* THE SOFTWARE.
*/
-#include "sysemu/sysemu.h"
+#include "sysemu/numa.h"
#include "exec/cpu-common.h"
#include "qemu/bitmap.h"
#include "qom/cpu.h"
#include "qapi/qmp/qerror.h"
#include "hw/boards.h"
#include "sysemu/hostmem.h"
+#include "qmp-commands.h"
+#include "hw/mem/pc-dimm.h"
+#include "qemu/option.h"
+#include "qemu/config-file.h"
QemuOptsList qemu_numa_opts = {
.name = "numa",
};
static int have_memdevs = -1;
+static int max_numa_nodeid; /* Highest specified NUMA node ID, plus one.
+ * For all nodes, nodeid < max_numa_nodeid
+ */
+int nb_numa_nodes;
+NodeInfo numa_info[MAX_NODES];
static void numa_node_parse(NumaNodeOptions *node, QemuOpts *opts, Error **errp)
{
if (nodenr >= MAX_NODES) {
error_setg(errp, "Max number of NUMA nodes reached: %"
- PRIu16 "\n", nodenr);
+ PRIu16 "", nodenr);
+ return;
+ }
+
+ if (numa_info[nodenr].present) {
+ error_setg(errp, "Duplicate NUMA nodeid: %" PRIu16, nodenr);
return;
}
for (cpus = node->cpus; cpus; cpus = cpus->next) {
- if (cpus->value > MAX_CPUMASK_BITS) {
- error_setg(errp, "CPU number %" PRIu16 " is bigger than %d",
- cpus->value, MAX_CPUMASK_BITS);
+ if (cpus->value >= max_cpus) {
+ error_setg(errp,
+ "CPU index (%" PRIu16 ")"
+ " should be smaller than maxcpus (%d)",
+ cpus->value, max_cpus);
return;
}
bitmap_set(numa_info[nodenr].node_cpu, cpus->value, 1);
}
if (node->has_mem && node->has_memdev) {
- error_setg(errp, "qemu: cannot specify both mem= and memdev=\n");
+ error_setg(errp, "qemu: cannot specify both mem= and memdev=");
return;
}
}
if (node->has_memdev != have_memdevs) {
error_setg(errp, "qemu: memdev option must be specified for either "
- "all or no nodes\n");
+ "all or no nodes");
return;
}
numa_info[nodenr].node_mem = object_property_get_int(o, "size", NULL);
numa_info[nodenr].node_memdev = MEMORY_BACKEND(o);
}
+ numa_info[nodenr].present = true;
+ max_numa_nodeid = MAX(max_numa_nodeid, nodenr + 1);
}
-int numa_init_func(QemuOpts *opts, void *opaque)
+static int parse_numa(QemuOpts *opts, void *opaque)
{
NumaOptions *object = NULL;
Error *err = NULL;
return 0;
error:
- qerror_report_err(err);
- error_free(err);
+ error_report_err(err);
if (object) {
QapiDeallocVisitor *dv = qapi_dealloc_visitor_new();
return -1;
}
-void set_numa_nodes(void)
+static char *enumerate_cpus(unsigned long *cpus, int max_cpus)
+{
+ int cpu;
+ bool first = true;
+ GString *s = g_string_new(NULL);
+
+ for (cpu = find_first_bit(cpus, max_cpus);
+ cpu < max_cpus;
+ cpu = find_next_bit(cpus, max_cpus, cpu + 1)) {
+ g_string_append_printf(s, "%s%d", first ? "" : " ", cpu);
+ first = false;
+ }
+ return g_string_free(s, FALSE);
+}
+
+static void validate_numa_cpus(void)
{
+ int i;
+ DECLARE_BITMAP(seen_cpus, MAX_CPUMASK_BITS);
+
+ bitmap_zero(seen_cpus, MAX_CPUMASK_BITS);
+ for (i = 0; i < nb_numa_nodes; i++) {
+ if (bitmap_intersects(seen_cpus, numa_info[i].node_cpu,
+ MAX_CPUMASK_BITS)) {
+ bitmap_and(seen_cpus, seen_cpus,
+ numa_info[i].node_cpu, MAX_CPUMASK_BITS);
+ error_report("CPU(s) present in multiple NUMA nodes: %s",
+ enumerate_cpus(seen_cpus, max_cpus));;
+ exit(EXIT_FAILURE);
+ }
+ bitmap_or(seen_cpus, seen_cpus,
+ numa_info[i].node_cpu, MAX_CPUMASK_BITS);
