#include "qemu/bitmap.h"
#include "qom/cpu.h"
#include "qemu/error-report.h"
-#include "include/exec/cpu-common.h" /* for RAM_ADDR_FMT */
#include "qapi-visit.h"
#include "qapi/opts-visitor.h"
#include "hw/boards.h"
#include "hw/mem/pc-dimm.h"
#include "qemu/option.h"
#include "qemu/config-file.h"
+#include "qemu/cutils.h"
QemuOptsList qemu_numa_opts = {
.name = "numa",
bool have_numa_distance;
NodeInfo numa_info[MAX_NODES];
-void numa_set_mem_node_id(ram_addr_t addr, uint64_t size, uint32_t node)
-{
- struct numa_addr_range *range;
-
- /*
- * Memory-less nodes can come here with 0 size in which case,
- * there is nothing to do.
- */
- if (!size) {
- return;
- }
-
- range = g_malloc0(sizeof(*range));
- range->mem_start = addr;
- range->mem_end = addr + size - 1;
- QLIST_INSERT_HEAD(&numa_info[node].addr, range, entry);
-}
-
-void numa_unset_mem_node_id(ram_addr_t addr, uint64_t size, uint32_t node)
-{
- struct numa_addr_range *range, *next;
-
- QLIST_FOREACH_SAFE(range, &numa_info[node].addr, entry, next) {
- if (addr == range->mem_start && (addr + size - 1) == range->mem_end) {
- QLIST_REMOVE(range, entry);
- g_free(range);
- return;
- }
- }
-}
-
-static void numa_set_mem_ranges(void)
-{
- int i;
- ram_addr_t mem_start = 0;
-
- /*
- * Deduce start address of each node and use it to store
- * the address range info in numa_info address range list
- */
- for (i = 0; i < nb_numa_nodes; i++) {
- numa_set_mem_node_id(mem_start, numa_info[i].node_mem, i);
- mem_start += numa_info[i].node_mem;
- }
-}
-
-/*
- * Check if @addr falls under NUMA @node.
- */
-static bool numa_addr_belongs_to_node(ram_addr_t addr, uint32_t node)
-{
- struct numa_addr_range *range;
-
- QLIST_FOREACH(range, &numa_info[node].addr, entry) {
- if (addr >= range->mem_start && addr <= range->mem_end) {
- return true;
- }
- }
- return false;
-}
-
-/*
- * Given an address, return the index of the NUMA node to which the
- * address belongs to.
- */
-uint32_t numa_get_node(ram_addr_t addr, Error **errp)
-{
- uint32_t i;
-
- /* For non NUMA configurations, check if the addr falls under node 0 */
- if (!nb_numa_nodes) {
- if (numa_addr_belongs_to_node(addr, 0)) {
- return 0;
- }
- }
-
- for (i = 0; i < nb_numa_nodes; i++) {
- if (numa_addr_belongs_to_node(addr, i)) {
- return i;
- }
- }
-
- error_setg(errp, "Address 0x" RAM_ADDR_FMT " doesn't belong to any "
- "NUMA node", addr);
- return -1;
-}
static void parse_numa_node(MachineState *ms, NumaNodeOptions *node,
- QemuOpts *opts, Error **errp)
+ Error **errp)
{
uint16_t nodenr;
uint16List *cpus = NULL;
}
if (node->has_mem && node->has_memdev) {
- error_setg(errp, "qemu: cannot specify both mem= and memdev=");
+ error_setg(errp, "cannot specify both mem= and memdev=");
return;
}
have_memdevs = node->has_memdev;
}
if (node->has_memdev != have_memdevs) {
- error_setg(errp, "qemu: memdev option must be specified for either "
+ error_setg(errp, "memdev option must be specified for either "
"all or no nodes");
return;
}
if (node->has_mem) {
- uint64_t mem_size = node->mem;
- const char *mem_str = qemu_opt_get(opts, "mem");
- /* Fix up legacy suffix-less format */
- if (g_ascii_isdigit(mem_str[strlen(mem_str) - 1])) {
- mem_size <<= 20;
- }
- numa_info[nodenr].node_mem = mem_size;
+ numa_info[nodenr].node_mem = node->mem;
}
if (node->has_memdev) {
Object *o;
}
object_ref(o);
- numa_info[nodenr].node_mem = object_property_get_int(o, "size", NULL);
