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/boards.h"
  28 #include "hw/hw.h"
  29 #include "hw/qdev.h"
  30 #include "net/net.h"
  31 #include "monitor/monitor.h"
  32 #include "sysemu/sysemu.h"
  33 #include "qemu/timer.h"
  34 #include "audio/audio.h"
  35 #include "migration/migration.h"
  36 #include "qemu/sockets.h"
  37 #include "qemu/queue.h"
  38 #include "sysemu/cpus.h"
  39 #include "exec/memory.h"
  40 #include "qmp-commands.h"
  41 #include "trace.h"
  42 #include "qemu/iov.h"
  43 #include "block/snapshot.h"
  44 #include "block/qapi.h"
  45
  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                   self_announce_delay(count));
 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 void dump_vmstate_vmsd(FILE *out_file,
 245                               const VMStateDescription *vmsd, int indent,
 246                               bool is_subsection);
 247
 248 static void dump_vmstate_vmsf(FILE *out_file, const VMStateField *field,
 249                               int indent)
 250 {
 251     fprintf(out_file, "%*s{\n", indent, "");
 252     indent += 2;
 253     fprintf(out_file, "%*s\"field\": \"%s\",\n", indent, "", field->name);
 254     fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "",
 255             field->version_id);
 256     fprintf(out_file, "%*s\"field_exists\": %s,\n", indent, "",
 257             field->field_exists ? "true" : "false");
 258     fprintf(out_file, "%*s\"size\": %zu", indent, "", field->size);
 259     if (field->vmsd != NULL) {
 260         fprintf(out_file, ",\n");
 261         dump_vmstate_vmsd(out_file, field->vmsd, indent, false);
 262     }
 263     fprintf(out_file, "\n%*s}", indent - 2, "");
 264 }
 265
 266 static void dump_vmstate_vmss(FILE *out_file,
 267                               const VMStateSubsection *subsection,
 268                               int indent)
 269 {
 270     if (subsection->vmsd != NULL) {
 271         dump_vmstate_vmsd(out_file, subsection->vmsd, indent, true);
 272     }
 273 }
 274
 275 static void dump_vmstate_vmsd(FILE *out_file,
 276                               const VMStateDescription *vmsd, int indent,
 277                               bool is_subsection)
 278 {
 279     if (is_subsection) {
 280         fprintf(out_file, "%*s{\n", indent, "");
 281     } else {
 282         fprintf(out_file, "%*s\"%s\": {\n", indent, "", "Description");
 283     }
 284     indent += 2;
 285     fprintf(out_file, "%*s\"name\": \"%s\",\n", indent, "", vmsd->name);
 286     fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "",
 287             vmsd->version_id);
 288     fprintf(out_file, "%*s\"minimum_version_id\": %d", indent, "",
 289             vmsd->minimum_version_id);
 290     if (vmsd->fields != NULL) {
 291         const VMStateField *field = vmsd->fields;
 292         bool first;
 293
 294         fprintf(out_file, ",\n%*s\"Fields\": [\n", indent, "");
 295         first = true;
 296         while (field->name != NULL) {
 297             if (field->flags & VMS_MUST_EXIST) {
 298                 /* Ignore VMSTATE_VALIDATE bits; these don't get migrated */
 299                 field++;
 300                 continue;
 301             }
 302             if (!first) {
 303                 fprintf(out_file, ",\n");
 304             }
 305             dump_vmstate_vmsf(out_file, field, indent + 2);
 306             field++;
 307             first = false;
 308         }
 309         fprintf(out_file, "\n%*s]", indent, "");
 310     }
 311     if (vmsd->subsections != NULL) {
 312         const VMStateSubsection *subsection = vmsd->subsections;
 313         bool first;
 314
 315         fprintf(out_file, ",\n%*s\"Subsections\": [\n", indent, "");
 316         first = true;
 317         while (subsection->vmsd != NULL) {
 318             if (!first) {
 319                 fprintf(out_file, ",\n");
 320             }
 321             dump_vmstate_vmss(out_file, subsection, indent + 2);
 322             subsection++;
 323             first = false;
 324         }
 325         fprintf(out_file, "\n%*s]", indent, "");
 326     }
 327     fprintf(out_file, "\n%*s}", indent - 2, "");
