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
  27 #ifndef _WIN32
  28 #include <arpa/inet.h>
  29 #endif
  30
  31 #ifdef _WIN32
  32 #include <windows.h>
  33 #endif
  34
  35 #include "qemu-common.h"
  36 #include "hw/hw.h"
  37 #include "hw/qdev.h"
  38 #include "net/net.h"
  39 #include "monitor/monitor.h"
  40 #include "sysemu/sysemu.h"
  41 #include "qemu/timer.h"
  42 #include "audio/audio.h"
  43 #include "migration/migration.h"
  44 #include "qemu/sockets.h"
  45 #include "qemu/queue.h"
  46 #include "sysemu/cpus.h"
  47 #include "exec/memory.h"
  48 #include "qmp-commands.h"
  49 #include "trace.h"
  50 #include "qemu/bitops.h"
  51
  52 #define SELF_ANNOUNCE_ROUNDS 5
  53
  54 #ifndef ETH_P_RARP
  55 #define ETH_P_RARP 0x8035
  56 #endif
  57 #define ARP_HTYPE_ETH 0x0001
  58 #define ARP_PTYPE_IP 0x0800
  59 #define ARP_OP_REQUEST_REV 0x3
  60
  61 static int announce_self_create(uint8_t *buf,
  62                                 uint8_t *mac_addr)
  63 {
  64     /* Ethernet header. */
  65     memset(buf, 0xff, 6);         /* destination MAC addr */
  66     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
  67     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
  68
  69     /* RARP header. */
  70     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
  71     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
  72     *(buf + 18) = 6; /* hardware addr length (ethernet) */
  73     *(buf + 19) = 4; /* protocol addr length (IPv4) */
  74     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
  75     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
  76     memset(buf + 28, 0x00, 4);     /* source protocol addr */
  77     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
  78     memset(buf + 38, 0x00, 4);     /* target protocol addr */
  79
  80     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
  81     memset(buf + 42, 0x00, 18);
  82
  83     return 60; /* len (FCS will be added by hardware) */
  84 }
  85
  86 static void qemu_announce_self_iter(NICState *nic, void *opaque)
  87 {
  88     uint8_t buf[60];
  89     int len;
  90
  91     len = announce_self_create(buf, nic->conf->macaddr.a);
  92
  93     qemu_send_packet_raw(&nic->nc, buf, len);
  94 }
  95
  96
  97 static void qemu_announce_self_once(void *opaque)
  98 {
  99     static int count = SELF_ANNOUNCE_ROUNDS;
 100     QEMUTimer *timer = *(QEMUTimer **)opaque;
 101
 102     qemu_foreach_nic(qemu_announce_self_iter, NULL);
 103
 104     if (--count) {
 105         /* delay 50ms, 150ms, 250ms, ... */
 106         qemu_mod_timer(timer, qemu_get_clock_ms(rt_clock) +
 107                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
 108     } else {
 109             qemu_del_timer(timer);
 110             qemu_free_timer(timer);
 111     }
 112 }
 113
 114 void qemu_announce_self(void)
 115 {
 116         static QEMUTimer *timer;
 117         timer = qemu_new_timer_ms(rt_clock, qemu_announce_self_once, &timer);
 118         qemu_announce_self_once(&timer);
 119 }
 120
 121 /***********************************************************/
 122 /* savevm/loadvm support */
 123
 124 #define IO_BUF_SIZE 32768
 125
 126 struct QEMUFile {
 127     const QEMUFileOps *ops;
 128     void *opaque;
 129     int is_write;
 130
 131     int64_t buf_offset; /* start of buffer when writing, end of buffer
 132                            when reading */
 133     int buf_index;
 134     int buf_size; /* 0 when writing */
 135     uint8_t buf[IO_BUF_SIZE];
 136
 137     int last_error;
 138 };
 139
 140 typedef struct QEMUFileStdio
 141 {
 142     FILE *stdio_file;
 143     QEMUFile *file;
 144 } QEMUFileStdio;
 145
 146 typedef struct QEMUFileSocket
 147 {
 148     int fd;
 149     QEMUFile *file;
 150 } QEMUFileSocket;
 151
 152 static int socket_get_fd(void *opaque)
 153 {
 154     QEMUFileSocket *s = opaque;
 155
 156     return s->fd;
 157 }
 158
 159 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 160 {
 161     QEMUFileSocket *s = opaque;
 162     ssize_t len;
 163
 164     for (;;) {
 165         len = qemu_recv(s->fd, buf, size, 0);
 166         if (len != -1) {
 167             break;
 168         }
 169         if (socket_error() == EAGAIN) {
 170             assert(qemu_in_coroutine());
 171             qemu_coroutine_yield();
 172         } else if (socket_error() != EINTR) {
 173             break;
 174         }
 175     }
 176
 177     if (len == -1) {
 178         len = -socket_error();
 179     }
 180     return len;
 181 }
 182
 183 static int socket_close(void *opaque)
 184 {
 185     QEMUFileSocket *s = opaque;
 186     closesocket(s->fd);
 187     g_free(s);
 188     return 0;
 189 }
 190
 191 static int stdio_get_fd(void *opaque)
 192 {
 193     QEMUFileStdio *s = opaque;
 194
 195     return fileno(s->stdio_file);
 196 }
 197
 198 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 199 {
 200     QEMUFileStdio *s = opaque;
 201     return fwrite(buf, 1, size, s->stdio_file);
 202 }
 203
 204 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 205 {
 206     QEMUFileStdio *s = opaque;
 207     FILE *fp = s->stdio_file;
 208     int bytes;
 209
 210     for (;;) {
 211         clearerr(fp);
 212         bytes = fread(buf, 1, size, fp);
 213         if (bytes != 0 || !ferror(fp)) {
 214             break;
 215         }
 216         if (errno == EAGAIN) {
 217             assert(qemu_in_coroutine());
 218             qemu_coroutine_yield();
 219         } else if (errno != EINTR) {
 220             break;
 221         }
 222     }
 223     return bytes;
 224 }
 225
 226 static int stdio_pclose(void *opaque)
 227 {
 228     QEMUFileStdio *s = opaque;
 229     int ret;
 230     ret = pclose(s->stdio_file);
 231     if (ret == -1) {
 232         ret = -errno;
 233     }
 234     g_free(s);
 235     return ret;
 236 }
 237
 238 static int stdio_fclose(void *opaque)
 239 {
 240     QEMUFileStdio *s = opaque;
 241     int ret = 0;
 242     if (fclose(s->stdio_file) == EOF) {
 243         ret = -errno;
 244     }
 245     g_free(s);
 246     return ret;
 247 }
 248
 249 static const QEMUFileOps stdio_pipe_read_ops = {
 250     .get_fd =     stdio_get_fd,
 251     .get_buffer = stdio_get_buffer,
 252     .close =      stdio_pclose
 253 };
 254
 255 static const QEMUFileOps stdio_pipe_write_ops = {
 256     .get_fd =     stdio_get_fd,
 257     .put_buffer = stdio_put_buffer,
 258     .close =      stdio_pclose
 259 };
 260
 261 QEMUFile *qemu_popen(FILE *stdio_file, const char *mode)
 262 {
 263     QEMUFileStdio *s;
 264
 265     if (stdio_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 266         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 267         return NULL;
 268     }
 269
 270     s = g_malloc0(sizeof(QEMUFileStdio));
 271
 272     s->stdio_file = stdio_file;
 273
 274     if(mode[0] == 'r') {
 275         s->file = qemu_fopen_ops(s, &stdio_pipe_read_ops);
 276     } else {
 277         s->file = qemu_fopen_ops(s, &stdio_pipe_write_ops);
 278     }
 279     return s->file;
 280 }
 281
 282 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 283 {
 284     FILE *popen_file;
 285
 286     popen_file = popen(command, mode);
 287     if(popen_file == NULL) {
 288         return NULL;
 289     }
 290
 291     return qemu_popen(popen_file, mode);
 292 }
 293
 294 static const QEMUFileOps stdio_file_read_ops = {
 295     .get_fd =     stdio_get_fd,
 296     .get_buffer = stdio_get_buffer,
 297     .close =      stdio_fclose
 298 };
 299
 300 static const QEMUFileOps stdio_file_write_ops = {
 301     .get_fd =     stdio_get_fd,
 302     .put_buffer = stdio_put_buffer,
 303     .close =      stdio_fclose
 304 };
 305
 306 QEMUFile *qemu_fdopen(int fd, const char *mode)
 307 {
 308     QEMUFileStdio *s;
 309
 310     if (mode == NULL ||
 311         (mode[0] != 'r' && mode[0] != 'w') ||
 312         mode[1] != 'b' || mode[2] != 0) {
 313         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 314         return NULL;
 315     }
 316
