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/bitops.h"
  42
  43 #define SELF_ANNOUNCE_ROUNDS 5
  44
  45 #ifndef ETH_P_RARP
  46 #define ETH_P_RARP 0x8035
  47 #endif
  48 #define ARP_HTYPE_ETH 0x0001
  49 #define ARP_PTYPE_IP 0x0800
  50 #define ARP_OP_REQUEST_REV 0x3
  51
  52 static int announce_self_create(uint8_t *buf,
  53                                 uint8_t *mac_addr)
  54 {
  55     /* Ethernet header. */
  56     memset(buf, 0xff, 6);         /* destination MAC addr */
  57     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
  58     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
  59
  60     /* RARP header. */
  61     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
  62     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
  63     *(buf + 18) = 6; /* hardware addr length (ethernet) */
  64     *(buf + 19) = 4; /* protocol addr length (IPv4) */
  65     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
  66     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
  67     memset(buf + 28, 0x00, 4);     /* source protocol addr */
  68     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
  69     memset(buf + 38, 0x00, 4);     /* target protocol addr */
  70
  71     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
  72     memset(buf + 42, 0x00, 18);
  73
  74     return 60; /* len (FCS will be added by hardware) */
  75 }
  76
  77 static void qemu_announce_self_iter(NICState *nic, void *opaque)
  78 {
  79     uint8_t buf[60];
  80     int len;
  81
  82     len = announce_self_create(buf, nic->conf->macaddr.a);
  83
  84     qemu_send_packet_raw(qemu_get_queue(nic), buf, len);
  85 }
  86
  87
  88 static void qemu_announce_self_once(void *opaque)
  89 {
  90     static int count = SELF_ANNOUNCE_ROUNDS;
  91     QEMUTimer *timer = *(QEMUTimer **)opaque;
  92
  93     qemu_foreach_nic(qemu_announce_self_iter, NULL);
  94
  95     if (--count) {
  96         /* delay 50ms, 150ms, 250ms, ... */
  97         qemu_mod_timer(timer, qemu_get_clock_ms(rt_clock) +
  98                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
  99     } else {
 100             qemu_del_timer(timer);
 101             qemu_free_timer(timer);
 102     }
 103 }
 104
 105 void qemu_announce_self(void)
 106 {
 107         static QEMUTimer *timer;
 108         timer = qemu_new_timer_ms(rt_clock, qemu_announce_self_once, &timer);
 109         qemu_announce_self_once(&timer);
 110 }
 111
 112 /***********************************************************/
 113 /* savevm/loadvm support */
 114
 115 #define IO_BUF_SIZE 32768
 116
 117 struct QEMUFile {
 118     const QEMUFileOps *ops;
 119     void *opaque;
 120     int is_write;
 121
 122     int64_t buf_offset; /* start of buffer when writing, end of buffer
 123                            when reading */
 124     int buf_index;
 125     int buf_size; /* 0 when writing */
 126     uint8_t buf[IO_BUF_SIZE];
 127
 128     int last_error;
 129 };
 130
 131 typedef struct QEMUFileStdio
 132 {
 133     FILE *stdio_file;
 134     QEMUFile *file;
 135 } QEMUFileStdio;
 136
 137 typedef struct QEMUFileSocket
 138 {
 139     int fd;
 140     QEMUFile *file;
 141 } QEMUFileSocket;
 142
 143 typedef struct {
 144     Coroutine *co;
 145     int fd;
 146 } FDYieldUntilData;
 147
 148 static void fd_coroutine_enter(void *opaque)
 149 {
 150     FDYieldUntilData *data = opaque;
 151     qemu_set_fd_handler(data->fd, NULL, NULL, NULL);
 152     qemu_coroutine_enter(data->co, NULL);
 153 }
 154
 155 /**
 156  * Yield until a file descriptor becomes readable
 157  *
 158  * Note that this function clobbers the handlers for the file descriptor.
 159  */
 160 static void coroutine_fn yield_until_fd_readable(int fd)
 161 {
 162     FDYieldUntilData data;
 163
 164     assert(qemu_in_coroutine());
 165     data.co = qemu_coroutine_self();
 166     data.fd = fd;
 167     qemu_set_fd_handler(fd, fd_coroutine_enter, NULL, &data);
 168     qemu_coroutine_yield();
 169 }
 170
 171 static int socket_get_fd(void *opaque)
 172 {
 173     QEMUFileSocket *s = opaque;
 174
 175     return s->fd;
 176 }
 177
 178 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 179 {
 180     QEMUFileSocket *s = opaque;
 181     ssize_t len;
 182
 183     for (;;) {
 184         len = qemu_recv(s->fd, buf, size, 0);
 185         if (len != -1) {
 186             break;
 187         }
 188         if (socket_error() == EAGAIN) {
 189             yield_until_fd_readable(s->fd);
 190         } else if (socket_error() != EINTR) {
 191             break;
 192         }
 193     }
 194
 195     if (len == -1) {
 196         len = -socket_error();
 197     }
 198     return len;
 199 }
 200
 201 static int socket_close(void *opaque)
 202 {
 203     QEMUFileSocket *s = opaque;
 204     closesocket(s->fd);
 205     g_free(s);
 206     return 0;
 207 }
 208
 209 static int stdio_get_fd(void *opaque)
 210 {
 211     QEMUFileStdio *s = opaque;
 212
 213     return fileno(s->stdio_file);
 214 }
 215
 216 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 217 {
 218     QEMUFileStdio *s = opaque;
 219     return fwrite(buf, 1, size, s->stdio_file);
 220 }
 221
 222 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 223 {
 224     QEMUFileStdio *s = opaque;
 225     FILE *fp = s->stdio_file;
 226     int bytes;
 227
 228     for (;;) {
 229         clearerr(fp);
 230         bytes = fread(buf, 1, size, fp);
 231         if (bytes != 0 || !ferror(fp)) {
 232             break;
 233         }
 234         if (errno == EAGAIN) {
 235             yield_until_fd_readable(fileno(fp));
 236         } else if (errno != EINTR) {
 237             break;
 238         }
 239     }
 240     return bytes;
 241 }
 242
 243 static int stdio_pclose(void *opaque)
 244 {
 245     QEMUFileStdio *s = opaque;
 246     int ret;
 247     ret = pclose(s->stdio_file);
 248     if (ret == -1) {
 249         ret = -errno;
 250     } else if (!WIFEXITED(ret) || WEXITSTATUS(ret) != 0) {
 251         /* close succeeded, but non-zero exit code: */
 252         ret = -EIO; /* fake errno value */
 253     }
 254     g_free(s);
 255     return ret;
 256 }
 257
 258 static int stdio_fclose(void *opaque)
 259 {
 260     QEMUFileStdio *s = opaque;
 261     int ret = 0;
 262
 263     if (s->file->ops->put_buffer) {
 264         int fd = fileno(s->stdio_file);
 265         struct stat st;
 266
 267         ret = fstat(fd, &st);
 268         if (ret == 0 && S_ISREG(st.st_mode)) {
 269             /*
 270              * If the file handle is a regular file make sure the
 271              * data is flushed to disk before signaling success.
