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 "hw/qdev.h"
  76 #include "net.h"
  77 #include "monitor.h"
  78 #include "sysemu.h"
  79 #include "qemu-timer.h"
  80 #include "qemu-char.h"
  81 #include "audio/audio.h"
  82 #include "migration.h"
  83 #include "qemu_socket.h"
  84 #include "qemu-queue.h"
  85
  86 #define SELF_ANNOUNCE_ROUNDS 5
  87
  88 #ifndef ETH_P_RARP
  89 #define ETH_P_RARP 0x8035
  90 #endif
  91 #define ARP_HTYPE_ETH 0x0001
  92 #define ARP_PTYPE_IP 0x0800
  93 #define ARP_OP_REQUEST_REV 0x3
  94
  95 static int announce_self_create(uint8_t *buf,
  96                                 uint8_t *mac_addr)
  97 {
  98     /* Ethernet header. */
  99     memset(buf, 0xff, 6);         /* destination MAC addr */
 100     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
 101     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
 102
 103     /* RARP header. */
 104     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
 105     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
 106     *(buf + 18) = 6; /* hardware addr length (ethernet) */
 107     *(buf + 19) = 4; /* protocol addr length (IPv4) */
 108     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
 109     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
 110     memset(buf + 28, 0x00, 4);     /* source protocol addr */
 111     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
 112     memset(buf + 38, 0x00, 4);     /* target protocol addr */
 113
 114     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
 115     memset(buf + 42, 0x00, 18);
 116
 117     return 60; /* len (FCS will be added by hardware) */
 118 }
 119
 120 static void qemu_announce_self_iter(NICState *nic, void *opaque)
 121 {
 122     uint8_t buf[60];
 123     int len;
 124
 125     len = announce_self_create(buf, nic->conf->macaddr.a);
 126
 127     qemu_send_packet_raw(&nic->nc, buf, len);
 128 }
 129
 130
 131 static void qemu_announce_self_once(void *opaque)
 132 {
 133     static int count = SELF_ANNOUNCE_ROUNDS;
 134     QEMUTimer *timer = *(QEMUTimer **)opaque;
 135
 136     qemu_foreach_nic(qemu_announce_self_iter, NULL);
 137
 138     if (--count) {
 139         /* delay 50ms, 150ms, 250ms, ... */
 140         qemu_mod_timer(timer, qemu_get_clock(rt_clock) +
 141                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
 142     } else {
 143             qemu_del_timer(timer);
 144             qemu_free_timer(timer);
 145     }
 146 }
 147
 148 void qemu_announce_self(void)
 149 {
 150         static QEMUTimer *timer;
 151         timer = qemu_new_timer(rt_clock, qemu_announce_self_once, &timer);
 152         qemu_announce_self_once(&timer);
 153 }
 154
 155 /***********************************************************/
 156 /* savevm/loadvm support */
 157
 158 #define IO_BUF_SIZE 32768
 159
 160 struct QEMUFile {
 161     QEMUFilePutBufferFunc *put_buffer;
 162     QEMUFileGetBufferFunc *get_buffer;
 163     QEMUFileCloseFunc *close;
 164     QEMUFileRateLimit *rate_limit;
 165     QEMUFileSetRateLimit *set_rate_limit;
 166     QEMUFileGetRateLimit *get_rate_limit;
 167     void *opaque;
 168     int is_write;
 169
 170     int64_t buf_offset; /* start of buffer when writing, end of buffer
 171                            when reading */
 172     int buf_index;
 173     int buf_size; /* 0 when writing */
 174     uint8_t buf[IO_BUF_SIZE];
 175
 176     int has_error;
 177 };
 178
 179 typedef struct QEMUFileStdio
 180 {
 181     FILE *stdio_file;
 182     QEMUFile *file;
 183 } QEMUFileStdio;
 184
 185 typedef struct QEMUFileSocket
 186 {
 187     int fd;
 188     QEMUFile *file;
 189 } QEMUFileSocket;
 190
 191 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 192 {
 193     QEMUFileSocket *s = opaque;
 194     ssize_t len;
 195
 196     do {
 197         len = recv(s->fd, (void *)buf, size, 0);
 198     } while (len == -1 && socket_error() == EINTR);
 199
 200     if (len == -1)
 201         len = -socket_error();
 202
 203     return len;
 204 }
 205
 206 static int socket_close(void *opaque)
 207 {
 208     QEMUFileSocket *s = opaque;
 209     qemu_free(s);
 210     return 0;
 211 }
 212
 213 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 214 {
 215     QEMUFileStdio *s = opaque;
 216     return fwrite(buf, 1, size, s->stdio_file);
 217 }
 218
 219 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 220 {
 221     QEMUFileStdio *s = opaque;
 222     FILE *fp = s->stdio_file;
 223     int bytes;
 224
 225     do {
 226         clearerr(fp);
 227         bytes = fread(buf, 1, size, fp);
 228     } while ((bytes == 0) && ferror(fp) && (errno == EINTR));
 229     return bytes;
 230 }
 231
 232 static int stdio_pclose(void *opaque)
 233 {
 234     QEMUFileStdio *s = opaque;
 235     int ret;
 236     ret = pclose(s->stdio_file);
 237     qemu_free(s);
 238     return ret;
 239 }
 240
 241 static int stdio_fclose(void *opaque)
 242 {
 243     QEMUFileStdio *s = opaque;
 244     fclose(s->stdio_file);
 245     qemu_free(s);
 246     return 0;
 247 }
 248
 249 QEMUFile *qemu_popen(FILE *stdio_file, const char *mode)
 250 {
 251     QEMUFileStdio *s;
 252
 253     if (stdio_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 254         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 255         return NULL;
 256     }
 257
 258     s = qemu_mallocz(sizeof(QEMUFileStdio));
 259
 260     s->stdio_file = stdio_file;
 261
 262     if(mode[0] == 'r') {
 263         s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_pclose,
 264                                  NULL, NULL, NULL);
 265     } else {
 266         s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_pclose,
 267                                  NULL, NULL, NULL);
 268     }
 269     return s->file;
 270 }
 271
 272 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 273 {
 274     FILE *popen_file;
 275
 276     popen_file = popen(command, mode);
 277     if(popen_file == NULL) {
 278         return NULL;
 279     }
 280
 281     return qemu_popen(popen_file, mode);
 282 }
 283
 284 int qemu_stdio_fd(QEMUFile *f)
 285 {
 286     QEMUFileStdio *p;
 287     int fd;
 288
 289     p = (QEMUFileStdio *)f->opaque;
 290     fd = fileno(p->stdio_file);
 291
 292     return fd;
 293 }
 294
 295 QEMUFile *qemu_fdopen(int fd, const char *mode)
 296 {
 297     QEMUFileStdio *s;
 298
 299     if (mode == NULL ||
 300         (mode[0] != 'r' && mode[0] != 'w') ||
 301         mode[1] != 'b' || mode[2] != 0) {
 302         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 303         return NULL;
 304     }
 305
 306     s = qemu_mallocz(sizeof(QEMUFileStdio));
 307     s->stdio_file = fdopen(fd, mode);
 308     if (!s->stdio_file)
 309         goto fail;
 310
