savevm.c

   1 /*
   2  * QEMU System Emulator
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a copy
   7  * of this software and associated documentation files (the "Software"), to deal
   8  * in the Software without restriction, including without limitation the rights
   9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  10  * copies of the Software, and to permit persons to whom the Software is
  11  * furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice shall be included in
  14  * all copies or substantial portions of the Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  22  * THE SOFTWARE.
  23  */
  24 #include <unistd.h>
  25 #include <fcntl.h>
  26 #include <time.h>
  27 #include <errno.h>
  28 #include <sys/time.h>
  29 #include <zlib.h>
  30
  31 /* Needed early for CONFIG_BSD etc. */
  32 #include "config-host.h"
  33
  34 #ifndef _WIN32
  35 #include <sys/times.h>
  36 #include <sys/wait.h>
  37 #include <termios.h>
  38 #include <sys/mman.h>
  39 #include <sys/ioctl.h>
  40 #include <sys/resource.h>
  41 #include <sys/socket.h>
  42 #include <netinet/in.h>
  43 #include <net/if.h>
  44 #include <arpa/inet.h>
  45 #include <dirent.h>
  46 #include <netdb.h>
  47 #include <sys/select.h>
  48 #ifdef CONFIG_BSD
  49 #include <sys/stat.h>
  50 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) || defined(__DragonFly__)
  51 #include <libutil.h>
  52 #else
  53 #include <util.h>
  54 #endif
  55 #ifdef __linux__
  56 #include <pty.h>
  57 #include <malloc.h>
  58 #include <linux/rtc.h>
  59 #endif
  60 #endif
  61 #endif
  62
  63 #ifdef _WIN32
  64 #include <windows.h>
  65 #include <malloc.h>
  66 #include <sys/timeb.h>
  67 #include <mmsystem.h>
  68 #define getopt_long_only getopt_long
  69 #define memalign(align, size) malloc(size)
  70 #endif
  71
  72 #include "qemu-common.h"
  73 #include "hw/hw.h"
  74 #include "hw/qdev.h"
  75 #include "net/net.h"
  76 #include "monitor/monitor.h"
  77 #include "sysemu/sysemu.h"
  78 #include "qemu/timer.h"
  79 #include "audio/audio.h"
  80 #include "migration/migration.h"
  81 #include "qemu/sockets.h"
  82 #include "qemu/queue.h"
  83 #include "qemu/timer.h"
  84 #include "sysemu/cpus.h"
  85 #include "exec/memory.h"
  86 #include "qmp-commands.h"
  87 #include "trace.h"
  88 #include "qemu/bitops.h"
  89
  90 #define SELF_ANNOUNCE_ROUNDS 5
  91
  92 #ifndef ETH_P_RARP
  93 #define ETH_P_RARP 0x8035
  94 #endif
  95 #define ARP_HTYPE_ETH 0x0001
  96 #define ARP_PTYPE_IP 0x0800
  97 #define ARP_OP_REQUEST_REV 0x3
  98
  99 static int announce_self_create(uint8_t *buf,
 100                                 uint8_t *mac_addr)
 101 {
 102     /* Ethernet header. */
 103     memset(buf, 0xff, 6);         /* destination MAC addr */
 104     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
 105     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
 106
 107     /* RARP header. */
 108     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
 109     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
 110     *(buf + 18) = 6; /* hardware addr length (ethernet) */
 111     *(buf + 19) = 4; /* protocol addr length (IPv4) */
 112     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
 113     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
 114     memset(buf + 28, 0x00, 4);     /* source protocol addr */
 115     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
 116     memset(buf + 38, 0x00, 4);     /* target protocol addr */
 117
 118     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
 119     memset(buf + 42, 0x00, 18);
 120
 121     return 60; /* len (FCS will be added by hardware) */
 122 }
 123
 124 static void qemu_announce_self_iter(NICState *nic, void *opaque)
 125 {
 126     uint8_t buf[60];
 127     int len;
 128
 129     len = announce_self_create(buf, nic->conf->macaddr.a);
 130
 131     qemu_send_packet_raw(&nic->nc, buf, len);
 132 }
 133
 134
 135 static void qemu_announce_self_once(void *opaque)
 136 {
 137     static int count = SELF_ANNOUNCE_ROUNDS;
 138     QEMUTimer *timer = *(QEMUTimer **)opaque;
 139
 140     qemu_foreach_nic(qemu_announce_self_iter, NULL);
 141
 142     if (--count) {
 143         /* delay 50ms, 150ms, 250ms, ... */
 144         qemu_mod_timer(timer, qemu_get_clock_ms(rt_clock) +
 145                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
 146     } else {
 147             qemu_del_timer(timer);
 148             qemu_free_timer(timer);
 149     }
 150 }
 151
 152 void qemu_announce_self(void)
 153 {
 154         static QEMUTimer *timer;
 155         timer = qemu_new_timer_ms(rt_clock, qemu_announce_self_once, &timer);
 156         qemu_announce_self_once(&timer);
 157 }
 158
 159 /***********************************************************/
 160 /* savevm/loadvm support */
 161
 162 #define IO_BUF_SIZE 32768
 163
 164 struct QEMUFile {
 165     const QEMUFileOps *ops;
 166     void *opaque;
 167     int is_write;
 168
 169     int64_t buf_offset; /* start of buffer when writing, end of buffer
 170                            when reading */
 171     int buf_index;
 172     int buf_size; /* 0 when writing */
 173     uint8_t buf[IO_BUF_SIZE];
 174
 175     int last_error;
 176 };
 177
 178 typedef struct QEMUFileStdio
 179 {
 180     FILE *stdio_file;
 181     QEMUFile *file;
 182 } QEMUFileStdio;
 183
 184 typedef struct QEMUFileSocket
 185 {
 186     int fd;
 187     QEMUFile *file;
 188 } QEMUFileSocket;
 189
 190 static int socket_get_fd(void *opaque)
 191 {
 192     QEMUFileSocket *s = opaque;
 193
 194     return s->fd;
 195 }
 196
 197 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 198 {
 199     QEMUFileSocket *s = opaque;
 200     ssize_t len;
 201
 202     for (;;) {
 203         len = qemu_recv(s->fd, buf, size, 0);
 204         if (len != -1) {
 205             break;
 206         }
 207         if (socket_error() == EAGAIN) {
 208             assert(qemu_in_coroutine());
 209             qemu_coroutine_yield();
 210         } else if (socket_error() != EINTR) {
 211             break;
 212         }
 213     }
 214
 215     if (len == -1) {
 216         len = -socket_error();
 217     }
 218     return len;
 219 }
 220
 221 static int socket_close(void *opaque)
 222 {
 223     QEMUFileSocket *s = opaque;
 224     closesocket(s->fd);
 225     g_free(s);
 226     return 0;
 227 }
 228
 229 static int stdio_get_fd(void *opaque)
 230 {
 231     QEMUFileStdio *s = opaque;
 232
 233     return fileno(s->stdio_file);
 234 }
 235
 236 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 237 {
 238     QEMUFileStdio *s = opaque;
 239     return fwrite(buf, 1, size, s->stdio_file);
 240 }
 241
 242 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 243 {
 244     QEMUFileStdio *s = opaque;
 245     FILE *fp = s->stdio_file;
 246     int bytes;
 247
 248     for (;;) {
 249         clearerr(fp);
 250         bytes = fread(buf, 1, size, fp);
 251         if (bytes != 0 || !ferror(fp)) {
 252             break;
 253         }
 254         if (errno == EAGAIN) {
 255             assert(qemu_in_coroutine());
 256             qemu_coroutine_yield();
 257         } else if (errno != EINTR) {
 258             break;
 259         }
 260     }
 261     return bytes;
 262 }
 263
 264 static int stdio_pclose(void *opaque)
 265 {
 266     QEMUFileStdio *s = opaque;
 267     int ret;
 268     ret = pclose(s->stdio_file);
 269     if (ret == -1) {
 270         ret = -errno;
 271     }
 272     g_free(s);
 273     return ret;
 274 }
 275
 276 static int stdio_fclose(void *opaque)
 277 {
 278     QEMUFileStdio *s = opaque;
 279     int ret = 0;
 280     if (fclose(s->stdio_file) == EOF) {
 281         ret = -errno;
 282     }
 283     g_free(s);
 284     return ret;
 285 }
 286
 287 static const QEMUFileOps stdio_pipe_read_ops = {
 288     .get_fd =     stdio_get_fd,
 289     .get_buffer = stdio_get_buffer,
 290     .close =      stdio_pclose
 291 };
 292
 293 static const QEMUFileOps stdio_pipe_write_ops = {
 294     .get_fd =     stdio_get_fd,
 295     .put_buffer = stdio_put_buffer,
 296     .close =      stdio_pclose
 297 };
 298
 299 QEMUFile *qemu_popen(FILE *stdio_file, const char *mode)
 300 {
 301     QEMUFileStdio *s;
 302
 303     if (stdio_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 304         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 305         return NULL;
 306     }
 307
 308     s = g_malloc0(sizeof(QEMUFileStdio));
 309
