Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
[qemu.git] / savevm.c
1 /*
2 * QEMU System Emulator
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24
25 #include "config-host.h"
26 #include "qemu-common.h"
27 #include "hw/hw.h"
28 #include "hw/qdev.h"
29 #include "net/net.h"
30 #include "monitor/monitor.h"
31 #include "sysemu/sysemu.h"
32 #include "qemu/timer.h"
33 #include "audio/audio.h"
34 #include "migration/migration.h"
35 #include "qemu/sockets.h"
36 #include "qemu/queue.h"
37 #include "sysemu/cpus.h"
38 #include "exec/memory.h"
39 #include "qmp-commands.h"
40 #include "trace.h"
41 #include "qemu/iov.h"
42 #include "block/snapshot.h"
43 #include "block/qapi.h"
44
45 #define SELF_ANNOUNCE_ROUNDS 5
46
47 #ifndef ETH_P_RARP
48 #define ETH_P_RARP 0x8035
49 #endif
50 #define ARP_HTYPE_ETH 0x0001
51 #define ARP_PTYPE_IP 0x0800
52 #define ARP_OP_REQUEST_REV 0x3
53
54 static int announce_self_create(uint8_t *buf,
55 uint8_t *mac_addr)
56 {
57 /* Ethernet header. */
58 memset(buf, 0xff, 6); /* destination MAC addr */
59 memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
60 *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
61
62 /* RARP header. */
63 *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
64 *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
65 *(buf + 18) = 6; /* hardware addr length (ethernet) */
66 *(buf + 19) = 4; /* protocol addr length (IPv4) */
67 *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
68 memcpy(buf + 22, mac_addr, 6); /* source hw addr */
69 memset(buf + 28, 0x00, 4); /* source protocol addr */
70 memcpy(buf + 32, mac_addr, 6); /* target hw addr */
71 memset(buf + 38, 0x00, 4); /* target protocol addr */
72
73 /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
74 memset(buf + 42, 0x00, 18);
75
76 return 60; /* len (FCS will be added by hardware) */
77 }
78
79 static void qemu_announce_self_iter(NICState *nic, void *opaque)
80 {
81 uint8_t buf[60];
82 int len;
83
84 len = announce_self_create(buf, nic->conf->macaddr.a);
85
86 qemu_send_packet_raw(qemu_get_queue(nic), buf, len);
87 }
88
89
90 static void qemu_announce_self_once(void *opaque)
91 {
92 static int count = SELF_ANNOUNCE_ROUNDS;
93 QEMUTimer *timer = *(QEMUTimer **)opaque;
94
95 qemu_foreach_nic(qemu_announce_self_iter, NULL);
96
97 if (--count) {
98 /* delay 50ms, 150ms, 250ms, ... */
99 timer_mod(timer, qemu_clock_get_ms(QEMU_CLOCK_REALTIME) +
100 50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
101 } else {
102 timer_del(timer);
103 timer_free(timer);
104 }
105 }
106
107 void qemu_announce_self(void)
108 {
109 static QEMUTimer *timer;
110 timer = timer_new_ms(QEMU_CLOCK_REALTIME, qemu_announce_self_once, &timer);
111 qemu_announce_self_once(&timer);
112 }
113
114 /***********************************************************/
115 /* savevm/loadvm support */
116
117 static ssize_t block_writev_buffer(void *opaque, struct iovec *iov, int iovcnt,
118 int64_t pos)
119 {
120 int ret;
121 QEMUIOVector qiov;
122
123 qemu_iovec_init_external(&qiov, iov, iovcnt);
124 ret = bdrv_writev_vmstate(opaque, &qiov, pos);
125 if (ret < 0) {
126 return ret;
127 }
128
129 return qiov.size;
130 }
131
132 static int block_put_buffer(void *opaque, const uint8_t *buf,
133 int64_t pos, int size)
134 {
135 bdrv_save_vmstate(opaque, buf, pos, size);
136 return size;
137 }
138
139 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
140 {
141 return bdrv_load_vmstate(opaque, buf, pos, size);
142 }
143
144 static int bdrv_fclose(void *opaque)
145 {
146 return bdrv_flush(opaque);
147 }
148
149 static const QEMUFileOps bdrv_read_ops = {
150 .get_buffer = block_get_buffer,
151 .close = bdrv_fclose
152 };
153
154 static const QEMUFileOps bdrv_write_ops = {
155 .put_buffer = block_put_buffer,
156 .writev_buffer = block_writev_buffer,
157 .close = bdrv_fclose
158 };
159
160 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
161 {
162 if (is_writable) {
163 return qemu_fopen_ops(bs, &bdrv_write_ops);
164 }
165 return qemu_fopen_ops(bs, &bdrv_read_ops);
166 }
167
168
169 /* QEMUFile timer support.