 311     if(mode[0] == 'r') {
 312         s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_fclose,
 313                                  NULL, NULL, NULL);
 314     } else {
 315         s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_fclose,
 316                                  NULL, NULL, NULL);
 317     }
 318     return s->file;
 319
 320 fail:
 321     qemu_free(s);
 322     return NULL;
 323 }
 324
 325 QEMUFile *qemu_fopen_socket(int fd)
 326 {
 327     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 328
 329     s->fd = fd;
 330     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close,
 331                              NULL, NULL, NULL);
 332     return s->file;
 333 }
 334
 335 static int file_put_buffer(void *opaque, const uint8_t *buf,
 336                             int64_t pos, int size)
 337 {
 338     QEMUFileStdio *s = opaque;
 339     fseek(s->stdio_file, pos, SEEK_SET);
 340     return fwrite(buf, 1, size, s->stdio_file);
 341 }
 342
 343 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 344 {
 345     QEMUFileStdio *s = opaque;
 346     fseek(s->stdio_file, pos, SEEK_SET);
 347     return fread(buf, 1, size, s->stdio_file);
 348 }
 349
 350 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 351 {
 352     QEMUFileStdio *s;
 353
 354     if (mode == NULL ||
 355         (mode[0] != 'r' && mode[0] != 'w') ||
 356         mode[1] != 'b' || mode[2] != 0) {
 357         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 358         return NULL;
 359     }
 360
 361     s = qemu_mallocz(sizeof(QEMUFileStdio));
 362
 363     s->stdio_file = fopen(filename, mode);
 364     if (!s->stdio_file)
 365         goto fail;
 366
 367     if(mode[0] == 'w') {
 368         s->file = qemu_fopen_ops(s, file_put_buffer, NULL, stdio_fclose,
 369                                  NULL, NULL, NULL);
 370     } else {
 371         s->file = qemu_fopen_ops(s, NULL, file_get_buffer, stdio_fclose,
 372                                NULL, NULL, NULL);
 373     }
 374     return s->file;
 375 fail:
 376     qemu_free(s);
 377     return NULL;
 378 }
 379
 380 static int block_put_buffer(void *opaque, const uint8_t *buf,
 381                            int64_t pos, int size)
 382 {
 383     bdrv_save_vmstate(opaque, buf, pos, size);
 384     return size;
 385 }
 386
 387 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 388 {
 389     return bdrv_load_vmstate(opaque, buf, pos, size);
 390 }
 391
 392 static int bdrv_fclose(void *opaque)
 393 {
 394     return 0;
 395 }
 396
 397 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 398 {
 399     if (is_writable)
 400         return qemu_fopen_ops(bs, block_put_buffer, NULL, bdrv_fclose,
 401                               NULL, NULL, NULL);
 402     return qemu_fopen_ops(bs, NULL, block_get_buffer, bdrv_fclose, NULL, NULL, NULL);
 403 }
 404
 405 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 406                          QEMUFileGetBufferFunc *get_buffer,
 407                          QEMUFileCloseFunc *close,
 408                          QEMUFileRateLimit *rate_limit,
 409                          QEMUFileSetRateLimit *set_rate_limit,
 410                          QEMUFileGetRateLimit *get_rate_limit)
 411 {
 412     QEMUFile *f;
 413
 414     f = qemu_mallocz(sizeof(QEMUFile));
 415
 416     f->opaque = opaque;
 417     f->put_buffer = put_buffer;
 418     f->get_buffer = get_buffer;
 419     f->close = close;
 420     f->rate_limit = rate_limit;
 421     f->set_rate_limit = set_rate_limit;
 422     f->get_rate_limit = get_rate_limit;
 423     f->is_write = 0;
 424
 425     return f;
 426 }
 427
 428 int qemu_file_has_error(QEMUFile *f)
 429 {
 430     return f->has_error;
 431 }
 432
 433 void qemu_file_set_error(QEMUFile *f)
 434 {
 435     f->has_error = 1;
 436 }
 437
 438 void qemu_fflush(QEMUFile *f)
 439 {
 440     if (!f->put_buffer)
 441         return;
 442
 443     if (f->is_write && f->buf_index > 0) {
 444         int len;
 445
 446         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 447         if (len > 0)
 448             f->buf_offset += f->buf_index;
 449         else
 450             f->has_error = 1;
 451         f->buf_index = 0;
 452     }
 453 }
 454
 455 static void qemu_fill_buffer(QEMUFile *f)
 456 {
 457     int len;
 458
 459     if (!f->get_buffer)
 460         return;
 461
 462     if (f->is_write)
 463         abort();
 464
 465     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 466     if (len > 0) {
 467         f->buf_index = 0;
 468         f->buf_size = len;
 469         f->buf_offset += len;
 470     } else if (len != -EAGAIN)
 471         f->has_error = 1;
 472 }
 473
 474 int qemu_fclose(QEMUFile *f)
 475 {
 476     int ret = 0;
 477     qemu_fflush(f);
 478     if (f->close)
 479         ret = f->close(f->opaque);
 480     qemu_free(f);
 481     return ret;
 482 }
 483
 484 void qemu_file_put_notify(QEMUFile *f)
 485 {
 486     f->put_buffer(f->opaque, NULL, 0, 0);
 487 }
 488
 489 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 490 {
 491     int l;
 492
 493     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 494         fprintf(stderr,
 495                 "Attempted to write to buffer while read buffer is not empty\n");
 496         abort();
 497     }
 498
 499     while (!f->has_error && size > 0) {
 500         l = IO_BUF_SIZE - f->buf_index;
 501         if (l > size)
 502             l = size;
 503         memcpy(f->buf + f->buf_index, buf, l);
 504         f->is_write = 1;
 505         f->buf_index += l;
 506         buf += l;
 507         size -= l;
 508         if (f->buf_index >= IO_BUF_SIZE)
 509             qemu_fflush(f);
 510     }
 511 }
 512
 513 void qemu_put_byte(QEMUFile *f, int v)
 514 {
 515     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 516         fprintf(stderr,
 517                 "Attempted to write to buffer while read buffer is not empty\n");
 518         abort();
 519     }
 520
 521     f->buf[f->buf_index++] = v;
 522     f->is_write = 1;
 523     if (f->buf_index >= IO_BUF_SIZE)
 524         qemu_fflush(f);
 525 }
 526
 527 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 528 {
 529     int size, l;
 530
 531     if (f->is_write)
 532         abort();
 533
 534     size = size1;
 535     while (size > 0) {
 536         l = f->buf_size - f->buf_index;
 537         if (l == 0) {
 538             qemu_fill_buffer(f);
 539             l = f->buf_size - f->buf_index;
 540             if (l == 0)
 541                 break;
 542         }
 543         if (l > size)
 544             l = size;
 545         memcpy(buf, f->buf + f->buf_index, l);
 546         f->buf_index += l;
 547         buf += l;
 548         size -= l;
 549     }
 550     return size1 - size;
