expected, so we have to fix it up here */
-        object_property_set_int(OBJECT(dev), mapped_node_id, "node-id", errp);
-    } else if (node_id != mapped_node_id) {
+        if (slot->props.has_node_id) {
+            object_property_set_int(OBJECT(dev), slot->props.node_id,
+                                    "node-id", errp);
+        }
+    } else if (node_id != slot->props.node_id) {
         error_setg(errp, "node-id=%d must match numa node specified "
                    "with -numa option", node_id);
     }
-- 
2.7.4