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     trace_qemu_announce_self_iter(qemu_ether_ntoa(&nic->conf->macaddr));
  85     len = announce_self_create(buf, nic->conf->macaddr.a);
  86
  87     qemu_send_packet_raw(qemu_get_queue(nic), buf, len);
  88 }
  89
  90
  91 static void qemu_announce_self_once(void *opaque)
  92 {
  93     static int count = SELF_ANNOUNCE_ROUNDS;
  94     QEMUTimer *timer = *(QEMUTimer **)opaque;
  95
  96     qemu_foreach_nic(qemu_announce_self_iter, NULL);
  97
  98     if (--count) {
  99         /* delay 50ms, 150ms, 250ms, ... */
 100         timer_mod(timer, qemu_clock_get_ms(QEMU_CLOCK_REALTIME) +
 101                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
 102     } else {
 103             timer_del(timer);
 104             timer_free(timer);
 105     }
 106 }
 107
 108 void qemu_announce_self(void)
 109 {
 110     static QEMUTimer *timer;
 111     timer = timer_new_ms(QEMU_CLOCK_REALTIME, qemu_announce_self_once, &timer);
 112     qemu_announce_self_once(&timer);
 113 }
 114
 115 /***********************************************************/
 116 /* savevm/loadvm support */
 117
 118 static ssize_t block_writev_buffer(void *opaque, struct iovec *iov, int iovcnt,
 119                                    int64_t pos)
 120 {
 121     int ret;
 122     QEMUIOVector qiov;
 123
 124     qemu_iovec_init_external(&qiov, iov, iovcnt);
 125     ret = bdrv_writev_vmstate(opaque, &qiov, pos);
 126     if (ret < 0) {
 127         return ret;
 128     }
 129
 130     return qiov.size;
 131 }
 132
 133 static int block_put_buffer(void *opaque, const uint8_t *buf,
 134                            int64_t pos, int size)
 135 {
 136     bdrv_save_vmstate(opaque, buf, pos, size);
 137     return size;
 138 }
 139
 140 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 141 {
 142     return bdrv_load_vmstate(opaque, buf, pos, size);
 143 }
 144
 145 static int bdrv_fclose(void *opaque)
 146 {
 147     return bdrv_flush(opaque);
 148 }
 149
 150 static const QEMUFileOps bdrv_read_ops = {
 151     .get_buffer = block_get_buffer,
 152     .close =      bdrv_fclose
 153 };
 154
 155 static const QEMUFileOps bdrv_write_ops = {
 156     .put_buffer     = block_put_buffer,
 157     .writev_buffer  = block_writev_buffer,
 158     .close          = bdrv_fclose
 159 };
 160
 161 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 162 {
 163     if (is_writable) {
 164         return qemu_fopen_ops(bs, &bdrv_write_ops);
 165     }
 166     return qemu_fopen_ops(bs, &bdrv_read_ops);
 167 }
 168
 169
 170 /* QEMUFile timer support.
 171  * Not in qemu-file.c to not add qemu-timer.c as dependency to qemu-file.c
 172  */
 173
 174 void timer_put(QEMUFile *f, QEMUTimer *ts)
 175 {
 176     uint64_t expire_time;
 177
 178     expire_time = timer_expire_time_ns(ts);
 179     qemu_put_be64(f, expire_time);
 180 }
 181
 182 void timer_get(QEMUFile *f, QEMUTimer *ts)
 183 {
 184     uint64_t expire_time;
 185
 186     expire_time = qemu_get_be64(f);
 187     if (expire_time != -1) {
 188         timer_mod_ns(ts, expire_time);
 189     } else {
 190         timer_del(ts);
 191     }
 192 }
 193
 194
 195 /* VMState timer support.