170 * Not in qemu-file.c to not add qemu-timer.c as dependency to qemu-file.c
171 */
172
173 void timer_put(QEMUFile *f, QEMUTimer *ts)
174 {
175 uint64_t expire_time;
176
177 expire_time = timer_expire_time_ns(ts);
178 qemu_put_be64(f, expire_time);
179 }
180
181 void timer_get(QEMUFile *f, QEMUTimer *ts)
182 {
183 uint64_t expire_time;
184
185 expire_time = qemu_get_be64(f);
186 if (expire_time != -1) {
187 timer_mod_ns(ts, expire_time);
188 } else {
189 timer_del(ts);
190 }
191 }
192
193
194 /* VMState timer support.
195 * Not in vmstate.c to not add qemu-timer.c as dependency to vmstate.c
196 */
197
198 static int get_timer(QEMUFile *f, void *pv, size_t size)
199 {
200 QEMUTimer *v = pv;
201 timer_get(f, v);
202 return 0;
203 }
204
205 static void put_timer(QEMUFile *f, void *pv, size_t size)
206 {
207 QEMUTimer *v = pv;
208 timer_put(f, v);
209 }
210
211 const VMStateInfo vmstate_info_timer = {
212 .name = "timer",
213 .get = get_timer,
214 .put = put_timer,
215 };
216
217
218 typedef struct CompatEntry {
219 char idstr[256];
220 int instance_id;
221 } CompatEntry;
222
223 typedef struct SaveStateEntry {
224 QTAILQ_ENTRY(SaveStateEntry) entry;
225 char idstr[256];
226 int instance_id;
227 int alias_id;
228 int version_id;
229 int section_id;
230 SaveVMHandlers *ops;
231 const VMStateDescription *vmsd;
232 void *opaque;
233 CompatEntry *compat;
234 int no_migrate;
235 int is_ram;
236 } SaveStateEntry;
237
238
239 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
240 QTAILQ_HEAD_INITIALIZER(savevm_handlers);
241 static int global_section_id;
242
243 static int calculate_new_instance_id(const char *idstr)
244 {
245 SaveStateEntry *se;
246 int instance_id = 0;
247
248 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
249 if (strcmp(idstr, se->idstr) == 0
250 && instance_id <= se->instance_id) {
251 instance_id = se->instance_id + 1;
252 }
253 }
254 return instance_id;
255 }
256
257 static int calculate_compat_instance_id(const char *idstr)
258 {
259 SaveStateEntry *se;
260 int instance_id = 0;
261
262 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
263 if (!se->compat) {
264 continue;
265 }
266
267 if (strcmp(idstr, se->compat->idstr) == 0
268 && instance_id <= se->compat->instance_id) {
269 instance_id = se->compat->instance_id + 1;
270 }
271 }
272 return instance_id;
273 }
274
275 /* TODO: Individual devices generally have very little idea about the rest
276 of the system, so instance_id should be removed/replaced.
277 Meanwhile pass -1 as instance_id if you do not already have a clearly
278 distinguishing id for all instances of your device class. */
279 int register_savevm_live(DeviceState *dev,
280 const char *idstr,
281 int instance_id,
282 int version_id,
283 SaveVMHandlers *ops,
284 void *opaque)
285 {
286 SaveStateEntry *se;
287
288 se = g_malloc0(sizeof(SaveStateEntry));
289 se->version_id = version_id;
290 se->section_id = global_section_id++;
291 se->ops = ops;
292 se->opaque = opaque;
293 se->vmsd = NULL;
294 se->no_migrate = 0;
295 /* if this is a live_savem then set is_ram */
296 if (ops->save_live_setup != NULL) {
297 se->is_ram = 1;
298 }
299
300 if (dev) {
301 char *id = qdev_get_dev_path(dev);
302 if (id) {
303 pstrcpy(se->idstr, sizeof(se->idstr), id);
304 pstrcat(se->idstr, sizeof(se->idstr), "/");
305 g_free(id);
306
307 se->compat = g_malloc0(sizeof(CompatEntry));
308 pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
309 se->compat->instance_id = instance_id == -1 ?
