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_setg(errp, "State blocked by non-migratable device '%s'",
 457                        se->idstr);
 458             return true;
 459         }
 460     }
 461     return false;
 462 }
 463
 464 void qemu_savevm_state_begin(QEMUFile *f,
 465                              const MigrationParams *params)
 466 {
 467     SaveStateEntry *se;
 468     int ret;
 469
 470     trace_savevm_state_begin();
 471     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 472         if (!se->ops || !se->ops->set_params) {
 473             continue;
 474         }
 475         se->ops->set_params(params, se->opaque);
 476     }
 477
 478     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 479     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 480
 481     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 482         int len;
 483
 484         if (!se->ops || !se->ops->save_live_setup) {
 485             continue;
 486         }
 487         if (se->ops && se->ops->is_active) {
 488             if (!se->ops->is_active(se->opaque)) {
 489                 continue;
 490             }
 491         }
 492         /* Section type */
 493         qemu_put_byte(f, QEMU_VM_SECTION_START);
 494         qemu_put_be32(f, se->section_id);
 495
 496         /* ID string */
 497         len = strlen(se->idstr);
 498         qemu_put_byte(f, len);
 499         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 500
 501         qemu_put_be32(f, se->instance_id);
 502         qemu_put_be32(f, se->version_id);
 503
 504         ret = se->ops->save_live_setup(f, se->opaque);
 505         if (ret < 0) {
 506             qemu_file_set_error(f, ret);
 507             break;
 508         }
 509     }
 510 }
 511
 512 /*
 513  * this function has three return values:
 514  *   negative: there was one error, and we have -errno.
 515  *   0 : We haven't finished, caller have to go again
 516  *   1 : We have finished, we can go to complete phase
 517  */
 518 int qemu_savevm_state_iterate(QEMUFile *f)
 519 {
 520     SaveStateEntry *se;
 521     int ret = 1;
 522
 523     trace_savevm_state_iterate();
 524     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 525         if (!se->ops || !se->ops->save_live_iterate) {
 526             continue;
 527         }
 528         if (se->ops && se->ops->is_active) {
 529             if (!se->ops->is_active(se->opaque)) {
 530                 continue;
 531             }
 532         }
 533         if (qemu_file_rate_limit(f)) {
 534             return 0;
 535         }
 536         trace_savevm_section_start(se->idstr, se->section_id);
 537         /* Section type */
 538         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 539         qemu_put_be32(f, se->section_id);
 540
 541         ret = se->ops->save_live_iterate(f, se->opaque);
 542         trace_savevm_section_end(se->idstr, se->section_id);
 543
 544         if (ret < 0) {
 545             qemu_file_set_error(f, ret);
 546         }
 547         if (ret <= 0) {
 548             /* Do not proceed to the next vmstate before this one reported
 549                completion of the current stage. This serializes the migration
 550                and reduces the probability that a faster changing state is
 551                synchronized over and over again. */
 552             break;
 553         }
 554     }
 555     return ret;
 556 }
 557
 558 void qemu_savevm_state_complete(QEMUFile *f)
 559 {
 560     SaveStateEntry *se;
 561     int ret;
 562
 563     trace_savevm_state_complete();
 564
 565     cpu_synchronize_all_states();
 566
 567     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 568         if (!se->ops || !se->ops->save_live_complete) {
 569             continue;
 570         }
 571         if (se->ops && se->ops->is_active) {
 572             if (!se->ops->is_active(se->opaque)) {
 573                 continue;
 574             }
 575         }
 576         trace_savevm_section_start(se->idstr, se->section_id);
 577         /* Section type */
 578         qemu_put_byte(f, QEMU_VM_SECTION_END);
 579         qemu_put_be32(f, se->section_id);
 580
 581         ret = se->ops->save_live_complete(f, se->opaque);
 582         trace_savevm_section_end(se->idstr, se->section_id);
 583         if (ret < 0) {
 584             qemu_file_set_error(f, ret);
 585             return;
 586         }
 587     }
 588
 589     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 590         int len;
 591
 592         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
 593             continue;
 594         }
 595         trace_savevm_section_start(se->idstr, se->section_id);
 596         /* Section type */
 597         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 598         qemu_put_be32(f, se->section_id);