 196  * Not in vmstate.c to not add qemu-timer.c as dependency to vmstate.c
 197  */
 198
 199 static int get_timer(QEMUFile *f, void *pv, size_t size)
 200 {
 201     QEMUTimer *v = pv;
 202     timer_get(f, v);
 203     return 0;
 204 }
 205
 206 static void put_timer(QEMUFile *f, void *pv, size_t size)
 207 {
 208     QEMUTimer *v = pv;
 209     timer_put(f, v);
 210 }
 211
 212 const VMStateInfo vmstate_info_timer = {
 213     .name = "timer",
 214     .get  = get_timer,
 215     .put  = put_timer,
 216 };
 217
 218
 219 typedef struct CompatEntry {
 220     char idstr[256];
 221     int instance_id;
 222 } CompatEntry;
 223
 224 typedef struct SaveStateEntry {
 225     QTAILQ_ENTRY(SaveStateEntry) entry;
 226     char idstr[256];
 227     int instance_id;
 228     int alias_id;
 229     int version_id;
 230     int section_id;
 231     SaveVMHandlers *ops;
 232     const VMStateDescription *vmsd;
 233     void *opaque;
 234     CompatEntry *compat;
 235     int no_migrate;
 236     int is_ram;
 237 } SaveStateEntry;
 238
 239
 240 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
 241     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
 242 static int global_section_id;
 243
 244 static int calculate_new_instance_id(const char *idstr)
 245 {
 246     SaveStateEntry *se;
 247     int instance_id = 0;
 248
 249     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 250         if (strcmp(idstr, se->idstr) == 0
 251             && instance_id <= se->instance_id) {
 252             instance_id = se->instance_id + 1;
 253         }
 254     }
 255     return instance_id;
 256 }
 257
 258 static int calculate_compat_instance_id(const char *idstr)
 259 {
 260     SaveStateEntry *se;
 261     int instance_id = 0;
 262
 263     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 264         if (!se->compat) {
 265             continue;
 266         }
 267
 268         if (strcmp(idstr, se->compat->idstr) == 0
 269             && instance_id <= se->compat->instance_id) {
 270             instance_id = se->compat->instance_id + 1;
 271         }
 272     }
 273     return instance_id;
 274 }
 275
 276 /* TODO: Individual devices generally have very little idea about the rest
 277    of the system, so instance_id should be removed/replaced.
 278    Meanwhile pass -1 as instance_id if you do not already have a clearly
 279    distinguishing id for all instances of your device class. */
 280 int register_savevm_live(DeviceState *dev,
 281                          const char *idstr,
 282                          int instance_id,
 283                          int version_id,
 284                          SaveVMHandlers *ops,
 285                          void *opaque)
 286 {
 287     SaveStateEntry *se;
 288
 289     se = g_malloc0(sizeof(SaveStateEntry));
 290     se->version_id = version_id;
 291     se->section_id = global_section_id++;
 292     se->ops = ops;
 293     se->opaque = opaque;
 294     se->vmsd = NULL;
 295     se->no_migrate = 0;
 296     /* if this is a live_savem then set is_ram */
 297     if (ops->save_live_setup != NULL) {
 298         se->is_ram = 1;
 299     }
 300
 301     if (dev) {
 302         char *id = qdev_get_dev_path(dev);
 303         if (id) {
 304             pstrcpy(se->idstr, sizeof(se->idstr), id);
 305             pstrcat(se->idstr, sizeof(se->idstr), "/");
 306             g_free(id);
 307
 308             se->compat = g_malloc0(sizeof(CompatEntry));
 309             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
 310             se->compat->instance_id = instance_id == -1 ?