 551 }
 552
 553 static int qemu_peek_byte(QEMUFile *f)
 554 {
 555     if (f->is_write)
 556         abort();
 557
 558     if (f->buf_index >= f->buf_size) {
 559         qemu_fill_buffer(f);
 560         if (f->buf_index >= f->buf_size)
 561             return 0;
 562     }
 563     return f->buf[f->buf_index];
 564 }
 565
 566 int qemu_get_byte(QEMUFile *f)
 567 {
 568     if (f->is_write)
 569         abort();
 570
 571     if (f->buf_index >= f->buf_size) {
 572         qemu_fill_buffer(f);
 573         if (f->buf_index >= f->buf_size)
 574             return 0;
 575     }
 576     return f->buf[f->buf_index++];
 577 }
 578
 579 int64_t qemu_ftell(QEMUFile *f)
 580 {
 581     return f->buf_offset - f->buf_size + f->buf_index;
 582 }
 583
 584 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 585 {
 586     if (whence == SEEK_SET) {
 587         /* nothing to do */
 588     } else if (whence == SEEK_CUR) {
 589         pos += qemu_ftell(f);
 590     } else {
 591         /* SEEK_END not supported */
 592         return -1;
 593     }
 594     if (f->put_buffer) {
 595         qemu_fflush(f);
 596         f->buf_offset = pos;
 597     } else {
 598         f->buf_offset = pos;
 599         f->buf_index = 0;
 600         f->buf_size = 0;
 601     }
 602     return pos;
 603 }
 604
 605 int qemu_file_rate_limit(QEMUFile *f)
 606 {
 607     if (f->rate_limit)
 608         return f->rate_limit(f->opaque);
 609
 610     return 0;
 611 }
 612
 613 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 614 {
 615     if (f->get_rate_limit)
 616         return f->get_rate_limit(f->opaque);
 617
 618     return 0;
 619 }
 620
 621 int64_t qemu_file_set_rate_limit(QEMUFile *f, int64_t new_rate)
 622 {
 623     /* any failed or completed migration keeps its state to allow probing of
 624      * migration data, but has no associated file anymore */
 625     if (f && f->set_rate_limit)
 626         return f->set_rate_limit(f->opaque, new_rate);
 627
 628     return 0;
 629 }
 630
 631 void qemu_put_be16(QEMUFile *f, unsigned int v)
 632 {
 633     qemu_put_byte(f, v >> 8);
 634     qemu_put_byte(f, v);
 635 }
 636
 637 void qemu_put_be32(QEMUFile *f, unsigned int v)
 638 {
 639     qemu_put_byte(f, v >> 24);
 640     qemu_put_byte(f, v >> 16);
 641     qemu_put_byte(f, v >> 8);
 642     qemu_put_byte(f, v);
 643 }
 644
 645 void qemu_put_be64(QEMUFile *f, uint64_t v)
 646 {
 647     qemu_put_be32(f, v >> 32);
 648     qemu_put_be32(f, v);
 649 }
 650
 651 unsigned int qemu_get_be16(QEMUFile *f)
 652 {
 653     unsigned int v;
 654     v = qemu_get_byte(f) << 8;
 655     v |= qemu_get_byte(f);
 656     return v;
 657 }
 658
 659 unsigned int qemu_get_be32(QEMUFile *f)
 660 {
 661     unsigned int v;
 662     v = qemu_get_byte(f) << 24;
 663     v |= qemu_get_byte(f) << 16;
 664     v |= qemu_get_byte(f) << 8;
 665     v |= qemu_get_byte(f);
 666     return v;
 667 }
 668
 669 uint64_t qemu_get_be64(QEMUFile *f)
 670 {
 671     uint64_t v;
 672     v = (uint64_t)qemu_get_be32(f) << 32;
 673     v |= qemu_get_be32(f);
 674     return v;
 675 }
 676
 677 /* bool */
 678
 679 static int get_bool(QEMUFile *f, void *pv, size_t size)
 680 {
 681     bool *v = pv;
 682     *v = qemu_get_byte(f);
 683     return 0;
 684 }
 685
 686 static void put_bool(QEMUFile *f, void *pv, size_t size)
 687 {
 688     bool *v = pv;
 689     qemu_put_byte(f, *v);
 690 }
 691
 692 const VMStateInfo vmstate_info_bool = {
 693     .name = "bool",
 694     .get  = get_bool,
 695     .put  = put_bool,
 696 };
 697
 698 /* 8 bit int */
 699
 700 static int get_int8(QEMUFile *f, void *pv, size_t size)
 701 {
 702     int8_t *v = pv;
 703     qemu_get_s8s(f, v);
 704     return 0;
 705 }
 706
 707 static void put_int8(QEMUFile *f, void *pv, size_t size)
 708 {
 709     int8_t *v = pv;
 710     qemu_put_s8s(f, v);
 711 }
 712
 713 const VMStateInfo vmstate_info_int8 = {
 714     .name = "int8",
 715     .get  = get_int8,
 716     .put  = put_int8,
 717 };
 718
 719 /* 16 bit int */
 720
 721 static int get_int16(QEMUFile *f, void *pv, size_t size)
 722 {
 723     int16_t *v = pv;
 724     qemu_get_sbe16s(f, v);
 725     return 0;
 726 }
 727
 728 static void put_int16(QEMUFile *f, void *pv, size_t size)
 729 {
 730     int16_t *v = pv;
 731     qemu_put_sbe16s(f, v);
 732 }
 733
 734 const VMStateInfo vmstate_info_int16 = {
 735     .name = "int16",
 736     .get  = get_int16,
 737     .put  = put_int16,
 738 };
 739
 740 /* 32 bit int */
 741
 742 static int get_int32(QEMUFile *f, void *pv, size_t size)
 743 {
 744     int32_t *v = pv;
 745     qemu_get_sbe32s(f, v);
 746     return 0;
 747 }
 748
 749 static void put_int32(QEMUFile *f, void *pv, size_t size)
 750 {
 751     int32_t *v = pv;
 752     qemu_put_sbe32s(f, v);
 753 }
 754
 755 const VMStateInfo vmstate_info_int32 = {
 756     .name = "int32",
 757     .get  = get_int32,
 758     .put  = put_int32,
 759 };
 760
 761 /* 32 bit int. See that the received value is the same than the one
 762    in the field */
 763
 764 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 765 {
 766     int32_t *v = pv;
 767     int32_t v2;
 768     qemu_get_sbe32s(f, &v2);
 769
 770     if (*v == v2)
 771         return 0;
 772     return -EINVAL;
 773 }
 774
 775 const VMStateInfo vmstate_info_int32_equal = {
 776     .name = "int32 equal",
 777     .get  = get_int32_equal,
 778     .put  = put_int32,
 779 };
 780
 781 /* 32 bit int. See that the received value is the less or the same
 782    than the one in the field */
 783
 784 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 785 {
 786     int32_t *old = pv;
 787     int32_t new;
 788     qemu_get_sbe32s(f, &new);
 789
 790     if (*old <= new)
 791         return 0;
 792     return -EINVAL;
 793 }
 794
 795 const VMStateInfo vmstate_info_int32_le = {
 796     .name = "int32 equal",
 797     .get  = get_int32_le,
 798     .put  = put_int32,
 799 };
 800
 801 /* 64 bit int */
 802
 803 static int get_int64(QEMUFile *f, void *pv, size_t size)
 804 {
 805     int64_t *v = pv;
 806     qemu_get_sbe64s(f, v);
 807     return 0;
 808 }
 809
 810 static void put_int64(QEMUFile *f, void *pv, size_t size)
 811 {
 812     int64_t *v = pv;
 813     qemu_put_sbe64s(f, v);
 814 }
 815
 816 const VMStateInfo vmstate_info_int64 = {
 817     .name = "int64",
 818     .get  = get_int64,
 819     .put  = put_int64,
 820 };
 821
 822 /* 8 bit unsigned int */
 823
 824 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 825 {
 826     uint8_t *v = pv;
 827     qemu_get_8s(f, v);
 828     return 0;
 829 }
 830
 831 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 832 {
 833     uint8_t *v = pv;
 834     qemu_put_8s(f, v);
 835 }
 836
 837 const VMStateInfo vmstate_info_uint8 = {
 838     .name = "uint8",
