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 <signal.h>
  27 #include <time.h>
  28 #include <errno.h>
  29 #include <sys/time.h>
  30 #include <zlib.h>
  31
  32 /* Needed early for CONFIG_BSD etc. */
  33 #include "config-host.h"
  34
  35 #ifndef _WIN32
  36 #include <sys/times.h>
  37 #include <sys/wait.h>
  38 #include <termios.h>
  39 #include <sys/mman.h>
  40 #include <sys/ioctl.h>
  41 #include <sys/resource.h>
  42 #include <sys/socket.h>
  43 #include <netinet/in.h>
  44 #include <net/if.h>
  45 #include <arpa/inet.h>
  46 #include <dirent.h>
  47 #include <netdb.h>
  48 #include <sys/select.h>
  49 #ifdef CONFIG_BSD
  50 #include <sys/stat.h>
  51 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) || defined(__DragonFly__)
  52 #include <libutil.h>
  53 #else
  54 #include <util.h>
  55 #endif
  56 #ifdef __linux__
  57 #include <pty.h>
  58 #include <malloc.h>
  59 #include <linux/rtc.h>
  60 #endif
  61 #endif
  62 #endif
  63
  64 #ifdef _WIN32
  65 #include <windows.h>
  66 #include <malloc.h>
  67 #include <sys/timeb.h>
  68 #include <mmsystem.h>
  69 #define getopt_long_only getopt_long
  70 #define memalign(align, size) malloc(size)
  71 #endif
  72
  73 #include "qemu-common.h"
  74 #include "hw/hw.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 "block.h"
  81 #include "audio/audio.h"
  82 #include "migration.h"
  83 #include "qemu_socket.h"
  84 #include "qemu-queue.h"
  85
  86 /* point to the block driver where the snapshots are managed */
  87 static BlockDriverState *bs_snapshots;
  88
  89 #define SELF_ANNOUNCE_ROUNDS 5
  90
  91 #ifndef ETH_P_RARP
  92 #define ETH_P_RARP 0x8035
  93 #endif
  94 #define ARP_HTYPE_ETH 0x0001
  95 #define ARP_PTYPE_IP 0x0800
  96 #define ARP_OP_REQUEST_REV 0x3
  97
  98 static int announce_self_create(uint8_t *buf,
  99                                 uint8_t *mac_addr)
 100 {
 101     /* Ethernet header. */
 102     memset(buf, 0xff, 6);         /* destination MAC addr */
 103     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
 104     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
 105
 106     /* RARP header. */
 107     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
 108     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
 109     *(buf + 18) = 6; /* hardware addr length (ethernet) */
 110     *(buf + 19) = 4; /* protocol addr length (IPv4) */
 111     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
 112     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
 113     memset(buf + 28, 0x00, 4);     /* source protocol addr */
 114     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
 115     memset(buf + 38, 0x00, 4);     /* target protocol addr */
 116
 117     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
 118     memset(buf + 42, 0x00, 18);
 119
 120     return 60; /* len (FCS will be added by hardware) */
 121 }
 122
 123 static void qemu_announce_self_iter(NICState *nic, void *opaque)
 124 {
 125     uint8_t buf[60];
 126     int len;
 127
 128     len = announce_self_create(buf, nic->conf->macaddr.a);
 129
 130     qemu_send_packet_raw(&nic->nc, buf, len);
 131 }
 132
 133
 134 static void qemu_announce_self_once(void *opaque)
 135 {
 136     static int count = SELF_ANNOUNCE_ROUNDS;
 137     QEMUTimer *timer = *(QEMUTimer **)opaque;
 138
 139     qemu_foreach_nic(qemu_announce_self_iter, NULL);
 140
 141     if (--count) {
 142         /* delay 50ms, 150ms, 250ms, ... */
 143         qemu_mod_timer(timer, qemu_get_clock(rt_clock) +
 144                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
 145     } else {
 146             qemu_del_timer(timer);
 147             qemu_free_timer(timer);
 148     }
 149 }
 150
 151 void qemu_announce_self(void)
 152 {
 153         static QEMUTimer *timer;
 154         timer = qemu_new_timer(rt_clock, qemu_announce_self_once, &timer);
 155         qemu_announce_self_once(&timer);
 156 }
 157
 158 /***********************************************************/
 159 /* savevm/loadvm support */
 160
 161 #define IO_BUF_SIZE 32768
 162
 163 struct QEMUFile {
 164     QEMUFilePutBufferFunc *put_buffer;
 165     QEMUFileGetBufferFunc *get_buffer;
 166     QEMUFileCloseFunc *close;
 167     QEMUFileRateLimit *rate_limit;
 168     QEMUFileSetRateLimit *set_rate_limit;
 169     QEMUFileGetRateLimit *get_rate_limit;
 170     void *opaque;
 171     int is_write;
 172
 173     int64_t buf_offset; /* start of buffer when writing, end of buffer
 174                            when reading */
 175     int buf_index;
 176     int buf_size; /* 0 when writing */
 177     uint8_t buf[IO_BUF_SIZE];
 178
 179     int has_error;
 180 };
 181
 182 typedef struct QEMUFileStdio
 183 {
 184     FILE *stdio_file;
 185     QEMUFile *file;
 186 } QEMUFileStdio;
 187
 188 typedef struct QEMUFileSocket
 189 {
 190     int fd;
 191     QEMUFile *file;
 192 } QEMUFileSocket;
 193
 194 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 195 {
 196     QEMUFileSocket *s = opaque;
 197     ssize_t len;
 198
 199     do {
 200         len = recv(s->fd, (void *)buf, size, 0);
 201     } while (len == -1 && socket_error() == EINTR);
 202
 203     if (len == -1)
 204         len = -socket_error();
 205
 206     return len;
 207 }
 208
 209 static int socket_close(void *opaque)
 210 {
 211     QEMUFileSocket *s = opaque;
 212     qemu_free(s);
 213     return 0;
 214 }
 215
 216 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 217 {
 218     QEMUFileStdio *s = opaque;
 219     return fwrite(buf, 1, size, s->stdio_file);
 220 }
 221
 222 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 223 {
 224     QEMUFileStdio *s = opaque;
 225     FILE *fp = s->stdio_file;
 226     int bytes;
 227
 228     do {
 229         clearerr(fp);
 230         bytes = fread(buf, 1, size, fp);
 231     } while ((bytes == 0) && ferror(fp) && (errno == EINTR));
 232     return bytes;
 233 }
 234
 235 static int stdio_pclose(void *opaque)
 236 {
 237     QEMUFileStdio *s = opaque;
 238     pclose(s->stdio_file);
 239     qemu_free(s);
 240     return 0;
 241 }
 242
 243 static int stdio_fclose(void *opaque)
 244 {
 245     QEMUFileStdio *s = opaque;
 246     fclose(s->stdio_file);
 247     qemu_free(s);
 248     return 0;
 249 }
 250
 251 QEMUFile *qemu_popen(FILE *stdio_file, const char *mode)
 252 {
 253     QEMUFileStdio *s;
 254
 255     if (stdio_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 256         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 257         return NULL;
 258     }
 259
 260     s = qemu_mallocz(sizeof(QEMUFileStdio));
 261
 262     s->stdio_file = stdio_file;
 263
 264     if(mode[0] == 'r') {
 265         s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_pclose,