+ numa_info[nodenr].node_mem = object_property_get_uint(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);
+ nb_numa_nodes++;
}
static void parse_numa_distance(NumaDistOptions *dist, Error **errp)
if (src >= MAX_NODES || dst >= MAX_NODES) {
error_setg(errp,
- "Invalid node %" PRIu16
- ", max possible could be %" PRIu16,
+ "Invalid node %d, max possible could be %d",
MAX(src, dst), MAX_NODES);
return;
}
goto end;
}
+ /* Fix up legacy suffix-less format */
+ if ((object->type == NUMA_OPTIONS_TYPE_NODE) && object->u.node.has_mem) {
+ const char *mem_str = qemu_opt_get(opts, "mem");
+ qemu_strtosz_MiB(mem_str, NULL, &object->u.node.mem);
+ }
+
switch (object->type) {
case NUMA_OPTIONS_TYPE_NODE:
- parse_numa_node(ms, &object->u.node, opts, &err);
+ parse_numa_node(ms, &object->u.node, &err);
if (err) {
goto end;
}
- nb_numa_nodes++;
break;
case NUMA_OPTIONS_TYPE_DIST:
parse_numa_distance(&object->u.dist, &err);
goto end;
}
break;
+ case NUMA_OPTIONS_TYPE_CPU:
+ if (!object->u.cpu.has_node_id) {
+ error_setg(&err, "Missing mandatory node-id property");
+ goto end;
+ }
+ if (!numa_info[object->u.cpu.node_id].present) {
+ error_setg(&err, "Invalid node-id=%" PRId64 ", NUMA node must be "
+ "defined with -numa node,nodeid=ID before it's used with "
+ "-numa cpu,node-id=ID", object->u.cpu.node_id);
+ goto end;
+ }
+
+ machine_set_cpu_numa_node(ms, qapi_NumaCpuOptions_base(&object->u.cpu),
+ &err);
+ break;
default:
abort();
}
void parse_numa_opts(MachineState *ms)
{
int i;
- const CPUArchIdList *possible_cpus;
MachineClass *mc = MACHINE_GET_CLASS(ms);
if (qemu_opts_foreach(qemu_find_opts("numa"), parse_numa, ms, NULL)) {
exit(1);
}
+ /*
+ * If memory hotplug is enabled (slots > 0) but without '-numa'
+ * options explicitly on CLI, guestes will break.
+ *
+ * Windows: won't enable memory hotplug without SRAT table at all
+ *
+ * Linux: if QEMU is started with initial memory all below 4Gb
+ * and no SRAT table present, guest kernel will use nommu DMA ops,
+ * which breaks 32bit hw drivers when memory is hotplugged and
+ * guest tries to use it with that drivers.
+ *
+ * Enable NUMA implicitly by adding a new NUMA node automatically.
+ */
+ if (ms->ram_slots > 0 && nb_numa_nodes == 0 &&
+ mc->auto_enable_numa_with_memhp) {
+ NumaNodeOptions node = { };
+ parse_numa_node(ms, &node, NULL);
+ }
+
assert(max_numa_nodeid <= MAX_NODES);
/* No support for sparse NUMA node IDs yet: */
exit(1);
}
- for (i = 0; i < nb_numa_nodes; i++) {
- QLIST_INIT(&numa_info[i].addr);
- }
-
- numa_set_mem_ranges();
-
- /* assign CPUs to nodes using board provided default mapping */
- if (!mc->cpu_index_to_instance_props || !mc->possible_cpu_arch_ids) {
- error_report("default CPUs to NUMA node mapping isn't supported");
- exit(1);
- }
-
- possible_cpus = mc->possible_cpu_arch_ids(ms);
- for (i = 0; i < possible_cpus->len; i++) {
- if (possible_cpus->cpus[i].props.has_node_id) {
- break;
- }
- }
-
- /* no CPUs are assigned to NUMA nodes */
- if (i == possible_cpus->len) {
- for (i = 0; i < max_cpus; i++) {
- CpuInstanceProperties props;
- /* fetch default mapping from board and enable it */
- props = mc->cpu_index_to_instance_props(ms, i);
- props.has_node_id = true;
-
- machine_set_cpu_numa_node(ms, &props, &error_fatal);
- }
- }
-
/* QEMU needs at least all unique node pair distances to build
* the whole NUMA distance table. QEMU treats the distance table
* as symmetric by default, i.e. distance A->B == distance B->A.