 839     .get  = get_uint8,
 840     .put  = put_uint8,
 841 };
 842
 843 /* 16 bit unsigned int */
 844
 845 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 846 {
 847     uint16_t *v = pv;
 848     qemu_get_be16s(f, v);
 849     return 0;
 850 }
 851
 852 static void put_uint16(QEMUFile *f, void *pv, size_t size)
 853 {
 854     uint16_t *v = pv;
 855     qemu_put_be16s(f, v);
 856 }
 857
 858 const VMStateInfo vmstate_info_uint16 = {
 859     .name = "uint16",
 860     .get  = get_uint16,
 861     .put  = put_uint16,
 862 };
 863
 864 /* 32 bit unsigned int */
 865
 866 static int get_uint32(QEMUFile *f, void *pv, size_t size)
 867 {
 868     uint32_t *v = pv;
 869     qemu_get_be32s(f, v);
 870     return 0;
 871 }
 872
 873 static void put_uint32(QEMUFile *f, void *pv, size_t size)
 874 {
 875     uint32_t *v = pv;
 876     qemu_put_be32s(f, v);
 877 }
 878
 879 const VMStateInfo vmstate_info_uint32 = {
 880     .name = "uint32",
 881     .get  = get_uint32,
 882     .put  = put_uint32,
 883 };
 884
 885 /* 32 bit uint. See that the received value is the same than the one
 886    in the field */
 887
 888 static int get_uint32_equal(QEMUFile *f, void *pv, size_t size)
 889 {
 890     uint32_t *v = pv;
 891     uint32_t v2;
 892     qemu_get_be32s(f, &v2);
 893
 894     if (*v == v2) {
 895         return 0;
 896     }
 897     return -EINVAL;
 898 }
 899
 900 const VMStateInfo vmstate_info_uint32_equal = {
 901     .name = "uint32 equal",
 902     .get  = get_uint32_equal,
 903     .put  = put_uint32,
 904 };
 905
 906 /* 64 bit unsigned int */
 907
 908 static int get_uint64(QEMUFile *f, void *pv, size_t size)
 909 {
 910     uint64_t *v = pv;
 911     qemu_get_be64s(f, v);
 912     return 0;
 913 }
 914
 915 static void put_uint64(QEMUFile *f, void *pv, size_t size)
 916 {
 917     uint64_t *v = pv;
 918     qemu_put_be64s(f, v);
 919 }
 920
 921 const VMStateInfo vmstate_info_uint64 = {
 922     .name = "uint64",
 923     .get  = get_uint64,
 924     .put  = put_uint64,
 925 };
 926
 927 /* 8 bit int. See that the received value is the same than the one
 928    in the field */
 929
 930 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
 931 {
 932     uint8_t *v = pv;
 933     uint8_t v2;
 934     qemu_get_8s(f, &v2);
 935
 936     if (*v == v2)
 937         return 0;
 938     return -EINVAL;
 939 }
 940
 941 const VMStateInfo vmstate_info_uint8_equal = {
 942     .name = "uint8 equal",
 943     .get  = get_uint8_equal,
 944     .put  = put_uint8,
 945 };
 946
 947 /* 16 bit unsigned int int. See that the received value is the same than the one
 948    in the field */
 949
 950 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
 951 {
 952     uint16_t *v = pv;
 953     uint16_t v2;
 954     qemu_get_be16s(f, &v2);
 955
 956     if (*v == v2)
 957         return 0;
 958     return -EINVAL;
 959 }
 960
 961 const VMStateInfo vmstate_info_uint16_equal = {
 962     .name = "uint16 equal",
 963     .get  = get_uint16_equal,
 964     .put  = put_uint16,
 965 };
 966
 967 /* timers  */
 968
 969 static int get_timer(QEMUFile *f, void *pv, size_t size)
 970 {
 971     QEMUTimer *v = pv;
 972     qemu_get_timer(f, v);
 973     return 0;
 974 }
 975
 976 static void put_timer(QEMUFile *f, void *pv, size_t size)
 977 {
 978     QEMUTimer *v = pv;
 979     qemu_put_timer(f, v);
 980 }
 981
 982 const VMStateInfo vmstate_info_timer = {
 983     .name = "timer",
 984     .get  = get_timer,
 985     .put  = put_timer,
 986 };
 987
 988 /* uint8_t buffers */
 989
 990 static int get_buffer(QEMUFile *f, void *pv, size_t size)
 991 {
 992     uint8_t *v = pv;
 993     qemu_get_buffer(f, v, size);
 994     return 0;
 995 }
 996
 997 static void put_buffer(QEMUFile *f, void *pv, size_t size)
 998 {
 999     uint8_t *v = pv;
1000     qemu_put_buffer(f, v, size);
1001 }
1002
1003 const VMStateInfo vmstate_info_buffer = {
1004     .name = "buffer",
1005     .get  = get_buffer,
1006     .put  = put_buffer,
1007 };
1008
1009 /* unused buffers: space that was used for some fields that are
1010    not usefull anymore */
1011
1012 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
1013 {
1014     uint8_t buf[1024];
1015     int block_len;
1016
1017     while (size > 0) {
1018         block_len = MIN(sizeof(buf), size);
1019         size -= block_len;
1020         qemu_get_buffer(f, buf, block_len);
1021     }
1022    return 0;
1023 }
1024
1025 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1026 {
1027     static const uint8_t buf[1024];
1028     int block_len;
1029
1030     while (size > 0) {
1031         block_len = MIN(sizeof(buf), size);
1032         size -= block_len;
1033         qemu_put_buffer(f, buf, block_len);
1034     }
1035 }
1036
1037 const VMStateInfo vmstate_info_unused_buffer = {
1038     .name = "unused_buffer",
1039     .get  = get_unused_buffer,
1040     .put  = put_unused_buffer,
1041 };
1042
1043 typedef struct CompatEntry {
1044     char idstr[256];
1045     int instance_id;
1046 } CompatEntry;
1047
1048 typedef struct SaveStateEntry {
1049     QTAILQ_ENTRY(SaveStateEntry) entry;
1050     char idstr[256];
1051     int instance_id;
1052     int alias_id;
1053     int version_id;
1054     int section_id;
1055     SaveSetParamsHandler *set_params;
1056     SaveLiveStateHandler *save_live_state;
1057     SaveStateHandler *save_state;
1058     LoadStateHandler *load_state;
1059     const VMStateDescription *vmsd;
1060     void *opaque;
1061     CompatEntry *compat;
1062     int no_migrate;
1063 } SaveStateEntry;
1064
1065
1066 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1067     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1068 static int global_section_id;
1069
1070 static int calculate_new_instance_id(const char *idstr)
1071 {
1072     SaveStateEntry *se;
1073     int instance_id = 0;
1074
1075     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1076         if (strcmp(idstr, se->idstr) == 0
1077             && instance_id <= se->instance_id) {
1078             instance_id = se->instance_id + 1;
1079         }
1080     }
1081     return instance_id;
1082 }
1083
1084 static int calculate_compat_instance_id(const char *idstr)
1085 {
1086     SaveStateEntry *se;
1087     int instance_id = 0;
1088
1089     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1090         if (!se->compat)
1091             continue;
1092
1093         if (strcmp(idstr, se->compat->idstr) == 0
1094             && instance_id <= se->compat->instance_id) {
1095             instance_id = se->compat->instance_id + 1;
1096         }
1097     }
1098     return instance_id;
1099 }
1100
1101 /* TODO: Individual devices generally have very little idea about the rest
1102    of the system, so instance_id should be removed/replaced.