 317     s = g_malloc0(sizeof(QEMUFileStdio));
 318     s->stdio_file = fdopen(fd, mode);
 319     if (!s->stdio_file)
 320         goto fail;
 321
 322     if(mode[0] == 'r') {
 323         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 324     } else {
 325         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 326     }
 327     return s->file;
 328
 329 fail:
 330     g_free(s);
 331     return NULL;
 332 }
 333
 334 static const QEMUFileOps socket_read_ops = {
 335     .get_fd =     socket_get_fd,
 336     .get_buffer = socket_get_buffer,
 337     .close =      socket_close
 338 };
 339
 340 QEMUFile *qemu_fopen_socket(int fd)
 341 {
 342     QEMUFileSocket *s = g_malloc0(sizeof(QEMUFileSocket));
 343
 344     s->fd = fd;
 345     s->file = qemu_fopen_ops(s, &socket_read_ops);
 346     return s->file;
 347 }
 348
 349 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 350 {
 351     QEMUFileStdio *s;
 352
 353     if (mode == NULL ||
 354         (mode[0] != 'r' && mode[0] != 'w') ||
 355         mode[1] != 'b' || mode[2] != 0) {
 356         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 357         return NULL;
 358     }
 359
 360     s = g_malloc0(sizeof(QEMUFileStdio));
 361
 362     s->stdio_file = fopen(filename, mode);
 363     if (!s->stdio_file)
 364         goto fail;
 365
 366     if(mode[0] == 'w') {
 367         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 368     } else {
 369         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 370     }
 371     return s->file;
 372 fail:
 373     g_free(s);
 374     return NULL;
 375 }
 376
 377 static int block_put_buffer(void *opaque, const uint8_t *buf,
 378                            int64_t pos, int size)
 379 {
 380     bdrv_save_vmstate(opaque, buf, pos, size);
 381     return size;
 382 }
 383
 384 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 385 {
 386     return bdrv_load_vmstate(opaque, buf, pos, size);
 387 }
 388
 389 static int bdrv_fclose(void *opaque)
 390 {
 391     return bdrv_flush(opaque);
 392 }
 393
 394 static const QEMUFileOps bdrv_read_ops = {
 395     .get_buffer = block_get_buffer,
 396     .close =      bdrv_fclose
 397 };
 398
 399 static const QEMUFileOps bdrv_write_ops = {
 400     .put_buffer = block_put_buffer,
 401     .close =      bdrv_fclose
 402 };
 403
 404 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 405 {
 406     if (is_writable)
 407         return qemu_fopen_ops(bs, &bdrv_write_ops);
 408     return qemu_fopen_ops(bs, &bdrv_read_ops);
 409 }
 410
 411 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
 412 {
 413     QEMUFile *f;
 414
 415     f = g_malloc0(sizeof(QEMUFile));
 416
 417     f->opaque = opaque;
 418     f->ops = ops;
 419     f->is_write = 0;
 420
 421     return f;
 422 }
 423
 424 int qemu_file_get_error(QEMUFile *f)
 425 {
 426     return f->last_error;
 427 }
 428
 429 static void qemu_file_set_error(QEMUFile *f, int ret)
 430 {
 431     f->last_error = ret;
 432 }
 433
 434 /** Flushes QEMUFile buffer
 435  *
 436  */
 437 static int qemu_fflush(QEMUFile *f)
 438 {
 439     int ret = 0;
 440
 441     if (!f->ops->put_buffer)
 442         return 0;
 443
 444     if (f->is_write && f->buf_index > 0) {
 445         ret = f->ops->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 446         if (ret >= 0) {
 447             f->buf_offset += f->buf_index;
 448         }
 449         f->buf_index = 0;
 450     }
 451     return ret;
 452 }
 453
 454 static void qemu_fill_buffer(QEMUFile *f)
 455 {
 456     int len;
 457     int pending;
 458
 459     if (!f->ops->get_buffer)
 460         return;
 461
 462     if (f->is_write)
 463         abort();
 464
 465     pending = f->buf_size - f->buf_index;
 466     if (pending > 0) {
 467         memmove(f->buf, f->buf + f->buf_index, pending);
 468     }
 469     f->buf_index = 0;
 470     f->buf_size = pending;
 471
 472     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->buf_offset,
 473                         IO_BUF_SIZE - pending);
 474     if (len > 0) {
 475         f->buf_size += len;
 476         f->buf_offset += len;
 477     } else if (len == 0) {
 478         qemu_file_set_error(f, -EIO);
 479     } else if (len != -EAGAIN)
 480         qemu_file_set_error(f, len);
 481 }
 482
 483 int qemu_get_fd(QEMUFile *f)
 484 {
 485     if (f->ops->get_fd) {
 486         return f->ops->get_fd(f->opaque);
 487     }
 488     return -1;
 489 }
 490
 491 /** Closes the file
 492  *
 493  * Returns negative error value if any error happened on previous operations or
 494  * while closing the file. Returns 0 or positive number on success.
 495  *
 496  * The meaning of return value on success depends on the specific backend
 497  * being used.
 498  */
 499 int qemu_fclose(QEMUFile *f)
 500 {
 501     int ret;
 502     ret = qemu_fflush(f);
 503
 504     if (f->ops->close) {
 505         int ret2 = f->ops->close(f->opaque);
 506         if (ret >= 0) {
 507             ret = ret2;
 508         }
 509     }
 510     /* If any error was spotted before closing, we should report it
 511      * instead of the close() return value.
 512      */
 513     if (f->last_error) {
 514         ret = f->last_error;
 515     }
 516     g_free(f);
 517     return ret;
 518 }
 519
 520 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 521 {
 522     int l;
 523
 524     if (f->last_error) {
 525         return;
 526     }
 527
 528     if (f->is_write == 0 && f->buf_index > 0) {
 529         fprintf(stderr,
 530                 "Attempted to write to buffer while read buffer is not empty\n");
 531         abort();
 532     }
 533
 534     while (size > 0) {
 535         l = IO_BUF_SIZE - f->buf_index;
 536         if (l > size)
 537             l = size;
 538         memcpy(f->buf + f->buf_index, buf, l);
 539         f->is_write = 1;
 540         f->buf_index += l;
 541         buf += l;
 542         size -= l;
 543         if (f->buf_index >= IO_BUF_SIZE) {
 544             int ret = qemu_fflush(f);
 545             if (ret < 0) {
 546                 qemu_file_set_error(f, ret);
 547                 break;
 548             }
 549         }
 550     }
 551 }
 552
 553 void qemu_put_byte(QEMUFile *f, int v)
 554 {
 555     if (f->last_error) {
 556         return;
 557     }
 558
 559     if (f->is_write == 0 && f->buf_index > 0) {
 560         fprintf(stderr,
 561                 "Attempted to write to buffer while read buffer is not empty\n");
 562         abort();
 563     }
 564
 565     f->buf[f->buf_index++] = v;
 566     f->is_write = 1;
 567     if (f->buf_index >= IO_BUF_SIZE) {
 568         int ret = qemu_fflush(f);
 569         if (ret < 0) {
 570             qemu_file_set_error(f, ret);
 571         }
 572     }
 573 }
 574
 575 static void qemu_file_skip(QEMUFile *f, int size)
 576 {
 577     if (f->buf_index + size <= f->buf_size) {
 578         f->buf_index += size;
 579     }
 580 }
 581
 582 static int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset)
 583 {
 584     int pending;
 585     int index;
 586
 587     if (f->is_write) {
 588         abort();
 589     }
 590
 591     index = f->buf_index + offset;
 592     pending = f->buf_size - index;
 593     if (pending < size) {
 594         qemu_fill_buffer(f);
 595         index = f->buf_index + offset;
 596         pending = f->buf_size - index;
 597     }
 598
 599     if (pending <= 0) {
 600         return 0;
 601     }
 602     if (size > pending) {
 603         size = pending;
 604     }
 605
 606     memcpy(buf, f->buf + index, size);
 607     return size;
 608 }
 609
 610 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size)
 611 {
 612     int pending = size;
 613     int done = 0;
 614
 615     while (pending > 0) {
 616         int res;
 617
 618         res = qemu_peek_buffer(f, buf, pending, 0);
 619         if (res == 0) {
 620             return done;
 621         }
 622         qemu_file_skip(f, res);
 623         buf += res;
 624         pending -= res;
 625         done += res;
 626     }
 627     return done;