 599
 600         /* ID string */
 601         len = strlen(se->idstr);
 602         qemu_put_byte(f, len);
 603         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 604
 605         qemu_put_be32(f, se->instance_id);
 606         qemu_put_be32(f, se->version_id);
 607
 608         vmstate_save(f, se);
 609         trace_savevm_section_end(se->idstr, se->section_id);
 610     }
 611
 612     qemu_put_byte(f, QEMU_VM_EOF);
 613     qemu_fflush(f);
 614 }
 615
 616 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
 617 {
 618     SaveStateEntry *se;
 619     uint64_t ret = 0;
 620
 621     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 622         if (!se->ops || !se->ops->save_live_pending) {
 623             continue;
 624         }
 625         if (se->ops && se->ops->is_active) {
 626             if (!se->ops->is_active(se->opaque)) {
 627                 continue;
 628             }
 629         }
 630         ret += se->ops->save_live_pending(f, se->opaque, max_size);
 631     }
 632     return ret;
 633 }
 634
 635 void qemu_savevm_state_cancel(void)
 636 {
 637     SaveStateEntry *se;
 638
 639     trace_savevm_state_cancel();
 640     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 641         if (se->ops && se->ops->cancel) {
 642             se->ops->cancel(se->opaque);
 643         }
 644     }
 645 }
 646
 647 static int qemu_savevm_state(QEMUFile *f)
 648 {
 649     int ret;
 650     MigrationParams params = {
 651         .blk = 0,
 652         .shared = 0
 653     };
 654
 655     if (qemu_savevm_state_blocked(NULL)) {
 656         return -EINVAL;
 657     }
 658
 659     qemu_mutex_unlock_iothread();
 660     qemu_savevm_state_begin(f, &params);
 661     qemu_mutex_lock_iothread();
 662
 663     while (qemu_file_get_error(f) == 0) {
 664         if (qemu_savevm_state_iterate(f) > 0) {
 665             break;
 666         }
 667     }
 668
 669     ret = qemu_file_get_error(f);
 670     if (ret == 0) {
 671         qemu_savevm_state_complete(f);
 672         ret = qemu_file_get_error(f);
 673     }
 674     if (ret != 0) {
 675         qemu_savevm_state_cancel();
 676     }
 677     return ret;
 678 }
 679
 680 static int qemu_save_device_state(QEMUFile *f)
 681 {
 682     SaveStateEntry *se;
 683
 684     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 685     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 686
 687     cpu_synchronize_all_states();
 688
 689     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 690         int len;
 691
 692         if (se->is_ram) {
 693             continue;
 694         }
 695         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
 696             continue;
 697         }
 698
 699         /* Section type */
 700         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 701         qemu_put_be32(f, se->section_id);
 702
 703         /* ID string */
 704         len = strlen(se->idstr);
 705         qemu_put_byte(f, len);
 706         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 707
 708         qemu_put_be32(f, se->instance_id);
 709         qemu_put_be32(f, se->version_id);
 710
 711         vmstate_save(f, se);
 712     }
 713
 714     qemu_put_byte(f, QEMU_VM_EOF);
 715
 716     return qemu_file_get_error(f);
 717 }
 718
 719 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 720 {
 721     SaveStateEntry *se;
 722
 723     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 724         if (!strcmp(se->idstr, idstr) &&
 725             (instance_id == se->instance_id ||
 726              instance_id == se->alias_id))
 727             return se;
 728         /* Migrating from an older version? */
 729         if (strstr(se->idstr, idstr) && se->compat) {
 730             if (!strcmp(se->compat->idstr, idstr) &&
 731                 (instance_id == se->compat->instance_id ||
 732                  instance_id == se->alias_id))
 733                 return se;
 734         }
 735     }
 736     return NULL;
 737 }
 738
 739 typedef struct LoadStateEntry {
 740     QLIST_ENTRY(LoadStateEntry) entry;
 741     SaveStateEntry *se;
 742     int section_id;
 743     int version_id;
 744 } LoadStateEntry;
 745
 746 int qemu_loadvm_state(QEMUFile *f)
 747 {
 748     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
 749         QLIST_HEAD_INITIALIZER(loadvm_handlers);
 750     LoadStateEntry *le, *new_le;
 751     uint8_t section_type;
 752     unsigned int v;
 753     int ret;
 754
 755     if (qemu_savevm_state_blocked(NULL)) {
 756         return -EINVAL;
 757     }
 758
 759     v = qemu_get_be32(f);
 760     if (v != QEMU_VM_FILE_MAGIC) {
 761         return -EINVAL;
 762     }
 763
 764     v = qemu_get_be32(f);
 765     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
 766         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
