2 * Dimm device for Memory Hotplug
4 * Copyright ProfitBricks GmbH 2012
5 * Copyright (C) 2014 Red Hat Inc
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2 of the License, or (at your option) any later version.
12 * This library is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with this library; if not, see <http://www.gnu.org/licenses/>
21 #include "qemu/osdep.h"
22 #include "hw/mem/pc-dimm.h"
23 #include "qapi/error.h"
24 #include "qemu/config-file.h"
25 #include "qapi/visitor.h"
26 #include "qemu/range.h"
27 #include "sysemu/numa.h"
28 #include "sysemu/kvm.h"
30 #include "hw/virtio/vhost.h"
32 typedef struct pc_dimms_capacity {
37 void pc_dimm_memory_plug(DeviceState *dev, MemoryHotplugState *hpms,
38 MemoryRegion *mr, uint64_t align, Error **errp)
41 MachineState *machine = MACHINE(qdev_get_machine());
42 PCDIMMDevice *dimm = PC_DIMM(dev);
43 Error *local_err = NULL;
44 uint64_t existing_dimms_capacity = 0;
47 addr = object_property_get_int(OBJECT(dimm), PC_DIMM_ADDR_PROP, &local_err);
52 addr = pc_dimm_get_free_addr(hpms->base,
53 memory_region_size(&hpms->mr),
54 !addr ? NULL : &addr, align,
55 memory_region_size(mr), &local_err);
60 existing_dimms_capacity = pc_existing_dimms_capacity(&local_err);
65 if (existing_dimms_capacity + memory_region_size(mr) >
66 machine->maxram_size - machine->ram_size) {
67 error_setg(&local_err, "not enough space, currently 0x%" PRIx64
68 " in use of total hot pluggable 0x" RAM_ADDR_FMT,
69 existing_dimms_capacity,
70 machine->maxram_size - machine->ram_size);
74 object_property_set_int(OBJECT(dev), addr, PC_DIMM_ADDR_PROP, &local_err);
78 trace_mhp_pc_dimm_assigned_address(addr);
80 slot = object_property_get_int(OBJECT(dev), PC_DIMM_SLOT_PROP, &local_err);
85 slot = pc_dimm_get_free_slot(slot == PC_DIMM_UNASSIGNED_SLOT ? NULL : &slot,
86 machine->ram_slots, &local_err);
90 object_property_set_int(OBJECT(dev), slot, PC_DIMM_SLOT_PROP, &local_err);
94 trace_mhp_pc_dimm_assigned_slot(slot);
96 if (kvm_enabled() && !kvm_has_free_slot(machine)) {
97 error_setg(&local_err, "hypervisor has no free memory slots left");
101 if (!vhost_has_free_slot()) {
102 error_setg(&local_err, "a used vhost backend has no free"
103 " memory slots left");
107 memory_region_add_subregion(&hpms->mr, addr - hpms->base, mr);
108 vmstate_register_ram(mr, dev);
109 numa_set_mem_node_id(addr, memory_region_size(mr), dimm->node);
112 error_propagate(errp, local_err);
115 void pc_dimm_memory_unplug(DeviceState *dev, MemoryHotplugState *hpms,
118 PCDIMMDevice *dimm = PC_DIMM(dev);
120 numa_unset_mem_node_id(dimm->addr, memory_region_size(mr), dimm->node);
121 memory_region_del_subregion(&hpms->mr, mr);
122 vmstate_unregister_ram(mr, dev);
125 static int pc_existing_dimms_capacity_internal(Object *obj, void *opaque)
127 pc_dimms_capacity *cap = opaque;
128 uint64_t *size = &cap->size;
130 if (object_dynamic_cast(obj, TYPE_PC_DIMM)) {
131 DeviceState *dev = DEVICE(obj);
134 (*size) += object_property_get_int(obj, PC_DIMM_SIZE_PROP,
138 if (cap->errp && *cap->errp) {
142 object_child_foreach(obj, pc_existing_dimms_capacity_internal, opaque);
146 uint64_t pc_existing_dimms_capacity(Error **errp)
148 pc_dimms_capacity cap;
153 pc_existing_dimms_capacity_internal(qdev_get_machine(), &cap);
157 int qmp_pc_dimm_device_list(Object *obj, void *opaque)
159 MemoryDeviceInfoList ***prev = opaque;
161 if (object_dynamic_cast(obj, TYPE_PC_DIMM)) {