1103    Meanwhile pass -1 as instance_id if you do not already have a clearly
1104    distinguishing id for all instances of your device class. */
1105 int register_savevm_live(DeviceState *dev,
1106                          const char *idstr,
1107                          int instance_id,
1108                          int version_id,
1109                          SaveSetParamsHandler *set_params,
1110                          SaveLiveStateHandler *save_live_state,
1111                          SaveStateHandler *save_state,
1112                          LoadStateHandler *load_state,
1113                          void *opaque)
1114 {
1115     SaveStateEntry *se;
1116
1117     se = qemu_mallocz(sizeof(SaveStateEntry));
1118     se->version_id = version_id;
1119     se->section_id = global_section_id++;
1120     se->set_params = set_params;
1121     se->save_live_state = save_live_state;
1122     se->save_state = save_state;
1123     se->load_state = load_state;
1124     se->opaque = opaque;
1125     se->vmsd = NULL;
1126     se->no_migrate = 0;
1127
1128     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1129         char *id = dev->parent_bus->info->get_dev_path(dev);
1130         if (id) {
1131             pstrcpy(se->idstr, sizeof(se->idstr), id);
1132             pstrcat(se->idstr, sizeof(se->idstr), "/");
1133             qemu_free(id);
1134
1135             se->compat = qemu_mallocz(sizeof(CompatEntry));
1136             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1137             se->compat->instance_id = instance_id == -1 ?
1138                          calculate_compat_instance_id(idstr) : instance_id;
1139             instance_id = -1;
1140         }
1141     }
1142     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1143
1144     if (instance_id == -1) {
1145         se->instance_id = calculate_new_instance_id(se->idstr);
1146     } else {
1147         se->instance_id = instance_id;
1148     }
1149     assert(!se->compat || se->instance_id == 0);
1150     /* add at the end of list */
1151     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1152     return 0;
1153 }
1154
1155 int register_savevm(DeviceState *dev,
1156                     const char *idstr,
1157                     int instance_id,
1158                     int version_id,
1159                     SaveStateHandler *save_state,
1160                     LoadStateHandler *load_state,
1161                     void *opaque)
1162 {
1163     return register_savevm_live(dev, idstr, instance_id, version_id,
1164                                 NULL, NULL, save_state, load_state, opaque);
1165 }
1166
1167 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1168 {
1169     SaveStateEntry *se, *new_se;
1170     char id[256] = "";
1171
1172     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1173         char *path = dev->parent_bus->info->get_dev_path(dev);
1174         if (path) {
1175             pstrcpy(id, sizeof(id), path);
1176             pstrcat(id, sizeof(id), "/");
1177             qemu_free(path);
1178         }
1179     }
1180     pstrcat(id, sizeof(id), idstr);
1181
1182     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1183         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1184             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1185             if (se->compat) {
1186                 qemu_free(se->compat);
1187             }
1188             qemu_free(se);
1189         }
1190     }
1191 }
1192
1193 /* mark a device as not to be migrated, that is the device should be
1194    unplugged before migration */
1195 void register_device_unmigratable(DeviceState *dev, const char *idstr,
1196                                                             void *opaque)
1197 {
1198     SaveStateEntry *se;
1199     char id[256] = "";
1200
1201     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1202         char *path = dev->parent_bus->info->get_dev_path(dev);
1203         if (path) {
1204             pstrcpy(id, sizeof(id), path);
1205             pstrcat(id, sizeof(id), "/");
1206             qemu_free(path);
1207         }
1208     }
1209     pstrcat(id, sizeof(id), idstr);
1210
1211     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1212         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1213             se->no_migrate = 1;
1214         }
1215     }
1216 }
1217
1218 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1219                                    const VMStateDescription *vmsd,
1220                                    void *opaque, int alias_id,
1221                                    int required_for_version)
1222 {
1223     SaveStateEntry *se;
1224
1225     /* If this triggers, alias support can be dropped for the vmsd. */
1226     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1227
1228     se = qemu_mallocz(sizeof(SaveStateEntry));
1229     se->version_id = vmsd->version_id;
1230     se->section_id = global_section_id++;
1231     se->save_live_state = NULL;
1232     se->save_state = NULL;
1233     se->load_state = NULL;
1234     se->opaque = opaque;
1235     se->vmsd = vmsd;
1236     se->alias_id = alias_id;
1237
1238     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1239         char *id = dev->parent_bus->info->get_dev_path(dev);
1240         if (id) {
1241             pstrcpy(se->idstr, sizeof(se->idstr), id);
1242             pstrcat(se->idstr, sizeof(se->idstr), "/");
1243             qemu_free(id);
1244
1245             se->compat = qemu_mallocz(sizeof(CompatEntry));
1246             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1247             se->compat->instance_id = instance_id == -1 ?
1248                          calculate_compat_instance_id(vmsd->name) : instance_id;
1249             instance_id = -1;
1250         }
1251     }
1252     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1253
1254     if (instance_id == -1) {
1255         se->instance_id = calculate_new_instance_id(se->idstr);
1256     } else {
1257         se->instance_id = instance_id;
1258     }
1259     assert(!se->compat || se->instance_id == 0);
1260     /* add at the end of list */
1261     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1262     return 0;
1263 }
1264
1265 int vmstate_register(DeviceState *dev, int instance_id,
1266                      const VMStateDescription *vmsd, void *opaque)
1267 {
1268     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1269                                           opaque, -1, 0);
1270 }
1271
1272 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1273                         void *opaque)
1274 {
1275     SaveStateEntry *se, *new_se;
1276
1277     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1278         if (se->vmsd == vmsd && se->opaque == opaque) {
1279             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1280             if (se->compat) {
1281                 qemu_free(se->compat);
1282             }
1283             qemu_free(se);
1284         }
1285     }
1286 }
1287
1288 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1289                                     void *opaque);
1290 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1291                                    void *opaque);
1292
1293 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1294                        void *opaque, int version_id)
1295 {
1296     VMStateField *field = vmsd->fields;
1297     int ret;
1298
1299     if (version_id > vmsd->version_id) {
1300         return -EINVAL;
1301     }
1302     if (version_id < vmsd->minimum_version_id_old) {
1303         return -EINVAL;
1304     }
1305     if  (version_id < vmsd->minimum_version_id) {
1306         return vmsd->load_state_old(f, opaque, version_id);
1307     }
1308     if (vmsd->pre_load) {
1309         int ret = vmsd->pre_load(opaque);
1310         if (ret)
1311             return ret;
1312     }
1313     while(field->name) {
1314         if ((field->field_exists &&
1315              field->field_exists(opaque, version_id)) ||
1316             (!field->field_exists &&
1317              field->version_id <= version_id)) {
1318             void *base_addr = opaque + field->offset;
1319             int i, n_elems = 1;
1320             int size = field->size;
1321
1322             if (field->flags & VMS_VBUFFER) {
1323                 size = *(int32_t *)(opaque+field->size_offset);
1324                 if (field->flags & VMS_MULTIPLY) {
1325                     size *= field->size;
1326                 }
1327             }
1328             if (field->flags & VMS_ARRAY) {