 311                          calculate_compat_instance_id(idstr) : instance_id;
 312             instance_id = -1;
 313         }
 314     }
 315     pstrcat(se->idstr, sizeof(se->idstr), idstr);
 316
 317     if (instance_id == -1) {
 318         se->instance_id = calculate_new_instance_id(se->idstr);
 319     } else {
 320         se->instance_id = instance_id;
 321     }
 322     assert(!se->compat || se->instance_id == 0);
 323     /* add at the end of list */
 324     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
 325     return 0;
 326 }
 327
 328 int register_savevm(DeviceState *dev,
 329                     const char *idstr,
 330                     int instance_id,
 331                     int version_id,
 332                     SaveStateHandler *save_state,
 333                     LoadStateHandler *load_state,
 334                     void *opaque)
 335 {
 336     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
 337     ops->save_state = save_state;
 338     ops->load_state = load_state;
 339     return register_savevm_live(dev, idstr, instance_id, version_id,
 340                                 ops, opaque);
 341 }
 342
 343 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
 344 {
 345     SaveStateEntry *se, *new_se;
 346     char id[256] = "";
 347
 348     if (dev) {
 349         char *path = qdev_get_dev_path(dev);
 350         if (path) {
 351             pstrcpy(id, sizeof(id), path);
 352             pstrcat(id, sizeof(id), "/");
 353             g_free(path);
 354         }
 355     }
 356     pstrcat(id, sizeof(id), idstr);
 357
 358     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
 359         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
 360             QTAILQ_REMOVE(&savevm_handlers, se, entry);
 361             if (se->compat) {
 362                 g_free(se->compat);
 363             }
 364             g_free(se->ops);
 365             g_free(se);
 366         }
 367     }
 368 }
 369
 370 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
 371                                    const VMStateDescription *vmsd,
 372                                    void *opaque, int alias_id,
 373                                    int required_for_version)
 374 {
 375     SaveStateEntry *se;
 376
 377     /* If this triggers, alias support can be dropped for the vmsd. */
 378     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
 379
 380     se = g_malloc0(sizeof(SaveStateEntry));
 381     se->version_id = vmsd->version_id;
 382     se->section_id = global_section_id++;
 383     se->opaque = opaque;
 384     se->vmsd = vmsd;
 385     se->alias_id = alias_id;
 386     se->no_migrate = vmsd->unmigratable;
 387
 388     if (dev) {
 389         char *id = qdev_get_dev_path(dev);
 390         if (id) {
 391             pstrcpy(se->idstr, sizeof(se->idstr), id);
 392             pstrcat(se->idstr, sizeof(se->idstr), "/");
 393             g_free(id);
 394
 395             se->compat = g_malloc0(sizeof(CompatEntry));
 396             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
 397             se->compat->instance_id = instance_id == -1 ?
 398                          calculate_compat_instance_id(vmsd->name) : instance_id;
 399             instance_id = -1;
 400         }
 401     }
 402     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
 403
 404     if (instance_id == -1) {
 405         se->instance_id = calculate_new_instance_id(se->idstr);
 406     } else {
 407         se->instance_id = instance_id;
 408     }
 409     assert(!se->compat || se->instance_id == 0);
 410     /* add at the end of list */
 411     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
 412     return 0;
 413 }
 414
 415 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
 416                         void *opaque)
 417 {
 418     SaveStateEntry *se, *new_se;
 419
 420     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
 421         if (se->vmsd == vmsd && se->opaque == opaque) {
 422             QTAILQ_REMOVE(&savevm_handlers, se, entry);
 423             if (se->compat) {
 424                 g_free(se->compat);
 425             }
 426             g_free(se);
 427         }
 428     }
 429 }
 430
 431 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
 432 {
 433     trace_vmstate_load(se->idstr, se->vmsd ? se->vmsd->name : "(old)");
 434     if (!se->vmsd) {         /* Old style */
 435         return se->ops->load_state(f, se->opaque, version_id);
 436     }
 437     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
 438 }
 439
 440 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
 441 {
 442     trace_vmstate_save(se->idstr, se->vmsd ? se->vmsd->name : "(old)");
 443     if (!se->vmsd) {         /* Old style */
 444         se->ops->save_state(f, se->opaque);
 445         return;
 446     }
 447     vmstate_save_state(f, se->vmsd, se->opaque);
 448 }
 449
 450 bool qemu_savevm_state_blocked(Error **errp)
 451 {
 452     SaveStateEntry *se;
 453
 454     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 455         if (se->no_migrate) {
 456             error_set(errp, QERR_MIGRATION_NOT_SUPPORTED, se->idstr);
 457             return true;
 458         }
 459     }
 460     return false;
 461 }
 462
 463 void qemu_savevm_state_begin(QEMUFile *f,
 464                              const MigrationParams *params)
 465 {
 466     SaveStateEntry *se;
 467     int ret;
 468
 469     trace_savevm_state_begin();
 470     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 471         if (!se->ops || !se->ops->set_params) {
 472             continue;
 473         }
 474         se->ops->set_params(params, se->opaque);
 475     }
 476
 477     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 478     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 479
 480     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 481         int len;
 482
 483         if (!se->ops || !se->ops->save_live_setup) {
 484             continue;
 485         }
 486         if (se->ops && se->ops->is_active) {
 487             if (!se->ops->is_active(se->opaque)) {
 488                 continue;
 489             }
 490         }
 491         /* Section type */
 492         qemu_put_byte(f, QEMU_VM_SECTION_START);
 493         qemu_put_be32(f, se->section_id);
 494
 495         /* ID string */
 496         len = strlen(se->idstr);
 497         qemu_put_byte(f, len);
 498         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 499
 500         qemu_put_be32(f, se->instance_id);
 501         qemu_put_be32(f, se->version_id);
 502
 503         ret = se->ops->save_live_setup(f, se->opaque);
 504         if (ret < 0) {
 505             qemu_file_set_error(f, ret);
 506             break;
 507         }
 508     }
 509 }
 510
 511 /*
 512  * this function has three return values:
 513  *   negative: there was one error, and we have -errno.
