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 #include <unistd.h>
  25 #include <fcntl.h>
  26 #include <time.h>
  27 #include <errno.h>
  28 #include <sys/time.h>
  29 #include <zlib.h>
  30
  31 /* Needed early for CONFIG_BSD etc. */
  32 #include "config-host.h"
  33
  34 #ifndef _WIN32
  35 #include <sys/times.h>
  36 #include <sys/wait.h>
  37 #include <termios.h>
  38 #include <sys/mman.h>
  39 #include <sys/ioctl.h>
  40 #include <sys/resource.h>
  41 #include <sys/socket.h>
  42 #include <netinet/in.h>
  43 #include <net/if.h>
  44 #include <arpa/inet.h>
  45 #include <dirent.h>
  46 #include <netdb.h>
  47 #include <sys/select.h>
  48 #ifdef CONFIG_BSD
  49 #include <sys/stat.h>
  50 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) || defined(__DragonFly__)
  51 #include <libutil.h>
  52 #else
  53 #include <util.h>
  54 #endif
  55 #ifdef __linux__
  56 #include <pty.h>
  57 #include <malloc.h>
  58 #include <linux/rtc.h>
  59 #endif
  60 #endif
  61 #endif
  62
  63 #ifdef _WIN32
  64 #include <windows.h>
  65 #include <malloc.h>
  66 #include <sys/timeb.h>
  67 #include <mmsystem.h>
  68 #define getopt_long_only getopt_long
  69 #define memalign(align, size) malloc(size)
  70 #endif
  71
  72 #include "qemu-common.h"
  73 #include "hw/hw.h"
  74 #include "hw/qdev.h"
  75 #include "net.h"
  76 #include "monitor.h"
  77 #include "sysemu.h"
  78 #include "qemu-timer.h"
  79 #include "qemu-char.h"
  80 #include "audio/audio.h"
  81 #include "migration.h"
  82 #include "qemu_socket.h"
  83 #include "qemu-queue.h"
  84 #include "qemu-timer.h"
  85 #include "cpus.h"
  86 #include "memory.h"
  87 #include "qmp-commands.h"
  88 #include "trace.h"
  89 #include "bitops.h"
  90
  91 #define SELF_ANNOUNCE_ROUNDS 5
  92
  93 #ifndef ETH_P_RARP
  94 #define ETH_P_RARP 0x8035
  95 #endif
  96 #define ARP_HTYPE_ETH 0x0001
  97 #define ARP_PTYPE_IP 0x0800
  98 #define ARP_OP_REQUEST_REV 0x3
  99
 100 static int announce_self_create(uint8_t *buf,
 101                                 uint8_t *mac_addr)
 102 {
 103     /* Ethernet header. */
 104     memset(buf, 0xff, 6);         /* destination MAC addr */
 105     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
 106     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
 107
 108     /* RARP header. */
 109     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
 110     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
 111     *(buf + 18) = 6; /* hardware addr length (ethernet) */
 112     *(buf + 19) = 4; /* protocol addr length (IPv4) */
 113     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
 114     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
 115     memset(buf + 28, 0x00, 4);     /* source protocol addr */
 116     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
 117     memset(buf + 38, 0x00, 4);     /* target protocol addr */
 118
 119     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
 120     memset(buf + 42, 0x00, 18);
 121
 122     return 60; /* len (FCS will be added by hardware) */
 123 }
 124
 125 static void qemu_announce_self_iter(NICState *nic, void *opaque)
 126 {
 127     uint8_t buf[60];
 128     int len;
 129
 130     len = announce_self_create(buf, nic->conf->macaddr.a);
 131
 132     qemu_send_packet_raw(&nic->nc, buf, len);
 133 }
 134
 135
 136 static void qemu_announce_self_once(void *opaque)
 137 {
 138     static int count = SELF_ANNOUNCE_ROUNDS;
 139     QEMUTimer *timer = *(QEMUTimer **)opaque;
 140
 141     qemu_foreach_nic(qemu_announce_self_iter, NULL);
 142
 143     if (--count) {
 144         /* delay 50ms, 150ms, 250ms, ... */
 145         qemu_mod_timer(timer, qemu_get_clock_ms(rt_clock) +
 146                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
 147     } else {
 148             qemu_del_timer(timer);
 149             qemu_free_timer(timer);
 150     }
 151 }
 152
 153 void qemu_announce_self(void)
 154 {
 155         static QEMUTimer *timer;
 156         timer = qemu_new_timer_ms(rt_clock, qemu_announce_self_once, &timer);
 157         qemu_announce_self_once(&timer);
 158 }
 159
 160 /***********************************************************/
 161 /* savevm/loadvm support */
 162
 163 #define IO_BUF_SIZE 32768
 164
 165 struct QEMUFile {
 166     const QEMUFileOps *ops;
 167     void *opaque;
 168     int is_write;
 169
 170     int64_t buf_offset; /* start of buffer when writing, end of buffer
 171                            when reading */
 172     int buf_index;
 173     int buf_size; /* 0 when writing */
 174     uint8_t buf[IO_BUF_SIZE];
 175
 176     int last_error;
 177 };
 178
 179 typedef struct QEMUFileStdio
 180 {
 181     FILE *stdio_file;
 182     QEMUFile *file;
 183 } QEMUFileStdio;
 184
 185 typedef struct QEMUFileSocket
 186 {
 187     int fd;
 188     QEMUFile *file;
 189 } QEMUFileSocket;
 190
 191 static int socket_get_fd(void *opaque)
 192 {
 193     QEMUFileSocket *s = opaque;
 194
 195     return s->fd;
 196 }
 197
 198 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 199 {
 200     QEMUFileSocket *s = opaque;
 201     ssize_t len;
 202
 203     do {
 204         len = qemu_recv(s->fd, buf, size, 0);
 205     } while (len == -1 && socket_error() == EINTR);
 206
 207     if (len == -1)
 208         len = -socket_error();
 209
 210     return len;
 211 }
 212
 213 static int socket_close(void *opaque)
 214 {
 215     QEMUFileSocket *s = opaque;
 216     g_free(s);
 217     return 0;
 218 }
 219
 220 static int stdio_get_fd(void *opaque)
 221 {
 222     QEMUFileStdio *s = opaque;
 223
 224     return fileno(s->stdio_file);
 225 }
 226
 227 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 228 {
 229     QEMUFileStdio *s = opaque;
 230     return fwrite(buf, 1, size, s->stdio_file);
 231 }
 232
 233 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 234 {
 235     QEMUFileStdio *s = opaque;
 236     FILE *fp = s->stdio_file;
 237     int bytes;
 238
 239     do {
 240         clearerr(fp);
 241         bytes = fread(buf, 1, size, fp);
 242     } while ((bytes == 0) && ferror(fp) && (errno == EINTR));
 243     return bytes;
 244 }
 245
 246 static int stdio_pclose(void *opaque)
 247 {
 248     QEMUFileStdio *s = opaque;
 249     int ret;
 250     ret = pclose(s->stdio_file);
 251     if (ret == -1) {
 252         ret = -errno;
 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     if (fclose(s->stdio_file) == EOF) {
 263         ret = -errno;
 264     }
 265     g_free(s);
 266     return ret;
 267 }
 268
 269 static const QEMUFileOps stdio_pipe_read_ops = {
 270     .get_fd =     stdio_get_fd,
 271     .get_buffer = stdio_get_buffer,
 272     .close =      stdio_pclose
 273 };
 274
 275 static const QEMUFileOps stdio_pipe_write_ops = {
 276     .get_fd =     stdio_get_fd,
 277     .put_buffer = stdio_put_buffer,
 278     .close =      stdio_pclose
 279 };
 280
 281 QEMUFile *qemu_popen(FILE *stdio_file, const char *mode)
 282 {
 283     QEMUFileStdio *s;
 284
 285     if (stdio_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 286         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 287         return NULL;
 288     }
 289
 290     s = g_malloc0(sizeof(QEMUFileStdio));
 291
 292     s->stdio_file = stdio_file;
 293
 294     if(mode[0] == 'r') {
 295         s->file = qemu_fopen_ops(s, &stdio_pipe_read_ops);
 296     } else {
 297         s->file = qemu_fopen_ops(s, &stdio_pipe_write_ops);
 298     }
 299     return s->file;
 300 }
 301
 302 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 303 {
 304     FILE *popen_file;
 305
 306     popen_file = popen(command, mode);
 307     if(popen_file == NULL) {
 308         return NULL;
 309     }
 310
 311     return qemu_popen(popen_file, mode);
 312 }
 313
 314 static const QEMUFileOps stdio_file_read_ops = {
 315     .get_fd =     stdio_get_fd,
 316     .get_buffer = stdio_get_buffer,
 317     .close =      stdio_fclose
 318 };
 319
 320 static const QEMUFileOps stdio_file_write_ops = {
 321     .get_fd =     stdio_get_fd,
 322     .put_buffer = stdio_put_buffer,
 323     .close =      stdio_fclose
 324 };
 325
 326 QEMUFile *qemu_fdopen(int fd, const char *mode)
