savevm.c

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