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 #include "qemu/iov.h"
  43 #include "block/snapshot.h"
  44 #include "block/qapi.h"
  45
  46 #define SELF_ANNOUNCE_ROUNDS 5
  47
  48 #ifndef ETH_P_RARP
  49 #define ETH_P_RARP 0x8035
  50 #endif
  51 #define ARP_HTYPE_ETH 0x0001
  52 #define ARP_PTYPE_IP 0x0800
  53 #define ARP_OP_REQUEST_REV 0x3
  54
  55 static int announce_self_create(uint8_t *buf,
  56                                 uint8_t *mac_addr)
  57 {
  58     /* Ethernet header. */
  59     memset(buf, 0xff, 6);         /* destination MAC addr */
  60     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
  61     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
  62
  63     /* RARP header. */
  64     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
  65     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
  66     *(buf + 18) = 6; /* hardware addr length (ethernet) */
  67     *(buf + 19) = 4; /* protocol addr length (IPv4) */
  68     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
  69     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
  70     memset(buf + 28, 0x00, 4);     /* source protocol addr */
  71     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
  72     memset(buf + 38, 0x00, 4);     /* target protocol addr */
  73
  74     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
  75     memset(buf + 42, 0x00, 18);
  76
  77     return 60; /* len (FCS will be added by hardware) */
  78 }
  79
  80 static void qemu_announce_self_iter(NICState *nic, void *opaque)
  81 {
  82     uint8_t buf[60];
  83     int len;
  84
  85     len = announce_self_create(buf, nic->conf->macaddr.a);
  86
  87     qemu_send_packet_raw(qemu_get_queue(nic), buf, len);
  88 }
  89
  90
  91 static void qemu_announce_self_once(void *opaque)
  92 {
  93     static int count = SELF_ANNOUNCE_ROUNDS;
  94     QEMUTimer *timer = *(QEMUTimer **)opaque;
  95
  96     qemu_foreach_nic(qemu_announce_self_iter, NULL);
  97
  98     if (--count) {
  99         /* delay 50ms, 150ms, 250ms, ... */
 100         timer_mod(timer, qemu_clock_get_ms(QEMU_CLOCK_REALTIME) +
 101                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
 102     } else {
 103             timer_del(timer);
 104             timer_free(timer);
 105     }
 106 }
 107
 108 void qemu_announce_self(void)
 109 {
 110         static QEMUTimer *timer;
 111         timer = timer_new_ms(QEMU_CLOCK_REALTIME, qemu_announce_self_once, &timer);
 112         qemu_announce_self_once(&timer);
 113 }
 114
 115 /***********************************************************/
 116 /* savevm/loadvm support */
 117
 118 #define IO_BUF_SIZE 32768
 119 #define MAX_IOV_SIZE MIN(IOV_MAX, 64)
 120
 121 struct QEMUFile {
 122     const QEMUFileOps *ops;
 123     void *opaque;
 124
 125     int64_t bytes_xfer;
 126     int64_t xfer_limit;
 127
 128     int64_t pos; /* start of buffer when writing, end of buffer
 129                     when reading */
 130     int buf_index;
 131     int buf_size; /* 0 when writing */
 132     uint8_t buf[IO_BUF_SIZE];
 133
 134     struct iovec iov[MAX_IOV_SIZE];
 135     unsigned int iovcnt;
 136
 137     int last_error;
 138 };
 139
 140 typedef struct QEMUFileStdio
 141 {
 142     FILE *stdio_file;
 143     QEMUFile *file;
 144 } QEMUFileStdio;
 145
 146 typedef struct QEMUFileSocket
 147 {
 148     int fd;
 149     QEMUFile *file;
 150 } QEMUFileSocket;
 151
 152 static ssize_t socket_writev_buffer(void *opaque, struct iovec *iov, int iovcnt,
 153                                     int64_t pos)
 154 {
 155     QEMUFileSocket *s = opaque;
 156     ssize_t len;
 157     ssize_t size = iov_size(iov, iovcnt);
 158
 159     len = iov_send(s->fd, iov, iovcnt, 0, size);
 160     if (len < size) {
 161         len = -socket_error();
 162     }
 163     return len;
 164 }
 165
 166 static int socket_get_fd(void *opaque)
 167 {
 168     QEMUFileSocket *s = opaque;
 169
 170     return s->fd;
 171 }
 172
 173 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 174 {
 175     QEMUFileSocket *s = opaque;
 176     ssize_t len;
 177
 178     for (;;) {
 179         len = qemu_recv(s->fd, buf, size, 0);
 180         if (len != -1) {
 181             break;
 182         }
 183         if (socket_error() == EAGAIN) {
 184             yield_until_fd_readable(s->fd);
 185         } else if (socket_error() != EINTR) {
 186             break;
 187         }
 188     }
 189
 190     if (len == -1) {
 191         len = -socket_error();
 192     }
 193     return len;
 194 }
 195
 196 static int socket_close(void *opaque)
 197 {
 198     QEMUFileSocket *s = opaque;
 199     closesocket(s->fd);
 200     g_free(s);
 201     return 0;
 202 }
 203
 204 static int stdio_get_fd(void *opaque)
 205 {
 206     QEMUFileStdio *s = opaque;
 207
 208     return fileno(s->stdio_file);
 209 }
 210
 211 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 212 {
 213     QEMUFileStdio *s = opaque;
 214     return fwrite(buf, 1, size, s->stdio_file);
 215 }
 216
 217 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 218 {
 219     QEMUFileStdio *s = opaque;
 220     FILE *fp = s->stdio_file;
 221     int bytes;
 222
 223     for (;;) {
 224         clearerr(fp);
 225         bytes = fread(buf, 1, size, fp);
 226         if (bytes != 0 || !ferror(fp)) {
 227             break;
 228         }
 229         if (errno == EAGAIN) {
 230             yield_until_fd_readable(fileno(fp));
 231         } else if (errno != EINTR) {
 232             break;
 233         }
 234     }
 235     return bytes;
 236 }
 237
 238 static int stdio_pclose(void *opaque)
 239 {
 240     QEMUFileStdio *s = opaque;
 241     int ret;
 242     ret = pclose(s->stdio_file);
 243     if (ret == -1) {
 244         ret = -errno;
 245     } else if (!WIFEXITED(ret) || WEXITSTATUS(ret) != 0) {
 246         /* close succeeded, but non-zero exit code: */
 247         ret = -EIO; /* fake errno value */
 248     }
 249     g_free(s);
 250     return ret;
 251 }
 252
 253 static int stdio_fclose(void *opaque)
 254 {
 255     QEMUFileStdio *s = opaque;
 256     int ret = 0;
 257
 258     if (s->file->ops->put_buffer || s->file->ops->writev_buffer) {
 259         int fd = fileno(s->stdio_file);
 260         struct stat st;
 261
 262         ret = fstat(fd, &st);
 263         if (ret == 0 && S_ISREG(st.st_mode)) {
 264             /*
 265              * If the file handle is a regular file make sure the
 266              * data is flushed to disk before signaling success.