1329                 n_elems = field->num;
1330             } else if (field->flags & VMS_VARRAY_INT32) {
1331                 n_elems = *(int32_t *)(opaque+field->num_offset);
1332             } else if (field->flags & VMS_VARRAY_UINT16) {
1333                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1334             }
1335             if (field->flags & VMS_POINTER) {
1336                 base_addr = *(void **)base_addr + field->start;
1337             }
1338             for (i = 0; i < n_elems; i++) {
1339                 void *addr = base_addr + size * i;
1340
1341                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1342                     addr = *(void **)addr;
1343                 }
1344                 if (field->flags & VMS_STRUCT) {
1345                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1346                 } else {
1347                     ret = field->info->get(f, addr, size);
1348
1349                 }
1350                 if (ret < 0) {
1351                     return ret;
1352                 }
1353             }
1354         }
1355         field++;
1356     }
1357     ret = vmstate_subsection_load(f, vmsd, opaque);
1358     if (ret != 0) {
1359         return ret;
1360     }
1361     if (vmsd->post_load) {
1362         return vmsd->post_load(opaque, version_id);
1363     }
1364     return 0;
1365 }
1366
1367 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1368                         void *opaque)
1369 {
1370     VMStateField *field = vmsd->fields;
1371
1372     if (vmsd->pre_save) {
1373         vmsd->pre_save(opaque);
1374     }
1375     while(field->name) {
1376         if (!field->field_exists ||
1377             field->field_exists(opaque, vmsd->version_id)) {
1378             void *base_addr = opaque + field->offset;
1379             int i, n_elems = 1;
1380             int size = field->size;
1381
1382             if (field->flags & VMS_VBUFFER) {
1383                 size = *(int32_t *)(opaque+field->size_offset);
1384                 if (field->flags & VMS_MULTIPLY) {
1385                     size *= field->size;
1386                 }
1387             }
1388             if (field->flags & VMS_ARRAY) {
1389                 n_elems = field->num;
1390             } else if (field->flags & VMS_VARRAY_INT32) {
1391                 n_elems = *(int32_t *)(opaque+field->num_offset);
1392             } else if (field->flags & VMS_VARRAY_UINT16) {
1393                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1394             }
1395             if (field->flags & VMS_POINTER) {
1396                 base_addr = *(void **)base_addr + field->start;
1397             }
1398             for (i = 0; i < n_elems; i++) {
1399                 void *addr = base_addr + size * i;
1400
1401                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1402                     addr = *(void **)addr;
1403                 }
1404                 if (field->flags & VMS_STRUCT) {
1405                     vmstate_save_state(f, field->vmsd, addr);
1406                 } else {
1407                     field->info->put(f, addr, size);
1408                 }
1409             }
1410         }
1411         field++;
1412     }
1413     vmstate_subsection_save(f, vmsd, opaque);
1414 }
1415
1416 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1417 {
1418     if (!se->vmsd) {         /* Old style */
1419         return se->load_state(f, se->opaque, version_id);
1420     }
1421     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1422 }
1423
1424 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1425 {
1426     if (!se->vmsd) {         /* Old style */
1427         se->save_state(f, se->opaque);
1428         return;
1429     }
1430     vmstate_save_state(f,se->vmsd, se->opaque);
1431 }
1432
1433 #define QEMU_VM_FILE_MAGIC           0x5145564d
1434 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1435 #define QEMU_VM_FILE_VERSION         0x00000003
1436
1437 #define QEMU_VM_EOF                  0x00
1438 #define QEMU_VM_SECTION_START        0x01
1439 #define QEMU_VM_SECTION_PART         0x02
1440 #define QEMU_VM_SECTION_END          0x03
1441 #define QEMU_VM_SECTION_FULL         0x04
1442 #define QEMU_VM_SUBSECTION           0x05
1443
1444 bool qemu_savevm_state_blocked(Monitor *mon)
1445 {
1446     SaveStateEntry *se;
1447
1448     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1449         if (se->no_migrate) {
1450             monitor_printf(mon, "state blocked by non-migratable device '%s'\n",
1451                            se->idstr);
1452             return true;
1453         }
1454     }
1455     return false;
1456 }
1457
1458 int qemu_savevm_state_begin(Monitor *mon, QEMUFile *f, int blk_enable,
1459                             int shared)
1460 {
1461     SaveStateEntry *se;
1462
1463     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1464         if(se->set_params == NULL) {
1465             continue;
1466         }
1467         se->set_params(blk_enable, shared, se->opaque);
1468     }
1469
1470     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1471     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1472
1473     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1474         int len;
1475
1476         if (se->save_live_state == NULL)
1477             continue;
1478
1479         /* Section type */
1480         qemu_put_byte(f, QEMU_VM_SECTION_START);
1481         qemu_put_be32(f, se->section_id);
1482
1483         /* ID string */
1484         len = strlen(se->idstr);
1485         qemu_put_byte(f, len);
1486         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1487
1488         qemu_put_be32(f, se->instance_id);
1489         qemu_put_be32(f, se->version_id);
1490
1491         se->save_live_state(mon, f, QEMU_VM_SECTION_START, se->opaque);
1492     }
1493
1494     if (qemu_file_has_error(f)) {
1495         qemu_savevm_state_cancel(mon, f);
1496         return -EIO;
1497     }
1498
1499     return 0;
1500 }
1501
1502 int qemu_savevm_state_iterate(Monitor *mon, QEMUFile *f)
1503 {
1504     SaveStateEntry *se;
1505     int ret = 1;
1506
1507     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1508         if (se->save_live_state == NULL)
1509             continue;
1510
1511         /* Section type */
1512         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1513         qemu_put_be32(f, se->section_id);
1514
1515         ret = se->save_live_state(mon, f, QEMU_VM_SECTION_PART, se->opaque);
1516         if (!ret) {
1517             /* Do not proceed to the next vmstate before this one reported
1518                completion of the current stage. This serializes the migration
1519                and reduces the probability that a faster changing state is
1520                synchronized over and over again. */
1521             break;
1522         }
1523     }
1524
1525     if (ret)
1526         return 1;
1527
1528     if (qemu_file_has_error(f)) {
1529         qemu_savevm_state_cancel(mon, f);
1530         return -EIO;
1531     }
1532
1533     return 0;
1534 }
1535
1536 int qemu_savevm_state_complete(Monitor *mon, QEMUFile *f)
1537 {
1538     SaveStateEntry *se;
1539
1540     cpu_synchronize_all_states();
1541
1542     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1543         if (se->save_live_state == NULL)
1544             continue;
1545
1546         /* Section type */
1547         qemu_put_byte(f, QEMU_VM_SECTION_END);
1548         qemu_put_be32(f, se->section_id);
1549
1550         se->save_live_state(mon, f, QEMU_VM_SECTION_END, se->opaque);
1551     }
1552
1553     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1554         int len;
1555
1556         if (se->save_state == NULL && se->vmsd == NULL)
1557             continue;
1558
1559         /* Section type */
1560         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1561         qemu_put_be32(f, se->section_id);
1562
1563         /* ID string */
1564         len = strlen(se->idstr);
1565         qemu_put_byte(f, len);
1566         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1567
1568         qemu_put_be32(f, se->instance_id);
1569         qemu_put_be32(f, se->version_id);
1570
1571         vmstate_save(f, se);
1572     }
1573
1574     qemu_put_byte(f, QEMU_VM_EOF);
1575
1576     if (qemu_file_has_error(f))
1577         return -EIO;
1578
1579     return 0;
1580 }
1581
1582 void qemu_savevm_state_cancel(Monitor *mon, QEMUFile *f)
1583 {
1584     SaveStateEntry *se;
1585
1586     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1587         if (se->save_live_state) {
