savevm.c

   1 /*
   2  * QEMU System Emulator
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a copy
   7  * of this software and associated documentation files (the "Software"), to deal
   8  * in the Software without restriction, including without limitation the rights
   9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  10  * copies of the Software, and to permit persons to whom the Software is
  11  * furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice shall be included in
  14  * all copies or substantial portions of the Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  22  * THE SOFTWARE.
  23  */
  24
  25 #include "config-host.h"
  26 #include "qemu-common.h"
  27 #include "hw/hw.h"
  28 #include "hw/qdev.h"
  29 #include "net/net.h"
  30 #include "monitor/monitor.h"
  31 #include "sysemu/sysemu.h"
  32 #include "qemu/timer.h"
  33 #include "audio/audio.h"
  34 #include "migration/migration.h"
  35 #include "qemu/sockets.h"
  36 #include "qemu/queue.h"
  37 #include "sysemu/cpus.h"
  38 #include "exec/memory.h"
  39 #include "qmp-commands.h"
  40 #include "trace.h"
  41 #include "qemu/bitops.h"
  42
  43 #define SELF_ANNOUNCE_ROUNDS 5
  44
  45 #ifndef ETH_P_RARP
  46 #define ETH_P_RARP 0x8035
  47 #endif
  48 #define ARP_HTYPE_ETH 0x0001
  49 #define ARP_PTYPE_IP 0x0800
  50 #define ARP_OP_REQUEST_REV 0x3
  51
  52 static int announce_self_create(uint8_t *buf,
  53                                 uint8_t *mac_addr)
  54 {
  55     /* Ethernet header. */
  56     memset(buf, 0xff, 6);         /* destination MAC addr */
  57     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
  58     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
  59
  60     /* RARP header. */
  61     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
  62     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
  63     *(buf + 18) = 6; /* hardware addr length (ethernet) */
  64     *(buf + 19) = 4; /* protocol addr length (IPv4) */
  65     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
  66     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
  67     memset(buf + 28, 0x00, 4);     /* source protocol addr */
  68     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
  69     memset(buf + 38, 0x00, 4);     /* target protocol addr */
  70
  71     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
  72     memset(buf + 42, 0x00, 18);
  73
  74     return 60; /* len (FCS will be added by hardware) */
  75 }
  76
  77 static void qemu_announce_self_iter(NICState *nic, void *opaque)
  78 {
  79     uint8_t buf[60];
  80     int len;
  81
  82     len = announce_self_create(buf, nic->conf->macaddr.a);
  83
  84     qemu_send_packet_raw(&nic->nc, buf, len);
  85 }
  86
  87
  88 static void qemu_announce_self_once(void *opaque)
  89 {
  90     static int count = SELF_ANNOUNCE_ROUNDS;
  91     QEMUTimer *timer = *(QEMUTimer **)opaque;
  92
  93     qemu_foreach_nic(qemu_announce_self_iter, NULL);
  94
  95     if (--count) {
  96         /* delay 50ms, 150ms, 250ms, ... */
  97         qemu_mod_timer(timer, qemu_get_clock_ms(rt_clock) +
  98                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
  99     } else {
 100             qemu_del_timer(timer);
 101             qemu_free_timer(timer);
 102     }
 103 }
 104
 105 void qemu_announce_self(void)
 106 {
 107         static QEMUTimer *timer;
 108         timer = qemu_new_timer_ms(rt_clock, qemu_announce_self_once, &timer);
 109         qemu_announce_self_once(&timer);
 110 }
 111
 112 /***********************************************************/
 113 /* savevm/loadvm support */
 114
 115 #define IO_BUF_SIZE 32768
 116
 117 struct QEMUFile {
 118     const QEMUFileOps *ops;
 119     void *opaque;
 120     int is_write;
 121
 122     int64_t buf_offset; /* start of buffer when writing, end of buffer
 123                            when reading */
 124     int buf_index;
 125     int buf_size; /* 0 when writing */
 126     uint8_t buf[IO_BUF_SIZE];
 127
 128     int last_error;
 129 };
 130
 131 typedef struct QEMUFileStdio
 132 {
 133     FILE *stdio_file;
 134     QEMUFile *file;
 135 } QEMUFileStdio;
 136
 137 typedef struct QEMUFileSocket
 138 {
 139     int fd;
 140     QEMUFile *file;
 141 } QEMUFileSocket;
 142
 143 static int socket_get_fd(void *opaque)
 144 {
 145     QEMUFileSocket *s = opaque;
 146
 147     return s->fd;
 148 }
 149
 150 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 151 {
 152     QEMUFileSocket *s = opaque;
 153     ssize_t len;
 154
 155     for (;;) {
 156         len = qemu_recv(s->fd, buf, size, 0);
 157         if (len != -1) {
 158             break;
 159         }
 160         if (socket_error() == EAGAIN) {
 161             assert(qemu_in_coroutine());
 162             qemu_coroutine_yield();
 163         } else if (socket_error() != EINTR) {
 164             break;
 165         }
 166     }
 167
 168     if (len == -1) {
 169         len = -socket_error();
 170     }
 171     return len;
 172 }
 173
 174 static int socket_close(void *opaque)
 175 {
 176     QEMUFileSocket *s = opaque;
 177     closesocket(s->fd);
 178     g_free(s);
 179     return 0;
 180 }
 181
 182 static int stdio_get_fd(void *opaque)
 183 {
 184     QEMUFileStdio *s = opaque;
 185
 186     return fileno(s->stdio_file);
 187 }
 188
 189 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 190 {
 191     QEMUFileStdio *s = opaque;
 192     return fwrite(buf, 1, size, s->stdio_file);
 193 }
 194
 195 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 196 {
 197     QEMUFileStdio *s = opaque;
 198     FILE *fp = s->stdio_file;
 199     int bytes;
 200
 201     for (;;) {
 202         clearerr(fp);
 203         bytes = fread(buf, 1, size, fp);
 204         if (bytes != 0 || !ferror(fp)) {
 205             break;
 206         }
 207         if (errno == EAGAIN) {
 208             assert(qemu_in_coroutine());
 209             qemu_coroutine_yield();
 210         } else if (errno != EINTR) {
 211             break;
 212         }
 213     }
 214     return bytes;
 215 }
 216
 217 static int stdio_pclose(void *opaque)
 218 {
 219     QEMUFileStdio *s = opaque;
 220     int ret;
 221     ret = pclose(s->stdio_file);
 222     if (ret == -1) {
 223         ret = -errno;
 224     }
 225     g_free(s);
 226     return ret;
 227 }
 228
 229 static int stdio_fclose(void *opaque)
 230 {
 231     QEMUFileStdio *s = opaque;
 232     int ret = 0;
 233     if (fclose(s->stdio_file) == EOF) {
 234         ret = -errno;
 235     }
 236     g_free(s);
 237     return ret;
 238 }
 239
 240 static const QEMUFileOps stdio_pipe_read_ops = {
 241     .get_fd =     stdio_get_fd,
 242     .get_buffer = stdio_get_buffer,
 243     .close =      stdio_pclose
 244 };
 245
 246 static const QEMUFileOps stdio_pipe_write_ops = {
 247     .get_fd =     stdio_get_fd,
 248     .put_buffer = stdio_put_buffer,
 249     .close =      stdio_pclose
 250 };
 251
 252 QEMUFile *qemu_popen(FILE *stdio_file, const char *mode)
 253 {
 254     QEMUFileStdio *s;
 255
 256     if (stdio_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 257         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 258         return NULL;
 259     }
 260
 261     s = g_malloc0(sizeof(QEMUFileStdio));
 262
 263     s->stdio_file = stdio_file;
 264
 265     if(mode[0] == 'r') {
 266         s->file = qemu_fopen_ops(s, &stdio_pipe_read_ops);
 267     } else {
 268         s->file = qemu_fopen_ops(s, &stdio_pipe_write_ops);
 269     }
 270     return s->file;
 271 }
 272
 273 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 274 {
 275     FILE *popen_file;
 276
 277     popen_file = popen(command, mode);
 278     if(popen_file == NULL) {
 279         return NULL;
 280     }
 281
 282     return qemu_popen(popen_file, mode);
 283 }
 284
 285 static const QEMUFileOps stdio_file_read_ops = {
 286     .get_fd =     stdio_get_fd,
 287     .get_buffer = stdio_get_buffer,
 288     .close =      stdio_fclose
 289 };
 290
 291 static const QEMUFileOps stdio_file_write_ops = {
 292     .get_fd =     stdio_get_fd,
 293     .put_buffer = stdio_put_buffer,
 294     .close =      stdio_fclose