 267              */
 268             ret = fsync(fd);
 269             if (ret != 0) {
 270                 ret = -errno;
 271                 return ret;
 272             }
 273         }
 274     }
 275     if (fclose(s->stdio_file) == EOF) {
 276         ret = -errno;
 277     }
 278     g_free(s);
 279     return ret;
 280 }
 281
 282 static const QEMUFileOps stdio_pipe_read_ops = {
 283     .get_fd =     stdio_get_fd,
 284     .get_buffer = stdio_get_buffer,
 285     .close =      stdio_pclose
 286 };
 287
 288 static const QEMUFileOps stdio_pipe_write_ops = {
 289     .get_fd =     stdio_get_fd,
 290     .put_buffer = stdio_put_buffer,
 291     .close =      stdio_pclose
 292 };
 293
 294 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 295 {
 296     FILE *stdio_file;
 297     QEMUFileStdio *s;
 298
 299     if (mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 300         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 301         return NULL;
 302     }
 303
 304     stdio_file = popen(command, mode);
 305     if (stdio_file == NULL) {
 306         return NULL;
 307     }
 308
 309     s = g_malloc0(sizeof(QEMUFileStdio));
 310
 311     s->stdio_file = stdio_file;
 312
 313     if(mode[0] == 'r') {
 314         s->file = qemu_fopen_ops(s, &stdio_pipe_read_ops);
 315     } else {
 316         s->file = qemu_fopen_ops(s, &stdio_pipe_write_ops);
 317     }
 318     return s->file;
 319 }
 320
 321 static const QEMUFileOps stdio_file_read_ops = {
 322     .get_fd =     stdio_get_fd,
 323     .get_buffer = stdio_get_buffer,
 324     .close =      stdio_fclose
 325 };
 326
 327 static const QEMUFileOps stdio_file_write_ops = {
 328     .get_fd =     stdio_get_fd,
 329     .put_buffer = stdio_put_buffer,
 330     .close =      stdio_fclose
 331 };
 332
 333 static ssize_t unix_writev_buffer(void *opaque, struct iovec *iov, int iovcnt,
 334                                   int64_t pos)
 335 {
 336     QEMUFileSocket *s = opaque;
 337     ssize_t len, offset;
 338     ssize_t size = iov_size(iov, iovcnt);
 339     ssize_t total = 0;
 340
 341     assert(iovcnt > 0);
 342     offset = 0;
 343     while (size > 0) {
 344         /* Find the next start position; skip all full-sized vector elements  */
 345         while (offset >= iov[0].iov_len) {
 346             offset -= iov[0].iov_len;
 347             iov++, iovcnt--;
 348         }
 349
 350         /* skip `offset' bytes from the (now) first element, undo it on exit */
 351         assert(iovcnt > 0);
 352         iov[0].iov_base += offset;
 353         iov[0].iov_len -= offset;
 354
 355         do {
 356             len = writev(s->fd, iov, iovcnt);
 357         } while (len == -1 && errno == EINTR);
 358         if (len == -1) {
 359             return -errno;
 360         }
 361
 362         /* Undo the changes above */
 363         iov[0].iov_base -= offset;
 364         iov[0].iov_len += offset;
 365
 366         /* Prepare for the next iteration */
 367         offset += len;
 368         total += len;
 369         size -= len;
 370     }
 371
 372     return total;
 373 }
 374
 375 static int unix_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 376 {
 377     QEMUFileSocket *s = opaque;
 378     ssize_t len;
 379
 380     for (;;) {
 381         len = read(s->fd, buf, size);
 382         if (len != -1) {
 383             break;
 384         }
 385         if (errno == EAGAIN) {
 386             yield_until_fd_readable(s->fd);
 387         } else if (errno != EINTR) {
 388             break;
 389         }
 390     }
 391
 392     if (len == -1) {
 393         len = -errno;
 394     }
 395     return len;
 396 }
 397
 398 static int unix_close(void *opaque)
 399 {
 400     QEMUFileSocket *s = opaque;
 401     close(s->fd);
 402     g_free(s);
 403     return 0;
 404 }
 405
 406 static const QEMUFileOps unix_read_ops = {
 407     .get_fd =     socket_get_fd,
 408     .get_buffer = unix_get_buffer,
 409     .close =      unix_close
 410 };
 411
 412 static const QEMUFileOps unix_write_ops = {
 413     .get_fd =     socket_get_fd,
 414     .writev_buffer = unix_writev_buffer,
 415     .close =      unix_close
 416 };
 417
 418 QEMUFile *qemu_fdopen(int fd, const char *mode)
 419 {
 420     QEMUFileSocket *s;
 421
 422     if (mode == NULL ||
 423         (mode[0] != 'r' && mode[0] != 'w') ||
 424         mode[1] != 'b' || mode[2] != 0) {
 425         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 426         return NULL;
 427     }
 428
 429     s = g_malloc0(sizeof(QEMUFileSocket));
 430     s->fd = fd;
 431
 432     if(mode[0] == 'r') {
 433         s->file = qemu_fopen_ops(s, &unix_read_ops);
 434     } else {
 435         s->file = qemu_fopen_ops(s, &unix_write_ops);
 436     }
 437     return s->file;
 438 }
 439
 440 static const QEMUFileOps socket_read_ops = {
 441     .get_fd =     socket_get_fd,
 442     .get_buffer = socket_get_buffer,
 443     .close =      socket_close
 444 };
 445
 446 static const QEMUFileOps socket_write_ops = {
 447     .get_fd =     socket_get_fd,
 448     .writev_buffer = socket_writev_buffer,
 449     .close =      socket_close
 450 };
 451
 452 bool qemu_file_mode_is_not_valid(const char *mode)
 453 {
 454     if (mode == NULL ||
 455         (mode[0] != 'r' && mode[0] != 'w') ||
 456         mode[1] != 'b' || mode[2] != 0) {
 457         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 458         return true;
 459     }
 460
 461     return false;
 462 }
 463
 464 QEMUFile *qemu_fopen_socket(int fd, const char *mode)
 465 {
 466     QEMUFileSocket *s;
 467
 468     if (qemu_file_mode_is_not_valid(mode)) {
 469         return NULL;
 470     }
 471
 472     s = g_malloc0(sizeof(QEMUFileSocket));
 473     s->fd = fd;
 474     if (mode[0] == 'w') {
 475         qemu_set_block(s->fd);
 476         s->file = qemu_fopen_ops(s, &socket_write_ops);
 477     } else {
 478         s->file = qemu_fopen_ops(s, &socket_read_ops);
 479     }
 480     return s->file;
 481 }
 482
 483 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 484 {
 485     QEMUFileStdio *s;
 486
 487     if (qemu_file_mode_is_not_valid(mode)) {
 488         return NULL;
 489     }
 490
 491     s = g_malloc0(sizeof(QEMUFileStdio));
 492
 493     s->stdio_file = fopen(filename, mode);
 494     if (!s->stdio_file)
 495         goto fail;
 496
 497     if(mode[0] == 'w') {
 498         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 499     } else {
 500         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 501     }
 502     return s->file;
 503 fail:
 504     g_free(s);
 505     return NULL;
 506 }
 507
 508 static ssize_t block_writev_buffer(void *opaque, struct iovec *iov, int iovcnt,
 509                                    int64_t pos)
 510 {
 511     int ret;
 512     QEMUIOVector qiov;
 513
 514     qemu_iovec_init_external(&qiov, iov, iovcnt);
 515     ret = bdrv_writev_vmstate(opaque, &qiov, pos);
 516     if (ret < 0) {
 517         return ret;
 518     }
 519
 520     return qiov.size;
 521 }
 522
 523 static int block_put_buffer(void *opaque, const uint8_t *buf,
 524                            int64_t pos, int size)
 525 {
 526     bdrv_save_vmstate(opaque, buf, pos, size);
 527     return size;
 528 }
 529
 530 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 531 {
 532     return bdrv_load_vmstate(opaque, buf, pos, size);
 533 }
 534
 535 static int bdrv_fclose(void *opaque)
 536 {
 537     return bdrv_flush(opaque);
 538 }
 539
 540 static const QEMUFileOps bdrv_read_ops = {
 541     .get_buffer = block_get_buffer,
 542     .close =      bdrv_fclose
 543 };
 544
 545 static const QEMUFileOps bdrv_write_ops = {
 546     .put_buffer     = block_put_buffer,
 547     .writev_buffer  = block_writev_buffer,
 548     .close          = bdrv_fclose
 549 };
 550
 551 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 552 {
 553     if (is_writable)
 554         return qemu_fopen_ops(bs, &bdrv_write_ops);
 555     return qemu_fopen_ops(bs, &bdrv_read_ops);
 556 }
 557
 558 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
 559 {
 560     QEMUFile *f;
 561
 562     f = g_malloc0(sizeof(QEMUFile));
 563
 564     f->opaque = opaque;
 565     f->ops = ops;
 566     return f;
 567 }
 568
 569 /*
 570  * Get last error for stream f
 571  *
 572  * Return negative error value if there has been an error on previous
 573  * operations, return 0 if no error happened.
 574  *
 575  */
 576 int qemu_file_get_error(QEMUFile *f)
 577 {
 578     return f->last_error;
 579 }
 580
 581 static void qemu_file_set_error(QEMUFile *f, int ret)
 582 {
 583     if (f->last_error == 0) {
 584         f->last_error = ret;
 585     }
 586 }
 587
 588 static inline bool qemu_file_is_writable(QEMUFile *f)
 589 {
 590     return f->ops->writev_buffer || f->ops->put_buffer;
 591 }
 592
 593 /**
 594  * Flushes QEMUFile buffer
 595  *
 596  * If there is writev_buffer QEMUFileOps it uses it otherwise uses
 597  * put_buffer ops.
 598  */
 599 void qemu_fflush(QEMUFile *f)
 600 {
 601     ssize_t ret = 0;
 602
 603     if (!qemu_file_is_writable(f)) {
 604         return;
 605     }
 606
 607     if (f->ops->writev_buffer) {
 608         if (f->iovcnt > 0) {
 609             ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt, f->pos);
 610         }
 611     } else {
 612         if (f->buf_index > 0) {
 613             ret = f->ops->put_buffer(f->opaque, f->buf, f->pos, f->buf_index);
 614         }
 615     }
 616     if (ret >= 0) {
 617         f->pos += ret;
 618     }
 619     f->buf_index = 0;
 620     f->iovcnt = 0;
 621     if (ret < 0) {
 622         qemu_file_set_error(f, ret);
 623     }
 624 }
 625
 626 void ram_control_before_iterate(QEMUFile *f, uint64_t flags)
 627 {
 628     int ret = 0;
 629
 630     if (f->ops->before_ram_iterate) {
 631         ret = f->ops->before_ram_iterate(f, f->opaque, flags);
 632         if (ret < 0) {
 633             qemu_file_set_error(f, ret);
 634         }
 635     }
 636 }
 637
 638 void ram_control_after_iterate(QEMUFile *f, uint64_t flags)
 639 {
 640     int ret = 0;
 641
 642     if (f->ops->after_ram_iterate) {
 643         ret = f->ops->after_ram_iterate(f, f->opaque, flags);
 644         if (ret < 0) {
 645             qemu_file_set_error(f, ret);
 646         }
 647     }
 648 }
 649
 650 void ram_control_load_hook(QEMUFile *f, uint64_t flags)
 651 {
 652     int ret = -EINVAL;
 653
 654     if (f->ops->hook_ram_load) {
 655         ret = f->ops->hook_ram_load(f, f->opaque, flags);
 656         if (ret < 0) {
 657             qemu_file_set_error(f, ret);
 658         }
 659     } else {
 660         qemu_file_set_error(f, ret);
 661     }
 662 }
 663
 664 size_t ram_control_save_page(QEMUFile *f, ram_addr_t block_offset,
 665                          ram_addr_t offset, size_t size, int *bytes_sent)
 666 {
 667     if (f->ops->save_page) {
 668         int ret = f->ops->save_page(f, f->opaque, block_offset,
 669                                     offset, size, bytes_sent);
 670
 671         if (ret != RAM_SAVE_CONTROL_DELAYED) {
 672             if (bytes_sent && *bytes_sent > 0) {
 673                 qemu_update_position(f, *bytes_sent);
 674             } else if (ret < 0) {
 675                 qemu_file_set_error(f, ret);
 676             }
 677         }
 678
 679         return ret;
 680     }
 681
 682     return RAM_SAVE_CONTROL_NOT_SUPP;
 683 }
 684
 685 static void qemu_fill_buffer(QEMUFile *f)
 686 {
 687     int len;
 688     int pending;
 689
 690     assert(!qemu_file_is_writable(f));
 691
 692     pending = f->buf_size - f->buf_index;
 693     if (pending > 0) {
 694         memmove(f->buf, f->buf + f->buf_index, pending);
 695     }
 696     f->buf_index = 0;
 697     f->buf_size = pending;
 698
 699     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->pos,
 700                         IO_BUF_SIZE - pending);
 701     if (len > 0) {
 702         f->buf_size += len;
 703         f->pos += len;
 704     } else if (len == 0) {
 705         qemu_file_set_error(f, -EIO);
 706     } else if (len != -EAGAIN)
 707         qemu_file_set_error(f, len);
 708 }
 709
 710 int qemu_get_fd(QEMUFile *f)
 711 {
 712     if (f->ops->get_fd) {
 713         return f->ops->get_fd(f->opaque);
 714     }
 715     return -1;
 716 }
 717
 718 void qemu_update_position(QEMUFile *f, size_t size)
 719 {
 720     f->pos += size;
 721 }
 722
 723 /** Closes the file
 724  *
 725  * Returns negative error value if any error happened on previous operations or
 726  * while closing the file. Returns 0 or positive number on success.
