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 is_ram;
 236 } SaveStateEntry;
 237
 238
 239 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
 240     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
 241 static int global_section_id;
 242
 243 static void dump_vmstate_vmsd(FILE *out_file,
 244                               const VMStateDescription *vmsd, int indent,
 245                               bool is_subsection);
 246
 247 static void dump_vmstate_vmsf(FILE *out_file, const VMStateField *field,
 248                               int indent)
 249 {
 250     fprintf(out_file, "%*s{\n", indent, "");
 251     indent += 2;
 252     fprintf(out_file, "%*s\"field\": \"%s\",\n", indent, "", field->name);
 253     fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "",
 254             field->version_id);
 255     fprintf(out_file, "%*s\"field_exists\": %s,\n", indent, "",
 256             field->field_exists ? "true" : "false");
 257     fprintf(out_file, "%*s\"size\": %zu", indent, "", field->size);
 258     if (field->vmsd != NULL) {
 259         fprintf(out_file, ",\n");
 260         dump_vmstate_vmsd(out_file, field->vmsd, indent, false);
 261     }
 262     fprintf(out_file, "\n%*s}", indent - 2, "");
 263 }
 264
 265 static void dump_vmstate_vmss(FILE *out_file,
 266                               const VMStateSubsection *subsection,
 267                               int indent)
 268 {
 269     if (subsection->vmsd != NULL) {
 270         dump_vmstate_vmsd(out_file, subsection->vmsd, indent, true);
 271     }
 272 }
 273
 274 static void dump_vmstate_vmsd(FILE *out_file,
 275                               const VMStateDescription *vmsd, int indent,
 276                               bool is_subsection)
 277 {
 278     if (is_subsection) {
 279         fprintf(out_file, "%*s{\n", indent, "");
 280     } else {
 281         fprintf(out_file, "%*s\"%s\": {\n", indent, "", "Description");
 282     }
 283     indent += 2;
 284     fprintf(out_file, "%*s\"name\": \"%s\",\n", indent, "", vmsd->name);
 285     fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "",
 286             vmsd->version_id);
 287     fprintf(out_file, "%*s\"minimum_version_id\": %d", indent, "",
 288             vmsd->minimum_version_id);
 289     if (vmsd->fields != NULL) {
 290         const VMStateField *field = vmsd->fields;
 291         bool first;
 292
 293         fprintf(out_file, ",\n%*s\"Fields\": [\n", indent, "");
 294         first = true;
 295         while (field->name != NULL) {
 296             if (field->flags & VMS_MUST_EXIST) {
 297                 /* Ignore VMSTATE_VALIDATE bits; these don't get migrated */
 298                 field++;
 299                 continue;
 300             }
 301             if (!first) {
 302                 fprintf(out_file, ",\n");
 303             }
 304             dump_vmstate_vmsf(out_file, field, indent + 2);
 305             field++;
 306             first = false;
 307         }
 308         fprintf(out_file, "\n%*s]", indent, "");
 309     }
 310     if (vmsd->subsections != NULL) {
 311         const VMStateSubsection *subsection = vmsd->subsections;
 312         bool first;
 313
 314         fprintf(out_file, ",\n%*s\"Subsections\": [\n", indent, "");
 315         first = true;
 316         while (subsection->vmsd != NULL) {
 317             if (!first) {
 318                 fprintf(out_file, ",\n");
 319             }
 320             dump_vmstate_vmss(out_file, subsection, indent + 2);
 321             subsection++;
 322             first = false;
 323         }
 324         fprintf(out_file, "\n%*s]", indent, "");
 325     }
 326     fprintf(out_file, "\n%*s}", indent - 2, "");
 327 }
 328
 329 static void dump_machine_type(FILE *out_file)
 330 {
 331     MachineClass *mc;
 332
 333     mc = MACHINE_GET_CLASS(current_machine);
 334
 335     fprintf(out_file, "  \"vmschkmachine\": {\n");