 310     s->stdio_file = stdio_file;
 311
 312     if(mode[0] == 'r') {
 313         s->file = qemu_fopen_ops(s, &stdio_pipe_read_ops);
 314     } else {
 315         s->file = qemu_fopen_ops(s, &stdio_pipe_write_ops);
 316     }
 317     return s->file;
 318 }
 319
 320 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 321 {
 322     FILE *popen_file;
 323
 324     popen_file = popen(command, mode);
 325     if(popen_file == NULL) {
 326         return NULL;
 327     }
 328
 329     return qemu_popen(popen_file, mode);
 330 }
 331
 332 static const QEMUFileOps stdio_file_read_ops = {
 333     .get_fd =     stdio_get_fd,
 334     .get_buffer = stdio_get_buffer,
 335     .close =      stdio_fclose
 336 };
 337
 338 static const QEMUFileOps stdio_file_write_ops = {
 339     .get_fd =     stdio_get_fd,
 340     .put_buffer = stdio_put_buffer,
 341     .close =      stdio_fclose
 342 };
 343
 344 QEMUFile *qemu_fdopen(int fd, const char *mode)
 345 {
 346     QEMUFileStdio *s;
 347
 348     if (mode == NULL ||
 349         (mode[0] != 'r' && mode[0] != 'w') ||
 350         mode[1] != 'b' || mode[2] != 0) {
 351         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 352         return NULL;
 353     }
 354
 355     s = g_malloc0(sizeof(QEMUFileStdio));
 356     s->stdio_file = fdopen(fd, mode);
 357     if (!s->stdio_file)
 358         goto fail;
 359
 360     if(mode[0] == 'r') {
 361         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 362     } else {
 363         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 364     }
 365     return s->file;
 366
 367 fail:
 368     g_free(s);
 369     return NULL;
 370 }
 371
 372 static const QEMUFileOps socket_read_ops = {
 373     .get_fd =     socket_get_fd,
 374     .get_buffer = socket_get_buffer,
 375     .close =      socket_close
 376 };
 377
 378 QEMUFile *qemu_fopen_socket(int fd)
 379 {
 380     QEMUFileSocket *s = g_malloc0(sizeof(QEMUFileSocket));
 381
 382     s->fd = fd;
 383     s->file = qemu_fopen_ops(s, &socket_read_ops);
 384     return s->file;
 385 }
 386
 387 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 388 {
 389     QEMUFileStdio *s;
 390
 391     if (mode == NULL ||
 392         (mode[0] != 'r' && mode[0] != 'w') ||
 393         mode[1] != 'b' || mode[2] != 0) {
 394         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 395         return NULL;
 396     }
 397
 398     s = g_malloc0(sizeof(QEMUFileStdio));
 399
 400     s->stdio_file = fopen(filename, mode);
 401     if (!s->stdio_file)
 402         goto fail;
 403
 404     if(mode[0] == 'w') {
 405         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 406     } else {
 407         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 408     }
 409     return s->file;
 410 fail:
 411     g_free(s);
 412     return NULL;
 413 }
 414
 415 static int block_put_buffer(void *opaque, const uint8_t *buf,
 416                            int64_t pos, int size)
 417 {
 418     bdrv_save_vmstate(opaque, buf, pos, size);
 419     return size;
 420 }
 421
 422 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 423 {
 424     return bdrv_load_vmstate(opaque, buf, pos, size);
 425 }
 426
 427 static int bdrv_fclose(void *opaque)
 428 {
 429     return bdrv_flush(opaque);
 430 }
 431
 432 static const QEMUFileOps bdrv_read_ops = {
 433     .get_buffer = block_get_buffer,
 434     .close =      bdrv_fclose
 435 };
 436
 437 static const QEMUFileOps bdrv_write_ops = {
 438     .put_buffer = block_put_buffer,
 439     .close =      bdrv_fclose
 440 };
 441
 442 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 443 {
 444     if (is_writable)
 445         return qemu_fopen_ops(bs, &bdrv_write_ops);
 446     return qemu_fopen_ops(bs, &bdrv_read_ops);
 447 }
 448
 449 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
 450 {
 451     QEMUFile *f;
 452
 453     f = g_malloc0(sizeof(QEMUFile));
 454
 455     f->opaque = opaque;
 456     f->ops = ops;
 457     f->is_write = 0;
 458
 459     return f;
 460 }
 461
 462 int qemu_file_get_error(QEMUFile *f)
 463 {
 464     return f->last_error;
 465 }
 466
 467 static void qemu_file_set_error(QEMUFile *f, int ret)
 468 {
 469     f->last_error = ret;
 470 }
 471
 472 /** Flushes QEMUFile buffer
 473  *
 474  */
 475 static int qemu_fflush(QEMUFile *f)
 476 {
 477     int ret = 0;
 478
 479     if (!f->ops->put_buffer)
 480         return 0;
 481
 482     if (f->is_write && f->buf_index > 0) {
 483         ret = f->ops->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 484         if (ret >= 0) {
 485             f->buf_offset += f->buf_index;
 486         }
 487         f->buf_index = 0;
 488     }
 489     return ret;
 490 }
 491
 492 static void qemu_fill_buffer(QEMUFile *f)
 493 {
 494     int len;
 495     int pending;
 496
 497     if (!f->ops->get_buffer)
 498         return;
 499
 500     if (f->is_write)
 501         abort();
 502
 503     pending = f->buf_size - f->buf_index;
 504     if (pending > 0) {
 505         memmove(f->buf, f->buf + f->buf_index, pending);
 506     }
 507     f->buf_index = 0;
 508     f->buf_size = pending;
 509
 510     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->buf_offset,
 511                         IO_BUF_SIZE - pending);
 512     if (len > 0) {
 513         f->buf_size += len;
 514         f->buf_offset += len;
 515     } else if (len == 0) {
 516         qemu_file_set_error(f, -EIO);
 517     } else if (len != -EAGAIN)
 518         qemu_file_set_error(f, len);
 519 }
 520
 521 int qemu_get_fd(QEMUFile *f)
 522 {
 523     if (f->ops->get_fd) {
 524         return f->ops->get_fd(f->opaque);
 525     }
 526     return -1;
 527 }
 528
 529 /** Closes the file
 530  *
 531  * Returns negative error value if any error happened on previous operations or
 532  * while closing the file. Returns 0 or positive number on success.
 533  *
 534  * The meaning of return value on success depends on the specific backend
 535  * being used.
 536  */
 537 int qemu_fclose(QEMUFile *f)
 538 {
 539     int ret;
 540     ret = qemu_fflush(f);
 541
 542     if (f->ops->close) {
 543         int ret2 = f->ops->close(f->opaque);
 544         if (ret >= 0) {
 545             ret = ret2;
 546         }
 547     }
 548     /* If any error was spotted before closing, we should report it
 549      * instead of the close() return value.
 550      */
 551     if (f->last_error) {
 552         ret = f->last_error;
 553     }
 554     g_free(f);
 555     return ret;
 556 }
 557
 558 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 559 {
 560     int l;
 561
 562     if (f->last_error) {
 563         return;
 564     }
 565
 566     if (f->is_write == 0 && f->buf_index > 0) {
 567         fprintf(stderr,
 568                 "Attempted to write to buffer while read buffer is not empty\n");
 569         abort();
 570     }
 571
 572     while (size > 0) {
 573         l = IO_BUF_SIZE - f->buf_index;
 574         if (l > size)
 575             l = size;
 576         memcpy(f->buf + f->buf_index, buf, l);
 577         f->is_write = 1;
 578         f->buf_index += l;
 579         buf += l;
 580         size -= l;
 581         if (f->buf_index >= IO_BUF_SIZE) {
 582             int ret = qemu_fflush(f);
 583             if (ret < 0) {
 584                 qemu_file_set_error(f, ret);
 585                 break;
 586             }
 587         }
 588     }
 589 }
 590
 591 void qemu_put_byte(QEMUFile *f, int v)
 592 {
 593     if (f->last_error) {
 594         return;
 595     }
 596
 597     if (f->is_write == 0 && f->buf_index > 0) {
 598         fprintf(stderr,
 599                 "Attempted to write to buffer while read buffer is not empty\n");
 600         abort();
 601     }
 602
 603     f->buf[f->buf_index++] = v;
 604     f->is_write = 1;
 605     if (f->buf_index >= IO_BUF_SIZE) {
 606         int ret = qemu_fflush(f);
 607         if (ret < 0) {
 608             qemu_file_set_error(f, ret);
 609         }
 610     }
 611 }
 612
 613 static void qemu_file_skip(QEMUFile *f, int size)
 614 {
 615     if (f->buf_index + size <= f->buf_size) {
 616         f->buf_index += size;
 617     }
 618 }
 619
 620 static int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset)
 621 {
 622     int pending;
 623     int index;
 624
 625     if (f->is_write) {
 626         abort();
 627     }
 628
 629     index = f->buf_index + offset;
 630     pending = f->buf_size - index;
 631     if (pending < size) {