 727  *
 728  * The meaning of return value on success depends on the specific backend
 729  * being used.
 730  */
 731 int qemu_fclose(QEMUFile *f)
 732 {
 733     int ret;
 734     qemu_fflush(f);
 735     ret = qemu_file_get_error(f);
 736
 737     if (f->ops->close) {
 738         int ret2 = f->ops->close(f->opaque);
 739         if (ret >= 0) {
 740             ret = ret2;
 741         }
 742     }
 743     /* If any error was spotted before closing, we should report it
 744      * instead of the close() return value.
 745      */
 746     if (f->last_error) {
 747         ret = f->last_error;
 748     }
 749     g_free(f);
 750     return ret;
 751 }
 752
 753 static void add_to_iovec(QEMUFile *f, const uint8_t *buf, int size)
 754 {
 755     /* check for adjacent buffer and coalesce them */
 756     if (f->iovcnt > 0 && buf == f->iov[f->iovcnt - 1].iov_base +
 757         f->iov[f->iovcnt - 1].iov_len) {
 758         f->iov[f->iovcnt - 1].iov_len += size;
 759     } else {
 760         f->iov[f->iovcnt].iov_base = (uint8_t *)buf;
 761         f->iov[f->iovcnt++].iov_len = size;
 762     }
 763
 764     if (f->iovcnt >= MAX_IOV_SIZE) {
 765         qemu_fflush(f);
 766     }
 767 }
 768
 769 void qemu_put_buffer_async(QEMUFile *f, const uint8_t *buf, int size)
 770 {
 771     if (!f->ops->writev_buffer) {
 772         qemu_put_buffer(f, buf, size);
 773         return;
 774     }
 775
 776     if (f->last_error) {
 777         return;
 778     }
 779
 780     f->bytes_xfer += size;
 781     add_to_iovec(f, buf, size);
 782 }
 783
 784 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 785 {
 786     int l;
 787
 788     if (f->last_error) {
 789         return;
 790     }
 791
 792     while (size > 0) {
 793         l = IO_BUF_SIZE - f->buf_index;
 794         if (l > size)
 795             l = size;
 796         memcpy(f->buf + f->buf_index, buf, l);
 797         f->bytes_xfer += size;
 798         if (f->ops->writev_buffer) {
 799             add_to_iovec(f, f->buf + f->buf_index, l);
 800         }
 801         f->buf_index += l;
 802         if (f->buf_index == IO_BUF_SIZE) {
 803             qemu_fflush(f);
 804         }
 805         if (qemu_file_get_error(f)) {
 806             break;
 807         }
 808         buf += l;
 809         size -= l;
 810     }
 811 }
 812
 813 void qemu_put_byte(QEMUFile *f, int v)
 814 {
 815     if (f->last_error) {
 816         return;
 817     }
 818
 819     f->buf[f->buf_index] = v;
 820     f->bytes_xfer++;
 821     if (f->ops->writev_buffer) {
 822         add_to_iovec(f, f->buf + f->buf_index, 1);
 823     }
 824     f->buf_index++;
 825     if (f->buf_index == IO_BUF_SIZE) {
 826         qemu_fflush(f);
 827     }
 828 }
 829
 830 static void qemu_file_skip(QEMUFile *f, int size)
 831 {
 832     if (f->buf_index + size <= f->buf_size) {
 833         f->buf_index += size;
 834     }
 835 }
 836
 837 static int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset)
 838 {
 839     int pending;
 840     int index;
 841
 842     assert(!qemu_file_is_writable(f));
 843
 844     index = f->buf_index + offset;
 845     pending = f->buf_size - index;
 846     if (pending < size) {
 847         qemu_fill_buffer(f);
 848         index = f->buf_index + offset;
 849         pending = f->buf_size - index;
 850     }
 851
 852     if (pending <= 0) {
 853         return 0;
 854     }
 855     if (size > pending) {
 856         size = pending;
 857     }
 858
 859     memcpy(buf, f->buf + index, size);
 860     return size;
 861 }
 862
 863 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size)
 864 {
 865     int pending = size;
 866     int done = 0;
 867
 868     while (pending > 0) {
 869         int res;
 870
 871         res = qemu_peek_buffer(f, buf, pending, 0);
 872         if (res == 0) {
 873             return done;
 874         }
 875         qemu_file_skip(f, res);
 876         buf += res;
 877         pending -= res;
 878         done += res;
 879     }
 880     return done;
 881 }
 882
 883 static int qemu_peek_byte(QEMUFile *f, int offset)
 884 {
 885     int index = f->buf_index + offset;
 886
 887     assert(!qemu_file_is_writable(f));
 888
 889     if (index >= f->buf_size) {
 890         qemu_fill_buffer(f);
 891         index = f->buf_index + offset;
 892         if (index >= f->buf_size) {
 893             return 0;
 894         }
 895     }
 896     return f->buf[index];
 897 }
 898
 899 int qemu_get_byte(QEMUFile *f)
 900 {
 901     int result;
 902
 903     result = qemu_peek_byte(f, 0);
 904     qemu_file_skip(f, 1);
 905     return result;
 906 }
 907
 908 int64_t qemu_ftell(QEMUFile *f)
 909 {
 910     qemu_fflush(f);
 911     return f->pos;
 912 }
 913
 914 int qemu_file_rate_limit(QEMUFile *f)
 915 {
 916     if (qemu_file_get_error(f)) {
 917         return 1;
 918     }
 919     if (f->xfer_limit > 0 && f->bytes_xfer > f->xfer_limit) {
 920         return 1;
 921     }
 922     return 0;
 923 }
 924
 925 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 926 {
 927     return f->xfer_limit;
 928 }
 929
 930 void qemu_file_set_rate_limit(QEMUFile *f, int64_t limit)
 931 {
 932     f->xfer_limit = limit;
 933 }
 934
 935 void qemu_file_reset_rate_limit(QEMUFile *f)
 936 {
 937     f->bytes_xfer = 0;
 938 }
 939
 940 void qemu_put_be16(QEMUFile *f, unsigned int v)
 941 {
 942     qemu_put_byte(f, v >> 8);
 943     qemu_put_byte(f, v);
 944 }
 945
 946 void qemu_put_be32(QEMUFile *f, unsigned int v)
 947 {
 948     qemu_put_byte(f, v >> 24);
 949     qemu_put_byte(f, v >> 16);
 950     qemu_put_byte(f, v >> 8);
 951     qemu_put_byte(f, v);
 952 }
 953
 954 void qemu_put_be64(QEMUFile *f, uint64_t v)
 955 {
 956     qemu_put_be32(f, v >> 32);
 957     qemu_put_be32(f, v);
 958 }
 959
 960 unsigned int qemu_get_be16(QEMUFile *f)
 961 {
 962     unsigned int v;
 963     v = qemu_get_byte(f) << 8;
 964     v |= qemu_get_byte(f);
 965     return v;
 966 }
 967
 968 unsigned int qemu_get_be32(QEMUFile *f)
 969 {
 970     unsigned int v;
 971     v = qemu_get_byte(f) << 24;
 972     v |= qemu_get_byte(f) << 16;
 973     v |= qemu_get_byte(f) << 8;
 974     v |= qemu_get_byte(f);
 975     return v;
 976 }
 977
 978 uint64_t qemu_get_be64(QEMUFile *f)
 979 {
 980     uint64_t v;
 981     v = (uint64_t)qemu_get_be32(f) << 32;
 982     v |= qemu_get_be32(f);
 983     return v;
 984 }
 985
 986
 987 /* timer */
 988
 989 void timer_put(QEMUFile *f, QEMUTimer *ts)
 990 {
 991     uint64_t expire_time;
 992
 993     expire_time = timer_expire_time_ns(ts);
 994     qemu_put_be64(f, expire_time);
 995 }
 996
 997 void timer_get(QEMUFile *f, QEMUTimer *ts)
 998 {
 999     uint64_t expire_time;
1000
1001     expire_time = qemu_get_be64(f);
1002     if (expire_time != -1) {
1003         timer_mod_ns(ts, expire_time);
1004     } else {
1005         timer_del(ts);
1006     }
1007 }
1008
1009
1010 /* bool */
1011
1012 static int get_bool(QEMUFile *f, void *pv, size_t size)
1013 {
1014     bool *v = pv;
1015     *v = qemu_get_byte(f);
1016     return 0;
1017 }
1018
1019 static void put_bool(QEMUFile *f, void *pv, size_t size)
1020 {
1021     bool *v = pv;
1022     qemu_put_byte(f, *v);
1023 }
1024
1025 const VMStateInfo vmstate_info_bool = {
1026     .name = "bool",
1027     .get  = get_bool,
1028     .put  = put_bool,
1029 };
1030
1031 /* 8 bit int */
1032
1033 static int get_int8(QEMUFile *f, void *pv, size_t size)
1034 {
1035     int8_t *v = pv;
1036     qemu_get_s8s(f, v);
1037     return 0;
1038 }
1039
1040 static void put_int8(QEMUFile *f, void *pv, size_t size)
1041 {
1042     int8_t *v = pv;
1043     qemu_put_s8s(f, v);
1044 }
1045
1046 const VMStateInfo vmstate_info_int8 = {
1047     .name = "int8",
1048     .get  = get_int8,
1049     .put  = put_int8,
1050 };
1051
1052 /* 16 bit int */
1053
1054 static int get_int16(QEMUFile *f, void *pv, size_t size)
1055 {
1056     int16_t *v = pv;
1057     qemu_get_sbe16s(f, v);
1058     return 0;
1059 }
1060
1061 static void put_int16(QEMUFile *f, void *pv, size_t size)
1062 {
1063     int16_t *v = pv;
1064     qemu_put_sbe16s(f, v);
1065 }
1066
1067 const VMStateInfo vmstate_info_int16 = {
1068     .name = "int16",
1069     .get  = get_int16,
1070     .put  = put_int16,
1071 };
1072
1073 /* 32 bit int */
1074
1075 static int get_int32(QEMUFile *f, void *pv, size_t size)
1076 {
1077     int32_t *v = pv;
1078     qemu_get_sbe32s(f, v);
1079     return 0;