1588             se->save_live_state(mon, f, -1, se->opaque);
1589         }
1590     }
1591 }
1592
1593 static int qemu_savevm_state(Monitor *mon, QEMUFile *f)
1594 {
1595     int saved_vm_running;
1596     int ret;
1597
1598     saved_vm_running = vm_running;
1599     vm_stop(VMSTOP_SAVEVM);
1600
1601     if (qemu_savevm_state_blocked(mon)) {
1602         ret = -EINVAL;
1603         goto out;
1604     }
1605
1606     ret = qemu_savevm_state_begin(mon, f, 0, 0);
1607     if (ret < 0)
1608         goto out;
1609
1610     do {
1611         ret = qemu_savevm_state_iterate(mon, f);
1612         if (ret < 0)
1613             goto out;
1614     } while (ret == 0);
1615
1616     ret = qemu_savevm_state_complete(mon, f);
1617
1618 out:
1619     if (qemu_file_has_error(f))
1620         ret = -EIO;
1621
1622     if (!ret && saved_vm_running)
1623         vm_start();
1624
1625     return ret;
1626 }
1627
1628 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1629 {
1630     SaveStateEntry *se;
1631
1632     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1633         if (!strcmp(se->idstr, idstr) &&
1634             (instance_id == se->instance_id ||
1635              instance_id == se->alias_id))
1636             return se;
1637         /* Migrating from an older version? */
1638         if (strstr(se->idstr, idstr) && se->compat) {
1639             if (!strcmp(se->compat->idstr, idstr) &&
1640                 (instance_id == se->compat->instance_id ||
1641                  instance_id == se->alias_id))
1642                 return se;
1643         }
1644     }
1645     return NULL;
1646 }
1647
1648 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1649 {
1650     while(sub && sub->needed) {
1651         if (strcmp(idstr, sub->vmsd->name) == 0) {
1652             return sub->vmsd;
1653         }
1654         sub++;
1655     }
1656     return NULL;
1657 }
1658
1659 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1660                                    void *opaque)
1661 {
1662     const VMStateSubsection *sub = vmsd->subsections;
1663
1664     if (!sub || !sub->needed) {
1665         return 0;
1666     }
1667
1668     while (qemu_peek_byte(f) == QEMU_VM_SUBSECTION) {
1669         char idstr[256];
1670         int ret;
1671         uint8_t version_id, len;
1672         const VMStateDescription *sub_vmsd;
1673
1674         qemu_get_byte(f); /* subsection */
1675         len = qemu_get_byte(f);
1676         qemu_get_buffer(f, (uint8_t *)idstr, len);
1677         idstr[len] = 0;
1678         version_id = qemu_get_be32(f);
1679
1680         sub_vmsd = vmstate_get_subsection(sub, idstr);
1681         if (sub_vmsd == NULL) {
1682             return -ENOENT;
1683         }
1684         assert(!sub_vmsd->subsections);
1685         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1686         if (ret) {
1687             return ret;
1688         }
1689     }
1690     return 0;
1691 }
1692
1693 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1694                                     void *opaque)
1695 {
1696     const VMStateSubsection *sub = vmsd->subsections;
1697
1698     while (sub && sub->needed) {
1699         if (sub->needed(opaque)) {
1700             const VMStateDescription *vmsd = sub->vmsd;
1701             uint8_t len;
1702
1703             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1704             len = strlen(vmsd->name);
1705             qemu_put_byte(f, len);
1706             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1707             qemu_put_be32(f, vmsd->version_id);
1708             assert(!vmsd->subsections);
1709             vmstate_save_state(f, vmsd, opaque);
1710         }
1711         sub++;
1712     }
1713 }
1714
1715 typedef struct LoadStateEntry {
1716     QLIST_ENTRY(LoadStateEntry) entry;
1717     SaveStateEntry *se;
1718     int section_id;
1719     int version_id;
1720 } LoadStateEntry;
1721
1722 int qemu_loadvm_state(QEMUFile *f)
1723 {
1724     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1725         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1726     LoadStateEntry *le, *new_le;
1727     uint8_t section_type;
1728     unsigned int v;
1729     int ret;
1730
1731     if (qemu_savevm_state_blocked(default_mon)) {
1732         return -EINVAL;
1733     }
1734
1735     v = qemu_get_be32(f);
1736     if (v != QEMU_VM_FILE_MAGIC)
1737         return -EINVAL;
1738
1739     v = qemu_get_be32(f);
1740     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1741         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1742         return -ENOTSUP;
1743     }
1744     if (v != QEMU_VM_FILE_VERSION)
1745         return -ENOTSUP;
1746
1747     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1748         uint32_t instance_id, version_id, section_id;
1749         SaveStateEntry *se;
1750         char idstr[257];
1751         int len;
1752
1753         switch (section_type) {
1754         case QEMU_VM_SECTION_START:
1755         case QEMU_VM_SECTION_FULL:
1756             /* Read section start */
1757             section_id = qemu_get_be32(f);
1758             len = qemu_get_byte(f);
1759             qemu_get_buffer(f, (uint8_t *)idstr, len);
1760             idstr[len] = 0;
1761             instance_id = qemu_get_be32(f);
1762             version_id = qemu_get_be32(f);
1763
1764             /* Find savevm section */
1765             se = find_se(idstr, instance_id);
1766             if (se == NULL) {
1767                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1768                 ret = -EINVAL;
1769                 goto out;
1770             }
1771
1772             /* Validate version */
1773             if (version_id > se->version_id) {
1774                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
1775                         version_id, idstr, se->version_id);
1776                 ret = -EINVAL;
1777                 goto out;
1778             }
1779
1780             /* Add entry */
1781             le = qemu_mallocz(sizeof(*le));
1782
1783             le->se = se;
1784             le->section_id = section_id;
1785             le->version_id = version_id;
1786             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
1787
1788             ret = vmstate_load(f, le->se, le->version_id);
1789             if (ret < 0) {
1790                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
1791                         instance_id, idstr);
1792                 goto out;
1793             }
1794             break;
1795         case QEMU_VM_SECTION_PART:
1796         case QEMU_VM_SECTION_END:
1797             section_id = qemu_get_be32(f);
1798
1799             QLIST_FOREACH(le, &loadvm_handlers, entry) {
1800                 if (le->section_id == section_id) {
1801                     break;
1802                 }
1803             }
1804             if (le == NULL) {
1805                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
1806                 ret = -EINVAL;
1807                 goto out;
1808             }
1809
1810             ret = vmstate_load(f, le->se, le->version_id);
1811             if (ret < 0) {
1812                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
1813                         section_id);
1814                 goto out;
1815             }
1816             break;
1817         default:
1818             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
1819             ret = -EINVAL;
1820             goto out;
1821         }
1822     }
1823
1824     cpu_synchronize_all_post_init();
1825
1826     ret = 0;
1827
1828 out:
1829     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
1830         QLIST_REMOVE(le, entry);
1831         qemu_free(le);
1832     }
1833
1834     if (qemu_file_has_error(f))
1835         ret = -EIO;
1836
1837     return ret;
1838 }
1839
1840 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1841                               const char *name)
1842 {
1843     QEMUSnapshotInfo *sn_tab, *sn;
1844     int nb_sns, i, ret;
1845
1846     ret = -ENOENT;
1847     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1848     if (nb_sns < 0)
1849         return ret;
1850     for(i = 0; i < nb_sns; i++) {
1851         sn = &sn_tab[i];
1852         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1853             *sn_info = *sn;
1854             ret = 0;
1855             break;
1856         }
1857     }
1858     qemu_free(sn_tab);
1859     return ret;
1860 }
1861
1862 /*
1863  * Deletes snapshots of a given name in all opened images.