310 calculate_compat_instance_id(idstr) : instance_id;
311 instance_id = -1;
312 }
313 }
314 pstrcat(se->idstr, sizeof(se->idstr), idstr);
315
316 if (instance_id == -1) {
317 se->instance_id = calculate_new_instance_id(se->idstr);
318 } else {
319 se->instance_id = instance_id;
320 }
321 assert(!se->compat || se->instance_id == 0);
322 /* add at the end of list */
323 QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
324 return 0;
325 }
326
327 int register_savevm(DeviceState *dev,
328 const char *idstr,
329 int instance_id,
330 int version_id,
331 SaveStateHandler *save_state,
332 LoadStateHandler *load_state,
333 void *opaque)
334 {
335 SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
336 ops->save_state = save_state;
337 ops->load_state = load_state;
338 return register_savevm_live(dev, idstr, instance_id, version_id,
339 ops, opaque);
340 }
341
342 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
343 {
344 SaveStateEntry *se, *new_se;
345 char id[256] = "";
346
347 if (dev) {
348 char *path = qdev_get_dev_path(dev);
349 if (path) {
350 pstrcpy(id, sizeof(id), path);
351 pstrcat(id, sizeof(id), "/");
352 g_free(path);
353 }
354 }
355 pstrcat(id, sizeof(id), idstr);
356
357 QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
358 if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
359 QTAILQ_REMOVE(&savevm_handlers, se, entry);
360 if (se->compat) {
361 g_free(se->compat);
362 }
363 g_free(se->ops);
364 g_free(se);
365 }
366 }
367 }
368
369 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
370 const VMStateDescription *vmsd,
371 void *opaque, int alias_id,
372 int required_for_version)
373 {
374 SaveStateEntry *se;
375
376 /* If this triggers, alias support can be dropped for the vmsd. */
377 assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
378
379 se = g_malloc0(sizeof(SaveStateEntry));
380 se->version_id = vmsd->version_id;
381 se->section_id = global_section_id++;
382 se->opaque = opaque;
383 se->vmsd = vmsd;
384 se->alias_id = alias_id;
385 se->no_migrate = vmsd->unmigratable;
386
387 if (dev) {
388 char *id = qdev_get_dev_path(dev);
389 if (id) {
390 pstrcpy(se->idstr, sizeof(se->idstr), id);
391 pstrcat(se->idstr, sizeof(se->idstr), "/");
392 g_free(id);
393
394 se->compat = g_malloc0(sizeof(CompatEntry));
395 pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
396 se->compat->instance_id = instance_id == -1 ?
397 calculate_compat_instance_id(vmsd->name) : instance_id;
398 instance_id = -1;
399 }
400 }
401 pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
402
403 if (instance_id == -1) {
404 se->instance_id = calculate_new_instance_id(se->idstr);
405 } else {
406 se->instance_id = instance_id;
407 }
408 assert(!se->compat || se->instance_id == 0);
409 /* add at the end of list */
410 QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
411 return 0;
412 }
413
414 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
415 void *opaque)
416 {
417 SaveStateEntry *se, *new_se;
418
419 QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
420 if (se->vmsd == vmsd && se->opaque == opaque) {
421 QTAILQ_REMOVE(&savevm_handlers, se, entry);
422 if (se->compat) {
423 g_free(se->compat);
424 }
425 g_free(se);
426 }
427 }
428 }
429
430 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
431 {
432 if (!se->vmsd) { /* Old style */
433 return se->ops->load_state(f, se->opaque, version_id);
434 }
435 return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
436 }
437
438 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
439 {
440 if (!se->vmsd) { /* Old style */
441 se->ops->save_state(f, se->opaque);
442 return;
443 }
444 vmstate_save_state(f, se->vmsd, se->opaque);
445 }
446
447 bool qemu_savevm_state_blocked(Error **errp)
448 {