1080 }
1081
1082 static void put_int32(QEMUFile *f, void *pv, size_t size)
1083 {
1084     int32_t *v = pv;
1085     qemu_put_sbe32s(f, v);
1086 }
1087
1088 const VMStateInfo vmstate_info_int32 = {
1089     .name = "int32",
1090     .get  = get_int32,
1091     .put  = put_int32,
1092 };
1093
1094 /* 32 bit int. See that the received value is the same than the one
1095    in the field */
1096
1097 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
1098 {
1099     int32_t *v = pv;
1100     int32_t v2;
1101     qemu_get_sbe32s(f, &v2);
1102
1103     if (*v == v2)
1104         return 0;
1105     return -EINVAL;
1106 }
1107
1108 const VMStateInfo vmstate_info_int32_equal = {
1109     .name = "int32 equal",
1110     .get  = get_int32_equal,
1111     .put  = put_int32,
1112 };
1113
1114 /* 32 bit int. See that the received value is the less or the same
1115    than the one in the field */
1116
1117 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
1118 {
1119     int32_t *old = pv;
1120     int32_t new;
1121     qemu_get_sbe32s(f, &new);
1122
1123     if (*old <= new)
1124         return 0;
1125     return -EINVAL;
1126 }
1127
1128 const VMStateInfo vmstate_info_int32_le = {
1129     .name = "int32 equal",
1130     .get  = get_int32_le,
1131     .put  = put_int32,
1132 };
1133
1134 /* 64 bit int */
1135
1136 static int get_int64(QEMUFile *f, void *pv, size_t size)
1137 {
1138     int64_t *v = pv;
1139     qemu_get_sbe64s(f, v);
1140     return 0;
1141 }
1142
1143 static void put_int64(QEMUFile *f, void *pv, size_t size)
1144 {
1145     int64_t *v = pv;
1146     qemu_put_sbe64s(f, v);
1147 }
1148
1149 const VMStateInfo vmstate_info_int64 = {
1150     .name = "int64",
1151     .get  = get_int64,
1152     .put  = put_int64,
1153 };
1154
1155 /* 8 bit unsigned int */
1156
1157 static int get_uint8(QEMUFile *f, void *pv, size_t size)
1158 {
1159     uint8_t *v = pv;
1160     qemu_get_8s(f, v);
1161     return 0;
1162 }
1163
1164 static void put_uint8(QEMUFile *f, void *pv, size_t size)
1165 {
1166     uint8_t *v = pv;
1167     qemu_put_8s(f, v);
1168 }
1169
1170 const VMStateInfo vmstate_info_uint8 = {
1171     .name = "uint8",
1172     .get  = get_uint8,
1173     .put  = put_uint8,
1174 };
1175
1176 /* 16 bit unsigned int */
1177
1178 static int get_uint16(QEMUFile *f, void *pv, size_t size)
1179 {
1180     uint16_t *v = pv;
1181     qemu_get_be16s(f, v);
1182     return 0;
1183 }
1184
1185 static void put_uint16(QEMUFile *f, void *pv, size_t size)
1186 {
1187     uint16_t *v = pv;
1188     qemu_put_be16s(f, v);
1189 }
1190
1191 const VMStateInfo vmstate_info_uint16 = {
1192     .name = "uint16",
1193     .get  = get_uint16,
1194     .put  = put_uint16,
1195 };
1196
1197 /* 32 bit unsigned int */
1198
1199 static int get_uint32(QEMUFile *f, void *pv, size_t size)
1200 {
1201     uint32_t *v = pv;
1202     qemu_get_be32s(f, v);
1203     return 0;
1204 }
1205
1206 static void put_uint32(QEMUFile *f, void *pv, size_t size)
1207 {
1208     uint32_t *v = pv;
1209     qemu_put_be32s(f, v);
1210 }
1211
1212 const VMStateInfo vmstate_info_uint32 = {
1213     .name = "uint32",
1214     .get  = get_uint32,
1215     .put  = put_uint32,
1216 };
1217
1218 /* 32 bit uint. See that the received value is the same than the one
1219    in the field */
1220
1221 static int get_uint32_equal(QEMUFile *f, void *pv, size_t size)
1222 {
1223     uint32_t *v = pv;
1224     uint32_t v2;
1225     qemu_get_be32s(f, &v2);
1226
1227     if (*v == v2) {
1228         return 0;
1229     }
1230     return -EINVAL;
1231 }
1232
1233 const VMStateInfo vmstate_info_uint32_equal = {
1234     .name = "uint32 equal",
1235     .get  = get_uint32_equal,
1236     .put  = put_uint32,
1237 };
1238
1239 /* 64 bit unsigned int */
1240
1241 static int get_uint64(QEMUFile *f, void *pv, size_t size)
1242 {
1243     uint64_t *v = pv;
1244     qemu_get_be64s(f, v);
1245     return 0;
1246 }
1247
1248 static void put_uint64(QEMUFile *f, void *pv, size_t size)
1249 {
1250     uint64_t *v = pv;
1251     qemu_put_be64s(f, v);
1252 }
1253
1254 const VMStateInfo vmstate_info_uint64 = {
1255     .name = "uint64",
1256     .get  = get_uint64,
1257     .put  = put_uint64,
1258 };
1259
1260 /* 64 bit unsigned int. See that the received value is the same than the one
1261    in the field */
1262
1263 static int get_uint64_equal(QEMUFile *f, void *pv, size_t size)
1264 {
1265     uint64_t *v = pv;
1266     uint64_t v2;
1267     qemu_get_be64s(f, &v2);
1268
1269     if (*v == v2) {
1270         return 0;
1271     }
1272     return -EINVAL;
1273 }
1274
1275 const VMStateInfo vmstate_info_uint64_equal = {
1276     .name = "int64 equal",
1277     .get  = get_uint64_equal,
1278     .put  = put_uint64,
1279 };
1280
1281 /* 8 bit int. See that the received value is the same than the one
1282    in the field */
1283
1284 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
1285 {
1286     uint8_t *v = pv;
1287     uint8_t v2;
1288     qemu_get_8s(f, &v2);
1289
1290     if (*v == v2)
1291         return 0;
1292     return -EINVAL;
1293 }
1294
1295 const VMStateInfo vmstate_info_uint8_equal = {
1296     .name = "uint8 equal",
1297     .get  = get_uint8_equal,
1298     .put  = put_uint8,
1299 };
1300
1301 /* 16 bit unsigned int int. See that the received value is the same than the one
1302    in the field */
1303
1304 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
1305 {
1306     uint16_t *v = pv;
1307     uint16_t v2;
1308     qemu_get_be16s(f, &v2);
1309
1310     if (*v == v2)
1311         return 0;
1312     return -EINVAL;
1313 }
1314
1315 const VMStateInfo vmstate_info_uint16_equal = {
1316     .name = "uint16 equal",
1317     .get  = get_uint16_equal,
1318     .put  = put_uint16,
1319 };
1320
1321 /* floating point */
1322
1323 static int get_float64(QEMUFile *f, void *pv, size_t size)
1324 {
1325     float64 *v = pv;
1326
1327     *v = make_float64(qemu_get_be64(f));
1328     return 0;
1329 }
1330
1331 static void put_float64(QEMUFile *f, void *pv, size_t size)
1332 {
1333     uint64_t *v = pv;
1334
1335     qemu_put_be64(f, float64_val(*v));
1336 }
1337
1338 const VMStateInfo vmstate_info_float64 = {
1339     .name = "float64",
1340     .get  = get_float64,
1341     .put  = put_float64,
1342 };
1343
1344 /* timers  */
1345
1346 static int get_timer(QEMUFile *f, void *pv, size_t size)
1347 {
1348     QEMUTimer *v = pv;
1349     timer_get(f, v);
1350     return 0;
1351 }
1352
1353 static void put_timer(QEMUFile *f, void *pv, size_t size)
1354 {
1355     QEMUTimer *v = pv;
1356     timer_put(f, v);
1357 }
1358
1359 const VMStateInfo vmstate_info_timer = {
1360     .name = "timer",
1361     .get  = get_timer,
1362     .put  = put_timer,
1363 };
1364
1365 /* uint8_t buffers */
1366
1367 static int get_buffer(QEMUFile *f, void *pv, size_t size)
1368 {
1369     uint8_t *v = pv;
1370     qemu_get_buffer(f, v, size);
1371     return 0;
1372 }
1373
1374 static void put_buffer(QEMUFile *f, void *pv, size_t size)
1375 {
1376     uint8_t *v = pv;
1377     qemu_put_buffer(f, v, size);
1378 }
1379
1380 const VMStateInfo vmstate_info_buffer = {
1381     .name = "buffer",
1382     .get  = get_buffer,
1383     .put  = put_buffer,
1384 };
1385
1386 /* unused buffers: space that was used for some fields that are
1387    not useful anymore */
1388
1389 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
1390 {
1391     uint8_t buf[1024];
1392     int block_len;
1393
1394     while (size > 0) {
1395         block_len = MIN(sizeof(buf), size);
1396         size -= block_len;
1397         qemu_get_buffer(f, buf, block_len);
1398     }
1399    return 0;
1400 }
1401
1402 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1403 {
1404     static const uint8_t buf[1024];
1405     int block_len;
1406
1407     while (size > 0) {
1408         block_len = MIN(sizeof(buf), size);
1409         size -= block_len;
1410         qemu_put_buffer(f, buf, block_len);
1411     }
1412 }
1413
1414 const VMStateInfo vmstate_info_unused_buffer = {
1415     .name = "unused_buffer",
1416     .get  = get_unused_buffer,
1417     .put  = put_unused_buffer,
1418 };
1419
1420 /* bitmaps (as defined by bitmap.h). Note that size here is the size
1421  * of the bitmap in bits. The on-the-wire format of a bitmap is 64
1422  * bit words with the bits in big endian order. The in-memory format
1423  * is an array of 'unsigned long', which may be either 32 or 64 bits.