 628 }
 629
 630 static int qemu_peek_byte(QEMUFile *f, int offset)
 631 {
 632     int index = f->buf_index + offset;
 633
 634     if (f->is_write) {
 635         abort();
 636     }
 637
 638     if (index >= f->buf_size) {
 639         qemu_fill_buffer(f);
 640         index = f->buf_index + offset;
 641         if (index >= f->buf_size) {
 642             return 0;
 643         }
 644     }
 645     return f->buf[index];
 646 }
 647
 648 int qemu_get_byte(QEMUFile *f)
 649 {
 650     int result;
 651
 652     result = qemu_peek_byte(f, 0);
 653     qemu_file_skip(f, 1);
 654     return result;
 655 }
 656
 657 static int64_t qemu_ftell(QEMUFile *f)
 658 {
 659     return f->buf_offset - f->buf_size + f->buf_index;
 660 }
 661
 662 int qemu_file_rate_limit(QEMUFile *f)
 663 {
 664     if (f->ops->rate_limit)
 665         return f->ops->rate_limit(f->opaque);
 666
 667     return 0;
 668 }
 669
 670 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 671 {
 672     if (f->ops->get_rate_limit)
 673         return f->ops->get_rate_limit(f->opaque);
 674
 675     return 0;
 676 }
 677
 678 int64_t qemu_file_set_rate_limit(QEMUFile *f, int64_t new_rate)
 679 {
 680     /* any failed or completed migration keeps its state to allow probing of
 681      * migration data, but has no associated file anymore */
 682     if (f && f->ops->set_rate_limit)
 683         return f->ops->set_rate_limit(f->opaque, new_rate);
 684
 685     return 0;
 686 }
 687
 688 void qemu_put_be16(QEMUFile *f, unsigned int v)
 689 {
 690     qemu_put_byte(f, v >> 8);
 691     qemu_put_byte(f, v);
 692 }
 693
 694 void qemu_put_be32(QEMUFile *f, unsigned int v)
 695 {
 696     qemu_put_byte(f, v >> 24);
 697     qemu_put_byte(f, v >> 16);
 698     qemu_put_byte(f, v >> 8);
 699     qemu_put_byte(f, v);
 700 }
 701
 702 void qemu_put_be64(QEMUFile *f, uint64_t v)
 703 {
 704     qemu_put_be32(f, v >> 32);
 705     qemu_put_be32(f, v);
 706 }
 707
 708 unsigned int qemu_get_be16(QEMUFile *f)
 709 {
 710     unsigned int v;
 711     v = qemu_get_byte(f) << 8;
 712     v |= qemu_get_byte(f);
 713     return v;
 714 }
 715
 716 unsigned int qemu_get_be32(QEMUFile *f)
 717 {
 718     unsigned int v;
 719     v = qemu_get_byte(f) << 24;
 720     v |= qemu_get_byte(f) << 16;
 721     v |= qemu_get_byte(f) << 8;
 722     v |= qemu_get_byte(f);
 723     return v;
 724 }
 725
 726 uint64_t qemu_get_be64(QEMUFile *f)
 727 {
 728     uint64_t v;
 729     v = (uint64_t)qemu_get_be32(f) << 32;
 730     v |= qemu_get_be32(f);
 731     return v;
 732 }
 733
 734
 735 /* timer */
 736
 737 void qemu_put_timer(QEMUFile *f, QEMUTimer *ts)
 738 {
 739     uint64_t expire_time;
 740
 741     expire_time = qemu_timer_expire_time_ns(ts);
 742     qemu_put_be64(f, expire_time);
 743 }
 744
 745 void qemu_get_timer(QEMUFile *f, QEMUTimer *ts)
 746 {
 747     uint64_t expire_time;
 748
 749     expire_time = qemu_get_be64(f);
 750     if (expire_time != -1) {
 751         qemu_mod_timer_ns(ts, expire_time);
 752     } else {
 753         qemu_del_timer(ts);
 754     }
 755 }
 756
 757
 758 /* bool */
 759
 760 static int get_bool(QEMUFile *f, void *pv, size_t size)
 761 {
 762     bool *v = pv;
 763     *v = qemu_get_byte(f);
 764     return 0;
 765 }
 766
 767 static void put_bool(QEMUFile *f, void *pv, size_t size)
 768 {
 769     bool *v = pv;
 770     qemu_put_byte(f, *v);
 771 }
 772
 773 const VMStateInfo vmstate_info_bool = {
 774     .name = "bool",
 775     .get  = get_bool,
 776     .put  = put_bool,
 777 };
 778
 779 /* 8 bit int */
 780
 781 static int get_int8(QEMUFile *f, void *pv, size_t size)
 782 {
 783     int8_t *v = pv;
 784     qemu_get_s8s(f, v);
 785     return 0;
 786 }
 787
 788 static void put_int8(QEMUFile *f, void *pv, size_t size)
 789 {
 790     int8_t *v = pv;
 791     qemu_put_s8s(f, v);
 792 }
 793
 794 const VMStateInfo vmstate_info_int8 = {
 795     .name = "int8",
 796     .get  = get_int8,
 797     .put  = put_int8,
 798 };
 799
 800 /* 16 bit int */
 801
 802 static int get_int16(QEMUFile *f, void *pv, size_t size)
 803 {
 804     int16_t *v = pv;
 805     qemu_get_sbe16s(f, v);
 806     return 0;
 807 }
 808
 809 static void put_int16(QEMUFile *f, void *pv, size_t size)
 810 {
 811     int16_t *v = pv;
 812     qemu_put_sbe16s(f, v);
 813 }
 814
 815 const VMStateInfo vmstate_info_int16 = {
 816     .name = "int16",
 817     .get  = get_int16,
 818     .put  = put_int16,
 819 };
 820
 821 /* 32 bit int */
 822
 823 static int get_int32(QEMUFile *f, void *pv, size_t size)
 824 {
 825     int32_t *v = pv;
 826     qemu_get_sbe32s(f, v);
 827     return 0;
 828 }
 829
 830 static void put_int32(QEMUFile *f, void *pv, size_t size)
 831 {
 832     int32_t *v = pv;
 833     qemu_put_sbe32s(f, v);
 834 }
 835
 836 const VMStateInfo vmstate_info_int32 = {
 837     .name = "int32",
 838     .get  = get_int32,
 839     .put  = put_int32,
 840 };
 841
 842 /* 32 bit int. See that the received value is the same than the one
 843    in the field */
 844
 845 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 846 {
 847     int32_t *v = pv;
 848     int32_t v2;
 849     qemu_get_sbe32s(f, &v2);
 850
 851     if (*v == v2)
 852         return 0;
 853     return -EINVAL;
 854 }
 855
 856 const VMStateInfo vmstate_info_int32_equal = {
 857     .name = "int32 equal",
 858     .get  = get_int32_equal,
 859     .put  = put_int32,
 860 };
 861
 862 /* 32 bit int. See that the received value is the less or the same
 863    than the one in the field */
 864
 865 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 866 {
 867     int32_t *old = pv;
 868     int32_t new;
 869     qemu_get_sbe32s(f, &new);
 870
 871     if (*old <= new)
 872         return 0;
 873     return -EINVAL;
 874 }
 875
 876 const VMStateInfo vmstate_info_int32_le = {
 877     .name = "int32 equal",
 878     .get  = get_int32_le,
 879     .put  = put_int32,
 880 };
 881
 882 /* 64 bit int */
 883
 884 static int get_int64(QEMUFile *f, void *pv, size_t size)
 885 {
 886     int64_t *v = pv;
 887     qemu_get_sbe64s(f, v);
 888     return 0;
 889 }
 890
 891 static void put_int64(QEMUFile *f, void *pv, size_t size)
 892 {
 893     int64_t *v = pv;
 894     qemu_put_sbe64s(f, v);
 895 }
 896
 897 const VMStateInfo vmstate_info_int64 = {
 898     .name = "int64",
 899     .get  = get_int64,
 900     .put  = put_int64,
 901 };
 902
 903 /* 8 bit unsigned int */
 904
 905 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 906 {
 907     uint8_t *v = pv;
 908     qemu_get_8s(f, v);
 909     return 0;
 910 }
 911
 912 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 913 {
 914     uint8_t *v = pv;
 915     qemu_put_8s(f, v);
 916 }
 917
 918 const VMStateInfo vmstate_info_uint8 = {
 919     .name = "uint8",
 920     .get  = get_uint8,
 921     .put  = put_uint8,
 922 };
 923
 924 /* 16 bit unsigned int */
 925
 926 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 927 {
 928     uint16_t *v = pv;
 929     qemu_get_be16s(f, v);
 930     return 0;
 931 }
 932
 933 static void put_uint16(QEMUFile *f, void *pv, size_t size)
 934 {
 935     uint16_t *v = pv;
 936     qemu_put_be16s(f, v);
 937 }
 938
 939 const VMStateInfo vmstate_info_uint16 = {
 940     .name = "uint16",
 941     .get  = get_uint16,
 942     .put  = put_uint16,
 943 };
 944
 945 /* 32 bit unsigned int */
 946
 947 static int get_uint32(QEMUFile *f, void *pv, size_t size)
 948 {
 949     uint32_t *v = pv;
 950     qemu_get_be32s(f, v);
 951     return 0;
 952 }
 953
 954 static void put_uint32(QEMUFile *f, void *pv, size_t size)
 955 {
 956     uint32_t *v = pv;
 957     qemu_put_be32s(f, v);
 958 }
 959
 960 const VMStateInfo vmstate_info_uint32 = {
 961     .name = "uint32",
 962     .get  = get_uint32,
 963     .put  = put_uint32,
 964 };
 965
 966 /* 32 bit uint. See that the received value is the same than the one
 967    in the field */
 968
 969 static int get_uint32_equal(QEMUFile *f, void *pv, size_t size)
 970 {
 971     uint32_t *v = pv;
 972     uint32_t v2;
 973     qemu_get_be32s(f, &v2);