449 SaveStateEntry *se;
450
451 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
452 if (se->no_migrate) {
453 error_set(errp, QERR_MIGRATION_NOT_SUPPORTED, se->idstr);
454 return true;
455 }
456 }
457 return false;
458 }
459
460 void qemu_savevm_state_begin(QEMUFile *f,
461 const MigrationParams *params)
462 {
463 SaveStateEntry *se;
464 int ret;
465
466 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
467 if (!se->ops || !se->ops->set_params) {
468 continue;
469 }
470 se->ops->set_params(params, se->opaque);
471 }
472
473 qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
474 qemu_put_be32(f, QEMU_VM_FILE_VERSION);
475
476 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
477 int len;
478
479 if (!se->ops || !se->ops->save_live_setup) {
480 continue;
481 }
482 if (se->ops && se->ops->is_active) {
483 if (!se->ops->is_active(se->opaque)) {
484 continue;
485 }
486 }
487 /* Section type */
488 qemu_put_byte(f, QEMU_VM_SECTION_START);
489 qemu_put_be32(f, se->section_id);
490
491 /* ID string */
492 len = strlen(se->idstr);
493 qemu_put_byte(f, len);
494 qemu_put_buffer(f, (uint8_t *)se->idstr, len);
495
496 qemu_put_be32(f, se->instance_id);
497 qemu_put_be32(f, se->version_id);
498
499 ret = se->ops->save_live_setup(f, se->opaque);
500 if (ret < 0) {
501 qemu_file_set_error(f, ret);
502 break;
503 }
504 }
505 }
506
507 /*
508 * this function has three return values:
509 * negative: there was one error, and we have -errno.
510 * 0 : We haven't finished, caller have to go again
511 * 1 : We have finished, we can go to complete phase
512 */
513 int qemu_savevm_state_iterate(QEMUFile *f)
514 {
515 SaveStateEntry *se;
516 int ret = 1;
517
518 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
519 if (!se->ops || !se->ops->save_live_iterate) {
520 continue;
521 }
522 if (se->ops && se->ops->is_active) {
523 if (!se->ops->is_active(se->opaque)) {
524 continue;
525 }
526 }
527 if (qemu_file_rate_limit(f)) {
528 return 0;
529 }
530 trace_savevm_section_start(se->idstr, se->section_id);
531 /* Section type */
532 qemu_put_byte(f, QEMU_VM_SECTION_PART);
533 qemu_put_be32(f, se->section_id);
534
535 ret = se->ops->save_live_iterate(f, se->opaque);
536 trace_savevm_section_end(se->idstr, se->section_id);
537
538 if (ret < 0) {
539 qemu_file_set_error(f, ret);
540 }
541 if (ret <= 0) {
542 /* Do not proceed to the next vmstate before this one reported
543 completion of the current stage. This serializes the migration
544 and reduces the probability that a faster changing state is
545 synchronized over and over again. */
546 break;
547 }
548 }
549 return ret;
550 }
551
552 void qemu_savevm_state_complete(QEMUFile *f)
553 {
554 SaveStateEntry *se;
555 int ret;
556
557 cpu_synchronize_all_states();
558
559 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
560 if (!se->ops || !se->ops->save_live_complete) {
561 continue;
562 }
563 if (se->ops && se->ops->is_active) {
564 if (!se->ops->is_active(se->opaque)) {
565 continue;
566 }
567 }
568 trace_savevm_section_start(se->idstr, se->section_id);
569 /* Section type */
570 qemu_put_byte(f, QEMU_VM_SECTION_END);
571 qemu_put_be32(f, se->section_id);
572
573 ret = se->ops->save_live_complete(f, se->opaque);
574 trace_savevm_section_end(se->idstr, se->section_id);
575 if (ret < 0) {
576 qemu_file_set_error(f, ret);
577 return;
578 }
579 }
580
581 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
582 int len;
583
584 if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
585 continue;
586 }
587 trace_savevm_section_start(se->idstr, se->section_id);
588 /* Section type */
589 qemu_put_byte(f, QEMU_VM_SECTION_FULL);
590 qemu_put_be32(f, se->section_id);
591
592 /* ID string */
593 len = strlen(se->idstr);
594 qemu_put_byte(f, len);
