savevm.c

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