 974
 975     if (*v == v2) {
 976         return 0;
 977     }
 978     return -EINVAL;
 979 }
 980
 981 const VMStateInfo vmstate_info_uint32_equal = {
 982     .name = "uint32 equal",
 983     .get  = get_uint32_equal,
 984     .put  = put_uint32,
 985 };
 986
 987 /* 64 bit unsigned int */
 988
 989 static int get_uint64(QEMUFile *f, void *pv, size_t size)
 990 {
 991     uint64_t *v = pv;
 992     qemu_get_be64s(f, v);
 993     return 0;
 994 }
 995
 996 static void put_uint64(QEMUFile *f, void *pv, size_t size)
 997 {
 998     uint64_t *v = pv;
 999     qemu_put_be64s(f, v);
1000 }
1001
1002 const VMStateInfo vmstate_info_uint64 = {
1003     .name = "uint64",
1004     .get  = get_uint64,
1005     .put  = put_uint64,
1006 };
1007
1008 /* 8 bit int. See that the received value is the same than the one
1009    in the field */
1010
1011 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
1012 {
1013     uint8_t *v = pv;
1014     uint8_t v2;
1015     qemu_get_8s(f, &v2);
1016
1017     if (*v == v2)
1018         return 0;
1019     return -EINVAL;
1020 }
1021
1022 const VMStateInfo vmstate_info_uint8_equal = {
1023     .name = "uint8 equal",
1024     .get  = get_uint8_equal,
1025     .put  = put_uint8,
1026 };
1027
1028 /* 16 bit unsigned int int. See that the received value is the same than the one
1029    in the field */
1030
1031 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
1032 {
1033     uint16_t *v = pv;
1034     uint16_t v2;
1035     qemu_get_be16s(f, &v2);
1036
1037     if (*v == v2)
1038         return 0;
1039     return -EINVAL;
1040 }
1041
1042 const VMStateInfo vmstate_info_uint16_equal = {
1043     .name = "uint16 equal",
1044     .get  = get_uint16_equal,
1045     .put  = put_uint16,
1046 };
1047
1048 /* timers  */
1049
1050 static int get_timer(QEMUFile *f, void *pv, size_t size)
1051 {
1052     QEMUTimer *v = pv;
1053     qemu_get_timer(f, v);
1054     return 0;
1055 }
1056
1057 static void put_timer(QEMUFile *f, void *pv, size_t size)
1058 {
1059     QEMUTimer *v = pv;
1060     qemu_put_timer(f, v);
1061 }
1062
1063 const VMStateInfo vmstate_info_timer = {
1064     .name = "timer",
1065     .get  = get_timer,
1066     .put  = put_timer,
1067 };
1068
1069 /* uint8_t buffers */
1070
1071 static int get_buffer(QEMUFile *f, void *pv, size_t size)
1072 {
1073     uint8_t *v = pv;
1074     qemu_get_buffer(f, v, size);
1075     return 0;
1076 }
1077
1078 static void put_buffer(QEMUFile *f, void *pv, size_t size)
1079 {
1080     uint8_t *v = pv;
1081     qemu_put_buffer(f, v, size);
1082 }
1083
1084 const VMStateInfo vmstate_info_buffer = {
1085     .name = "buffer",
1086     .get  = get_buffer,
1087     .put  = put_buffer,
1088 };
1089
1090 /* unused buffers: space that was used for some fields that are
1091    not useful anymore */
1092
1093 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
1094 {
1095     uint8_t buf[1024];
1096     int block_len;
1097
1098     while (size > 0) {
1099         block_len = MIN(sizeof(buf), size);
1100         size -= block_len;
1101         qemu_get_buffer(f, buf, block_len);
1102     }
1103    return 0;
1104 }
1105
1106 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1107 {
1108     static const uint8_t buf[1024];
1109     int block_len;
1110
1111     while (size > 0) {
1112         block_len = MIN(sizeof(buf), size);
1113         size -= block_len;
1114         qemu_put_buffer(f, buf, block_len);
1115     }
1116 }
1117
1118 const VMStateInfo vmstate_info_unused_buffer = {
1119     .name = "unused_buffer",
1120     .get  = get_unused_buffer,
1121     .put  = put_unused_buffer,
1122 };
1123
1124 /* bitmaps (as defined by bitmap.h). Note that size here is the size
1125  * of the bitmap in bits. The on-the-wire format of a bitmap is 64
1126  * bit words with the bits in big endian order. The in-memory format
1127  * is an array of 'unsigned long', which may be either 32 or 64 bits.
1128  */
1129 /* This is the number of 64 bit words sent over the wire */
1130 #define BITS_TO_U64S(nr) DIV_ROUND_UP(nr, 64)
1131 static int get_bitmap(QEMUFile *f, void *pv, size_t size)
1132 {
1133     unsigned long *bmp = pv;
1134     int i, idx = 0;
1135     for (i = 0; i < BITS_TO_U64S(size); i++) {
1136         uint64_t w = qemu_get_be64(f);
1137         bmp[idx++] = w;
1138         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1139             bmp[idx++] = w >> 32;
1140         }
1141     }
1142     return 0;
1143 }
1144
1145 static void put_bitmap(QEMUFile *f, void *pv, size_t size)
1146 {
1147     unsigned long *bmp = pv;
1148     int i, idx = 0;
1149     for (i = 0; i < BITS_TO_U64S(size); i++) {
1150         uint64_t w = bmp[idx++];
1151         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1152             w |= ((uint64_t)bmp[idx++]) << 32;
1153         }
1154         qemu_put_be64(f, w);
1155     }
1156 }
1157
1158 const VMStateInfo vmstate_info_bitmap = {
1159     .name = "bitmap",
1160     .get = get_bitmap,
1161     .put = put_bitmap,
1162 };
1163
1164 typedef struct CompatEntry {
1165     char idstr[256];
1166     int instance_id;
1167 } CompatEntry;
1168
1169 typedef struct SaveStateEntry {
1170     QTAILQ_ENTRY(SaveStateEntry) entry;
1171     char idstr[256];
1172     int instance_id;
1173     int alias_id;
1174     int version_id;
1175     int section_id;
1176     SaveVMHandlers *ops;
1177     const VMStateDescription *vmsd;
1178     void *opaque;
1179     CompatEntry *compat;
1180     int no_migrate;
1181     int is_ram;
1182 } SaveStateEntry;
1183
1184
1185 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1186     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1187 static int global_section_id;
1188
1189 static int calculate_new_instance_id(const char *idstr)
1190 {
1191     SaveStateEntry *se;
1192     int instance_id = 0;
1193
1194     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1195         if (strcmp(idstr, se->idstr) == 0
1196             && instance_id <= se->instance_id) {
1197             instance_id = se->instance_id + 1;
1198         }
1199     }
1200     return instance_id;
1201 }
1202
1203 static int calculate_compat_instance_id(const char *idstr)
1204 {
1205     SaveStateEntry *se;
1206     int instance_id = 0;
1207
1208     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1209         if (!se->compat)
1210             continue;
1211
1212         if (strcmp(idstr, se->compat->idstr) == 0
1213             && instance_id <= se->compat->instance_id) {
1214             instance_id = se->compat->instance_id + 1;
1215         }
1216     }
1217     return instance_id;
1218 }
1219
1220 /* TODO: Individual devices generally have very little idea about the rest
1221    of the system, so instance_id should be removed/replaced.
1222    Meanwhile pass -1 as instance_id if you do not already have a clearly
1223    distinguishing id for all instances of your device class. */
1224 int register_savevm_live(DeviceState *dev,
1225                          const char *idstr,
1226                          int instance_id,
1227                          int version_id,
1228                          SaveVMHandlers *ops,
1229                          void *opaque)
1230 {
1231     SaveStateEntry *se;
1232
1233     se = g_malloc0(sizeof(SaveStateEntry));
1234     se->version_id = version_id;
1235     se->section_id = global_section_id++;
1236     se->ops = ops;
1237     se->opaque = opaque;
1238     se->vmsd = NULL;
1239     se->no_migrate = 0;
1240     /* if this is a live_savem then set is_ram */
1241     if (ops->save_live_setup != NULL) {
1242         se->is_ram = 1;
1243     }
1244
1245     if (dev) {
1246         char *id = qdev_get_dev_path(dev);
1247         if (id) {
1248             pstrcpy(se->idstr, sizeof(se->idstr), id);
1249             pstrcat(se->idstr, sizeof(se->idstr), "/");
1250             g_free(id);
1251
1252             se->compat = g_malloc0(sizeof(CompatEntry));
1253             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1254             se->compat->instance_id = instance_id == -1 ?