 327 {
 328     QEMUFileStdio *s;
 329
 330     if (mode == NULL ||
 331         (mode[0] != 'r' && mode[0] != 'w') ||
 332         mode[1] != 'b' || mode[2] != 0) {
 333         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 334         return NULL;
 335     }
 336
 337     s = g_malloc0(sizeof(QEMUFileStdio));
 338     s->stdio_file = fdopen(fd, mode);
 339     if (!s->stdio_file)
 340         goto fail;
 341
 342     if(mode[0] == 'r') {
 343         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 344     } else {
 345         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 346     }
 347     return s->file;
 348
 349 fail:
 350     g_free(s);
 351     return NULL;
 352 }
 353
 354 static const QEMUFileOps socket_read_ops = {
 355     .get_fd =     socket_get_fd,
 356     .get_buffer = socket_get_buffer,
 357     .close =      socket_close
 358 };
 359
 360 QEMUFile *qemu_fopen_socket(int fd)
 361 {
 362     QEMUFileSocket *s = g_malloc0(sizeof(QEMUFileSocket));
 363
 364     s->fd = fd;
 365     s->file = qemu_fopen_ops(s, &socket_read_ops);
 366     return s->file;
 367 }
 368
 369 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 370 {
 371     QEMUFileStdio *s;
 372
 373     if (mode == NULL ||
 374         (mode[0] != 'r' && mode[0] != 'w') ||
 375         mode[1] != 'b' || mode[2] != 0) {
 376         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 377         return NULL;
 378     }
 379
 380     s = g_malloc0(sizeof(QEMUFileStdio));
 381
 382     s->stdio_file = fopen(filename, mode);
 383     if (!s->stdio_file)
 384         goto fail;
 385
 386     if(mode[0] == 'w') {
 387         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 388     } else {
 389         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 390     }
 391     return s->file;
 392 fail:
 393     g_free(s);
 394     return NULL;
 395 }
 396
 397 static int block_put_buffer(void *opaque, const uint8_t *buf,
 398                            int64_t pos, int size)
 399 {
 400     bdrv_save_vmstate(opaque, buf, pos, size);
 401     return size;
 402 }
 403
 404 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 405 {
 406     return bdrv_load_vmstate(opaque, buf, pos, size);
 407 }
 408
 409 static int bdrv_fclose(void *opaque)
 410 {
 411     return bdrv_flush(opaque);
 412 }
 413
 414 static const QEMUFileOps bdrv_read_ops = {
 415     .get_buffer = block_get_buffer,
 416     .close =      bdrv_fclose
 417 };
 418
 419 static const QEMUFileOps bdrv_write_ops = {
 420     .put_buffer = block_put_buffer,
 421     .close =      bdrv_fclose
 422 };
 423
 424 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 425 {
 426     if (is_writable)
 427         return qemu_fopen_ops(bs, &bdrv_write_ops);
 428     return qemu_fopen_ops(bs, &bdrv_read_ops);
 429 }
 430
 431 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
 432 {
 433     QEMUFile *f;
 434
 435     f = g_malloc0(sizeof(QEMUFile));
 436
 437     f->opaque = opaque;
 438     f->ops = ops;
 439     f->is_write = 0;
 440
 441     return f;
 442 }
 443
 444 int qemu_file_get_error(QEMUFile *f)
 445 {
 446     return f->last_error;
 447 }
 448
 449 static void qemu_file_set_error(QEMUFile *f, int ret)
 450 {
 451     f->last_error = ret;
 452 }
 453
 454 /** Flushes QEMUFile buffer
 455  *
 456  */
 457 static int qemu_fflush(QEMUFile *f)
 458 {
 459     int ret = 0;
 460
 461     if (!f->ops->put_buffer)
 462         return 0;
 463
 464     if (f->is_write && f->buf_index > 0) {
 465         ret = f->ops->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 466         if (ret >= 0) {
 467             f->buf_offset += f->buf_index;
 468         }
 469         f->buf_index = 0;
 470     }
 471     return ret;
 472 }
 473
 474 static void qemu_fill_buffer(QEMUFile *f)
 475 {
 476     int len;
 477     int pending;
 478
 479     if (!f->ops->get_buffer)
 480         return;
 481
 482     if (f->is_write)
 483         abort();
 484
 485     pending = f->buf_size - f->buf_index;
 486     if (pending > 0) {
 487         memmove(f->buf, f->buf + f->buf_index, pending);
 488     }
 489     f->buf_index = 0;
 490     f->buf_size = pending;
 491
 492     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->buf_offset,
 493                         IO_BUF_SIZE - pending);
 494     if (len > 0) {
 495         f->buf_size += len;
 496         f->buf_offset += len;
 497     } else if (len == 0) {
 498         qemu_file_set_error(f, -EIO);
 499     } else if (len != -EAGAIN)
 500         qemu_file_set_error(f, len);
 501 }
 502
 503 int qemu_get_fd(QEMUFile *f)
 504 {
 505     if (f->ops->get_fd) {
 506         return f->ops->get_fd(f->opaque);
 507     }
 508     return -1;
 509 }
 510
 511 /** Closes the file
 512  *
 513  * Returns negative error value if any error happened on previous operations or
 514  * while closing the file. Returns 0 or positive number on success.
 515  *
 516  * The meaning of return value on success depends on the specific backend
 517  * being used.
 518  */
 519 int qemu_fclose(QEMUFile *f)
 520 {
 521     int ret;
 522     ret = qemu_fflush(f);
 523
 524     if (f->ops->close) {
 525         int ret2 = f->ops->close(f->opaque);
 526         if (ret >= 0) {
 527             ret = ret2;
 528         }
 529     }
 530     /* If any error was spotted before closing, we should report it
 531      * instead of the close() return value.
 532      */
 533     if (f->last_error) {
 534         ret = f->last_error;
 535     }
 536     g_free(f);
 537     return ret;
 538 }
 539
 540 int qemu_file_put_notify(QEMUFile *f)
 541 {
 542     return f->ops->put_buffer(f->opaque, NULL, 0, 0);
 543 }
 544
 545 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 546 {
 547     int l;
 548
 549     if (f->last_error) {
 550         return;
 551     }
 552
 553     if (f->is_write == 0 && f->buf_index > 0) {
 554         fprintf(stderr,
 555                 "Attempted to write to buffer while read buffer is not empty\n");
 556         abort();
 557     }
 558
 559     while (size > 0) {
 560         l = IO_BUF_SIZE - f->buf_index;
 561         if (l > size)
 562             l = size;
 563         memcpy(f->buf + f->buf_index, buf, l);
 564         f->is_write = 1;
 565         f->buf_index += l;
 566         buf += l;
 567         size -= l;
 568         if (f->buf_index >= IO_BUF_SIZE) {
 569             int ret = qemu_fflush(f);
 570             if (ret < 0) {
 571                 qemu_file_set_error(f, ret);
 572                 break;
 573             }
 574         }
 575     }
 576 }
 577
 578 void qemu_put_byte(QEMUFile *f, int v)
 579 {
 580     if (f->last_error) {
 581         return;
 582     }
 583
 584     if (f->is_write == 0 && f->buf_index > 0) {
 585         fprintf(stderr,
 586                 "Attempted to write to buffer while read buffer is not empty\n");
 587         abort();
 588     }
 589
 590     f->buf[f->buf_index++] = v;
 591     f->is_write = 1;
 592     if (f->buf_index >= IO_BUF_SIZE) {
 593         int ret = qemu_fflush(f);
 594         if (ret < 0) {
 595             qemu_file_set_error(f, ret);
 596         }
 597     }
 598 }
 599
 600 static void qemu_file_skip(QEMUFile *f, int size)
 601 {
 602     if (f->buf_index + size <= f->buf_size) {
 603         f->buf_index += size;
 604     }
 605 }
 606
 607 static int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset)
 608 {
 609     int pending;
 610     int index;
 611
 612     if (f->is_write) {
 613         abort();
 614     }
 615
 616     index = f->buf_index + offset;
 617     pending = f->buf_size - index;
 618     if (pending < size) {
 619         qemu_fill_buffer(f);
 620         index = f->buf_index + offset;
 621         pending = f->buf_size - index;
 622     }
 623
 624     if (pending <= 0) {
 625         return 0;
 626     }
 627     if (size > pending) {
 628         size = pending;
 629     }
 630
 631     memcpy(buf, f->buf + index, size);
 632     return size;
 633 }
 634
 635 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size)
 636 {
 637     int pending = size;
 638     int done = 0;
 639