 272              */
 273             ret = fsync(fd);
 274             if (ret != 0) {
 275                 ret = -errno;
 276                 return ret;
 277             }
 278         }
 279     }
 280     if (fclose(s->stdio_file) == EOF) {
 281         ret = -errno;
 282     }
 283     g_free(s);
 284     return ret;
 285 }
 286
 287 static const QEMUFileOps stdio_pipe_read_ops = {
 288     .get_fd =     stdio_get_fd,
 289     .get_buffer = stdio_get_buffer,
 290     .close =      stdio_pclose
 291 };
 292
 293 static const QEMUFileOps stdio_pipe_write_ops = {
 294     .get_fd =     stdio_get_fd,
 295     .put_buffer = stdio_put_buffer,
 296     .close =      stdio_pclose
 297 };
 298
 299 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 300 {
 301     FILE *stdio_file;
 302     QEMUFileStdio *s;
 303
 304     stdio_file = popen(command, mode);
 305     if (stdio_file == NULL) {
 306         return NULL;
 307     }
 308
 309     if (mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 310         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 311         return NULL;
 312     }
 313
 314     s = g_malloc0(sizeof(QEMUFileStdio));
 315
 316     s->stdio_file = stdio_file;
 317
 318     if(mode[0] == 'r') {
 319         s->file = qemu_fopen_ops(s, &stdio_pipe_read_ops);
 320     } else {
 321         s->file = qemu_fopen_ops(s, &stdio_pipe_write_ops);
 322     }
 323     return s->file;
 324 }
 325
 326 static const QEMUFileOps stdio_file_read_ops = {
 327     .get_fd =     stdio_get_fd,
 328     .get_buffer = stdio_get_buffer,
 329     .close =      stdio_fclose
 330 };
 331
 332 static const QEMUFileOps stdio_file_write_ops = {
 333     .get_fd =     stdio_get_fd,
 334     .put_buffer = stdio_put_buffer,
 335     .close =      stdio_fclose
 336 };
 337
 338 QEMUFile *qemu_fdopen(int fd, const char *mode)
 339 {
 340     QEMUFileStdio *s;
 341
 342     if (mode == NULL ||
 343         (mode[0] != 'r' && mode[0] != 'w') ||
 344         mode[1] != 'b' || mode[2] != 0) {
 345         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 346         return NULL;
 347     }
 348
 349     s = g_malloc0(sizeof(QEMUFileStdio));
 350     s->stdio_file = fdopen(fd, mode);
 351     if (!s->stdio_file)
 352         goto fail;
 353
 354     if(mode[0] == 'r') {
 355         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 356     } else {
 357         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 358     }
 359     return s->file;
 360
 361 fail:
 362     g_free(s);
 363     return NULL;
 364 }
 365
 366 static const QEMUFileOps socket_read_ops = {
 367     .get_fd =     socket_get_fd,
 368     .get_buffer = socket_get_buffer,
 369     .close =      socket_close
 370 };
 371
 372 QEMUFile *qemu_fopen_socket(int fd)
 373 {
 374     QEMUFileSocket *s = g_malloc0(sizeof(QEMUFileSocket));
 375
 376     s->fd = fd;
 377     s->file = qemu_fopen_ops(s, &socket_read_ops);
 378     return s->file;
 379 }
 380
 381 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 382 {
 383     QEMUFileStdio *s;
 384
 385     if (mode == NULL ||
 386         (mode[0] != 'r' && mode[0] != 'w') ||
 387         mode[1] != 'b' || mode[2] != 0) {
 388         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 389         return NULL;
 390     }
 391
 392     s = g_malloc0(sizeof(QEMUFileStdio));
 393
 394     s->stdio_file = fopen(filename, mode);
 395     if (!s->stdio_file)
 396         goto fail;
 397
 398     if(mode[0] == 'w') {
 399         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 400     } else {
 401         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 402     }
 403     return s->file;
 404 fail:
 405     g_free(s);
 406     return NULL;
 407 }
 408
 409 static int block_put_buffer(void *opaque, const uint8_t *buf,
 410                            int64_t pos, int size)
 411 {
 412     bdrv_save_vmstate(opaque, buf, pos, size);
 413     return size;
 414 }
 415
 416 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 417 {
 418     return bdrv_load_vmstate(opaque, buf, pos, size);
 419 }
 420
 421 static int bdrv_fclose(void *opaque)
 422 {
 423     return bdrv_flush(opaque);
 424 }
 425
 426 static const QEMUFileOps bdrv_read_ops = {
 427     .get_buffer = block_get_buffer,
 428     .close =      bdrv_fclose
 429 };
 430
 431 static const QEMUFileOps bdrv_write_ops = {
 432     .put_buffer = block_put_buffer,
 433     .close =      bdrv_fclose
 434 };
 435
 436 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 437 {
 438     if (is_writable)
 439         return qemu_fopen_ops(bs, &bdrv_write_ops);
 440     return qemu_fopen_ops(bs, &bdrv_read_ops);
 441 }
 442
 443 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
 444 {
 445     QEMUFile *f;
 446
 447     f = g_malloc0(sizeof(QEMUFile));
 448
 449     f->opaque = opaque;
 450     f->ops = ops;
 451     f->is_write = 0;
 452
 453     return f;
 454 }
 455
 456 int qemu_file_get_error(QEMUFile *f)
 457 {
 458     return f->last_error;
 459 }
 460
 461 static void qemu_file_set_error(QEMUFile *f, int ret)
 462 {
 463     if (f->last_error == 0) {
 464         f->last_error = ret;
 465     }
 466 }
 467
 468 /** Flushes QEMUFile buffer
 469  *
 470  */
 471 static void qemu_fflush(QEMUFile *f)
 472 {
 473     int ret = 0;
 474
 475     if (!f->ops->put_buffer) {
 476         return;
 477     }
 478     if (f->is_write && f->buf_index > 0) {
 479         ret = f->ops->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 480         if (ret >= 0) {
 481             f->buf_offset += f->buf_index;
 482         }
 483         f->buf_index = 0;
 484     }
 485     if (ret < 0) {
 486         qemu_file_set_error(f, ret);
 487     }
 488 }
 489
 490 static void qemu_fill_buffer(QEMUFile *f)
 491 {
 492     int len;
 493     int pending;
 494
 495     if (!f->ops->get_buffer)
 496         return;
 497
 498     if (f->is_write)
 499         abort();
 500
 501     pending = f->buf_size - f->buf_index;
 502     if (pending > 0) {
 503         memmove(f->buf, f->buf + f->buf_index, pending);
 504     }
 505     f->buf_index = 0;
 506     f->buf_size = pending;
 507
 508     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->buf_offset,
 509                         IO_BUF_SIZE - pending);
 510     if (len > 0) {
 511         f->buf_size += len;
 512         f->buf_offset += len;
 513     } else if (len == 0) {
 514         qemu_file_set_error(f, -EIO);
 515     } else if (len != -EAGAIN)
 516         qemu_file_set_error(f, len);
 517 }
 518
 519 int qemu_get_fd(QEMUFile *f)
 520 {
 521     if (f->ops->get_fd) {
 522         return f->ops->get_fd(f->opaque);
 523     }
 524     return -1;
 525 }
 526
 527 /** Closes the file
 528  *
 529  * Returns negative error value if any error happened on previous operations or
 530  * while closing the file. Returns 0 or positive number on success.
 531  *
 532  * The meaning of return value on success depends on the specific backend
 533  * being used.
 534  */
 535 int qemu_fclose(QEMUFile *f)
 536 {
 537     int ret;
 538     qemu_fflush(f);
 539     ret = qemu_file_get_error(f);
 540
 541     if (f->ops->close) {
 542         int ret2 = f->ops->close(f->opaque);
 543         if (ret >= 0) {
 544             ret = ret2;
 545         }
 546     }
 547     /* If any error was spotted before closing, we should report it
 548      * instead of the close() return value.