 514  *   0 : We haven't finished, caller have to go again
 515  *   1 : We have finished, we can go to complete phase
 516  */
 517 int qemu_savevm_state_iterate(QEMUFile *f)
 518 {
 519     SaveStateEntry *se;
 520     int ret = 1;
 521
 522     trace_savevm_state_iterate();
 523     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 524         if (!se->ops || !se->ops->save_live_iterate) {
 525             continue;
 526         }
 527         if (se->ops && se->ops->is_active) {
 528             if (!se->ops->is_active(se->opaque)) {
 529                 continue;
 530             }
 531         }
 532         if (qemu_file_rate_limit(f)) {
 533             return 0;
 534         }
 535         trace_savevm_section_start(se->idstr, se->section_id);
 536         /* Section type */
 537         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 538         qemu_put_be32(f, se->section_id);
 539
 540         ret = se->ops->save_live_iterate(f, se->opaque);
 541         trace_savevm_section_end(se->idstr, se->section_id);
 542
 543         if (ret < 0) {
 544             qemu_file_set_error(f, ret);
 545         }
 546         if (ret <= 0) {
 547             /* Do not proceed to the next vmstate before this one reported
 548                completion of the current stage. This serializes the migration
 549                and reduces the probability that a faster changing state is
 550                synchronized over and over again. */
 551             break;
 552         }
 553     }
 554     return ret;
 555 }
 556
 557 void qemu_savevm_state_complete(QEMUFile *f)
 558 {
 559     SaveStateEntry *se;
 560     int ret;
 561
 562     trace_savevm_state_complete();
 563
 564     cpu_synchronize_all_states();
 565
 566     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 567         if (!se->ops || !se->ops->save_live_complete) {
 568             continue;
 569         }
 570         if (se->ops && se->ops->is_active) {
 571             if (!se->ops->is_active(se->opaque)) {
 572                 continue;
 573             }
 574         }
 575         trace_savevm_section_start(se->idstr, se->section_id);
 576         /* Section type */
 577         qemu_put_byte(f, QEMU_VM_SECTION_END);
 578         qemu_put_be32(f, se->section_id);
 579
 580         ret = se->ops->save_live_complete(f, se->opaque);
 581         trace_savevm_section_end(se->idstr, se->section_id);
 582         if (ret < 0) {
 583             qemu_file_set_error(f, ret);
 584             return;
 585         }
 586     }
 587
 588     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 589         int len;
 590
 591         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
 592             continue;
 593         }
 594         trace_savevm_section_start(se->idstr, se->section_id);
 595         /* Section type */
 596         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 597         qemu_put_be32(f, se->section_id);
 598
 599         /* ID string */
 600         len = strlen(se->idstr);
 601         qemu_put_byte(f, len);
 602         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 603
 604         qemu_put_be32(f, se->instance_id);
 605         qemu_put_be32(f, se->version_id);
 606
 607         vmstate_save(f, se);
 608         trace_savevm_section_end(se->idstr, se->section_id);
 609     }
 610
 611     qemu_put_byte(f, QEMU_VM_EOF);
 612     qemu_fflush(f);
 613 }
 614
 615 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
 616 {
 617     SaveStateEntry *se;
 618     uint64_t ret = 0;
 619
 620     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 621         if (!se->ops || !se->ops->save_live_pending) {
 622             continue;
 623         }
 624         if (se->ops && se->ops->is_active) {
 625             if (!se->ops->is_active(se->opaque)) {
 626                 continue;
 627             }
 628         }
 629         ret += se->ops->save_live_pending(f, se->opaque, max_size);
 630     }
 631     return ret;
 632 }
 633
 634 void qemu_savevm_state_cancel(void)
 635 {
 636     SaveStateEntry *se;
 637
 638     trace_savevm_state_cancel();
 639     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 640         if (se->ops && se->ops->cancel) {
 641             se->ops->cancel(se->opaque);
 642         }
 643     }
 644 }
 645
 646 static int qemu_savevm_state(QEMUFile *f)
 647 {
 648     int ret;
 649     MigrationParams params = {
 650         .blk = 0,