/* Validation succeeded, now fill in any missing distances. */
complete_init_numa_distance();
}
- } else {
- numa_set_mem_node_id(0, ram_size, 0);
+ }
+}
+
+void numa_cpu_pre_plug(const CPUArchId *slot, DeviceState *dev, Error **errp)
+{
+ int node_id = object_property_get_int(OBJECT(dev), "node-id", &error_abort);
+
+ 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 */
+ 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);
}
}
/* Legacy behavior: if allocation failed, fall back to
* regular RAM allocation.
*/
- memory_region_init_ram(mr, owner, name, ram_size, &error_fatal);
+ memory_region_init_ram_nomigrate(mr, owner, name, ram_size, &error_fatal);
}
#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_fatal);
+ memory_region_init_ram_nomigrate(mr, owner, name, ram_size, &error_fatal);
}
vmstate_register_ram_global(mr);
}
}
memory_region_init(mr, owner, name, ram_size);
- for (i = 0; i < MAX_NODES; i++) {
+ for (i = 0; i < nb_numa_nodes; i++) {
uint64_t size = numa_info[i].node_mem;
HostMemoryBackend *backend = numa_info[i].node_memdev;
if (!backend) {
}
}
-static void numa_stat_memory_devices(uint64_t node_mem[])
+static void numa_stat_memory_devices(NumaNodeMem node_mem[])
{
MemoryDeviceInfoList *info_list = NULL;
MemoryDeviceInfoList **prev = &info_list;
MemoryDeviceInfoList *info;
+ PCDIMMDeviceInfo *pcdimm_info;
qmp_pc_dimm_device_list(qdev_get_machine(), &prev);
for (info = info_list; info; info = info->next) {
if (value) {
switch (value->type) {
- case MEMORY_DEVICE_INFO_KIND_DIMM:
- node_mem[value->u.dimm.data->node] += value->u.dimm.data->size;
+ case MEMORY_DEVICE_INFO_KIND_DIMM: {
+ pcdimm_info = value->u.dimm.data;
+ node_mem[pcdimm_info->node].node_mem += pcdimm_info->size;
+ if (pcdimm_info->hotpluggable && pcdimm_info->hotplugged) {
+ node_mem[pcdimm_info->node].node_plugged_mem +=
+ pcdimm_info->size;
+ }
break;
+ }
+
default:
break;
}
qapi_free_MemoryDeviceInfoList(info_list);
}
-void query_numa_node_mem(uint64_t node_mem[])
+void query_numa_node_mem(NumaNodeMem node_mem[])
{
int i;
numa_stat_memory_devices(node_mem);
for (i = 0; i < nb_numa_nodes; i++) {
- node_mem[i] += numa_info[i].node_mem;
+ node_mem[i].node_mem += numa_info[i].node_mem;
}
}
m->value->id = object_property_get_str(obj, "id", NULL);
m->value->has_id = !!m->value->id;
- m->value->size = object_property_get_int(obj, "size",
- &error_abort);
+ m->value->size = object_property_get_uint(obj, "size",
+ &error_abort);
m->value->merge = object_property_get_bool(obj, "merge",
&error_abort);
m->value->dump = object_property_get_bool(obj, "dump",
projects / qemu.git / blobdiff