 266                                  NULL, NULL, NULL);
 267     } else {
 268         s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_pclose,
 269                                  NULL, NULL, NULL);
 270     }
 271     return s->file;
 272 }
 273
 274 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 275 {
 276     FILE *popen_file;
 277
 278     popen_file = popen(command, mode);
 279     if(popen_file == NULL) {
 280         return NULL;
 281     }
 282
 283     return qemu_popen(popen_file, mode);
 284 }
 285
 286 int qemu_stdio_fd(QEMUFile *f)
 287 {
 288     QEMUFileStdio *p;
 289     int fd;
 290
 291     p = (QEMUFileStdio *)f->opaque;
 292     fd = fileno(p->stdio_file);
 293
 294     return fd;
 295 }
 296
 297 QEMUFile *qemu_fdopen(int fd, const char *mode)
 298 {
 299     QEMUFileStdio *s;
 300
 301     if (mode == NULL ||
 302         (mode[0] != 'r' && mode[0] != 'w') ||
 303         mode[1] != 'b' || mode[2] != 0) {
 304         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 305         return NULL;
 306     }
 307
 308     s = qemu_mallocz(sizeof(QEMUFileStdio));
 309     s->stdio_file = fdopen(fd, mode);
 310     if (!s->stdio_file)
 311         goto fail;
 312
 313     if(mode[0] == 'r') {
 314         s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_fclose,
 315                                  NULL, NULL, NULL);
 316     } else {
 317         s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_fclose,
 318                                  NULL, NULL, NULL);
 319     }
 320     return s->file;
 321
 322 fail:
 323     qemu_free(s);
 324     return NULL;
 325 }
 326
 327 QEMUFile *qemu_fopen_socket(int fd)
 328 {
 329     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 330
 331     s->fd = fd;
 332     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close,
 333                              NULL, NULL, NULL);
 334     return s->file;
 335 }
 336
 337 static int file_put_buffer(void *opaque, const uint8_t *buf,
 338                             int64_t pos, int size)
 339 {
 340     QEMUFileStdio *s = opaque;
 341     fseek(s->stdio_file, pos, SEEK_SET);
 342     return fwrite(buf, 1, size, s->stdio_file);
 343 }
 344
 345 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 346 {
 347     QEMUFileStdio *s = opaque;
 348     fseek(s->stdio_file, pos, SEEK_SET);
 349     return fread(buf, 1, size, s->stdio_file);
 350 }
 351
 352 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 353 {
 354     QEMUFileStdio *s;
 355
 356     if (mode == NULL ||
 357         (mode[0] != 'r' && mode[0] != 'w') ||
 358         mode[1] != 'b' || mode[2] != 0) {
 359         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 360         return NULL;
 361     }
 362
 363     s = qemu_mallocz(sizeof(QEMUFileStdio));
 364
 365     s->stdio_file = fopen(filename, mode);
 366     if (!s->stdio_file)
 367         goto fail;
 368
 369     if(mode[0] == 'w') {
 370         s->file = qemu_fopen_ops(s, file_put_buffer, NULL, stdio_fclose,
 371                                  NULL, NULL, NULL);
 372     } else {
 373         s->file = qemu_fopen_ops(s, NULL, file_get_buffer, stdio_fclose,
 374                                NULL, NULL, NULL);
 375     }
 376     return s->file;
 377 fail:
 378     qemu_free(s);
 379     return NULL;
 380 }
 381
 382 static int block_put_buffer(void *opaque, const uint8_t *buf,
 383                            int64_t pos, int size)
 384 {
 385     bdrv_save_vmstate(opaque, buf, pos, size);
 386     return size;
 387 }
 388
 389 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 390 {
 391     return bdrv_load_vmstate(opaque, buf, pos, size);
 392 }
 393
 394 static int bdrv_fclose(void *opaque)
 395 {
 396     return 0;
 397 }
 398
 399 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 400 {
 401     if (is_writable)
 402         return qemu_fopen_ops(bs, block_put_buffer, NULL, bdrv_fclose,
 403                               NULL, NULL, NULL);
 404     return qemu_fopen_ops(bs, NULL, block_get_buffer, bdrv_fclose, NULL, NULL, NULL);
 405 }
 406
 407 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 408                          QEMUFileGetBufferFunc *get_buffer,
 409                          QEMUFileCloseFunc *close,
 410                          QEMUFileRateLimit *rate_limit,
 411                          QEMUFileSetRateLimit *set_rate_limit,
 412                          QEMUFileGetRateLimit *get_rate_limit)
 413 {
 414     QEMUFile *f;
 415
 416     f = qemu_mallocz(sizeof(QEMUFile));
 417
 418     f->opaque = opaque;
 419     f->put_buffer = put_buffer;
 420     f->get_buffer = get_buffer;
 421     f->close = close;
 422     f->rate_limit = rate_limit;
 423     f->set_rate_limit = set_rate_limit;
 424     f->get_rate_limit = get_rate_limit;
 425     f->is_write = 0;
 426
 427     return f;
 428 }
 429
 430 int qemu_file_has_error(QEMUFile *f)
 431 {
 432     return f->has_error;
 433 }
 434
 435 void qemu_file_set_error(QEMUFile *f)
 436 {
 437     f->has_error = 1;
 438 }
 439
 440 void qemu_fflush(QEMUFile *f)
 441 {
 442     if (!f->put_buffer)
 443         return;
 444
 445     if (f->is_write && f->buf_index > 0) {
 446         int len;
 447
 448         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 449         if (len > 0)
 450             f->buf_offset += f->buf_index;
 451         else
 452             f->has_error = 1;
 453         f->buf_index = 0;
 454     }
 455 }
 456
 457 static void qemu_fill_buffer(QEMUFile *f)
 458 {
 459     int len;
 460
 461     if (!f->get_buffer)
 462         return;
 463
 464     if (f->is_write)
 465         abort();
 466
 467     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 468     if (len > 0) {
 469         f->buf_index = 0;
 470         f->buf_size = len;
 471         f->buf_offset += len;
 472     } else if (len != -EAGAIN)
 473         f->has_error = 1;
 474 }
 475
 476 int qemu_fclose(QEMUFile *f)
 477 {
 478     int ret = 0;
 479     qemu_fflush(f);
 480     if (f->close)
 481         ret = f->close(f->opaque);
 482     qemu_free(f);
 483     return ret;
 484 }
 485
 486 void qemu_file_put_notify(QEMUFile *f)
 487 {
 488     f->put_buffer(f->opaque, NULL, 0, 0);
 489 }
 490
 491 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 492 {
 493     int l;
 494
 495     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 496         fprintf(stderr,
 497                 "Attempted to write to buffer while read buffer is not empty\n");
 498         abort();
 499     }
 500
 501     while (!f->has_error && size > 0) {
 502         l = IO_BUF_SIZE - f->buf_index;
 503         if (l > size)
 504             l = size;