 632         qemu_fill_buffer(f);
 633         index = f->buf_index + offset;
 634         pending = f->buf_size - index;
 635     }
 636
 637     if (pending <= 0) {
 638         return 0;
 639     }
 640     if (size > pending) {
 641         size = pending;
 642     }
 643
 644     memcpy(buf, f->buf + index, size);
 645     return size;
 646 }
 647
 648 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size)
 649 {
 650     int pending = size;
 651     int done = 0;
 652
 653     while (pending > 0) {
 654         int res;
 655
 656         res = qemu_peek_buffer(f, buf, pending, 0);
 657         if (res == 0) {
 658             return done;
 659         }
 660         qemu_file_skip(f, res);
 661         buf += res;
 662         pending -= res;
 663         done += res;
 664     }
 665     return done;
 666 }
 667
 668 static int qemu_peek_byte(QEMUFile *f, int offset)
 669 {
 670     int index = f->buf_index + offset;
 671
 672     if (f->is_write) {
 673         abort();
 674     }
 675
 676     if (index >= f->buf_size) {
 677         qemu_fill_buffer(f);
 678         index = f->buf_index + offset;
 679         if (index >= f->buf_size) {
 680             return 0;
 681         }
 682     }
 683     return f->buf[index];
 684 }
 685
 686 int qemu_get_byte(QEMUFile *f)
 687 {
 688     int result;
 689
 690     result = qemu_peek_byte(f, 0);
 691     qemu_file_skip(f, 1);
 692     return result;
 693 }
 694
 695 static int64_t qemu_ftell(QEMUFile *f)
 696 {
 697     return f->buf_offset - f->buf_size + f->buf_index;
 698 }
 699
 700 int qemu_file_rate_limit(QEMUFile *f)
 701 {
 702     if (f->ops->rate_limit)
 703         return f->ops->rate_limit(f->opaque);
 704
 705     return 0;
 706 }
 707
 708 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 709 {
 710     if (f->ops->get_rate_limit)
 711         return f->ops->get_rate_limit(f->opaque);
 712
 713     return 0;
 714 }
 715
 716 int64_t qemu_file_set_rate_limit(QEMUFile *f, int64_t new_rate)
 717 {
 718     /* any failed or completed migration keeps its state to allow probing of
 719      * migration data, but has no associated file anymore */
 720     if (f && f->ops->set_rate_limit)
 721         return f->ops->set_rate_limit(f->opaque, new_rate);
 722
 723     return 0;
 724 }
 725
 726 void qemu_put_be16(QEMUFile *f, unsigned int v)
 727 {
 728     qemu_put_byte(f, v >> 8);
 729     qemu_put_byte(f, v);
 730 }
 731
 732 void qemu_put_be32(QEMUFile *f, unsigned int v)
 733 {
 734     qemu_put_byte(f, v >> 24);
 735     qemu_put_byte(f, v >> 16);
 736     qemu_put_byte(f, v >> 8);
 737     qemu_put_byte(f, v);
 738 }
 739
 740 void qemu_put_be64(QEMUFile *f, uint64_t v)
 741 {
 742     qemu_put_be32(f, v >> 32);
 743     qemu_put_be32(f, v);
 744 }
 745
 746 unsigned int qemu_get_be16(QEMUFile *f)
 747 {
 748     unsigned int v;
 749     v = qemu_get_byte(f) << 8;
 750     v |= qemu_get_byte(f);
 751     return v;
 752 }
 753
 754 unsigned int qemu_get_be32(QEMUFile *f)
 755 {
 756     unsigned int v;
 757     v = qemu_get_byte(f) << 24;
 758     v |= qemu_get_byte(f) << 16;
 759     v |= qemu_get_byte(f) << 8;
 760     v |= qemu_get_byte(f);
 761     return v;
 762 }
 763
 764 uint64_t qemu_get_be64(QEMUFile *f)
 765 {
 766     uint64_t v;
 767     v = (uint64_t)qemu_get_be32(f) << 32;
 768     v |= qemu_get_be32(f);
 769     return v;
 770 }
 771
 772
 773 /* timer */
 774
 775 void qemu_put_timer(QEMUFile *f, QEMUTimer *ts)
 776 {
 777     uint64_t expire_time;
 778
 779     expire_time = qemu_timer_expire_time_ns(ts);
 780     qemu_put_be64(f, expire_time);
 781 }
 782
 783 void qemu_get_timer(QEMUFile *f, QEMUTimer *ts)
 784 {
 785     uint64_t expire_time;
 786
 787     expire_time = qemu_get_be64(f);
 788     if (expire_time != -1) {
 789         qemu_mod_timer_ns(ts, expire_time);
 790     } else {
 791         qemu_del_timer(ts);
 792     }
 793 }
 794
 795
 796 /* bool */
 797
 798 static int get_bool(QEMUFile *f, void *pv, size_t size)
 799 {
 800     bool *v = pv;
 801     *v = qemu_get_byte(f);
 802     return 0;
 803 }
 804
 805 static void put_bool(QEMUFile *f, void *pv, size_t size)
 806 {
 807     bool *v = pv;
 808     qemu_put_byte(f, *v);
 809 }
 810
 811 const VMStateInfo vmstate_info_bool = {
 812     .name = "bool",
 813     .get  = get_bool,
 814     .put  = put_bool,
 815 };
 816
 817 /* 8 bit int */
 818
 819 static int get_int8(QEMUFile *f, void *pv, size_t size)
 820 {
 821     int8_t *v = pv;
 822     qemu_get_s8s(f, v);
 823     return 0;
 824 }
 825
 826 static void put_int8(QEMUFile *f, void *pv, size_t size)
 827 {
 828     int8_t *v = pv;
 829     qemu_put_s8s(f, v);
 830 }
 831
 832 const VMStateInfo vmstate_info_int8 = {
 833     .name = "int8",
 834     .get  = get_int8,
 835     .put  = put_int8,
 836 };
 837
 838 /* 16 bit int */
 839
 840 static int get_int16(QEMUFile *f, void *pv, size_t size)
 841 {
 842     int16_t *v = pv;
 843     qemu_get_sbe16s(f, v);
 844     return 0;
 845 }
 846
 847 static void put_int16(QEMUFile *f, void *pv, size_t size)
 848 {
 849     int16_t *v = pv;
 850     qemu_put_sbe16s(f, v);
 851 }
 852
 853 const VMStateInfo vmstate_info_int16 = {
 854     .name = "int16",
 855     .get  = get_int16,
 856     .put  = put_int16,
 857 };
 858
 859 /* 32 bit int */
 860
 861 static int get_int32(QEMUFile *f, void *pv, size_t size)
 862 {
 863     int32_t *v = pv;
 864     qemu_get_sbe32s(f, v);
 865     return 0;
 866 }
 867
 868 static void put_int32(QEMUFile *f, void *pv, size_t size)
 869 {
 870     int32_t *v = pv;
 871     qemu_put_sbe32s(f, v);
 872 }
 873
 874 const VMStateInfo vmstate_info_int32 = {
 875     .name = "int32",
 876     .get  = get_int32,
 877     .put  = put_int32,
 878 };
 879
 880 /* 32 bit int. See that the received value is the same than the one
 881    in the field */
 882
 883 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 884 {
 885     int32_t *v = pv;
 886     int32_t v2;
 887     qemu_get_sbe32s(f, &v2);
 888
 889     if (*v == v2)
 890         return 0;
 891     return -EINVAL;
 892 }
 893
 894 const VMStateInfo vmstate_info_int32_equal = {
 895     .name = "int32 equal",
 896     .get  = get_int32_equal,
 897     .put  = put_int32,
 898 };
 899
 900 /* 32 bit int. See that the received value is the less or the same
 901    than the one in the field */
 902
 903 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 904 {
 905     int32_t *old = pv;
 906     int32_t new;
 907     qemu_get_sbe32s(f, &new);
 908
 909     if (*old <= new)
 910         return 0;
 911     return -EINVAL;
 912 }
 913
 914 const VMStateInfo vmstate_info_int32_le = {
 915     .name = "int32 equal",
 916     .get  = get_int32_le,
 917     .put  = put_int32,
 918 };
 919
 920 /* 64 bit int */
 921
 922 static int get_int64(QEMUFile *f, void *pv, size_t size)
 923 {
 924     int64_t *v = pv;
 925     qemu_get_sbe64s(f, v);
 926     return 0;
 927 }
 928
 929 static void put_int64(QEMUFile *f, void *pv, size_t size)
 930 {
 931     int64_t *v = pv;
 932     qemu_put_sbe64s(f, v);
 933 }
 934
 935 const VMStateInfo vmstate_info_int64 = {
 936     .name = "int64",
 937     .get  = get_int64,
 938     .put  = put_int64,
 939 };
 940
 941 /* 8 bit unsigned int */
 942
 943 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 944 {
 945     uint8_t *v = pv;
 946     qemu_get_8s(f, v);
 947     return 0;
 948 }
 949
 950 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 951 {
 952     uint8_t *v = pv;
 953     qemu_put_8s(f, v);
 954 }
 955
 956 const VMStateInfo vmstate_info_uint8 = {
 957     .name = "uint8",
 958     .get  = get_uint8,
 959     .put  = put_uint8,
 960 };
 961
 962 /* 16 bit unsigned int */
 963
 964 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 965 {
 966     uint16_t *v = pv;
 967     qemu_get_be16s(f, v);
 968     return 0;
 969 }
 970
 971 static void put_uint16(QEMUFile *f, void *pv, size_t size)
 972 {
 973     uint16_t *v = pv;
 974     qemu_put_be16s(f, v);
 975 }
 976
 977 const VMStateInfo vmstate_info_uint16 = {
 978     .name = "uint16",
 979     .get  = get_uint16,
 980     .put  = put_uint16,
 981 };
 982
 983 /* 32 bit unsigned int */
 984
 985 static int get_uint32(QEMUFile *f, void *pv, size_t size)
 986 {
 987     uint32_t *v = pv;
 988     qemu_get_be32s(f, v);
 989     return 0;
 990 }
 991
 992 static void put_uint32(QEMUFile *f, void *pv, size_t size)
 993 {
 994     uint32_t *v = pv;
 995     qemu_put_be32s(f, v);
