1/*
2 * Memory Device Interface
3 *
4 * Copyright ProfitBricks GmbH 2012
5 * Copyright (C) 2014 Red Hat Inc
6 * Copyright (c) 2018 Red Hat Inc
7 *
8 * This work is licensed under the terms of the GNU GPL, version 2 or later.
9 * See the COPYING file in the top-level directory.
10 */
11
12#include "qemu/osdep.h"
13#include "hw/mem/memory-device.h"
14#include "qapi/error.h"
15#include "hw/boards.h"
16#include "qemu/range.h"
17#include "hw/virtio/vhost.h"
18#include "sysemu/kvm.h"
19#include "trace.h"
20
21static gint memory_device_addr_sort(gconstpointer a, gconstpointer b)
22{
23 const MemoryDeviceState *md_a = MEMORY_DEVICE(a);
24 const MemoryDeviceState *md_b = MEMORY_DEVICE(b);
25 const MemoryDeviceClass *mdc_a = MEMORY_DEVICE_GET_CLASS(a);
26 const MemoryDeviceClass *mdc_b = MEMORY_DEVICE_GET_CLASS(b);
27 const uint64_t addr_a = mdc_a->get_addr(md_a);
28 const uint64_t addr_b = mdc_b->get_addr(md_b);
29
30 if (addr_a > addr_b) {
31 return 1;
32 } else if (addr_a < addr_b) {
33 return -1;
34 }
35 return 0;
36}
37
38static int memory_device_build_list(Object *obj, void *opaque)
39{
40 GSList **list = opaque;
41
42 if (object_dynamic_cast(obj, TYPE_MEMORY_DEVICE)) {
43 DeviceState *dev = DEVICE(obj);
44 if (dev->realized) { /* only realized memory devices matter */
45 *list = g_slist_insert_sorted(*list, dev, memory_device_addr_sort);
46 }
47 }
48
49 object_child_foreach(obj, memory_device_build_list, opaque);
50 return 0;
51}
52
53static int memory_device_used_region_size(Object *obj, void *opaque)
54{
55 uint64_t *size = opaque;
56
57 if (object_dynamic_cast(obj, TYPE_MEMORY_DEVICE)) {
58 const DeviceState *dev = DEVICE(obj);
59 const MemoryDeviceState *md = MEMORY_DEVICE(obj);
60
61 if (dev->realized) {
62 *size += memory_device_get_region_size(md, &error_abort);
63 }
64 }
65
66 object_child_foreach(obj, memory_device_used_region_size, opaque);
67 return 0;
68}
69
70static void memory_device_check_addable(MachineState *ms, uint64_t size,
71 Error **errp)
72{
73 uint64_t used_region_size = 0;
74
75 /* we will need a new memory slot for kvm and vhost */
76 if (kvm_enabled() && !kvm_has_free_slot(ms)) {
77 error_setg(errp, "hypervisor has no free memory slots left");
78 return;
79 }
80 if (!vhost_has_free_slot()) {
81 error_setg(errp, "a used vhost backend has no free memory slots left");
82 return;
83 }
84
85 /* will we exceed the total amount of memory specified */
86 memory_device_used_region_size(OBJECT(ms), &used_region_size);
87 if (used_region_size + size < used_region_size ||
88 used_region_size + size > ms->maxram_size - ms->ram_size) {
89 error_setg(errp, "not enough space, currently 0x%" PRIx64
90 " in use of total space for memory devices 0x" RAM_ADDR_FMT,
91 used_region_size, ms->maxram_size - ms->ram_size);
92 return;
93 }
94
95}
96
97static uint64_t memory_device_get_free_addr(MachineState *ms,
98 const uint64_t *hint,
99 uint64_t align, uint64_t size,
100 Error **errp)
101{
102 GSList *list = NULL, *item;
103 Range as, new = range_empty;
104
105 if (!ms->device_memory) {
106 error_setg(errp, "memory devices (e.g. for memory hotplug) are not "
107 "supported by the machine");
108 return 0;
109 }
110
111 if (!memory_region_size(&ms->device_memory->mr)) {
112 error_setg(errp, "memory devices (e.g. for memory hotplug) are not "