 328 }
 329
 330 static void dump_machine_type(FILE *out_file)
 331 {
 332     MachineClass *mc;
 333
 334     mc = MACHINE_GET_CLASS(current_machine);
 335
 336     fprintf(out_file, "  \"vmschkmachine\": {\n");
 337     fprintf(out_file, "    \"Name\": \"%s\"\n", mc->name);
 338     fprintf(out_file, "  },\n");
 339 }
 340
 341 void dump_vmstate_json_to_file(FILE *out_file)
 342 {
 343     GSList *list, *elt;
 344     bool first;
 345
 346     fprintf(out_file, "{\n");
 347     dump_machine_type(out_file);
 348
 349     first = true;
 350     list = object_class_get_list(TYPE_DEVICE, true);
 351     for (elt = list; elt; elt = elt->next) {
 352         DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, elt->data,
 353                                              TYPE_DEVICE);
 354         const char *name;
 355         int indent = 2;
 356
 357         if (!dc->vmsd) {
 358             continue;
 359         }
 360
 361         if (!first) {
 362             fprintf(out_file, ",\n");
 363         }
 364         name = object_class_get_name(OBJECT_CLASS(dc));
 365         fprintf(out_file, "%*s\"%s\": {\n", indent, "", name);
 366         indent += 2;
 367         fprintf(out_file, "%*s\"Name\": \"%s\",\n", indent, "", name);
 368         fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "",
 369                 dc->vmsd->version_id);
 370         fprintf(out_file, "%*s\"minimum_version_id\": %d,\n", indent, "",
 371                 dc->vmsd->minimum_version_id);
 372
 373         dump_vmstate_vmsd(out_file, dc->vmsd, indent, false);
 374
 375         fprintf(out_file, "\n%*s}", indent - 2, "");
 376         first = false;
 377     }
 378     fprintf(out_file, "\n}\n");
 379     fclose(out_file);
 380 }
 381
 382 static int calculate_new_instance_id(const char *idstr)
 383 {
 384     SaveStateEntry *se;
 385     int instance_id = 0;
 386
 387     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 388         if (strcmp(idstr, se->idstr) == 0
 389             && instance_id <= se->instance_id) {
 390             instance_id = se->instance_id + 1;
 391         }
 392     }
 393     return instance_id;
 394 }
 395
 396 static int calculate_compat_instance_id(const char *idstr)
 397 {
 398     SaveStateEntry *se;
 399     int instance_id = 0;
 400
 401     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 402         if (!se->compat) {
 403             continue;
 404         }
 405
 406         if (strcmp(idstr, se->compat->idstr) == 0
 407             && instance_id <= se->compat->instance_id) {
 408             instance_id = se->compat->instance_id + 1;
 409         }
 410     }
 411     return instance_id;
 412 }
 413
 414 /* TODO: Individual devices generally have very little idea about the rest
 415    of the system, so instance_id should be removed/replaced.
 416    Meanwhile pass -1 as instance_id if you do not already have a clearly
 417    distinguishing id for all instances of your device class. */
 418 int register_savevm_live(DeviceState *dev,
 419                          const char *idstr,
 420                          int instance_id,
 421                          int version_id,
 422                          SaveVMHandlers *ops,
 423                          void *opaque)
 424 {
 425     SaveStateEntry *se;
 426
 427     se = g_malloc0(sizeof(SaveStateEntry));
 428     se->version_id = version_id;
 429     se->section_id = global_section_id++;
 430     se->ops = ops;
 431     se->opaque = opaque;
 432     se->vmsd = NULL;
 433     se->no_migrate = 0;
 434     /* if this is a live_savem then set is_ram */
 435     if (ops->save_live_setup != NULL) {
 436         se->is_ram = 1;
 437     }
 438
 439     if (dev) {
 440         char *id = qdev_get_dev_path(dev);
 441         if (id) {
 442             pstrcpy(se->idstr, sizeof(se->idstr), id);
 443             pstrcat(se->idstr, sizeof(se->idstr), "/");
 444             g_free(id);
 445
 446             se->compat = g_malloc0(sizeof(CompatEntry));
 447             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
 448             se->compat->instance_id = instance_id == -1 ?