1255                          calculate_compat_instance_id(idstr) : instance_id;
1256             instance_id = -1;
1257         }
1258     }
1259     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1260
1261     if (instance_id == -1) {
1262         se->instance_id = calculate_new_instance_id(se->idstr);
1263     } else {
1264         se->instance_id = instance_id;
1265     }
1266     assert(!se->compat || se->instance_id == 0);
1267     /* add at the end of list */
1268     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1269     return 0;
1270 }
1271
1272 int register_savevm(DeviceState *dev,
1273                     const char *idstr,
1274                     int instance_id,
1275                     int version_id,
1276                     SaveStateHandler *save_state,
1277                     LoadStateHandler *load_state,
1278                     void *opaque)
1279 {
1280     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
1281     ops->save_state = save_state;
1282     ops->load_state = load_state;
1283     return register_savevm_live(dev, idstr, instance_id, version_id,
1284                                 ops, opaque);
1285 }
1286
1287 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1288 {
1289     SaveStateEntry *se, *new_se;
1290     char id[256] = "";
1291
1292     if (dev) {
1293         char *path = qdev_get_dev_path(dev);
1294         if (path) {
1295             pstrcpy(id, sizeof(id), path);
1296             pstrcat(id, sizeof(id), "/");
1297             g_free(path);
1298         }
1299     }
1300     pstrcat(id, sizeof(id), idstr);
1301
1302     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1303         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1304             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1305             if (se->compat) {
1306                 g_free(se->compat);
1307             }
1308             g_free(se->ops);
1309             g_free(se);
1310         }
1311     }
1312 }
1313
1314 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1315                                    const VMStateDescription *vmsd,
1316                                    void *opaque, int alias_id,
1317                                    int required_for_version)
1318 {
1319     SaveStateEntry *se;
1320
1321     /* If this triggers, alias support can be dropped for the vmsd. */
1322     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1323
1324     se = g_malloc0(sizeof(SaveStateEntry));
1325     se->version_id = vmsd->version_id;
1326     se->section_id = global_section_id++;
1327     se->opaque = opaque;
1328     se->vmsd = vmsd;
1329     se->alias_id = alias_id;
1330     se->no_migrate = vmsd->unmigratable;
1331
1332     if (dev) {
1333         char *id = qdev_get_dev_path(dev);
1334         if (id) {
1335             pstrcpy(se->idstr, sizeof(se->idstr), id);
1336             pstrcat(se->idstr, sizeof(se->idstr), "/");
1337             g_free(id);
1338
1339             se->compat = g_malloc0(sizeof(CompatEntry));
1340             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1341             se->compat->instance_id = instance_id == -1 ?
1342                          calculate_compat_instance_id(vmsd->name) : instance_id;
1343             instance_id = -1;
1344         }
1345     }
1346     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1347
1348     if (instance_id == -1) {
1349         se->instance_id = calculate_new_instance_id(se->idstr);
1350     } else {
1351         se->instance_id = instance_id;
1352     }
1353     assert(!se->compat || se->instance_id == 0);
1354     /* add at the end of list */
1355     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1356     return 0;
1357 }
1358
1359 int vmstate_register(DeviceState *dev, int instance_id,
1360                      const VMStateDescription *vmsd, void *opaque)
1361 {
1362     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1363                                           opaque, -1, 0);
1364 }
1365
1366 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1367                         void *opaque)
1368 {
1369     SaveStateEntry *se, *new_se;
1370
1371     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1372         if (se->vmsd == vmsd && se->opaque == opaque) {
1373             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1374             if (se->compat) {
1375                 g_free(se->compat);
1376             }
1377             g_free(se);
1378         }
1379     }
1380 }
1381
1382 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1383                                     void *opaque);
1384 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1385                                    void *opaque);
1386
1387 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1388                        void *opaque, int version_id)
1389 {
1390     VMStateField *field = vmsd->fields;
1391     int ret;
1392
1393     if (version_id > vmsd->version_id) {
1394         return -EINVAL;
1395     }
1396     if (version_id < vmsd->minimum_version_id_old) {
1397         return -EINVAL;
1398     }
1399     if  (version_id < vmsd->minimum_version_id) {
1400         return vmsd->load_state_old(f, opaque, version_id);
1401     }
1402     if (vmsd->pre_load) {
1403         int ret = vmsd->pre_load(opaque);
1404         if (ret)
1405             return ret;
1406     }
1407     while(field->name) {
1408         if ((field->field_exists &&
1409              field->field_exists(opaque, version_id)) ||
1410             (!field->field_exists &&
1411              field->version_id <= version_id)) {
1412             void *base_addr = opaque + field->offset;
1413             int i, n_elems = 1;
1414             int size = field->size;
1415
1416             if (field->flags & VMS_VBUFFER) {
1417                 size = *(int32_t *)(opaque+field->size_offset);
1418                 if (field->flags & VMS_MULTIPLY) {
1419                     size *= field->size;
1420                 }
1421             }
1422             if (field->flags & VMS_ARRAY) {
1423                 n_elems = field->num;
1424             } else if (field->flags & VMS_VARRAY_INT32) {
1425                 n_elems = *(int32_t *)(opaque+field->num_offset);
1426             } else if (field->flags & VMS_VARRAY_UINT32) {
1427                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1428             } else if (field->flags & VMS_VARRAY_UINT16) {
1429                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1430             } else if (field->flags & VMS_VARRAY_UINT8) {
1431                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1432             }
1433             if (field->flags & VMS_POINTER) {
1434                 base_addr = *(void **)base_addr + field->start;
1435             }
1436             for (i = 0; i < n_elems; i++) {
1437                 void *addr = base_addr + size * i;
1438
1439                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1440                     addr = *(void **)addr;
1441                 }
1442                 if (field->flags & VMS_STRUCT) {
1443                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1444                 } else {
1445                     ret = field->info->get(f, addr, size);
1446
1447                 }
1448                 if (ret < 0) {
1449                     return ret;
1450                 }
1451             }
1452         }
1453         field++;
1454     }
1455     ret = vmstate_subsection_load(f, vmsd, opaque);
1456     if (ret != 0) {
1457         return ret;
1458     }
1459     if (vmsd->post_load) {
1460         return vmsd->post_load(opaque, version_id);
1461     }
1462     return 0;
1463 }
1464
1465 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1466                         void *opaque)
1467 {
1468     VMStateField *field = vmsd->fields;
1469
1470     if (vmsd->pre_save) {
1471         vmsd->pre_save(opaque);
1472     }
1473     while(field->name) {
1474         if (!field->field_exists ||
1475             field->field_exists(opaque, vmsd->version_id)) {
1476             void *base_addr = opaque + field->offset;
1477             int i, n_elems = 1;
1478             int size = field->size;
1479
1480             if (field->flags & VMS_VBUFFER) {
1481                 size = *(int32_t *)(opaque+field->size_offset);
1482                 if (field->flags & VMS_MULTIPLY) {
1483                     size *= field->size;
1484                 }
1485             }
1486             if (field->flags & VMS_ARRAY) {
1487                 n_elems = field->num;
1488             } else if (field->flags & VMS_VARRAY_INT32) {
1489                 n_elems = *(int32_t *)(opaque+field->num_offset);
1490             } else if (field->flags & VMS_VARRAY_UINT32) {
1491                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1492             } else if (field->flags & VMS_VARRAY_UINT16) {
1493                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1494             } else if (field->flags & VMS_VARRAY_UINT8) {
1495                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1496             }
1497             if (field->flags & VMS_POINTER) {
1498                 base_addr = *(void **)base_addr + field->start;
1499             }
1500             for (i = 0; i < n_elems; i++) {
1501                 void *addr = base_addr + size * i;
1502
1503                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1504                     addr = *(void **)addr;
1505                 }
1506                 if (field->flags & VMS_STRUCT) {
1507                     vmstate_save_state(f, field->vmsd, addr);
1508                 } else {
1509                     field->info->put(f, addr, size);
1510                 }
1511             }
1512         }
1513         field++;
1514     }
1515     vmstate_subsection_save(f, vmsd, opaque);
1516 }
1517
1518 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1519 {
1520     if (!se->vmsd) {         /* Old style */
1521         return se->ops->load_state(f, se->opaque, version_id);
1522     }
1523     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1524 }
1525
1526 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1527 {
1528     if (!se->vmsd) {         /* Old style */
1529         se->ops->save_state(f, se->opaque);
1530         return;
1531     }
1532     vmstate_save_state(f,se->vmsd, se->opaque);
1533 }
1534
1535 #define QEMU_VM_FILE_MAGIC           0x5145564d
1536 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1537 #define QEMU_VM_FILE_VERSION         0x00000003
1538
1539 #define QEMU_VM_EOF                  0x00
1540 #define QEMU_VM_SECTION_START        0x01
1541 #define QEMU_VM_SECTION_PART         0x02
1542 #define QEMU_VM_SECTION_END          0x03
1543 #define QEMU_VM_SECTION_FULL         0x04
1544 #define QEMU_VM_SUBSECTION           0x05
1545
1546 bool qemu_savevm_state_blocked(Error **errp)
1547 {
1548     SaveStateEntry *se;
1549
1550     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1551         if (se->no_migrate) {
1552             error_set(errp, QERR_MIGRATION_NOT_SUPPORTED, se->idstr);
1553             return true;
1554         }
1555     }
1556     return false;
1557 }
1558
1559 int qemu_savevm_state_begin(QEMUFile *f,
1560                             const MigrationParams *params)
1561 {
1562     SaveStateEntry *se;
1563     int ret;
1564
1565     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1566         if (!se->ops || !se->ops->set_params) {
1567             continue;
1568         }
1569         se->ops->set_params(params, se->opaque);
1570     }
1571
1572     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1573     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1574
1575     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1576         int len;
1577
1578         if (!se->ops || !se->ops->save_live_setup) {
1579             continue;
1580         }
1581         if (se->ops && se->ops->is_active) {
1582             if (!se->ops->is_active(se->opaque)) {
1583                 continue;
1584             }
1585         }
1586         /* Section type */
1587         qemu_put_byte(f, QEMU_VM_SECTION_START);
1588         qemu_put_be32(f, se->section_id);
1589
1590         /* ID string */
1591         len = strlen(se->idstr);
1592         qemu_put_byte(f, len);
1593         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1594
1595         qemu_put_be32(f, se->instance_id);
1596         qemu_put_be32(f, se->version_id);
1597
1598         ret = se->ops->save_live_setup(f, se->opaque);
1599         if (ret < 0) {
1600             qemu_savevm_state_cancel(f);
1601             return ret;
1602         }
1603     }
1604     ret = qemu_file_get_error(f);
1605     if (ret != 0) {
1606         qemu_savevm_state_cancel(f);
1607     }
1608
1609     return ret;
1610
1611 }
1612
1613 /*
1614  * this function has three return values:
1615  *   negative: there was one error, and we have -errno.