 651         .shared = 0
 652     };
 653
 654     if (qemu_savevm_state_blocked(NULL)) {
 655         return -EINVAL;
 656     }
 657
 658     qemu_mutex_unlock_iothread();
 659     qemu_savevm_state_begin(f, &params);
 660     qemu_mutex_lock_iothread();
 661
 662     while (qemu_file_get_error(f) == 0) {
 663         if (qemu_savevm_state_iterate(f) > 0) {
 664             break;
 665         }
 666     }
 667
 668     ret = qemu_file_get_error(f);
 669     if (ret == 0) {
 670         qemu_savevm_state_complete(f);
 671         ret = qemu_file_get_error(f);
 672     }
 673     if (ret != 0) {
 674         qemu_savevm_state_cancel();
 675     }
 676     return ret;
 677 }
 678
 679 static int qemu_save_device_state(QEMUFile *f)
 680 {
 681     SaveStateEntry *se;
 682
 683     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 684     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 685
 686     cpu_synchronize_all_states();
 687
 688     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 689         int len;
 690
 691         if (se->is_ram) {
 692             continue;
 693         }
 694         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
 695             continue;
 696         }
 697
 698         /* Section type */
 699         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 700         qemu_put_be32(f, se->section_id);
 701
 702         /* ID string */
 703         len = strlen(se->idstr);
 704         qemu_put_byte(f, len);
 705         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 706
 707         qemu_put_be32(f, se->instance_id);
 708         qemu_put_be32(f, se->version_id);
 709
 710         vmstate_save(f, se);
 711     }
 712
 713     qemu_put_byte(f, QEMU_VM_EOF);
 714
 715     return qemu_file_get_error(f);
 716 }
 717
 718 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 719 {
 720     SaveStateEntry *se;
 721
 722     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 723         if (!strcmp(se->idstr, idstr) &&
 724             (instance_id == se->instance_id ||
 725              instance_id == se->alias_id))
 726             return se;
 727         /* Migrating from an older version? */
 728         if (strstr(se->idstr, idstr) && se->compat) {
 729             if (!strcmp(se->compat->idstr, idstr) &&
 730                 (instance_id == se->compat->instance_id ||
 731                  instance_id == se->alias_id))
 732                 return se;
 733         }
 734     }
 735     return NULL;
 736 }
 737
 738 typedef struct LoadStateEntry {
 739     QLIST_ENTRY(LoadStateEntry) entry;
 740     SaveStateEntry *se;
 741     int section_id;
 742     int version_id;
 743 } LoadStateEntry;
 744
 745 int qemu_loadvm_state(QEMUFile *f)
 746 {
 747     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
 748         QLIST_HEAD_INITIALIZER(loadvm_handlers);
 749     LoadStateEntry *le, *new_le;
 750     uint8_t section_type;
 751     unsigned int v;
 752     int ret;
 753
 754     if (qemu_savevm_state_blocked(NULL)) {
 755         return -EINVAL;
 756     }
 757
 758     v = qemu_get_be32(f);
 759     if (v != QEMU_VM_FILE_MAGIC) {
 760         return -EINVAL;
 761     }
 762
 763     v = qemu_get_be32(f);
 764     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
 765         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
 766         return -ENOTSUP;
 767     }
 768     if (v != QEMU_VM_FILE_VERSION) {
 769         return -ENOTSUP;
 770     }
 771
 772     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 773         uint32_t instance_id, version_id, section_id;
 774         SaveStateEntry *se;
 775         char idstr[257];
 776         int len;
 777
 778         switch (section_type) {
 779         case QEMU_VM_SECTION_START:
 780         case QEMU_VM_SECTION_FULL:
 781             /* Read section start */
 782             section_id = qemu_get_be32(f);
 783             len = qemu_get_byte(f);
 784             qemu_get_buffer(f, (uint8_t *)idstr, len);
 785             idstr[len] = 0;
 786             instance_id = qemu_get_be32(f);
 787             version_id = qemu_get_be32(f);
 788
 789             /* Find savevm section */
 790             se = find_se(idstr, instance_id);
 791             if (se == NULL) {
 792                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 793                 ret = -EINVAL;
 794                 goto out;
 795             }
 796
 797             /* Validate version */