 505         memcpy(f->buf + f->buf_index, buf, l);
 506         f->is_write = 1;
 507         f->buf_index += l;
 508         buf += l;
 509         size -= l;
 510         if (f->buf_index >= IO_BUF_SIZE)
 511             qemu_fflush(f);
 512     }
 513 }
 514
 515 void qemu_put_byte(QEMUFile *f, int v)
 516 {
 517     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 518         fprintf(stderr,
 519                 "Attempted to write to buffer while read buffer is not empty\n");
 520         abort();
 521     }
 522
 523     f->buf[f->buf_index++] = v;
 524     f->is_write = 1;
 525     if (f->buf_index >= IO_BUF_SIZE)
 526         qemu_fflush(f);
 527 }
 528
 529 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 530 {
 531     int size, l;
 532
 533     if (f->is_write)
 534         abort();
 535
 536     size = size1;
 537     while (size > 0) {
 538         l = f->buf_size - f->buf_index;
 539         if (l == 0) {
 540             qemu_fill_buffer(f);
 541             l = f->buf_size - f->buf_index;
 542             if (l == 0)
 543                 break;
 544         }
 545         if (l > size)
 546             l = size;
 547         memcpy(buf, f->buf + f->buf_index, l);
 548         f->buf_index += l;
 549         buf += l;
 550         size -= l;
 551     }
 552     return size1 - size;
 553 }
 554
 555 int qemu_get_byte(QEMUFile *f)
 556 {
 557     if (f->is_write)
 558         abort();
 559
 560     if (f->buf_index >= f->buf_size) {
 561         qemu_fill_buffer(f);
 562         if (f->buf_index >= f->buf_size)
 563             return 0;
 564     }
 565     return f->buf[f->buf_index++];
 566 }
 567
 568 int64_t qemu_ftell(QEMUFile *f)
 569 {
 570     return f->buf_offset - f->buf_size + f->buf_index;
 571 }
 572
 573 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 574 {
 575     if (whence == SEEK_SET) {
 576         /* nothing to do */
 577     } else if (whence == SEEK_CUR) {
 578         pos += qemu_ftell(f);
 579     } else {
 580         /* SEEK_END not supported */
 581         return -1;
 582     }
 583     if (f->put_buffer) {
 584         qemu_fflush(f);
 585         f->buf_offset = pos;
 586     } else {
 587         f->buf_offset = pos;
 588         f->buf_index = 0;
 589         f->buf_size = 0;
 590     }
 591     return pos;
 592 }
 593
 594 int qemu_file_rate_limit(QEMUFile *f)
 595 {
 596     if (f->rate_limit)
 597         return f->rate_limit(f->opaque);
 598
 599     return 0;
 600 }
 601
 602 size_t qemu_file_get_rate_limit(QEMUFile *f)
 603 {
 604     if (f->get_rate_limit)
 605         return f->get_rate_limit(f->opaque);
 606
 607     return 0;
 608 }
 609
 610 size_t qemu_file_set_rate_limit(QEMUFile *f, size_t new_rate)
 611 {
 612     /* any failed or completed migration keeps its state to allow probing of
 613      * migration data, but has no associated file anymore */
 614     if (f && f->set_rate_limit)
 615         return f->set_rate_limit(f->opaque, new_rate);
 616
 617     return 0;
 618 }
 619
 620 void qemu_put_be16(QEMUFile *f, unsigned int v)
 621 {
 622     qemu_put_byte(f, v >> 8);
 623     qemu_put_byte(f, v);
 624 }
 625
 626 void qemu_put_be32(QEMUFile *f, unsigned int v)
 627 {
 628     qemu_put_byte(f, v >> 24);
 629     qemu_put_byte(f, v >> 16);
 630     qemu_put_byte(f, v >> 8);
 631     qemu_put_byte(f, v);
 632 }
 633
 634 void qemu_put_be64(QEMUFile *f, uint64_t v)
 635 {
 636     qemu_put_be32(f, v >> 32);
 637     qemu_put_be32(f, v);
 638 }
 639
 640 unsigned int qemu_get_be16(QEMUFile *f)
 641 {
 642     unsigned int v;
 643     v = qemu_get_byte(f) << 8;
 644     v |= qemu_get_byte(f);
 645     return v;
 646 }
 647
 648 unsigned int qemu_get_be32(QEMUFile *f)
 649 {
 650     unsigned int v;
 651     v = qemu_get_byte(f) << 24;
 652     v |= qemu_get_byte(f) << 16;
 653     v |= qemu_get_byte(f) << 8;
 654     v |= qemu_get_byte(f);
 655     return v;
 656 }
 657
 658 uint64_t qemu_get_be64(QEMUFile *f)
 659 {
 660     uint64_t v;
 661     v = (uint64_t)qemu_get_be32(f) << 32;
 662     v |= qemu_get_be32(f);
 663     return v;
 664 }
 665
 666 /* 8 bit int */
 667
 668 static int get_int8(QEMUFile *f, void *pv, size_t size)
 669 {
 670     int8_t *v = pv;
 671     qemu_get_s8s(f, v);
 672     return 0;
 673 }
 674
 675 static void put_int8(QEMUFile *f, void *pv, size_t size)
 676 {
 677     int8_t *v = pv;
 678     qemu_put_s8s(f, v);
 679 }
 680
 681 const VMStateInfo vmstate_info_int8 = {
 682     .name = "int8",
 683     .get  = get_int8,
 684     .put  = put_int8,
 685 };
 686
 687 /* 16 bit int */
 688
 689 static int get_int16(QEMUFile *f, void *pv, size_t size)
 690 {
 691     int16_t *v = pv;
 692     qemu_get_sbe16s(f, v);
 693     return 0;
 694 }
 695
 696 static void put_int16(QEMUFile *f, void *pv, size_t size)
 697 {
 698     int16_t *v = pv;
 699     qemu_put_sbe16s(f, v);
 700 }
 701
 702 const VMStateInfo vmstate_info_int16 = {
 703     .name = "int16",
 704     .get  = get_int16,
 705     .put  = put_int16,
 706 };
 707
 708 /* 32 bit int */
 709
 710 static int get_int32(QEMUFile *f, void *pv, size_t size)
 711 {
 712     int32_t *v = pv;
 713     qemu_get_sbe32s(f, v);
 714     return 0;
 715 }
 716
 717 static void put_int32(QEMUFile *f, void *pv, size_t size)
 718 {
 719     int32_t *v = pv;
 720     qemu_put_sbe32s(f, v);
 721 }
 722
 723 const VMStateInfo vmstate_info_int32 = {
 724     .name = "int32",
 725     .get  = get_int32,
 726     .put  = put_int32,
 727 };
 728
 729 /* 32 bit int. See that the received value is the same than the one
 730    in the field */
 731
 732 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 733 {
 734     int32_t *v = pv;
 735     int32_t v2;
 736     qemu_get_sbe32s(f, &v2);
 737
 738     if (*v == v2)
 739         return 0;
 740     return -EINVAL;
 741 }
 742
 743 const VMStateInfo vmstate_info_int32_equal = {
 744     .name = "int32 equal",
 745     .get  = get_int32_equal,
 746     .put  = put_int32,
 747 };
 748
 749 /* 32 bit int. See that the received value is the less or the same
 750    than the one in the field */
 751
 752 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 753 {
 754     int32_t *old = pv;
 755     int32_t new;
 756     qemu_get_sbe32s(f, &new);
 757
 758     if (*old <= new)
 759         return 0;
 760     return -EINVAL;
 761 }
 762
 763 const VMStateInfo vmstate_info_int32_le = {
 764     .name = "int32 equal",
 765     .get  = get_int32_le,
 766     .put  = put_int32,
 767 };
 768
 769 /* 64 bit int */
 770
 771 static int get_int64(QEMUFile *f, void *pv, size_t size)
 772 {
 773     int64_t *v = pv;
 774     qemu_get_sbe64s(f, v);
 775     return 0;
 776 }
 777
 778 static void put_int64(QEMUFile *f, void *pv, size_t size)