 640     while (pending > 0) {
 641         int res;
 642
 643         res = qemu_peek_buffer(f, buf, pending, 0);
 644         if (res == 0) {
 645             return done;
 646         }
 647         qemu_file_skip(f, res);
 648         buf += res;
 649         pending -= res;
 650         done += res;
 651     }
 652     return done;
 653 }
 654
 655 static int qemu_peek_byte(QEMUFile *f, int offset)
 656 {
 657     int index = f->buf_index + offset;
 658
 659     if (f->is_write) {
 660         abort();
 661     }
 662
 663     if (index >= f->buf_size) {
 664         qemu_fill_buffer(f);
 665         index = f->buf_index + offset;
 666         if (index >= f->buf_size) {
 667             return 0;
 668         }
 669     }
 670     return f->buf[index];
 671 }
 672
 673 int qemu_get_byte(QEMUFile *f)
 674 {
 675     int result;
 676
 677     result = qemu_peek_byte(f, 0);
 678     qemu_file_skip(f, 1);
 679     return result;
 680 }
 681
 682 static int64_t qemu_ftell(QEMUFile *f)
 683 {
 684     return f->buf_offset - f->buf_size + f->buf_index;
 685 }
 686
 687 int qemu_file_rate_limit(QEMUFile *f)
 688 {
 689     if (f->ops->rate_limit)
 690         return f->ops->rate_limit(f->opaque);
 691
 692     return 0;
 693 }
 694
 695 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 696 {
 697     if (f->ops->get_rate_limit)
 698         return f->ops->get_rate_limit(f->opaque);
 699
 700     return 0;
 701 }
 702
 703 int64_t qemu_file_set_rate_limit(QEMUFile *f, int64_t new_rate)
 704 {
 705     /* any failed or completed migration keeps its state to allow probing of
 706      * migration data, but has no associated file anymore */
 707     if (f && f->ops->set_rate_limit)
 708         return f->ops->set_rate_limit(f->opaque, new_rate);
 709
 710     return 0;
 711 }
 712
 713 void qemu_put_be16(QEMUFile *f, unsigned int v)
 714 {
 715     qemu_put_byte(f, v >> 8);
 716     qemu_put_byte(f, v);
 717 }
 718
 719 void qemu_put_be32(QEMUFile *f, unsigned int v)
 720 {
 721     qemu_put_byte(f, v >> 24);
 722     qemu_put_byte(f, v >> 16);
 723     qemu_put_byte(f, v >> 8);
 724     qemu_put_byte(f, v);
 725 }
 726
 727 void qemu_put_be64(QEMUFile *f, uint64_t v)
 728 {
 729     qemu_put_be32(f, v >> 32);
 730     qemu_put_be32(f, v);
 731 }
 732
 733 unsigned int qemu_get_be16(QEMUFile *f)
 734 {
 735     unsigned int v;
 736     v = qemu_get_byte(f) << 8;
 737     v |= qemu_get_byte(f);
 738     return v;
 739 }
 740
 741 unsigned int qemu_get_be32(QEMUFile *f)
 742 {
 743     unsigned int v;
 744     v = qemu_get_byte(f) << 24;
 745     v |= qemu_get_byte(f) << 16;
 746     v |= qemu_get_byte(f) << 8;
 747     v |= qemu_get_byte(f);
 748     return v;
 749 }
 750
 751 uint64_t qemu_get_be64(QEMUFile *f)
 752 {
 753     uint64_t v;
 754     v = (uint64_t)qemu_get_be32(f) << 32;
 755     v |= qemu_get_be32(f);
 756     return v;
 757 }
 758
 759
 760 /* timer */
 761
 762 void qemu_put_timer(QEMUFile *f, QEMUTimer *ts)
 763 {
 764     uint64_t expire_time;
 765
 766     expire_time = qemu_timer_expire_time_ns(ts);
 767     qemu_put_be64(f, expire_time);
 768 }
 769
 770 void qemu_get_timer(QEMUFile *f, QEMUTimer *ts)
 771 {
 772     uint64_t expire_time;
 773
 774     expire_time = qemu_get_be64(f);
 775     if (expire_time != -1) {
 776         qemu_mod_timer_ns(ts, expire_time);
 777     } else {
 778         qemu_del_timer(ts);
 779     }
 780 }
 781
 782
 783 /* bool */
 784
 785 static int get_bool(QEMUFile *f, void *pv, size_t size)
 786 {
 787     bool *v = pv;
 788     *v = qemu_get_byte(f);
 789     return 0;
 790 }
 791
 792 static void put_bool(QEMUFile *f, void *pv, size_t size)
 793 {
 794     bool *v = pv;
 795     qemu_put_byte(f, *v);
 796 }
 797
 798 const VMStateInfo vmstate_info_bool = {
 799     .name = "bool",
 800     .get  = get_bool,
 801     .put  = put_bool,
 802 };
 803
 804 /* 8 bit int */
 805
 806 static int get_int8(QEMUFile *f, void *pv, size_t size)
 807 {
 808     int8_t *v = pv;
 809     qemu_get_s8s(f, v);
 810     return 0;
 811 }
 812
 813 static void put_int8(QEMUFile *f, void *pv, size_t size)
 814 {
 815     int8_t *v = pv;
 816     qemu_put_s8s(f, v);
 817 }
 818
 819 const VMStateInfo vmstate_info_int8 = {
 820     .name = "int8",
 821     .get  = get_int8,
 822     .put  = put_int8,
 823 };
 824
 825 /* 16 bit int */
 826
 827 static int get_int16(QEMUFile *f, void *pv, size_t size)
 828 {
 829     int16_t *v = pv;
 830     qemu_get_sbe16s(f, v);
 831     return 0;
 832 }
 833
 834 static void put_int16(QEMUFile *f, void *pv, size_t size)
 835 {
 836     int16_t *v = pv;
 837     qemu_put_sbe16s(f, v);
 838 }
 839
 840 const VMStateInfo vmstate_info_int16 = {
 841     .name = "int16",
 842     .get  = get_int16,
 843     .put  = put_int16,
 844 };
 845
 846 /* 32 bit int */
 847
 848 static int get_int32(QEMUFile *f, void *pv, size_t size)
 849 {
 850     int32_t *v = pv;
 851     qemu_get_sbe32s(f, v);
 852     return 0;
 853 }
 854
 855 static void put_int32(QEMUFile *f, void *pv, size_t size)
 856 {
 857     int32_t *v = pv;
 858     qemu_put_sbe32s(f, v);
 859 }
 860
 861 const VMStateInfo vmstate_info_int32 = {
 862     .name = "int32",
 863     .get  = get_int32,
 864     .put  = put_int32,
 865 };
 866
 867 /* 32 bit int. See that the received value is the same than the one
 868    in the field */
 869
 870 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 871 {
 872     int32_t *v = pv;
 873     int32_t v2;
 874     qemu_get_sbe32s(f, &v2);
 875
 876     if (*v == v2)
 877         return 0;
 878     return -EINVAL;
 879 }
 880
 881 const VMStateInfo vmstate_info_int32_equal = {
 882     .name = "int32 equal",
 883     .get  = get_int32_equal,
 884     .put  = put_int32,
 885 };
 886
 887 /* 32 bit int. See that the received value is the less or the same
 888    than the one in the field */
 889
 890 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 891 {
 892     int32_t *old = pv;
 893     int32_t new;
 894     qemu_get_sbe32s(f, &new);
 895
 896     if (*old <= new)
 897         return 0;
 898     return -EINVAL;
 899 }
 900
 901 const VMStateInfo vmstate_info_int32_le = {
 902     .name = "int32 equal",
 903     .get  = get_int32_le,
 904     .put  = put_int32,
 905 };
 906
 907 /* 64 bit int */
 908
 909 static int get_int64(QEMUFile *f, void *pv, size_t size)
 910 {
 911     int64_t *v = pv;
 912     qemu_get_sbe64s(f, v);
 913     return 0;
 914 }
 915
 916 static void put_int64(QEMUFile *f, void *pv, size_t size)
 917 {
 918     int64_t *v = pv;
 919     qemu_put_sbe64s(f, v);
 920 }
 921
 922 const VMStateInfo vmstate_info_int64 = {
 923     .name = "int64",
 924     .get  = get_int64,
 925     .put  = put_int64,
 926 };
 927
 928 /* 8 bit unsigned int */
 929
 930 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 931 {
 932     uint8_t *v = pv;
 933     qemu_get_8s(f, v);
 934     return 0;
 935 }
 936
 937 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 938 {
 939     uint8_t *v = pv;
 940     qemu_put_8s(f, v);
 941 }
 942
 943 const VMStateInfo vmstate_info_uint8 = {
 944     .name = "uint8",
 945     .get  = get_uint8,
 946     .put  = put_uint8,
 947 };
 948
 949 /* 16 bit unsigned int */
 950
 951 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 952 {
 953     uint16_t *v = pv;
 954     qemu_get_be16s(f, v);
 955     return 0;
 956 }
 957
 958 static void put_uint16(QEMUFile *f, void *pv, size_t size)
 959 {
 960     uint16_t *v = pv;
 961     qemu_put_be16s(f, v);
 962 }
 963
 964 const VMStateInfo vmstate_info_uint16 = {
 965     .name = "uint16",
 966     .get  = get_uint16,
 967     .put  = put_uint16,
 968 };
 969
 970 /* 32 bit unsigned int */
 971
 972 static int get_uint32(QEMUFile *f, void *pv, size_t size)
 973 {
 974     uint32_t *v = pv;
 975     qemu_get_be32s(f, v);
 976     return 0;
 977 }
 978
 979 static void put_uint32(QEMUFile *f, void *pv, size_t size)
 980 {
 981     uint32_t *v = pv;
 982     qemu_put_be32s(f, v);
 983 }
 984
 985 const VMStateInfo vmstate_info_uint32 = {
 986     .name = "uint32",
 987     .get  = get_uint32,
 988     .put  = put_uint32,
 989 };
 990