 798             if (version_id > se->version_id) {
 799                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 800                         version_id, idstr, se->version_id);
 801                 ret = -EINVAL;
 802                 goto out;
 803             }
 804
 805             /* Add entry */
 806             le = g_malloc0(sizeof(*le));
 807
 808             le->se = se;
 809             le->section_id = section_id;
 810             le->version_id = version_id;
 811             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
 812
 813             ret = vmstate_load(f, le->se, le->version_id);
 814             if (ret < 0) {
 815                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 816                         instance_id, idstr);
 817                 goto out;
 818             }
 819             break;
 820         case QEMU_VM_SECTION_PART:
 821         case QEMU_VM_SECTION_END:
 822             section_id = qemu_get_be32(f);
 823
 824             QLIST_FOREACH(le, &loadvm_handlers, entry) {
 825                 if (le->section_id == section_id) {
 826                     break;
 827                 }
 828             }
 829             if (le == NULL) {
 830                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 831                 ret = -EINVAL;
 832                 goto out;
 833             }
 834
 835             ret = vmstate_load(f, le->se, le->version_id);
 836             if (ret < 0) {
 837                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
 838                         section_id);
 839                 goto out;
 840             }
 841             break;
 842         default:
 843             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 844             ret = -EINVAL;
 845             goto out;
 846         }
 847     }
 848
 849     cpu_synchronize_all_post_init();
 850
 851     ret = 0;
 852
 853 out:
 854     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
 855         QLIST_REMOVE(le, entry);
 856         g_free(le);
 857     }
 858
 859     if (ret == 0) {
 860         ret = qemu_file_get_error(f);
 861     }
 862
 863     return ret;
 864 }
 865
 866 static BlockDriverState *find_vmstate_bs(void)
 867 {
 868     BlockDriverState *bs = NULL;
 869     while ((bs = bdrv_next(bs))) {
 870         if (bdrv_can_snapshot(bs)) {
 871             return bs;
 872         }
 873     }
 874     return NULL;
 875 }
 876
 877 /*
 878  * Deletes snapshots of a given name in all opened images.
 879  */
 880 static int del_existing_snapshots(Monitor *mon, const char *name)
 881 {
 882     BlockDriverState *bs;
 883     QEMUSnapshotInfo sn1, *snapshot = &sn1;
 884     Error *err = NULL;
 885
 886     bs = NULL;
 887     while ((bs = bdrv_next(bs))) {
 888         if (bdrv_can_snapshot(bs) &&
 889             bdrv_snapshot_find(bs, snapshot, name) >= 0) {
 890             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
 891             if (err) {
 892                 monitor_printf(mon,
 893                                "Error while deleting snapshot on device '%s':"
 894                                " %s\n",
 895                                bdrv_get_device_name(bs),
 896                                error_get_pretty(err));
 897                 error_free(err);
 898                 return -1;
 899             }
 900         }
 901     }
 902
 903     return 0;
 904 }
 905
 906 void do_savevm(Monitor *mon, const QDict *qdict)
 907 {
 908     BlockDriverState *bs, *bs1;
 909     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
 910     int ret;
 911     QEMUFile *f;
 912     int saved_vm_running;
 913     uint64_t vm_state_size;
 914     qemu_timeval tv;
 915     struct tm tm;
 916     const char *name = qdict_get_try_str(qdict, "name");
 917
 918     /* Verify if there is a device that doesn't support snapshots and is writable */
 919     bs = NULL;
 920     while ((bs = bdrv_next(bs))) {
 921
 922         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
 923             continue;
 924         }
 925
 926         if (!bdrv_can_snapshot(bs)) {
 927             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
 928                                bdrv_get_device_name(bs));
 929             return;
 930         }
 931     }
 932
 933     bs = find_vmstate_bs();
 934     if (!bs) {
 935         monitor_printf(mon, "No block device can accept snapshots\n");
 936         return;
 937     }