595 qemu_put_buffer(f, (uint8_t *)se->idstr, len);
596
597 qemu_put_be32(f, se->instance_id);
598 qemu_put_be32(f, se->version_id);
599
600 vmstate_save(f, se);
601 trace_savevm_section_end(se->idstr, se->section_id);
602 }
603
604 qemu_put_byte(f, QEMU_VM_EOF);
605 qemu_fflush(f);
606 }
607
608 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
609 {
610 SaveStateEntry *se;
611 uint64_t ret = 0;
612
613 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
614 if (!se->ops || !se->ops->save_live_pending) {
615 continue;
616 }
617 if (se->ops && se->ops->is_active) {
618 if (!se->ops->is_active(se->opaque)) {
619 continue;
620 }
621 }
622 ret += se->ops->save_live_pending(f, se->opaque, max_size);
623 }
624 return ret;
625 }
626
627 void qemu_savevm_state_cancel(void)
628 {
629 SaveStateEntry *se;
630
631 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
632 if (se->ops && se->ops->cancel) {
633 se->ops->cancel(se->opaque);
634 }
635 }
636 }
637
638 static int qemu_savevm_state(QEMUFile *f)
639 {
640 int ret;
641 MigrationParams params = {
642 .blk = 0,
643 .shared = 0
644 };
645
646 if (qemu_savevm_state_blocked(NULL)) {
647 return -EINVAL;
648 }
649
650 qemu_mutex_unlock_iothread();
651 qemu_savevm_state_begin(f, &params);
652 qemu_mutex_lock_iothread();
653
654 while (qemu_file_get_error(f) == 0) {
655 if (qemu_savevm_state_iterate(f) > 0) {
656 break;
657 }
658 }
659
660 ret = qemu_file_get_error(f);
661 if (ret == 0) {
662 qemu_savevm_state_complete(f);
663 ret = qemu_file_get_error(f);
664 }
665 if (ret != 0) {
666 qemu_savevm_state_cancel();
667 }
668 return ret;
669 }
670
671 static int qemu_save_device_state(QEMUFile *f)
672 {
673 SaveStateEntry *se;
674
675 qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
676 qemu_put_be32(f, QEMU_VM_FILE_VERSION);
677
678 cpu_synchronize_all_states();
679
680 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
681 int len;
682
683 if (se->is_ram) {
684 continue;
685 }
686 if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
687 continue;
688 }
689
690 /* Section type */
691 qemu_put_byte(f, QEMU_VM_SECTION_FULL);
692 qemu_put_be32(f, se->section_id);
693
694 /* ID string */
695 len = strlen(se->idstr);
696 qemu_put_byte(f, len);
697 qemu_put_buffer(f, (uint8_t *)se->idstr, len);
698
699 qemu_put_be32(f, se->instance_id);
700 qemu_put_be32(f, se->version_id);
701
702 vmstate_save(f, se);
703 }
704
705 qemu_put_byte(f, QEMU_VM_EOF);
706
707 return qemu_file_get_error(f);
708 }
709
710 static SaveStateEntry *find_se(const char *idstr, int instance_id)
711 {
712 SaveStateEntry *se;
713
714 QTAILQ_FOREACH(se, &savevm_handlers, entry) {
715 if (!strcmp(se->idstr, idstr) &&
716 (instance_id == se->instance_id ||
717 instance_id == se->alias_id))
718 return se;
719 /* Migrating from an older version? */
720 if (strstr(se->idstr, idstr) && se->compat) {
721 if (!strcmp(se->compat->idstr, idstr) &&
722 (instance_id == se->compat->instance_id ||
723 instance_id == se->alias_id))
724 return se;
725 }
726 }
727 return NULL;
728 }
729
730 typedef struct LoadStateEntry {
731 QLIST_ENTRY(LoadStateEntry) entry;
732 SaveStateEntry *se;
733 int section_id;
734 int version_id;
735 } LoadStateEntry;
736
737 int qemu_loadvm_state(QEMUFile *f)
738 {
739 QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
740 QLIST_HEAD_INITIALIZER(loadvm_handlers);
741 LoadStateEntry *le, *new_le;
742 uint8_t section_type;
743 unsigned int v;
744 int ret;
745
746 if (qemu_savevm_state_blocked(NULL)) {
747 return -EINVAL;
748 }
749
750 v = qemu_get_be32(f);
751 if (v != QEMU_VM_FILE_MAGIC) {
752 return -EINVAL;
753 }
754
755 v = qemu_get_be32(f);
756 if (v == QEMU_VM_FILE_VERSION_COMPAT) {
757 fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