 996 }
 997
 998 const VMStateInfo vmstate_info_uint32 = {
 999     .name = "uint32",
1000     .get  = get_uint32,
1001     .put  = put_uint32,
1002 };
1003
1004 /* 32 bit uint. See that the received value is the same than the one
1005    in the field */
1006
1007 static int get_uint32_equal(QEMUFile *f, void *pv, size_t size)
1008 {
1009     uint32_t *v = pv;
1010     uint32_t v2;
1011     qemu_get_be32s(f, &v2);
1012
1013     if (*v == v2) {
1014         return 0;
1015     }
1016     return -EINVAL;
1017 }
1018
1019 const VMStateInfo vmstate_info_uint32_equal = {
1020     .name = "uint32 equal",
1021     .get  = get_uint32_equal,
1022     .put  = put_uint32,
1023 };
1024
1025 /* 64 bit unsigned int */
1026
1027 static int get_uint64(QEMUFile *f, void *pv, size_t size)
1028 {
1029     uint64_t *v = pv;
1030     qemu_get_be64s(f, v);
1031     return 0;
1032 }
1033
1034 static void put_uint64(QEMUFile *f, void *pv, size_t size)
1035 {
1036     uint64_t *v = pv;
1037     qemu_put_be64s(f, v);
1038 }
1039
1040 const VMStateInfo vmstate_info_uint64 = {
1041     .name = "uint64",
1042     .get  = get_uint64,
1043     .put  = put_uint64,
1044 };
1045
1046 /* 8 bit int. See that the received value is the same than the one
1047    in the field */
1048
1049 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
1050 {
1051     uint8_t *v = pv;
1052     uint8_t v2;
1053     qemu_get_8s(f, &v2);
1054
1055     if (*v == v2)
1056         return 0;
1057     return -EINVAL;
1058 }
1059
1060 const VMStateInfo vmstate_info_uint8_equal = {
1061     .name = "uint8 equal",
1062     .get  = get_uint8_equal,
1063     .put  = put_uint8,
1064 };
1065
1066 /* 16 bit unsigned int int. See that the received value is the same than the one
1067    in the field */
1068
1069 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
1070 {
1071     uint16_t *v = pv;
1072     uint16_t v2;
1073     qemu_get_be16s(f, &v2);
1074
1075     if (*v == v2)
1076         return 0;
1077     return -EINVAL;
1078 }
1079
1080 const VMStateInfo vmstate_info_uint16_equal = {
1081     .name = "uint16 equal",
1082     .get  = get_uint16_equal,
1083     .put  = put_uint16,
1084 };
1085
1086 /* timers  */
1087
1088 static int get_timer(QEMUFile *f, void *pv, size_t size)
1089 {
1090     QEMUTimer *v = pv;
1091     qemu_get_timer(f, v);
1092     return 0;
1093 }
1094
1095 static void put_timer(QEMUFile *f, void *pv, size_t size)
1096 {
1097     QEMUTimer *v = pv;
1098     qemu_put_timer(f, v);
1099 }
1100
1101 const VMStateInfo vmstate_info_timer = {
1102     .name = "timer",
1103     .get  = get_timer,
1104     .put  = put_timer,
1105 };
1106
1107 /* uint8_t buffers */
1108
1109 static int get_buffer(QEMUFile *f, void *pv, size_t size)
1110 {
1111     uint8_t *v = pv;
1112     qemu_get_buffer(f, v, size);
1113     return 0;
1114 }
1115
1116 static void put_buffer(QEMUFile *f, void *pv, size_t size)
1117 {
1118     uint8_t *v = pv;
1119     qemu_put_buffer(f, v, size);
1120 }
1121
1122 const VMStateInfo vmstate_info_buffer = {
1123     .name = "buffer",
1124     .get  = get_buffer,
1125     .put  = put_buffer,
1126 };
1127
1128 /* unused buffers: space that was used for some fields that are
1129    not useful anymore */
1130
1131 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
1132 {
1133     uint8_t buf[1024];
1134     int block_len;
1135
1136     while (size > 0) {
1137         block_len = MIN(sizeof(buf), size);
1138         size -= block_len;
1139         qemu_get_buffer(f, buf, block_len);
1140     }
1141    return 0;
1142 }
1143
1144 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1145 {
1146     static const uint8_t buf[1024];
1147     int block_len;
1148
1149     while (size > 0) {
1150         block_len = MIN(sizeof(buf), size);
1151         size -= block_len;
1152         qemu_put_buffer(f, buf, block_len);
1153     }
1154 }
1155
1156 const VMStateInfo vmstate_info_unused_buffer = {
1157     .name = "unused_buffer",
1158     .get  = get_unused_buffer,
1159     .put  = put_unused_buffer,
1160 };
1161
1162 /* bitmaps (as defined by bitmap.h). Note that size here is the size
1163  * of the bitmap in bits. The on-the-wire format of a bitmap is 64
1164  * bit words with the bits in big endian order. The in-memory format
1165  * is an array of 'unsigned long', which may be either 32 or 64 bits.
1166  */
1167 /* This is the number of 64 bit words sent over the wire */
1168 #define BITS_TO_U64S(nr) DIV_ROUND_UP(nr, 64)
1169 static int get_bitmap(QEMUFile *f, void *pv, size_t size)
1170 {
1171     unsigned long *bmp = pv;
1172     int i, idx = 0;
1173     for (i = 0; i < BITS_TO_U64S(size); i++) {
1174         uint64_t w = qemu_get_be64(f);
1175         bmp[idx++] = w;
1176         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1177             bmp[idx++] = w >> 32;
1178         }
1179     }
1180     return 0;
1181 }
1182
1183 static void put_bitmap(QEMUFile *f, void *pv, size_t size)
1184 {
1185     unsigned long *bmp = pv;
1186     int i, idx = 0;
1187     for (i = 0; i < BITS_TO_U64S(size); i++) {
1188         uint64_t w = bmp[idx++];
1189         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1190             w |= ((uint64_t)bmp[idx++]) << 32;
1191         }
1192         qemu_put_be64(f, w);
1193     }
1194 }
1195
1196 const VMStateInfo vmstate_info_bitmap = {
1197     .name = "bitmap",
1198     .get = get_bitmap,
1199     .put = put_bitmap,
1200 };
1201
1202 typedef struct CompatEntry {
1203     char idstr[256];
1204     int instance_id;
1205 } CompatEntry;
1206
1207 typedef struct SaveStateEntry {
1208     QTAILQ_ENTRY(SaveStateEntry) entry;
1209     char idstr[256];
1210     int instance_id;
1211     int alias_id;
1212     int version_id;
1213     int section_id;
1214     SaveVMHandlers *ops;
1215     const VMStateDescription *vmsd;
1216     void *opaque;
1217     CompatEntry *compat;
1218     int no_migrate;
1219     int is_ram;
1220 } SaveStateEntry;
1221
1222
1223 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1224     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1225 static int global_section_id;
1226
1227 static int calculate_new_instance_id(const char *idstr)
1228 {
1229     SaveStateEntry *se;
1230     int instance_id = 0;
1231
1232     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1233         if (strcmp(idstr, se->idstr) == 0
1234             && instance_id <= se->instance_id) {
1235             instance_id = se->instance_id + 1;
1236         }
1237     }
1238     return instance_id;
1239 }
1240
1241 static int calculate_compat_instance_id(const char *idstr)
1242 {
1243     SaveStateEntry *se;
1244     int instance_id = 0;
1245
1246     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1247         if (!se->compat)
1248             continue;
1249
1250         if (strcmp(idstr, se->compat->idstr) == 0
1251             && instance_id <= se->compat->instance_id) {
1252             instance_id = se->compat->instance_id + 1;
1253         }
1254     }
1255     return instance_id;
1256 }
1257
1258 /* TODO: Individual devices generally have very little idea about the rest
1259    of the system, so instance_id should be removed/replaced.
1260    Meanwhile pass -1 as instance_id if you do not already have a clearly
1261    distinguishing id for all instances of your device class. */
1262 int register_savevm_live(DeviceState *dev,
1263                          const char *idstr,
1264                          int instance_id,
1265                          int version_id,
1266                          SaveVMHandlers *ops,
1267                          void *opaque)
1268 {
1269     SaveStateEntry *se;
1270
1271     se = g_malloc0(sizeof(SaveStateEntry));
1272     se->version_id = version_id;
1273     se->section_id = global_section_id++;
1274     se->ops = ops;
1275     se->opaque = opaque;
1276     se->vmsd = NULL;
1277     se->no_migrate = 0;
1278     /* if this is a live_savem then set is_ram */
1279     if (ops->save_live_setup != NULL) {
1280         se->is_ram = 1;
1281     }
1282
1283     if (dev) {
1284         char *id = qdev_get_dev_path(dev);
1285         if (id) {
1286             pstrcpy(se->idstr, sizeof(se->idstr), id);
1287             pstrcat(se->idstr, sizeof(se->idstr), "/");
1288             g_free(id);
1289
1290             se->compat = g_malloc0(sizeof(CompatEntry));
1291             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1292             se->compat->instance_id = instance_id == -1 ?