+ }
+
+ if (!bitmap_full(seen_cpus, max_cpus)) {
+ char *msg;
+ bitmap_complement(seen_cpus, seen_cpus, max_cpus);
+ msg = enumerate_cpus(seen_cpus, max_cpus);
+ error_report("warning: CPU(s) not present in any NUMA nodes: %s", msg);
+ error_report("warning: All CPU(s) up to maxcpus should be described "
+ "in NUMA config");
+ g_free(msg);
+ }
+}
+
+void parse_numa_opts(MachineClass *mc)
+{
+ int i;
+
+ if (qemu_opts_foreach(qemu_find_opts("numa"), parse_numa,
+ NULL, 1) != 0) {
+ exit(1);
+ }
+
+ assert(max_numa_nodeid <= MAX_NODES);
+
+ /* No support for sparse NUMA node IDs yet: */
+ for (i = max_numa_nodeid - 1; i >= 0; i--) {
+ /* Report large node IDs first, to make mistakes easier to spot */
+ if (!numa_info[i].present) {
+ error_report("numa: Node ID missing: %d", i);
+ exit(1);
+ }
+ }
+
+ /* This must be always true if all nodes are present: */
+ assert(nb_numa_nodes == max_numa_nodeid);
+
if (nb_numa_nodes > 0) {
uint64_t numa_total;
- int i;
if (nb_numa_nodes > MAX_NODES) {
nb_numa_nodes = MAX_NODES;
}
- /* If no memory size if given for any node, assume the default case
+ /* If no memory size is given for any node, assume the default case
* and distribute the available memory equally across all nodes
*/
for (i = 0; i < nb_numa_nodes; i++) {
if (i == nb_numa_nodes) {
uint64_t usedmem = 0;
- /* On Linux, the each node's border has to be 8MB aligned,
+ /* On Linux, each node's border has to be 8MB aligned,
* the final node gets the rest.
*/
for (i = 0; i < nb_numa_nodes - 1; i++) {
numa_total += numa_info[i].node_mem;
}
if (numa_total != ram_size) {
- error_report("total memory for NUMA nodes (%" PRIu64 ")"
- " should equal RAM size (" RAM_ADDR_FMT ")",
+ error_report("total memory for NUMA nodes (0x%" PRIx64 ")"
+ " should equal RAM size (0x" RAM_ADDR_FMT ")",
numa_total, ram_size);
exit(1);
}
break;
}
}
- /* assigning the VCPUs round-robin is easier to implement, guest OSes
- * must cope with this anyway, because there are BIOSes out there in
- * real machines which also use this scheme.
+ /* Historically VCPUs were assigned in round-robin order to NUMA
+ * nodes. However it causes issues with guest not handling it nice
+ * in case where cores/threads from a multicore CPU appear on
+ * different nodes. So allow boards to override default distribution
+ * rule grouping VCPUs by socket so that VCPUs from the same socket
+ * would be on the same node.
*/
if (i == nb_numa_nodes) {
for (i = 0; i < max_cpus; i++) {
- set_bit(i, numa_info[i % nb_numa_nodes].node_cpu);
+ unsigned node_id = i % nb_numa_nodes;
+ if (mc->cpu_index_to_socket_id) {
+ node_id = mc->cpu_index_to_socket_id(i) % nb_numa_nodes;
+ }
+
+ set_bit(i, numa_info[node_id].node_cpu);
}
}
+
+ validate_numa_cpus();
}
}
-void set_numa_modes(void)
+void numa_post_machine_init(void)
{
CPUState *cpu;
int i;
const char *name,
uint64_t ram_size)
{
- memory_region_init_ram(mr, owner, name, ram_size);
+ if (mem_path) {
+#ifdef __linux__
+ Error *err = NULL;
+ memory_region_init_ram_from_file(mr, owner, name, ram_size, false,
+ mem_path, &err);
+
+ /* Legacy behavior: if allocation failed, fall back to
+ * regular RAM allocation.