1616  *   0 : We haven't finished, caller have to go again
1617  *   1 : We have finished, we can go to complete phase
1618  */
1619 int qemu_savevm_state_iterate(QEMUFile *f)
1620 {
1621     SaveStateEntry *se;
1622     int ret = 1;
1623
1624     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1625         if (!se->ops || !se->ops->save_live_iterate) {
1626             continue;
1627         }
1628         if (se->ops && se->ops->is_active) {
1629             if (!se->ops->is_active(se->opaque)) {
1630                 continue;
1631             }
1632         }
1633         if (qemu_file_rate_limit(f)) {
1634             return 0;
1635         }
1636         trace_savevm_section_start();
1637         /* Section type */
1638         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1639         qemu_put_be32(f, se->section_id);
1640
1641         ret = se->ops->save_live_iterate(f, se->opaque);
1642         trace_savevm_section_end(se->section_id);
1643
1644         if (ret <= 0) {
1645             /* Do not proceed to the next vmstate before this one reported
1646                completion of the current stage. This serializes the migration
1647                and reduces the probability that a faster changing state is
1648                synchronized over and over again. */
1649             break;
1650         }
1651     }
1652     if (ret != 0) {
1653         return ret;
1654     }
1655     ret = qemu_file_get_error(f);
1656     if (ret != 0) {
1657         qemu_savevm_state_cancel(f);
1658     }
1659     return ret;
1660 }
1661
1662 int qemu_savevm_state_complete(QEMUFile *f)
1663 {
1664     SaveStateEntry *se;
1665     int ret;
1666
1667     cpu_synchronize_all_states();
1668
1669     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1670         if (!se->ops || !se->ops->save_live_complete) {
1671             continue;
1672         }
1673         if (se->ops && se->ops->is_active) {
1674             if (!se->ops->is_active(se->opaque)) {
1675                 continue;
1676             }
1677         }
1678         trace_savevm_section_start();
1679         /* Section type */
1680         qemu_put_byte(f, QEMU_VM_SECTION_END);
1681         qemu_put_be32(f, se->section_id);
1682
1683         ret = se->ops->save_live_complete(f, se->opaque);
1684         trace_savevm_section_end(se->section_id);
1685         if (ret < 0) {
1686             return ret;
1687         }
1688     }
1689
1690     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1691         int len;
1692
1693         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1694             continue;
1695         }
1696         trace_savevm_section_start();
1697         /* Section type */
1698         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1699         qemu_put_be32(f, se->section_id);
1700
1701         /* ID string */
1702         len = strlen(se->idstr);
1703         qemu_put_byte(f, len);
1704         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1705
1706         qemu_put_be32(f, se->instance_id);
1707         qemu_put_be32(f, se->version_id);
1708
1709         vmstate_save(f, se);
1710         trace_savevm_section_end(se->section_id);
1711     }
1712
1713     qemu_put_byte(f, QEMU_VM_EOF);
1714
1715     return qemu_file_get_error(f);
1716 }
1717
1718 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
1719 {
1720     SaveStateEntry *se;
1721     uint64_t ret = 0;
1722
1723     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1724         if (!se->ops || !se->ops->save_live_pending) {
1725             continue;
1726         }
1727         if (se->ops && se->ops->is_active) {
1728             if (!se->ops->is_active(se->opaque)) {
1729                 continue;
1730             }
1731         }
1732         ret += se->ops->save_live_pending(f, se->opaque, max_size);
1733     }
1734     return ret;
1735 }
1736
1737 void qemu_savevm_state_cancel(QEMUFile *f)
1738 {
1739     SaveStateEntry *se;
1740
1741     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1742         if (se->ops && se->ops->cancel) {
1743             se->ops->cancel(se->opaque);
1744         }
1745     }
1746 }
1747
1748 static int qemu_savevm_state(QEMUFile *f)
1749 {
1750     int ret;
1751     MigrationParams params = {
1752         .blk = 0,
1753         .shared = 0
1754     };
1755
1756     if (qemu_savevm_state_blocked(NULL)) {
1757         ret = -EINVAL;
1758         goto out;
1759     }
1760
1761     ret = qemu_savevm_state_begin(f, &params);
1762     if (ret < 0)
1763         goto out;
1764
1765     do {
1766         ret = qemu_savevm_state_iterate(f);
1767         if (ret < 0)
1768             goto out;
1769     } while (ret == 0);
1770
1771     ret = qemu_savevm_state_complete(f);
1772
1773 out:
1774     if (ret == 0) {
1775         ret = qemu_file_get_error(f);
1776     }
1777
1778     return ret;
1779 }
1780
1781 static int qemu_save_device_state(QEMUFile *f)
1782 {
1783     SaveStateEntry *se;
1784
1785     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1786     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1787
1788     cpu_synchronize_all_states();
1789
1790     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1791         int len;
1792
1793         if (se->is_ram) {
1794             continue;
1795         }
1796         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1797             continue;
1798         }
1799
1800         /* Section type */
1801         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1802         qemu_put_be32(f, se->section_id);
1803
1804         /* ID string */
1805         len = strlen(se->idstr);
1806         qemu_put_byte(f, len);
1807         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1808
1809         qemu_put_be32(f, se->instance_id);
1810         qemu_put_be32(f, se->version_id);
1811
1812         vmstate_save(f, se);
1813     }
1814
1815     qemu_put_byte(f, QEMU_VM_EOF);
1816
1817     return qemu_file_get_error(f);
1818 }
1819
1820 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1821 {
1822     SaveStateEntry *se;
1823
1824     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1825         if (!strcmp(se->idstr, idstr) &&
1826             (instance_id == se->instance_id ||
1827              instance_id == se->alias_id))
1828             return se;
1829         /* Migrating from an older version? */
1830         if (strstr(se->idstr, idstr) && se->compat) {
1831             if (!strcmp(se->compat->idstr, idstr) &&
1832                 (instance_id == se->compat->instance_id ||
1833                  instance_id == se->alias_id))
1834                 return se;
1835         }
1836     }
1837     return NULL;
1838 }
1839
1840 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1841 {
1842     while(sub && sub->needed) {
1843         if (strcmp(idstr, sub->vmsd->name) == 0) {
1844             return sub->vmsd;
1845         }
1846         sub++;
1847     }
1848     return NULL;
1849 }
1850
1851 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1852                                    void *opaque)
1853 {
1854     while (qemu_peek_byte(f, 0) == QEMU_VM_SUBSECTION) {
1855         char idstr[256];
1856         int ret;
1857         uint8_t version_id, len, size;
1858         const VMStateDescription *sub_vmsd;
1859
1860         len = qemu_peek_byte(f, 1);
1861         if (len < strlen(vmsd->name) + 1) {
1862             /* subsection name has be be "section_name/a" */
1863             return 0;
1864         }
1865         size = qemu_peek_buffer(f, (uint8_t *)idstr, len, 2);
1866         if (size != len) {
1867             return 0;
1868         }
1869         idstr[size] = 0;
1870
1871         if (strncmp(vmsd->name, idstr, strlen(vmsd->name)) != 0) {
1872             /* it don't have a valid subsection name */
1873             return 0;
1874         }
1875         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
1876         if (sub_vmsd == NULL) {
1877             return -ENOENT;
1878         }
1879         qemu_file_skip(f, 1); /* subsection */
1880         qemu_file_skip(f, 1); /* len */
1881         qemu_file_skip(f, len); /* idstr */
1882         version_id = qemu_get_be32(f);
1883
1884         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1885         if (ret) {
1886             return ret;
1887         }
1888     }
1889     return 0;
1890 }
1891
1892 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1893                                     void *opaque)
1894 {
1895     const VMStateSubsection *sub = vmsd->subsections;
1896
1897     while (sub && sub->needed) {
1898         if (sub->needed(opaque)) {
1899             const VMStateDescription *vmsd = sub->vmsd;
1900             uint8_t len;
1901
1902             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1903             len = strlen(vmsd->name);
1904             qemu_put_byte(f, len);
1905             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1906             qemu_put_be32(f, vmsd->version_id);
1907             vmstate_save_state(f, vmsd, opaque);
1908         }
1909         sub++;
1910     }
1911 }
1912
1913 typedef struct LoadStateEntry {
1914     QLIST_ENTRY(LoadStateEntry) entry;
1915     SaveStateEntry *se;
1916     int section_id;
1917     int version_id;
1918 } LoadStateEntry;
1919
1920 int qemu_loadvm_state(QEMUFile *f)
1921 {
1922     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1923         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1924     LoadStateEntry *le, *new_le;
1925     uint8_t section_type;
1926     unsigned int v;
1927     int ret;
1928
1929     if (qemu_savevm_state_blocked(NULL)) {
1930         return -EINVAL;
1931     }
1932
1933     v = qemu_get_be32(f);
1934     if (v != QEMU_VM_FILE_MAGIC)
1935         return -EINVAL;
1936
1937     v = qemu_get_be32(f);