 549      */
 550     if (f->last_error) {
 551         ret = f->last_error;
 552     }
 553     g_free(f);
 554     return ret;
 555 }
 556
 557 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 558 {
 559     int l;
 560
 561     if (f->last_error) {
 562         return;
 563     }
 564
 565     if (f->is_write == 0 && f->buf_index > 0) {
 566         fprintf(stderr,
 567                 "Attempted to write to buffer while read buffer is not empty\n");
 568         abort();
 569     }
 570
 571     while (size > 0) {
 572         l = IO_BUF_SIZE - f->buf_index;
 573         if (l > size)
 574             l = size;
 575         memcpy(f->buf + f->buf_index, buf, l);
 576         f->is_write = 1;
 577         f->buf_index += l;
 578         buf += l;
 579         size -= l;
 580         if (f->buf_index >= IO_BUF_SIZE) {
 581             qemu_fflush(f);
 582             if (qemu_file_get_error(f)) {
 583                 break;
 584             }
 585         }
 586     }
 587 }
 588
 589 void qemu_put_byte(QEMUFile *f, int v)
 590 {
 591     if (f->last_error) {
 592         return;
 593     }
 594
 595     if (f->is_write == 0 && f->buf_index > 0) {
 596         fprintf(stderr,
 597                 "Attempted to write to buffer while read buffer is not empty\n");
 598         abort();
 599     }
 600
 601     f->buf[f->buf_index++] = v;
 602     f->is_write = 1;
 603     if (f->buf_index >= IO_BUF_SIZE) {
 604         qemu_fflush(f);
 605     }
 606 }
 607
 608 static void qemu_file_skip(QEMUFile *f, int size)
 609 {
 610     if (f->buf_index + size <= f->buf_size) {
 611         f->buf_index += size;
 612     }
 613 }
 614
 615 static int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset)
 616 {
 617     int pending;
 618     int index;
 619
 620     if (f->is_write) {
 621         abort();
 622     }
 623
 624     index = f->buf_index + offset;
 625     pending = f->buf_size - index;
 626     if (pending < size) {
 627         qemu_fill_buffer(f);
 628         index = f->buf_index + offset;
 629         pending = f->buf_size - index;
 630     }
 631
 632     if (pending <= 0) {
 633         return 0;
 634     }
 635     if (size > pending) {
 636         size = pending;
 637     }
 638
 639     memcpy(buf, f->buf + index, size);
 640     return size;
 641 }
 642
 643 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size)
 644 {
 645     int pending = size;
 646     int done = 0;
 647
 648     while (pending > 0) {
 649         int res;
 650
 651         res = qemu_peek_buffer(f, buf, pending, 0);
 652         if (res == 0) {
 653             return done;
 654         }
 655         qemu_file_skip(f, res);
 656         buf += res;
 657         pending -= res;
 658         done += res;
 659     }
 660     return done;
 661 }
 662
 663 static int qemu_peek_byte(QEMUFile *f, int offset)
 664 {
 665     int index = f->buf_index + offset;
 666
 667     if (f->is_write) {
 668         abort();
 669     }
 670
 671     if (index >= f->buf_size) {
 672         qemu_fill_buffer(f);
 673         index = f->buf_index + offset;
 674         if (index >= f->buf_size) {
 675             return 0;
 676         }
 677     }
 678     return f->buf[index];
 679 }
 680
 681 int qemu_get_byte(QEMUFile *f)
 682 {
 683     int result;
 684
 685     result = qemu_peek_byte(f, 0);
 686     qemu_file_skip(f, 1);
 687     return result;
 688 }
 689
 690 int64_t qemu_ftell(QEMUFile *f)
 691 {
 692     /* buf_offset excludes buffer for writing but includes it for reading */
 693     if (f->is_write) {
 694         return f->buf_offset + f->buf_index;
 695     } else {
 696         return f->buf_offset - f->buf_size + f->buf_index;
 697     }
 698 }
 699
 700 int qemu_file_rate_limit(QEMUFile *f)
 701 {
 702     if (f->ops->rate_limit)
 703         return f->ops->rate_limit(f->opaque);
 704
 705     return 0;
 706 }
 707
 708 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 709 {
 710     if (f->ops->get_rate_limit)
 711         return f->ops->get_rate_limit(f->opaque);
 712
 713     return 0;
 714 }
 715
 716 int64_t qemu_file_set_rate_limit(QEMUFile *f, int64_t new_rate)
 717 {
 718     /* any failed or completed migration keeps its state to allow probing of
 719      * migration data, but has no associated file anymore */
 720     if (f && f->ops->set_rate_limit)
 721         return f->ops->set_rate_limit(f->opaque, new_rate);
 722
 723     return 0;
 724 }
 725
 726 void qemu_put_be16(QEMUFile *f, unsigned int v)
 727 {
 728     qemu_put_byte(f, v >> 8);
 729     qemu_put_byte(f, v);
 730 }
 731
 732 void qemu_put_be32(QEMUFile *f, unsigned int v)
 733 {
 734     qemu_put_byte(f, v >> 24);
 735     qemu_put_byte(f, v >> 16);
 736     qemu_put_byte(f, v >> 8);
 737     qemu_put_byte(f, v);
 738 }
 739
 740 void qemu_put_be64(QEMUFile *f, uint64_t v)
 741 {
 742     qemu_put_be32(f, v >> 32);
 743     qemu_put_be32(f, v);
 744 }
 745
 746 unsigned int qemu_get_be16(QEMUFile *f)
 747 {
 748     unsigned int v;
 749     v = qemu_get_byte(f) << 8;
 750     v |= qemu_get_byte(f);
 751     return v;
 752 }
 753
 754 unsigned int qemu_get_be32(QEMUFile *f)
 755 {
 756     unsigned int v;
 757     v = qemu_get_byte(f) << 24;
 758     v |= qemu_get_byte(f) << 16;
 759     v |= qemu_get_byte(f) << 8;
 760     v |= qemu_get_byte(f);
 761     return v;
 762 }
 763
 764 uint64_t qemu_get_be64(QEMUFile *f)
 765 {
 766     uint64_t v;
 767     v = (uint64_t)qemu_get_be32(f) << 32;
 768     v |= qemu_get_be32(f);
 769     return v;
 770 }
 771
 772
 773 /* timer */
 774
 775 void qemu_put_timer(QEMUFile *f, QEMUTimer *ts)
 776 {
 777     uint64_t expire_time;
 778
 779     expire_time = qemu_timer_expire_time_ns(ts);
 780     qemu_put_be64(f, expire_time);
 781 }
 782
 783 void qemu_get_timer(QEMUFile *f, QEMUTimer *ts)
 784 {
 785     uint64_t expire_time;
 786
 787     expire_time = qemu_get_be64(f);
 788     if (expire_time != -1) {
 789         qemu_mod_timer_ns(ts, expire_time);
 790     } else {
 791         qemu_del_timer(ts);
 792     }
 793 }
 794
 795
 796 /* bool */
 797
 798 static int get_bool(QEMUFile *f, void *pv, size_t size)
 799 {
 800     bool *v = pv;
 801     *v = qemu_get_byte(f);
 802     return 0;
 803 }
 804
 805 static void put_bool(QEMUFile *f, void *pv, size_t size)
 806 {
 807     bool *v = pv;
 808     qemu_put_byte(f, *v);
 809 }
 810
 811 const VMStateInfo vmstate_info_bool = {
 812     .name = "bool",
 813     .get  = get_bool,
 814     .put  = put_bool,
 815 };
 816
 817 /* 8 bit int */
 818
 819 static int get_int8(QEMUFile *f, void *pv, size_t size)
 820 {
 821     int8_t *v = pv;
 822     qemu_get_s8s(f, v);
 823     return 0;
 824 }
 825
 826 static void put_int8(QEMUFile *f, void *pv, size_t size)
 827 {
 828     int8_t *v = pv;
 829     qemu_put_s8s(f, v);
 830 }
 831
 832 const VMStateInfo vmstate_info_int8 = {
 833     .name = "int8",
 834     .get  = get_int8,
 835     .put  = put_int8,
 836 };
 837
 838 /* 16 bit int */
 839
 840 static int get_int16(QEMUFile *f, void *pv, size_t size)
 841 {
 842     int16_t *v = pv;
 843     qemu_get_sbe16s(f, v);
 844     return 0;
 845 }
 846
 847 static void put_int16(QEMUFile *f, void *pv, size_t size)
 848 {
 849     int16_t *v = pv;
 850     qemu_put_sbe16s(f, v);
 851 }
 852
 853 const VMStateInfo vmstate_info_int16 = {
 854     .name = "int16",
 855     .get  = get_int16,
 856     .put  = put_int16,
 857 };
 858
 859 /* 32 bit int */
 860
 861 static int get_int32(QEMUFile *f, void *pv, size_t size)
 862 {
 863     int32_t *v = pv;
 864     qemu_get_sbe32s(f, v);
 865     return 0;
 866 }
 867
 868 static void put_int32(QEMUFile *f, void *pv, size_t size)
 869 {
 870     int32_t *v = pv;
 871     qemu_put_sbe32s(f, v);
 872 }
 873
 874 const VMStateInfo vmstate_info_int32 = {
 875     .name = "int32",
 876     .get  = get_int32,
 877     .put  = put_int32,
 878 };
 879
 880 /* 32 bit int. See that the received value is the same than the one
 881    in the field */
 882
 883 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 884 {
 885     int32_t *v = pv;
 886     int32_t v2;
 887     qemu_get_sbe32s(f, &v2);
 888
 889     if (*v == v2)
 890         return 0;
 891     return -EINVAL;
 892 }
 893
 894 const VMStateInfo vmstate_info_int32_equal = {
 895     .name = "int32 equal",
 896     .get  = get_int32_equal,
 897     .put  = put_int32,
 898 };
 899
 900 /* 32 bit int. See that the received value is the less or the same
 901    than the one in the field */
 902
 903 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 904 {
 905     int32_t *old = pv;
 906     int32_t new;
 907     qemu_get_sbe32s(f, &new);
 908
 909     if (*old <= new)
 910         return 0;
 911     return -EINVAL;
 912 }
 913
 914 const VMStateInfo vmstate_info_int32_le = {
 915     .name = "int32 equal",
 916     .get  = get_int32_le,
 917     .put  = put_int32,
 918 };
 919