 449                          calculate_compat_instance_id(idstr) : instance_id;
 450             instance_id = -1;
 451         }
 452     }
 453     pstrcat(se->idstr, sizeof(se->idstr), idstr);
 454
 455     if (instance_id == -1) {
 456         se->instance_id = calculate_new_instance_id(se->idstr);
 457     } else {
 458         se->instance_id = instance_id;
 459     }
 460     assert(!se->compat || se->instance_id == 0);
 461     /* add at the end of list */
 462     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
 463     return 0;
 464 }
 465
 466 int register_savevm(DeviceState *dev,
 467                     const char *idstr,
 468                     int instance_id,
 469                     int version_id,
 470                     SaveStateHandler *save_state,
 471                     LoadStateHandler *load_state,
 472                     void *opaque)
 473 {
 474     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
 475     ops->save_state = save_state;
 476     ops->load_state = load_state;
 477     return register_savevm_live(dev, idstr, instance_id, version_id,
 478                                 ops, opaque);
 479 }
 480
 481 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
 482 {
 483     SaveStateEntry *se, *new_se;
 484     char id[256] = "";
 485
 486     if (dev) {
 487         char *path = qdev_get_dev_path(dev);
 488         if (path) {
 489             pstrcpy(id, sizeof(id), path);
 490             pstrcat(id, sizeof(id), "/");
 491             g_free(path);
 492         }
 493     }
 494     pstrcat(id, sizeof(id), idstr);
 495
 496     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
 497         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
 498             QTAILQ_REMOVE(&savevm_handlers, se, entry);
 499             if (se->compat) {
 500                 g_free(se->compat);
 501             }
 502             g_free(se->ops);
 503             g_free(se);
 504         }
 505     }
 506 }
 507
 508 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
 509                                    const VMStateDescription *vmsd,
 510                                    void *opaque, int alias_id,
 511                                    int required_for_version)
 512 {
 513     SaveStateEntry *se;
 514
 515     /* If this triggers, alias support can be dropped for the vmsd. */
 516     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
 517
 518     se = g_malloc0(sizeof(SaveStateEntry));
 519     se->version_id = vmsd->version_id;
 520     se->section_id = global_section_id++;
 521     se->opaque = opaque;
 522     se->vmsd = vmsd;
 523     se->alias_id = alias_id;
 524     se->no_migrate = vmsd->unmigratable;
 525
 526     if (dev) {
 527         char *id = qdev_get_dev_path(dev);
 528         if (id) {
 529             pstrcpy(se->idstr, sizeof(se->idstr), id);
 530             pstrcat(se->idstr, sizeof(se->idstr), "/");
 531             g_free(id);
 532
 533             se->compat = g_malloc0(sizeof(CompatEntry));
 534             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
 535             se->compat->instance_id = instance_id == -1 ?
 536                          calculate_compat_instance_id(vmsd->name) : instance_id;
 537             instance_id = -1;
 538         }
 539     }
 540     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
 541
 542     if (instance_id == -1) {
 543         se->instance_id = calculate_new_instance_id(se->idstr);
 544     } else {
 545         se->instance_id = instance_id;
 546     }
 547     assert(!se->compat || se->instance_id == 0);
 548     /* add at the end of list */
 549     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
 550     return 0;
 551 }
 552
 553 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
 554                         void *opaque)
 555 {
 556     SaveStateEntry *se, *new_se;
 557
 558     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
 559         if (se->vmsd == vmsd && se->opaque == opaque) {
 560             QTAILQ_REMOVE(&savevm_handlers, se, entry);
 561             if (se->compat) {
 562                 g_free(se->compat);
 563             }
 564             g_free(se);
 565         }
 566     }
 567 }
 568
 569 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
 570 {
 571     trace_vmstate_load(se->idstr, se->vmsd ? se->vmsd->name : "(old)");
 572     if (!se->vmsd) {         /* Old style */
 573         return se->ops->load_state(f, se->opaque, version_id);
 574     }
 575     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
 576 }
 577
 578 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
 579 {
 580     trace_vmstate_save(se->idstr, se->vmsd ? se->vmsd->name : "(old)");
 581     if (!se->vmsd) {         /* Old style */
 582         se->ops->save_state(f, se->opaque);
 583         return;
 584     }
 585     vmstate_save_state(f, se->vmsd, se->opaque);