 991 /* 32 bit uint. See that the received value is the same than the one
 992    in the field */
 993
 994 static int get_uint32_equal(QEMUFile *f, void *pv, size_t size)
 995 {
 996     uint32_t *v = pv;
 997     uint32_t v2;
 998     qemu_get_be32s(f, &v2);
 999
1000     if (*v == v2) {
1001         return 0;
1002     }
1003     return -EINVAL;
1004 }
1005
1006 const VMStateInfo vmstate_info_uint32_equal = {
1007     .name = "uint32 equal",
1008     .get  = get_uint32_equal,
1009     .put  = put_uint32,
1010 };
1011
1012 /* 64 bit unsigned int */
1013
1014 static int get_uint64(QEMUFile *f, void *pv, size_t size)
1015 {
1016     uint64_t *v = pv;
1017     qemu_get_be64s(f, v);
1018     return 0;
1019 }
1020
1021 static void put_uint64(QEMUFile *f, void *pv, size_t size)
1022 {
1023     uint64_t *v = pv;
1024     qemu_put_be64s(f, v);
1025 }
1026
1027 const VMStateInfo vmstate_info_uint64 = {
1028     .name = "uint64",
1029     .get  = get_uint64,
1030     .put  = put_uint64,
1031 };
1032
1033 /* 8 bit int. See that the received value is the same than the one
1034    in the field */
1035
1036 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
1037 {
1038     uint8_t *v = pv;
1039     uint8_t v2;
1040     qemu_get_8s(f, &v2);
1041
1042     if (*v == v2)
1043         return 0;
1044     return -EINVAL;
1045 }
1046
1047 const VMStateInfo vmstate_info_uint8_equal = {
1048     .name = "uint8 equal",
1049     .get  = get_uint8_equal,
1050     .put  = put_uint8,
1051 };
1052
1053 /* 16 bit unsigned int int. See that the received value is the same than the one
1054    in the field */
1055
1056 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
1057 {
1058     uint16_t *v = pv;
1059     uint16_t v2;
1060     qemu_get_be16s(f, &v2);
1061
1062     if (*v == v2)
1063         return 0;
1064     return -EINVAL;
1065 }
1066
1067 const VMStateInfo vmstate_info_uint16_equal = {
1068     .name = "uint16 equal",
1069     .get  = get_uint16_equal,
1070     .put  = put_uint16,
1071 };
1072
1073 /* timers  */
1074
1075 static int get_timer(QEMUFile *f, void *pv, size_t size)
1076 {
1077     QEMUTimer *v = pv;
1078     qemu_get_timer(f, v);
1079     return 0;
1080 }
1081
1082 static void put_timer(QEMUFile *f, void *pv, size_t size)
1083 {
1084     QEMUTimer *v = pv;
1085     qemu_put_timer(f, v);
1086 }
1087
1088 const VMStateInfo vmstate_info_timer = {
1089     .name = "timer",
1090     .get  = get_timer,
1091     .put  = put_timer,
1092 };
1093
1094 /* uint8_t buffers */
1095
1096 static int get_buffer(QEMUFile *f, void *pv, size_t size)
1097 {
1098     uint8_t *v = pv;
1099     qemu_get_buffer(f, v, size);
1100     return 0;
1101 }
1102
1103 static void put_buffer(QEMUFile *f, void *pv, size_t size)
1104 {
1105     uint8_t *v = pv;
1106     qemu_put_buffer(f, v, size);
1107 }
1108
1109 const VMStateInfo vmstate_info_buffer = {
1110     .name = "buffer",
1111     .get  = get_buffer,
1112     .put  = put_buffer,
1113 };
1114
1115 /* unused buffers: space that was used for some fields that are
1116    not useful anymore */
1117
1118 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
1119 {
1120     uint8_t buf[1024];
1121     int block_len;
1122
1123     while (size > 0) {
1124         block_len = MIN(sizeof(buf), size);
1125         size -= block_len;
1126         qemu_get_buffer(f, buf, block_len);
1127     }
1128    return 0;
1129 }
1130
1131 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1132 {
1133     static const 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_put_buffer(f, buf, block_len);
1140     }
1141 }
1142
1143 const VMStateInfo vmstate_info_unused_buffer = {
1144     .name = "unused_buffer",
1145     .get  = get_unused_buffer,
1146     .put  = put_unused_buffer,
1147 };
1148
1149 /* bitmaps (as defined by bitmap.h). Note that size here is the size
1150  * of the bitmap in bits. The on-the-wire format of a bitmap is 64
1151  * bit words with the bits in big endian order. The in-memory format
1152  * is an array of 'unsigned long', which may be either 32 or 64 bits.
1153  */
1154 /* This is the number of 64 bit words sent over the wire */
1155 #define BITS_TO_U64S(nr) DIV_ROUND_UP(nr, 64)
1156 static int get_bitmap(QEMUFile *f, void *pv, size_t size)
1157 {
1158     unsigned long *bmp = pv;
1159     int i, idx = 0;
1160     for (i = 0; i < BITS_TO_U64S(size); i++) {
1161         uint64_t w = qemu_get_be64(f);
1162         bmp[idx++] = w;
1163         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1164             bmp[idx++] = w >> 32;
1165         }
1166     }
1167     return 0;
1168 }
1169
1170 static void put_bitmap(QEMUFile *f, void *pv, size_t size)
1171 {
1172     unsigned long *bmp = pv;
1173     int i, idx = 0;
1174     for (i = 0; i < BITS_TO_U64S(size); i++) {
1175         uint64_t w = bmp[idx++];
1176         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1177             w |= ((uint64_t)bmp[idx++]) << 32;
1178         }
1179         qemu_put_be64(f, w);
1180     }
1181 }
1182
1183 const VMStateInfo vmstate_info_bitmap = {
1184     .name = "bitmap",
1185     .get = get_bitmap,
1186     .put = put_bitmap,
1187 };
1188
1189 typedef struct CompatEntry {
1190     char idstr[256];
1191     int instance_id;
1192 } CompatEntry;
1193
1194 typedef struct SaveStateEntry {
1195     QTAILQ_ENTRY(SaveStateEntry) entry;
1196     char idstr[256];
1197     int instance_id;
1198     int alias_id;
1199     int version_id;
1200     int section_id;
1201     SaveVMHandlers *ops;
1202     const VMStateDescription *vmsd;
1203     void *opaque;
1204     CompatEntry *compat;
1205     int no_migrate;
1206     int is_ram;
1207 } SaveStateEntry;
1208
1209
1210 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1211     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1212 static int global_section_id;
1213
1214 static int calculate_new_instance_id(const char *idstr)
1215 {
1216     SaveStateEntry *se;
1217     int instance_id = 0;
1218
1219     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1220         if (strcmp(idstr, se->idstr) == 0
1221             && instance_id <= se->instance_id) {
1222             instance_id = se->instance_id + 1;
1223         }
1224     }
1225     return instance_id;
1226 }
1227
1228 static int calculate_compat_instance_id(const char *idstr)
1229 {
1230     SaveStateEntry *se;
1231     int instance_id = 0;
1232
1233     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1234         if (!se->compat)
1235             continue;
1236
1237         if (strcmp(idstr, se->compat->idstr) == 0
1238             && instance_id <= se->compat->instance_id) {
1239             instance_id = se->compat->instance_id + 1;
1240         }
1241     }
1242     return instance_id;
1243 }
1244
1245 /* TODO: Individual devices generally have very little idea about the rest
1246    of the system, so instance_id should be removed/replaced.
1247    Meanwhile pass -1 as instance_id if you do not already have a clearly
1248    distinguishing id for all instances of your device class. */
1249 int register_savevm_live(DeviceState *dev,
1250                          const char *idstr,
1251                          int instance_id,
1252                          int version_id,
1253                          SaveVMHandlers *ops,
1254                          void *opaque)
1255 {
1256     SaveStateEntry *se;
1257
1258     se = g_malloc0(sizeof(SaveStateEntry));
1259     se->version_id = version_id;
1260     se->section_id = global_section_id++;
1261     se->ops = ops;
1262     se->opaque = opaque;
1263     se->vmsd = NULL;
1264     se->no_migrate = 0;
1265     /* if this is a live_savem then set is_ram */
1266     if (ops->save_live_setup != NULL) {
1267         se->is_ram = 1;
1268     }
1269
1270     if (dev) {
1271         char *id = qdev_get_dev_path(dev);
1272         if (id) {
1273             pstrcpy(se->idstr, sizeof(se->idstr), id);
1274             pstrcat(se->idstr, sizeof(se->idstr), "/");
1275             g_free(id);
1276
1277             se->compat = g_malloc0(sizeof(CompatEntry));
1278             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1279             se->compat->instance_id = instance_id == -1 ?