1293                          calculate_compat_instance_id(idstr) : instance_id;
1294             instance_id = -1;
1295         }
1296     }
1297     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1298
1299     if (instance_id == -1) {
1300         se->instance_id = calculate_new_instance_id(se->idstr);
1301     } else {
1302         se->instance_id = instance_id;
1303     }
1304     assert(!se->compat || se->instance_id == 0);
1305     /* add at the end of list */
1306     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1307     return 0;
1308 }
1309
1310 int register_savevm(DeviceState *dev,
1311                     const char *idstr,
1312                     int instance_id,
1313                     int version_id,
1314                     SaveStateHandler *save_state,
1315                     LoadStateHandler *load_state,
1316                     void *opaque)
1317 {
1318     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
1319     ops->save_state = save_state;
1320     ops->load_state = load_state;
1321     return register_savevm_live(dev, idstr, instance_id, version_id,
1322                                 ops, opaque);
1323 }
1324
1325 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1326 {
1327     SaveStateEntry *se, *new_se;
1328     char id[256] = "";
1329
1330     if (dev) {
1331         char *path = qdev_get_dev_path(dev);
1332         if (path) {
1333             pstrcpy(id, sizeof(id), path);
1334             pstrcat(id, sizeof(id), "/");
1335             g_free(path);
1336         }
1337     }
1338     pstrcat(id, sizeof(id), idstr);
1339
1340     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1341         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1342             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1343             if (se->compat) {
1344                 g_free(se->compat);
1345             }
1346             g_free(se->ops);
1347             g_free(se);
1348         }
1349     }
1350 }
1351
1352 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1353                                    const VMStateDescription *vmsd,
1354                                    void *opaque, int alias_id,
1355                                    int required_for_version)
1356 {
1357     SaveStateEntry *se;
1358
1359     /* If this triggers, alias support can be dropped for the vmsd. */
1360     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1361
1362     se = g_malloc0(sizeof(SaveStateEntry));
1363     se->version_id = vmsd->version_id;
1364     se->section_id = global_section_id++;
1365     se->opaque = opaque;
1366     se->vmsd = vmsd;
1367     se->alias_id = alias_id;
1368     se->no_migrate = vmsd->unmigratable;
1369
1370     if (dev) {
1371         char *id = qdev_get_dev_path(dev);
1372         if (id) {
1373             pstrcpy(se->idstr, sizeof(se->idstr), id);
1374             pstrcat(se->idstr, sizeof(se->idstr), "/");
1375             g_free(id);
1376
1377             se->compat = g_malloc0(sizeof(CompatEntry));
1378             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1379             se->compat->instance_id = instance_id == -1 ?
1380                          calculate_compat_instance_id(vmsd->name) : instance_id;
1381             instance_id = -1;
1382         }
1383     }
1384     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1385
1386     if (instance_id == -1) {
1387         se->instance_id = calculate_new_instance_id(se->idstr);
1388     } else {
1389         se->instance_id = instance_id;
1390     }
1391     assert(!se->compat || se->instance_id == 0);
1392     /* add at the end of list */
1393     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1394     return 0;
1395 }
1396
1397 int vmstate_register(DeviceState *dev, int instance_id,
1398                      const VMStateDescription *vmsd, void *opaque)
1399 {
1400     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1401                                           opaque, -1, 0);
1402 }
1403
1404 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1405                         void *opaque)
1406 {
1407     SaveStateEntry *se, *new_se;
1408
1409     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1410         if (se->vmsd == vmsd && se->opaque == opaque) {
1411             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1412             if (se->compat) {
1413                 g_free(se->compat);
1414             }
1415             g_free(se);
1416         }
1417     }
1418 }
1419
1420 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1421                                     void *opaque);
1422 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1423                                    void *opaque);
1424
1425 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1426                        void *opaque, int version_id)
1427 {
1428     VMStateField *field = vmsd->fields;
1429     int ret;
1430
1431     if (version_id > vmsd->version_id) {
1432         return -EINVAL;
1433     }
1434     if (version_id < vmsd->minimum_version_id_old) {
1435         return -EINVAL;
1436     }
1437     if  (version_id < vmsd->minimum_version_id) {
1438         return vmsd->load_state_old(f, opaque, version_id);
1439     }
1440     if (vmsd->pre_load) {
1441         int ret = vmsd->pre_load(opaque);
1442         if (ret)
1443             return ret;
1444     }
1445     while(field->name) {
1446         if ((field->field_exists &&
1447              field->field_exists(opaque, version_id)) ||
1448             (!field->field_exists &&
1449              field->version_id <= version_id)) {
1450             void *base_addr = opaque + field->offset;
1451             int i, n_elems = 1;
1452             int size = field->size;
1453
1454             if (field->flags & VMS_VBUFFER) {
1455                 size = *(int32_t *)(opaque+field->size_offset);
1456                 if (field->flags & VMS_MULTIPLY) {
1457                     size *= field->size;
1458                 }
1459             }
1460             if (field->flags & VMS_ARRAY) {
1461                 n_elems = field->num;
1462             } else if (field->flags & VMS_VARRAY_INT32) {
1463                 n_elems = *(int32_t *)(opaque+field->num_offset);
1464             } else if (field->flags & VMS_VARRAY_UINT32) {
1465                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1466             } else if (field->flags & VMS_VARRAY_UINT16) {
1467                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1468             } else if (field->flags & VMS_VARRAY_UINT8) {
1469                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1470             }
1471             if (field->flags & VMS_POINTER) {
1472                 base_addr = *(void **)base_addr + field->start;
1473             }
1474             for (i = 0; i < n_elems; i++) {
1475                 void *addr = base_addr + size * i;
1476
1477                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1478                     addr = *(void **)addr;
1479                 }
1480                 if (field->flags & VMS_STRUCT) {
1481                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1482                 } else {
1483                     ret = field->info->get(f, addr, size);
1484
1485                 }
1486                 if (ret < 0) {
1487                     return ret;
1488                 }
1489             }
1490         }
1491         field++;
1492     }
1493     ret = vmstate_subsection_load(f, vmsd, opaque);
1494     if (ret != 0) {
1495         return ret;
1496     }
1497     if (vmsd->post_load) {
1498         return vmsd->post_load(opaque, version_id);
1499     }
1500     return 0;
1501 }
1502
1503 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1504                         void *opaque)
1505 {
1506     VMStateField *field = vmsd->fields;
1507
1508     if (vmsd->pre_save) {
1509         vmsd->pre_save(opaque);
1510     }
1511     while(field->name) {
1512         if (!field->field_exists ||
1513             field->field_exists(opaque, vmsd->version_id)) {
1514             void *base_addr = opaque + field->offset;
1515             int i, n_elems = 1;
1516             int size = field->size;
1517
1518             if (field->flags & VMS_VBUFFER) {
1519                 size = *(int32_t *)(opaque+field->size_offset);
1520                 if (field->flags & VMS_MULTIPLY) {
1521                     size *= field->size;
1522                 }
1523             }
1524             if (field->flags & VMS_ARRAY) {
1525                 n_elems = field->num;
1526             } else if (field->flags & VMS_VARRAY_INT32) {
1527                 n_elems = *(int32_t *)(opaque+field->num_offset);
1528             } else if (field->flags & VMS_VARRAY_UINT32) {
1529                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1530             } else if (field->flags & VMS_VARRAY_UINT16) {
1531                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1532             } else if (field->flags & VMS_VARRAY_UINT8) {
1533                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1534             }
1535             if (field->flags & VMS_POINTER) {
1536                 base_addr = *(void **)base_addr + field->start;
1537             }
1538             for (i = 0; i < n_elems; i++) {
1539                 void *addr = base_addr + size * i;
1540
1541                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1542                     addr = *(void **)addr;
1543                 }
1544                 if (field->flags & VMS_STRUCT) {
1545                     vmstate_save_state(f, field->vmsd, addr);
1546                 } else {
1547                     field->info->put(f, addr, size);
1548                 }
1549             }
1550         }
1551         field++;
1552     }
1553     vmstate_subsection_save(f, vmsd, opaque);
1554 }
1555
1556 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1557 {
1558     if (!se->vmsd) {         /* Old style */
1559         return se->ops->load_state(f, se->opaque, version_id);
1560     }
1561     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1562 }
1563
1564 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1565 {
1566     if (!se->vmsd) {         /* Old style */
1567         se->ops->save_state(f, se->opaque);
1568         return;
1569     }
1570     vmstate_save_state(f,se->vmsd, se->opaque);
1571 }
1572
1573 #define QEMU_VM_FILE_MAGIC           0x5145564d
1574 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1575 #define QEMU_VM_FILE_VERSION         0x00000003
1576
1577 #define QEMU_VM_EOF                  0x00
1578 #define QEMU_VM_SECTION_START        0x01
1579 #define QEMU_VM_SECTION_PART         0x02
1580 #define QEMU_VM_SECTION_END          0x03
1581 #define QEMU_VM_SECTION_FULL         0x04
1582 #define QEMU_VM_SUBSECTION           0x05
1583
1584 bool qemu_savevm_state_blocked(Error **errp)
1585 {
1586     SaveStateEntry *se;
1587
1588     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1589         if (se->no_migrate) {
1590             error_set(errp, QERR_MIGRATION_NOT_SUPPORTED, se->idstr);
1591             return true;
1592         }
1593     }
1594     return false;
1595 }
1596
1597 int qemu_savevm_state_begin(QEMUFile *f,
1598                             const MigrationParams *params)
1599 {
1600     SaveStateEntry *se;
1601     int ret;
1602
1603     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1604         if (!se->ops || !se->ops->set_params) {
1605             continue;
1606         }
1607         se->ops->set_params(params, se->opaque);
1608     }
1609
1610     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1611     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1612
1613     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1614         int len;
1615
1616         if (!se->ops || !se->ops->save_live_setup) {
1617             continue;
1618         }
1619         if (se->ops && se->ops->is_active) {
1620             if (!se->ops->is_active(se->opaque)) {
1621                 continue;
1622             }
1623         }
1624         /* Section type */
1625         qemu_put_byte(f, QEMU_VM_SECTION_START);
1626         qemu_put_be32(f, se->section_id);
1627
1628         /* ID string */
1629         len = strlen(se->idstr);
1630         qemu_put_byte(f, len);
1631         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1632
1633         qemu_put_be32(f, se->instance_id);
1634         qemu_put_be32(f, se->version_id);
1635
1636         ret = se->ops->save_live_setup(f, se->opaque);
1637         if (ret < 0) {
1638             qemu_savevm_state_cancel(f);
1639             return ret;
1640         }
1641     }
1642     ret = qemu_file_get_error(f);
1643     if (ret != 0) {
1644         qemu_savevm_state_cancel(f);
1645     }
1646
1647     return ret;
1648
1649 }
1650
1651 /*
1652  * this function has three return values:
1653  *   negative: there was one error, and we have -errno.