1424  */
1425 /* This is the number of 64 bit words sent over the wire */
1426 #define BITS_TO_U64S(nr) DIV_ROUND_UP(nr, 64)
1427 static int get_bitmap(QEMUFile *f, void *pv, size_t size)
1428 {
1429     unsigned long *bmp = pv;
1430     int i, idx = 0;
1431     for (i = 0; i < BITS_TO_U64S(size); i++) {
1432         uint64_t w = qemu_get_be64(f);
1433         bmp[idx++] = w;
1434         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1435             bmp[idx++] = w >> 32;
1436         }
1437     }
1438     return 0;
1439 }
1440
1441 static void put_bitmap(QEMUFile *f, void *pv, size_t size)
1442 {
1443     unsigned long *bmp = pv;
1444     int i, idx = 0;
1445     for (i = 0; i < BITS_TO_U64S(size); i++) {
1446         uint64_t w = bmp[idx++];
1447         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1448             w |= ((uint64_t)bmp[idx++]) << 32;
1449         }
1450         qemu_put_be64(f, w);
1451     }
1452 }
1453
1454 const VMStateInfo vmstate_info_bitmap = {
1455     .name = "bitmap",
1456     .get = get_bitmap,
1457     .put = put_bitmap,
1458 };
1459
1460 typedef struct CompatEntry {
1461     char idstr[256];
1462     int instance_id;
1463 } CompatEntry;
1464
1465 typedef struct SaveStateEntry {
1466     QTAILQ_ENTRY(SaveStateEntry) entry;
1467     char idstr[256];
1468     int instance_id;
1469     int alias_id;
1470     int version_id;
1471     int section_id;
1472     SaveVMHandlers *ops;
1473     const VMStateDescription *vmsd;
1474     void *opaque;
1475     CompatEntry *compat;
1476     int no_migrate;
1477     int is_ram;
1478 } SaveStateEntry;
1479
1480
1481 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1482     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1483 static int global_section_id;
1484
1485 static int calculate_new_instance_id(const char *idstr)
1486 {
1487     SaveStateEntry *se;
1488     int instance_id = 0;
1489
1490     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1491         if (strcmp(idstr, se->idstr) == 0
1492             && instance_id <= se->instance_id) {
1493             instance_id = se->instance_id + 1;
1494         }
1495     }
1496     return instance_id;
1497 }
1498
1499 static int calculate_compat_instance_id(const char *idstr)
1500 {
1501     SaveStateEntry *se;
1502     int instance_id = 0;
1503
1504     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1505         if (!se->compat)
1506             continue;
1507
1508         if (strcmp(idstr, se->compat->idstr) == 0
1509             && instance_id <= se->compat->instance_id) {
1510             instance_id = se->compat->instance_id + 1;
1511         }
1512     }
1513     return instance_id;
1514 }
1515
1516 /* TODO: Individual devices generally have very little idea about the rest
1517    of the system, so instance_id should be removed/replaced.
1518    Meanwhile pass -1 as instance_id if you do not already have a clearly
1519    distinguishing id for all instances of your device class. */
1520 int register_savevm_live(DeviceState *dev,
1521                          const char *idstr,
1522                          int instance_id,
1523                          int version_id,
1524                          SaveVMHandlers *ops,
1525                          void *opaque)
1526 {
1527     SaveStateEntry *se;
1528
1529     se = g_malloc0(sizeof(SaveStateEntry));
1530     se->version_id = version_id;
1531     se->section_id = global_section_id++;
1532     se->ops = ops;
1533     se->opaque = opaque;
1534     se->vmsd = NULL;
1535     se->no_migrate = 0;
1536     /* if this is a live_savem then set is_ram */
1537     if (ops->save_live_setup != NULL) {
1538         se->is_ram = 1;
1539     }
1540
1541     if (dev) {
1542         char *id = qdev_get_dev_path(dev);
1543         if (id) {
1544             pstrcpy(se->idstr, sizeof(se->idstr), id);
1545             pstrcat(se->idstr, sizeof(se->idstr), "/");
1546             g_free(id);
1547
1548             se->compat = g_malloc0(sizeof(CompatEntry));
1549             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1550             se->compat->instance_id = instance_id == -1 ?
1551                          calculate_compat_instance_id(idstr) : instance_id;
1552             instance_id = -1;
1553         }
1554     }
1555     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1556
1557     if (instance_id == -1) {
1558         se->instance_id = calculate_new_instance_id(se->idstr);
1559     } else {
1560         se->instance_id = instance_id;
1561     }
1562     assert(!se->compat || se->instance_id == 0);
1563     /* add at the end of list */
1564     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1565     return 0;
1566 }
1567
1568 int register_savevm(DeviceState *dev,
1569                     const char *idstr,
1570                     int instance_id,
1571                     int version_id,
1572                     SaveStateHandler *save_state,
1573                     LoadStateHandler *load_state,
1574                     void *opaque)
1575 {
1576     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
1577     ops->save_state = save_state;
1578     ops->load_state = load_state;
1579     return register_savevm_live(dev, idstr, instance_id, version_id,
1580                                 ops, opaque);
1581 }
1582
1583 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1584 {
1585     SaveStateEntry *se, *new_se;
1586     char id[256] = "";
1587
1588     if (dev) {
1589         char *path = qdev_get_dev_path(dev);
1590         if (path) {
1591             pstrcpy(id, sizeof(id), path);
1592             pstrcat(id, sizeof(id), "/");
1593             g_free(path);
1594         }
1595     }
1596     pstrcat(id, sizeof(id), idstr);
1597
1598     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1599         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1600             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1601             if (se->compat) {
1602                 g_free(se->compat);
1603             }
1604             g_free(se->ops);
1605             g_free(se);
1606         }
1607     }
1608 }
1609
1610 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1611                                    const VMStateDescription *vmsd,
1612                                    void *opaque, int alias_id,
1613                                    int required_for_version)
1614 {
1615     SaveStateEntry *se;
1616
1617     /* If this triggers, alias support can be dropped for the vmsd. */
1618     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1619
1620     se = g_malloc0(sizeof(SaveStateEntry));
1621     se->version_id = vmsd->version_id;
1622     se->section_id = global_section_id++;
1623     se->opaque = opaque;
1624     se->vmsd = vmsd;
1625     se->alias_id = alias_id;
1626     se->no_migrate = vmsd->unmigratable;
1627
1628     if (dev) {
1629         char *id = qdev_get_dev_path(dev);
1630         if (id) {
1631             pstrcpy(se->idstr, sizeof(se->idstr), id);
1632             pstrcat(se->idstr, sizeof(se->idstr), "/");
1633             g_free(id);
1634
1635             se->compat = g_malloc0(sizeof(CompatEntry));
1636             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1637             se->compat->instance_id = instance_id == -1 ?