1280                          calculate_compat_instance_id(idstr) : instance_id;
1281             instance_id = -1;
1282         }
1283     }
1284     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1285
1286     if (instance_id == -1) {
1287         se->instance_id = calculate_new_instance_id(se->idstr);
1288     } else {
1289         se->instance_id = instance_id;
1290     }
1291     assert(!se->compat || se->instance_id == 0);
1292     /* add at the end of list */
1293     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1294     return 0;
1295 }
1296
1297 int register_savevm(DeviceState *dev,
1298                     const char *idstr,
1299                     int instance_id,
1300                     int version_id,
1301                     SaveStateHandler *save_state,
1302                     LoadStateHandler *load_state,
1303                     void *opaque)
1304 {
1305     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
1306     ops->save_state = save_state;
1307     ops->load_state = load_state;
1308     return register_savevm_live(dev, idstr, instance_id, version_id,
1309                                 ops, opaque);
1310 }
1311
1312 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1313 {
1314     SaveStateEntry *se, *new_se;
1315     char id[256] = "";
1316
1317     if (dev) {
1318         char *path = qdev_get_dev_path(dev);
1319         if (path) {
1320             pstrcpy(id, sizeof(id), path);
1321             pstrcat(id, sizeof(id), "/");
1322             g_free(path);
1323         }
1324     }
1325     pstrcat(id, sizeof(id), idstr);
1326
1327     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1328         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1329             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1330             if (se->compat) {
1331                 g_free(se->compat);
1332             }
1333             g_free(se->ops);
1334             g_free(se);
1335         }
1336     }
1337 }
1338
1339 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1340                                    const VMStateDescription *vmsd,
1341                                    void *opaque, int alias_id,
1342                                    int required_for_version)
1343 {
1344     SaveStateEntry *se;
1345
1346     /* If this triggers, alias support can be dropped for the vmsd. */
1347     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1348
1349     se = g_malloc0(sizeof(SaveStateEntry));
1350     se->version_id = vmsd->version_id;
1351     se->section_id = global_section_id++;
1352     se->opaque = opaque;
1353     se->vmsd = vmsd;
1354     se->alias_id = alias_id;
1355     se->no_migrate = vmsd->unmigratable;
1356
1357     if (dev) {
1358         char *id = qdev_get_dev_path(dev);
1359         if (id) {
1360             pstrcpy(se->idstr, sizeof(se->idstr), id);
1361             pstrcat(se->idstr, sizeof(se->idstr), "/");
1362             g_free(id);
1363
1364             se->compat = g_malloc0(sizeof(CompatEntry));
1365             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1366             se->compat->instance_id = instance_id == -1 ?
1367                          calculate_compat_instance_id(vmsd->name) : instance_id;
1368             instance_id = -1;
1369         }
1370     }
1371     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1372
1373     if (instance_id == -1) {
1374         se->instance_id = calculate_new_instance_id(se->idstr);
1375     } else {
1376         se->instance_id = instance_id;
1377     }
1378     assert(!se->compat || se->instance_id == 0);
1379     /* add at the end of list */
1380     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1381     return 0;
1382 }
1383
1384 int vmstate_register(DeviceState *dev, int instance_id,
1385                      const VMStateDescription *vmsd, void *opaque)
1386 {
1387     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1388                                           opaque, -1, 0);
1389 }
1390
1391 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1392                         void *opaque)
1393 {
1394     SaveStateEntry *se, *new_se;
1395
1396     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1397         if (se->vmsd == vmsd && se->opaque == opaque) {
1398             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1399             if (se->compat) {
1400                 g_free(se->compat);
1401             }
1402             g_free(se);
1403         }
1404     }
1405 }
1406
1407 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1408                                     void *opaque);
1409 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1410                                    void *opaque);
1411
1412 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1413                        void *opaque, int version_id)
1414 {
1415     VMStateField *field = vmsd->fields;
1416     int ret;
1417
1418     if (version_id > vmsd->version_id) {
1419         return -EINVAL;
1420     }
1421     if (version_id < vmsd->minimum_version_id_old) {
1422         return -EINVAL;
1423     }
1424     if  (version_id < vmsd->minimum_version_id) {
1425         return vmsd->load_state_old(f, opaque, version_id);
1426     }
1427     if (vmsd->pre_load) {
1428         int ret = vmsd->pre_load(opaque);
1429         if (ret)
1430             return ret;
1431     }
1432     while(field->name) {
1433         if ((field->field_exists &&
1434              field->field_exists(opaque, version_id)) ||
1435             (!field->field_exists &&
1436              field->version_id <= version_id)) {
1437             void *base_addr = opaque + field->offset;
1438             int i, n_elems = 1;
1439             int size = field->size;
1440
1441             if (field->flags & VMS_VBUFFER) {
1442                 size = *(int32_t *)(opaque+field->size_offset);
1443                 if (field->flags & VMS_MULTIPLY) {
1444                     size *= field->size;
1445                 }
1446             }
1447             if (field->flags & VMS_ARRAY) {
1448                 n_elems = field->num;
1449             } else if (field->flags & VMS_VARRAY_INT32) {
1450                 n_elems = *(int32_t *)(opaque+field->num_offset);
1451             } else if (field->flags & VMS_VARRAY_UINT32) {
1452                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1453             } else if (field->flags & VMS_VARRAY_UINT16) {
1454                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1455             } else if (field->flags & VMS_VARRAY_UINT8) {
1456                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1457             }
1458             if (field->flags & VMS_POINTER) {
1459                 base_addr = *(void **)base_addr + field->start;
1460             }
1461             for (i = 0; i < n_elems; i++) {
1462                 void *addr = base_addr + size * i;
1463
1464                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1465                     addr = *(void **)addr;
1466                 }
1467                 if (field->flags & VMS_STRUCT) {
1468                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1469                 } else {
1470                     ret = field->info->get(f, addr, size);
1471
1472                 }
1473                 if (ret < 0) {
1474                     return ret;
1475                 }
1476             }
1477         }
1478         field++;
1479     }
1480     ret = vmstate_subsection_load(f, vmsd, opaque);
1481     if (ret != 0) {
1482         return ret;
1483     }
1484     if (vmsd->post_load) {
1485         return vmsd->post_load(opaque, version_id);
1486     }
1487     return 0;
1488 }
1489
1490 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1491                         void *opaque)
1492 {
1493     VMStateField *field = vmsd->fields;
1494
1495     if (vmsd->pre_save) {
1496         vmsd->pre_save(opaque);
1497     }
1498     while(field->name) {
1499         if (!field->field_exists ||
1500             field->field_exists(opaque, vmsd->version_id)) {
1501             void *base_addr = opaque + field->offset;
1502             int i, n_elems = 1;
1503             int size = field->size;
1504
1505             if (field->flags & VMS_VBUFFER) {
1506                 size = *(int32_t *)(opaque+field->size_offset);
1507                 if (field->flags & VMS_MULTIPLY) {
1508                     size *= field->size;
1509                 }
1510             }
1511             if (field->flags & VMS_ARRAY) {
1512                 n_elems = field->num;
1513             } else if (field->flags & VMS_VARRAY_INT32) {
1514                 n_elems = *(int32_t *)(opaque+field->num_offset);
1515             } else if (field->flags & VMS_VARRAY_UINT32) {
1516                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1517             } else if (field->flags & VMS_VARRAY_UINT16) {
1518                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1519             } else if (field->flags & VMS_VARRAY_UINT8) {
1520                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1521             }
1522             if (field->flags & VMS_POINTER) {
1523                 base_addr = *(void **)base_addr + field->start;
1524             }
1525             for (i = 0; i < n_elems; i++) {
1526                 void *addr = base_addr + size * i;
1527
1528                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1529                     addr = *(void **)addr;
1530                 }
1531                 if (field->flags & VMS_STRUCT) {
1532                     vmstate_save_state(f, field->vmsd, addr);
1533                 } else {
1534                     field->info->put(f, addr, size);
1535                 }
1536             }
1537         }
1538         field++;
1539     }
1540     vmstate_subsection_save(f, vmsd, opaque);
1541 }
1542
1543 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1544 {
1545     if (!se->vmsd) {         /* Old style */
1546         return se->ops->load_state(f, se->opaque, version_id);
1547     }
1548     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1549 }
1550
1551 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1552 {
1553     if (!se->vmsd) {         /* Old style */
1554         se->ops->save_state(f, se->opaque);
1555         return;
1556     }
1557     vmstate_save_state(f,se->vmsd, se->opaque);
1558 }
1559
1560 #define QEMU_VM_FILE_MAGIC           0x5145564d
1561 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1562 #define QEMU_VM_FILE_VERSION         0x00000003
1563
1564 #define QEMU_VM_EOF                  0x00
1565 #define QEMU_VM_SECTION_START        0x01
1566 #define QEMU_VM_SECTION_PART         0x02
1567 #define QEMU_VM_SECTION_END          0x03
1568 #define QEMU_VM_SECTION_FULL         0x04
1569 #define QEMU_VM_SUBSECTION           0x05
1570
1571 bool qemu_savevm_state_blocked(Error **errp)
1572 {
1573     SaveStateEntry *se;
1574
1575     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1576         if (se->no_migrate) {
1577             error_set(errp, QERR_MIGRATION_NOT_SUPPORTED, se->idstr);
1578             return true;
1579         }
1580     }
1581     return false;
1582 }
1583
1584 int qemu_savevm_state_begin(QEMUFile *f,
1585                             const MigrationParams *params)
1586 {
1587     SaveStateEntry *se;
1588     int ret;
1589
1590     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1591         if (!se->ops || !se->ops->set_params) {
1592             continue;
1593         }
1594         se->ops->set_params(params, se->opaque);
1595     }
1596
1597     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1598     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1599
1600     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1601         int len;
1602
1603         if (!se->ops || !se->ops->save_live_setup) {
1604             continue;
1605         }
1606         if (se->ops && se->ops->is_active) {
1607             if (!se->ops->is_active(se->opaque)) {
1608                 continue;
1609             }
1610         }
1611         /* Section type */
1612         qemu_put_byte(f, QEMU_VM_SECTION_START);
1613         qemu_put_be32(f, se->section_id);
1614
1615         /* ID string */
1616         len = strlen(se->idstr);
1617         qemu_put_byte(f, len);
1618         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1619
1620         qemu_put_be32(f, se->instance_id);
1621         qemu_put_be32(f, se->version_id);
1622
1623         ret = se->ops->save_live_setup(f, se->opaque);
1624         if (ret < 0) {
1625             qemu_savevm_state_cancel(f);
1626             return ret;
1627         }
1628     }
1629     ret = qemu_file_get_error(f);
1630     if (ret != 0) {
1631         qemu_savevm_state_cancel(f);
1632     }
1633
1634     return ret;
1635
1636 }
1637
1638 /*
1639  * this function has three return values:
1640  *   negative: there was one error, and we have -errno.