1938     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1939         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1940         return -ENOTSUP;
1941     }
1942     if (v != QEMU_VM_FILE_VERSION)
1943         return -ENOTSUP;
1944
1945     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1946         uint32_t instance_id, version_id, section_id;
1947         SaveStateEntry *se;
1948         char idstr[257];
1949         int len;
1950
1951         switch (section_type) {
1952         case QEMU_VM_SECTION_START:
1953         case QEMU_VM_SECTION_FULL:
1954             /* Read section start */
1955             section_id = qemu_get_be32(f);
1956             len = qemu_get_byte(f);
1957             qemu_get_buffer(f, (uint8_t *)idstr, len);
1958             idstr[len] = 0;
1959             instance_id = qemu_get_be32(f);
1960             version_id = qemu_get_be32(f);
1961
1962             /* Find savevm section */
1963             se = find_se(idstr, instance_id);
1964             if (se == NULL) {
1965                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1966                 ret = -EINVAL;
1967                 goto out;
1968             }
1969
1970             /* Validate version */
1971             if (version_id > se->version_id) {
1972                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
1973                         version_id, idstr, se->version_id);
1974                 ret = -EINVAL;
1975                 goto out;
1976             }
1977
1978             /* Add entry */
1979             le = g_malloc0(sizeof(*le));
1980
1981             le->se = se;
1982             le->section_id = section_id;
1983             le->version_id = version_id;
1984             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
1985
1986             ret = vmstate_load(f, le->se, le->version_id);
1987             if (ret < 0) {
1988                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
1989                         instance_id, idstr);
1990                 goto out;
1991             }
1992             break;
1993         case QEMU_VM_SECTION_PART:
1994         case QEMU_VM_SECTION_END:
1995             section_id = qemu_get_be32(f);
1996
1997             QLIST_FOREACH(le, &loadvm_handlers, entry) {
1998                 if (le->section_id == section_id) {
1999                     break;
2000                 }
2001             }
2002             if (le == NULL) {
2003                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
2004                 ret = -EINVAL;
2005                 goto out;
2006             }
2007
2008             ret = vmstate_load(f, le->se, le->version_id);
2009             if (ret < 0) {
2010                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
2011                         section_id);
2012                 goto out;
2013             }
2014             break;
2015         default:
2016             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
2017             ret = -EINVAL;
2018             goto out;
2019         }
2020     }
2021
2022     cpu_synchronize_all_post_init();
2023
2024     ret = 0;
2025
2026 out:
2027     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
2028         QLIST_REMOVE(le, entry);
2029         g_free(le);
2030     }
2031
2032     if (ret == 0) {
2033         ret = qemu_file_get_error(f);
2034     }
2035
2036     return ret;
2037 }
2038
2039 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
2040                               const char *name)
2041 {
2042     QEMUSnapshotInfo *sn_tab, *sn;
2043     int nb_sns, i, ret;
2044
2045     ret = -ENOENT;
2046     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2047     if (nb_sns < 0)
2048         return ret;
2049     for(i = 0; i < nb_sns; i++) {
2050         sn = &sn_tab[i];
2051         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
2052             *sn_info = *sn;
2053             ret = 0;
2054             break;
2055         }
2056     }
2057     g_free(sn_tab);
2058     return ret;
2059 }
2060
2061 /*
2062  * Deletes snapshots of a given name in all opened images.
2063  */
2064 static int del_existing_snapshots(Monitor *mon, const char *name)
2065 {
2066     BlockDriverState *bs;
2067     QEMUSnapshotInfo sn1, *snapshot = &sn1;
2068     int ret;
2069
2070     bs = NULL;
2071     while ((bs = bdrv_next(bs))) {
2072         if (bdrv_can_snapshot(bs) &&
2073             bdrv_snapshot_find(bs, snapshot, name) >= 0)
2074         {
2075             ret = bdrv_snapshot_delete(bs, name);
2076             if (ret < 0) {
2077                 monitor_printf(mon,
2078                                "Error while deleting snapshot on '%s'\n",
2079                                bdrv_get_device_name(bs));
2080                 return -1;
2081             }
2082         }
2083     }
2084
2085     return 0;
2086 }
2087
2088 void do_savevm(Monitor *mon, const QDict *qdict)
2089 {
2090     BlockDriverState *bs, *bs1;
2091     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
2092     int ret;
2093     QEMUFile *f;
2094     int saved_vm_running;
2095     uint64_t vm_state_size;
2096 #ifdef _WIN32
2097     struct _timeb tb;
2098     struct tm *ptm;
2099 #else
2100     struct timeval tv;
2101     struct tm tm;
2102 #endif
2103     const char *name = qdict_get_try_str(qdict, "name");
2104
2105     /* Verify if there is a device that doesn't support snapshots and is writable */
2106     bs = NULL;
2107     while ((bs = bdrv_next(bs))) {
2108
2109         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2110             continue;
2111         }
2112
2113         if (!bdrv_can_snapshot(bs)) {
2114             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
2115                                bdrv_get_device_name(bs));
2116             return;
2117         }
2118     }
2119
2120     bs = bdrv_snapshots();
2121     if (!bs) {
2122         monitor_printf(mon, "No block device can accept snapshots\n");
2123         return;
2124     }
2125
2126     saved_vm_running = runstate_is_running();
2127     vm_stop(RUN_STATE_SAVE_VM);
2128
2129     memset(sn, 0, sizeof(*sn));
2130
2131     /* fill auxiliary fields */
2132 #ifdef _WIN32
2133     _ftime(&tb);
2134     sn->date_sec = tb.time;
2135     sn->date_nsec = tb.millitm * 1000000;
2136 #else
2137     gettimeofday(&tv, NULL);
2138     sn->date_sec = tv.tv_sec;
2139     sn->date_nsec = tv.tv_usec * 1000;
2140 #endif
2141     sn->vm_clock_nsec = qemu_get_clock_ns(vm_clock);
2142
2143     if (name) {
2144         ret = bdrv_snapshot_find(bs, old_sn, name);
2145         if (ret >= 0) {
2146             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
2147             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
2148         } else {
2149             pstrcpy(sn->name, sizeof(sn->name), name);
2150         }
2151     } else {
2152 #ifdef _WIN32
2153         time_t t = tb.time;
2154         ptm = localtime(&t);
2155         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", ptm);
2156 #else
2157         /* cast below needed for OpenBSD where tv_sec is still 'long' */
2158         localtime_r((const time_t *)&tv.tv_sec, &tm);
2159         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
2160 #endif
2161     }
2162
2163     /* Delete old snapshots of the same name */
2164     if (name && del_existing_snapshots(mon, name) < 0) {
2165         goto the_end;
2166     }
2167
2168     /* save the VM state */
2169     f = qemu_fopen_bdrv(bs, 1);
2170     if (!f) {
2171         monitor_printf(mon, "Could not open VM state file\n");
2172         goto the_end;
2173     }
2174     ret = qemu_savevm_state(f);
2175     vm_state_size = qemu_ftell(f);
2176     qemu_fclose(f);
2177     if (ret < 0) {
2178         monitor_printf(mon, "Error %d while writing VM\n", ret);
2179         goto the_end;
2180     }
2181
2182     /* create the snapshots */
2183
2184     bs1 = NULL;
2185     while ((bs1 = bdrv_next(bs1))) {
2186         if (bdrv_can_snapshot(bs1)) {
2187             /* Write VM state size only to the image that contains the state */
2188             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
2189             ret = bdrv_snapshot_create(bs1, sn);
2190             if (ret < 0) {
2191                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
2192                                bdrv_get_device_name(bs1));
2193             }
2194         }
2195     }
2196
2197  the_end:
2198     if (saved_vm_running)
2199         vm_start();
2200 }
2201
2202 void qmp_xen_save_devices_state(const char *filename, Error **errp)
2203 {
2204     QEMUFile *f;
2205     int saved_vm_running;
2206     int ret;
2207
2208     saved_vm_running = runstate_is_running();
2209     vm_stop(RUN_STATE_SAVE_VM);
2210
2211     f = qemu_fopen(filename, "wb");
2212     if (!f) {
2213         error_set(errp, QERR_OPEN_FILE_FAILED, filename);
2214         goto the_end;
2215     }
2216     ret = qemu_save_device_state(f);
2217     qemu_fclose(f);
2218     if (ret < 0) {
2219         error_set(errp, QERR_IO_ERROR);
2220     }
2221
2222  the_end:
2223     if (saved_vm_running)
2224         vm_start();
2225 }
2226
2227 int load_vmstate(const char *name)
2228 {
2229     BlockDriverState *bs, *bs_vm_state;
2230     QEMUSnapshotInfo sn;
2231     QEMUFile *f;
2232     int ret;
2233
2234     bs_vm_state = bdrv_snapshots();
2235     if (!bs_vm_state) {
2236         error_report("No block device supports snapshots");
2237         return -ENOTSUP;
2238     }
2239
2240     /* Don't even try to load empty VM states */
2241     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
2242     if (ret < 0) {
2243         return ret;
2244     } else if (sn.vm_state_size == 0) {
2245         error_report("This is a disk-only snapshot. Revert to it offline "
2246             "using qemu-img.");