 295 };
 296
 297 QEMUFile *qemu_fdopen(int fd, const char *mode)
 298 {
 299     QEMUFileStdio *s;
 300
 301     if (mode == NULL ||
 302         (mode[0] != 'r' && mode[0] != 'w') ||
 303         mode[1] != 'b' || mode[2] != 0) {
 304         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 305         return NULL;
 306     }
 307
 308     s = g_malloc0(sizeof(QEMUFileStdio));
 309     s->stdio_file = fdopen(fd, mode);
 310     if (!s->stdio_file)
 311         goto fail;
 312
 313     if(mode[0] == 'r') {
 314         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 315     } else {
 316         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 317     }
 318     return s->file;
 319
 320 fail:
 321     g_free(s);
 322     return NULL;
 323 }
 324
 325 static const QEMUFileOps socket_read_ops = {
 326     .get_fd =     socket_get_fd,
 327     .get_buffer = socket_get_buffer,
 328     .close =      socket_close
 329 };
 330
 331 QEMUFile *qemu_fopen_socket(int fd)
 332 {
 333     QEMUFileSocket *s = g_malloc0(sizeof(QEMUFileSocket));
 334
 335     s->fd = fd;
 336     s->file = qemu_fopen_ops(s, &socket_read_ops);
 337     return s->file;
 338 }
 339
 340 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 341 {
 342     QEMUFileStdio *s;
 343
 344     if (mode == NULL ||
 345         (mode[0] != 'r' && mode[0] != 'w') ||
 346         mode[1] != 'b' || mode[2] != 0) {
 347         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 348         return NULL;
 349     }
 350
 351     s = g_malloc0(sizeof(QEMUFileStdio));
 352
 353     s->stdio_file = fopen(filename, mode);
 354     if (!s->stdio_file)
 355         goto fail;
 356
 357     if(mode[0] == 'w') {
 358         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 359     } else {
 360         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 361     }
 362     return s->file;
 363 fail:
 364     g_free(s);
 365     return NULL;
 366 }
 367
 368 static int block_put_buffer(void *opaque, const uint8_t *buf,
 369                            int64_t pos, int size)
 370 {
 371     bdrv_save_vmstate(opaque, buf, pos, size);
 372     return size;
 373 }
 374
 375 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 376 {
 377     return bdrv_load_vmstate(opaque, buf, pos, size);
 378 }
 379
 380 static int bdrv_fclose(void *opaque)
 381 {
 382     return bdrv_flush(opaque);
 383 }
 384
 385 static const QEMUFileOps bdrv_read_ops = {
 386     .get_buffer = block_get_buffer,
 387     .close =      bdrv_fclose
 388 };
 389
 390 static const QEMUFileOps bdrv_write_ops = {
 391     .put_buffer = block_put_buffer,
 392     .close =      bdrv_fclose
 393 };
 394
 395 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 396 {
 397     if (is_writable)
 398         return qemu_fopen_ops(bs, &bdrv_write_ops);
 399     return qemu_fopen_ops(bs, &bdrv_read_ops);
 400 }
 401
 402 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
 403 {
 404     QEMUFile *f;
 405
 406     f = g_malloc0(sizeof(QEMUFile));
 407
 408     f->opaque = opaque;
 409     f->ops = ops;
 410     f->is_write = 0;
 411
 412     return f;
 413 }
 414
 415 int qemu_file_get_error(QEMUFile *f)
 416 {
 417     return f->last_error;
 418 }
 419
 420 static void qemu_file_set_error(QEMUFile *f, int ret)
 421 {
 422     f->last_error = ret;
 423 }
 424
 425 /** Flushes QEMUFile buffer
 426  *
 427  */
 428 static int qemu_fflush(QEMUFile *f)
 429 {
 430     int ret = 0;
 431
 432     if (!f->ops->put_buffer)
 433         return 0;
 434
 435     if (f->is_write && f->buf_index > 0) {
 436         ret = f->ops->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 437         if (ret >= 0) {
 438             f->buf_offset += f->buf_index;
 439         }
 440         f->buf_index = 0;
 441     }
 442     return ret;
 443 }
 444
 445 static void qemu_fill_buffer(QEMUFile *f)
 446 {
 447     int len;
 448     int pending;
 449
 450     if (!f->ops->get_buffer)
 451         return;
 452
 453     if (f->is_write)
 454         abort();
 455
 456     pending = f->buf_size - f->buf_index;
 457     if (pending > 0) {
 458         memmove(f->buf, f->buf + f->buf_index, pending);
 459     }
 460     f->buf_index = 0;
 461     f->buf_size = pending;
 462
 463     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->buf_offset,
 464                         IO_BUF_SIZE - pending);
 465     if (len > 0) {
 466         f->buf_size += len;
 467         f->buf_offset += len;
 468     } else if (len == 0) {
 469         qemu_file_set_error(f, -EIO);
 470     } else if (len != -EAGAIN)
 471         qemu_file_set_error(f, len);
 472 }
 473
 474 int qemu_get_fd(QEMUFile *f)
 475 {
 476     if (f->ops->get_fd) {
 477         return f->ops->get_fd(f->opaque);
 478     }
 479     return -1;
 480 }
 481
 482 /** Closes the file
 483  *
 484  * Returns negative error value if any error happened on previous operations or
 485  * while closing the file. Returns 0 or positive number on success.
 486  *
 487  * The meaning of return value on success depends on the specific backend
 488  * being used.
 489  */
 490 int qemu_fclose(QEMUFile *f)
 491 {
 492     int ret;
 493     ret = qemu_fflush(f);
 494
 495     if (f->ops->close) {
 496         int ret2 = f->ops->close(f->opaque);
 497         if (ret >= 0) {
 498             ret = ret2;
 499         }
 500     }
 501     /* If any error was spotted before closing, we should report it
 502      * instead of the close() return value.
 503      */
 504     if (f->last_error) {
 505         ret = f->last_error;
 506     }
 507     g_free(f);
 508     return ret;
 509 }
 510
 511 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 512 {
 513     int l;
 514
 515     if (f->last_error) {
 516         return;
 517     }
 518
 519     if (f->is_write == 0 && f->buf_index > 0) {
 520         fprintf(stderr,
 521                 "Attempted to write to buffer while read buffer is not empty\n");
 522         abort();
 523     }
 524
 525     while (size > 0) {
 526         l = IO_BUF_SIZE - f->buf_index;
 527         if (l > size)
 528             l = size;
 529         memcpy(f->buf + f->buf_index, buf, l);
 530         f->is_write = 1;
 531         f->buf_index += l;
 532         buf += l;
 533         size -= l;
 534         if (f->buf_index >= IO_BUF_SIZE) {
 535             int ret = qemu_fflush(f);
 536             if (ret < 0) {
 537                 qemu_file_set_error(f, ret);
 538                 break;
 539             }
 540         }
 541     }
 542 }
 543
 544 void qemu_put_byte(QEMUFile *f, int v)
 545 {
 546     if (f->last_error) {
 547         return;
 548     }
 549
 550     if (f->is_write == 0 && f->buf_index > 0) {
 551         fprintf(stderr,
 552                 "Attempted to write to buffer while read buffer is not empty\n");
 553         abort();
 554     }
 555
 556     f->buf[f->buf_index++] = v;
 557     f->is_write = 1;
 558     if (f->buf_index >= IO_BUF_SIZE) {
 559         int ret = qemu_fflush(f);
 560         if (ret < 0) {
 561             qemu_file_set_error(f, ret);
 562         }
 563     }
 564 }
 565
 566 static void qemu_file_skip(QEMUFile *f, int size)
 567 {
 568     if (f->buf_index + size <= f->buf_size) {
 569         f->buf_index += size;
 570     }
 571 }
 572
 573 static int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset)
 574 {
 575     int pending;
 576     int index;
 577
 578     if (f->is_write) {
 579         abort();
 580     }
 581
 582     index = f->buf_index + offset;
 583     pending = f->buf_size - index;
 584     if (pending < size) {
 585         qemu_fill_buffer(f);
 586         index = f->buf_index + offset;
 587         pending = f->buf_size - index;
 588     }
 589
 590     if (pending <= 0) {
 591         return 0;
 592     }
 593     if (size > pending) {
 594         size = pending;
 595     }
 596
 597     memcpy(buf, f->buf + index, size);
 598     return size;
 599 }
 600
 601 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size)
 602 {
 603     int pending = size;
 604     int done = 0;
 605
 606     while (pending > 0) {
 607         int res;
 608
 609         res = qemu_peek_buffer(f, buf, pending, 0);