 938
 939     saved_vm_running = runstate_is_running();
 940     vm_stop(RUN_STATE_SAVE_VM);
 941
 942     memset(sn, 0, sizeof(*sn));
 943
 944     /* fill auxiliary fields */
 945     qemu_gettimeofday(&tv);
 946     sn->date_sec = tv.tv_sec;
 947     sn->date_nsec = tv.tv_usec * 1000;
 948     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
 949
 950     if (name) {
 951         ret = bdrv_snapshot_find(bs, old_sn, name);
 952         if (ret >= 0) {
 953             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
 954             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
 955         } else {
 956             pstrcpy(sn->name, sizeof(sn->name), name);
 957         }
 958     } else {
 959         /* cast below needed for OpenBSD where tv_sec is still 'long' */
 960         localtime_r((const time_t *)&tv.tv_sec, &tm);
 961         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
 962     }
 963
 964     /* Delete old snapshots of the same name */
 965     if (name && del_existing_snapshots(mon, name) < 0) {
 966         goto the_end;
 967     }
 968
 969     /* save the VM state */
 970     f = qemu_fopen_bdrv(bs, 1);
 971     if (!f) {
 972         monitor_printf(mon, "Could not open VM state file\n");
 973         goto the_end;
 974     }
 975     ret = qemu_savevm_state(f);
 976     vm_state_size = qemu_ftell(f);
 977     qemu_fclose(f);
 978     if (ret < 0) {
 979         monitor_printf(mon, "Error %d while writing VM\n", ret);
 980         goto the_end;
 981     }
 982
 983     /* create the snapshots */
 984
 985     bs1 = NULL;
 986     while ((bs1 = bdrv_next(bs1))) {
 987         if (bdrv_can_snapshot(bs1)) {
 988             /* Write VM state size only to the image that contains the state */
 989             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
 990             ret = bdrv_snapshot_create(bs1, sn);
 991             if (ret < 0) {
 992                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
 993                                bdrv_get_device_name(bs1));
 994             }
 995         }
 996     }
 997
 998  the_end:
 999     if (saved_vm_running) {
1000         vm_start();
1001     }
1002 }
1003
1004 void qmp_xen_save_devices_state(const char *filename, Error **errp)
1005 {
1006     QEMUFile *f;
1007     int saved_vm_running;
1008     int ret;
1009
1010     saved_vm_running = runstate_is_running();
1011     vm_stop(RUN_STATE_SAVE_VM);
1012
1013     f = qemu_fopen(filename, "wb");
1014     if (!f) {
1015         error_setg_file_open(errp, errno, filename);
1016         goto the_end;
1017     }
1018     ret = qemu_save_device_state(f);
1019     qemu_fclose(f);
1020     if (ret < 0) {
1021         error_set(errp, QERR_IO_ERROR);
1022     }
1023
1024  the_end:
1025     if (saved_vm_running) {
1026         vm_start();
1027     }
1028 }
1029
1030 int load_vmstate(const char *name)
1031 {
1032     BlockDriverState *bs, *bs_vm_state;
1033     QEMUSnapshotInfo sn;
1034     QEMUFile *f;
1035     int ret;
1036
1037     bs_vm_state = find_vmstate_bs();
1038     if (!bs_vm_state) {
1039         error_report("No block device supports snapshots");
1040         return -ENOTSUP;
1041     }
1042
1043     /* Don't even try to load empty VM states */
1044     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
1045     if (ret < 0) {
1046         return ret;
1047     } else if (sn.vm_state_size == 0) {
1048         error_report("This is a disk-only snapshot. Revert to it offline "
1049             "using qemu-img.");
1050         return -EINVAL;
1051     }
1052
1053     /* Verify if there is any device that doesn't support snapshots and is
1054     writable and check if the requested snapshot is available too. */
1055     bs = NULL;
1056     while ((bs = bdrv_next(bs))) {
1057
1058         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
1059             continue;
1060         }
1061
1062         if (!bdrv_can_snapshot(bs)) {
1063             error_report("Device '%s' is writable but does not support snapshots.",
1064                                bdrv_get_device_name(bs));
1065             return -ENOTSUP;
1066         }
1067
1068         ret = bdrv_snapshot_find(bs, &sn, name);
1069         if (ret < 0) {
1070             error_report("Device '%s' does not have the requested snapshot '%s'",
1071                            bdrv_get_device_name(bs), name);
1072             return ret;
1073         }
1074     }
1075