 779 {
 780     int64_t *v = pv;
 781     qemu_put_sbe64s(f, v);
 782 }
 783
 784 const VMStateInfo vmstate_info_int64 = {
 785     .name = "int64",
 786     .get  = get_int64,
 787     .put  = put_int64,
 788 };
 789
 790 /* 8 bit unsigned int */
 791
 792 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 793 {
 794     uint8_t *v = pv;
 795     qemu_get_8s(f, v);
 796     return 0;
 797 }
 798
 799 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 800 {
 801     uint8_t *v = pv;
 802     qemu_put_8s(f, v);
 803 }
 804
 805 const VMStateInfo vmstate_info_uint8 = {
 806     .name = "uint8",
 807     .get  = get_uint8,
 808     .put  = put_uint8,
 809 };
 810
 811 /* 16 bit unsigned int */
 812
 813 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 814 {
 815     uint16_t *v = pv;
 816     qemu_get_be16s(f, v);
 817     return 0;
 818 }
 819
 820 static void put_uint16(QEMUFile *f, void *pv, size_t size)
 821 {
 822     uint16_t *v = pv;
 823     qemu_put_be16s(f, v);
 824 }
 825
 826 const VMStateInfo vmstate_info_uint16 = {
 827     .name = "uint16",
 828     .get  = get_uint16,
 829     .put  = put_uint16,
 830 };
 831
 832 /* 32 bit unsigned int */
 833
 834 static int get_uint32(QEMUFile *f, void *pv, size_t size)
 835 {
 836     uint32_t *v = pv;
 837     qemu_get_be32s(f, v);
 838     return 0;
 839 }
 840
 841 static void put_uint32(QEMUFile *f, void *pv, size_t size)
 842 {
 843     uint32_t *v = pv;
 844     qemu_put_be32s(f, v);
 845 }
 846
 847 const VMStateInfo vmstate_info_uint32 = {
 848     .name = "uint32",
 849     .get  = get_uint32,
 850     .put  = put_uint32,
 851 };
 852
 853 /* 64 bit unsigned int */
 854
 855 static int get_uint64(QEMUFile *f, void *pv, size_t size)
 856 {
 857     uint64_t *v = pv;
 858     qemu_get_be64s(f, v);
 859     return 0;
 860 }
 861
 862 static void put_uint64(QEMUFile *f, void *pv, size_t size)
 863 {
 864     uint64_t *v = pv;
 865     qemu_put_be64s(f, v);
 866 }
 867
 868 const VMStateInfo vmstate_info_uint64 = {
 869     .name = "uint64",
 870     .get  = get_uint64,
 871     .put  = put_uint64,
 872 };
 873
 874 /* 64 bit linux kernel unsigned int */
 875
 876 #ifdef __linux__
 877 static int get_u64(QEMUFile *f, void *pv, size_t size)
 878 {
 879     uint64_t *v = pv;
 880     qemu_get_be64s(f, v);
 881     return 0;
 882 }
 883
 884 static void put_u64(QEMUFile *f, void *pv, size_t size)
 885 {
 886     uint64_t *v = pv;
 887     qemu_put_be64s(f, v);
 888 }
 889
 890 const VMStateInfo vmstate_info_u64 = {
 891     .name = "__u64",
 892     .get  = get_u64,
 893     .put  = put_u64,
 894 };
 895 #endif /* __linux__ */
 896
 897 /* 8 bit int. See that the received value is the same than the one
 898    in the field */
 899
 900 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
 901 {
 902     uint8_t *v = pv;
 903     uint8_t v2;
 904     qemu_get_8s(f, &v2);
 905
 906     if (*v == v2)
 907         return 0;
 908     return -EINVAL;
 909 }
 910
 911 const VMStateInfo vmstate_info_uint8_equal = {
 912     .name = "uint8 equal",
 913     .get  = get_uint8_equal,
 914     .put  = put_uint8,
 915 };
 916
 917 /* 16 bit unsigned int int. See that the received value is the same than the one
 918    in the field */
 919
 920 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
 921 {
 922     uint16_t *v = pv;
 923     uint16_t v2;
 924     qemu_get_be16s(f, &v2);
 925
 926     if (*v == v2)
 927         return 0;
 928     return -EINVAL;
 929 }
 930
 931 const VMStateInfo vmstate_info_uint16_equal = {
 932     .name = "uint16 equal",
 933     .get  = get_uint16_equal,
 934     .put  = put_uint16,
 935 };
 936
 937 /* timers  */
 938
 939 static int get_timer(QEMUFile *f, void *pv, size_t size)
 940 {
 941     QEMUTimer *v = pv;
 942     qemu_get_timer(f, v);
 943     return 0;
 944 }
 945
 946 static void put_timer(QEMUFile *f, void *pv, size_t size)
 947 {
 948     QEMUTimer *v = pv;
 949     qemu_put_timer(f, v);
 950 }
 951
 952 const VMStateInfo vmstate_info_timer = {
 953     .name = "timer",
 954     .get  = get_timer,
 955     .put  = put_timer,
 956 };
 957
 958 /* uint8_t buffers */
 959
 960 static int get_buffer(QEMUFile *f, void *pv, size_t size)
 961 {
 962     uint8_t *v = pv;
 963     qemu_get_buffer(f, v, size);
 964     return 0;
 965 }
 966
 967 static void put_buffer(QEMUFile *f, void *pv, size_t size)
 968 {
 969     uint8_t *v = pv;
 970     qemu_put_buffer(f, v, size);
 971 }
 972
 973 const VMStateInfo vmstate_info_buffer = {
 974     .name = "buffer",
 975     .get  = get_buffer,
 976     .put  = put_buffer,
 977 };
 978
 979 /* unused buffers: space that was used for some fields that are
 980    not usefull anymore */
 981
 982 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
 983 {
 984     uint8_t buf[1024];
 985     int block_len;
 986
 987     while (size > 0) {
 988         block_len = MIN(sizeof(buf), size);
 989         size -= block_len;
 990         qemu_get_buffer(f, buf, block_len);
 991     }
 992    return 0;
 993 }
 994
 995 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
 996 {
 997     static const uint8_t buf[1024];
 998     int block_len;
 999
1000     while (size > 0) {
1001         block_len = MIN(sizeof(buf), size);
1002         size -= block_len;
1003         qemu_put_buffer(f, buf, block_len);
1004     }
1005 }
1006
1007 const VMStateInfo vmstate_info_unused_buffer = {
1008     .name = "unused_buffer",
1009     .get  = get_unused_buffer,
1010     .put  = put_unused_buffer,
1011 };
1012
1013 typedef struct SaveStateEntry {
1014     QTAILQ_ENTRY(SaveStateEntry) entry;
1015     char idstr[256];
1016     int instance_id;
1017     int alias_id;
1018     int version_id;
1019     int section_id;
1020     SaveSetParamsHandler *set_params;
1021     SaveLiveStateHandler *save_live_state;
1022     SaveStateHandler *save_state;
1023     LoadStateHandler *load_state;
1024     const VMStateDescription *vmsd;
1025     void *opaque;
1026 } SaveStateEntry;
1027
1028
1029 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1030     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1031 static int global_section_id;
1032
1033 static int calculate_new_instance_id(const char *idstr)
1034 {
1035     SaveStateEntry *se;
1036     int instance_id = 0;
1037
1038     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1039         if (strcmp(idstr, se->idstr) == 0
1040             && instance_id <= se->instance_id) {
1041             instance_id = se->instance_id + 1;
1042         }
1043     }
1044     return instance_id;
1045 }
1046
1047 /* TODO: Individual devices generally have very little idea about the rest
1048    of the system, so instance_id should be removed/replaced.