 610         if (res == 0) {
 611             return done;
 612         }
 613         qemu_file_skip(f, res);
 614         buf += res;
 615         pending -= res;
 616         done += res;
 617     }
 618     return done;
 619 }
 620
 621 static int qemu_peek_byte(QEMUFile *f, int offset)
 622 {
 623     int index = f->buf_index + offset;
 624
 625     if (f->is_write) {
 626         abort();
 627     }
 628
 629     if (index >= f->buf_size) {
 630         qemu_fill_buffer(f);
 631         index = f->buf_index + offset;
 632         if (index >= f->buf_size) {
 633             return 0;
 634         }
 635     }
 636     return f->buf[index];
 637 }
 638
 639 int qemu_get_byte(QEMUFile *f)
 640 {
 641     int result;
 642
 643     result = qemu_peek_byte(f, 0);
 644     qemu_file_skip(f, 1);
 645     return result;
 646 }
 647
 648 static int64_t qemu_ftell(QEMUFile *f)
 649 {
 650     return f->buf_offset - f->buf_size + f->buf_index;
 651 }
 652
 653 int qemu_file_rate_limit(QEMUFile *f)
 654 {
 655     if (f->ops->rate_limit)
 656         return f->ops->rate_limit(f->opaque);
 657
 658     return 0;
 659 }
 660
 661 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 662 {
 663     if (f->ops->get_rate_limit)
 664         return f->ops->get_rate_limit(f->opaque);
 665
 666     return 0;
 667 }
 668
 669 int64_t qemu_file_set_rate_limit(QEMUFile *f, int64_t new_rate)
 670 {
 671     /* any failed or completed migration keeps its state to allow probing of
 672      * migration data, but has no associated file anymore */
 673     if (f && f->ops->set_rate_limit)
 674         return f->ops->set_rate_limit(f->opaque, new_rate);
 675
 676     return 0;
 677 }
 678
 679 void qemu_put_be16(QEMUFile *f, unsigned int v)
 680 {
 681     qemu_put_byte(f, v >> 8);
 682     qemu_put_byte(f, v);
 683 }
 684
 685 void qemu_put_be32(QEMUFile *f, unsigned int v)
 686 {
 687     qemu_put_byte(f, v >> 24);
 688     qemu_put_byte(f, v >> 16);
 689     qemu_put_byte(f, v >> 8);
 690     qemu_put_byte(f, v);
 691 }
 692
 693 void qemu_put_be64(QEMUFile *f, uint64_t v)
 694 {
 695     qemu_put_be32(f, v >> 32);
 696     qemu_put_be32(f, v);
 697 }
 698
 699 unsigned int qemu_get_be16(QEMUFile *f)
 700 {
 701     unsigned int v;
 702     v = qemu_get_byte(f) << 8;
 703     v |= qemu_get_byte(f);
 704     return v;
 705 }
 706
 707 unsigned int qemu_get_be32(QEMUFile *f)
 708 {
 709     unsigned int v;
 710     v = qemu_get_byte(f) << 24;
 711     v |= qemu_get_byte(f) << 16;
 712     v |= qemu_get_byte(f) << 8;
 713     v |= qemu_get_byte(f);
 714     return v;
 715 }
 716
 717 uint64_t qemu_get_be64(QEMUFile *f)
 718 {
 719     uint64_t v;
 720     v = (uint64_t)qemu_get_be32(f) << 32;
 721     v |= qemu_get_be32(f);
 722     return v;
 723 }
 724
 725
 726 /* timer */
 727
 728 void qemu_put_timer(QEMUFile *f, QEMUTimer *ts)
 729 {
 730     uint64_t expire_time;
 731
 732     expire_time = qemu_timer_expire_time_ns(ts);
 733     qemu_put_be64(f, expire_time);
 734 }
 735
 736 void qemu_get_timer(QEMUFile *f, QEMUTimer *ts)
 737 {
 738     uint64_t expire_time;
 739
 740     expire_time = qemu_get_be64(f);
 741     if (expire_time != -1) {
 742         qemu_mod_timer_ns(ts, expire_time);
 743     } else {
 744         qemu_del_timer(ts);
 745     }
 746 }
 747
 748
 749 /* bool */
 750
 751 static int get_bool(QEMUFile *f, void *pv, size_t size)
 752 {
 753     bool *v = pv;
 754     *v = qemu_get_byte(f);
 755     return 0;
 756 }
 757
 758 static void put_bool(QEMUFile *f, void *pv, size_t size)
 759 {
 760     bool *v = pv;
 761     qemu_put_byte(f, *v);
 762 }
 763
 764 const VMStateInfo vmstate_info_bool = {
 765     .name = "bool",
 766     .get  = get_bool,
 767     .put  = put_bool,
 768 };
 769
 770 /* 8 bit int */
 771
 772 static int get_int8(QEMUFile *f, void *pv, size_t size)
 773 {
 774     int8_t *v = pv;
 775     qemu_get_s8s(f, v);
 776     return 0;
 777 }
 778
 779 static void put_int8(QEMUFile *f, void *pv, size_t size)
 780 {
 781     int8_t *v = pv;
 782     qemu_put_s8s(f, v);
 783 }
 784
 785 const VMStateInfo vmstate_info_int8 = {
 786     .name = "int8",
 787     .get  = get_int8,
 788     .put  = put_int8,
 789 };
 790
 791 /* 16 bit int */
 792
 793 static int get_int16(QEMUFile *f, void *pv, size_t size)
 794 {
 795     int16_t *v = pv;
 796     qemu_get_sbe16s(f, v);
 797     return 0;
 798 }
 799
 800 static void put_int16(QEMUFile *f, void *pv, size_t size)
 801 {
 802     int16_t *v = pv;
 803     qemu_put_sbe16s(f, v);
 804 }
 805
 806 const VMStateInfo vmstate_info_int16 = {
 807     .name = "int16",
 808     .get  = get_int16,
 809     .put  = put_int16,
 810 };
 811
 812 /* 32 bit int */
 813
 814 static int get_int32(QEMUFile *f, void *pv, size_t size)
 815 {
 816     int32_t *v = pv;
 817     qemu_get_sbe32s(f, v);
 818     return 0;
 819 }
 820
 821 static void put_int32(QEMUFile *f, void *pv, size_t size)
 822 {
 823     int32_t *v = pv;
 824     qemu_put_sbe32s(f, v);
 825 }
 826
 827 const VMStateInfo vmstate_info_int32 = {
 828     .name = "int32",
 829     .get  = get_int32,
 830     .put  = put_int32,
 831 };
 832
 833 /* 32 bit int. See that the received value is the same than the one
 834    in the field */
 835
 836 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 837 {
 838     int32_t *v = pv;
 839     int32_t v2;
 840     qemu_get_sbe32s(f, &v2);
 841
 842     if (*v == v2)
 843         return 0;
 844     return -EINVAL;
 845 }
 846
 847 const VMStateInfo vmstate_info_int32_equal = {
 848     .name = "int32 equal",
 849     .get  = get_int32_equal,
 850     .put  = put_int32,
 851 };
 852
 853 /* 32 bit int. See that the received value is the less or the same
 854    than the one in the field */
 855
 856 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 857 {
 858     int32_t *old = pv;
 859     int32_t new;
 860     qemu_get_sbe32s(f, &new);
 861
 862     if (*old <= new)
 863         return 0;
 864     return -EINVAL;
 865 }
 866
 867 const VMStateInfo vmstate_info_int32_le = {
 868     .name = "int32 equal",
 869     .get  = get_int32_le,
 870     .put  = put_int32,
 871 };
 872
 873 /* 64 bit int */
 874
 875 static int get_int64(QEMUFile *f, void *pv, size_t size)
 876 {
 877     int64_t *v = pv;
 878     qemu_get_sbe64s(f, v);
 879     return 0;
 880 }
 881
 882 static void put_int64(QEMUFile *f, void *pv, size_t size)
 883 {
 884     int64_t *v = pv;
 885     qemu_put_sbe64s(f, v);
 886 }
 887
 888 const VMStateInfo vmstate_info_int64 = {
 889     .name = "int64",
 890     .get  = get_int64,
 891     .put  = put_int64,
 892 };
 893
 894 /* 8 bit unsigned int */
 895
 896 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 897 {
 898     uint8_t *v = pv;
 899     qemu_get_8s(f, v);
 900     return 0;
 901 }
 902
 903 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 904 {
 905     uint8_t *v = pv;
 906     qemu_put_8s(f, v);
 907 }
 908
 909 const VMStateInfo vmstate_info_uint8 = {
 910     .name = "uint8",
 911     .get  = get_uint8,
 912     .put  = put_uint8,
 913 };
 914
 915 /* 16 bit unsigned int */
 916
 917 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 918 {
 919     uint16_t *v = pv;
 920     qemu_get_be16s(f, v);
 921     return 0;
 922 }
 923
 924 static void put_uint16(QEMUFile *f, void *pv, size_t size)
 925 {
 926     uint16_t *v = pv;
 927     qemu_put_be16s(f, v);
 928 }
 929
 930 const VMStateInfo vmstate_info_uint16 = {
 931     .name = "uint16",
 932     .get  = get_uint16,
 933     .put  = put_uint16,
 934 };
 935
 936 /* 32 bit unsigned int */
 937
 938 static int get_uint32(QEMUFile *f, void *pv, size_t size)
 939 {
 940     uint32_t *v = pv;
 941     qemu_get_be32s(f, v);
 942     return 0;
 943 }
 944
 945 static void put_uint32(QEMUFile *f, void *pv, size_t size)
 946 {
 947     uint32_t *v = pv;
 948     qemu_put_be32s(f, v);
 949 }
 950
 951 const VMStateInfo vmstate_info_uint32 = {
 952     .name = "uint32",
 953     .get  = get_uint32,
 954     .put  = put_uint32,
 955 };
 956
 957 /* 32 bit uint. See that the received value is the same than the one
 958    in the field */
 959