 767         return -ENOTSUP;
 768     }
 769     if (v != QEMU_VM_FILE_VERSION) {
 770         return -ENOTSUP;
 771     }
 772
 773     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 774         uint32_t instance_id, version_id, section_id;
 775         SaveStateEntry *se;
 776         char idstr[257];
 777         int len;
 778
 779         switch (section_type) {
 780         case QEMU_VM_SECTION_START:
 781         case QEMU_VM_SECTION_FULL:
 782             /* Read section start */
 783             section_id = qemu_get_be32(f);
 784             len = qemu_get_byte(f);
 785             qemu_get_buffer(f, (uint8_t *)idstr, len);
 786             idstr[len] = 0;
 787             instance_id = qemu_get_be32(f);
 788             version_id = qemu_get_be32(f);
 789
 790             /* Find savevm section */
 791             se = find_se(idstr, instance_id);
 792             if (se == NULL) {
 793                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 794                 ret = -EINVAL;
 795                 goto out;
 796             }
 797
 798             /* Validate version */
 799             if (version_id > se->version_id) {
 800                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 801                         version_id, idstr, se->version_id);
 802                 ret = -EINVAL;
 803                 goto out;
 804             }
 805
 806             /* Add entry */
 807             le = g_malloc0(sizeof(*le));
 808
 809             le->se = se;
 810             le->section_id = section_id;
 811             le->version_id = version_id;
 812             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
 813
 814             ret = vmstate_load(f, le->se, le->version_id);
 815             if (ret < 0) {
 816                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 817                         instance_id, idstr);
 818                 goto out;
 819             }
 820             break;
 821         case QEMU_VM_SECTION_PART:
 822         case QEMU_VM_SECTION_END:
 823             section_id = qemu_get_be32(f);
 824
 825             QLIST_FOREACH(le, &loadvm_handlers, entry) {
 826                 if (le->section_id == section_id) {
 827                     break;
 828                 }
 829             }
 830             if (le == NULL) {
 831                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 832                 ret = -EINVAL;
 833                 goto out;
 834             }
 835
 836             ret = vmstate_load(f, le->se, le->version_id);
 837             if (ret < 0) {
 838                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
 839                         section_id);
 840                 goto out;
 841             }
 842             break;
 843         default:
 844             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 845             ret = -EINVAL;
 846             goto out;
 847         }
 848     }
 849
 850     cpu_synchronize_all_post_init();
 851
 852     ret = 0;
 853
 854 out:
 855     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
 856         QLIST_REMOVE(le, entry);
 857         g_free(le);
 858     }
 859
 860     if (ret == 0) {
 861         ret = qemu_file_get_error(f);
 862     }
 863
 864     return ret;
 865 }
 866
 867 static BlockDriverState *find_vmstate_bs(void)
 868 {
 869     BlockDriverState *bs = NULL;
 870     while ((bs = bdrv_next(bs))) {
 871         if (bdrv_can_snapshot(bs)) {
 872             return bs;
 873         }
 874     }
 875     return NULL;
 876 }
 877
 878 /*
 879  * Deletes snapshots of a given name in all opened images.
 880  */
 881 static int del_existing_snapshots(Monitor *mon, const char *name)
 882 {
 883     BlockDriverState *bs;
 884     QEMUSnapshotInfo sn1, *snapshot = &sn1;
 885     Error *err = NULL;
 886
 887     bs = NULL;
 888     while ((bs = bdrv_next(bs))) {
 889         if (bdrv_can_snapshot(bs) &&
 890             bdrv_snapshot_find(bs, snapshot, name) >= 0) {
 891             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
 892             if (err) {
 893                 monitor_printf(mon,
 894                                "Error while deleting snapshot on device '%s':"
 895                                " %s\n",
 896                                bdrv_get_device_name(bs),
 897                                error_get_pretty(err));
 898                 error_free(err);
 899                 return -1;
 900             }
 901         }
 902     }
 903
 904     return 0;
 905 }
 906
 907 void do_savevm(Monitor *mon, const QDict *qdict)
 908 {
 909     BlockDriverState *bs, *bs1;
 910     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
 911     int ret;
 912     QEMUFile *f;
 913     int saved_vm_running;
 914     uint64_t vm_state_size;
 915     qemu_timeval tv;