1049    Meanwhile pass -1 as instance_id if you do not already have a clearly
1050    distinguishing id for all instances of your device class. */
1051 int register_savevm_live(const char *idstr,
1052                          int instance_id,
1053                          int version_id,
1054                          SaveSetParamsHandler *set_params,
1055                          SaveLiveStateHandler *save_live_state,
1056                          SaveStateHandler *save_state,
1057                          LoadStateHandler *load_state,
1058                          void *opaque)
1059 {
1060     SaveStateEntry *se;
1061
1062     se = qemu_mallocz(sizeof(SaveStateEntry));
1063     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
1064     se->version_id = version_id;
1065     se->section_id = global_section_id++;
1066     se->set_params = set_params;
1067     se->save_live_state = save_live_state;
1068     se->save_state = save_state;
1069     se->load_state = load_state;
1070     se->opaque = opaque;
1071     se->vmsd = NULL;
1072
1073     if (instance_id == -1) {
1074         se->instance_id = calculate_new_instance_id(idstr);
1075     } else {
1076         se->instance_id = instance_id;
1077     }
1078     /* add at the end of list */
1079     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1080     return 0;
1081 }
1082
1083 int register_savevm(const char *idstr,
1084                     int instance_id,
1085                     int version_id,
1086                     SaveStateHandler *save_state,
1087                     LoadStateHandler *load_state,
1088                     void *opaque)
1089 {
1090     return register_savevm_live(idstr, instance_id, version_id,
1091                                 NULL, NULL, save_state, load_state, opaque);
1092 }
1093
1094 void unregister_savevm(const char *idstr, void *opaque)
1095 {
1096     SaveStateEntry *se, *new_se;
1097
1098     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1099         if (strcmp(se->idstr, idstr) == 0 && se->opaque == opaque) {
1100             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1101             qemu_free(se);
1102         }
1103     }
1104 }
1105
1106 int vmstate_register_with_alias_id(int instance_id,
1107                                    const VMStateDescription *vmsd,
1108                                    void *opaque, int alias_id,
1109                                    int required_for_version)
1110 {
1111     SaveStateEntry *se;
1112
1113     /* If this triggers, alias support can be dropped for the vmsd. */
1114     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1115
1116     se = qemu_mallocz(sizeof(SaveStateEntry));
1117     pstrcpy(se->idstr, sizeof(se->idstr), vmsd->name);
1118     se->version_id = vmsd->version_id;
1119     se->section_id = global_section_id++;
1120     se->save_live_state = NULL;
1121     se->save_state = NULL;
1122     se->load_state = NULL;
1123     se->opaque = opaque;
1124     se->vmsd = vmsd;
1125     se->alias_id = alias_id;
1126
1127     if (instance_id == -1) {
1128         se->instance_id = calculate_new_instance_id(vmsd->name);
1129     } else {
1130         se->instance_id = instance_id;
1131     }
1132     /* add at the end of list */
1133     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1134     return 0;
1135 }
1136
1137 int vmstate_register(int instance_id, const VMStateDescription *vmsd,
1138                      void *opaque)
1139 {
1140     return vmstate_register_with_alias_id(instance_id, vmsd, opaque, -1, 0);
1141 }
1142
1143 void vmstate_unregister(const VMStateDescription *vmsd, void *opaque)
1144 {
1145     SaveStateEntry *se, *new_se;
1146
1147     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1148         if (se->vmsd == vmsd && se->opaque == opaque) {
1149             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1150             qemu_free(se);
1151         }
1152     }
1153 }
1154
1155 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1156                        void *opaque, int version_id)
1157 {
1158     VMStateField *field = vmsd->fields;
1159
1160     if (version_id > vmsd->version_id) {
1161         return -EINVAL;
1162     }
1163     if (version_id < vmsd->minimum_version_id_old) {
1164         return -EINVAL;
1165     }
1166     if  (version_id < vmsd->minimum_version_id) {
1167         return vmsd->load_state_old(f, opaque, version_id);
1168     }
1169     if (vmsd->pre_load) {
1170         int ret = vmsd->pre_load(opaque);
1171         if (ret)
1172             return ret;
1173     }
1174     while(field->name) {
1175         if ((field->field_exists &&
1176              field->field_exists(opaque, version_id)) ||
1177             (!field->field_exists &&
1178              field->version_id <= version_id)) {
1179             void *base_addr = opaque + field->offset;
1180             int ret, i, n_elems = 1;
1181             int size = field->size;
1182
1183             if (field->flags & VMS_VBUFFER) {
1184                 size = *(int32_t *)(opaque+field->size_offset);
1185                 if (field->flags & VMS_MULTIPLY) {
1186                     size *= field->size;
1187                 }
1188             }
1189             if (field->flags & VMS_ARRAY) {
1190                 n_elems = field->num;
1191             } else if (field->flags & VMS_VARRAY_INT32) {
1192                 n_elems = *(int32_t *)(opaque+field->num_offset);
1193             } else if (field->flags & VMS_VARRAY_UINT16) {
1194                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1195             }
1196             if (field->flags & VMS_POINTER) {
1197                 base_addr = *(void **)base_addr + field->start;
1198             }
1199             for (i = 0; i < n_elems; i++) {
1200                 void *addr = base_addr + size * i;
1201
1202                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1203                     addr = *(void **)addr;
1204                 }
1205                 if (field->flags & VMS_STRUCT) {
1206                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1207                 } else {
1208                     ret = field->info->get(f, addr, size);
1209
1210                 }
1211                 if (ret < 0) {
1212                     return ret;
1213                 }
1214             }
1215         }
1216         field++;
1217     }
1218     if (vmsd->post_load) {
1219         return vmsd->post_load(opaque, version_id);
1220     }
1221     return 0;
1222 }
1223
1224 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1225                         void *opaque)
1226 {
1227     VMStateField *field = vmsd->fields;
1228