 920 /* 64 bit int */
 921
 922 static int get_int64(QEMUFile *f, void *pv, size_t size)
 923 {
 924     int64_t *v = pv;
 925     qemu_get_sbe64s(f, v);
 926     return 0;
 927 }
 928
 929 static void put_int64(QEMUFile *f, void *pv, size_t size)
 930 {
 931     int64_t *v = pv;
 932     qemu_put_sbe64s(f, v);
 933 }
 934
 935 const VMStateInfo vmstate_info_int64 = {
 936     .name = "int64",
 937     .get  = get_int64,
 938     .put  = put_int64,
 939 };
 940
 941 /* 8 bit unsigned int */
 942
 943 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 944 {
 945     uint8_t *v = pv;
 946     qemu_get_8s(f, v);
 947     return 0;
 948 }
 949
 950 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 951 {
 952     uint8_t *v = pv;
 953     qemu_put_8s(f, v);
 954 }
 955
 956 const VMStateInfo vmstate_info_uint8 = {
 957     .name = "uint8",
 958     .get  = get_uint8,
 959     .put  = put_uint8,
 960 };
 961
 962 /* 16 bit unsigned int */
 963
 964 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 965 {
 966     uint16_t *v = pv;
 967     qemu_get_be16s(f, v);
 968     return 0;
 969 }
 970
 971 static void put_uint16(QEMUFile *f, void *pv, size_t size)
 972 {
 973     uint16_t *v = pv;
 974     qemu_put_be16s(f, v);
 975 }
 976
 977 const VMStateInfo vmstate_info_uint16 = {
 978     .name = "uint16",
 979     .get  = get_uint16,
 980     .put  = put_uint16,
 981 };
 982
 983 /* 32 bit unsigned int */
 984
 985 static int get_uint32(QEMUFile *f, void *pv, size_t size)
 986 {
 987     uint32_t *v = pv;
 988     qemu_get_be32s(f, v);
 989     return 0;
 990 }
 991
 992 static void put_uint32(QEMUFile *f, void *pv, size_t size)
 993 {
 994     uint32_t *v = pv;
 995     qemu_put_be32s(f, v);
 996 }
 997
 998 const VMStateInfo vmstate_info_uint32 = {
 999     .name = "uint32",
1000     .get  = get_uint32,
1001     .put  = put_uint32,
1002 };
1003
1004 /* 32 bit uint. See that the received value is the same than the one
1005    in the field */
1006
1007 static int get_uint32_equal(QEMUFile *f, void *pv, size_t size)
1008 {
1009     uint32_t *v = pv;
1010     uint32_t v2;
1011     qemu_get_be32s(f, &v2);
1012
1013     if (*v == v2) {
1014         return 0;
1015     }
1016     return -EINVAL;
1017 }
1018
1019 const VMStateInfo vmstate_info_uint32_equal = {
1020     .name = "uint32 equal",
1021     .get  = get_uint32_equal,
1022     .put  = put_uint32,
1023 };
1024
1025 /* 64 bit unsigned int */
1026
1027 static int get_uint64(QEMUFile *f, void *pv, size_t size)
1028 {
1029     uint64_t *v = pv;
1030     qemu_get_be64s(f, v);
1031     return 0;
1032 }
1033
1034 static void put_uint64(QEMUFile *f, void *pv, size_t size)
1035 {
1036     uint64_t *v = pv;
1037     qemu_put_be64s(f, v);
1038 }
1039
1040 const VMStateInfo vmstate_info_uint64 = {
1041     .name = "uint64",
1042     .get  = get_uint64,
1043     .put  = put_uint64,
1044 };
1045
1046 /* 8 bit int. See that the received value is the same than the one
1047    in the field */
1048
1049 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
1050 {
1051     uint8_t *v = pv;
1052     uint8_t v2;
1053     qemu_get_8s(f, &v2);
1054
1055     if (*v == v2)
1056         return 0;
1057     return -EINVAL;
1058 }
1059
1060 const VMStateInfo vmstate_info_uint8_equal = {
1061     .name = "uint8 equal",
1062     .get  = get_uint8_equal,
1063     .put  = put_uint8,
1064 };
1065
1066 /* 16 bit unsigned int int. See that the received value is the same than the one
1067    in the field */
1068
1069 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
1070 {
1071     uint16_t *v = pv;
1072     uint16_t v2;
1073     qemu_get_be16s(f, &v2);
1074
1075     if (*v == v2)
1076         return 0;
1077     return -EINVAL;
1078 }
1079
1080 const VMStateInfo vmstate_info_uint16_equal = {
1081     .name = "uint16 equal",
1082     .get  = get_uint16_equal,
1083     .put  = put_uint16,
1084 };
1085
1086 /* timers  */
1087
1088 static int get_timer(QEMUFile *f, void *pv, size_t size)
1089 {
1090     QEMUTimer *v = pv;
1091     qemu_get_timer(f, v);
1092     return 0;
1093 }
1094
1095 static void put_timer(QEMUFile *f, void *pv, size_t size)
1096 {
1097     QEMUTimer *v = pv;
1098     qemu_put_timer(f, v);
1099 }
1100
1101 const VMStateInfo vmstate_info_timer = {
1102     .name = "timer",
1103     .get  = get_timer,
1104     .put  = put_timer,
1105 };
1106
1107 /* uint8_t buffers */
1108
1109 static int get_buffer(QEMUFile *f, void *pv, size_t size)
1110 {
1111     uint8_t *v = pv;
1112     qemu_get_buffer(f, v, size);
1113     return 0;
1114 }
1115
1116 static void put_buffer(QEMUFile *f, void *pv, size_t size)
1117 {
1118     uint8_t *v = pv;
1119     qemu_put_buffer(f, v, size);
1120 }
1121
1122 const VMStateInfo vmstate_info_buffer = {
1123     .name = "buffer",
1124     .get  = get_buffer,
1125     .put  = put_buffer,
1126 };
1127
1128 /* unused buffers: space that was used for some fields that are
1129    not useful anymore */
1130
1131 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
1132 {
1133     uint8_t buf[1024];
1134     int block_len;
1135
1136     while (size > 0) {
1137         block_len = MIN(sizeof(buf), size);
1138         size -= block_len;
1139         qemu_get_buffer(f, buf, block_len);
1140     }
1141    return 0;
1142 }
1143
1144 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1145 {
1146     static const uint8_t buf[1024];
1147     int block_len;
1148
1149     while (size > 0) {
1150         block_len = MIN(sizeof(buf), size);
1151         size -= block_len;
1152         qemu_put_buffer(f, buf, block_len);
1153     }
1154 }
1155
1156 const VMStateInfo vmstate_info_unused_buffer = {
1157     .name = "unused_buffer",
1158     .get  = get_unused_buffer,
1159     .put  = put_unused_buffer,
1160 };
1161
1162 /* bitmaps (as defined by bitmap.h). Note that size here is the size
1163  * of the bitmap in bits. The on-the-wire format of a bitmap is 64
1164  * bit words with the bits in big endian order. The in-memory format
1165  * is an array of 'unsigned long', which may be either 32 or 64 bits.
1166  */
1167 /* This is the number of 64 bit words sent over the wire */
1168 #define BITS_TO_U64S(nr) DIV_ROUND_UP(nr, 64)
1169 static int get_bitmap(QEMUFile *f, void *pv, size_t size)
1170 {
1171     unsigned long *bmp = pv;
1172     int i, idx = 0;
1173     for (i = 0; i < BITS_TO_U64S(size); i++) {
1174         uint64_t w = qemu_get_be64(f);
1175         bmp[idx++] = w;
1176         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1177             bmp[idx++] = w >> 32;
1178         }
1179     }
1180     return 0;
1181 }
1182
1183 static void put_bitmap(QEMUFile *f, void *pv, size_t size)
1184 {
1185     unsigned long *bmp = pv;
1186     int i, idx = 0;
1187     for (i = 0; i < BITS_TO_U64S(size); i++) {
1188         uint64_t w = bmp[idx++];
1189         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1190             w |= ((uint64_t)bmp[idx++]) << 32;
1191         }
1192         qemu_put_be64(f, w);
1193     }
1194 }
1195
1196 const VMStateInfo vmstate_info_bitmap = {
1197     .name = "bitmap",
1198     .get = get_bitmap,
1199     .put = put_bitmap,
1200 };
1201
1202 typedef struct CompatEntry {
1203     char idstr[256];
1204     int instance_id;
1205 } CompatEntry;
1206
1207 typedef struct SaveStateEntry {
1208     QTAILQ_ENTRY(SaveStateEntry) entry;
1209     char idstr[256];
1210     int instance_id;
1211     int alias_id;
1212     int version_id;
1213     int section_id;
1214     SaveVMHandlers *ops;
1215     const VMStateDescription *vmsd;
1216     void *opaque;
1217     CompatEntry *compat;
1218     int no_migrate;
1219     int is_ram;
1220 } SaveStateEntry;
1221
1222
1223 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1224     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1225 static int global_section_id;
1226
1227 static int calculate_new_instance_id(const char *idstr)
1228 {
1229     SaveStateEntry *se;
1230     int instance_id = 0;
1231
1232     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1233         if (strcmp(idstr, se->idstr) == 0
1234             && instance_id <= se->instance_id) {
1235             instance_id = se->instance_id + 1;
1236         }
1237     }
1238     return instance_id;
1239 }
1240
1241 static int calculate_compat_instance_id(const char *idstr)
1242 {
1243     SaveStateEntry *se;
1244     int instance_id = 0;
1245
1246     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1247         if (!se->compat)
1248             continue;
1249
1250         if (strcmp(idstr, se->compat->idstr) == 0
1251             && instance_id <= se->compat->instance_id) {
1252             instance_id = se->compat->instance_id + 1;
1253         }
1254     }
1255     return instance_id;
1256 }
1257
1258 /* TODO: Individual devices generally have very little idea about the rest
1259    of the system, so instance_id should be removed/replaced.