 336     fprintf(out_file, "    \"Name\": \"%s\"\n", mc->name);
 337     fprintf(out_file, "  },\n");
 338 }
 339
 340 void dump_vmstate_json_to_file(FILE *out_file)
 341 {
 342     GSList *list, *elt;
 343     bool first;
 344
 345     fprintf(out_file, "{\n");
 346     dump_machine_type(out_file);
 347
 348     first = true;
 349     list = object_class_get_list(TYPE_DEVICE, true);
 350     for (elt = list; elt; elt = elt->next) {
 351         DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, elt->data,
 352                                              TYPE_DEVICE);
 353         const char *name;
 354         int indent = 2;
 355
 356         if (!dc->vmsd) {
 357             continue;
 358         }
 359
 360         if (!first) {
 361             fprintf(out_file, ",\n");
 362         }
 363         name = object_class_get_name(OBJECT_CLASS(dc));
 364         fprintf(out_file, "%*s\"%s\": {\n", indent, "", name);
 365         indent += 2;
 366         fprintf(out_file, "%*s\"Name\": \"%s\",\n", indent, "", name);
 367         fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "",
 368                 dc->vmsd->version_id);
 369         fprintf(out_file, "%*s\"minimum_version_id\": %d,\n", indent, "",
 370                 dc->vmsd->minimum_version_id);
 371
 372         dump_vmstate_vmsd(out_file, dc->vmsd, indent, false);
 373
 374         fprintf(out_file, "\n%*s}", indent - 2, "");
 375         first = false;
 376     }
 377     fprintf(out_file, "\n}\n");
 378     fclose(out_file);
 379 }
 380
 381 static int calculate_new_instance_id(const char *idstr)
 382 {
 383     SaveStateEntry *se;
 384     int instance_id = 0;
 385
 386     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 387         if (strcmp(idstr, se->idstr) == 0
 388             && instance_id <= se->instance_id) {
 389             instance_id = se->instance_id + 1;
 390         }
 391     }
 392     return instance_id;
 393 }
 394
 395 static int calculate_compat_instance_id(const char *idstr)
 396 {
 397     SaveStateEntry *se;
 398     int instance_id = 0;
 399
 400     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 401         if (!se->compat) {
 402             continue;
 403         }
 404
 405         if (strcmp(idstr, se->compat->idstr) == 0
 406             && instance_id <= se->compat->instance_id) {
 407             instance_id = se->compat->instance_id + 1;
 408         }
 409     }
 410     return instance_id;
 411 }
 412
 413 /* TODO: Individual devices generally have very little idea about the rest
 414    of the system, so instance_id should be removed/replaced.
 415    Meanwhile pass -1 as instance_id if you do not already have a clearly
 416    distinguishing id for all instances of your device class. */
 417 int register_savevm_live(DeviceState *dev,
 418                          const char *idstr,
 419                          int instance_id,
 420                          int version_id,
 421                          SaveVMHandlers *ops,
 422                          void *opaque)
 423 {
 424     SaveStateEntry *se;
 425
 426     se = g_malloc0(sizeof(SaveStateEntry));
 427     se->version_id = version_id;
 428     se->section_id = global_section_id++;
 429     se->ops = ops;
 430     se->opaque = opaque;
 431     se->vmsd = NULL;
 432     /* if this is a live_savem then set is_ram */
 433     if (ops->save_live_setup != NULL) {
 434         se->is_ram = 1;
 435     }
 436
 437     if (dev) {
 438         char *id = qdev_get_dev_path(dev);
 439         if (id) {
 440             pstrcpy(se->idstr, sizeof(se->idstr), id);
 441             pstrcat(se->idstr, sizeof(se->idstr), "/");
 442             g_free(id);
 443
 444             se->compat = g_malloc0(sizeof(CompatEntry));
 445             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
 446             se->compat->instance_id = instance_id == -1 ?