1641  *   0 : We haven't finished, caller have to go again
1642  *   1 : We have finished, we can go to complete phase
1643  */
1644 int qemu_savevm_state_iterate(QEMUFile *f)
1645 {
1646     SaveStateEntry *se;
1647     int ret = 1;
1648
1649     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1650         if (!se->ops || !se->ops->save_live_iterate) {
1651             continue;
1652         }
1653         if (se->ops && se->ops->is_active) {
1654             if (!se->ops->is_active(se->opaque)) {
1655                 continue;
1656             }
1657         }
1658         if (qemu_file_rate_limit(f)) {
1659             return 0;
1660         }
1661         trace_savevm_section_start();
1662         /* Section type */
1663         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1664         qemu_put_be32(f, se->section_id);
1665
1666         ret = se->ops->save_live_iterate(f, se->opaque);
1667         trace_savevm_section_end(se->section_id);
1668
1669         if (ret <= 0) {
1670             /* Do not proceed to the next vmstate before this one reported
1671                completion of the current stage. This serializes the migration
1672                and reduces the probability that a faster changing state is
1673                synchronized over and over again. */
1674             break;
1675         }
1676     }
1677     if (ret != 0) {
1678         return ret;
1679     }
1680     ret = qemu_file_get_error(f);
1681     if (ret != 0) {
1682         qemu_savevm_state_cancel(f);
1683     }
1684     return ret;
1685 }
1686
1687 int qemu_savevm_state_complete(QEMUFile *f)
1688 {
1689     SaveStateEntry *se;
1690     int ret;
1691
1692     cpu_synchronize_all_states();
1693
1694     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1695         if (!se->ops || !se->ops->save_live_complete) {
1696             continue;
1697         }
1698         if (se->ops && se->ops->is_active) {
1699             if (!se->ops->is_active(se->opaque)) {
1700                 continue;
1701             }
1702         }
1703         trace_savevm_section_start();
1704         /* Section type */
1705         qemu_put_byte(f, QEMU_VM_SECTION_END);
1706         qemu_put_be32(f, se->section_id);
1707
1708         ret = se->ops->save_live_complete(f, se->opaque);
1709         trace_savevm_section_end(se->section_id);
1710         if (ret < 0) {
1711             return ret;
1712         }
1713     }
1714
1715     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1716         int len;
1717
1718         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1719             continue;
1720         }
1721         trace_savevm_section_start();
1722         /* Section type */
1723         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1724         qemu_put_be32(f, se->section_id);
1725
1726         /* ID string */
1727         len = strlen(se->idstr);
1728         qemu_put_byte(f, len);
1729         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1730
1731         qemu_put_be32(f, se->instance_id);
1732         qemu_put_be32(f, se->version_id);
1733
1734         vmstate_save(f, se);
1735         trace_savevm_section_end(se->section_id);
1736     }
1737
1738     qemu_put_byte(f, QEMU_VM_EOF);
1739
1740     return qemu_file_get_error(f);
1741 }
1742
1743 void qemu_savevm_state_cancel(QEMUFile *f)
1744 {
1745     SaveStateEntry *se;
1746
1747     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1748         if (se->ops && se->ops->cancel) {
1749             se->ops->cancel(se->opaque);
1750         }
1751     }
1752 }
1753
1754 static int qemu_savevm_state(QEMUFile *f)
1755 {
1756     int ret;
1757     MigrationParams params = {
1758         .blk = 0,
1759         .shared = 0
1760     };
1761
1762     if (qemu_savevm_state_blocked(NULL)) {
1763         ret = -EINVAL;
1764         goto out;
1765     }
1766
1767     ret = qemu_savevm_state_begin(f, &params);
1768     if (ret < 0)
1769         goto out;
1770
1771     do {
1772         ret = qemu_savevm_state_iterate(f);
1773         if (ret < 0)
1774             goto out;
1775     } while (ret == 0);
1776
1777     ret = qemu_savevm_state_complete(f);
1778
1779 out:
1780     if (ret == 0) {
1781         ret = qemu_file_get_error(f);
1782     }
1783
1784     return ret;
1785 }
1786
1787 static int qemu_save_device_state(QEMUFile *f)
1788 {
1789     SaveStateEntry *se;
1790
1791     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1792     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1793
1794     cpu_synchronize_all_states();
1795
1796     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1797         int len;
1798
1799         if (se->is_ram) {
1800             continue;
1801         }
1802         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1803             continue;
1804         }
1805
1806         /* Section type */
1807         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1808         qemu_put_be32(f, se->section_id);
1809
1810         /* ID string */
1811         len = strlen(se->idstr);
1812         qemu_put_byte(f, len);
1813         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1814
1815         qemu_put_be32(f, se->instance_id);
1816         qemu_put_be32(f, se->version_id);
1817
1818         vmstate_save(f, se);
1819     }
1820
1821     qemu_put_byte(f, QEMU_VM_EOF);
1822
1823     return qemu_file_get_error(f);
1824 }
1825
1826 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1827 {
1828     SaveStateEntry *se;
1829
1830     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1831         if (!strcmp(se->idstr, idstr) &&
1832             (instance_id == se->instance_id ||
1833              instance_id == se->alias_id))
1834             return se;
1835         /* Migrating from an older version? */
1836         if (strstr(se->idstr, idstr) && se->compat) {
1837             if (!strcmp(se->compat->idstr, idstr) &&
1838                 (instance_id == se->compat->instance_id ||
1839                  instance_id == se->alias_id))
1840                 return se;
1841         }
1842     }
1843     return NULL;
1844 }
1845
1846 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1847 {
1848     while(sub && sub->needed) {
1849         if (strcmp(idstr, sub->vmsd->name) == 0) {
1850             return sub->vmsd;
1851         }
1852         sub++;
1853     }
1854     return NULL;
1855 }
1856
1857 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1858                                    void *opaque)
1859 {
1860     while (qemu_peek_byte(f, 0) == QEMU_VM_SUBSECTION) {
1861         char idstr[256];
1862         int ret;
1863         uint8_t version_id, len, size;
1864         const VMStateDescription *sub_vmsd;
1865
1866         len = qemu_peek_byte(f, 1);
1867         if (len < strlen(vmsd->name) + 1) {
1868             /* subsection name has be be "section_name/a" */
1869             return 0;
1870         }
1871         size = qemu_peek_buffer(f, (uint8_t *)idstr, len, 2);
1872         if (size != len) {
1873             return 0;
1874         }
1875         idstr[size] = 0;
1876
1877         if (strncmp(vmsd->name, idstr, strlen(vmsd->name)) != 0) {
1878             /* it don't have a valid subsection name */
1879             return 0;
1880         }
1881         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
1882         if (sub_vmsd == NULL) {
1883             return -ENOENT;
1884         }
1885         qemu_file_skip(f, 1); /* subsection */
1886         qemu_file_skip(f, 1); /* len */
1887         qemu_file_skip(f, len); /* idstr */
1888         version_id = qemu_get_be32(f);
1889
1890         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1891         if (ret) {
1892             return ret;
1893         }
1894     }
1895     return 0;
1896 }
1897
1898 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1899                                     void *opaque)
1900 {
1901     const VMStateSubsection *sub = vmsd->subsections;
1902
1903     while (sub && sub->needed) {
1904         if (sub->needed(opaque)) {
1905             const VMStateDescription *vmsd = sub->vmsd;
1906             uint8_t len;
1907
1908             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1909             len = strlen(vmsd->name);
1910             qemu_put_byte(f, len);
1911             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1912             qemu_put_be32(f, vmsd->version_id);
1913             vmstate_save_state(f, vmsd, opaque);
1914         }
1915         sub++;
1916     }
1917 }
1918
1919 typedef struct LoadStateEntry {
1920     QLIST_ENTRY(LoadStateEntry) entry;
1921     SaveStateEntry *se;
1922     int section_id;
1923     int version_id;
1924 } LoadStateEntry;
1925
1926 int qemu_loadvm_state(QEMUFile *f)
1927 {
1928     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1929         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1930     LoadStateEntry *le, *new_le;
1931     uint8_t section_type;
1932     unsigned int v;
1933     int ret;
1934
1935     if (qemu_savevm_state_blocked(NULL)) {
1936         return -EINVAL;
1937     }
1938
1939     v = qemu_get_be32(f);
1940     if (v != QEMU_VM_FILE_MAGIC)
1941         return -EINVAL;
1942
1943     v = qemu_get_be32(f);
1944     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1945         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1946         return -ENOTSUP;
1947     }
1948     if (v != QEMU_VM_FILE_VERSION)
1949         return -ENOTSUP;
1950
1951     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1952         uint32_t instance_id, version_id, section_id;
1953         SaveStateEntry *se;