1260    Meanwhile pass -1 as instance_id if you do not already have a clearly
1261    distinguishing id for all instances of your device class. */
1262 int register_savevm_live(DeviceState *dev,
1263                          const char *idstr,
1264                          int instance_id,
1265                          int version_id,
1266                          SaveVMHandlers *ops,
1267                          void *opaque)
1268 {
1269     SaveStateEntry *se;
1270
1271     se = g_malloc0(sizeof(SaveStateEntry));
1272     se->version_id = version_id;
1273     se->section_id = global_section_id++;
1274     se->ops = ops;
1275     se->opaque = opaque;
1276     se->vmsd = NULL;
1277     se->no_migrate = 0;
1278     /* if this is a live_savem then set is_ram */
1279     if (ops->save_live_setup != NULL) {
1280         se->is_ram = 1;
1281     }
1282
1283     if (dev) {
1284         char *id = qdev_get_dev_path(dev);
1285         if (id) {
1286             pstrcpy(se->idstr, sizeof(se->idstr), id);
1287             pstrcat(se->idstr, sizeof(se->idstr), "/");
1288             g_free(id);
1289
1290             se->compat = g_malloc0(sizeof(CompatEntry));
1291             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1292             se->compat->instance_id = instance_id == -1 ?
1293                          calculate_compat_instance_id(idstr) : instance_id;
1294             instance_id = -1;
1295         }
1296     }
1297     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1298
1299     if (instance_id == -1) {
1300         se->instance_id = calculate_new_instance_id(se->idstr);
1301     } else {
1302         se->instance_id = instance_id;
1303     }
1304     assert(!se->compat || se->instance_id == 0);
1305     /* add at the end of list */
1306     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1307     return 0;
1308 }
1309
1310 int register_savevm(DeviceState *dev,
1311                     const char *idstr,
1312                     int instance_id,
1313                     int version_id,
1314                     SaveStateHandler *save_state,
1315                     LoadStateHandler *load_state,
1316                     void *opaque)
1317 {
1318     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
1319     ops->save_state = save_state;
1320     ops->load_state = load_state;
1321     return register_savevm_live(dev, idstr, instance_id, version_id,
1322                                 ops, opaque);
1323 }
1324
1325 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1326 {
1327     SaveStateEntry *se, *new_se;
1328     char id[256] = "";
1329
1330     if (dev) {
1331         char *path = qdev_get_dev_path(dev);
1332         if (path) {
1333             pstrcpy(id, sizeof(id), path);
1334             pstrcat(id, sizeof(id), "/");
1335             g_free(path);
1336         }
1337     }
1338     pstrcat(id, sizeof(id), idstr);
1339
1340     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1341         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1342             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1343             if (se->compat) {
1344                 g_free(se->compat);
1345             }
1346             g_free(se->ops);
1347             g_free(se);
1348         }
1349     }
1350 }
1351
1352 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1353                                    const VMStateDescription *vmsd,
1354                                    void *opaque, int alias_id,
1355                                    int required_for_version)
1356 {
1357     SaveStateEntry *se;
1358
1359     /* If this triggers, alias support can be dropped for the vmsd. */
1360     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1361
1362     se = g_malloc0(sizeof(SaveStateEntry));
1363     se->version_id = vmsd->version_id;
1364     se->section_id = global_section_id++;
1365     se->opaque = opaque;
1366     se->vmsd = vmsd;
1367     se->alias_id = alias_id;
1368     se->no_migrate = vmsd->unmigratable;
1369
1370     if (dev) {
1371         char *id = qdev_get_dev_path(dev);
1372         if (id) {
1373             pstrcpy(se->idstr, sizeof(se->idstr), id);
1374             pstrcat(se->idstr, sizeof(se->idstr), "/");
1375             g_free(id);
1376
1377             se->compat = g_malloc0(sizeof(CompatEntry));
1378             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1379             se->compat->instance_id = instance_id == -1 ?
1380                          calculate_compat_instance_id(vmsd->name) : instance_id;
1381             instance_id = -1;
1382         }
1383     }
1384     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1385
1386     if (instance_id == -1) {
1387         se->instance_id = calculate_new_instance_id(se->idstr);
1388     } else {
1389         se->instance_id = instance_id;
1390     }
1391     assert(!se->compat || se->instance_id == 0);
1392     /* add at the end of list */
1393     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1394     return 0;
1395 }
1396
1397 int vmstate_register(DeviceState *dev, int instance_id,
1398                      const VMStateDescription *vmsd, void *opaque)
1399 {
1400     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1401                                           opaque, -1, 0);
1402 }
1403
1404 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1405                         void *opaque)
1406 {
1407     SaveStateEntry *se, *new_se;
1408
1409     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1410         if (se->vmsd == vmsd && se->opaque == opaque) {
1411             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1412             if (se->compat) {
1413                 g_free(se->compat);
1414             }
1415             g_free(se);
1416         }
1417     }
1418 }
1419
1420 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1421                                     void *opaque);
1422 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1423                                    void *opaque);
1424
1425 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1426                        void *opaque, int version_id)
1427 {
1428     VMStateField *field = vmsd->fields;
1429     int ret;
1430
1431     if (version_id > vmsd->version_id) {
1432         return -EINVAL;
1433     }
1434     if (version_id < vmsd->minimum_version_id_old) {
1435         return -EINVAL;
1436     }
1437     if  (version_id < vmsd->minimum_version_id) {
1438         return vmsd->load_state_old(f, opaque, version_id);
1439     }
1440     if (vmsd->pre_load) {
1441         int ret = vmsd->pre_load(opaque);
1442         if (ret)
1443             return ret;
1444     }
1445     while(field->name) {
1446         if ((field->field_exists &&
1447              field->field_exists(opaque, version_id)) ||
1448             (!field->field_exists &&
1449              field->version_id <= version_id)) {
1450             void *base_addr = opaque + field->offset;
1451             int i, n_elems = 1;
1452             int size = field->size;
1453
1454             if (field->flags & VMS_VBUFFER) {
1455                 size = *(int32_t *)(opaque+field->size_offset);
1456                 if (field->flags & VMS_MULTIPLY) {
1457                     size *= field->size;
1458                 }
1459             }
1460             if (field->flags & VMS_ARRAY) {
1461                 n_elems = field->num;
1462             } else if (field->flags & VMS_VARRAY_INT32) {
1463                 n_elems = *(int32_t *)(opaque+field->num_offset);
1464             } else if (field->flags & VMS_VARRAY_UINT32) {
1465                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1466             } else if (field->flags & VMS_VARRAY_UINT16) {
1467                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1468             } else if (field->flags & VMS_VARRAY_UINT8) {
1469                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1470             }
1471             if (field->flags & VMS_POINTER) {
1472                 base_addr = *(void **)base_addr + field->start;
1473             }
1474             for (i = 0; i < n_elems; i++) {
1475                 void *addr = base_addr + size * i;
1476
1477                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1478                     addr = *(void **)addr;
1479                 }
1480                 if (field->flags & VMS_STRUCT) {
1481                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1482                 } else {
1483                     ret = field->info->get(f, addr, size);
1484
1485                 }
1486                 if (ret < 0) {
1487                     return ret;
1488                 }
1489             }
1490         }
1491         field++;
1492     }
1493     ret = vmstate_subsection_load(f, vmsd, opaque);
1494     if (ret != 0) {
1495         return ret;
1496     }
1497     if (vmsd->post_load) {
1498         return vmsd->post_load(opaque, version_id);
1499     }
1500     return 0;
1501 }
1502
1503 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1504                         void *opaque)
1505 {
1506     VMStateField *field = vmsd->fields;
1507
1508     if (vmsd->pre_save) {
1509         vmsd->pre_save(opaque);
1510     }
1511     while(field->name) {
1512         if (!field->field_exists ||
1513             field->field_exists(opaque, vmsd->version_id)) {