1229     if (vmsd->pre_save) {
1230         vmsd->pre_save(opaque);
1231     }
1232     while(field->name) {
1233         if (!field->field_exists ||
1234             field->field_exists(opaque, vmsd->version_id)) {
1235             void *base_addr = opaque + field->offset;
1236             int i, n_elems = 1;
1237             int size = field->size;
1238
1239             if (field->flags & VMS_VBUFFER) {
1240                 size = *(int32_t *)(opaque+field->size_offset);
1241                 if (field->flags & VMS_MULTIPLY) {
1242                     size *= field->size;
1243                 }
1244             }
1245             if (field->flags & VMS_ARRAY) {
1246                 n_elems = field->num;
1247             } else if (field->flags & VMS_VARRAY_INT32) {
1248                 n_elems = *(int32_t *)(opaque+field->num_offset);
1249             } else if (field->flags & VMS_VARRAY_UINT16) {
1250                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1251             }
1252             if (field->flags & VMS_POINTER) {
1253                 base_addr = *(void **)base_addr + field->start;
1254             }
1255             for (i = 0; i < n_elems; i++) {
1256                 void *addr = base_addr + size * i;
1257
1258                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1259                     addr = *(void **)addr;
1260                 }
1261                 if (field->flags & VMS_STRUCT) {
1262                     vmstate_save_state(f, field->vmsd, addr);
1263                 } else {
1264                     field->info->put(f, addr, size);
1265                 }
1266             }
1267         }
1268         field++;
1269     }
1270 }
1271
1272 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1273 {
1274     if (!se->vmsd) {         /* Old style */
1275         return se->load_state(f, se->opaque, version_id);
1276     }
1277     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1278 }
1279
1280 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1281 {
1282     if (!se->vmsd) {         /* Old style */
1283         se->save_state(f, se->opaque);
1284         return;
1285     }
1286     vmstate_save_state(f,se->vmsd, se->opaque);
1287 }
1288
1289 #define QEMU_VM_FILE_MAGIC           0x5145564d
1290 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1291 #define QEMU_VM_FILE_VERSION         0x00000003
1292
1293 #define QEMU_VM_EOF                  0x00
1294 #define QEMU_VM_SECTION_START        0x01
1295 #define QEMU_VM_SECTION_PART         0x02
1296 #define QEMU_VM_SECTION_END          0x03
1297 #define QEMU_VM_SECTION_FULL         0x04
1298
1299 int qemu_savevm_state_begin(Monitor *mon, QEMUFile *f, int blk_enable,
1300                             int shared)
1301 {
1302     SaveStateEntry *se;
1303
1304     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1305         if(se->set_params == NULL) {
1306             continue;
1307         }
1308         se->set_params(blk_enable, shared, se->opaque);
1309     }
1310
1311     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1312     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1313
1314     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1315         int len;
1316
1317         if (se->save_live_state == NULL)
1318             continue;
1319
1320         /* Section type */
1321         qemu_put_byte(f, QEMU_VM_SECTION_START);
1322         qemu_put_be32(f, se->section_id);
1323
1324         /* ID string */
1325         len = strlen(se->idstr);
1326         qemu_put_byte(f, len);
1327         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1328
1329         qemu_put_be32(f, se->instance_id);
1330         qemu_put_be32(f, se->version_id);
1331
1332         se->save_live_state(mon, f, QEMU_VM_SECTION_START, se->opaque);
1333     }
1334
1335     if (qemu_file_has_error(f)) {
1336         qemu_savevm_state_cancel(mon, f);
1337         return -EIO;
1338     }
1339
1340     return 0;
1341 }
1342
1343 int qemu_savevm_state_iterate(Monitor *mon, QEMUFile *f)
1344 {
1345     SaveStateEntry *se;
1346     int ret = 1;
1347
1348     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1349         if (se->save_live_state == NULL)
1350             continue;
1351
1352         /* Section type */
1353         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1354         qemu_put_be32(f, se->section_id);
1355
1356         ret = se->save_live_state(mon, f, QEMU_VM_SECTION_PART, se->opaque);
1357         if (!ret) {
1358             /* Do not proceed to the next vmstate before this one reported
1359                completion of the current stage. This serializes the migration
1360                and reduces the probability that a faster changing state is
1361                synchronized over and over again. */
1362             break;
1363         }
1364     }
1365
1366     if (ret)
1367         return 1;
1368
1369     if (qemu_file_has_error(f)) {
1370         qemu_savevm_state_cancel(mon, f);
1371         return -EIO;
1372     }
1373
1374     return 0;
1375 }
1376
1377 int qemu_savevm_state_complete(Monitor *mon, QEMUFile *f)
1378 {
1379     SaveStateEntry *se;
1380
1381     cpu_synchronize_all_states();
1382
1383     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1384         if (se->save_live_state == NULL)
1385             continue;
1386
1387         /* Section type */
1388         qemu_put_byte(f, QEMU_VM_SECTION_END);
1389         qemu_put_be32(f, se->section_id);
1390
1391         se->save_live_state(mon, f, QEMU_VM_SECTION_END, se->opaque);
1392     }
1393
1394     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1395         int len;
1396
1397         if (se->save_state == NULL && se->vmsd == NULL)
1398             continue;
1399
1400         /* Section type */
1401         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1402         qemu_put_be32(f, se->section_id);
1403
1404         /* ID string */
1405         len = strlen(se->idstr);
1406         qemu_put_byte(f, len);
1407         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1408
1409         qemu_put_be32(f, se->instance_id);
1410         qemu_put_be32(f, se->version_id);
1411
1412         vmstate_save(f, se);
1413     }
1414
1415     qemu_put_byte(f, QEMU_VM_EOF);
1416
1417     if (qemu_file_has_error(f))
1418         return -EIO;
1419
1420     return 0;
1421 }
1422
1423 void qemu_savevm_state_cancel(Monitor *mon, QEMUFile *f)
1424 {
1425     SaveStateEntry *se;
1426
1427     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1428         if (se->save_live_state) {
1429             se->save_live_state(mon, f, -1, se->opaque);
1430         }
1431     }
1432 }
1433
1434 static int qemu_savevm_state(Monitor *mon, QEMUFile *f)