 586 }
 587
 588 bool qemu_savevm_state_blocked(Error **errp)
 589 {
 590     SaveStateEntry *se;
 591
 592     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 593         if (se->no_migrate) {
 594             error_setg(errp, "State blocked by non-migratable device '%s'",
 595                        se->idstr);
 596             return true;
 597         }
 598     }
 599     return false;
 600 }
 601
 602 void qemu_savevm_state_begin(QEMUFile *f,
 603                              const MigrationParams *params)
 604 {
 605     SaveStateEntry *se;
 606     int ret;
 607
 608     trace_savevm_state_begin();
 609     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 610         if (!se->ops || !se->ops->set_params) {
 611             continue;
 612         }
 613         se->ops->set_params(params, se->opaque);
 614     }
 615
 616     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 617     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 618
 619     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 620         int len;
 621
 622         if (!se->ops || !se->ops->save_live_setup) {
 623             continue;
 624         }
 625         if (se->ops && se->ops->is_active) {
 626             if (!se->ops->is_active(se->opaque)) {
 627                 continue;
 628             }
 629         }
 630         /* Section type */
 631         qemu_put_byte(f, QEMU_VM_SECTION_START);
 632         qemu_put_be32(f, se->section_id);
 633
 634         /* ID string */
 635         len = strlen(se->idstr);
 636         qemu_put_byte(f, len);
 637         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 638
 639         qemu_put_be32(f, se->instance_id);
 640         qemu_put_be32(f, se->version_id);
 641
 642         ret = se->ops->save_live_setup(f, se->opaque);
 643         if (ret < 0) {
 644             qemu_file_set_error(f, ret);
 645             break;
 646         }
 647     }
 648 }
 649
 650 /*
 651  * this function has three return values:
 652  *   negative: there was one error, and we have -errno.
 653  *   0 : We haven't finished, caller have to go again
 654  *   1 : We have finished, we can go to complete phase
 655  */
 656 int qemu_savevm_state_iterate(QEMUFile *f)
 657 {
 658     SaveStateEntry *se;
 659     int ret = 1;
 660
 661     trace_savevm_state_iterate();
 662     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 663         if (!se->ops || !se->ops->save_live_iterate) {
 664             continue;
 665         }
 666         if (se->ops && se->ops->is_active) {
 667             if (!se->ops->is_active(se->opaque)) {
 668                 continue;
 669             }
 670         }
 671         if (qemu_file_rate_limit(f)) {
 672             return 0;
 673         }
 674         trace_savevm_section_start(se->idstr, se->section_id);
 675         /* Section type */
 676         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 677         qemu_put_be32(f, se->section_id);
 678
 679         ret = se->ops->save_live_iterate(f, se->opaque);
 680         trace_savevm_section_end(se->idstr, se->section_id);
 681
 682         if (ret < 0) {
 683             qemu_file_set_error(f, ret);
 684         }
 685         if (ret <= 0) {
 686             /* Do not proceed to the next vmstate before this one reported
 687                completion of the current stage. This serializes the migration
 688                and reduces the probability that a faster changing state is
 689                synchronized over and over again. */
 690             break;
 691         }
 692     }
 693     return ret;
 694 }
 695
 696 void qemu_savevm_state_complete(QEMUFile *f)
 697 {
 698     SaveStateEntry *se;
 699     int ret;
 700
 701     trace_savevm_state_complete();
 702
 703     cpu_synchronize_all_states();
 704
 705     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 706         if (!se->ops || !se->ops->save_live_complete) {
 707             continue;
 708         }
 709         if (se->ops && se->ops->is_active) {
 710             if (!se->ops->is_active(se->opaque)) {
 711                 continue;
 712             }
 713         }
 714         trace_savevm_section_start(se->idstr, se->section_id);
 715         /* Section type */
 716         qemu_put_byte(f, QEMU_VM_SECTION_END);
 717         qemu_put_be32(f, se->section_id);
 718
 719         ret = se->ops->save_live_complete(f, se->opaque);
 720         trace_savevm_section_end(se->idstr, se->section_id);
 721         if (ret < 0) {
 722             qemu_file_set_error(f, ret);
 723             return;
 724         }
 725     }
 726
 727     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 728         int len;
 729
 730         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
 731             continue;
 732         }
 733         trace_savevm_section_start(se->idstr, se->section_id);