 960 static int get_uint32_equal(QEMUFile *f, void *pv, size_t size)
 961 {
 962     uint32_t *v = pv;
 963     uint32_t v2;
 964     qemu_get_be32s(f, &v2);
 965
 966     if (*v == v2) {
 967         return 0;
 968     }
 969     return -EINVAL;
 970 }
 971
 972 const VMStateInfo vmstate_info_uint32_equal = {
 973     .name = "uint32 equal",
 974     .get  = get_uint32_equal,
 975     .put  = put_uint32,
 976 };
 977
 978 /* 64 bit unsigned int */
 979
 980 static int get_uint64(QEMUFile *f, void *pv, size_t size)
 981 {
 982     uint64_t *v = pv;
 983     qemu_get_be64s(f, v);
 984     return 0;
 985 }
 986
 987 static void put_uint64(QEMUFile *f, void *pv, size_t size)
 988 {
 989     uint64_t *v = pv;
 990     qemu_put_be64s(f, v);
 991 }
 992
 993 const VMStateInfo vmstate_info_uint64 = {
 994     .name = "uint64",
 995     .get  = get_uint64,
 996     .put  = put_uint64,
 997 };
 998
 999 /* 8 bit int. See that the received value is the same than the one
1000    in the field */
1001
1002 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
1003 {
1004     uint8_t *v = pv;
1005     uint8_t v2;
1006     qemu_get_8s(f, &v2);
1007
1008     if (*v == v2)
1009         return 0;
1010     return -EINVAL;
1011 }
1012
1013 const VMStateInfo vmstate_info_uint8_equal = {
1014     .name = "uint8 equal",
1015     .get  = get_uint8_equal,
1016     .put  = put_uint8,
1017 };
1018
1019 /* 16 bit unsigned int int. See that the received value is the same than the one
1020    in the field */
1021
1022 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
1023 {
1024     uint16_t *v = pv;
1025     uint16_t v2;
1026     qemu_get_be16s(f, &v2);
1027
1028     if (*v == v2)
1029         return 0;
1030     return -EINVAL;
1031 }
1032
1033 const VMStateInfo vmstate_info_uint16_equal = {
1034     .name = "uint16 equal",
1035     .get  = get_uint16_equal,
1036     .put  = put_uint16,
1037 };
1038
1039 /* timers  */
1040
1041 static int get_timer(QEMUFile *f, void *pv, size_t size)
1042 {
1043     QEMUTimer *v = pv;
1044     qemu_get_timer(f, v);
1045     return 0;
1046 }
1047
1048 static void put_timer(QEMUFile *f, void *pv, size_t size)
1049 {
1050     QEMUTimer *v = pv;
1051     qemu_put_timer(f, v);
1052 }
1053
1054 const VMStateInfo vmstate_info_timer = {
1055     .name = "timer",
1056     .get  = get_timer,
1057     .put  = put_timer,
1058 };
1059
1060 /* uint8_t buffers */
1061
1062 static int get_buffer(QEMUFile *f, void *pv, size_t size)
1063 {
1064     uint8_t *v = pv;
1065     qemu_get_buffer(f, v, size);
1066     return 0;
1067 }
1068
1069 static void put_buffer(QEMUFile *f, void *pv, size_t size)
1070 {
1071     uint8_t *v = pv;
1072     qemu_put_buffer(f, v, size);
1073 }
1074
1075 const VMStateInfo vmstate_info_buffer = {
1076     .name = "buffer",
1077     .get  = get_buffer,
1078     .put  = put_buffer,
1079 };
1080
1081 /* unused buffers: space that was used for some fields that are
1082    not useful anymore */
1083
1084 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
1085 {
1086     uint8_t buf[1024];
1087     int block_len;
1088
1089     while (size > 0) {
1090         block_len = MIN(sizeof(buf), size);
1091         size -= block_len;
1092         qemu_get_buffer(f, buf, block_len);
1093     }
1094    return 0;
1095 }
1096
1097 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1098 {
1099     static const uint8_t buf[1024];
1100     int block_len;
1101
1102     while (size > 0) {
1103         block_len = MIN(sizeof(buf), size);
1104         size -= block_len;
1105         qemu_put_buffer(f, buf, block_len);
1106     }
1107 }
1108
1109 const VMStateInfo vmstate_info_unused_buffer = {
1110     .name = "unused_buffer",
1111     .get  = get_unused_buffer,
1112     .put  = put_unused_buffer,
1113 };
1114
1115 /* bitmaps (as defined by bitmap.h). Note that size here is the size
1116  * of the bitmap in bits. The on-the-wire format of a bitmap is 64
1117  * bit words with the bits in big endian order. The in-memory format
1118  * is an array of 'unsigned long', which may be either 32 or 64 bits.
1119  */
1120 /* This is the number of 64 bit words sent over the wire */
1121 #define BITS_TO_U64S(nr) DIV_ROUND_UP(nr, 64)
1122 static int get_bitmap(QEMUFile *f, void *pv, size_t size)
1123 {
1124     unsigned long *bmp = pv;
1125     int i, idx = 0;
1126     for (i = 0; i < BITS_TO_U64S(size); i++) {
1127         uint64_t w = qemu_get_be64(f);
1128         bmp[idx++] = w;
1129         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1130             bmp[idx++] = w >> 32;
1131         }
1132     }
1133     return 0;
1134 }
1135
1136 static void put_bitmap(QEMUFile *f, void *pv, size_t size)
1137 {
1138     unsigned long *bmp = pv;
1139     int i, idx = 0;
1140     for (i = 0; i < BITS_TO_U64S(size); i++) {
1141         uint64_t w = bmp[idx++];
1142         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1143             w |= ((uint64_t)bmp[idx++]) << 32;
1144         }
1145         qemu_put_be64(f, w);
1146     }
1147 }
1148
1149 const VMStateInfo vmstate_info_bitmap = {
1150     .name = "bitmap",
1151     .get = get_bitmap,
1152     .put = put_bitmap,
1153 };
1154
1155 typedef struct CompatEntry {
1156     char idstr[256];
1157     int instance_id;
1158 } CompatEntry;
1159
1160 typedef struct SaveStateEntry {
1161     QTAILQ_ENTRY(SaveStateEntry) entry;
1162     char idstr[256];
1163     int instance_id;
1164     int alias_id;
1165     int version_id;
1166     int section_id;
1167     SaveVMHandlers *ops;
1168     const VMStateDescription *vmsd;
1169     void *opaque;
1170     CompatEntry *compat;
1171     int no_migrate;
1172     int is_ram;
1173 } SaveStateEntry;
1174
1175
1176 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1177     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1178 static int global_section_id;
1179
1180 static int calculate_new_instance_id(const char *idstr)
1181 {
1182     SaveStateEntry *se;
1183     int instance_id = 0;
1184
1185     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1186         if (strcmp(idstr, se->idstr) == 0
1187             && instance_id <= se->instance_id) {
1188             instance_id = se->instance_id + 1;
1189         }
1190     }
1191     return instance_id;
1192 }
1193
1194 static int calculate_compat_instance_id(const char *idstr)
1195 {
1196     SaveStateEntry *se;
1197     int instance_id = 0;
1198
1199     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1200         if (!se->compat)
1201             continue;
1202
1203         if (strcmp(idstr, se->compat->idstr) == 0
1204             && instance_id <= se->compat->instance_id) {
1205             instance_id = se->compat->instance_id + 1;
1206         }
1207     }
1208     return instance_id;
1209 }
1210
1211 /* TODO: Individual devices generally have very little idea about the rest
1212    of the system, so instance_id should be removed/replaced.
1213    Meanwhile pass -1 as instance_id if you do not already have a clearly
1214    distinguishing id for all instances of your device class. */
1215 int register_savevm_live(DeviceState *dev,
1216                          const char *idstr,
1217                          int instance_id,
1218                          int version_id,
1219                          SaveVMHandlers *ops,
1220                          void *opaque)
1221 {
1222     SaveStateEntry *se;
1223
1224     se = g_malloc0(sizeof(SaveStateEntry));
1225     se->version_id = version_id;
1226     se->section_id = global_section_id++;
1227     se->ops = ops;
1228     se->opaque = opaque;
1229     se->vmsd = NULL;
1230     se->no_migrate = 0;
1231     /* if this is a live_savem then set is_ram */
1232     if (ops->save_live_setup != NULL) {
1233         se->is_ram = 1;
1234     }
1235
1236     if (dev) {
1237         char *id = qdev_get_dev_path(dev);
1238         if (id) {
1239             pstrcpy(se->idstr, sizeof(se->idstr), id);
1240             pstrcat(se->idstr, sizeof(se->idstr), "/");
1241             g_free(id);
1242
1243             se->compat = g_malloc0(sizeof(CompatEntry));
1244             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1245             se->compat->instance_id = instance_id == -1 ?