 447                          calculate_compat_instance_id(idstr) : instance_id;
 448             instance_id = -1;
 449         }
 450     }
 451     pstrcat(se->idstr, sizeof(se->idstr), idstr);
 452
 453     if (instance_id == -1) {
 454         se->instance_id = calculate_new_instance_id(se->idstr);
 455     } else {
 456         se->instance_id = instance_id;
 457     }
 458     assert(!se->compat || se->instance_id == 0);
 459     /* add at the end of list */
 460     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
 461     return 0;
 462 }
 463
 464 int register_savevm(DeviceState *dev,
 465                     const char *idstr,
 466                     int instance_id,
 467                     int version_id,
 468                     SaveStateHandler *save_state,
 469                     LoadStateHandler *load_state,
 470                     void *opaque)
 471 {
 472     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
 473     ops->save_state = save_state;
 474     ops->load_state = load_state;
 475     return register_savevm_live(dev, idstr, instance_id, version_id,
 476                                 ops, opaque);
 477 }
 478
 479 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
 480 {
 481     SaveStateEntry *se, *new_se;
 482     char id[256] = "";
 483
 484     if (dev) {
 485         char *path = qdev_get_dev_path(dev);
 486         if (path) {
 487             pstrcpy(id, sizeof(id), path);
 488             pstrcat(id, sizeof(id), "/");
 489             g_free(path);
 490         }
 491     }
 492     pstrcat(id, sizeof(id), idstr);
 493
 494     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
 495         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
 496             QTAILQ_REMOVE(&savevm_handlers, se, entry);
 497             if (se->compat) {
 498                 g_free(se->compat);
 499             }
 500             g_free(se->ops);
 501             g_free(se);
 502         }
 503     }
 504 }
 505
 506 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
 507                                    const VMStateDescription *vmsd,
 508                                    void *opaque, int alias_id,
 509                                    int required_for_version)
 510 {
 511     SaveStateEntry *se;
 512
 513     /* If this triggers, alias support can be dropped for the vmsd. */
 514     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
 515
 516     se = g_malloc0(sizeof(SaveStateEntry));
 517     se->version_id = vmsd->version_id;
 518     se->section_id = global_section_id++;
 519     se->opaque = opaque;
 520     se->vmsd = vmsd;
 521     se->alias_id = alias_id;
 522
 523     if (dev) {
 524         char *id = qdev_get_dev_path(dev);
 525         if (id) {
 526             pstrcpy(se->idstr, sizeof(se->idstr), id);
 527             pstrcat(se->idstr, sizeof(se->idstr), "/");
 528             g_free(id);
 529
 530             se->compat = g_malloc0(sizeof(CompatEntry));
 531             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
 532             se->compat->instance_id = instance_id == -1 ?
 533                          calculate_compat_instance_id(vmsd->name) : instance_id;
 534             instance_id = -1;
 535         }
 536     }
 537     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
 538
 539     if (instance_id == -1) {
 540         se->instance_id = calculate_new_instance_id(se->idstr);
 541     } else {
 542         se->instance_id = instance_id;
 543     }
 544     assert(!se->compat || se->instance_id == 0);
 545     /* add at the end of list */
 546     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
 547     return 0;
 548 }
 549
 550 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
 551                         void *opaque)
 552 {
 553     SaveStateEntry *se, *new_se;
 554
 555     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
 556         if (se->vmsd == vmsd && se->opaque == opaque) {
 557             QTAILQ_REMOVE(&savevm_handlers, se, entry);
 558             if (se->compat) {
 559                 g_free(se->compat);
 560             }
 561             g_free(se);
 562         }
 563     }
 564 }
 565
 566 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
 567 {
 568     trace_vmstate_load(se->idstr, se->vmsd ? se->vmsd->name : "(old)");
 569     if (!se->vmsd) {         /* Old style */
 570         return se->ops->load_state(f, se->opaque, version_id);
 571     }
 572     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
 573 }
 574
 575 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
 576 {
 577     trace_vmstate_save(se->idstr, se->vmsd ? se->vmsd->name : "(old)");
 578     if (!se->vmsd) {         /* Old style */
 579         se->ops->save_state(f, se->opaque);
 580         return;
 581     }
 582     vmstate_save_state(f, se->vmsd, se->opaque);
 583 }
 584
 585 bool qemu_savevm_state_blocked(Error **errp)
 586 {
 587     SaveStateEntry *se;
 588
 589     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 590         if (se->vmsd && se->vmsd->unmigratable) {
 591             error_setg(errp, "State blocked by non-migratable device '%s'",
 592                        se->idstr);
 593             return true;
 594         }
 595     }
 596     return false;
 597 }
 598
 599 void qemu_savevm_state_begin(QEMUFile *f,
 600                              const MigrationParams *params)
 601 {
 602     SaveStateEntry *se;
 603     int ret;
 604
 605     trace_savevm_state_begin();
 606     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 607         if (!se->ops || !se->ops->set_params) {
 608             continue;
 609         }
 610         se->ops->set_params(params, se->opaque);
 611     }
 612
 613     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 614     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 615
 616     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 617         int len;
 618
 619         if (!se->ops || !se->ops->save_live_setup) {
 620             continue;
 621         }
 622         if (se->ops && se->ops->is_active) {
 623             if (!se->ops->is_active(se->opaque)) {
 624                 continue;
 625             }
 626         }
 627         /* Section type */
 628         qemu_put_byte(f, QEMU_VM_SECTION_START);
 629         qemu_put_be32(f, se->section_id);
 630
 631         /* ID string */
 632         len = strlen(se->idstr);
 633         qemu_put_byte(f, len);
 634         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 635
 636         qemu_put_be32(f, se->instance_id);
 637         qemu_put_be32(f, se->version_id);
 638
 639         ret = se->ops->save_live_setup(f, se->opaque);
 640         if (ret < 0) {
 641             qemu_file_set_error(f, ret);
 642             break;
 643         }
 644     }
 645 }
 646
 647 /*
 648  * this function has three return values:
 649  *   negative: there was one error, and we have -errno.