 734         /* Section type */
 735         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 736         qemu_put_be32(f, se->section_id);
 737
 738         /* ID string */
 739         len = strlen(se->idstr);
 740         qemu_put_byte(f, len);
 741         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 742
 743         qemu_put_be32(f, se->instance_id);
 744         qemu_put_be32(f, se->version_id);
 745
 746         vmstate_save(f, se);
 747         trace_savevm_section_end(se->idstr, se->section_id);
 748     }
 749
 750     qemu_put_byte(f, QEMU_VM_EOF);
 751     qemu_fflush(f);
 752 }
 753
 754 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
 755 {
 756     SaveStateEntry *se;
 757     uint64_t ret = 0;
 758
 759     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 760         if (!se->ops || !se->ops->save_live_pending) {
 761             continue;
 762         }
 763         if (se->ops && se->ops->is_active) {
 764             if (!se->ops->is_active(se->opaque)) {
 765                 continue;
 766             }
 767         }
 768         ret += se->ops->save_live_pending(f, se->opaque, max_size);
 769     }
 770     return ret;
 771 }
 772
 773 void qemu_savevm_state_cancel(void)
 774 {
 775     SaveStateEntry *se;
 776
 777     trace_savevm_state_cancel();
 778     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 779         if (se->ops && se->ops->cancel) {
 780             se->ops->cancel(se->opaque);
 781         }
 782     }
 783 }
 784
 785 static int qemu_savevm_state(QEMUFile *f)
 786 {
 787     int ret;
 788     MigrationParams params = {
 789         .blk = 0,
 790         .shared = 0
 791     };
 792
 793     if (qemu_savevm_state_blocked(NULL)) {
 794         return -EINVAL;
 795     }
 796
 797     qemu_mutex_unlock_iothread();
 798     qemu_savevm_state_begin(f, &params);
 799     qemu_mutex_lock_iothread();
 800
 801     while (qemu_file_get_error(f) == 0) {
 802         if (qemu_savevm_state_iterate(f) > 0) {
 803             break;
 804         }
 805     }
 806
 807     ret = qemu_file_get_error(f);
 808     if (ret == 0) {
 809         qemu_savevm_state_complete(f);
 810         ret = qemu_file_get_error(f);
 811     }
 812     if (ret != 0) {
 813         qemu_savevm_state_cancel();
 814     }
 815     return ret;
 816 }
 817
 818 static int qemu_save_device_state(QEMUFile *f)
 819 {
 820     SaveStateEntry *se;
 821
 822     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 823     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 824
 825     cpu_synchronize_all_states();
 826
 827     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 828         int len;
 829
 830         if (se->is_ram) {
 831             continue;
 832         }
 833         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
 834             continue;
 835         }
 836
 837         /* Section type */
 838         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 839         qemu_put_be32(f, se->section_id);
 840
 841         /* ID string */
 842         len = strlen(se->idstr);
 843         qemu_put_byte(f, len);
 844         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 845
 846         qemu_put_be32(f, se->instance_id);
 847         qemu_put_be32(f, se->version_id);
 848
 849         vmstate_save(f, se);
 850     }
 851
 852     qemu_put_byte(f, QEMU_VM_EOF);
 853
 854     return qemu_file_get_error(f);
 855 }
 856
 857 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 858 {
 859     SaveStateEntry *se;
 860
 861     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 862         if (!strcmp(se->idstr, idstr) &&
 863             (instance_id == se->instance_id ||
 864              instance_id == se->alias_id))
 865             return se;
 866         /* Migrating from an older version? */
 867         if (strstr(se->idstr, idstr) && se->compat) {
 868             if (!strcmp(se->compat->idstr, idstr) &&
 869                 (instance_id == se->compat->instance_id ||
 870                  instance_id == se->alias_id))
 871                 return se;
 872         }
 873     }
 874     return NULL;
 875 }
 876
 877 typedef struct LoadStateEntry {
 878     QLIST_ENTRY(LoadStateEntry) entry;
 879     SaveStateEntry *se;
 880     int section_id;
 881     int version_id;
 882 } LoadStateEntry;
 883
 884 int qemu_loadvm_state(QEMUFile *f)
 885 {
 886     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
 887         QLIST_HEAD_INITIALIZER(loadvm_handlers);
 888     LoadStateEntry *le, *new_le;
 889     uint8_t section_type;
 890     unsigned int v;
 891     int ret;
 892
 893     if (qemu_savevm_state_blocked(NULL)) {
 894         return -EINVAL;
 895     }
 896
 897     v = qemu_get_be32(f);
 898     if (v != QEMU_VM_FILE_MAGIC) {
 899         return -EINVAL;
 900     }
 901
 902     v = qemu_get_be32(f);