1435 {
1436     int saved_vm_running;
1437     int ret;
1438
1439     saved_vm_running = vm_running;
1440     vm_stop(0);
1441
1442     bdrv_flush_all();
1443
1444     ret = qemu_savevm_state_begin(mon, f, 0, 0);
1445     if (ret < 0)
1446         goto out;
1447
1448     do {
1449         ret = qemu_savevm_state_iterate(mon, f);
1450         if (ret < 0)
1451             goto out;
1452     } while (ret == 0);
1453
1454     ret = qemu_savevm_state_complete(mon, f);
1455
1456 out:
1457     if (qemu_file_has_error(f))
1458         ret = -EIO;
1459
1460     if (!ret && saved_vm_running)
1461         vm_start();
1462
1463     return ret;
1464 }
1465
1466 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1467 {
1468     SaveStateEntry *se;
1469
1470     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1471         if (!strcmp(se->idstr, idstr) &&
1472             (instance_id == se->instance_id ||
1473              instance_id == se->alias_id))
1474             return se;
1475     }
1476     return NULL;
1477 }
1478
1479 typedef struct LoadStateEntry {
1480     QLIST_ENTRY(LoadStateEntry) entry;
1481     SaveStateEntry *se;
1482     int section_id;
1483     int version_id;
1484 } LoadStateEntry;
1485
1486 int qemu_loadvm_state(QEMUFile *f)
1487 {
1488     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1489         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1490     LoadStateEntry *le, *new_le;
1491     uint8_t section_type;
1492     unsigned int v;
1493     int ret;
1494
1495     v = qemu_get_be32(f);
1496     if (v != QEMU_VM_FILE_MAGIC)
1497         return -EINVAL;
1498
1499     v = qemu_get_be32(f);
1500     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1501         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1502         return -ENOTSUP;
1503     }
1504     if (v != QEMU_VM_FILE_VERSION)
1505         return -ENOTSUP;
1506
1507     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1508         uint32_t instance_id, version_id, section_id;
1509         SaveStateEntry *se;
1510         char idstr[257];
1511         int len;
1512
1513         switch (section_type) {
1514         case QEMU_VM_SECTION_START:
1515         case QEMU_VM_SECTION_FULL:
1516             /* Read section start */
1517             section_id = qemu_get_be32(f);
1518             len = qemu_get_byte(f);
1519             qemu_get_buffer(f, (uint8_t *)idstr, len);
1520             idstr[len] = 0;
1521             instance_id = qemu_get_be32(f);
1522             version_id = qemu_get_be32(f);
1523
1524             /* Find savevm section */
1525             se = find_se(idstr, instance_id);
1526             if (se == NULL) {
1527                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1528                 ret = -EINVAL;
1529                 goto out;
1530             }
1531
1532             /* Validate version */
1533             if (version_id > se->version_id) {
1534                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
1535                         version_id, idstr, se->version_id);
1536                 ret = -EINVAL;
1537                 goto out;
1538             }
1539
1540             /* Add entry */
1541             le = qemu_mallocz(sizeof(*le));
1542
1543             le->se = se;
1544             le->section_id = section_id;
1545             le->version_id = version_id;
1546             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
1547
1548             ret = vmstate_load(f, le->se, le->version_id);
1549             if (ret < 0) {
1550                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
1551                         instance_id, idstr);
1552                 goto out;
1553             }
1554             break;
1555         case QEMU_VM_SECTION_PART:
1556         case QEMU_VM_SECTION_END:
1557             section_id = qemu_get_be32(f);
1558
1559             QLIST_FOREACH(le, &loadvm_handlers, entry) {
1560                 if (le->section_id == section_id) {
1561                     break;
1562                 }
1563             }
1564             if (le == NULL) {
1565                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
1566                 ret = -EINVAL;
1567                 goto out;
1568             }
1569
1570             ret = vmstate_load(f, le->se, le->version_id);
1571             if (ret < 0) {
1572                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
1573                         section_id);
1574                 goto out;
1575             }
1576             break;
1577         default:
1578             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
1579             ret = -EINVAL;
1580             goto out;
1581         }
1582     }
1583
1584     cpu_synchronize_all_post_init();
1585
1586     ret = 0;
1587
1588 out:
1589     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
1590         QLIST_REMOVE(le, entry);
1591         qemu_free(le);
1592     }
1593
1594     if (qemu_file_has_error(f))
1595         ret = -EIO;
1596
1597     return ret;
1598 }
1599
1600 /* device can contain snapshots */
1601 static int bdrv_can_snapshot(BlockDriverState *bs)
1602 {
1603     return (bs &&
1604             !bdrv_is_removable(bs) &&
1605             !bdrv_is_read_only(bs));
1606 }
1607
1608 /* device must be snapshots in order to have a reliable snapshot */
1609 static int bdrv_has_snapshot(BlockDriverState *bs)
1610 {
1611     return (bs &&
1612             !bdrv_is_removable(bs) &&
1613             !bdrv_is_read_only(bs));
1614 }
1615
1616 static BlockDriverState *get_bs_snapshots(void)
1617 {
1618     BlockDriverState *bs;
1619     DriveInfo *dinfo;
1620
1621     if (bs_snapshots)
1622         return bs_snapshots;
1623     QTAILQ_FOREACH(dinfo, &drives, next) {
1624         bs = dinfo->bdrv;
1625         if (bdrv_can_snapshot(bs))
1626             goto ok;
1627     }
1628     return NULL;
1629  ok:
1630     bs_snapshots = bs;
1631     return bs;
1632 }
1633
1634 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1635                               const char *name)
1636 {
1637     QEMUSnapshotInfo *sn_tab, *sn;
1638     int nb_sns, i, ret;
1639
1640     ret = -ENOENT;
1641     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1642     if (nb_sns < 0)
1643         return ret;
1644     for(i = 0; i < nb_sns; i++) {
1645         sn = &sn_tab[i];
1646         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1647             *sn_info = *sn;
1648             ret = 0;
1649             break;
1650         }
1651     }
1652     qemu_free(sn_tab);
1653     return ret;
1654 }
1655
1656 /*
1657  * Deletes snapshots of a given name in all opened images.