1954         char idstr[257];
1955         int len;
1956
1957         switch (section_type) {
1958         case QEMU_VM_SECTION_START:
1959         case QEMU_VM_SECTION_FULL:
1960             /* Read section start */
1961             section_id = qemu_get_be32(f);
1962             len = qemu_get_byte(f);
1963             qemu_get_buffer(f, (uint8_t *)idstr, len);
1964             idstr[len] = 0;
1965             instance_id = qemu_get_be32(f);
1966             version_id = qemu_get_be32(f);
1967
1968             /* Find savevm section */
1969             se = find_se(idstr, instance_id);
1970             if (se == NULL) {
1971                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1972                 ret = -EINVAL;
1973                 goto out;
1974             }
1975
1976             /* Validate version */
1977             if (version_id > se->version_id) {
1978                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
1979                         version_id, idstr, se->version_id);
1980                 ret = -EINVAL;
1981                 goto out;
1982             }
1983
1984             /* Add entry */
1985             le = g_malloc0(sizeof(*le));
1986
1987             le->se = se;
1988             le->section_id = section_id;
1989             le->version_id = version_id;
1990             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
1991
1992             ret = vmstate_load(f, le->se, le->version_id);
1993             if (ret < 0) {
1994                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
1995                         instance_id, idstr);
1996                 goto out;
1997             }
1998             break;
1999         case QEMU_VM_SECTION_PART:
2000         case QEMU_VM_SECTION_END:
2001             section_id = qemu_get_be32(f);
2002
2003             QLIST_FOREACH(le, &loadvm_handlers, entry) {
2004                 if (le->section_id == section_id) {
2005                     break;
2006                 }
2007             }
2008             if (le == NULL) {
2009                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
2010                 ret = -EINVAL;
2011                 goto out;
2012             }
2013
2014             ret = vmstate_load(f, le->se, le->version_id);
2015             if (ret < 0) {
2016                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
2017                         section_id);
2018                 goto out;
2019             }
2020             break;
2021         default:
2022             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
2023             ret = -EINVAL;
2024             goto out;
2025         }
2026     }
2027
2028     cpu_synchronize_all_post_init();
2029
2030     ret = 0;
2031
2032 out:
2033     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
2034         QLIST_REMOVE(le, entry);
2035         g_free(le);
2036     }
2037
2038     if (ret == 0) {
2039         ret = qemu_file_get_error(f);
2040     }
2041
2042     return ret;
2043 }
2044
2045 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
2046                               const char *name)
2047 {
2048     QEMUSnapshotInfo *sn_tab, *sn;
2049     int nb_sns, i, ret;
2050
2051     ret = -ENOENT;
2052     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2053     if (nb_sns < 0)
2054         return ret;
2055     for(i = 0; i < nb_sns; i++) {
2056         sn = &sn_tab[i];
2057         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
2058             *sn_info = *sn;
2059             ret = 0;
2060             break;
2061         }
2062     }
2063     g_free(sn_tab);
2064     return ret;
2065 }
2066
2067 /*
2068  * Deletes snapshots of a given name in all opened images.
2069  */
2070 static int del_existing_snapshots(Monitor *mon, const char *name)
2071 {
2072     BlockDriverState *bs;
2073     QEMUSnapshotInfo sn1, *snapshot = &sn1;
2074     int ret;
2075
2076     bs = NULL;
2077     while ((bs = bdrv_next(bs))) {
2078         if (bdrv_can_snapshot(bs) &&
2079             bdrv_snapshot_find(bs, snapshot, name) >= 0)
2080         {
2081             ret = bdrv_snapshot_delete(bs, name);
2082             if (ret < 0) {
2083                 monitor_printf(mon,
2084                                "Error while deleting snapshot on '%s'\n",
2085                                bdrv_get_device_name(bs));
2086                 return -1;
2087             }
2088         }
2089     }
2090
2091     return 0;
2092 }
2093
2094 void do_savevm(Monitor *mon, const QDict *qdict)
2095 {
2096     BlockDriverState *bs, *bs1;
2097     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
2098     int ret;
2099     QEMUFile *f;
2100     int saved_vm_running;
2101     uint64_t vm_state_size;
2102 #ifdef _WIN32
2103     struct _timeb tb;
2104     struct tm *ptm;
2105 #else
2106     struct timeval tv;
2107     struct tm tm;
2108 #endif
2109     const char *name = qdict_get_try_str(qdict, "name");
2110
2111     /* Verify if there is a device that doesn't support snapshots and is writable */
2112     bs = NULL;
2113     while ((bs = bdrv_next(bs))) {
2114
2115         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2116             continue;
2117         }
2118
2119         if (!bdrv_can_snapshot(bs)) {
2120             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
2121                                bdrv_get_device_name(bs));
2122             return;
2123         }
2124     }
2125
2126     bs = bdrv_snapshots();
2127     if (!bs) {
2128         monitor_printf(mon, "No block device can accept snapshots\n");
2129         return;
2130     }
2131
2132     saved_vm_running = runstate_is_running();
2133     vm_stop(RUN_STATE_SAVE_VM);
2134
2135     memset(sn, 0, sizeof(*sn));
2136
2137     /* fill auxiliary fields */
2138 #ifdef _WIN32
2139     _ftime(&tb);
2140     sn->date_sec = tb.time;
2141     sn->date_nsec = tb.millitm * 1000000;
2142 #else
2143     gettimeofday(&tv, NULL);
2144     sn->date_sec = tv.tv_sec;
2145     sn->date_nsec = tv.tv_usec * 1000;
2146 #endif
2147     sn->vm_clock_nsec = qemu_get_clock_ns(vm_clock);
2148
2149     if (name) {
2150         ret = bdrv_snapshot_find(bs, old_sn, name);
2151         if (ret >= 0) {
2152             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
2153             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
2154         } else {
2155             pstrcpy(sn->name, sizeof(sn->name), name);
2156         }
2157     } else {
2158 #ifdef _WIN32
2159         time_t t = tb.time;
2160         ptm = localtime(&t);
2161         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", ptm);
2162 #else
2163         /* cast below needed for OpenBSD where tv_sec is still 'long' */
2164         localtime_r((const time_t *)&tv.tv_sec, &tm);
2165         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
2166 #endif
2167     }
2168
2169     /* Delete old snapshots of the same name */
2170     if (name && del_existing_snapshots(mon, name) < 0) {
2171         goto the_end;
2172     }
2173
2174     /* save the VM state */
2175     f = qemu_fopen_bdrv(bs, 1);
2176     if (!f) {
2177         monitor_printf(mon, "Could not open VM state file\n");
2178         goto the_end;
2179     }
2180     ret = qemu_savevm_state(f);
2181     vm_state_size = qemu_ftell(f);
2182     qemu_fclose(f);
2183     if (ret < 0) {
2184         monitor_printf(mon, "Error %d while writing VM\n", ret);
2185         goto the_end;
2186     }
2187
2188     /* create the snapshots */
2189
2190     bs1 = NULL;
2191     while ((bs1 = bdrv_next(bs1))) {
2192         if (bdrv_can_snapshot(bs1)) {
2193             /* Write VM state size only to the image that contains the state */
2194             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
2195             ret = bdrv_snapshot_create(bs1, sn);
2196             if (ret < 0) {
2197                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
2198                                bdrv_get_device_name(bs1));
2199             }
2200         }
2201     }
2202
2203  the_end:
2204     if (saved_vm_running)
2205         vm_start();
2206 }
2207
2208 void qmp_xen_save_devices_state(const char *filename, Error **errp)
2209 {
2210     QEMUFile *f;
2211     int saved_vm_running;
2212     int ret;
2213
2214     saved_vm_running = runstate_is_running();
2215     vm_stop(RUN_STATE_SAVE_VM);
2216
2217     f = qemu_fopen(filename, "wb");
2218     if (!f) {
2219         error_set(errp, QERR_OPEN_FILE_FAILED, filename);
2220         goto the_end;
2221     }
2222     ret = qemu_save_device_state(f);
2223     qemu_fclose(f);
2224     if (ret < 0) {
2225         error_set(errp, QERR_IO_ERROR);
2226     }
2227
2228  the_end:
2229     if (saved_vm_running)
2230         vm_start();
2231 }
2232
2233 int load_vmstate(const char *name)
2234 {
2235     BlockDriverState *bs, *bs_vm_state;
2236     QEMUSnapshotInfo sn;
2237     QEMUFile *f;
2238     int ret;
2239
2240     bs_vm_state = bdrv_snapshots();
2241     if (!bs_vm_state) {
2242         error_report("No block device supports snapshots");
2243         return -ENOTSUP;
2244     }
2245
2246     /* Don't even try to load empty VM states */
2247     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
2248     if (ret < 0) {
2249         return ret;
2250     } else if (sn.vm_state_size == 0) {
2251         error_report("This is a disk-only snapshot. Revert to it offline "
2252             "using qemu-img.");
2253         return -EINVAL;
2254     }
2255
2256     /* Verify if there is any device that doesn't support snapshots and is
2257     writable and check if the requested snapshot is available too. */