 903     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
 904         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
 905         return -ENOTSUP;
 906     }
 907     if (v != QEMU_VM_FILE_VERSION) {
 908         return -ENOTSUP;
 909     }
 910
 911     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 912         uint32_t instance_id, version_id, section_id;
 913         SaveStateEntry *se;
 914         char idstr[257];
 915         int len;
 916
 917         switch (section_type) {
 918         case QEMU_VM_SECTION_START:
 919         case QEMU_VM_SECTION_FULL:
 920             /* Read section start */
 921             section_id = qemu_get_be32(f);
 922             len = qemu_get_byte(f);
 923             qemu_get_buffer(f, (uint8_t *)idstr, len);
 924             idstr[len] = 0;
 925             instance_id = qemu_get_be32(f);
 926             version_id = qemu_get_be32(f);
 927
 928             /* Find savevm section */
 929             se = find_se(idstr, instance_id);
 930             if (se == NULL) {
 931                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 932                 ret = -EINVAL;
 933                 goto out;
 934             }
 935
 936             /* Validate version */
 937             if (version_id > se->version_id) {
 938                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 939                         version_id, idstr, se->version_id);
 940                 ret = -EINVAL;
 941                 goto out;
 942             }
 943
 944             /* Add entry */
 945             le = g_malloc0(sizeof(*le));
 946
 947             le->se = se;
 948             le->section_id = section_id;
 949             le->version_id = version_id;
 950             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
 951
 952             ret = vmstate_load(f, le->se, le->version_id);
 953             if (ret < 0) {
 954                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 955                         instance_id, idstr);
 956                 goto out;
 957             }
 958             break;
 959         case QEMU_VM_SECTION_PART:
 960         case QEMU_VM_SECTION_END:
 961             section_id = qemu_get_be32(f);
 962
 963             QLIST_FOREACH(le, &loadvm_handlers, entry) {
 964                 if (le->section_id == section_id) {
 965                     break;
 966                 }
 967             }
 968             if (le == NULL) {
 969                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 970                 ret = -EINVAL;
 971                 goto out;
 972             }
 973
 974             ret = vmstate_load(f, le->se, le->version_id);
 975             if (ret < 0) {
 976                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
 977                         section_id);
 978                 goto out;
 979             }
 980             break;
 981         default:
 982             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 983             ret = -EINVAL;
 984             goto out;
 985         }
 986     }
 987
 988     cpu_synchronize_all_post_init();
 989
 990     ret = 0;
 991
 992 out:
 993     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
 994         QLIST_REMOVE(le, entry);
 995         g_free(le);
 996     }
 997
 998     if (ret == 0) {
 999         ret = qemu_file_get_error(f);
1000     }
1001
1002     return ret;
1003 }
1004
1005 static BlockDriverState *find_vmstate_bs(void)
1006 {
1007     BlockDriverState *bs = NULL;
1008     while ((bs = bdrv_next(bs))) {
1009         if (bdrv_can_snapshot(bs)) {
1010             return bs;
1011         }
1012     }
1013     return NULL;
1014 }
1015
1016 /*
1017  * Deletes snapshots of a given name in all opened images.
1018  */
1019 static int del_existing_snapshots(Monitor *mon, const char *name)
1020 {
1021     BlockDriverState *bs;
1022     QEMUSnapshotInfo sn1, *snapshot = &sn1;
1023     Error *err = NULL;
1024
1025     bs = NULL;
1026     while ((bs = bdrv_next(bs))) {
1027         if (bdrv_can_snapshot(bs) &&
1028             bdrv_snapshot_find(bs, snapshot, name) >= 0) {
1029             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
1030             if (err) {
1031                 monitor_printf(mon,
1032                                "Error while deleting snapshot on device '%s':"
1033                                " %s\n",
1034                                bdrv_get_device_name(bs),
1035                                error_get_pretty(err));
1036                 error_free(err);
1037                 return -1;
1038             }
1039         }
1040     }
1041
1042     return 0;
1043 }
1044
1045 void do_savevm(Monitor *mon, const QDict *qdict)
1046 {
1047     BlockDriverState *bs, *bs1;
1048     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1049     int ret;
1050     QEMUFile *f;