1638                          calculate_compat_instance_id(vmsd->name) : instance_id;
1639             instance_id = -1;
1640         }
1641     }
1642     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1643
1644     if (instance_id == -1) {
1645         se->instance_id = calculate_new_instance_id(se->idstr);
1646     } else {
1647         se->instance_id = instance_id;
1648     }
1649     assert(!se->compat || se->instance_id == 0);
1650     /* add at the end of list */
1651     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1652     return 0;
1653 }
1654
1655 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1656                         void *opaque)
1657 {
1658     SaveStateEntry *se, *new_se;
1659
1660     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1661         if (se->vmsd == vmsd && se->opaque == opaque) {
1662             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1663             if (se->compat) {
1664                 g_free(se->compat);
1665             }
1666             g_free(se);
1667         }
1668     }
1669 }
1670
1671 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1672                                     void *opaque);
1673 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1674                                    void *opaque);
1675
1676 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1677                        void *opaque, int version_id)
1678 {
1679     VMStateField *field = vmsd->fields;
1680     int ret;
1681
1682     if (version_id > vmsd->version_id) {
1683         return -EINVAL;
1684     }
1685     if (version_id < vmsd->minimum_version_id_old) {
1686         return -EINVAL;
1687     }
1688     if  (version_id < vmsd->minimum_version_id) {
1689         return vmsd->load_state_old(f, opaque, version_id);
1690     }
1691     if (vmsd->pre_load) {
1692         int ret = vmsd->pre_load(opaque);
1693         if (ret)
1694             return ret;
1695     }
1696     while(field->name) {
1697         if ((field->field_exists &&
1698              field->field_exists(opaque, version_id)) ||
1699             (!field->field_exists &&
1700              field->version_id <= version_id)) {
1701             void *base_addr = opaque + field->offset;
1702             int i, n_elems = 1;
1703             int size = field->size;
1704
1705             if (field->flags & VMS_VBUFFER) {
1706                 size = *(int32_t *)(opaque+field->size_offset);
1707                 if (field->flags & VMS_MULTIPLY) {
1708                     size *= field->size;
1709                 }
1710             }
1711             if (field->flags & VMS_ARRAY) {
1712                 n_elems = field->num;
1713             } else if (field->flags & VMS_VARRAY_INT32) {
1714                 n_elems = *(int32_t *)(opaque+field->num_offset);
1715             } else if (field->flags & VMS_VARRAY_UINT32) {
1716                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1717             } else if (field->flags & VMS_VARRAY_UINT16) {
1718                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1719             } else if (field->flags & VMS_VARRAY_UINT8) {
1720                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1721             }
1722             if (field->flags & VMS_POINTER) {
1723                 base_addr = *(void **)base_addr + field->start;
1724             }
1725             for (i = 0; i < n_elems; i++) {
1726                 void *addr = base_addr + size * i;
1727
1728                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1729                     addr = *(void **)addr;
1730                 }
1731                 if (field->flags & VMS_STRUCT) {
1732                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1733                 } else {
1734                     ret = field->info->get(f, addr, size);
1735
1736                 }
1737                 if (ret < 0) {
1738                     return ret;
1739                 }
1740             }
1741         }
1742         field++;
1743     }
1744     ret = vmstate_subsection_load(f, vmsd, opaque);
1745     if (ret != 0) {
1746         return ret;
1747     }
1748     if (vmsd->post_load) {
1749         return vmsd->post_load(opaque, version_id);
1750     }
1751     return 0;
1752 }
1753
1754 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1755                         void *opaque)
1756 {
1757     VMStateField *field = vmsd->fields;
1758
1759     if (vmsd->pre_save) {
1760         vmsd->pre_save(opaque);
1761     }
1762     while(field->name) {
1763         if (!field->field_exists ||
1764             field->field_exists(opaque, vmsd->version_id)) {
1765             void *base_addr = opaque + field->offset;
1766             int i, n_elems = 1;
1767             int size = field->size;
1768
1769             if (field->flags & VMS_VBUFFER) {
1770                 size = *(int32_t *)(opaque+field->size_offset);
1771                 if (field->flags & VMS_MULTIPLY) {
1772                     size *= field->size;
1773                 }
1774             }
1775             if (field->flags & VMS_ARRAY) {
1776                 n_elems = field->num;
1777             } else if (field->flags & VMS_VARRAY_INT32) {
1778                 n_elems = *(int32_t *)(opaque+field->num_offset);
1779             } else if (field->flags & VMS_VARRAY_UINT32) {
1780                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1781             } else if (field->flags & VMS_VARRAY_UINT16) {
1782                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1783             } else if (field->flags & VMS_VARRAY_UINT8) {
1784                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1785             }
1786             if (field->flags & VMS_POINTER) {
1787                 base_addr = *(void **)base_addr + field->start;
1788             }
1789             for (i = 0; i < n_elems; i++) {
1790                 void *addr = base_addr + size * i;
1791
1792                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1793                     addr = *(void **)addr;
1794                 }
1795                 if (field->flags & VMS_STRUCT) {
1796                     vmstate_save_state(f, field->vmsd, addr);
1797                 } else {
1798                     field->info->put(f, addr, size);
1799                 }
1800             }
1801         }
1802         field++;
1803     }
1804     vmstate_subsection_save(f, vmsd, opaque);
1805 }
1806
1807 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1808 {
1809     if (!se->vmsd) {         /* Old style */
1810         return se->ops->load_state(f, se->opaque, version_id);
1811     }
1812     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1813 }
1814
1815 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1816 {
1817     if (!se->vmsd) {         /* Old style */
1818         se->ops->save_state(f, se->opaque);
1819         return;
1820     }
1821     vmstate_save_state(f,se->vmsd, se->opaque);
1822 }
1823
1824 #define QEMU_VM_FILE_MAGIC           0x5145564d
1825 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1826 #define QEMU_VM_FILE_VERSION         0x00000003
1827
1828 #define QEMU_VM_EOF                  0x00
1829 #define QEMU_VM_SECTION_START        0x01
1830 #define QEMU_VM_SECTION_PART         0x02
1831 #define QEMU_VM_SECTION_END          0x03
1832 #define QEMU_VM_SECTION_FULL         0x04
1833 #define QEMU_VM_SUBSECTION           0x05
1834
1835 bool qemu_savevm_state_blocked(Error **errp)
1836 {
1837     SaveStateEntry *se;
1838
1839     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1840         if (se->no_migrate) {
1841             error_set(errp, QERR_MIGRATION_NOT_SUPPORTED, se->idstr);
1842             return true;
1843         }
1844     }
1845     return false;
1846 }
1847
1848 void qemu_savevm_state_begin(QEMUFile *f,
1849                              const MigrationParams *params)
1850 {
1851     SaveStateEntry *se;
1852     int ret;
1853
1854     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1855         if (!se->ops || !se->ops->set_params) {
1856             continue;
1857         }
1858         se->ops->set_params(params, se->opaque);
1859     }
1860
1861     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1862     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1863
1864     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1865         int len;
1866
1867         if (!se->ops || !se->ops->save_live_setup) {
1868             continue;
1869         }
1870         if (se->ops && se->ops->is_active) {
1871             if (!se->ops->is_active(se->opaque)) {
1872                 continue;
1873             }
1874         }
1875         /* Section type */
1876         qemu_put_byte(f, QEMU_VM_SECTION_START);
1877         qemu_put_be32(f, se->section_id);
1878
1879         /* ID string */
1880         len = strlen(se->idstr);
1881         qemu_put_byte(f, len);
1882         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1883
1884         qemu_put_be32(f, se->instance_id);
1885         qemu_put_be32(f, se->version_id);
1886
1887         ret = se->ops->save_live_setup(f, se->opaque);
1888         if (ret < 0) {
1889             qemu_file_set_error(f, ret);
1890             break;
1891         }
1892     }
1893 }
1894
1895 /*
1896  * this function has three return values:
1897  *   negative: there was one error, and we have -errno.
1898  *   0 : We haven't finished, caller have to go again
1899  *   1 : We have finished, we can go to complete phase
1900  */
1901 int qemu_savevm_state_iterate(QEMUFile *f)
1902 {
1903     SaveStateEntry *se;
1904     int ret = 1;
1905
1906     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1907         if (!se->ops || !se->ops->save_live_iterate) {
1908             continue;
1909         }
1910         if (se->ops && se->ops->is_active) {
1911             if (!se->ops->is_active(se->opaque)) {
1912                 continue;
1913             }
1914         }
1915         if (qemu_file_rate_limit(f)) {
1916             return 0;
1917         }
1918         trace_savevm_section_start();
1919         /* Section type */
1920         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1921         qemu_put_be32(f, se->section_id);
1922
1923         ret = se->ops->save_live_iterate(f, se->opaque);
1924         trace_savevm_section_end(se->section_id);
1925
1926         if (ret < 0) {
1927             qemu_file_set_error(f, ret);
1928         }
1929         if (ret <= 0) {
1930             /* Do not proceed to the next vmstate before this one reported
1931                completion of the current stage. This serializes the migration
1932                and reduces the probability that a faster changing state is
1933                synchronized over and over again. */
1934             break;
1935         }
1936     }
1937     return ret;
1938 }
1939
1940 void qemu_savevm_state_complete(QEMUFile *f)
1941 {
1942     SaveStateEntry *se;
1943     int ret;
1944
1945     cpu_synchronize_all_states();
1946
1947     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1948         if (!se->ops || !se->ops->save_live_complete) {
1949             continue;
1950         }
1951         if (se->ops && se->ops->is_active) {
1952             if (!se->ops->is_active(se->opaque)) {
1953                 continue;
1954             }
1955         }
1956         trace_savevm_section_start();
1957         /* Section type */
1958         qemu_put_byte(f, QEMU_VM_SECTION_END);