1246                          calculate_compat_instance_id(idstr) : instance_id;
1247             instance_id = -1;
1248         }
1249     }
1250     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1251
1252     if (instance_id == -1) {
1253         se->instance_id = calculate_new_instance_id(se->idstr);
1254     } else {
1255         se->instance_id = instance_id;
1256     }
1257     assert(!se->compat || se->instance_id == 0);
1258     /* add at the end of list */
1259     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1260     return 0;
1261 }
1262
1263 int register_savevm(DeviceState *dev,
1264                     const char *idstr,
1265                     int instance_id,
1266                     int version_id,
1267                     SaveStateHandler *save_state,
1268                     LoadStateHandler *load_state,
1269                     void *opaque)
1270 {
1271     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
1272     ops->save_state = save_state;
1273     ops->load_state = load_state;
1274     return register_savevm_live(dev, idstr, instance_id, version_id,
1275                                 ops, opaque);
1276 }
1277
1278 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1279 {
1280     SaveStateEntry *se, *new_se;
1281     char id[256] = "";
1282
1283     if (dev) {
1284         char *path = qdev_get_dev_path(dev);
1285         if (path) {
1286             pstrcpy(id, sizeof(id), path);
1287             pstrcat(id, sizeof(id), "/");
1288             g_free(path);
1289         }
1290     }
1291     pstrcat(id, sizeof(id), idstr);
1292
1293     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1294         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1295             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1296             if (se->compat) {
1297                 g_free(se->compat);
1298             }
1299             g_free(se->ops);
1300             g_free(se);
1301         }
1302     }
1303 }
1304
1305 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1306                                    const VMStateDescription *vmsd,
1307                                    void *opaque, int alias_id,
1308                                    int required_for_version)
1309 {
1310     SaveStateEntry *se;
1311
1312     /* If this triggers, alias support can be dropped for the vmsd. */
1313     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1314
1315     se = g_malloc0(sizeof(SaveStateEntry));
1316     se->version_id = vmsd->version_id;
1317     se->section_id = global_section_id++;
1318     se->opaque = opaque;
1319     se->vmsd = vmsd;
1320     se->alias_id = alias_id;
1321     se->no_migrate = vmsd->unmigratable;
1322
1323     if (dev) {
1324         char *id = qdev_get_dev_path(dev);
1325         if (id) {
1326             pstrcpy(se->idstr, sizeof(se->idstr), id);
1327             pstrcat(se->idstr, sizeof(se->idstr), "/");
1328             g_free(id);
1329
1330             se->compat = g_malloc0(sizeof(CompatEntry));
1331             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1332             se->compat->instance_id = instance_id == -1 ?
1333                          calculate_compat_instance_id(vmsd->name) : instance_id;
1334             instance_id = -1;
1335         }
1336     }
1337     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1338
1339     if (instance_id == -1) {
1340         se->instance_id = calculate_new_instance_id(se->idstr);
1341     } else {
1342         se->instance_id = instance_id;
1343     }
1344     assert(!se->compat || se->instance_id == 0);
1345     /* add at the end of list */
1346     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1347     return 0;
1348 }
1349
1350 int vmstate_register(DeviceState *dev, int instance_id,
1351                      const VMStateDescription *vmsd, void *opaque)
1352 {
1353     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1354                                           opaque, -1, 0);
1355 }
1356
1357 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1358                         void *opaque)
1359 {
1360     SaveStateEntry *se, *new_se;
1361
1362     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1363         if (se->vmsd == vmsd && se->opaque == opaque) {
1364             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1365             if (se->compat) {
1366                 g_free(se->compat);
1367             }
1368             g_free(se);
1369         }
1370     }
1371 }
1372
1373 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1374                                     void *opaque);
1375 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1376                                    void *opaque);
1377
1378 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1379                        void *opaque, int version_id)
1380 {
1381     VMStateField *field = vmsd->fields;
1382     int ret;
1383
1384     if (version_id > vmsd->version_id) {
1385         return -EINVAL;
1386     }
1387     if (version_id < vmsd->minimum_version_id_old) {
1388         return -EINVAL;
1389     }
1390     if  (version_id < vmsd->minimum_version_id) {
1391         return vmsd->load_state_old(f, opaque, version_id);
1392     }
1393     if (vmsd->pre_load) {
1394         int ret = vmsd->pre_load(opaque);
1395         if (ret)
1396             return ret;
1397     }
1398     while(field->name) {
1399         if ((field->field_exists &&
1400              field->field_exists(opaque, version_id)) ||
1401             (!field->field_exists &&
1402              field->version_id <= version_id)) {
1403             void *base_addr = opaque + field->offset;
1404             int i, n_elems = 1;
1405             int size = field->size;
1406
1407             if (field->flags & VMS_VBUFFER) {
1408                 size = *(int32_t *)(opaque+field->size_offset);
1409                 if (field->flags & VMS_MULTIPLY) {
1410                     size *= field->size;
1411                 }
1412             }
1413             if (field->flags & VMS_ARRAY) {
1414                 n_elems = field->num;
1415             } else if (field->flags & VMS_VARRAY_INT32) {
1416                 n_elems = *(int32_t *)(opaque+field->num_offset);
1417             } else if (field->flags & VMS_VARRAY_UINT32) {
1418                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1419             } else if (field->flags & VMS_VARRAY_UINT16) {
1420                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1421             } else if (field->flags & VMS_VARRAY_UINT8) {
1422                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1423             }
1424             if (field->flags & VMS_POINTER) {
1425                 base_addr = *(void **)base_addr + field->start;
1426             }
1427             for (i = 0; i < n_elems; i++) {
1428                 void *addr = base_addr + size * i;
1429
1430                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1431                     addr = *(void **)addr;
1432                 }
1433                 if (field->flags & VMS_STRUCT) {
1434                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1435                 } else {
1436                     ret = field->info->get(f, addr, size);
1437
1438                 }
1439                 if (ret < 0) {
1440                     return ret;
1441                 }
1442             }
1443         }
1444         field++;
1445     }
1446     ret = vmstate_subsection_load(f, vmsd, opaque);
1447     if (ret != 0) {
1448         return ret;
1449     }
1450     if (vmsd->post_load) {
1451         return vmsd->post_load(opaque, version_id);
1452     }
1453     return 0;
1454 }
1455
1456 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1457                         void *opaque)
1458 {
1459     VMStateField *field = vmsd->fields;
1460
1461     if (vmsd->pre_save) {
1462         vmsd->pre_save(opaque);
1463     }
1464     while(field->name) {
1465         if (!field->field_exists ||
1466             field->field_exists(opaque, vmsd->version_id)) {
1467             void *base_addr = opaque + field->offset;
1468             int i, n_elems = 1;
1469             int size = field->size;
1470
1471             if (field->flags & VMS_VBUFFER) {
1472                 size = *(int32_t *)(opaque+field->size_offset);
1473                 if (field->flags & VMS_MULTIPLY) {
1474                     size *= field->size;
1475                 }
1476             }
1477             if (field->flags & VMS_ARRAY) {
1478                 n_elems = field->num;
1479             } else if (field->flags & VMS_VARRAY_INT32) {
1480                 n_elems = *(int32_t *)(opaque+field->num_offset);
1481             } else if (field->flags & VMS_VARRAY_UINT32) {
1482                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1483             } else if (field->flags & VMS_VARRAY_UINT16) {
1484                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1485             } else if (field->flags & VMS_VARRAY_UINT8) {
1486                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1487             }
1488             if (field->flags & VMS_POINTER) {
1489                 base_addr = *(void **)base_addr + field->start;
1490             }
1491             for (i = 0; i < n_elems; i++) {
1492                 void *addr = base_addr + size * i;
1493
1494                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1495                     addr = *(void **)addr;
1496                 }
1497                 if (field->flags & VMS_STRUCT) {
1498                     vmstate_save_state(f, field->vmsd, addr);
1499                 } else {
1500                     field->info->put(f, addr, size);
1501                 }
1502             }
1503         }
1504         field++;
1505     }
1506     vmstate_subsection_save(f, vmsd, opaque);
1507 }
1508
1509 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1510 {
1511     if (!se->vmsd) {         /* Old style */
1512         return se->ops->load_state(f, se->opaque, version_id);
1513     }
1514     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1515 }
1516
1517 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1518 {
1519     if (!se->vmsd) {         /* Old style */
1520         se->ops->save_state(f, se->opaque);
1521         return;
1522     }
1523     vmstate_save_state(f,se->vmsd, se->opaque);
1524 }
1525
1526 #define QEMU_VM_FILE_MAGIC           0x5145564d
1527 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1528 #define QEMU_VM_FILE_VERSION         0x00000003
1529
1530 #define QEMU_VM_EOF                  0x00
1531 #define QEMU_VM_SECTION_START        0x01
1532 #define QEMU_VM_SECTION_PART         0x02
1533 #define QEMU_VM_SECTION_END          0x03
1534 #define QEMU_VM_SECTION_FULL         0x04
1535 #define QEMU_VM_SUBSECTION           0x05
1536
1537 bool qemu_savevm_state_blocked(Error **errp)
1538 {
1539     SaveStateEntry *se;
1540
1541     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1542         if (se->no_migrate) {
1543             error_set(errp, QERR_MIGRATION_NOT_SUPPORTED, se->idstr);
1544             return true;
1545         }
1546     }
1547     return false;
1548 }
1549
1550 int qemu_savevm_state_begin(QEMUFile *f,
1551                             const MigrationParams *params)
1552 {
1553     SaveStateEntry *se;
1554     int ret;
1555
1556     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1557         if (!se->ops || !se->ops->set_params) {
1558             continue;
1559         }
1560         se->ops->set_params(params, se->opaque);
1561     }
1562
1563     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1564     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1565
1566     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1567         int len;
1568
1569         if (!se->ops || !se->ops->save_live_setup) {
1570             continue;
1571         }
1572         if (se->ops && se->ops->is_active) {
1573             if (!se->ops->is_active(se->opaque)) {
1574                 continue;
1575             }
1576         }
1577         /* Section type */
1578         qemu_put_byte(f, QEMU_VM_SECTION_START);
1579         qemu_put_be32(f, se->section_id);
1580
1581         /* ID string */
1582         len = strlen(se->idstr);
1583         qemu_put_byte(f, len);
1584         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1585
1586         qemu_put_be32(f, se->instance_id);
1587         qemu_put_be32(f, se->version_id);
1588
1589         ret = se->ops->save_live_setup(f, se->opaque);
1590         if (ret < 0) {
1591             qemu_savevm_state_cancel(f);
1592             return ret;
1593         }
1594     }
1595     ret = qemu_file_get_error(f);
1596     if (ret != 0) {
1597         qemu_savevm_state_cancel(f);
1598     }
1599
1600     return ret;
1601
1602 }
1603
1604 /*
1605  * this function has three return values:
1606  *   negative: there was one error, and we have -errno.