+ */
+ if (err) {
+ error_report_err(err);
+ memory_region_init_ram(mr, owner, name, ram_size, &error_abort);
+ }
+#else
+ fprintf(stderr, "-mem-path not supported on this host\n");
+ exit(1);
+#endif
+ } else {
+ memory_region_init_ram(mr, owner, name, ram_size, &error_abort);
+ }
vmstate_register_ram_global(mr);
}
}
MemoryRegion *seg = host_memory_backend_get_memory(backend, &local_err);
if (local_err) {
- qerror_report_err(local_err);
+ error_report_err(local_err);
+ exit(1);
+ }
+
+ if (memory_region_is_mapped(seg)) {
+ char *path = object_get_canonical_path_component(OBJECT(backend));
+ error_report("memory backend %s is used multiple times. Each "
+ "-numa option must use a different memdev value.",
+ path);
exit(1);
}
addr += size;
}
}
+
+static void numa_stat_memory_devices(uint64_t node_mem[])
+{
+ MemoryDeviceInfoList *info_list = NULL;
+ MemoryDeviceInfoList **prev = &info_list;
+ MemoryDeviceInfoList *info;
+
+ qmp_pc_dimm_device_list(qdev_get_machine(), &prev);
+ for (info = info_list; info; info = info->next) {
+ MemoryDeviceInfo *value = info->value;
+
+ if (value) {
+ switch (value->kind) {
+ case MEMORY_DEVICE_INFO_KIND_DIMM:
+ node_mem[value->dimm->node] += value->dimm->size;
+ break;
+ default:
+ break;
+ }
+ }
+ }
+ qapi_free_MemoryDeviceInfoList(info_list);
+}
+
+void query_numa_node_mem(uint64_t node_mem[])
+{
+ int i;
+
+ if (nb_numa_nodes <= 0) {
+ return;
+ }
+
+ numa_stat_memory_devices(node_mem);
+ for (i = 0; i < nb_numa_nodes; i++) {
+ node_mem[i] += numa_info[i].node_mem;
+ }
+}
+
+static int query_memdev(Object *obj, void *opaque)
+{
+ MemdevList **list = opaque;
+ MemdevList *m = NULL;
+ Error *err = NULL;
+
+ if (object_dynamic_cast(obj, TYPE_MEMORY_BACKEND)) {
+ m = g_malloc0(sizeof(*m));
+
+ m->value = g_malloc0(sizeof(*m->value));
+
+ m->value->size = object_property_get_int(obj, "size",
+ &err);
+ if (err) {
+ goto error;
+ }
+
+ m->value->merge = object_property_get_bool(obj, "merge",
+ &err);
+ if (err) {
+ goto error;
+ }
+
+ m->value->dump = object_property_get_bool(obj, "dump",
+ &err);
+ if (err) {
+ goto error;
+ }
+
+ m->value->prealloc = object_property_get_bool(obj,
+ "prealloc", &err);
+ if (err) {
+ goto error;
+ }
+
+ m->value->policy = object_property_get_enum(obj,
+ "policy",
+ HostMemPolicy_lookup,
+ &err);
+ if (err) {
+ goto error;
+ }
+
+ object_property_get_uint16List(obj, "host-nodes",
+ &m->value->host_nodes, &err);
+ if (err) {
+ goto error;
+ }
+
+ m->next = *list;
+ *list = m;
+ }
+
+ return 0;
+error:
+ g_free(m->value);
+ g_free(m);
+
+ return -1;
+}
+
+MemdevList *qmp_query_memdev(Error **errp)
+{
+ Object *obj;
+ MemdevList *list = NULL;
+
+ obj = object_resolve_path("/objects", NULL);
+ if (obj == NULL) {
+ return NULL;
+ }
+
+ if (object_child_foreach(obj, query_memdev, &list) != 0) {
+ goto error;
+ }
+
+ return list;
+
+error:
+ qapi_free_MemdevList(list);
+ return NULL;
+}
projects / mirror_qemu.git / blobdiff