 650  *   0 : We haven't finished, caller have to go again
 651  *   1 : We have finished, we can go to complete phase
 652  */
 653 int qemu_savevm_state_iterate(QEMUFile *f)
 654 {
 655     SaveStateEntry *se;
 656     int ret = 1;
 657
 658     trace_savevm_state_iterate();
 659     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 660         if (!se->ops || !se->ops->save_live_iterate) {
 661             continue;
 662         }
 663         if (se->ops && se->ops->is_active) {
 664             if (!se->ops->is_active(se->opaque)) {
 665                 continue;
 666             }
 667         }
 668         if (qemu_file_rate_limit(f)) {
 669             return 0;
 670         }
 671         trace_savevm_section_start(se->idstr, se->section_id);
 672         /* Section type */
 673         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 674         qemu_put_be32(f, se->section_id);
 675
 676         ret = se->ops->save_live_iterate(f, se->opaque);
 677         trace_savevm_section_end(se->idstr, se->section_id);
 678
 679         if (ret < 0) {
 680             qemu_file_set_error(f, ret);
 681         }
 682         if (ret <= 0) {
 683             /* Do not proceed to the next vmstate before this one reported
 684                completion of the current stage. This serializes the migration
 685                and reduces the probability that a faster changing state is
 686                synchronized over and over again. */
 687             break;
 688         }
 689     }
 690     return ret;
 691 }
 692
 693 void qemu_savevm_state_complete(QEMUFile *f)
 694 {
 695     SaveStateEntry *se;
 696     int ret;
 697
 698     trace_savevm_state_complete();
 699
 700     cpu_synchronize_all_states();
 701
 702     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 703         if (!se->ops || !se->ops->save_live_complete) {
 704             continue;
 705         }
 706         if (se->ops && se->ops->is_active) {
 707             if (!se->ops->is_active(se->opaque)) {
 708                 continue;
 709             }
 710         }
 711         trace_savevm_section_start(se->idstr, se->section_id);
 712         /* Section type */
 713         qemu_put_byte(f, QEMU_VM_SECTION_END);
 714         qemu_put_be32(f, se->section_id);
 715
 716         ret = se->ops->save_live_complete(f, se->opaque);
 717         trace_savevm_section_end(se->idstr, se->section_id);
 718         if (ret < 0) {
 719             qemu_file_set_error(f, ret);
 720             return;
 721         }
 722     }
 723
 724     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 725         int len;
 726
 727         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
 728             continue;
 729         }
 730         trace_savevm_section_start(se->idstr, se->section_id);
 731         /* Section type */
 732         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 733         qemu_put_be32(f, se->section_id);
 734
 735         /* ID string */
 736         len = strlen(se->idstr);
 737         qemu_put_byte(f, len);
 738         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 739
 740         qemu_put_be32(f, se->instance_id);
 741         qemu_put_be32(f, se->version_id);
 742
 743         vmstate_save(f, se);
 744         trace_savevm_section_end(se->idstr, se->section_id);
 745     }
 746
 747     qemu_put_byte(f, QEMU_VM_EOF);
 748     qemu_fflush(f);
 749 }
 750
 751 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
 752 {
 753     SaveStateEntry *se;
 754     uint64_t ret = 0;
 755
 756     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 757         if (!se->ops || !se->ops->save_live_pending) {
 758             continue;
 759         }
 760         if (se->ops && se->ops->is_active) {
 761             if (!se->ops->is_active(se->opaque)) {
 762                 continue;
 763             }
 764         }
 765         ret += se->ops->save_live_pending(f, se->opaque, max_size);
 766     }
 767     return ret;
 768 }
 769
 770 void qemu_savevm_state_cancel(void)
 771 {
 772     SaveStateEntry *se;
 773
 774     trace_savevm_state_cancel();
 775     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 776         if (se->ops && se->ops->cancel) {
 777             se->ops->cancel(se->opaque);
 778         }
 779     }
 780 }
 781
 782 static int qemu_savevm_state(QEMUFile *f)
 783 {
 784     int ret;
 785     MigrationParams params = {