1959         qemu_put_be32(f, se->section_id);
1960
1961         ret = se->ops->save_live_complete(f, se->opaque);
1962         trace_savevm_section_end(se->section_id);
1963         if (ret < 0) {
1964             qemu_file_set_error(f, ret);
1965             return;
1966         }
1967     }
1968
1969     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1970         int len;
1971
1972         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1973             continue;
1974         }
1975         trace_savevm_section_start();
1976         /* Section type */
1977         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1978         qemu_put_be32(f, se->section_id);
1979
1980         /* ID string */
1981         len = strlen(se->idstr);
1982         qemu_put_byte(f, len);
1983         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1984
1985         qemu_put_be32(f, se->instance_id);
1986         qemu_put_be32(f, se->version_id);
1987
1988         vmstate_save(f, se);
1989         trace_savevm_section_end(se->section_id);
1990     }
1991
1992     qemu_put_byte(f, QEMU_VM_EOF);
1993     qemu_fflush(f);
1994 }
1995
1996 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
1997 {
1998     SaveStateEntry *se;
1999     uint64_t ret = 0;
2000
2001     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
2002         if (!se->ops || !se->ops->save_live_pending) {
2003             continue;
2004         }
2005         if (se->ops && se->ops->is_active) {
2006             if (!se->ops->is_active(se->opaque)) {
2007                 continue;
2008             }
2009         }
2010         ret += se->ops->save_live_pending(f, se->opaque, max_size);
2011     }
2012     return ret;
2013 }
2014
2015 void qemu_savevm_state_cancel(void)
2016 {
2017     SaveStateEntry *se;
2018
2019     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
2020         if (se->ops && se->ops->cancel) {
2021             se->ops->cancel(se->opaque);
2022         }
2023     }
2024 }
2025
2026 static int qemu_savevm_state(QEMUFile *f)
2027 {
2028     int ret;
2029     MigrationParams params = {
2030         .blk = 0,
2031         .shared = 0
2032     };
2033
2034     if (qemu_savevm_state_blocked(NULL)) {
2035         return -EINVAL;
2036     }
2037
2038     qemu_mutex_unlock_iothread();
2039     qemu_savevm_state_begin(f, &params);
2040     qemu_mutex_lock_iothread();
2041
2042     while (qemu_file_get_error(f) == 0) {
2043         if (qemu_savevm_state_iterate(f) > 0) {
2044             break;
2045         }
2046     }
2047
2048     ret = qemu_file_get_error(f);
2049     if (ret == 0) {
2050         qemu_savevm_state_complete(f);
2051         ret = qemu_file_get_error(f);
2052     }
2053     if (ret != 0) {
2054         qemu_savevm_state_cancel();
2055     }
2056     return ret;
2057 }
2058
2059 static int qemu_save_device_state(QEMUFile *f)
2060 {
2061     SaveStateEntry *se;
2062
2063     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
2064     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
2065
2066     cpu_synchronize_all_states();
2067
2068     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
2069         int len;
2070
2071         if (se->is_ram) {
2072             continue;
2073         }
2074         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
2075             continue;
2076         }
2077
2078         /* Section type */
2079         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
2080         qemu_put_be32(f, se->section_id);
2081
2082         /* ID string */
2083         len = strlen(se->idstr);
2084         qemu_put_byte(f, len);
2085         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
2086
2087         qemu_put_be32(f, se->instance_id);
2088         qemu_put_be32(f, se->version_id);
2089
2090         vmstate_save(f, se);
2091     }
2092
2093     qemu_put_byte(f, QEMU_VM_EOF);
2094
2095     return qemu_file_get_error(f);
2096 }
2097
2098 static SaveStateEntry *find_se(const char *idstr, int instance_id)
2099 {
2100     SaveStateEntry *se;
2101
2102     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
2103         if (!strcmp(se->idstr, idstr) &&
2104             (instance_id == se->instance_id ||
2105              instance_id == se->alias_id))
2106             return se;
2107         /* Migrating from an older version? */
2108         if (strstr(se->idstr, idstr) && se->compat) {
2109             if (!strcmp(se->compat->idstr, idstr) &&
2110                 (instance_id == se->compat->instance_id ||
2111                  instance_id == se->alias_id))
2112                 return se;
2113         }
2114     }
2115     return NULL;
2116 }
2117
2118 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
2119 {
2120     while(sub && sub->needed) {
2121         if (strcmp(idstr, sub->vmsd->name) == 0) {
2122             return sub->vmsd;
2123         }
2124         sub++;
2125     }
2126     return NULL;
2127 }
2128
2129 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
2130                                    void *opaque)
2131 {
2132     while (qemu_peek_byte(f, 0) == QEMU_VM_SUBSECTION) {
2133         char idstr[256];
2134         int ret;
2135         uint8_t version_id, len, size;
2136         const VMStateDescription *sub_vmsd;
2137
2138         len = qemu_peek_byte(f, 1);
2139         if (len < strlen(vmsd->name) + 1) {
2140             /* subsection name has be be "section_name/a" */
2141             return 0;
2142         }
2143         size = qemu_peek_buffer(f, (uint8_t *)idstr, len, 2);
2144         if (size != len) {
2145             return 0;
2146         }
2147         idstr[size] = 0;
2148
2149         if (strncmp(vmsd->name, idstr, strlen(vmsd->name)) != 0) {
2150             /* it don't have a valid subsection name */
2151             return 0;
2152         }
2153         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
2154         if (sub_vmsd == NULL) {
2155             return -ENOENT;
2156         }
2157         qemu_file_skip(f, 1); /* subsection */
2158         qemu_file_skip(f, 1); /* len */
2159         qemu_file_skip(f, len); /* idstr */
2160         version_id = qemu_get_be32(f);
2161
2162         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
2163         if (ret) {
2164             return ret;
2165         }
2166     }
2167     return 0;
2168 }
2169
2170 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
2171                                     void *opaque)
2172 {
2173     const VMStateSubsection *sub = vmsd->subsections;
2174
2175     while (sub && sub->needed) {
2176         if (sub->needed(opaque)) {
2177             const VMStateDescription *vmsd = sub->vmsd;
2178             uint8_t len;
2179
2180             qemu_put_byte(f, QEMU_VM_SUBSECTION);
2181             len = strlen(vmsd->name);
2182             qemu_put_byte(f, len);
2183             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
2184             qemu_put_be32(f, vmsd->version_id);
2185             vmstate_save_state(f, vmsd, opaque);
2186         }
2187         sub++;
2188     }
2189 }
2190
2191 typedef struct LoadStateEntry {
2192     QLIST_ENTRY(LoadStateEntry) entry;
2193     SaveStateEntry *se;
2194     int section_id;
2195     int version_id;
2196 } LoadStateEntry;
2197
2198 int qemu_loadvm_state(QEMUFile *f)
2199 {
2200     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
2201         QLIST_HEAD_INITIALIZER(loadvm_handlers);
2202     LoadStateEntry *le, *new_le;
2203     uint8_t section_type;
2204     unsigned int v;
2205     int ret;
2206
2207     if (qemu_savevm_state_blocked(NULL)) {
2208         return -EINVAL;
2209     }
2210
2211     v = qemu_get_be32(f);
2212     if (v != QEMU_VM_FILE_MAGIC)
2213         return -EINVAL;
2214
2215     v = qemu_get_be32(f);
2216     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
2217         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
2218         return -ENOTSUP;
2219     }
2220     if (v != QEMU_VM_FILE_VERSION)
2221         return -ENOTSUP;
2222
2223     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
2224         uint32_t instance_id, version_id, section_id;
2225         SaveStateEntry *se;
2226         char idstr[257];
2227         int len;
2228
2229         switch (section_type) {
2230         case QEMU_VM_SECTION_START:
2231         case QEMU_VM_SECTION_FULL:
2232             /* Read section start */
2233             section_id = qemu_get_be32(f);
2234             len = qemu_get_byte(f);
2235             qemu_get_buffer(f, (uint8_t *)idstr, len);
2236             idstr[len] = 0;
2237             instance_id = qemu_get_be32(f);
2238             version_id = qemu_get_be32(f);
2239
2240             /* Find savevm section */
2241             se = find_se(idstr, instance_id);
2242             if (se == NULL) {
2243                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
2244                 ret = -EINVAL;
2245                 goto out;
2246             }
2247
2248             /* Validate version */
2249             if (version_id > se->version_id) {
2250                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
2251                         version_id, idstr, se->version_id);
2252                 ret = -EINVAL;
2253                 goto out;
2254             }
2255
2256             /* Add entry */
2257             le = g_malloc0(sizeof(*le));
2258
2259             le->se = se;
2260             le->section_id = section_id;
2261             le->version_id = version_id;
2262             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
2263
2264             ret = vmstate_load(f, le->se, le->version_id);
2265             if (ret < 0) {
2266                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
2267                         instance_id, idstr);
2268                 goto out;
2269             }
2270             break;
2271         case QEMU_VM_SECTION_PART:
2272         case QEMU_VM_SECTION_END:
2273             section_id = qemu_get_be32(f);
2274
2275             QLIST_FOREACH(le, &loadvm_handlers, entry) {
2276                 if (le->section_id == section_id) {
2277                     break;
2278                 }
2279             }
2280             if (le == NULL) {
2281                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
2282                 ret = -EINVAL;
2283                 goto out;
2284             }
2285
2286             ret = vmstate_load(f, le->se, le->version_id);
2287             if (ret < 0) {
2288                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
2289                         section_id);
2290                 goto out;
2291             }
2292             break;
2293         default:
2294             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
2295             ret = -EINVAL;
2296             goto out;
2297         }
2298     }
2299
2300     cpu_synchronize_all_post_init();
2301
2302     ret = 0;
2303
2304 out:
2305     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
2306         QLIST_REMOVE(le, entry);
2307         g_free(le);
2308     }
2309
2310     if (ret == 0) {
2311         ret = qemu_file_get_error(f);
2312     }
2313
2314     return ret;
2315 }
2316
2317 static BlockDriverState *find_vmstate_bs(void)
2318 {
2319     BlockDriverState *bs = NULL;
2320     while ((bs = bdrv_next(bs))) {
2321         if (bdrv_can_snapshot(bs)) {
2322             return bs;
2323         }
2324     }
2325     return NULL;
2326 }
2327
2328 /*
2329  * Deletes snapshots of a given name in all opened images.