1051     int saved_vm_running;
1052     uint64_t vm_state_size;
1053     qemu_timeval tv;
1054     struct tm tm;
1055     const char *name = qdict_get_try_str(qdict, "name");
1056
1057     /* Verify if there is a device that doesn't support snapshots and is writable */
1058     bs = NULL;
1059     while ((bs = bdrv_next(bs))) {
1060
1061         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
1062             continue;
1063         }
1064
1065         if (!bdrv_can_snapshot(bs)) {
1066             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
1067                                bdrv_get_device_name(bs));
1068             return;
1069         }
1070     }
1071
1072     bs = find_vmstate_bs();
1073     if (!bs) {
1074         monitor_printf(mon, "No block device can accept snapshots\n");
1075         return;
1076     }
1077
1078     saved_vm_running = runstate_is_running();
1079     vm_stop(RUN_STATE_SAVE_VM);
1080
1081     memset(sn, 0, sizeof(*sn));
1082
1083     /* fill auxiliary fields */
1084     qemu_gettimeofday(&tv);
1085     sn->date_sec = tv.tv_sec;
1086     sn->date_nsec = tv.tv_usec * 1000;
1087     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1088
1089     if (name) {
1090         ret = bdrv_snapshot_find(bs, old_sn, name);
1091         if (ret >= 0) {
1092             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1093             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1094         } else {
1095             pstrcpy(sn->name, sizeof(sn->name), name);
1096         }
1097     } else {
1098         /* cast below needed for OpenBSD where tv_sec is still 'long' */
1099         localtime_r((const time_t *)&tv.tv_sec, &tm);
1100         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
1101     }
1102
1103     /* Delete old snapshots of the same name */
1104     if (name && del_existing_snapshots(mon, name) < 0) {
1105         goto the_end;
1106     }
1107
1108     /* save the VM state */
1109     f = qemu_fopen_bdrv(bs, 1);
1110     if (!f) {
1111         monitor_printf(mon, "Could not open VM state file\n");
1112         goto the_end;
1113     }
1114     ret = qemu_savevm_state(f);
1115     vm_state_size = qemu_ftell(f);
1116     qemu_fclose(f);
1117     if (ret < 0) {
1118         monitor_printf(mon, "Error %d while writing VM\n", ret);
1119         goto the_end;
1120     }
1121
1122     /* create the snapshots */
1123
1124     bs1 = NULL;
1125     while ((bs1 = bdrv_next(bs1))) {
1126         if (bdrv_can_snapshot(bs1)) {
1127             /* Write VM state size only to the image that contains the state */
1128             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1129             ret = bdrv_snapshot_create(bs1, sn);
1130             if (ret < 0) {
1131                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1132                                bdrv_get_device_name(bs1));
1133             }
1134         }
1135     }
1136
1137  the_end:
1138     if (saved_vm_running) {
1139         vm_start();
1140     }
1141 }
1142
1143 void qmp_xen_save_devices_state(const char *filename, Error **errp)
1144 {
1145     QEMUFile *f;
1146     int saved_vm_running;
1147     int ret;
1148
1149     saved_vm_running = runstate_is_running();
1150     vm_stop(RUN_STATE_SAVE_VM);
1151
1152     f = qemu_fopen(filename, "wb");
1153     if (!f) {
1154         error_setg_file_open(errp, errno, filename);
1155         goto the_end;
1156     }
1157     ret = qemu_save_device_state(f);
1158     qemu_fclose(f);
1159     if (ret < 0) {
1160         error_set(errp, QERR_IO_ERROR);
1161     }
1162
1163  the_end:
1164     if (saved_vm_running) {
1165         vm_start();
1166     }
1167 }
1168
1169 int load_vmstate(const char *name)
1170 {
1171     BlockDriverState *bs, *bs_vm_state;
1172     QEMUSnapshotInfo sn;
1173     QEMUFile *f;
1174     int ret;
1175
1176     bs_vm_state = find_vmstate_bs();
1177     if (!bs_vm_state) {
1178         error_report("No block device supports snapshots");
1179         return -ENOTSUP;
1180     }
1181
1182     /* Don't even try to load empty VM states */
1183     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
1184     if (ret < 0) {
1185         return ret;
1186     } else if (sn.vm_state_size == 0) {
1187         error_report("This is a disk-only snapshot. Revert to it offline "
1188             "using qemu-img.");
1189         return -EINVAL;
1190     }
1191
1192     /* Verify if there is any device that doesn't support snapshots and is
1193     writable and check if the requested snapshot is available too. */
1194     bs = NULL;
1195     while ((bs = bdrv_next(bs))) {
1196
1197         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
1198             continue;
1199         }
1200
1201         if (!bdrv_can_snapshot(bs)) {
1202             error_report("Device '%s' is writable but does not support snapshots.",