 786         .blk = 0,
 787         .shared = 0
 788     };
 789
 790     if (qemu_savevm_state_blocked(NULL)) {
 791         return -EINVAL;
 792     }
 793
 794     qemu_mutex_unlock_iothread();
 795     qemu_savevm_state_begin(f, &params);
 796     qemu_mutex_lock_iothread();
 797
 798     while (qemu_file_get_error(f) == 0) {
 799         if (qemu_savevm_state_iterate(f) > 0) {
 800             break;
 801         }
 802     }
 803
 804     ret = qemu_file_get_error(f);
 805     if (ret == 0) {
 806         qemu_savevm_state_complete(f);
 807         ret = qemu_file_get_error(f);
 808     }
 809     if (ret != 0) {
 810         qemu_savevm_state_cancel();
 811     }
 812     return ret;
 813 }
 814
 815 static int qemu_save_device_state(QEMUFile *f)
 816 {
 817     SaveStateEntry *se;
 818
 819     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 820     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 821
 822     cpu_synchronize_all_states();
 823
 824     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 825         int len;
 826
 827         if (se->is_ram) {
 828             continue;
 829         }
 830         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
 831             continue;
 832         }
 833
 834         /* Section type */
 835         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 836         qemu_put_be32(f, se->section_id);
 837
 838         /* ID string */
 839         len = strlen(se->idstr);
 840         qemu_put_byte(f, len);
 841         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 842
 843         qemu_put_be32(f, se->instance_id);
 844         qemu_put_be32(f, se->version_id);
 845
 846         vmstate_save(f, se);
 847     }
 848
 849     qemu_put_byte(f, QEMU_VM_EOF);
 850
 851     return qemu_file_get_error(f);
 852 }
 853
 854 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 855 {
 856     SaveStateEntry *se;
 857
 858     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
 859         if (!strcmp(se->idstr, idstr) &&
 860             (instance_id == se->instance_id ||
 861              instance_id == se->alias_id))
 862             return se;
 863         /* Migrating from an older version? */
 864         if (strstr(se->idstr, idstr) && se->compat) {
 865             if (!strcmp(se->compat->idstr, idstr) &&
 866                 (instance_id == se->compat->instance_id ||
 867                  instance_id == se->alias_id))
 868                 return se;
 869         }
 870     }
 871     return NULL;
 872 }
 873
 874 typedef struct LoadStateEntry {
 875     QLIST_ENTRY(LoadStateEntry) entry;
 876     SaveStateEntry *se;
 877     int section_id;
 878     int version_id;
 879 } LoadStateEntry;
 880
 881 int qemu_loadvm_state(QEMUFile *f)
 882 {
 883     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
 884         QLIST_HEAD_INITIALIZER(loadvm_handlers);
 885     LoadStateEntry *le, *new_le;
 886     uint8_t section_type;
 887     unsigned int v;
 888     int ret;
 889
 890     if (qemu_savevm_state_blocked(NULL)) {
 891         return -EINVAL;
 892     }
 893
 894     v = qemu_get_be32(f);
 895     if (v != QEMU_VM_FILE_MAGIC) {
 896         return -EINVAL;
 897     }
 898
 899     v = qemu_get_be32(f);
 900     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
 901         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
 902         return -ENOTSUP;
 903     }
 904     if (v != QEMU_VM_FILE_VERSION) {
 905         return -ENOTSUP;
 906     }
 907
 908     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 909         uint32_t instance_id, version_id, section_id;
 910         SaveStateEntry *se;
 911         char idstr[257];
 912         int len;
 913
 914         switch (section_type) {
 915         case QEMU_VM_SECTION_START:
 916         case QEMU_VM_SECTION_FULL:
 917             /* Read section start */
 918             section_id = qemu_get_be32(f);
 919             len = qemu_get_byte(f);
 920             qemu_get_buffer(f, (uint8_t *)idstr, len);
 921             idstr[len] = 0;
 922             instance_id = qemu_get_be32(f);
 923             version_id = qemu_get_be32(f);
 924
 925             /* Find savevm section */
 926             se = find_se(idstr, instance_id);
 927             if (se == NULL) {
 928                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 929                 ret = -EINVAL;
 930                 goto out;