1607  *   0 : We haven't finished, caller have to go again
1608  *   1 : We have finished, we can go to complete phase
1609  */
1610 int qemu_savevm_state_iterate(QEMUFile *f)
1611 {
1612     SaveStateEntry *se;
1613     int ret = 1;
1614
1615     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1616         if (!se->ops || !se->ops->save_live_iterate) {
1617             continue;
1618         }
1619         if (se->ops && se->ops->is_active) {
1620             if (!se->ops->is_active(se->opaque)) {
1621                 continue;
1622             }
1623         }
1624         if (qemu_file_rate_limit(f)) {
1625             return 0;
1626         }
1627         trace_savevm_section_start();
1628         /* Section type */
1629         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1630         qemu_put_be32(f, se->section_id);
1631
1632         ret = se->ops->save_live_iterate(f, se->opaque);
1633         trace_savevm_section_end(se->section_id);
1634
1635         if (ret <= 0) {
1636             /* Do not proceed to the next vmstate before this one reported
1637                completion of the current stage. This serializes the migration
1638                and reduces the probability that a faster changing state is
1639                synchronized over and over again. */
1640             break;
1641         }
1642     }
1643     if (ret != 0) {
1644         return ret;
1645     }
1646     ret = qemu_file_get_error(f);
1647     if (ret != 0) {
1648         qemu_savevm_state_cancel(f);
1649     }
1650     return ret;
1651 }
1652
1653 int qemu_savevm_state_complete(QEMUFile *f)
1654 {
1655     SaveStateEntry *se;
1656     int ret;
1657
1658     cpu_synchronize_all_states();
1659
1660     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1661         if (!se->ops || !se->ops->save_live_complete) {
1662             continue;
1663         }
1664         if (se->ops && se->ops->is_active) {
1665             if (!se->ops->is_active(se->opaque)) {
1666                 continue;
1667             }
1668         }
1669         trace_savevm_section_start();
1670         /* Section type */
1671         qemu_put_byte(f, QEMU_VM_SECTION_END);
1672         qemu_put_be32(f, se->section_id);
1673
1674         ret = se->ops->save_live_complete(f, se->opaque);
1675         trace_savevm_section_end(se->section_id);
1676         if (ret < 0) {
1677             return ret;
1678         }
1679     }
1680
1681     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1682         int len;
1683
1684         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1685             continue;
1686         }
1687         trace_savevm_section_start();
1688         /* Section type */
1689         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1690         qemu_put_be32(f, se->section_id);
1691
1692         /* ID string */
1693         len = strlen(se->idstr);
1694         qemu_put_byte(f, len);
1695         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1696
1697         qemu_put_be32(f, se->instance_id);
1698         qemu_put_be32(f, se->version_id);
1699
1700         vmstate_save(f, se);
1701         trace_savevm_section_end(se->section_id);
1702     }
1703
1704     qemu_put_byte(f, QEMU_VM_EOF);
1705
1706     return qemu_file_get_error(f);
1707 }
1708
1709 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
1710 {
1711     SaveStateEntry *se;
1712     uint64_t ret = 0;
1713
1714     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1715         if (!se->ops || !se->ops->save_live_pending) {
1716             continue;
1717         }
1718         if (se->ops && se->ops->is_active) {
1719             if (!se->ops->is_active(se->opaque)) {
1720                 continue;
1721             }
1722         }
1723         ret += se->ops->save_live_pending(f, se->opaque, max_size);
1724     }
1725     return ret;
1726 }
1727
1728 void qemu_savevm_state_cancel(QEMUFile *f)
1729 {
1730     SaveStateEntry *se;
1731
1732     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1733         if (se->ops && se->ops->cancel) {
1734             se->ops->cancel(se->opaque);
1735         }
1736     }
1737 }
1738
1739 static int qemu_savevm_state(QEMUFile *f)
1740 {
1741     int ret;
1742     MigrationParams params = {
1743         .blk = 0,
1744         .shared = 0
1745     };
1746
1747     if (qemu_savevm_state_blocked(NULL)) {
1748         ret = -EINVAL;
1749         goto out;
1750     }
1751
1752     ret = qemu_savevm_state_begin(f, &params);
1753     if (ret < 0)
1754         goto out;
1755
1756     do {
1757         ret = qemu_savevm_state_iterate(f);
1758         if (ret < 0)
1759             goto out;
1760     } while (ret == 0);
1761
1762     ret = qemu_savevm_state_complete(f);
1763
1764 out:
1765     if (ret == 0) {
1766         ret = qemu_file_get_error(f);
1767     }
1768
1769     return ret;
1770 }
1771
1772 static int qemu_save_device_state(QEMUFile *f)
1773 {
1774     SaveStateEntry *se;
1775
1776     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1777     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1778
1779     cpu_synchronize_all_states();
1780
1781     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1782         int len;
1783
1784         if (se->is_ram) {
1785             continue;
1786         }
1787         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1788             continue;
1789         }
1790
1791         /* Section type */
1792         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1793         qemu_put_be32(f, se->section_id);
1794
1795         /* ID string */
1796         len = strlen(se->idstr);
1797         qemu_put_byte(f, len);
1798         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1799
1800         qemu_put_be32(f, se->instance_id);
1801         qemu_put_be32(f, se->version_id);
1802
1803         vmstate_save(f, se);
1804     }
1805
1806     qemu_put_byte(f, QEMU_VM_EOF);
1807
1808     return qemu_file_get_error(f);
1809 }
1810
1811 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1812 {
1813     SaveStateEntry *se;
1814
1815     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1816         if (!strcmp(se->idstr, idstr) &&
1817             (instance_id == se->instance_id ||
1818              instance_id == se->alias_id))
1819             return se;
1820         /* Migrating from an older version? */
1821         if (strstr(se->idstr, idstr) && se->compat) {
1822             if (!strcmp(se->compat->idstr, idstr) &&
1823                 (instance_id == se->compat->instance_id ||
1824                  instance_id == se->alias_id))
1825                 return se;
1826         }
1827     }
1828     return NULL;
1829 }
1830
1831 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1832 {
1833     while(sub && sub->needed) {
1834         if (strcmp(idstr, sub->vmsd->name) == 0) {
1835             return sub->vmsd;
1836         }
1837         sub++;
1838     }
1839     return NULL;
1840 }
1841
1842 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1843                                    void *opaque)
1844 {
1845     while (qemu_peek_byte(f, 0) == QEMU_VM_SUBSECTION) {
1846         char idstr[256];
1847         int ret;
1848         uint8_t version_id, len, size;
1849         const VMStateDescription *sub_vmsd;
1850
1851         len = qemu_peek_byte(f, 1);
1852         if (len < strlen(vmsd->name) + 1) {
1853             /* subsection name has be be "section_name/a" */
1854             return 0;
1855         }
1856         size = qemu_peek_buffer(f, (uint8_t *)idstr, len, 2);
1857         if (size != len) {
1858             return 0;
1859         }
1860         idstr[size] = 0;
1861
1862         if (strncmp(vmsd->name, idstr, strlen(vmsd->name)) != 0) {
1863             /* it don't have a valid subsection name */
1864             return 0;
1865         }
1866         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
1867         if (sub_vmsd == NULL) {
1868             return -ENOENT;
1869         }
1870         qemu_file_skip(f, 1); /* subsection */
1871         qemu_file_skip(f, 1); /* len */
1872         qemu_file_skip(f, len); /* idstr */
1873         version_id = qemu_get_be32(f);
1874
1875         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1876         if (ret) {
1877             return ret;
1878         }
1879     }
1880     return 0;
1881 }
1882
1883 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1884                                     void *opaque)
1885 {
1886     const VMStateSubsection *sub = vmsd->subsections;
1887
1888     while (sub && sub->needed) {
1889         if (sub->needed(opaque)) {
1890             const VMStateDescription *vmsd = sub->vmsd;
1891             uint8_t len;
1892
1893             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1894             len = strlen(vmsd->name);
1895             qemu_put_byte(f, len);
1896             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1897             qemu_put_be32(f, vmsd->version_id);
1898             vmstate_save_state(f, vmsd, opaque);
1899         }
1900         sub++;
1901     }
1902 }
1903
1904 typedef struct LoadStateEntry {
1905     QLIST_ENTRY(LoadStateEntry) entry;
1906     SaveStateEntry *se;
1907     int section_id;
1908     int version_id;
1909 } LoadStateEntry;
1910
1911 int qemu_loadvm_state(QEMUFile *f)
1912 {
1913     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1914         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1915     LoadStateEntry *le, *new_le;
1916     uint8_t section_type;
1917     unsigned int v;
1918     int ret;
1919
1920     if (qemu_savevm_state_blocked(NULL)) {
1921         return -EINVAL;
1922     }
1923