113 "enabled, please specify the maxmem option");
114 return 0;
115 }
116 range_init_nofail(&as, ms->device_memory->base,
117 memory_region_size(&ms->device_memory->mr));
118
119 /* start of address space indicates the maximum alignment we expect */
120 if (!QEMU_IS_ALIGNED(range_lob(&as), align)) {
121 error_setg(errp, "the alignment (0x%" PRIx64 ") is not supported",
122 align);
123 return 0;
124 }
125
126 memory_device_check_addable(ms, size, errp);
127 if (*errp) {
128 return 0;
129 }
130
131 if (hint && !QEMU_IS_ALIGNED(*hint, align)) {
132 error_setg(errp, "address must be aligned to 0x%" PRIx64 " bytes",
133 align);
134 return 0;
135 }
136
137 if (!QEMU_IS_ALIGNED(size, align)) {
138 error_setg(errp, "backend memory size must be multiple of 0x%"
139 PRIx64, align);
140 return 0;
141 }
142
143 if (hint) {
144 if (range_init(&new, *hint, size) || !range_contains_range(&as, &new)) {
145 error_setg(errp, "can't add memory device [0x%" PRIx64 ":0x%" PRIx64
146 "], usable range for memory devices [0x%" PRIx64 ":0x%"
147 PRIx64 "]", *hint, size, range_lob(&as),
148 range_size(&as));
149 return 0;
150 }
151 } else {
152 if (range_init(&new, range_lob(&as), size)) {
153 error_setg(errp, "can't add memory device, device too big");
154 return 0;
155 }
156 }
157
158 /* find address range that will fit new memory device */
159 object_child_foreach(OBJECT(ms), memory_device_build_list, &list);
160 for (item = list; item; item = g_slist_next(item)) {
161 const MemoryDeviceState *md = item->data;
162 const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(OBJECT(md));
163 uint64_t next_addr;
164 Range tmp;
165
166 range_init_nofail(&tmp, mdc->get_addr(md),
167 memory_device_get_region_size(md, &error_abort));
168
169 if (range_overlaps_range(&tmp, &new)) {
170 if (hint) {
171 const DeviceState *d = DEVICE(md);
172 error_setg(errp, "address range conflicts with memory device"
173 " id='%s'", d->id ? d->id : "(unnamed)");
174 goto out;
175 }
176
177 next_addr = QEMU_ALIGN_UP(range_upb(&tmp) + 1, align);
178 if (!next_addr || range_init(&new, next_addr, range_size(&new))) {
179 range_make_empty(&new);
180 break;
181 }
182 }
183 }
184
185 if (!range_contains_range(&as, &new)) {
186 error_setg(errp, "could not find position in guest address space for "
187 "memory device - memory fragmented due to alignments");
188 goto out;
189 }
190out:
191 g_slist_free(list);
192 return range_lob(&new);
193}
194
195MemoryDeviceInfoList *qmp_memory_device_list(void)
196{
197 GSList *devices = NULL, *item;
198 MemoryDeviceInfoList *list = NULL, *prev = NULL;
199
200 object_child_foreach(qdev_get_machine(), memory_device_build_list,
201 &devices);
202
203 for (item = devices; item; item = g_slist_next(item)) {
204 const MemoryDeviceState *md = MEMORY_DEVICE(item->data);
205 const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(item->data);
206 MemoryDeviceInfoList *elem = g_new0(MemoryDeviceInfoList, 1);
207 MemoryDeviceInfo *info = g_new0(MemoryDeviceInfo, 1);
208
209 mdc->fill_device_info(md, info);
210
211 elem->value = info;
212 elem->next = NULL;
213 if (prev) {
214 prev->next = elem;
215 } else {
216 list = elem;
217 }
218 prev = elem;
219 }
220
221 g_slist_free(devices);
222
223 return list;
224}
225
226static int memory_device_plugged_size(Object *obj, void *opaque)
227{