 931             }
 932
 933             /* Validate version */
 934             if (version_id > se->version_id) {
 935                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 936                         version_id, idstr, se->version_id);
 937                 ret = -EINVAL;
 938                 goto out;
 939             }
 940
 941             /* Add entry */
 942             le = g_malloc0(sizeof(*le));
 943
 944             le->se = se;
 945             le->section_id = section_id;
 946             le->version_id = version_id;
 947             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
 948
 949             ret = vmstate_load(f, le->se, le->version_id);
 950             if (ret < 0) {
 951                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 952                         instance_id, idstr);
 953                 goto out;
 954             }
 955             break;
 956         case QEMU_VM_SECTION_PART:
 957         case QEMU_VM_SECTION_END:
 958             section_id = qemu_get_be32(f);
 959
 960             QLIST_FOREACH(le, &loadvm_handlers, entry) {
 961                 if (le->section_id == section_id) {
 962                     break;
 963                 }
 964             }
 965             if (le == NULL) {
 966                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 967                 ret = -EINVAL;
 968                 goto out;
 969             }
 970
 971             ret = vmstate_load(f, le->se, le->version_id);
 972             if (ret < 0) {
 973                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
 974                         section_id);
 975                 goto out;
 976             }
 977             break;
 978         default:
 979             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 980             ret = -EINVAL;
 981             goto out;
 982         }
 983     }
 984
 985     cpu_synchronize_all_post_init();
 986
 987     ret = 0;
 988
 989 out:
 990     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
 991         QLIST_REMOVE(le, entry);
 992         g_free(le);
 993     }
 994
 995     if (ret == 0) {
 996         ret = qemu_file_get_error(f);
 997     }
 998
 999     return ret;
1000 }
1001
1002 static BlockDriverState *find_vmstate_bs(void)
1003 {
1004     BlockDriverState *bs = NULL;
1005     while ((bs = bdrv_next(bs))) {
1006         if (bdrv_can_snapshot(bs)) {
1007             return bs;
1008         }
1009     }
1010     return NULL;
1011 }
1012
1013 /*
1014  * Deletes snapshots of a given name in all opened images.
1015  */
1016 static int del_existing_snapshots(Monitor *mon, const char *name)
1017 {
1018     BlockDriverState *bs;
1019     QEMUSnapshotInfo sn1, *snapshot = &sn1;
1020     Error *err = NULL;
1021
1022     bs = NULL;
1023     while ((bs = bdrv_next(bs))) {
1024         if (bdrv_can_snapshot(bs) &&
1025             bdrv_snapshot_find(bs, snapshot, name) >= 0) {
1026             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
1027             if (err) {
1028                 monitor_printf(mon,
1029                                "Error while deleting snapshot on device '%s':"
1030                                " %s\n",
1031                                bdrv_get_device_name(bs),
1032                                error_get_pretty(err));
1033                 error_free(err);
1034                 return -1;
1035             }
1036         }
1037     }
1038
1039     return 0;
1040 }
1041
1042 void do_savevm(Monitor *mon, const QDict *qdict)
1043 {
1044     BlockDriverState *bs, *bs1;
1045     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1046     int ret;
1047     QEMUFile *f;
1048     int saved_vm_running;
1049     uint64_t vm_state_size;
1050     qemu_timeval tv;
1051     struct tm tm;
1052     const char *name = qdict_get_try_str(qdict, "name");
1053
1054     /* Verify if there is a device that doesn't support snapshots and is writable */
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             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
1064                                bdrv_get_device_name(bs));
1065             return;
1066         }
1067     }
1068
1069     bs = find_vmstate_bs();
1070     if (!bs) {
1071         monitor_printf(mon, "No block device can accept snapshots\n");
1072         return;
1073     }
1074
1075     saved_vm_running = runstate_is_running();