1658  */
1659 static int del_existing_snapshots(Monitor *mon, const char *name)
1660 {
1661     BlockDriverState *bs;
1662     DriveInfo *dinfo;
1663     QEMUSnapshotInfo sn1, *snapshot = &sn1;
1664     int ret;
1665
1666     QTAILQ_FOREACH(dinfo, &drives, next) {
1667         bs = dinfo->bdrv;
1668         if (bdrv_can_snapshot(bs) &&
1669             bdrv_snapshot_find(bs, snapshot, name) >= 0)
1670         {
1671             ret = bdrv_snapshot_delete(bs, name);
1672             if (ret < 0) {
1673                 monitor_printf(mon,
1674                                "Error while deleting snapshot on '%s'\n",
1675                                bdrv_get_device_name(bs));
1676                 return -1;
1677             }
1678         }
1679     }
1680
1681     return 0;
1682 }
1683
1684 void do_savevm(Monitor *mon, const QDict *qdict)
1685 {
1686     DriveInfo *dinfo;
1687     BlockDriverState *bs, *bs1;
1688     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1689     int ret;
1690     QEMUFile *f;
1691     int saved_vm_running;
1692     uint32_t vm_state_size;
1693 #ifdef _WIN32
1694     struct _timeb tb;
1695 #else
1696     struct timeval tv;
1697 #endif
1698     const char *name = qdict_get_try_str(qdict, "name");
1699
1700     bs = get_bs_snapshots();
1701     if (!bs) {
1702         monitor_printf(mon, "No block device can accept snapshots\n");
1703         return;
1704     }
1705
1706     /* ??? Should this occur after vm_stop?  */
1707     qemu_aio_flush();
1708
1709     saved_vm_running = vm_running;
1710     vm_stop(0);
1711
1712     memset(sn, 0, sizeof(*sn));
1713     if (name) {
1714         ret = bdrv_snapshot_find(bs, old_sn, name);
1715         if (ret >= 0) {
1716             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1717             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1718         } else {
1719             pstrcpy(sn->name, sizeof(sn->name), name);
1720         }
1721     }
1722
1723     /* fill auxiliary fields */
1724 #ifdef _WIN32
1725     _ftime(&tb);
1726     sn->date_sec = tb.time;
1727     sn->date_nsec = tb.millitm * 1000000;
1728 #else
1729     gettimeofday(&tv, NULL);
1730     sn->date_sec = tv.tv_sec;
1731     sn->date_nsec = tv.tv_usec * 1000;
1732 #endif
1733     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1734
1735     /* Delete old snapshots of the same name */
1736     if (name && del_existing_snapshots(mon, name) < 0) {
1737         goto the_end;
1738     }
1739
1740     /* save the VM state */
1741     f = qemu_fopen_bdrv(bs, 1);
1742     if (!f) {
1743         monitor_printf(mon, "Could not open VM state file\n");
1744         goto the_end;
1745     }
1746     ret = qemu_savevm_state(mon, f);
1747     vm_state_size = qemu_ftell(f);
1748     qemu_fclose(f);
1749     if (ret < 0) {
1750         monitor_printf(mon, "Error %d while writing VM\n", ret);
1751         goto the_end;
1752     }
1753
1754     /* create the snapshots */
1755
1756     QTAILQ_FOREACH(dinfo, &drives, next) {
1757         bs1 = dinfo->bdrv;
1758         if (bdrv_has_snapshot(bs1)) {
1759             /* Write VM state size only to the image that contains the state */
1760             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1761             ret = bdrv_snapshot_create(bs1, sn);
1762             if (ret < 0) {
1763                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1764                                bdrv_get_device_name(bs1));
1765             }
1766         }
1767     }
1768
1769  the_end:
1770     if (saved_vm_running)
1771         vm_start();
1772 }
1773
1774 int load_vmstate(const char *name)
1775 {
1776     DriveInfo *dinfo;
1777     BlockDriverState *bs, *bs1;
1778     QEMUSnapshotInfo sn;
1779     QEMUFile *f;
1780     int ret;
1781
1782     bs = get_bs_snapshots();
1783     if (!bs) {
1784         error_report("No block device supports snapshots");
1785         return -EINVAL;
1786     }
1787
1788     /* Flush all IO requests so they don't interfere with the new state.  */
1789     qemu_aio_flush();
1790
1791     QTAILQ_FOREACH(dinfo, &drives, next) {
1792         bs1 = dinfo->bdrv;
1793         if (bdrv_has_snapshot(bs1)) {
1794             ret = bdrv_snapshot_goto(bs1, name);
1795             if (ret < 0) {
1796                 switch(ret) {
1797                 case -ENOTSUP:
1798                     error_report("%sSnapshots not supported on device '%s'",
1799                                  bs != bs1 ? "Warning: " : "",
1800                                  bdrv_get_device_name(bs1));
1801                     break;
1802                 case -ENOENT:
1803                     error_report("%sCould not find snapshot '%s' on device '%s'",
1804                                  bs != bs1 ? "Warning: " : "",
1805                                  name, bdrv_get_device_name(bs1));
1806                     break;
1807                 default:
1808                     error_report("%sError %d while activating snapshot on '%s'",
1809                                  bs != bs1 ? "Warning: " : "",
1810                                  ret, bdrv_get_device_name(bs1));
1811                     break;
1812                 }
1813                 /* fatal on snapshot block device */
1814                 if (bs == bs1)
1815                     return 0;
1816             }
1817         }
1818     }
1819
1820     /* Don't even try to load empty VM states */
1821     ret = bdrv_snapshot_find(bs, &sn, name);
1822     if ((ret >= 0) && (sn.vm_state_size == 0))
1823         return -EINVAL;
1824
1825     /* restore the VM state */
1826     f = qemu_fopen_bdrv(bs, 0);
1827     if (!f) {
1828         error_report("Could not open VM state file");
1829         return -EINVAL;
1830     }
1831     ret = qemu_loadvm_state(f);
1832     qemu_fclose(f);
1833     if (ret < 0) {
1834         error_report("Error %d while loading VM state", ret);
1835         return ret;
1836     }
1837     return 0;
1838 }
1839
1840 void do_delvm(Monitor *mon, const QDict *qdict)
1841 {
1842     DriveInfo *dinfo;
1843     BlockDriverState *bs, *bs1;
1844     int ret;
1845     const char *name = qdict_get_str(qdict, "name");
1846
1847     bs = get_bs_snapshots();
1848     if (!bs) {
1849         monitor_printf(mon, "No block device supports snapshots\n");
1850         return;
1851     }
1852
1853     QTAILQ_FOREACH(dinfo, &drives, next) {
1854         bs1 = dinfo->bdrv;
1855         if (bdrv_has_snapshot(bs1)) {
1856             ret = bdrv_snapshot_delete(bs1, name);
1857             if (ret < 0) {
1858                 if (ret == -ENOTSUP)
1859                     monitor_printf(mon,
1860                                    "Snapshots not supported on device '%s'\n",
1861                                    bdrv_get_device_name(bs1));
1862                 else
1863                     monitor_printf(mon, "Error %d while deleting snapshot on "
1864                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1865             }
1866         }
1867     }
1868 }
1869
1870 void do_info_snapshots(Monitor *mon)
1871 {
1872     DriveInfo *dinfo;
1873     BlockDriverState *bs, *bs1;
1874     QEMUSnapshotInfo *sn_tab, *sn;
1875     int nb_sns, i;
1876     char buf[256];
1877
1878     bs = get_bs_snapshots();
1879     if (!bs) {
1880         monitor_printf(mon, "No available block device supports snapshots\n");
1881         return;
1882     }
1883     monitor_printf(mon, "Snapshot devices:");
1884     QTAILQ_FOREACH(dinfo, &drives, next) {
1885         bs1 = dinfo->bdrv;
1886         if (bdrv_has_snapshot(bs1)) {
1887             if (bs == bs1)
1888                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1889         }
1890     }
1891     monitor_printf(mon, "\n");
1892
1893     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1894     if (nb_sns < 0) {
1895         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1896         return;
1897     }
1898     monitor_printf(mon, "Snapshot list (from %s):\n",
1899                    bdrv_get_device_name(bs));
1900     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1901     for(i = 0; i < nb_sns; i++) {
1902         sn = &sn_tab[i];
1903         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1904     }
1905     qemu_free(sn_tab);
1906 }