1654  *   0 : We haven't finished, caller have to go again
1655  *   1 : We have finished, we can go to complete phase
1656  */
1657 int qemu_savevm_state_iterate(QEMUFile *f)
1658 {
1659     SaveStateEntry *se;
1660     int ret = 1;
1661
1662     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1663         if (!se->ops || !se->ops->save_live_iterate) {
1664             continue;
1665         }
1666         if (se->ops && se->ops->is_active) {
1667             if (!se->ops->is_active(se->opaque)) {
1668                 continue;
1669             }
1670         }
1671         if (qemu_file_rate_limit(f)) {
1672             return 0;
1673         }
1674         trace_savevm_section_start();
1675         /* Section type */
1676         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1677         qemu_put_be32(f, se->section_id);
1678
1679         ret = se->ops->save_live_iterate(f, se->opaque);
1680         trace_savevm_section_end(se->section_id);
1681
1682         if (ret <= 0) {
1683             /* Do not proceed to the next vmstate before this one reported
1684                completion of the current stage. This serializes the migration
1685                and reduces the probability that a faster changing state is
1686                synchronized over and over again. */
1687             break;
1688         }
1689     }
1690     if (ret != 0) {
1691         return ret;
1692     }
1693     ret = qemu_file_get_error(f);
1694     if (ret != 0) {
1695         qemu_savevm_state_cancel(f);
1696     }
1697     return ret;
1698 }
1699
1700 int qemu_savevm_state_complete(QEMUFile *f)
1701 {
1702     SaveStateEntry *se;
1703     int ret;
1704
1705     cpu_synchronize_all_states();
1706
1707     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1708         if (!se->ops || !se->ops->save_live_complete) {
1709             continue;
1710         }
1711         if (se->ops && se->ops->is_active) {
1712             if (!se->ops->is_active(se->opaque)) {
1713                 continue;
1714             }
1715         }
1716         trace_savevm_section_start();
1717         /* Section type */
1718         qemu_put_byte(f, QEMU_VM_SECTION_END);
1719         qemu_put_be32(f, se->section_id);
1720
1721         ret = se->ops->save_live_complete(f, se->opaque);
1722         trace_savevm_section_end(se->section_id);
1723         if (ret < 0) {
1724             return ret;
1725         }
1726     }
1727
1728     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1729         int len;
1730
1731         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1732             continue;
1733         }
1734         trace_savevm_section_start();
1735         /* Section type */
1736         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1737         qemu_put_be32(f, se->section_id);
1738
1739         /* ID string */
1740         len = strlen(se->idstr);
1741         qemu_put_byte(f, len);
1742         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1743
1744         qemu_put_be32(f, se->instance_id);
1745         qemu_put_be32(f, se->version_id);
1746
1747         vmstate_save(f, se);
1748         trace_savevm_section_end(se->section_id);
1749     }
1750
1751     qemu_put_byte(f, QEMU_VM_EOF);
1752
1753     return qemu_file_get_error(f);
1754 }
1755
1756 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
1757 {
1758     SaveStateEntry *se;
1759     uint64_t ret = 0;
1760
1761     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1762         if (!se->ops || !se->ops->save_live_pending) {
1763             continue;
1764         }
1765         if (se->ops && se->ops->is_active) {
1766             if (!se->ops->is_active(se->opaque)) {
1767                 continue;
1768             }
1769         }
1770         ret += se->ops->save_live_pending(f, se->opaque, max_size);
1771     }
1772     return ret;
1773 }
1774
1775 void qemu_savevm_state_cancel(QEMUFile *f)
1776 {
1777     SaveStateEntry *se;
1778
1779     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1780         if (se->ops && se->ops->cancel) {
1781             se->ops->cancel(se->opaque);
1782         }
1783     }
1784 }
1785
1786 static int qemu_savevm_state(QEMUFile *f)
1787 {
1788     int ret;
1789     MigrationParams params = {
1790         .blk = 0,
1791         .shared = 0
1792     };
1793
1794     if (qemu_savevm_state_blocked(NULL)) {
1795         ret = -EINVAL;
1796         goto out;
1797     }
1798
1799     ret = qemu_savevm_state_begin(f, &params);
1800     if (ret < 0)
1801         goto out;
1802
1803     do {
1804         ret = qemu_savevm_state_iterate(f);
1805         if (ret < 0)
1806             goto out;
1807     } while (ret == 0);
1808
1809     ret = qemu_savevm_state_complete(f);
1810
1811 out:
1812     if (ret == 0) {
1813         ret = qemu_file_get_error(f);
1814     }
1815
1816     return ret;
1817 }
1818
1819 static int qemu_save_device_state(QEMUFile *f)
1820 {
1821     SaveStateEntry *se;
1822
1823     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1824     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1825
1826     cpu_synchronize_all_states();
1827
1828     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1829         int len;
1830
1831         if (se->is_ram) {
1832             continue;
1833         }
1834         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1835             continue;
1836         }
1837
1838         /* Section type */
1839         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1840         qemu_put_be32(f, se->section_id);
1841
1842         /* ID string */
1843         len = strlen(se->idstr);
1844         qemu_put_byte(f, len);
1845         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1846
1847         qemu_put_be32(f, se->instance_id);
1848         qemu_put_be32(f, se->version_id);
1849
1850         vmstate_save(f, se);
1851     }
1852
1853     qemu_put_byte(f, QEMU_VM_EOF);
1854
1855     return qemu_file_get_error(f);
1856 }
1857
1858 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1859 {
1860     SaveStateEntry *se;
1861
1862     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1863         if (!strcmp(se->idstr, idstr) &&
1864             (instance_id == se->instance_id ||
1865              instance_id == se->alias_id))
1866             return se;
1867         /* Migrating from an older version? */
1868         if (strstr(se->idstr, idstr) && se->compat) {
1869             if (!strcmp(se->compat->idstr, idstr) &&
1870                 (instance_id == se->compat->instance_id ||
1871                  instance_id == se->alias_id))
1872                 return se;
1873         }
1874     }
1875     return NULL;
1876 }
1877
1878 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1879 {
1880     while(sub && sub->needed) {
1881         if (strcmp(idstr, sub->vmsd->name) == 0) {
1882             return sub->vmsd;
1883         }
1884         sub++;
1885     }
1886     return NULL;
1887 }
1888
1889 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1890                                    void *opaque)
1891 {
1892     while (qemu_peek_byte(f, 0) == QEMU_VM_SUBSECTION) {
1893         char idstr[256];
1894         int ret;
1895         uint8_t version_id, len, size;
1896         const VMStateDescription *sub_vmsd;
1897
1898         len = qemu_peek_byte(f, 1);
1899         if (len < strlen(vmsd->name) + 1) {
1900             /* subsection name has be be "section_name/a" */
1901             return 0;
1902         }
1903         size = qemu_peek_buffer(f, (uint8_t *)idstr, len, 2);
1904         if (size != len) {
1905             return 0;
1906         }
1907         idstr[size] = 0;
1908
1909         if (strncmp(vmsd->name, idstr, strlen(vmsd->name)) != 0) {
1910             /* it don't have a valid subsection name */
1911             return 0;
1912         }
1913         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
1914         if (sub_vmsd == NULL) {
1915             return -ENOENT;
1916         }
1917         qemu_file_skip(f, 1); /* subsection */
1918         qemu_file_skip(f, 1); /* len */
1919         qemu_file_skip(f, len); /* idstr */
1920         version_id = qemu_get_be32(f);
1921
1922         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1923         if (ret) {
1924             return ret;
1925         }
1926     }
1927     return 0;
1928 }
1929
1930 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1931                                     void *opaque)
1932 {
1933     const VMStateSubsection *sub = vmsd->subsections;
1934
1935     while (sub && sub->needed) {
1936         if (sub->needed(opaque)) {
1937             const VMStateDescription *vmsd = sub->vmsd;
1938             uint8_t len;
1939
1940             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1941             len = strlen(vmsd->name);
1942             qemu_put_byte(f, len);
1943             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1944             qemu_put_be32(f, vmsd->version_id);
1945             vmstate_save_state(f, vmsd, opaque);
1946         }
1947         sub++;
1948     }
1949 }
1950
1951 typedef struct LoadStateEntry {
1952     QLIST_ENTRY(LoadStateEntry) entry;
1953     SaveStateEntry *se;
1954     int section_id;
1955     int version_id;
1956 } LoadStateEntry;
1957
1958 int qemu_loadvm_state(QEMUFile *f)
1959 {
1960     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1961         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1962     LoadStateEntry *le, *new_le;
1963     uint8_t section_type;
1964     unsigned int v;
1965     int ret;
1966
1967     if (qemu_savevm_state_blocked(NULL)) {
1968         return -EINVAL;
1969     }
1970
1971     v = qemu_get_be32(f);
1972     if (v != QEMU_VM_FILE_MAGIC)
1973         return -EINVAL;
1974
1975     v = qemu_get_be32(f);