758 return -ENOTSUP;
759 }
760 if (v != QEMU_VM_FILE_VERSION) {
761 return -ENOTSUP;
762 }
763
764 while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
765 uint32_t instance_id, version_id, section_id;
766 SaveStateEntry *se;
767 char idstr[257];
768 int len;
769
770 switch (section_type) {
771 case QEMU_VM_SECTION_START:
772 case QEMU_VM_SECTION_FULL:
773 /* Read section start */
774 section_id = qemu_get_be32(f);
775 len = qemu_get_byte(f);
776 qemu_get_buffer(f, (uint8_t *)idstr, len);
777 idstr[len] = 0;
778 instance_id = qemu_get_be32(f);
779 version_id = qemu_get_be32(f);
780
781 /* Find savevm section */
782 se = find_se(idstr, instance_id);
783 if (se == NULL) {
784 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
785 ret = -EINVAL;
786 goto out;
787 }
788
789 /* Validate version */
790 if (version_id > se->version_id) {
791 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
792 version_id, idstr, se->version_id);
793 ret = -EINVAL;
794 goto out;
795 }
796
797 /* Add entry */
798 le = g_malloc0(sizeof(*le));
799
800 le->se = se;
801 le->section_id = section_id;
802 le->version_id = version_id;
803 QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
804
805 ret = vmstate_load(f, le->se, le->version_id);
806 if (ret < 0) {
807 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
808 instance_id, idstr);
809 goto out;
810 }
811 break;
812 case QEMU_VM_SECTION_PART:
813 case QEMU_VM_SECTION_END:
814 section_id = qemu_get_be32(f);
815
816 QLIST_FOREACH(le, &loadvm_handlers, entry) {
817 if (le->section_id == section_id) {
818 break;
819 }
820 }
821 if (le == NULL) {
822 fprintf(stderr, "Unknown savevm section %d\n", section_id);
823 ret = -EINVAL;
824 goto out;
825 }
826
827 ret = vmstate_load(f, le->se, le->version_id);
828 if (ret < 0) {
829 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
830 section_id);
831 goto out;
832 }
833 break;
834 default:
835 fprintf(stderr, "Unknown savevm section type %d\n", section_type);
836 ret = -EINVAL;
837 goto out;
838 }
839 }
840
841 cpu_synchronize_all_post_init();
842
843 ret = 0;
844
845 out:
846 QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
847 QLIST_REMOVE(le, entry);
848 g_free(le);
849 }
850
851 if (ret == 0) {
852 ret = qemu_file_get_error(f);
853 }
854
855 return ret;
856 }
857
858 static BlockDriverState *find_vmstate_bs(void)
859 {
860 BlockDriverState *bs = NULL;
861 while ((bs = bdrv_next(bs))) {
862 if (bdrv_can_snapshot(bs)) {
863 return bs;
864 }
865 }
866 return NULL;
867 }
868
869 /*
870 * Deletes snapshots of a given name in all opened images.
871 */
872 static int del_existing_snapshots(Monitor *mon, const char *name)
873 {
874 BlockDriverState *bs;
875 QEMUSnapshotInfo sn1, *snapshot = &sn1;
876 Error *err = NULL;
877
878 bs = NULL;
879 while ((bs = bdrv_next(bs))) {
880 if (bdrv_can_snapshot(bs) &&
881 bdrv_snapshot_find(bs, snapshot, name) >= 0) {
882 bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
883 if (err) {
884 monitor_printf(mon,
885 "Error while deleting snapshot on device '%s':"
886 " %s\n",
887 bdrv_get_device_name(bs),
888 error_get_pretty(err));
889 error_free(err);
890 return -1;
891 }
892 }
893 }
894
895 return 0;
896 }
897
898 void do_savevm(Monitor *mon, const QDict *qdict)
899 {
900 BlockDriverState *bs, *bs1;
901 QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
902 int ret;
903 QEMUFile *f;
904 int saved_vm_running;
905 uint64_t vm_state_size;
906 qemu_timeval tv;
907 struct tm tm;
908 const char *name = qdict_get_try_str(qdict, "name");
909
910 /* Verify if there is a device that doesn't support snapshots and is writable */
911 bs = NULL;
912 while ((bs = bdrv_next(bs))) {
913
914 if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
915 continue;