1514             void *base_addr = opaque + field->offset;
1515             int i, n_elems = 1;
1516             int size = field->size;
1517
1518             if (field->flags & VMS_VBUFFER) {
1519                 size = *(int32_t *)(opaque+field->size_offset);
1520                 if (field->flags & VMS_MULTIPLY) {
1521                     size *= field->size;
1522                 }
1523             }
1524             if (field->flags & VMS_ARRAY) {
1525                 n_elems = field->num;
1526             } else if (field->flags & VMS_VARRAY_INT32) {
1527                 n_elems = *(int32_t *)(opaque+field->num_offset);
1528             } else if (field->flags & VMS_VARRAY_UINT32) {
1529                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1530             } else if (field->flags & VMS_VARRAY_UINT16) {
1531                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1532             } else if (field->flags & VMS_VARRAY_UINT8) {
1533                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1534             }
1535             if (field->flags & VMS_POINTER) {
1536                 base_addr = *(void **)base_addr + field->start;
1537             }
1538             for (i = 0; i < n_elems; i++) {
1539                 void *addr = base_addr + size * i;
1540
1541                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1542                     addr = *(void **)addr;
1543                 }
1544                 if (field->flags & VMS_STRUCT) {
1545                     vmstate_save_state(f, field->vmsd, addr);
1546                 } else {
1547                     field->info->put(f, addr, size);
1548                 }
1549             }
1550         }
1551         field++;
1552     }
1553     vmstate_subsection_save(f, vmsd, opaque);
1554 }
1555
1556 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1557 {
1558     if (!se->vmsd) {         /* Old style */
1559         return se->ops->load_state(f, se->opaque, version_id);
1560     }
1561     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1562 }
1563
1564 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1565 {
1566     if (!se->vmsd) {         /* Old style */
1567         se->ops->save_state(f, se->opaque);
1568         return;
1569     }
1570     vmstate_save_state(f,se->vmsd, se->opaque);
1571 }
1572
1573 #define QEMU_VM_FILE_MAGIC           0x5145564d
1574 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1575 #define QEMU_VM_FILE_VERSION         0x00000003
1576
1577 #define QEMU_VM_EOF                  0x00
1578 #define QEMU_VM_SECTION_START        0x01
1579 #define QEMU_VM_SECTION_PART         0x02
1580 #define QEMU_VM_SECTION_END          0x03
1581 #define QEMU_VM_SECTION_FULL         0x04
1582 #define QEMU_VM_SUBSECTION           0x05
1583
1584 bool qemu_savevm_state_blocked(Error **errp)
1585 {
1586     SaveStateEntry *se;
1587
1588     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1589         if (se->no_migrate) {
1590             error_set(errp, QERR_MIGRATION_NOT_SUPPORTED, se->idstr);
1591             return true;
1592         }
1593     }
1594     return false;
1595 }
1596
1597 void qemu_savevm_state_begin(QEMUFile *f,
1598                              const MigrationParams *params)
1599 {
1600     SaveStateEntry *se;
1601     int ret;
1602
1603     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1604         if (!se->ops || !se->ops->set_params) {
1605             continue;
1606         }
1607         se->ops->set_params(params, se->opaque);
1608     }
1609
1610     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1611     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1612
1613     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1614         int len;
1615
1616         if (!se->ops || !se->ops->save_live_setup) {
1617             continue;
1618         }
1619         if (se->ops && se->ops->is_active) {
1620             if (!se->ops->is_active(se->opaque)) {
1621                 continue;
1622             }
1623         }
1624         /* Section type */
1625         qemu_put_byte(f, QEMU_VM_SECTION_START);
1626         qemu_put_be32(f, se->section_id);
1627
1628         /* ID string */
1629         len = strlen(se->idstr);
1630         qemu_put_byte(f, len);
1631         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1632
1633         qemu_put_be32(f, se->instance_id);
1634         qemu_put_be32(f, se->version_id);
1635
1636         ret = se->ops->save_live_setup(f, se->opaque);
1637         if (ret < 0) {
1638             qemu_file_set_error(f, ret);
1639             break;
1640         }
1641     }
1642 }
1643
1644 /*
1645  * this function has three return values:
1646  *   negative: there was one error, and we have -errno.
1647  *   0 : We haven't finished, caller have to go again
1648  *   1 : We have finished, we can go to complete phase
1649  */
1650 int qemu_savevm_state_iterate(QEMUFile *f)
1651 {
1652     SaveStateEntry *se;
1653     int ret = 1;
1654
1655     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1656         if (!se->ops || !se->ops->save_live_iterate) {
1657             continue;
1658         }
1659         if (se->ops && se->ops->is_active) {
1660             if (!se->ops->is_active(se->opaque)) {
1661                 continue;
1662             }
1663         }
1664         if (qemu_file_rate_limit(f)) {
1665             return 0;
1666         }
1667         trace_savevm_section_start();
1668         /* Section type */
1669         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1670         qemu_put_be32(f, se->section_id);
1671
1672         ret = se->ops->save_live_iterate(f, se->opaque);
1673         trace_savevm_section_end(se->section_id);
1674
1675         if (ret < 0) {
1676             qemu_file_set_error(f, ret);
1677         }
1678         if (ret <= 0) {
1679             /* Do not proceed to the next vmstate before this one reported
1680                completion of the current stage. This serializes the migration
1681                and reduces the probability that a faster changing state is
1682                synchronized over and over again. */
1683             break;
1684         }
1685     }
1686     return ret;
1687 }
1688
1689 void qemu_savevm_state_complete(QEMUFile *f)
1690 {
1691     SaveStateEntry *se;
1692     int ret;
1693
1694     cpu_synchronize_all_states();
1695
1696     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1697         if (!se->ops || !se->ops->save_live_complete) {
1698             continue;
1699         }
1700         if (se->ops && se->ops->is_active) {
1701             if (!se->ops->is_active(se->opaque)) {
1702                 continue;
1703             }
1704         }
1705         trace_savevm_section_start();
1706         /* Section type */
1707         qemu_put_byte(f, QEMU_VM_SECTION_END);
1708         qemu_put_be32(f, se->section_id);
1709
1710         ret = se->ops->save_live_complete(f, se->opaque);
1711         trace_savevm_section_end(se->section_id);
1712         if (ret < 0) {
1713             qemu_file_set_error(f, ret);
1714             return;
1715         }
1716     }
1717
1718     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1719         int len;
1720
1721         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1722             continue;
1723         }
1724         trace_savevm_section_start();
1725         /* Section type */
1726         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1727         qemu_put_be32(f, se->section_id);
1728
1729         /* ID string */
1730         len = strlen(se->idstr);
1731         qemu_put_byte(f, len);
1732         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1733
1734         qemu_put_be32(f, se->instance_id);
1735         qemu_put_be32(f, se->version_id);
1736
1737         vmstate_save(f, se);
1738         trace_savevm_section_end(se->section_id);
1739     }
1740
1741     qemu_put_byte(f, QEMU_VM_EOF);
1742     qemu_fflush(f);
1743 }
1744
1745 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
1746 {
1747     SaveStateEntry *se;
1748     uint64_t ret = 0;
1749
1750     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1751         if (!se->ops || !se->ops->save_live_pending) {
1752             continue;
1753         }
1754         if (se->ops && se->ops->is_active) {
1755             if (!se->ops->is_active(se->opaque)) {
1756                 continue;
1757             }
1758         }
1759         ret += se->ops->save_live_pending(f, se->opaque, max_size);
1760     }
1761     return ret;
1762 }
1763
1764 void qemu_savevm_state_cancel(void)
1765 {
1766     SaveStateEntry *se;
1767
1768     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1769         if (se->ops && se->ops->cancel) {
1770             se->ops->cancel(se->opaque);
1771         }
1772     }
1773 }
1774
1775 static int qemu_savevm_state(QEMUFile *f)
1776 {
1777     int ret;
1778     MigrationParams params = {
1779         .blk = 0,
1780         .shared = 0
1781     };
1782
1783     if (qemu_savevm_state_blocked(NULL)) {
1784         return -EINVAL;
1785     }
1786
1787     qemu_mutex_unlock_iothread();
1788     qemu_savevm_state_begin(f, &params);
1789     qemu_mutex_lock_iothread();
1790
1791     while (qemu_file_get_error(f) == 0) {
1792         if (qemu_savevm_state_iterate(f) > 0) {
1793             break;