162 DeviceState *dev = DEVICE(obj);
165 MemoryDeviceInfoList *elem = g_new0(MemoryDeviceInfoList, 1);
166 MemoryDeviceInfo *info = g_new0(MemoryDeviceInfo, 1);
167 PCDIMMDeviceInfo *di = g_new0(PCDIMMDeviceInfo, 1);
168 DeviceClass *dc = DEVICE_GET_CLASS(obj);
169 PCDIMMDevice *dimm = PC_DIMM(obj);
173 di->id = g_strdup(dev->id);
175 di->hotplugged = dev->hotplugged;
176 di->hotpluggable = dc->hotpluggable;
177 di->addr = dimm->addr;
178 di->slot = dimm->slot;
179 di->node = dimm->node;
180 di->size = object_property_get_int(OBJECT(dimm), PC_DIMM_SIZE_PROP,
182 di->memdev = object_get_canonical_path(OBJECT(dimm->hostmem));
184 info->u.dimm.data = di;
192 object_child_foreach(obj, qmp_pc_dimm_device_list, opaque);
196 static int pc_dimm_slot2bitmap(Object *obj, void *opaque)
198 unsigned long *bitmap = opaque;
200 if (object_dynamic_cast(obj, TYPE_PC_DIMM)) {
201 DeviceState *dev = DEVICE(obj);
202 if (dev->realized) { /* count only realized DIMMs */
203 PCDIMMDevice *d = PC_DIMM(obj);
204 set_bit(d->slot, bitmap);
208 object_child_foreach(obj, pc_dimm_slot2bitmap, opaque);
212 int pc_dimm_get_free_slot(const int *hint, int max_slots, Error **errp)
214 unsigned long *bitmap = bitmap_new(max_slots);
217 object_child_foreach(qdev_get_machine(), pc_dimm_slot2bitmap, bitmap);
219 /* check if requested slot is not occupied */
221 if (*hint >= max_slots) {
222 error_setg(errp, "invalid slot# %d, should be less than %d",
224 } else if (!test_bit(*hint, bitmap)) {
227 error_setg(errp, "slot %d is busy", *hint);
232 /* search for free slot */
233 slot = find_first_zero_bit(bitmap, max_slots);
234 if (slot == max_slots) {
235 error_setg(errp, "no free slots available");
242 static gint pc_dimm_addr_sort(gconstpointer a, gconstpointer b)
244 PCDIMMDevice *x = PC_DIMM(a);
245 PCDIMMDevice *y = PC_DIMM(b);
246 Int128 diff = int128_sub(int128_make64(x->addr), int128_make64(y->addr));
248 if (int128_lt(diff, int128_zero())) {
250 } else if (int128_gt(diff, int128_zero())) {
256 static int pc_dimm_built_list(Object *obj, void *opaque)
258 GSList **list = opaque;
260 if (object_dynamic_cast(obj, TYPE_PC_DIMM)) {
261 DeviceState *dev = DEVICE(obj);
262 if (dev->realized) { /* only realized DIMMs matter */
263 *list = g_slist_insert_sorted(*list, dev, pc_dimm_addr_sort);
267 object_child_foreach(obj, pc_dimm_built_list, opaque);
271 uint64_t pc_dimm_get_free_addr(uint64_t address_space_start,
272 uint64_t address_space_size,
273 uint64_t *hint, uint64_t align, uint64_t size,
276 GSList *list = NULL, *item;
277 uint64_t new_addr, ret = 0;
278 uint64_t address_space_end = address_space_start + address_space_size;
280 g_assert(QEMU_ALIGN_UP(address_space_start, align) == address_space_start);
282 if (!address_space_size) {
283 error_setg(errp, "memory hotplug is not enabled, "
284 "please add maxmem option");
288 if (hint && QEMU_ALIGN_UP(*hint, align) != *hint) {
289 error_setg(errp, "address must be aligned to 0x%" PRIx64 " bytes",
294 if (QEMU_ALIGN_UP(size, align) != size) {
295 error_setg(errp, "backend memory size must be multiple of 0x%"
300 assert(address_space_end > address_space_start);
301 object_child_foreach(qdev_get_machine(), pc_dimm_built_list, &list);
306 new_addr = address_space_start;
309 /* find address range that will fit new DIMM */
310 for (item = list; item; item = g_slist_next(item)) {
311 PCDIMMDevice *dimm = item->data;