916 }
917
918 if (!bdrv_can_snapshot(bs)) {
919 monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
920 bdrv_get_device_name(bs));
921 return;
922 }
923 }
924
925 bs = find_vmstate_bs();
926 if (!bs) {
927 monitor_printf(mon, "No block device can accept snapshots\n");
928 return;
929 }
930
931 saved_vm_running = runstate_is_running();
932 vm_stop(RUN_STATE_SAVE_VM);
933
934 memset(sn, 0, sizeof(*sn));
935
936 /* fill auxiliary fields */
937 qemu_gettimeofday(&tv);
938 sn->date_sec = tv.tv_sec;
939 sn->date_nsec = tv.tv_usec * 1000;
940 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
941
942 if (name) {
943 ret = bdrv_snapshot_find(bs, old_sn, name);
944 if (ret >= 0) {
945 pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
946 pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
947 } else {
948 pstrcpy(sn->name, sizeof(sn->name), name);
949 }
950 } else {
951 /* cast below needed for OpenBSD where tv_sec is still 'long' */
952 localtime_r((const time_t *)&tv.tv_sec, &tm);
953 strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
954 }
955
956 /* Delete old snapshots of the same name */
957 if (name && del_existing_snapshots(mon, name) < 0) {
958 goto the_end;
959 }
960
961 /* save the VM state */
962 f = qemu_fopen_bdrv(bs, 1);
963 if (!f) {
964 monitor_printf(mon, "Could not open VM state file\n");
965 goto the_end;
966 }
967 ret = qemu_savevm_state(f);
968 vm_state_size = qemu_ftell(f);
969 qemu_fclose(f);
970 if (ret < 0) {
971 monitor_printf(mon, "Error %d while writing VM\n", ret);
972 goto the_end;
973 }
974
975 /* create the snapshots */
976
977 bs1 = NULL;
978 while ((bs1 = bdrv_next(bs1))) {
979 if (bdrv_can_snapshot(bs1)) {
980 /* Write VM state size only to the image that contains the state */
981 sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
982 ret = bdrv_snapshot_create(bs1, sn);
983 if (ret < 0) {
984 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
985 bdrv_get_device_name(bs1));
986 }
987 }
988 }
989
990 the_end:
991 if (saved_vm_running) {
992 vm_start();
993 }
994 }
995
996 void qmp_xen_save_devices_state(const char *filename, Error **errp)
997 {
998 QEMUFile *f;
999 int saved_vm_running;
1000 int ret;
1001
1002 saved_vm_running = runstate_is_running();
1003 vm_stop(RUN_STATE_SAVE_VM);
1004
1005 f = qemu_fopen(filename, "wb");
1006 if (!f) {
1007 error_setg_file_open(errp, errno, filename);
1008 goto the_end;
1009 }
1010 ret = qemu_save_device_state(f);
1011 qemu_fclose(f);
1012 if (ret < 0) {
1013 error_set(errp, QERR_IO_ERROR);
1014 }
1015
1016 the_end:
1017 if (saved_vm_running) {
1018 vm_start();
1019 }
1020 }
1021
1022 int load_vmstate(const char *name)
1023 {
1024 BlockDriverState *bs, *bs_vm_state;
1025 QEMUSnapshotInfo sn;
1026 QEMUFile *f;
1027 int ret;
1028
1029 bs_vm_state = find_vmstate_bs();
1030 if (!bs_vm_state) {
1031 error_report("No block device supports snapshots");
1032 return -ENOTSUP;
1033 }
1034
1035 /* Don't even try to load empty VM states */
1036 ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
1037 if (ret < 0) {
1038 return ret;
1039 } else if (sn.vm_state_size == 0) {
1040 error_report("This is a disk-only snapshot. Revert to it offline "
1041 "using qemu-img.");
1042 return -EINVAL;
1043 }
1044
1045 /* Verify if there is any device that doesn't support snapshots and is
1046 writable and check if the requested snapshot is available too. */
1047 bs = NULL;
1048 while ((bs = bdrv_next(bs))) {
1049
1050 if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
1051 continue;
1052 }
1053
1054 if (!bdrv_can_snapshot(bs)) {
1055 error_report("Device '%s' is writable but does not support snapshots.",
1056 bdrv_get_device_name(bs));
1057 return -ENOTSUP;
1058 }
1059
1060 ret = bdrv_snapshot_find(bs, &sn, name);