1794         }
1795     }
1796
1797     ret = qemu_file_get_error(f);
1798     if (ret == 0) {
1799         qemu_savevm_state_complete(f);
1800         ret = qemu_file_get_error(f);
1801     }
1802     if (ret != 0) {
1803         qemu_savevm_state_cancel();
1804     }
1805     return ret;
1806 }
1807
1808 static int qemu_save_device_state(QEMUFile *f)
1809 {
1810     SaveStateEntry *se;
1811
1812     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1813     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1814
1815     cpu_synchronize_all_states();
1816
1817     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1818         int len;
1819
1820         if (se->is_ram) {
1821             continue;
1822         }
1823         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1824             continue;
1825         }
1826
1827         /* Section type */
1828         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1829         qemu_put_be32(f, se->section_id);
1830
1831         /* ID string */
1832         len = strlen(se->idstr);
1833         qemu_put_byte(f, len);
1834         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1835
1836         qemu_put_be32(f, se->instance_id);
1837         qemu_put_be32(f, se->version_id);
1838
1839         vmstate_save(f, se);
1840     }
1841
1842     qemu_put_byte(f, QEMU_VM_EOF);
1843
1844     return qemu_file_get_error(f);
1845 }
1846
1847 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1848 {
1849     SaveStateEntry *se;
1850
1851     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1852         if (!strcmp(se->idstr, idstr) &&
1853             (instance_id == se->instance_id ||
1854              instance_id == se->alias_id))
1855             return se;
1856         /* Migrating from an older version? */
1857         if (strstr(se->idstr, idstr) && se->compat) {
1858             if (!strcmp(se->compat->idstr, idstr) &&
1859                 (instance_id == se->compat->instance_id ||
1860                  instance_id == se->alias_id))
1861                 return se;
1862         }
1863     }
1864     return NULL;
1865 }
1866
1867 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1868 {
1869     while(sub && sub->needed) {
1870         if (strcmp(idstr, sub->vmsd->name) == 0) {
1871             return sub->vmsd;
1872         }
1873         sub++;
1874     }
1875     return NULL;
1876 }
1877
1878 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1879                                    void *opaque)
1880 {
1881     while (qemu_peek_byte(f, 0) == QEMU_VM_SUBSECTION) {
1882         char idstr[256];
1883         int ret;
1884         uint8_t version_id, len, size;
1885         const VMStateDescription *sub_vmsd;
1886
1887         len = qemu_peek_byte(f, 1);
1888         if (len < strlen(vmsd->name) + 1) {
1889             /* subsection name has be be "section_name/a" */
1890             return 0;
1891         }
1892         size = qemu_peek_buffer(f, (uint8_t *)idstr, len, 2);
1893         if (size != len) {
1894             return 0;
1895         }
1896         idstr[size] = 0;
1897
1898         if (strncmp(vmsd->name, idstr, strlen(vmsd->name)) != 0) {
1899             /* it don't have a valid subsection name */
1900             return 0;
1901         }
1902         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
1903         if (sub_vmsd == NULL) {
1904             return -ENOENT;
1905         }
1906         qemu_file_skip(f, 1); /* subsection */
1907         qemu_file_skip(f, 1); /* len */
1908         qemu_file_skip(f, len); /* idstr */
1909         version_id = qemu_get_be32(f);
1910
1911         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1912         if (ret) {
1913             return ret;
1914         }
1915     }
1916     return 0;
1917 }
1918
1919 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1920                                     void *opaque)
1921 {
1922     const VMStateSubsection *sub = vmsd->subsections;
1923
1924     while (sub && sub->needed) {
1925         if (sub->needed(opaque)) {
1926             const VMStateDescription *vmsd = sub->vmsd;
1927             uint8_t len;
1928
1929             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1930             len = strlen(vmsd->name);
1931             qemu_put_byte(f, len);
1932             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1933             qemu_put_be32(f, vmsd->version_id);
1934             vmstate_save_state(f, vmsd, opaque);
1935         }
1936         sub++;
1937     }
1938 }
1939
1940 typedef struct LoadStateEntry {
1941     QLIST_ENTRY(LoadStateEntry) entry;
1942     SaveStateEntry *se;
1943     int section_id;
1944     int version_id;
1945 } LoadStateEntry;
1946
1947 int qemu_loadvm_state(QEMUFile *f)
1948 {
1949     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1950         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1951     LoadStateEntry *le, *new_le;
1952     uint8_t section_type;
1953     unsigned int v;
1954     int ret;
1955
1956     if (qemu_savevm_state_blocked(NULL)) {
1957         return -EINVAL;
1958     }
1959
1960     v = qemu_get_be32(f);
1961     if (v != QEMU_VM_FILE_MAGIC)
1962         return -EINVAL;
1963
1964     v = qemu_get_be32(f);
1965     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1966         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1967         return -ENOTSUP;
1968     }
1969     if (v != QEMU_VM_FILE_VERSION)
1970         return -ENOTSUP;
1971
1972     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1973         uint32_t instance_id, version_id, section_id;
1974         SaveStateEntry *se;
1975         char idstr[257];
1976         int len;
1977
1978         switch (section_type) {
1979         case QEMU_VM_SECTION_START:
1980         case QEMU_VM_SECTION_FULL:
1981             /* Read section start */
1982             section_id = qemu_get_be32(f);
1983             len = qemu_get_byte(f);
1984             qemu_get_buffer(f, (uint8_t *)idstr, len);
1985             idstr[len] = 0;
1986             instance_id = qemu_get_be32(f);
1987             version_id = qemu_get_be32(f);
1988
1989             /* Find savevm section */
1990             se = find_se(idstr, instance_id);
1991             if (se == NULL) {
1992                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1993                 ret = -EINVAL;
1994                 goto out;
1995             }
1996
1997             /* Validate version */
1998             if (version_id > se->version_id) {
1999                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
2000                         version_id, idstr, se->version_id);
2001                 ret = -EINVAL;
2002                 goto out;
2003             }
2004
2005             /* Add entry */
2006             le = g_malloc0(sizeof(*le));
2007
2008             le->se = se;
2009             le->section_id = section_id;
2010             le->version_id = version_id;
2011             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
2012
2013             ret = vmstate_load(f, le->se, le->version_id);
2014             if (ret < 0) {
2015                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
2016                         instance_id, idstr);
2017                 goto out;
2018             }
2019             break;
2020         case QEMU_VM_SECTION_PART:
2021         case QEMU_VM_SECTION_END:
2022             section_id = qemu_get_be32(f);
2023
2024             QLIST_FOREACH(le, &loadvm_handlers, entry) {
2025                 if (le->section_id == section_id) {
2026                     break;
2027                 }
2028             }
2029             if (le == NULL) {
2030                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
2031                 ret = -EINVAL;
2032                 goto out;
2033             }
2034
2035             ret = vmstate_load(f, le->se, le->version_id);
2036             if (ret < 0) {
2037                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
2038                         section_id);
2039                 goto out;
2040             }
2041             break;
2042         default:
2043             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
2044             ret = -EINVAL;
2045             goto out;
2046         }
2047     }
2048
2049     cpu_synchronize_all_post_init();
2050
2051     ret = 0;
2052
2053 out:
2054     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
2055         QLIST_REMOVE(le, entry);
2056         g_free(le);
2057     }
2058
2059     if (ret == 0) {
2060         ret = qemu_file_get_error(f);
2061     }
2062
2063     return ret;
2064 }
2065
2066 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
2067                               const char *name)
2068 {
2069     QEMUSnapshotInfo *sn_tab, *sn;
2070     int nb_sns, i, ret;
2071
2072     ret = -ENOENT;
2073     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2074     if (nb_sns < 0)
2075         return ret;
2076     for(i = 0; i < nb_sns; i++) {
2077         sn = &sn_tab[i];
2078         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
2079             *sn_info = *sn;
2080             ret = 0;
2081             break;
2082         }
2083     }
2084     g_free(sn_tab);
2085     return ret;
2086 }
2087
2088 /*
2089  * Deletes snapshots of a given name in all opened images.