1203                                bdrv_get_device_name(bs));
1204             return -ENOTSUP;
1205         }
1206
1207         ret = bdrv_snapshot_find(bs, &sn, name);
1208         if (ret < 0) {
1209             error_report("Device '%s' does not have the requested snapshot '%s'",
1210                            bdrv_get_device_name(bs), name);
1211             return ret;
1212         }
1213     }
1214
1215     /* Flush all IO requests so they don't interfere with the new state.  */
1216     bdrv_drain_all();
1217
1218     bs = NULL;
1219     while ((bs = bdrv_next(bs))) {
1220         if (bdrv_can_snapshot(bs)) {
1221             ret = bdrv_snapshot_goto(bs, name);
1222             if (ret < 0) {
1223                 error_report("Error %d while activating snapshot '%s' on '%s'",
1224                              ret, name, bdrv_get_device_name(bs));
1225                 return ret;
1226             }
1227         }
1228     }
1229
1230     /* restore the VM state */
1231     f = qemu_fopen_bdrv(bs_vm_state, 0);
1232     if (!f) {
1233         error_report("Could not open VM state file");
1234         return -EINVAL;
1235     }
1236
1237     qemu_system_reset(VMRESET_SILENT);
1238     ret = qemu_loadvm_state(f);
1239
1240     qemu_fclose(f);
1241     if (ret < 0) {
1242         error_report("Error %d while loading VM state", ret);
1243         return ret;
1244     }
1245
1246     return 0;
1247 }
1248
1249 void do_delvm(Monitor *mon, const QDict *qdict)
1250 {
1251     BlockDriverState *bs, *bs1;
1252     Error *err = NULL;
1253     const char *name = qdict_get_str(qdict, "name");
1254
1255     bs = find_vmstate_bs();
1256     if (!bs) {
1257         monitor_printf(mon, "No block device supports snapshots\n");
1258         return;
1259     }
1260
1261     bs1 = NULL;
1262     while ((bs1 = bdrv_next(bs1))) {
1263         if (bdrv_can_snapshot(bs1)) {
1264             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
1265             if (err) {
1266                 monitor_printf(mon,
1267                                "Error while deleting snapshot on device '%s':"
1268                                " %s\n",
1269                                bdrv_get_device_name(bs),
1270                                error_get_pretty(err));
1271                 error_free(err);
1272             }
1273         }
1274     }
1275 }
1276
1277 void do_info_snapshots(Monitor *mon, const QDict *qdict)
1278 {
1279     BlockDriverState *bs, *bs1;
1280     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
1281     int nb_sns, i, ret, available;
1282     int total;
1283     int *available_snapshots;
1284
1285     bs = find_vmstate_bs();
1286     if (!bs) {
1287         monitor_printf(mon, "No available block device supports snapshots\n");
1288         return;
1289     }
1290
1291     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1292     if (nb_sns < 0) {
1293         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1294         return;
1295     }
1296
1297     if (nb_sns == 0) {
1298         monitor_printf(mon, "There is no snapshot available.\n");
1299         return;
1300     }
1301
1302     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
1303     total = 0;
1304     for (i = 0; i < nb_sns; i++) {
1305         sn = &sn_tab[i];
1306         available = 1;
1307         bs1 = NULL;
1308
1309         while ((bs1 = bdrv_next(bs1))) {
1310             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
1311                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
1312                 if (ret < 0) {
1313                     available = 0;
1314                     break;
1315                 }
1316             }
1317         }
1318
1319         if (available) {
1320             available_snapshots[total] = i;
1321             total++;
1322         }
1323     }
1324
1325     if (total > 0) {
1326         bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, NULL);
1327         monitor_printf(mon, "\n");
1328         for (i = 0; i < total; i++) {
1329             sn = &sn_tab[available_snapshots[i]];
1330             bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, sn);
1331             monitor_printf(mon, "\n");
1332         }
1333     } else {
1334         monitor_printf(mon, "There is no suitable snapshot available\n");
1335     }
1336
1337     g_free(sn_tab);
1338     g_free(available_snapshots);
1339
1340 }
1341
1342 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
1343 {
1344     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
1345                        memory_region_name(mr), dev);
1346 }
1347
1348 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
1349 {
1350     qemu_ram_unset_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK);
1351 }
1352
1353 void vmstate_register_ram_global(MemoryRegion *mr)
1354 {
1355     vmstate_register_ram(mr, NULL);
1356 }