1924     v = qemu_get_be32(f);
1925     if (v != QEMU_VM_FILE_MAGIC)
1926         return -EINVAL;
1927
1928     v = qemu_get_be32(f);
1929     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1930         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1931         return -ENOTSUP;
1932     }
1933     if (v != QEMU_VM_FILE_VERSION)
1934         return -ENOTSUP;
1935
1936     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1937         uint32_t instance_id, version_id, section_id;
1938         SaveStateEntry *se;
1939         char idstr[257];
1940         int len;
1941
1942         switch (section_type) {
1943         case QEMU_VM_SECTION_START:
1944         case QEMU_VM_SECTION_FULL:
1945             /* Read section start */
1946             section_id = qemu_get_be32(f);
1947             len = qemu_get_byte(f);
1948             qemu_get_buffer(f, (uint8_t *)idstr, len);
1949             idstr[len] = 0;
1950             instance_id = qemu_get_be32(f);
1951             version_id = qemu_get_be32(f);
1952
1953             /* Find savevm section */
1954             se = find_se(idstr, instance_id);
1955             if (se == NULL) {
1956                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1957                 ret = -EINVAL;
1958                 goto out;
1959             }
1960
1961             /* Validate version */
1962             if (version_id > se->version_id) {
1963                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
1964                         version_id, idstr, se->version_id);
1965                 ret = -EINVAL;
1966                 goto out;
1967             }
1968
1969             /* Add entry */
1970             le = g_malloc0(sizeof(*le));
1971
1972             le->se = se;
1973             le->section_id = section_id;
1974             le->version_id = version_id;
1975             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
1976
1977             ret = vmstate_load(f, le->se, le->version_id);
1978             if (ret < 0) {
1979                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
1980                         instance_id, idstr);
1981                 goto out;
1982             }
1983             break;
1984         case QEMU_VM_SECTION_PART:
1985         case QEMU_VM_SECTION_END:
1986             section_id = qemu_get_be32(f);
1987
1988             QLIST_FOREACH(le, &loadvm_handlers, entry) {
1989                 if (le->section_id == section_id) {
1990                     break;
1991                 }
1992             }
1993             if (le == NULL) {
1994                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
1995                 ret = -EINVAL;
1996                 goto out;
1997             }
1998
1999             ret = vmstate_load(f, le->se, le->version_id);
2000             if (ret < 0) {
2001                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
2002                         section_id);
2003                 goto out;
2004             }
2005             break;
2006         default:
2007             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
2008             ret = -EINVAL;
2009             goto out;
2010         }
2011     }
2012
2013     cpu_synchronize_all_post_init();
2014
2015     ret = 0;
2016
2017 out:
2018     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
2019         QLIST_REMOVE(le, entry);
2020         g_free(le);
2021     }
2022
2023     if (ret == 0) {
2024         ret = qemu_file_get_error(f);
2025     }
2026
2027     return ret;
2028 }
2029
2030 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
2031                               const char *name)
2032 {
2033     QEMUSnapshotInfo *sn_tab, *sn;
2034     int nb_sns, i, ret;
2035
2036     ret = -ENOENT;
2037     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2038     if (nb_sns < 0)
2039         return ret;
2040     for(i = 0; i < nb_sns; i++) {
2041         sn = &sn_tab[i];
2042         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
2043             *sn_info = *sn;
2044             ret = 0;
2045             break;
2046         }
2047     }
2048     g_free(sn_tab);
2049     return ret;
2050 }
2051
2052 /*
2053  * Deletes snapshots of a given name in all opened images.
2054  */
2055 static int del_existing_snapshots(Monitor *mon, const char *name)
2056 {
2057     BlockDriverState *bs;
2058     QEMUSnapshotInfo sn1, *snapshot = &sn1;
2059     int ret;
2060
2061     bs = NULL;
2062     while ((bs = bdrv_next(bs))) {
2063         if (bdrv_can_snapshot(bs) &&
2064             bdrv_snapshot_find(bs, snapshot, name) >= 0)
2065         {
2066             ret = bdrv_snapshot_delete(bs, name);
2067             if (ret < 0) {
2068                 monitor_printf(mon,
2069                                "Error while deleting snapshot on '%s'\n",
2070                                bdrv_get_device_name(bs));
2071                 return -1;
2072             }
2073         }
2074     }
2075
2076     return 0;
2077 }
2078
2079 void do_savevm(Monitor *mon, const QDict *qdict)
2080 {
2081     BlockDriverState *bs, *bs1;
2082     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
2083     int ret;
2084     QEMUFile *f;
2085     int saved_vm_running;
2086     uint64_t vm_state_size;
2087     qemu_timeval tv;
2088     struct tm tm;
2089     const char *name = qdict_get_try_str(qdict, "name");
2090
2091     /* Verify if there is a device that doesn't support snapshots and is writable */
2092     bs = NULL;
2093     while ((bs = bdrv_next(bs))) {
2094
2095         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2096             continue;
2097         }
2098
2099         if (!bdrv_can_snapshot(bs)) {
2100             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
2101                                bdrv_get_device_name(bs));
2102             return;
2103         }
2104     }
2105
2106     bs = bdrv_snapshots();
2107     if (!bs) {
2108         monitor_printf(mon, "No block device can accept snapshots\n");
2109         return;
2110     }
2111
2112     saved_vm_running = runstate_is_running();
2113     vm_stop(RUN_STATE_SAVE_VM);
2114
2115     memset(sn, 0, sizeof(*sn));
2116
2117     /* fill auxiliary fields */
2118     qemu_gettimeofday(&tv);
2119     sn->date_sec = tv.tv_sec;
2120     sn->date_nsec = tv.tv_usec * 1000;
2121     sn->vm_clock_nsec = qemu_get_clock_ns(vm_clock);
2122
2123     if (name) {
2124         ret = bdrv_snapshot_find(bs, old_sn, name);
2125         if (ret >= 0) {
2126             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
2127             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
2128         } else {
2129             pstrcpy(sn->name, sizeof(sn->name), name);
2130         }
2131     } else {
2132         /* cast below needed for OpenBSD where tv_sec is still 'long' */
2133         localtime_r((const time_t *)&tv.tv_sec, &tm);
2134         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
2135     }
2136
2137     /* Delete old snapshots of the same name */
2138     if (name && del_existing_snapshots(mon, name) < 0) {
2139         goto the_end;
2140     }
2141
2142     /* save the VM state */
2143     f = qemu_fopen_bdrv(bs, 1);
2144     if (!f) {
2145         monitor_printf(mon, "Could not open VM state file\n");
2146         goto the_end;
2147     }
2148     ret = qemu_savevm_state(f);
2149     vm_state_size = qemu_ftell(f);
2150     qemu_fclose(f);
2151     if (ret < 0) {
2152         monitor_printf(mon, "Error %d while writing VM\n", ret);
2153         goto the_end;
2154     }
2155
2156     /* create the snapshots */
2157
2158     bs1 = NULL;
2159     while ((bs1 = bdrv_next(bs1))) {
2160         if (bdrv_can_snapshot(bs1)) {
2161             /* Write VM state size only to the image that contains the state */
2162             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
2163             ret = bdrv_snapshot_create(bs1, sn);
2164             if (ret < 0) {
2165                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
2166                                bdrv_get_device_name(bs1));
2167             }
2168         }
2169     }
2170
2171  the_end:
2172     if (saved_vm_running)
2173         vm_start();
2174 }
2175
2176 void qmp_xen_save_devices_state(const char *filename, Error **errp)
2177 {
2178     QEMUFile *f;
2179     int saved_vm_running;
2180     int ret;
2181
2182     saved_vm_running = runstate_is_running();
2183     vm_stop(RUN_STATE_SAVE_VM);
2184
2185     f = qemu_fopen(filename, "wb");
2186     if (!f) {
2187         error_set(errp, QERR_OPEN_FILE_FAILED, filename);
2188         goto the_end;
2189     }
2190     ret = qemu_save_device_state(f);
2191     qemu_fclose(f);
2192     if (ret < 0) {
2193         error_set(errp, QERR_IO_ERROR);
2194     }
2195
2196  the_end:
2197     if (saved_vm_running)
2198         vm_start();
2199 }
2200
2201 int load_vmstate(const char *name)
2202 {
2203     BlockDriverState *bs, *bs_vm_state;
2204     QEMUSnapshotInfo sn;
2205     QEMUFile *f;
2206     int ret;
2207
2208     bs_vm_state = bdrv_snapshots();
2209     if (!bs_vm_state) {
2210         error_report("No block device supports snapshots");
2211         return -ENOTSUP;
2212     }
2213
2214     /* Don't even try to load empty VM states */
2215     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
2216     if (ret < 0) {
2217         return ret;
2218     } else if (sn.vm_state_size == 0) {
2219         error_report("This is a disk-only snapshot. Revert to it offline "
2220             "using qemu-img.");
2221         return -EINVAL;
2222     }
2223
2224     /* Verify if there is any device that doesn't support snapshots and is