1076     /* Flush all IO requests so they don't interfere with the new state.  */
1077     bdrv_drain_all();
1078
1079     bs = NULL;
1080     while ((bs = bdrv_next(bs))) {
1081         if (bdrv_can_snapshot(bs)) {
1082             ret = bdrv_snapshot_goto(bs, name);
1083             if (ret < 0) {
1084                 error_report("Error %d while activating snapshot '%s' on '%s'",
1085                              ret, name, bdrv_get_device_name(bs));
1086                 return ret;
1087             }
1088         }
1089     }
1090
1091     /* restore the VM state */
1092     f = qemu_fopen_bdrv(bs_vm_state, 0);
1093     if (!f) {
1094         error_report("Could not open VM state file");
1095         return -EINVAL;
1096     }
1097
1098     qemu_system_reset(VMRESET_SILENT);
1099     ret = qemu_loadvm_state(f);
1100
1101     qemu_fclose(f);
1102     if (ret < 0) {
1103         error_report("Error %d while loading VM state", ret);
1104         return ret;
1105     }
1106
1107     return 0;
1108 }
1109
1110 void do_delvm(Monitor *mon, const QDict *qdict)
1111 {
1112     BlockDriverState *bs, *bs1;
1113     Error *err = NULL;
1114     const char *name = qdict_get_str(qdict, "name");
1115
1116     bs = find_vmstate_bs();
1117     if (!bs) {
1118         monitor_printf(mon, "No block device supports snapshots\n");
1119         return;
1120     }
1121
1122     bs1 = NULL;
1123     while ((bs1 = bdrv_next(bs1))) {
1124         if (bdrv_can_snapshot(bs1)) {
1125             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
1126             if (err) {
1127                 monitor_printf(mon,
1128                                "Error while deleting snapshot on device '%s':"
1129                                " %s\n",
1130                                bdrv_get_device_name(bs),
1131                                error_get_pretty(err));
1132                 error_free(err);
1133             }
1134         }
1135     }
1136 }
1137
1138 void do_info_snapshots(Monitor *mon, const QDict *qdict)
1139 {
1140     BlockDriverState *bs, *bs1;
1141     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
1142     int nb_sns, i, ret, available;
1143     int total;
1144     int *available_snapshots;
1145
1146     bs = find_vmstate_bs();
1147     if (!bs) {
1148         monitor_printf(mon, "No available block device supports snapshots\n");
1149         return;
1150     }
1151
1152     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1153     if (nb_sns < 0) {
1154         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1155         return;
1156     }
1157
1158     if (nb_sns == 0) {
1159         monitor_printf(mon, "There is no snapshot available.\n");
1160         return;
1161     }
1162
1163     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
1164     total = 0;
1165     for (i = 0; i < nb_sns; i++) {
1166         sn = &sn_tab[i];
1167         available = 1;
1168         bs1 = NULL;
1169
1170         while ((bs1 = bdrv_next(bs1))) {
1171             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
1172                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
1173                 if (ret < 0) {
1174                     available = 0;
1175                     break;
1176                 }
1177             }
1178         }
1179
1180         if (available) {
1181             available_snapshots[total] = i;
1182             total++;
1183         }
1184     }
1185
1186     if (total > 0) {
1187         bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, NULL);
1188         monitor_printf(mon, "\n");
1189         for (i = 0; i < total; i++) {
1190             sn = &sn_tab[available_snapshots[i]];
1191             bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, sn);
1192             monitor_printf(mon, "\n");
1193         }
1194     } else {
1195         monitor_printf(mon, "There is no suitable snapshot available\n");
1196     }
1197
1198     g_free(sn_tab);
1199     g_free(available_snapshots);
1200
1201 }
1202
1203 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
1204 {
1205     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
1206                        memory_region_name(mr), dev);
1207 }
1208
1209 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
1210 {
1211     /* Nothing do to while the implementation is in RAMBlock */
1212 }
1213
1214 void vmstate_register_ram_global(MemoryRegion *mr)
1215 {
1216     vmstate_register_ram(mr, NULL);
1217 }