 916     struct tm tm;
 917     const char *name = qdict_get_try_str(qdict, "name");
 918
 919     /* Verify if there is a device that doesn't support snapshots and is writable */
 920     bs = NULL;
 921     while ((bs = bdrv_next(bs))) {
 922
 923         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
 924             continue;
 925         }
 926
 927         if (!bdrv_can_snapshot(bs)) {
 928             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
 929                                bdrv_get_device_name(bs));
 930             return;
 931         }
 932     }
 933
 934     bs = find_vmstate_bs();
 935     if (!bs) {
 936         monitor_printf(mon, "No block device can accept snapshots\n");
 937         return;
 938     }
 939
 940     saved_vm_running = runstate_is_running();
 941     vm_stop(RUN_STATE_SAVE_VM);
 942
 943     memset(sn, 0, sizeof(*sn));
 944
 945     /* fill auxiliary fields */
 946     qemu_gettimeofday(&tv);
 947     sn->date_sec = tv.tv_sec;
 948     sn->date_nsec = tv.tv_usec * 1000;
 949     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
 950
 951     if (name) {
 952         ret = bdrv_snapshot_find(bs, old_sn, name);
 953         if (ret >= 0) {
 954             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
 955             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
 956         } else {
 957             pstrcpy(sn->name, sizeof(sn->name), name);
 958         }
 959     } else {
 960         /* cast below needed for OpenBSD where tv_sec is still 'long' */
 961         localtime_r((const time_t *)&tv.tv_sec, &tm);
 962         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
 963     }
 964
 965     /* Delete old snapshots of the same name */
 966     if (name && del_existing_snapshots(mon, name) < 0) {
 967         goto the_end;
 968     }
 969
 970     /* save the VM state */
 971     f = qemu_fopen_bdrv(bs, 1);
 972     if (!f) {
 973         monitor_printf(mon, "Could not open VM state file\n");
 974         goto the_end;
 975     }
 976     ret = qemu_savevm_state(f);
 977     vm_state_size = qemu_ftell(f);
 978     qemu_fclose(f);
 979     if (ret < 0) {
 980         monitor_printf(mon, "Error %d while writing VM\n", ret);
 981         goto the_end;
 982     }
 983
 984     /* create the snapshots */
 985
 986     bs1 = NULL;
 987     while ((bs1 = bdrv_next(bs1))) {
 988         if (bdrv_can_snapshot(bs1)) {
 989             /* Write VM state size only to the image that contains the state */
 990             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
 991             ret = bdrv_snapshot_create(bs1, sn);
 992             if (ret < 0) {
 993                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
 994                                bdrv_get_device_name(bs1));
 995             }
 996         }
 997     }
 998
 999  the_end:
1000     if (saved_vm_running) {
1001         vm_start();
1002     }
1003 }
1004
1005 void qmp_xen_save_devices_state(const char *filename, Error **errp)
1006 {
1007     QEMUFile *f;
1008     int saved_vm_running;
1009     int ret;
1010
1011     saved_vm_running = runstate_is_running();
1012     vm_stop(RUN_STATE_SAVE_VM);
1013
1014     f = qemu_fopen(filename, "wb");
1015     if (!f) {
1016         error_setg_file_open(errp, errno, filename);
1017         goto the_end;
1018     }
1019     ret = qemu_save_device_state(f);
1020     qemu_fclose(f);
1021     if (ret < 0) {
1022         error_set(errp, QERR_IO_ERROR);
1023     }
1024
1025  the_end:
1026     if (saved_vm_running) {
1027         vm_start();
1028     }
1029 }
1030
1031 int load_vmstate(const char *name)
1032 {
1033     BlockDriverState *bs, *bs_vm_state;
1034     QEMUSnapshotInfo sn;
1035     QEMUFile *f;
1036     int ret;
1037
1038     bs_vm_state = find_vmstate_bs();
1039     if (!bs_vm_state) {
1040         error_report("No block device supports snapshots");
1041         return -ENOTSUP;
1042     }
1043
1044     /* Don't even try to load empty VM states */
1045     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
1046     if (ret < 0) {
1047         return ret;
1048     } else if (sn.vm_state_size == 0) {
1049         error_report("This is a disk-only snapshot. Revert to it offline "
1050             "using qemu-img.");
1051         return -EINVAL;
1052     }
1053
1054     /* Verify if there is any device that doesn't support snapshots and is
1055     writable and check if the requested snapshot is available too. */
1056     bs = NULL;
1057     while ((bs = bdrv_next(bs))) {
1058
1059         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
1060             continue;
1061         }
1062
1063         if (!bdrv_can_snapshot(bs)) {
1064             error_report("Device '%s' is writable but does not support snapshots.",
1065                                bdrv_get_device_name(bs));