2225     writable and check if the requested snapshot is available too. */
2226     bs = NULL;
2227     while ((bs = bdrv_next(bs))) {
2228
2229         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2230             continue;
2231         }
2232
2233         if (!bdrv_can_snapshot(bs)) {
2234             error_report("Device '%s' is writable but does not support snapshots.",
2235                                bdrv_get_device_name(bs));
2236             return -ENOTSUP;
2237         }
2238
2239         ret = bdrv_snapshot_find(bs, &sn, name);
2240         if (ret < 0) {
2241             error_report("Device '%s' does not have the requested snapshot '%s'",
2242                            bdrv_get_device_name(bs), name);
2243             return ret;
2244         }
2245     }
2246
2247     /* Flush all IO requests so they don't interfere with the new state.  */
2248     bdrv_drain_all();
2249
2250     bs = NULL;
2251     while ((bs = bdrv_next(bs))) {
2252         if (bdrv_can_snapshot(bs)) {
2253             ret = bdrv_snapshot_goto(bs, name);
2254             if (ret < 0) {
2255                 error_report("Error %d while activating snapshot '%s' on '%s'",
2256                              ret, name, bdrv_get_device_name(bs));
2257                 return ret;
2258             }
2259         }
2260     }
2261
2262     /* restore the VM state */
2263     f = qemu_fopen_bdrv(bs_vm_state, 0);
2264     if (!f) {
2265         error_report("Could not open VM state file");
2266         return -EINVAL;
2267     }
2268
2269     qemu_system_reset(VMRESET_SILENT);
2270     ret = qemu_loadvm_state(f);
2271
2272     qemu_fclose(f);
2273     if (ret < 0) {
2274         error_report("Error %d while loading VM state", ret);
2275         return ret;
2276     }
2277
2278     return 0;
2279 }
2280
2281 void do_delvm(Monitor *mon, const QDict *qdict)
2282 {
2283     BlockDriverState *bs, *bs1;
2284     int ret;
2285     const char *name = qdict_get_str(qdict, "name");
2286
2287     bs = bdrv_snapshots();
2288     if (!bs) {
2289         monitor_printf(mon, "No block device supports snapshots\n");
2290         return;
2291     }
2292
2293     bs1 = NULL;
2294     while ((bs1 = bdrv_next(bs1))) {
2295         if (bdrv_can_snapshot(bs1)) {
2296             ret = bdrv_snapshot_delete(bs1, name);
2297             if (ret < 0) {
2298                 if (ret == -ENOTSUP)
2299                     monitor_printf(mon,
2300                                    "Snapshots not supported on device '%s'\n",
2301                                    bdrv_get_device_name(bs1));
2302                 else
2303                     monitor_printf(mon, "Error %d while deleting snapshot on "
2304                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2305             }
2306         }
2307     }
2308 }
2309
2310 void do_info_snapshots(Monitor *mon, const QDict *qdict)
2311 {
2312     BlockDriverState *bs, *bs1;
2313     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2314     int nb_sns, i, ret, available;
2315     int total;
2316     int *available_snapshots;
2317     char buf[256];
2318
2319     bs = bdrv_snapshots();
2320     if (!bs) {
2321         monitor_printf(mon, "No available block device supports snapshots\n");
2322         return;
2323     }
2324
2325     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2326     if (nb_sns < 0) {
2327         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2328         return;
2329     }
2330
2331     if (nb_sns == 0) {
2332         monitor_printf(mon, "There is no snapshot available.\n");
2333         return;
2334     }
2335
2336     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
2337     total = 0;
2338     for (i = 0; i < nb_sns; i++) {
2339         sn = &sn_tab[i];
2340         available = 1;
2341         bs1 = NULL;
2342
2343         while ((bs1 = bdrv_next(bs1))) {
2344             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2345                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2346                 if (ret < 0) {
2347                     available = 0;
2348                     break;
2349                 }
2350             }
2351         }
2352
2353         if (available) {
2354             available_snapshots[total] = i;
2355             total++;
2356         }
2357     }
2358
2359     if (total > 0) {
2360         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2361         for (i = 0; i < total; i++) {
2362             sn = &sn_tab[available_snapshots[i]];
2363             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2364         }
2365     } else {
2366         monitor_printf(mon, "There is no suitable snapshot available\n");
2367     }
2368
2369     g_free(sn_tab);
2370     g_free(available_snapshots);
2371
2372 }
2373
2374 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
2375 {
2376     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
2377                        memory_region_name(mr), dev);
2378 }
2379
2380 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
2381 {
2382     /* Nothing do to while the implementation is in RAMBlock */
2383 }
2384
2385 void vmstate_register_ram_global(MemoryRegion *mr)
2386 {
2387     vmstate_register_ram(mr, NULL);
2388 }
2389
2390 /*
2391   page = zrun nzrun
2392        | zrun nzrun page
2393
2394   zrun = length
2395
2396   nzrun = length byte...
2397
2398   length = uleb128 encoded integer
2399  */
2400 int xbzrle_encode_buffer(uint8_t *old_buf, uint8_t *new_buf, int slen,
2401                          uint8_t *dst, int dlen)
2402 {
2403     uint32_t zrun_len = 0, nzrun_len = 0;
2404     int d = 0, i = 0;
2405     long res, xor;
2406     uint8_t *nzrun_start = NULL;
2407
2408     g_assert(!(((uintptr_t)old_buf | (uintptr_t)new_buf | slen) %
2409                sizeof(long)));
2410
2411     while (i < slen) {
2412         /* overflow */
2413         if (d + 2 > dlen) {
2414             return -1;
2415         }
2416
2417         /* not aligned to sizeof(long) */
2418         res = (slen - i) % sizeof(long);
2419         while (res && old_buf[i] == new_buf[i]) {
2420             zrun_len++;
2421             i++;
2422             res--;
2423         }
2424
2425         /* word at a time for speed */
2426         if (!res) {
2427             while (i < slen &&
2428                    (*(long *)(old_buf + i)) == (*(long *)(new_buf + i))) {
2429                 i += sizeof(long);
2430                 zrun_len += sizeof(long);
2431             }
2432
2433             /* go over the rest */
2434             while (i < slen && old_buf[i] == new_buf[i]) {
2435                 zrun_len++;
2436                 i++;
2437             }
2438         }
2439
2440         /* buffer unchanged */
2441         if (zrun_len == slen) {
2442             return 0;
2443         }
2444
2445         /* skip last zero run */
2446         if (i == slen) {
2447             return d;
2448         }
2449
2450         d += uleb128_encode_small(dst + d, zrun_len);
2451
2452         zrun_len = 0;
2453         nzrun_start = new_buf + i;
2454
2455         /* overflow */
2456         if (d + 2 > dlen) {
2457             return -1;
2458         }
2459         /* not aligned to sizeof(long) */
2460         res = (slen - i) % sizeof(long);
2461         while (res && old_buf[i] != new_buf[i]) {
2462             i++;
2463             nzrun_len++;
2464             res--;
2465         }
2466
2467         /* word at a time for speed, use of 32-bit long okay */
2468         if (!res) {
2469             /* truncation to 32-bit long okay */
2470             long mask = (long)0x0101010101010101ULL;
2471             while (i < slen) {
2472                 xor = *(long *)(old_buf + i) ^ *(long *)(new_buf + i);
2473                 if ((xor - mask) & ~xor & (mask << 7)) {
2474                     /* found the end of an nzrun within the current long */
2475                     while (old_buf[i] != new_buf[i]) {
2476                         nzrun_len++;
2477                         i++;
2478                     }
2479                     break;
2480                 } else {
2481                     i += sizeof(long);
2482                     nzrun_len += sizeof(long);
2483                 }
2484             }
2485         }
2486
2487         d += uleb128_encode_small(dst + d, nzrun_len);
2488         /* overflow */
2489         if (d + nzrun_len > dlen) {
2490             return -1;
2491         }
2492         memcpy(dst + d, nzrun_start, nzrun_len);
2493         d += nzrun_len;
2494         nzrun_len = 0;
2495     }
2496
2497     return d;
2498 }
2499
2500 int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen)
2501 {
2502     int i = 0, d = 0;
2503     int ret;
2504     uint32_t count = 0;
2505
2506     while (i < slen) {
2507
2508         /* zrun */
2509         if ((slen - i) < 2) {
2510             return -1;
2511         }
2512
2513         ret = uleb128_decode_small(src + i, &count);
2514         if (ret < 0 || (i && !count)) {
2515             return -1;
2516         }
2517         i += ret;
2518         d += count;
2519
2520         /* overflow */
2521         if (d > dlen) {
2522             return -1;
2523         }
2524
2525         /* nzrun */
2526         if ((slen - i) < 2) {
2527             return -1;
2528         }
2529
2530         ret = uleb128_decode_small(src + i, &count);
2531         if (ret < 0 || !count) {
2532             return -1;
2533         }
2534         i += ret;
2535
2536         /* overflow */
2537         if (d + count > dlen || i + count > slen) {
2538             return -1;
2539         }
2540
2541         memcpy(dst + d, src + i, count);
2542         d += count;
2543         i += count;
2544     }
2545
2546     return d;
2547 }