2330  */
2331 static int del_existing_snapshots(Monitor *mon, const char *name)
2332 {
2333     BlockDriverState *bs;
2334     QEMUSnapshotInfo sn1, *snapshot = &sn1;
2335     Error *err = NULL;
2336
2337     bs = NULL;
2338     while ((bs = bdrv_next(bs))) {
2339         if (bdrv_can_snapshot(bs) &&
2340             bdrv_snapshot_find(bs, snapshot, name) >= 0)
2341         {
2342             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
2343             if (error_is_set(&err)) {
2344                 monitor_printf(mon,
2345                                "Error while deleting snapshot on device '%s':"
2346                                " %s\n",
2347                                bdrv_get_device_name(bs),
2348                                error_get_pretty(err));
2349                 error_free(err);
2350                 return -1;
2351             }
2352         }
2353     }
2354
2355     return 0;
2356 }
2357
2358 void do_savevm(Monitor *mon, const QDict *qdict)
2359 {
2360     BlockDriverState *bs, *bs1;
2361     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
2362     int ret;
2363     QEMUFile *f;
2364     int saved_vm_running;
2365     uint64_t vm_state_size;
2366     qemu_timeval tv;
2367     struct tm tm;
2368     const char *name = qdict_get_try_str(qdict, "name");
2369
2370     /* Verify if there is a device that doesn't support snapshots and is writable */
2371     bs = NULL;
2372     while ((bs = bdrv_next(bs))) {
2373
2374         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2375             continue;
2376         }
2377
2378         if (!bdrv_can_snapshot(bs)) {
2379             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
2380                                bdrv_get_device_name(bs));
2381             return;
2382         }
2383     }
2384
2385     bs = find_vmstate_bs();
2386     if (!bs) {
2387         monitor_printf(mon, "No block device can accept snapshots\n");
2388         return;
2389     }
2390
2391     saved_vm_running = runstate_is_running();
2392     vm_stop(RUN_STATE_SAVE_VM);
2393
2394     memset(sn, 0, sizeof(*sn));
2395
2396     /* fill auxiliary fields */
2397     qemu_gettimeofday(&tv);
2398     sn->date_sec = tv.tv_sec;
2399     sn->date_nsec = tv.tv_usec * 1000;
2400     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
2401
2402     if (name) {
2403         ret = bdrv_snapshot_find(bs, old_sn, name);
2404         if (ret >= 0) {
2405             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
2406             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
2407         } else {
2408             pstrcpy(sn->name, sizeof(sn->name), name);
2409         }
2410     } else {
2411         /* cast below needed for OpenBSD where tv_sec is still 'long' */
2412         localtime_r((const time_t *)&tv.tv_sec, &tm);
2413         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
2414     }
2415
2416     /* Delete old snapshots of the same name */
2417     if (name && del_existing_snapshots(mon, name) < 0) {
2418         goto the_end;
2419     }
2420
2421     /* save the VM state */
2422     f = qemu_fopen_bdrv(bs, 1);
2423     if (!f) {
2424         monitor_printf(mon, "Could not open VM state file\n");
2425         goto the_end;
2426     }
2427     ret = qemu_savevm_state(f);
2428     vm_state_size = qemu_ftell(f);
2429     qemu_fclose(f);
2430     if (ret < 0) {
2431         monitor_printf(mon, "Error %d while writing VM\n", ret);
2432         goto the_end;
2433     }
2434
2435     /* create the snapshots */
2436
2437     bs1 = NULL;
2438     while ((bs1 = bdrv_next(bs1))) {
2439         if (bdrv_can_snapshot(bs1)) {
2440             /* Write VM state size only to the image that contains the state */
2441             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
2442             ret = bdrv_snapshot_create(bs1, sn);
2443             if (ret < 0) {
2444                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
2445                                bdrv_get_device_name(bs1));
2446             }
2447         }
2448     }
2449
2450  the_end:
2451     if (saved_vm_running)
2452         vm_start();
2453 }
2454
2455 void qmp_xen_save_devices_state(const char *filename, Error **errp)
2456 {
2457     QEMUFile *f;
2458     int saved_vm_running;
2459     int ret;
2460
2461     saved_vm_running = runstate_is_running();
2462     vm_stop(RUN_STATE_SAVE_VM);
2463
2464     f = qemu_fopen(filename, "wb");
2465     if (!f) {
2466         error_setg_file_open(errp, errno, filename);
2467         goto the_end;
2468     }
2469     ret = qemu_save_device_state(f);
2470     qemu_fclose(f);
2471     if (ret < 0) {
2472         error_set(errp, QERR_IO_ERROR);
2473     }
2474
2475  the_end:
2476     if (saved_vm_running)
2477         vm_start();
2478 }
2479
2480 int load_vmstate(const char *name)
2481 {
2482     BlockDriverState *bs, *bs_vm_state;
2483     QEMUSnapshotInfo sn;
2484     QEMUFile *f;
2485     int ret;
2486
2487     bs_vm_state = find_vmstate_bs();
2488     if (!bs_vm_state) {
2489         error_report("No block device supports snapshots");
2490         return -ENOTSUP;
2491     }
2492
2493     /* Don't even try to load empty VM states */
2494     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
2495     if (ret < 0) {
2496         return ret;
2497     } else if (sn.vm_state_size == 0) {
2498         error_report("This is a disk-only snapshot. Revert to it offline "
2499             "using qemu-img.");
2500         return -EINVAL;
2501     }
2502
2503     /* Verify if there is any device that doesn't support snapshots and is
2504     writable and check if the requested snapshot is available too. */
2505     bs = NULL;
2506     while ((bs = bdrv_next(bs))) {
2507
2508         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2509             continue;
2510         }
2511
2512         if (!bdrv_can_snapshot(bs)) {
2513             error_report("Device '%s' is writable but does not support snapshots.",
2514                                bdrv_get_device_name(bs));
2515             return -ENOTSUP;
2516         }
2517
2518         ret = bdrv_snapshot_find(bs, &sn, name);
2519         if (ret < 0) {
2520             error_report("Device '%s' does not have the requested snapshot '%s'",
2521                            bdrv_get_device_name(bs), name);
2522             return ret;
2523         }
2524     }
2525
2526     /* Flush all IO requests so they don't interfere with the new state.  */
2527     bdrv_drain_all();
2528
2529     bs = NULL;
2530     while ((bs = bdrv_next(bs))) {
2531         if (bdrv_can_snapshot(bs)) {
2532             ret = bdrv_snapshot_goto(bs, name);
2533             if (ret < 0) {
2534                 error_report("Error %d while activating snapshot '%s' on '%s'",
2535                              ret, name, bdrv_get_device_name(bs));
2536                 return ret;
2537             }
2538         }
2539     }
2540
2541     /* restore the VM state */
2542     f = qemu_fopen_bdrv(bs_vm_state, 0);
2543     if (!f) {
2544         error_report("Could not open VM state file");
2545         return -EINVAL;
2546     }
2547
2548     qemu_system_reset(VMRESET_SILENT);
2549     ret = qemu_loadvm_state(f);
2550
2551     qemu_fclose(f);
2552     if (ret < 0) {
2553         error_report("Error %d while loading VM state", ret);
2554         return ret;
2555     }
2556
2557     return 0;
2558 }
2559
2560 void do_delvm(Monitor *mon, const QDict *qdict)
2561 {
2562     BlockDriverState *bs, *bs1;
2563     Error *err = NULL;
2564     const char *name = qdict_get_str(qdict, "name");
2565
2566     bs = find_vmstate_bs();
2567     if (!bs) {
2568         monitor_printf(mon, "No block device supports snapshots\n");
2569         return;
2570     }
2571
2572     bs1 = NULL;
2573     while ((bs1 = bdrv_next(bs1))) {
2574         if (bdrv_can_snapshot(bs1)) {
2575             bdrv_snapshot_delete_by_id_or_name(bs, name, &err);
2576             if (error_is_set(&err)) {
2577                 monitor_printf(mon,
2578                                "Error while deleting snapshot on device '%s':"
2579                                " %s\n",
2580                                bdrv_get_device_name(bs),
2581                                error_get_pretty(err));
2582                 error_free(err);
2583             }
2584         }
2585     }
2586 }
2587
2588 void do_info_snapshots(Monitor *mon, const QDict *qdict)
2589 {
2590     BlockDriverState *bs, *bs1;
2591     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2592     int nb_sns, i, ret, available;
2593     int total;
2594     int *available_snapshots;
2595
2596     bs = find_vmstate_bs();
2597     if (!bs) {
2598         monitor_printf(mon, "No available block device supports snapshots\n");
2599         return;
2600     }
2601
2602     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2603     if (nb_sns < 0) {
2604         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2605         return;
2606     }
2607
2608     if (nb_sns == 0) {
2609         monitor_printf(mon, "There is no snapshot available.\n");
2610         return;
2611     }
2612
2613     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
2614     total = 0;
2615     for (i = 0; i < nb_sns; i++) {
2616         sn = &sn_tab[i];
2617         available = 1;
2618         bs1 = NULL;
2619
2620         while ((bs1 = bdrv_next(bs1))) {
2621             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2622                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2623                 if (ret < 0) {
2624                     available = 0;
2625                     break;
2626                 }
2627             }
2628         }
2629
2630         if (available) {
2631             available_snapshots[total] = i;
2632             total++;
2633         }
2634     }
2635
2636     if (total > 0) {
2637         bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, NULL);
2638         monitor_printf(mon, "\n");
2639         for (i = 0; i < total; i++) {
2640             sn = &sn_tab[available_snapshots[i]];
2641             bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, sn);
2642             monitor_printf(mon, "\n");
2643         }
2644     } else {
2645         monitor_printf(mon, "There is no suitable snapshot available\n");
2646     }
2647
2648     g_free(sn_tab);
2649     g_free(available_snapshots);
2650
2651 }
2652
2653 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
2654 {
2655     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
2656                        memory_region_name(mr), dev);
2657 }
2658
2659 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
2660 {
2661     /* Nothing do to while the implementation is in RAMBlock */
2662 }
2663
2664 void vmstate_register_ram_global(MemoryRegion *mr)
2665 {
2666     vmstate_register_ram(mr, NULL);
2667 }