1066             return -ENOTSUP;
1067         }
1068
1069         ret = bdrv_snapshot_find(bs, &sn, name);
1070         if (ret < 0) {
1071             error_report("Device '%s' does not have the requested snapshot '%s'",
1072                            bdrv_get_device_name(bs), name);
1073             return ret;
1074         }
1075     }
1076
1077     /* Flush all IO requests so they don't interfere with the new state.  */
1078     bdrv_drain_all();
1079
1080     bs = NULL;
1081     while ((bs = bdrv_next(bs))) {
1082         if (bdrv_can_snapshot(bs)) {
1083             ret = bdrv_snapshot_goto(bs, name);
1084             if (ret < 0) {
1085                 error_report("Error %d while activating snapshot '%s' on '%s'",
1086                              ret, name, bdrv_get_device_name(bs));
1087                 return ret;
1088             }
1089         }
1090     }
1091
1092     /* restore the VM state */
1093     f = qemu_fopen_bdrv(bs_vm_state, 0);
1094     if (!f) {
1095         error_report("Could not open VM state file");
1096         return -EINVAL;
1097     }
1098
1099     qemu_system_reset(VMRESET_SILENT);
1100     ret = qemu_loadvm_state(f);
1101
1102     qemu_fclose(f);
1103     if (ret < 0) {
1104         error_report("Error %d while loading VM state", ret);
1105         return ret;
1106     }
1107
1108     return 0;
1109 }
1110
1111 void do_delvm(Monitor *mon, const QDict *qdict)
1112 {
1113     BlockDriverState *bs, *bs1;
1114     Error *err = NULL;
1115     const char *name = qdict_get_str(qdict, "name");
1116
1117     bs = find_vmstate_bs();
1118     if (!bs) {
1119         monitor_printf(mon, "No block device supports snapshots\n");
1120         return;
1121     }
1122
1123     bs1 = NULL;
1124     while ((bs1 = bdrv_next(bs1))) {
1125         if (bdrv_can_snapshot(bs1)) {
1126             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
1127             if (err) {
1128                 monitor_printf(mon,
1129                                "Error while deleting snapshot on device '%s':"
1130                                " %s\n",
1131                                bdrv_get_device_name(bs),
1132                                error_get_pretty(err));
1133                 error_free(err);
1134             }
1135         }
1136     }
1137 }
1138
1139 void do_info_snapshots(Monitor *mon, const QDict *qdict)
1140 {
1141     BlockDriverState *bs, *bs1;
1142     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
1143     int nb_sns, i, ret, available;
1144     int total;
1145     int *available_snapshots;
1146
1147     bs = find_vmstate_bs();
1148     if (!bs) {
1149         monitor_printf(mon, "No available block device supports snapshots\n");
1150         return;
1151     }
1152
1153     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1154     if (nb_sns < 0) {
1155         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1156         return;
1157     }
1158
1159     if (nb_sns == 0) {
1160         monitor_printf(mon, "There is no snapshot available.\n");
1161         return;
1162     }
1163
1164     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
1165     total = 0;
1166     for (i = 0; i < nb_sns; i++) {
1167         sn = &sn_tab[i];
1168         available = 1;
1169         bs1 = NULL;
1170
1171         while ((bs1 = bdrv_next(bs1))) {
1172             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
1173                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
1174                 if (ret < 0) {
1175                     available = 0;
1176                     break;
1177                 }
1178             }
1179         }
1180
1181         if (available) {
1182             available_snapshots[total] = i;
1183             total++;
1184         }
1185     }
1186
1187     if (total > 0) {
1188         bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, NULL);
1189         monitor_printf(mon, "\n");
1190         for (i = 0; i < total; i++) {
1191             sn = &sn_tab[available_snapshots[i]];
1192             bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, sn);
1193             monitor_printf(mon, "\n");
1194         }
1195     } else {
1196         monitor_printf(mon, "There is no suitable snapshot available\n");
1197     }
1198
1199     g_free(sn_tab);
1200     g_free(available_snapshots);
1201
1202 }
1203
1204 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
1205 {
1206     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
1207                        memory_region_name(mr), dev);
1208 }
1209
1210 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
1211 {
1212     /* Nothing do to while the implementation is in RAMBlock */
1213 }
1214
1215 void vmstate_register_ram_global(MemoryRegion *mr)
1216 {
1217     vmstate_register_ram(mr, NULL);
1218 }