2258     bs = NULL;
2259     while ((bs = bdrv_next(bs))) {
2260
2261         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2262             continue;
2263         }
2264
2265         if (!bdrv_can_snapshot(bs)) {
2266             error_report("Device '%s' is writable but does not support snapshots.",
2267                                bdrv_get_device_name(bs));
2268             return -ENOTSUP;
2269         }
2270
2271         ret = bdrv_snapshot_find(bs, &sn, name);
2272         if (ret < 0) {
2273             error_report("Device '%s' does not have the requested snapshot '%s'",
2274                            bdrv_get_device_name(bs), name);
2275             return ret;
2276         }
2277     }
2278
2279     /* Flush all IO requests so they don't interfere with the new state.  */
2280     bdrv_drain_all();
2281
2282     bs = NULL;
2283     while ((bs = bdrv_next(bs))) {
2284         if (bdrv_can_snapshot(bs)) {
2285             ret = bdrv_snapshot_goto(bs, name);
2286             if (ret < 0) {
2287                 error_report("Error %d while activating snapshot '%s' on '%s'",
2288                              ret, name, bdrv_get_device_name(bs));
2289                 return ret;
2290             }
2291         }
2292     }
2293
2294     /* restore the VM state */
2295     f = qemu_fopen_bdrv(bs_vm_state, 0);
2296     if (!f) {
2297         error_report("Could not open VM state file");
2298         return -EINVAL;
2299     }
2300
2301     qemu_system_reset(VMRESET_SILENT);
2302     ret = qemu_loadvm_state(f);
2303
2304     qemu_fclose(f);
2305     if (ret < 0) {
2306         error_report("Error %d while loading VM state", ret);
2307         return ret;
2308     }
2309
2310     return 0;
2311 }
2312
2313 void do_delvm(Monitor *mon, const QDict *qdict)
2314 {
2315     BlockDriverState *bs, *bs1;
2316     int ret;
2317     const char *name = qdict_get_str(qdict, "name");
2318
2319     bs = bdrv_snapshots();
2320     if (!bs) {
2321         monitor_printf(mon, "No block device supports snapshots\n");
2322         return;
2323     }
2324
2325     bs1 = NULL;
2326     while ((bs1 = bdrv_next(bs1))) {
2327         if (bdrv_can_snapshot(bs1)) {
2328             ret = bdrv_snapshot_delete(bs1, name);
2329             if (ret < 0) {
2330                 if (ret == -ENOTSUP)
2331                     monitor_printf(mon,
2332                                    "Snapshots not supported on device '%s'\n",
2333                                    bdrv_get_device_name(bs1));
2334                 else
2335                     monitor_printf(mon, "Error %d while deleting snapshot on "
2336                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2337             }
2338         }
2339     }
2340 }
2341
2342 void do_info_snapshots(Monitor *mon)
2343 {
2344     BlockDriverState *bs, *bs1;
2345     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2346     int nb_sns, i, ret, available;
2347     int total;
2348     int *available_snapshots;
2349     char buf[256];
2350
2351     bs = bdrv_snapshots();
2352     if (!bs) {
2353         monitor_printf(mon, "No available block device supports snapshots\n");
2354         return;
2355     }
2356
2357     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2358     if (nb_sns < 0) {
2359         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2360         return;
2361     }
2362
2363     if (nb_sns == 0) {
2364         monitor_printf(mon, "There is no snapshot available.\n");
2365         return;
2366     }
2367
2368     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
2369     total = 0;
2370     for (i = 0; i < nb_sns; i++) {
2371         sn = &sn_tab[i];
2372         available = 1;
2373         bs1 = NULL;
2374
2375         while ((bs1 = bdrv_next(bs1))) {
2376             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2377                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2378                 if (ret < 0) {
2379                     available = 0;
2380                     break;
2381                 }
2382             }
2383         }
2384
2385         if (available) {
2386             available_snapshots[total] = i;
2387             total++;
2388         }
2389     }
2390
2391     if (total > 0) {
2392         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2393         for (i = 0; i < total; i++) {
2394             sn = &sn_tab[available_snapshots[i]];
2395             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2396         }
2397     } else {
2398         monitor_printf(mon, "There is no suitable snapshot available\n");
2399     }
2400
2401     g_free(sn_tab);
2402     g_free(available_snapshots);
2403
2404 }
2405
2406 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
2407 {
2408     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
2409                        memory_region_name(mr), dev);
2410 }
2411
2412 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
2413 {
2414     /* Nothing do to while the implementation is in RAMBlock */
2415 }
2416
2417 void vmstate_register_ram_global(MemoryRegion *mr)
2418 {
2419     vmstate_register_ram(mr, NULL);
2420 }
2421
2422 /*
2423   page = zrun nzrun
2424        | zrun nzrun page
2425
2426   zrun = length
2427
2428   nzrun = length byte...
2429
2430   length = uleb128 encoded integer
2431  */
2432 int xbzrle_encode_buffer(uint8_t *old_buf, uint8_t *new_buf, int slen,
2433                          uint8_t *dst, int dlen)
2434 {
2435     uint32_t zrun_len = 0, nzrun_len = 0;
2436     int d = 0, i = 0;
2437     long res, xor;
2438     uint8_t *nzrun_start = NULL;
2439
2440     g_assert(!(((uintptr_t)old_buf | (uintptr_t)new_buf | slen) %
2441                sizeof(long)));
2442
2443     while (i < slen) {
2444         /* overflow */
2445         if (d + 2 > dlen) {
2446             return -1;
2447         }
2448
2449         /* not aligned to sizeof(long) */
2450         res = (slen - i) % sizeof(long);
2451         while (res && old_buf[i] == new_buf[i]) {
2452             zrun_len++;
2453             i++;
2454             res--;
2455         }
2456
2457         /* word at a time for speed */
2458         if (!res) {
2459             while (i < slen &&
2460                    (*(long *)(old_buf + i)) == (*(long *)(new_buf + i))) {
2461                 i += sizeof(long);
2462                 zrun_len += sizeof(long);
2463             }
2464
2465             /* go over the rest */
2466             while (i < slen && old_buf[i] == new_buf[i]) {
2467                 zrun_len++;
2468                 i++;
2469             }
2470         }
2471
2472         /* buffer unchanged */
2473         if (zrun_len == slen) {
2474             return 0;
2475         }
2476
2477         /* skip last zero run */
2478         if (i == slen) {
2479             return d;
2480         }
2481
2482         d += uleb128_encode_small(dst + d, zrun_len);
2483
2484         zrun_len = 0;
2485         nzrun_start = new_buf + i;
2486
2487         /* overflow */
2488         if (d + 2 > dlen) {
2489             return -1;
2490         }
2491         /* not aligned to sizeof(long) */
2492         res = (slen - i) % sizeof(long);
2493         while (res && old_buf[i] != new_buf[i]) {
2494             i++;
2495             nzrun_len++;
2496             res--;
2497         }
2498
2499         /* word at a time for speed, use of 32-bit long okay */
2500         if (!res) {
2501             /* truncation to 32-bit long okay */
2502             long mask = (long)0x0101010101010101ULL;
2503             while (i < slen) {
2504                 xor = *(long *)(old_buf + i) ^ *(long *)(new_buf + i);
2505                 if ((xor - mask) & ~xor & (mask << 7)) {
2506                     /* found the end of an nzrun within the current long */
2507                     while (old_buf[i] != new_buf[i]) {
2508                         nzrun_len++;
2509                         i++;
2510                     }
2511                     break;
2512                 } else {
2513                     i += sizeof(long);
2514                     nzrun_len += sizeof(long);
2515                 }
2516             }
2517         }
2518
2519         d += uleb128_encode_small(dst + d, nzrun_len);
2520         /* overflow */
2521         if (d + nzrun_len > dlen) {
2522             return -1;
2523         }
2524         memcpy(dst + d, nzrun_start, nzrun_len);
2525         d += nzrun_len;
2526         nzrun_len = 0;
2527     }
2528
2529     return d;
2530 }
2531
2532 int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen)
2533 {
2534     int i = 0, d = 0;
2535     int ret;
2536     uint32_t count = 0;
2537
2538     while (i < slen) {
2539
2540         /* zrun */
2541         if ((slen - i) < 2) {
2542             return -1;
2543         }
2544
2545         ret = uleb128_decode_small(src + i, &count);
2546         if (ret < 0 || (i && !count)) {
2547             return -1;
2548         }
2549         i += ret;
2550         d += count;
2551
2552         /* overflow */
2553         if (d > dlen) {
2554             return -1;
2555         }
2556
2557         /* nzrun */
2558         if ((slen - i) < 2) {
2559             return -1;
2560         }
2561
2562         ret = uleb128_decode_small(src + i, &count);
2563         if (ret < 0 || !count) {
2564             return -1;
2565         }
2566         i += ret;
2567
2568         /* overflow */
2569         if (d + count > dlen || i + count > slen) {
2570             return -1;
2571         }
2572
2573         memcpy(dst + d, src + i, count);
2574         d += count;
2575         i += count;
2576     }
2577
2578     return d;
2579 }