228 uint64_t *size = opaque;
229
230 if (object_dynamic_cast(obj, TYPE_MEMORY_DEVICE)) {
231 const DeviceState *dev = DEVICE(obj);
232 const MemoryDeviceState *md = MEMORY_DEVICE(obj);
233 const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(obj);
234
235 if (dev->realized) {
236 *size += mdc->get_plugged_size(md, &error_abort);
237 }
238 }
239
240 object_child_foreach(obj, memory_device_plugged_size, opaque);
241 return 0;
242}
243
244uint64_t get_plugged_memory_size(void)
245{
246 uint64_t size = 0;
247
248 memory_device_plugged_size(qdev_get_machine(), &size);
249
250 return size;
251}
252
253void memory_device_pre_plug(MemoryDeviceState *md, MachineState *ms,
254 const uint64_t *legacy_align, Error **errp)
255{
256 const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(md);
257 Error *local_err = NULL;
258 uint64_t addr, align;
259 MemoryRegion *mr;
260
261 mr = mdc->get_memory_region(md, &local_err);
262 if (local_err) {
263 goto out;
264 }
265
266 align = legacy_align ? *legacy_align : memory_region_get_alignment(mr);
267 addr = mdc->get_addr(md);
268 addr = memory_device_get_free_addr(ms, !addr ? NULL : &addr, align,
269 memory_region_size(mr), &local_err);
270 if (local_err) {
271 goto out;
272 }
273 mdc->set_addr(md, addr, &local_err);
274 if (!local_err) {
275 trace_memory_device_pre_plug(DEVICE(md)->id ? DEVICE(md)->id : "",
276 addr);
277 }
278out:
279 error_propagate(errp, local_err);
280}
281
282void memory_device_plug(MemoryDeviceState *md, MachineState *ms)
283{
284 const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(md);
285 const uint64_t addr = mdc->get_addr(md);
286 MemoryRegion *mr;
287
288 /*
289 * We expect that a previous call to memory_device_pre_plug() succeeded, so
290 * it can't fail at this point.
291 */
292 mr = mdc->get_memory_region(md, &error_abort);
293 g_assert(ms->device_memory);
294
295 memory_region_add_subregion(&ms->device_memory->mr,
296 addr - ms->device_memory->base, mr);
297 trace_memory_device_plug(DEVICE(md)->id ? DEVICE(md)->id : "", addr);
298}
299
300void memory_device_unplug(MemoryDeviceState *md, MachineState *ms)
301{
302 const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(md);
303 MemoryRegion *mr;
304
305 /*
306 * We expect that a previous call to memory_device_pre_plug() succeeded, so
307 * it can't fail at this point.
308 */
309 mr = mdc->get_memory_region(md, &error_abort);
310 g_assert(ms->device_memory);
311
312 memory_region_del_subregion(&ms->device_memory->mr, mr);
313 trace_memory_device_unplug(DEVICE(md)->id ? DEVICE(md)->id : "",
314 mdc->get_addr(md));
315}
316
317uint64_t memory_device_get_region_size(const MemoryDeviceState *md,
318 Error **errp)
319{
320 const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(md);
321 MemoryRegion *mr;
322
323 /* dropping const here is fine as we don't touch the memory region */
324 mr = mdc->get_memory_region((MemoryDeviceState *)md, errp);
325 if (!mr) {
326 return 0;
327 }
328
329 return memory_region_size(mr);
330}
331
332static const TypeInfo memory_device_info = {
333 .name = TYPE_MEMORY_DEVICE,
334 .parent = TYPE_INTERFACE,
335 .class_size = sizeof(MemoryDeviceClass),
336};
337
338static void memory_device_register_types(void)
339{
340 type_register_static(&memory_device_info);
341}
342
343type_init(memory_device_register_types)
344