1076     vm_stop(RUN_STATE_SAVE_VM);
1077
1078     memset(sn, 0, sizeof(*sn));
1079
1080     /* fill auxiliary fields */
1081     qemu_gettimeofday(&tv);
1082     sn->date_sec = tv.tv_sec;
1083     sn->date_nsec = tv.tv_usec * 1000;
1084     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1085
1086     if (name) {
1087         ret = bdrv_snapshot_find(bs, old_sn, name);
1088         if (ret >= 0) {
1089             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1090             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1091         } else {
1092             pstrcpy(sn->name, sizeof(sn->name), name);
1093         }
1094     } else {
1095         /* cast below needed for OpenBSD where tv_sec is still 'long' */
1096         localtime_r((const time_t *)&tv.tv_sec, &tm);
1097         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
1098     }
1099
1100     /* Delete old snapshots of the same name */
1101     if (name && del_existing_snapshots(mon, name) < 0) {
1102         goto the_end;
1103     }
1104
1105     /* save the VM state */
1106     f = qemu_fopen_bdrv(bs, 1);
1107     if (!f) {
1108         monitor_printf(mon, "Could not open VM state file\n");
1109         goto the_end;
1110     }
1111     ret = qemu_savevm_state(f);
1112     vm_state_size = qemu_ftell(f);
1113     qemu_fclose(f);
1114     if (ret < 0) {
1115         monitor_printf(mon, "Error %d while writing VM\n", ret);
1116         goto the_end;
1117     }
1118
1119     /* create the snapshots */
1120
1121     bs1 = NULL;
1122     while ((bs1 = bdrv_next(bs1))) {
1123         if (bdrv_can_snapshot(bs1)) {
1124             /* Write VM state size only to the image that contains the state */
1125             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1126             ret = bdrv_snapshot_create(bs1, sn);
1127             if (ret < 0) {
1128                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1129                                bdrv_get_device_name(bs1));
1130             }
1131         }
1132     }
1133
1134  the_end:
1135     if (saved_vm_running) {
1136         vm_start();
1137     }
1138 }
1139
1140 void qmp_xen_save_devices_state(const char *filename, Error **errp)
1141 {
1142     QEMUFile *f;
1143     int saved_vm_running;
1144     int ret;
1145
1146     saved_vm_running = runstate_is_running();
1147     vm_stop(RUN_STATE_SAVE_VM);
1148
1149     f = qemu_fopen(filename, "wb");
1150     if (!f) {
1151         error_setg_file_open(errp, errno, filename);
1152         goto the_end;
1153     }
1154     ret = qemu_save_device_state(f);
1155     qemu_fclose(f);
1156     if (ret < 0) {
1157         error_set(errp, QERR_IO_ERROR);
1158     }
1159
1160  the_end:
1161     if (saved_vm_running) {
1162         vm_start();
1163     }
1164 }
1165
1166 int load_vmstate(const char *name)
1167 {
1168     BlockDriverState *bs, *bs_vm_state;
1169     QEMUSnapshotInfo sn;
1170     QEMUFile *f;
1171     int ret;
1172
1173     bs_vm_state = find_vmstate_bs();
1174     if (!bs_vm_state) {
1175         error_report("No block device supports snapshots");
1176         return -ENOTSUP;
1177     }
1178
1179     /* Don't even try to load empty VM states */
1180     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
1181     if (ret < 0) {
1182         return ret;
1183     } else if (sn.vm_state_size == 0) {
1184         error_report("This is a disk-only snapshot. Revert to it offline "
1185             "using qemu-img.");
1186         return -EINVAL;
1187     }
1188
1189     /* Verify if there is any device that doesn't support snapshots and is
1190     writable and check if the requested snapshot is available too. */
1191     bs = NULL;
1192     while ((bs = bdrv_next(bs))) {
1193
1194         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
1195             continue;
1196         }
1197
1198         if (!bdrv_can_snapshot(bs)) {
1199             error_report("Device '%s' is writable but does not support snapshots.",
1200                                bdrv_get_device_name(bs));
1201             return -ENOTSUP;
1202         }
1203
1204         ret = bdrv_snapshot_find(bs, &sn, name);
1205         if (ret < 0) {
1206             error_report("Device '%s' does not have the requested snapshot '%s'",
1207                            bdrv_get_device_name(bs), name);
1208             return ret;
1209         }
1210     }
1211
1212     /* Flush all IO requests so they don't interfere with the new state.  */