2247         return -EINVAL;
2248     }
2249
2250     /* Verify if there is any device that doesn't support snapshots and is
2251     writable and check if the requested snapshot is available too. */
2252     bs = NULL;
2253     while ((bs = bdrv_next(bs))) {
2254
2255         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2256             continue;
2257         }
2258
2259         if (!bdrv_can_snapshot(bs)) {
2260             error_report("Device '%s' is writable but does not support snapshots.",
2261                                bdrv_get_device_name(bs));
2262             return -ENOTSUP;
2263         }
2264
2265         ret = bdrv_snapshot_find(bs, &sn, name);
2266         if (ret < 0) {
2267             error_report("Device '%s' does not have the requested snapshot '%s'",
2268                            bdrv_get_device_name(bs), name);
2269             return ret;
2270         }
2271     }
2272
2273     /* Flush all IO requests so they don't interfere with the new state.  */
2274     bdrv_drain_all();
2275
2276     bs = NULL;
2277     while ((bs = bdrv_next(bs))) {
2278         if (bdrv_can_snapshot(bs)) {
2279             ret = bdrv_snapshot_goto(bs, name);
2280             if (ret < 0) {
2281                 error_report("Error %d while activating snapshot '%s' on '%s'",
2282                              ret, name, bdrv_get_device_name(bs));
2283                 return ret;
2284             }
2285         }
2286     }
2287
2288     /* restore the VM state */
2289     f = qemu_fopen_bdrv(bs_vm_state, 0);
2290     if (!f) {
2291         error_report("Could not open VM state file");
2292         return -EINVAL;
2293     }
2294
2295     qemu_system_reset(VMRESET_SILENT);
2296     ret = qemu_loadvm_state(f);
2297
2298     qemu_fclose(f);
2299     if (ret < 0) {
2300         error_report("Error %d while loading VM state", ret);
2301         return ret;
2302     }
2303
2304     return 0;
2305 }
2306
2307 void do_delvm(Monitor *mon, const QDict *qdict)
2308 {
2309     BlockDriverState *bs, *bs1;
2310     int ret;
2311     const char *name = qdict_get_str(qdict, "name");
2312
2313     bs = bdrv_snapshots();
2314     if (!bs) {
2315         monitor_printf(mon, "No block device supports snapshots\n");
2316         return;
2317     }
2318
2319     bs1 = NULL;
2320     while ((bs1 = bdrv_next(bs1))) {
2321         if (bdrv_can_snapshot(bs1)) {
2322             ret = bdrv_snapshot_delete(bs1, name);
2323             if (ret < 0) {
2324                 if (ret == -ENOTSUP)
2325                     monitor_printf(mon,
2326                                    "Snapshots not supported on device '%s'\n",
2327                                    bdrv_get_device_name(bs1));
2328                 else
2329                     monitor_printf(mon, "Error %d while deleting snapshot on "
2330                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2331             }
2332         }
2333     }
2334 }
2335
2336 void do_info_snapshots(Monitor *mon)
2337 {
2338     BlockDriverState *bs, *bs1;
2339     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2340     int nb_sns, i, ret, available;
2341     int total;
2342     int *available_snapshots;
2343     char buf[256];
2344
2345     bs = bdrv_snapshots();
2346     if (!bs) {
2347         monitor_printf(mon, "No available block device supports snapshots\n");
2348         return;
2349     }
2350
2351     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2352     if (nb_sns < 0) {
2353         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2354         return;
2355     }
2356
2357     if (nb_sns == 0) {
2358         monitor_printf(mon, "There is no snapshot available.\n");
2359         return;
2360     }
2361
2362     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
2363     total = 0;
2364     for (i = 0; i < nb_sns; i++) {
2365         sn = &sn_tab[i];
2366         available = 1;
2367         bs1 = NULL;
2368
2369         while ((bs1 = bdrv_next(bs1))) {
2370             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2371                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2372                 if (ret < 0) {
2373                     available = 0;
2374                     break;
2375                 }
2376             }
2377         }
2378
2379         if (available) {
2380             available_snapshots[total] = i;
2381             total++;
2382         }
2383     }
2384
2385     if (total > 0) {
2386         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2387         for (i = 0; i < total; i++) {
2388             sn = &sn_tab[available_snapshots[i]];
2389             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2390         }
2391     } else {
2392         monitor_printf(mon, "There is no suitable snapshot available\n");
2393     }
2394
2395     g_free(sn_tab);
2396     g_free(available_snapshots);
2397
2398 }
2399
2400 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
2401 {
2402     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
2403                        memory_region_name(mr), dev);
2404 }
2405
2406 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
2407 {
2408     /* Nothing do to while the implementation is in RAMBlock */
2409 }
2410
2411 void vmstate_register_ram_global(MemoryRegion *mr)
2412 {
2413     vmstate_register_ram(mr, NULL);
2414 }
2415
2416 /*
2417   page = zrun nzrun
2418        | zrun nzrun page
2419
2420   zrun = length
2421
2422   nzrun = length byte...
2423
2424   length = uleb128 encoded integer
2425  */
2426 int xbzrle_encode_buffer(uint8_t *old_buf, uint8_t *new_buf, int slen,
2427                          uint8_t *dst, int dlen)
2428 {
2429     uint32_t zrun_len = 0, nzrun_len = 0;
2430     int d = 0, i = 0;
2431     long res, xor;
2432     uint8_t *nzrun_start = NULL;
2433
2434     g_assert(!(((uintptr_t)old_buf | (uintptr_t)new_buf | slen) %
2435                sizeof(long)));
2436
2437     while (i < slen) {
2438         /* overflow */
2439         if (d + 2 > dlen) {
2440             return -1;
2441         }
2442
2443         /* not aligned to sizeof(long) */
2444         res = (slen - i) % sizeof(long);
2445         while (res && old_buf[i] == new_buf[i]) {
2446             zrun_len++;
2447             i++;
2448             res--;
2449         }
2450
2451         /* word at a time for speed */
2452         if (!res) {
2453             while (i < slen &&
2454                    (*(long *)(old_buf + i)) == (*(long *)(new_buf + i))) {
2455                 i += sizeof(long);
2456                 zrun_len += sizeof(long);
2457             }
2458
2459             /* go over the rest */
2460             while (i < slen && old_buf[i] == new_buf[i]) {
2461                 zrun_len++;
2462                 i++;
2463             }
2464         }
2465
2466         /* buffer unchanged */
2467         if (zrun_len == slen) {
2468             return 0;
2469         }
2470
2471         /* skip last zero run */
2472         if (i == slen) {
2473             return d;
2474         }
2475
2476         d += uleb128_encode_small(dst + d, zrun_len);
2477
2478         zrun_len = 0;
2479         nzrun_start = new_buf + i;
2480
2481         /* overflow */
2482         if (d + 2 > dlen) {
2483             return -1;
2484         }
2485         /* not aligned to sizeof(long) */
2486         res = (slen - i) % sizeof(long);
2487         while (res && old_buf[i] != new_buf[i]) {
2488             i++;
2489             nzrun_len++;
2490             res--;
2491         }
2492
2493         /* word at a time for speed, use of 32-bit long okay */
2494         if (!res) {
2495             /* truncation to 32-bit long okay */
2496             long mask = (long)0x0101010101010101ULL;
2497             while (i < slen) {
2498                 xor = *(long *)(old_buf + i) ^ *(long *)(new_buf + i);
2499                 if ((xor - mask) & ~xor & (mask << 7)) {
2500                     /* found the end of an nzrun within the current long */
2501                     while (old_buf[i] != new_buf[i]) {
2502                         nzrun_len++;
2503                         i++;
2504                     }
2505                     break;
2506                 } else {
2507                     i += sizeof(long);
2508                     nzrun_len += sizeof(long);
2509                 }
2510             }
2511         }
2512
2513         d += uleb128_encode_small(dst + d, nzrun_len);
2514         /* overflow */
2515         if (d + nzrun_len > dlen) {
2516             return -1;
2517         }
2518         memcpy(dst + d, nzrun_start, nzrun_len);
2519         d += nzrun_len;
2520         nzrun_len = 0;
2521     }
2522
2523     return d;
2524 }
2525
2526 int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen)
2527 {
2528     int i = 0, d = 0;
2529     int ret;
2530     uint32_t count = 0;
2531
2532     while (i < slen) {
2533
2534         /* zrun */
2535         if ((slen - i) < 2) {
2536             return -1;
2537         }
2538
2539         ret = uleb128_decode_small(src + i, &count);
2540         if (ret < 0 || (i && !count)) {
2541             return -1;
2542         }
2543         i += ret;
2544         d += count;
2545
2546         /* overflow */
2547         if (d > dlen) {
2548             return -1;
2549         }
2550
2551         /* nzrun */
2552         if ((slen - i) < 2) {
2553             return -1;
2554         }
2555
2556         ret = uleb128_decode_small(src + i, &count);
2557         if (ret < 0 || !count) {
2558             return -1;
2559         }
2560         i += ret;
2561
2562         /* overflow */
2563         if (d + count > dlen || i + count > slen) {
2564             return -1;
2565         }
2566
2567         memcpy(dst + d, src + i, count);
2568         d += count;
2569         i += count;
2570     }
2571
2572     return d;
2573 }