1061 if (ret < 0) {
1062 error_report("Device '%s' does not have the requested snapshot '%s'",
1063 bdrv_get_device_name(bs), name);
1064 return ret;
1065 }
1066 }
1067
1068 /* Flush all IO requests so they don't interfere with the new state. */
1069 bdrv_drain_all();
1070
1071 bs = NULL;
1072 while ((bs = bdrv_next(bs))) {
1073 if (bdrv_can_snapshot(bs)) {
1074 ret = bdrv_snapshot_goto(bs, name);
1075 if (ret < 0) {
1076 error_report("Error %d while activating snapshot '%s' on '%s'",
1077 ret, name, bdrv_get_device_name(bs));
1078 return ret;
1079 }
1080 }
1081 }
1082
1083 /* restore the VM state */
1084 f = qemu_fopen_bdrv(bs_vm_state, 0);
1085 if (!f) {
1086 error_report("Could not open VM state file");
1087 return -EINVAL;
1088 }
1089
1090 qemu_system_reset(VMRESET_SILENT);
1091 ret = qemu_loadvm_state(f);
1092
1093 qemu_fclose(f);
1094 if (ret < 0) {
1095 error_report("Error %d while loading VM state", ret);
1096 return ret;
1097 }
1098
1099 return 0;
1100 }
1101
1102 void do_delvm(Monitor *mon, const QDict *qdict)
1103 {
1104 BlockDriverState *bs, *bs1;
1105 Error *err = NULL;
1106 const char *name = qdict_get_str(qdict, "name");
1107
1108 bs = find_vmstate_bs();
1109 if (!bs) {
1110 monitor_printf(mon, "No block device supports snapshots\n");
1111 return;
1112 }
1113
1114 bs1 = NULL;
1115 while ((bs1 = bdrv_next(bs1))) {
1116 if (bdrv_can_snapshot(bs1)) {
1117 bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
1118 if (err) {
1119 monitor_printf(mon,
1120 "Error while deleting snapshot on device '%s':"
1121 " %s\n",
1122 bdrv_get_device_name(bs),
1123 error_get_pretty(err));
1124 error_free(err);
1125 }
1126 }
1127 }
1128 }
1129
1130 void do_info_snapshots(Monitor *mon, const QDict *qdict)
1131 {
1132 BlockDriverState *bs, *bs1;
1133 QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
1134 int nb_sns, i, ret, available;
1135 int total;
1136 int *available_snapshots;
1137
1138 bs = find_vmstate_bs();
1139 if (!bs) {
1140 monitor_printf(mon, "No available block device supports snapshots\n");
1141 return;
1142 }
1143
1144 nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1145 if (nb_sns < 0) {
1146 monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1147 return;
1148 }
1149
1150 if (nb_sns == 0) {
1151 monitor_printf(mon, "There is no snapshot available.\n");
1152 return;
1153 }
1154
1155 available_snapshots = g_malloc0(sizeof(int) * nb_sns);
1156 total = 0;
1157 for (i = 0; i < nb_sns; i++) {
1158 sn = &sn_tab[i];
1159 available = 1;
1160 bs1 = NULL;
1161
1162 while ((bs1 = bdrv_next(bs1))) {
1163 if (bdrv_can_snapshot(bs1) && bs1 != bs) {
1164 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
1165 if (ret < 0) {
1166 available = 0;
1167 break;
1168 }
1169 }
1170 }
1171
1172 if (available) {
1173 available_snapshots[total] = i;
1174 total++;
1175 }
1176 }
1177
1178 if (total > 0) {
1179 bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, NULL);
1180 monitor_printf(mon, "\n");
1181 for (i = 0; i < total; i++) {
1182 sn = &sn_tab[available_snapshots[i]];
1183 bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, sn);
1184 monitor_printf(mon, "\n");
1185 }
1186 } else {
1187 monitor_printf(mon, "There is no suitable snapshot available\n");
1188 }
1189
1190 g_free(sn_tab);
1191 g_free(available_snapshots);
1192
1193 }
1194
1195 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
1196 {
1197 qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
1198 memory_region_name(mr), dev);
1199 }
1200
1201 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
1202 {
1203 /* Nothing do to while the implementation is in RAMBlock */
1204 }
1205
1206 void vmstate_register_ram_global(MemoryRegion *mr)
1207 {
1208 vmstate_register_ram(mr, NULL);
1209 }