1976     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1977         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1978         return -ENOTSUP;
1979     }
1980     if (v != QEMU_VM_FILE_VERSION)
1981         return -ENOTSUP;
1982
1983     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1984         uint32_t instance_id, version_id, section_id;
1985         SaveStateEntry *se;
1986         char idstr[257];
1987         int len;
1988
1989         switch (section_type) {
1990         case QEMU_VM_SECTION_START:
1991         case QEMU_VM_SECTION_FULL:
1992             /* Read section start */
1993             section_id = qemu_get_be32(f);
1994             len = qemu_get_byte(f);
1995             qemu_get_buffer(f, (uint8_t *)idstr, len);
1996             idstr[len] = 0;
1997             instance_id = qemu_get_be32(f);
1998             version_id = qemu_get_be32(f);
1999
2000             /* Find savevm section */
2001             se = find_se(idstr, instance_id);
2002             if (se == NULL) {
2003                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
2004                 ret = -EINVAL;
2005                 goto out;
2006             }
2007
2008             /* Validate version */
2009             if (version_id > se->version_id) {
2010                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
2011                         version_id, idstr, se->version_id);
2012                 ret = -EINVAL;
2013                 goto out;
2014             }
2015
2016             /* Add entry */
2017             le = g_malloc0(sizeof(*le));
2018
2019             le->se = se;
2020             le->section_id = section_id;
2021             le->version_id = version_id;
2022             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
2023
2024             ret = vmstate_load(f, le->se, le->version_id);
2025             if (ret < 0) {
2026                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
2027                         instance_id, idstr);
2028                 goto out;
2029             }
2030             break;
2031         case QEMU_VM_SECTION_PART:
2032         case QEMU_VM_SECTION_END:
2033             section_id = qemu_get_be32(f);
2034
2035             QLIST_FOREACH(le, &loadvm_handlers, entry) {
2036                 if (le->section_id == section_id) {
2037                     break;
2038                 }
2039             }
2040             if (le == NULL) {
2041                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
2042                 ret = -EINVAL;
2043                 goto out;
2044             }
2045
2046             ret = vmstate_load(f, le->se, le->version_id);
2047             if (ret < 0) {
2048                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
2049                         section_id);
2050                 goto out;
2051             }
2052             break;
2053         default:
2054             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
2055             ret = -EINVAL;
2056             goto out;
2057         }
2058     }
2059
2060     cpu_synchronize_all_post_init();
2061
2062     ret = 0;
2063
2064 out:
2065     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
2066         QLIST_REMOVE(le, entry);
2067         g_free(le);
2068     }
2069
2070     if (ret == 0) {
2071         ret = qemu_file_get_error(f);
2072     }
2073
2074     return ret;
2075 }
2076
2077 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
2078                               const char *name)
2079 {
2080     QEMUSnapshotInfo *sn_tab, *sn;
2081     int nb_sns, i, ret;
2082
2083     ret = -ENOENT;
2084     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2085     if (nb_sns < 0)
2086         return ret;
2087     for(i = 0; i < nb_sns; i++) {
2088         sn = &sn_tab[i];
2089         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
2090             *sn_info = *sn;
2091             ret = 0;
2092             break;
2093         }
2094     }
2095     g_free(sn_tab);
2096     return ret;
2097 }
2098
2099 /*
2100  * Deletes snapshots of a given name in all opened images.
2101  */
2102 static int del_existing_snapshots(Monitor *mon, const char *name)
2103 {
2104     BlockDriverState *bs;
2105     QEMUSnapshotInfo sn1, *snapshot = &sn1;
2106     int ret;
2107
2108     bs = NULL;
2109     while ((bs = bdrv_next(bs))) {
2110         if (bdrv_can_snapshot(bs) &&
2111             bdrv_snapshot_find(bs, snapshot, name) >= 0)
2112         {
2113             ret = bdrv_snapshot_delete(bs, name);
2114             if (ret < 0) {
2115                 monitor_printf(mon,
2116                                "Error while deleting snapshot on '%s'\n",
2117                                bdrv_get_device_name(bs));
2118                 return -1;
2119             }
2120         }
2121     }
2122
2123     return 0;
2124 }
2125
2126 void do_savevm(Monitor *mon, const QDict *qdict)
2127 {
2128     BlockDriverState *bs, *bs1;
2129     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
2130     int ret;
2131     QEMUFile *f;
2132     int saved_vm_running;
2133     uint64_t vm_state_size;
2134 #ifdef _WIN32
2135     struct _timeb tb;
2136     struct tm *ptm;
2137 #else
2138     struct timeval tv;
2139     struct tm tm;
2140 #endif
2141     const char *name = qdict_get_try_str(qdict, "name");
2142
2143     /* Verify if there is a device that doesn't support snapshots and is writable */
2144     bs = NULL;
2145     while ((bs = bdrv_next(bs))) {
2146
2147         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2148             continue;
2149         }
2150
2151         if (!bdrv_can_snapshot(bs)) {
2152             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
2153                                bdrv_get_device_name(bs));
2154             return;
2155         }
2156     }
2157
2158     bs = bdrv_snapshots();
2159     if (!bs) {
2160         monitor_printf(mon, "No block device can accept snapshots\n");
2161         return;
2162     }
2163
2164     saved_vm_running = runstate_is_running();
2165     vm_stop(RUN_STATE_SAVE_VM);
2166
2167     memset(sn, 0, sizeof(*sn));
2168
2169     /* fill auxiliary fields */
2170 #ifdef _WIN32
2171     _ftime(&tb);
2172     sn->date_sec = tb.time;
2173     sn->date_nsec = tb.millitm * 1000000;
2174 #else
2175     gettimeofday(&tv, NULL);
2176     sn->date_sec = tv.tv_sec;
2177     sn->date_nsec = tv.tv_usec * 1000;
2178 #endif
2179     sn->vm_clock_nsec = qemu_get_clock_ns(vm_clock);
2180
2181     if (name) {
2182         ret = bdrv_snapshot_find(bs, old_sn, name);
2183         if (ret >= 0) {
2184             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
2185             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
2186         } else {
2187             pstrcpy(sn->name, sizeof(sn->name), name);
2188         }
2189     } else {
2190 #ifdef _WIN32
2191         time_t t = tb.time;
2192         ptm = localtime(&t);
2193         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", ptm);
2194 #else
2195         /* cast below needed for OpenBSD where tv_sec is still 'long' */
2196         localtime_r((const time_t *)&tv.tv_sec, &tm);
2197         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
2198 #endif
2199     }
2200
2201     /* Delete old snapshots of the same name */
2202     if (name && del_existing_snapshots(mon, name) < 0) {
2203         goto the_end;
2204     }
2205
2206     /* save the VM state */
2207     f = qemu_fopen_bdrv(bs, 1);
2208     if (!f) {
2209         monitor_printf(mon, "Could not open VM state file\n");
2210         goto the_end;
2211     }
2212     ret = qemu_savevm_state(f);
2213     vm_state_size = qemu_ftell(f);
2214     qemu_fclose(f);
2215     if (ret < 0) {
2216         monitor_printf(mon, "Error %d while writing VM\n", ret);
2217         goto the_end;
2218     }
2219
2220     /* create the snapshots */
2221
2222     bs1 = NULL;
2223     while ((bs1 = bdrv_next(bs1))) {
2224         if (bdrv_can_snapshot(bs1)) {
2225             /* Write VM state size only to the image that contains the state */
2226             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
2227             ret = bdrv_snapshot_create(bs1, sn);
2228             if (ret < 0) {
2229                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
2230                                bdrv_get_device_name(bs1));
2231             }
2232         }
2233     }
2234
2235  the_end:
2236     if (saved_vm_running)
2237         vm_start();
2238 }
2239
2240 void qmp_xen_save_devices_state(const char *filename, Error **errp)
2241 {
2242     QEMUFile *f;
2243     int saved_vm_running;
2244     int ret;
2245
2246     saved_vm_running = runstate_is_running();
2247     vm_stop(RUN_STATE_SAVE_VM);
2248
2249     f = qemu_fopen(filename, "wb");
2250     if (!f) {
2251         error_set(errp, QERR_OPEN_FILE_FAILED, filename);
2252         goto the_end;
2253     }
2254     ret = qemu_save_device_state(f);
2255     qemu_fclose(f);
2256     if (ret < 0) {
2257         error_set(errp, QERR_IO_ERROR);
2258     }
2259
2260  the_end:
2261     if (saved_vm_running)
2262         vm_start();
2263 }
2264
2265 int load_vmstate(const char *name)
2266 {
2267     BlockDriverState *bs, *bs_vm_state;
2268     QEMUSnapshotInfo sn;
2269     QEMUFile *f;
2270     int ret;
2271
2272     bs_vm_state = bdrv_snapshots();
2273     if (!bs_vm_state) {
2274         error_report("No block device supports snapshots");
2275         return -ENOTSUP;
2276     }
2277
2278     /* Don't even try to load empty VM states */
2279     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
2280     if (ret < 0) {
2281         return ret;
2282     } else if (sn.vm_state_size == 0) {
2283         error_report("This is a disk-only snapshot. Revert to it offline "
2284             "using qemu-img.");