312 uint64_t dimm_size = object_property_get_int(OBJECT(dimm),
319 if (ranges_overlap(dimm->addr, dimm_size, new_addr, size)) {
321 DeviceState *d = DEVICE(dimm);
322 error_setg(errp, "address range conflicts with '%s'", d->id);
325 new_addr = QEMU_ALIGN_UP(dimm->addr + dimm_size, align);
330 if (new_addr < address_space_start) {
331 error_setg(errp, "can't add memory [0x%" PRIx64 ":0x%" PRIx64
332 "] at 0x%" PRIx64, new_addr, size, address_space_start);
333 } else if ((new_addr + size) > address_space_end) {
334 error_setg(errp, "can't add memory [0x%" PRIx64 ":0x%" PRIx64
335 "] beyond 0x%" PRIx64, new_addr, size, address_space_end);
343 static Property pc_dimm_properties[] = {
344 DEFINE_PROP_UINT64(PC_DIMM_ADDR_PROP, PCDIMMDevice, addr, 0),
345 DEFINE_PROP_UINT32(PC_DIMM_NODE_PROP, PCDIMMDevice, node, 0),
346 DEFINE_PROP_INT32(PC_DIMM_SLOT_PROP, PCDIMMDevice, slot,
347 PC_DIMM_UNASSIGNED_SLOT),
348 DEFINE_PROP_END_OF_LIST(),
351 static void pc_dimm_get_size(Object *obj, Visitor *v, const char *name,
352 void *opaque, Error **errp)
356 PCDIMMDevice *dimm = PC_DIMM(obj);
358 mr = host_memory_backend_get_memory(dimm->hostmem, errp);
359 value = memory_region_size(mr);
361 visit_type_int(v, name, &value, errp);
364 static void pc_dimm_check_memdev_is_busy(Object *obj, const char *name,
365 Object *val, Error **errp)
368 Error *local_err = NULL;
370 mr = host_memory_backend_get_memory(MEMORY_BACKEND(val), &local_err);
374 if (memory_region_is_mapped(mr)) {
375 char *path = object_get_canonical_path_component(val);
376 error_setg(&local_err, "can't use already busy memdev: %s", path);
379 qdev_prop_allow_set_link_before_realize(obj, name, val, &local_err);
383 error_propagate(errp, local_err);
386 static void pc_dimm_init(Object *obj)
388 PCDIMMDevice *dimm = PC_DIMM(obj);
390 object_property_add(obj, PC_DIMM_SIZE_PROP, "int", pc_dimm_get_size,
391 NULL, NULL, NULL, &error_abort);
392 object_property_add_link(obj, PC_DIMM_MEMDEV_PROP, TYPE_MEMORY_BACKEND,
393 (Object **)&dimm->hostmem,
394 pc_dimm_check_memdev_is_busy,
395 OBJ_PROP_LINK_UNREF_ON_RELEASE,
399 static void pc_dimm_realize(DeviceState *dev, Error **errp)
401 PCDIMMDevice *dimm = PC_DIMM(dev);
403 if (!dimm->hostmem) {
404 error_setg(errp, "'" PC_DIMM_MEMDEV_PROP "' property is not set");
407 if (((nb_numa_nodes > 0) && (dimm->node >= nb_numa_nodes)) ||
408 (!nb_numa_nodes && dimm->node)) {
409 error_setg(errp, "'DIMM property " PC_DIMM_NODE_PROP " has value %"
410 PRIu32 "' which exceeds the number of numa nodes: %d",
411 dimm->node, nb_numa_nodes ? nb_numa_nodes : 1);
416 static MemoryRegion *pc_dimm_get_memory_region(PCDIMMDevice *dimm)
418 return host_memory_backend_get_memory(dimm->hostmem, &error_abort);
421 static void pc_dimm_class_init(ObjectClass *oc, void *data)
423 DeviceClass *dc = DEVICE_CLASS(oc);
424 PCDIMMDeviceClass *ddc = PC_DIMM_CLASS(oc);
426 dc->realize = pc_dimm_realize;
427 dc->props = pc_dimm_properties;
428 dc->desc = "DIMM memory module";
430 ddc->get_memory_region = pc_dimm_get_memory_region;
433 static TypeInfo pc_dimm_info = {
434 .name = TYPE_PC_DIMM,
435 .parent = TYPE_DEVICE,
436 .instance_size = sizeof(PCDIMMDevice),
437 .instance_init = pc_dimm_init,
438 .class_init = pc_dimm_class_init,
439 .class_size = sizeof(PCDIMMDeviceClass),
442 static void pc_dimm_register_types(void)
444 type_register_static(&pc_dimm_info);
447 type_init(pc_dimm_register_types)