1213     bdrv_drain_all();
1214
1215     bs = NULL;
1216     while ((bs = bdrv_next(bs))) {
1217         if (bdrv_can_snapshot(bs)) {
1218             ret = bdrv_snapshot_goto(bs, name);
1219             if (ret < 0) {
1220                 error_report("Error %d while activating snapshot '%s' on '%s'",
1221                              ret, name, bdrv_get_device_name(bs));
1222                 return ret;
1223             }
1224         }
1225     }
1226
1227     /* restore the VM state */
1228     f = qemu_fopen_bdrv(bs_vm_state, 0);
1229     if (!f) {
1230         error_report("Could not open VM state file");
1231         return -EINVAL;
1232     }
1233
1234     qemu_system_reset(VMRESET_SILENT);
1235     ret = qemu_loadvm_state(f);
1236
1237     qemu_fclose(f);
1238     if (ret < 0) {
1239         error_report("Error %d while loading VM state", ret);
1240         return ret;
1241     }
1242
1243     return 0;
1244 }
1245
1246 void do_delvm(Monitor *mon, const QDict *qdict)
1247 {
1248     BlockDriverState *bs;
1249     Error *err;
1250     const char *name = qdict_get_str(qdict, "name");
1251
1252     if (!find_vmstate_bs()) {
1253         monitor_printf(mon, "No block device supports snapshots\n");
1254         return;
1255     }
1256
1257     bs = NULL;
1258     while ((bs = bdrv_next(bs))) {
1259         if (bdrv_can_snapshot(bs)) {
1260             err = NULL;
1261             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
1262             if (err) {
1263                 monitor_printf(mon,
1264                                "Error while deleting snapshot on device '%s':"
1265                                " %s\n",
1266                                bdrv_get_device_name(bs),
1267                                error_get_pretty(err));
1268                 error_free(err);
1269             }
1270         }
1271     }
1272 }
1273
1274 void do_info_snapshots(Monitor *mon, const QDict *qdict)
1275 {
1276     BlockDriverState *bs, *bs1;
1277     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
1278     int nb_sns, i, ret, available;
1279     int total;
1280     int *available_snapshots;
1281
1282     bs = find_vmstate_bs();
1283     if (!bs) {
1284         monitor_printf(mon, "No available block device supports snapshots\n");
1285         return;
1286     }
1287
1288     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1289     if (nb_sns < 0) {
1290         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1291         return;
1292     }
1293
1294     if (nb_sns == 0) {
1295         monitor_printf(mon, "There is no snapshot available.\n");
1296         return;
1297     }
1298
1299     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
1300     total = 0;
1301     for (i = 0; i < nb_sns; i++) {
1302         sn = &sn_tab[i];
1303         available = 1;
1304         bs1 = NULL;
1305
1306         while ((bs1 = bdrv_next(bs1))) {
1307             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
1308                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
1309                 if (ret < 0) {
1310                     available = 0;
1311                     break;
1312                 }
1313             }
1314         }
1315
1316         if (available) {
1317             available_snapshots[total] = i;
1318             total++;
1319         }
1320     }
1321
1322     if (total > 0) {
1323         bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, NULL);
1324         monitor_printf(mon, "\n");
1325         for (i = 0; i < total; i++) {
1326             sn = &sn_tab[available_snapshots[i]];
1327             bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, sn);
1328             monitor_printf(mon, "\n");
1329         }
1330     } else {
1331         monitor_printf(mon, "There is no suitable snapshot available\n");
1332     }
1333
1334     g_free(sn_tab);
1335     g_free(available_snapshots);
1336
1337 }
1338
1339 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
1340 {
1341     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
1342                        memory_region_name(mr), dev);
1343 }
1344
1345 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
1346 {
1347     qemu_ram_unset_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK);
1348 }
1349
1350 void vmstate_register_ram_global(MemoryRegion *mr)
1351 {
1352     vmstate_register_ram(mr, NULL);
1353 }