2285         return -EINVAL;
2286     }
2287
2288     /* Verify if there is any device that doesn't support snapshots and is
2289     writable and check if the requested snapshot is available too. */
2290     bs = NULL;
2291     while ((bs = bdrv_next(bs))) {
2292
2293         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2294             continue;
2295         }
2296
2297         if (!bdrv_can_snapshot(bs)) {
2298             error_report("Device '%s' is writable but does not support snapshots.",
2299                                bdrv_get_device_name(bs));
2300             return -ENOTSUP;
2301         }
2302
2303         ret = bdrv_snapshot_find(bs, &sn, name);
2304         if (ret < 0) {
2305             error_report("Device '%s' does not have the requested snapshot '%s'",
2306                            bdrv_get_device_name(bs), name);
2307             return ret;
2308         }
2309     }
2310
2311     /* Flush all IO requests so they don't interfere with the new state.  */
2312     bdrv_drain_all();
2313
2314     bs = NULL;
2315     while ((bs = bdrv_next(bs))) {
2316         if (bdrv_can_snapshot(bs)) {
2317             ret = bdrv_snapshot_goto(bs, name);
2318             if (ret < 0) {
2319                 error_report("Error %d while activating snapshot '%s' on '%s'",
2320                              ret, name, bdrv_get_device_name(bs));
2321                 return ret;
2322             }
2323         }
2324     }
2325
2326     /* restore the VM state */
2327     f = qemu_fopen_bdrv(bs_vm_state, 0);
2328     if (!f) {
2329         error_report("Could not open VM state file");
2330         return -EINVAL;
2331     }
2332
2333     qemu_system_reset(VMRESET_SILENT);
2334     ret = qemu_loadvm_state(f);
2335
2336     qemu_fclose(f);
2337     if (ret < 0) {
2338         error_report("Error %d while loading VM state", ret);
2339         return ret;
2340     }
2341
2342     return 0;
2343 }
2344
2345 void do_delvm(Monitor *mon, const QDict *qdict)
2346 {
2347     BlockDriverState *bs, *bs1;
2348     int ret;
2349     const char *name = qdict_get_str(qdict, "name");
2350
2351     bs = bdrv_snapshots();
2352     if (!bs) {
2353         monitor_printf(mon, "No block device supports snapshots\n");
2354         return;
2355     }
2356
2357     bs1 = NULL;
2358     while ((bs1 = bdrv_next(bs1))) {
2359         if (bdrv_can_snapshot(bs1)) {
2360             ret = bdrv_snapshot_delete(bs1, name);
2361             if (ret < 0) {
2362                 if (ret == -ENOTSUP)
2363                     monitor_printf(mon,
2364                                    "Snapshots not supported on device '%s'\n",
2365                                    bdrv_get_device_name(bs1));
2366                 else
2367                     monitor_printf(mon, "Error %d while deleting snapshot on "
2368                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2369             }
2370         }
2371     }
2372 }
2373
2374 void do_info_snapshots(Monitor *mon)
2375 {
2376     BlockDriverState *bs, *bs1;
2377     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2378     int nb_sns, i, ret, available;
2379     int total;
2380     int *available_snapshots;
2381     char buf[256];
2382
2383     bs = bdrv_snapshots();
2384     if (!bs) {
2385         monitor_printf(mon, "No available block device supports snapshots\n");
2386         return;
2387     }
2388
2389     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2390     if (nb_sns < 0) {
2391         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2392         return;
2393     }
2394
2395     if (nb_sns == 0) {
2396         monitor_printf(mon, "There is no snapshot available.\n");
2397         return;
2398     }
2399
2400     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
2401     total = 0;
2402     for (i = 0; i < nb_sns; i++) {
2403         sn = &sn_tab[i];
2404         available = 1;
2405         bs1 = NULL;
2406
2407         while ((bs1 = bdrv_next(bs1))) {
2408             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2409                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2410                 if (ret < 0) {
2411                     available = 0;
2412                     break;
2413                 }
2414             }
2415         }
2416
2417         if (available) {
2418             available_snapshots[total] = i;
2419             total++;
2420         }
2421     }
2422
2423     if (total > 0) {
2424         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2425         for (i = 0; i < total; i++) {
2426             sn = &sn_tab[available_snapshots[i]];
2427             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2428         }
2429     } else {
2430         monitor_printf(mon, "There is no suitable snapshot available\n");
2431     }
2432
2433     g_free(sn_tab);
2434     g_free(available_snapshots);
2435
2436 }
2437
2438 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
2439 {
2440     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
2441                        memory_region_name(mr), dev);
2442 }
2443
2444 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
2445 {
2446     /* Nothing do to while the implementation is in RAMBlock */
2447 }
2448
2449 void vmstate_register_ram_global(MemoryRegion *mr)
2450 {
2451     vmstate_register_ram(mr, NULL);
2452 }
2453
2454 /*
2455   page = zrun nzrun
2456        | zrun nzrun page
2457
2458   zrun = length
2459
2460   nzrun = length byte...
2461
2462   length = uleb128 encoded integer
2463  */
2464 int xbzrle_encode_buffer(uint8_t *old_buf, uint8_t *new_buf, int slen,
2465                          uint8_t *dst, int dlen)
2466 {
2467     uint32_t zrun_len = 0, nzrun_len = 0;
2468     int d = 0, i = 0;
2469     long res, xor;
2470     uint8_t *nzrun_start = NULL;
2471
2472     g_assert(!(((uintptr_t)old_buf | (uintptr_t)new_buf | slen) %
2473                sizeof(long)));
2474
2475     while (i < slen) {
2476         /* overflow */
2477         if (d + 2 > dlen) {
2478             return -1;
2479         }
2480
2481         /* not aligned to sizeof(long) */
2482         res = (slen - i) % sizeof(long);
2483         while (res && old_buf[i] == new_buf[i]) {
2484             zrun_len++;
2485             i++;
2486             res--;
2487         }
2488
2489         /* word at a time for speed */
2490         if (!res) {
2491             while (i < slen &&
2492                    (*(long *)(old_buf + i)) == (*(long *)(new_buf + i))) {
2493                 i += sizeof(long);
2494                 zrun_len += sizeof(long);
2495             }
2496
2497             /* go over the rest */
2498             while (i < slen && old_buf[i] == new_buf[i]) {
2499                 zrun_len++;
2500                 i++;
2501             }
2502         }
2503
2504         /* buffer unchanged */
2505         if (zrun_len == slen) {
2506             return 0;
2507         }
2508
2509         /* skip last zero run */
2510         if (i == slen) {
2511             return d;
2512         }
2513
2514         d += uleb128_encode_small(dst + d, zrun_len);
2515
2516         zrun_len = 0;
2517         nzrun_start = new_buf + i;
2518
2519         /* overflow */
2520         if (d + 2 > dlen) {
2521             return -1;
2522         }
2523         /* not aligned to sizeof(long) */
2524         res = (slen - i) % sizeof(long);
2525         while (res && old_buf[i] != new_buf[i]) {
2526             i++;
2527             nzrun_len++;
2528             res--;
2529         }
2530
2531         /* word at a time for speed, use of 32-bit long okay */
2532         if (!res) {
2533             /* truncation to 32-bit long okay */
2534             long mask = (long)0x0101010101010101ULL;
2535             while (i < slen) {
2536                 xor = *(long *)(old_buf + i) ^ *(long *)(new_buf + i);
2537                 if ((xor - mask) & ~xor & (mask << 7)) {
2538                     /* found the end of an nzrun within the current long */
2539                     while (old_buf[i] != new_buf[i]) {
2540                         nzrun_len++;
2541                         i++;
2542                     }
2543                     break;
2544                 } else {
2545                     i += sizeof(long);
2546                     nzrun_len += sizeof(long);
2547                 }
2548             }
2549         }
2550
2551         d += uleb128_encode_small(dst + d, nzrun_len);
2552         /* overflow */
2553         if (d + nzrun_len > dlen) {
2554             return -1;
2555         }
2556         memcpy(dst + d, nzrun_start, nzrun_len);
2557         d += nzrun_len;
2558         nzrun_len = 0;
2559     }
2560
2561     return d;
2562 }
2563
2564 int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen)
2565 {
2566     int i = 0, d = 0;
2567     int ret;
2568     uint32_t count = 0;
2569
2570     while (i < slen) {
2571
2572         /* zrun */
2573         if ((slen - i) < 2) {
2574             return -1;
2575         }
2576
2577         ret = uleb128_decode_small(src + i, &count);
2578         if (ret < 0 || (i && !count)) {
2579             return -1;
2580         }
2581         i += ret;
2582         d += count;
2583
2584         /* overflow */
2585         if (d > dlen) {
2586             return -1;
2587         }
2588
2589         /* nzrun */
2590         if ((slen - i) < 2) {
2591             return -1;
2592         }
2593
2594         ret = uleb128_decode_small(src + i, &count);
2595         if (ret < 0 || !count) {
2596             return -1;
2597         }
2598         i += ret;
2599
2600         /* overflow */
2601         if (d + count > dlen || i + count > slen) {
2602             return -1;
2603         }
2604
2605         memcpy(dst + d, src + i, count);
2606         d += count;
2607         i += count;
2608     }
2609
2610     return d;
2611 }