2090  */
2091 static int del_existing_snapshots(Monitor *mon, const char *name)
2092 {
2093     BlockDriverState *bs;
2094     QEMUSnapshotInfo sn1, *snapshot = &sn1;
2095     int ret;
2096
2097     bs = NULL;
2098     while ((bs = bdrv_next(bs))) {
2099         if (bdrv_can_snapshot(bs) &&
2100             bdrv_snapshot_find(bs, snapshot, name) >= 0)
2101         {
2102             ret = bdrv_snapshot_delete(bs, name);
2103             if (ret < 0) {
2104                 monitor_printf(mon,
2105                                "Error while deleting snapshot on '%s'\n",
2106                                bdrv_get_device_name(bs));
2107                 return -1;
2108             }
2109         }
2110     }
2111
2112     return 0;
2113 }
2114
2115 void do_savevm(Monitor *mon, const QDict *qdict)
2116 {
2117     BlockDriverState *bs, *bs1;
2118     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
2119     int ret;
2120     QEMUFile *f;
2121     int saved_vm_running;
2122     uint64_t vm_state_size;
2123     qemu_timeval tv;
2124     struct tm tm;
2125     const char *name = qdict_get_try_str(qdict, "name");
2126
2127     /* Verify if there is a device that doesn't support snapshots and is writable */
2128     bs = NULL;
2129     while ((bs = bdrv_next(bs))) {
2130
2131         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2132             continue;
2133         }
2134
2135         if (!bdrv_can_snapshot(bs)) {
2136             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
2137                                bdrv_get_device_name(bs));
2138             return;
2139         }
2140     }
2141
2142     bs = bdrv_snapshots();
2143     if (!bs) {
2144         monitor_printf(mon, "No block device can accept snapshots\n");
2145         return;
2146     }
2147
2148     saved_vm_running = runstate_is_running();
2149     vm_stop(RUN_STATE_SAVE_VM);
2150
2151     memset(sn, 0, sizeof(*sn));
2152
2153     /* fill auxiliary fields */
2154     qemu_gettimeofday(&tv);
2155     sn->date_sec = tv.tv_sec;
2156     sn->date_nsec = tv.tv_usec * 1000;
2157     sn->vm_clock_nsec = qemu_get_clock_ns(vm_clock);
2158
2159     if (name) {
2160         ret = bdrv_snapshot_find(bs, old_sn, name);
2161         if (ret >= 0) {
2162             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
2163             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
2164         } else {
2165             pstrcpy(sn->name, sizeof(sn->name), name);
2166         }
2167     } else {
2168         /* cast below needed for OpenBSD where tv_sec is still 'long' */
2169         localtime_r((const time_t *)&tv.tv_sec, &tm);
2170         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
2171     }
2172
2173     /* Delete old snapshots of the same name */
2174     if (name && del_existing_snapshots(mon, name) < 0) {
2175         goto the_end;
2176     }
2177
2178     /* save the VM state */
2179     f = qemu_fopen_bdrv(bs, 1);
2180     if (!f) {
2181         monitor_printf(mon, "Could not open VM state file\n");
2182         goto the_end;
2183     }
2184     ret = qemu_savevm_state(f);
2185     vm_state_size = qemu_ftell(f);
2186     qemu_fclose(f);
2187     if (ret < 0) {
2188         monitor_printf(mon, "Error %d while writing VM\n", ret);
2189         goto the_end;
2190     }
2191
2192     /* create the snapshots */
2193
2194     bs1 = NULL;
2195     while ((bs1 = bdrv_next(bs1))) {
2196         if (bdrv_can_snapshot(bs1)) {
2197             /* Write VM state size only to the image that contains the state */
2198             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
2199             ret = bdrv_snapshot_create(bs1, sn);
2200             if (ret < 0) {
2201                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
2202                                bdrv_get_device_name(bs1));
2203             }
2204         }
2205     }
2206
2207  the_end:
2208     if (saved_vm_running)
2209         vm_start();
2210 }
2211
2212 void qmp_xen_save_devices_state(const char *filename, Error **errp)
2213 {
2214     QEMUFile *f;
2215     int saved_vm_running;
2216     int ret;
2217
2218     saved_vm_running = runstate_is_running();
2219     vm_stop(RUN_STATE_SAVE_VM);
2220
2221     f = qemu_fopen(filename, "wb");
2222     if (!f) {
2223         error_set(errp, QERR_OPEN_FILE_FAILED, filename);
2224         goto the_end;
2225     }
2226     ret = qemu_save_device_state(f);
2227     qemu_fclose(f);
2228     if (ret < 0) {
2229         error_set(errp, QERR_IO_ERROR);
2230     }
2231
2232  the_end:
2233     if (saved_vm_running)
2234         vm_start();
2235 }
2236
2237 int load_vmstate(const char *name)
2238 {
2239     BlockDriverState *bs, *bs_vm_state;
2240     QEMUSnapshotInfo sn;
2241     QEMUFile *f;
2242     int ret;
2243
2244     bs_vm_state = bdrv_snapshots();
2245     if (!bs_vm_state) {
2246         error_report("No block device supports snapshots");
2247         return -ENOTSUP;
2248     }
2249
2250     /* Don't even try to load empty VM states */
2251     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
2252     if (ret < 0) {
2253         return ret;
2254     } else if (sn.vm_state_size == 0) {
2255         error_report("This is a disk-only snapshot. Revert to it offline "
2256             "using qemu-img.");
2257         return -EINVAL;
2258     }
2259
2260     /* Verify if there is any device that doesn't support snapshots and is
2261     writable and check if the requested snapshot is available too. */
2262     bs = NULL;
2263     while ((bs = bdrv_next(bs))) {
2264
2265         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2266             continue;
2267         }
2268
2269         if (!bdrv_can_snapshot(bs)) {
2270             error_report("Device '%s' is writable but does not support snapshots.",
2271                                bdrv_get_device_name(bs));
2272             return -ENOTSUP;
2273         }
2274
2275         ret = bdrv_snapshot_find(bs, &sn, name);
2276         if (ret < 0) {
2277             error_report("Device '%s' does not have the requested snapshot '%s'",
2278                            bdrv_get_device_name(bs), name);
2279             return ret;
2280         }
2281     }
2282
2283     /* Flush all IO requests so they don't interfere with the new state.  */
2284     bdrv_drain_all();
2285
2286     bs = NULL;
2287     while ((bs = bdrv_next(bs))) {
2288         if (bdrv_can_snapshot(bs)) {
2289             ret = bdrv_snapshot_goto(bs, name);
2290             if (ret < 0) {
2291                 error_report("Error %d while activating snapshot '%s' on '%s'",
2292                              ret, name, bdrv_get_device_name(bs));
2293                 return ret;
2294             }
2295         }
2296     }
2297
2298     /* restore the VM state */
2299     f = qemu_fopen_bdrv(bs_vm_state, 0);
2300     if (!f) {
2301         error_report("Could not open VM state file");
2302         return -EINVAL;
2303     }
2304
2305     qemu_system_reset(VMRESET_SILENT);
2306     ret = qemu_loadvm_state(f);
2307
2308     qemu_fclose(f);
2309     if (ret < 0) {
2310         error_report("Error %d while loading VM state", ret);
2311         return ret;
2312     }
2313
2314     return 0;
2315 }
2316
2317 void do_delvm(Monitor *mon, const QDict *qdict)
2318 {
2319     BlockDriverState *bs, *bs1;
2320     int ret;
2321     const char *name = qdict_get_str(qdict, "name");
2322
2323     bs = bdrv_snapshots();
2324     if (!bs) {
2325         monitor_printf(mon, "No block device supports snapshots\n");
2326         return;
2327     }
2328
2329     bs1 = NULL;
2330     while ((bs1 = bdrv_next(bs1))) {
2331         if (bdrv_can_snapshot(bs1)) {
2332             ret = bdrv_snapshot_delete(bs1, name);
2333             if (ret < 0) {
2334                 if (ret == -ENOTSUP)
2335                     monitor_printf(mon,
2336                                    "Snapshots not supported on device '%s'\n",
2337                                    bdrv_get_device_name(bs1));
2338                 else
2339                     monitor_printf(mon, "Error %d while deleting snapshot on "
2340                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2341             }
2342         }
2343     }
2344 }
2345
2346 void do_info_snapshots(Monitor *mon, const QDict *qdict)
2347 {
2348     BlockDriverState *bs, *bs1;
2349     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2350     int nb_sns, i, ret, available;
2351     int total;
2352     int *available_snapshots;
2353     char buf[256];
2354
2355     bs = bdrv_snapshots();
2356     if (!bs) {
2357         monitor_printf(mon, "No available block device supports snapshots\n");
2358         return;
2359     }
2360
2361     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2362     if (nb_sns < 0) {
2363         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2364         return;
2365     }
2366
2367     if (nb_sns == 0) {
2368         monitor_printf(mon, "There is no snapshot available.\n");
2369         return;
2370     }
2371
2372     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
2373     total = 0;
2374     for (i = 0; i < nb_sns; i++) {
2375         sn = &sn_tab[i];
2376         available = 1;
2377         bs1 = NULL;
2378
2379         while ((bs1 = bdrv_next(bs1))) {
2380             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2381                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2382                 if (ret < 0) {
2383                     available = 0;
2384                     break;
2385                 }
2386             }
2387         }
2388
2389         if (available) {
2390             available_snapshots[total] = i;
2391             total++;
2392         }
2393     }
2394
2395     if (total > 0) {
2396         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2397         for (i = 0; i < total; i++) {
2398             sn = &sn_tab[available_snapshots[i]];
2399             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2400         }
2401     } else {
2402         monitor_printf(mon, "There is no suitable snapshot available\n");
2403     }
2404
2405     g_free(sn_tab);
2406     g_free(available_snapshots);
2407
2408 }
2409
2410 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
2411 {
2412     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
2413                        memory_region_name(mr), dev);
2414 }
2415
2416 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
2417 {
2418     /* Nothing do to while the implementation is in RAMBlock */
2419 }
2420
2421 void vmstate_register_ram_global(MemoryRegion *mr)
2422 {
2423     vmstate_register_ram(mr, NULL);
2424 }