1864  */
1865 static int del_existing_snapshots(Monitor *mon, const char *name)
1866 {
1867     BlockDriverState *bs;
1868     QEMUSnapshotInfo sn1, *snapshot = &sn1;
1869     int ret;
1870
1871     bs = NULL;
1872     while ((bs = bdrv_next(bs))) {
1873         if (bdrv_can_snapshot(bs) &&
1874             bdrv_snapshot_find(bs, snapshot, name) >= 0)
1875         {
1876             ret = bdrv_snapshot_delete(bs, name);
1877             if (ret < 0) {
1878                 monitor_printf(mon,
1879                                "Error while deleting snapshot on '%s'\n",
1880                                bdrv_get_device_name(bs));
1881                 return -1;
1882             }
1883         }
1884     }
1885
1886     return 0;
1887 }
1888
1889 void do_savevm(Monitor *mon, const QDict *qdict)
1890 {
1891     BlockDriverState *bs, *bs1;
1892     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1893     int ret;
1894     QEMUFile *f;
1895     int saved_vm_running;
1896     uint32_t vm_state_size;
1897 #ifdef _WIN32
1898     struct _timeb tb;
1899     struct tm *ptm;
1900 #else
1901     struct timeval tv;
1902     struct tm tm;
1903 #endif
1904     const char *name = qdict_get_try_str(qdict, "name");
1905
1906     /* Verify if there is a device that doesn't support snapshots and is writable */
1907     bs = NULL;
1908     while ((bs = bdrv_next(bs))) {
1909
1910         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1911             continue;
1912         }
1913
1914         if (!bdrv_can_snapshot(bs)) {
1915             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
1916                                bdrv_get_device_name(bs));
1917             return;
1918         }
1919     }
1920
1921     bs = bdrv_snapshots();
1922     if (!bs) {
1923         monitor_printf(mon, "No block device can accept snapshots\n");
1924         return;
1925     }
1926
1927     saved_vm_running = vm_running;
1928     vm_stop(VMSTOP_SAVEVM);
1929
1930     memset(sn, 0, sizeof(*sn));
1931
1932     /* fill auxiliary fields */
1933 #ifdef _WIN32
1934     _ftime(&tb);
1935     sn->date_sec = tb.time;
1936     sn->date_nsec = tb.millitm * 1000000;
1937 #else
1938     gettimeofday(&tv, NULL);
1939     sn->date_sec = tv.tv_sec;
1940     sn->date_nsec = tv.tv_usec * 1000;
1941 #endif
1942     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1943
1944     if (name) {
1945         ret = bdrv_snapshot_find(bs, old_sn, name);
1946         if (ret >= 0) {
1947             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1948             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1949         } else {
1950             pstrcpy(sn->name, sizeof(sn->name), name);
1951         }
1952     } else {
1953 #ifdef _WIN32
1954         ptm = localtime(&tb.time);
1955         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", ptm);
1956 #else
1957         /* cast below needed for OpenBSD where tv_sec is still 'long' */
1958         localtime_r((const time_t *)&tv.tv_sec, &tm);
1959         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
1960 #endif
1961     }
1962
1963     /* Delete old snapshots of the same name */
1964     if (name && del_existing_snapshots(mon, name) < 0) {
1965         goto the_end;
1966     }
1967
1968     /* save the VM state */
1969     f = qemu_fopen_bdrv(bs, 1);
1970     if (!f) {
1971         monitor_printf(mon, "Could not open VM state file\n");
1972         goto the_end;
1973     }
1974     ret = qemu_savevm_state(mon, f);
1975     vm_state_size = qemu_ftell(f);
1976     qemu_fclose(f);
1977     if (ret < 0) {
1978         monitor_printf(mon, "Error %d while writing VM\n", ret);
1979         goto the_end;
1980     }
1981
1982     /* create the snapshots */
1983
1984     bs1 = NULL;
1985     while ((bs1 = bdrv_next(bs1))) {
1986         if (bdrv_can_snapshot(bs1)) {
1987             /* Write VM state size only to the image that contains the state */
1988             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1989             ret = bdrv_snapshot_create(bs1, sn);
1990             if (ret < 0) {
1991                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1992                                bdrv_get_device_name(bs1));
1993             }
1994         }
1995     }
1996
1997  the_end:
1998     if (saved_vm_running)
1999         vm_start();
2000 }
2001
2002 int load_vmstate(const char *name)
2003 {
2004     BlockDriverState *bs, *bs_vm_state;
2005     QEMUSnapshotInfo sn;
2006     QEMUFile *f;
2007     int ret;
2008
2009     bs_vm_state = bdrv_snapshots();
2010     if (!bs_vm_state) {
2011         error_report("No block device supports snapshots");
2012         return -ENOTSUP;
2013     }
2014
2015     /* Don't even try to load empty VM states */
2016     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
2017     if (ret < 0) {
2018         return ret;
2019     } else if (sn.vm_state_size == 0) {
2020         return -EINVAL;
2021     }
2022
2023     /* Verify if there is any device that doesn't support snapshots and is
2024     writable and check if the requested snapshot is available too. */
2025     bs = NULL;
2026     while ((bs = bdrv_next(bs))) {
2027
2028         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
2029             continue;
2030         }
2031
2032         if (!bdrv_can_snapshot(bs)) {
2033             error_report("Device '%s' is writable but does not support snapshots.",
2034                                bdrv_get_device_name(bs));
2035             return -ENOTSUP;
2036         }
2037
2038         ret = bdrv_snapshot_find(bs, &sn, name);
2039         if (ret < 0) {
2040             error_report("Device '%s' does not have the requested snapshot '%s'",
2041                            bdrv_get_device_name(bs), name);
2042             return ret;
2043         }
2044     }
2045
2046     /* Flush all IO requests so they don't interfere with the new state.  */
2047     qemu_aio_flush();
2048
2049     bs = NULL;
2050     while ((bs = bdrv_next(bs))) {
2051         if (bdrv_can_snapshot(bs)) {
2052             ret = bdrv_snapshot_goto(bs, name);
2053             if (ret < 0) {
2054                 error_report("Error %d while activating snapshot '%s' on '%s'",
2055                              ret, name, bdrv_get_device_name(bs));
2056                 return ret;
2057             }
2058         }
2059     }
2060
2061     /* restore the VM state */
2062     f = qemu_fopen_bdrv(bs_vm_state, 0);
2063     if (!f) {
2064         error_report("Could not open VM state file");
2065         return -EINVAL;
2066     }
2067
2068     ret = qemu_loadvm_state(f);
2069
2070     qemu_fclose(f);
2071     if (ret < 0) {
2072         error_report("Error %d while loading VM state", ret);
2073         return ret;
2074     }
2075
2076     return 0;
2077 }
2078
2079 void do_delvm(Monitor *mon, const QDict *qdict)
2080 {
2081     BlockDriverState *bs, *bs1;
2082     int ret;
2083     const char *name = qdict_get_str(qdict, "name");
2084
2085     bs = bdrv_snapshots();
2086     if (!bs) {
2087         monitor_printf(mon, "No block device supports snapshots\n");
2088         return;
2089     }
2090
2091     bs1 = NULL;
2092     while ((bs1 = bdrv_next(bs1))) {
2093         if (bdrv_can_snapshot(bs1)) {
2094             ret = bdrv_snapshot_delete(bs1, name);
2095             if (ret < 0) {
2096                 if (ret == -ENOTSUP)
2097                     monitor_printf(mon,
2098                                    "Snapshots not supported on device '%s'\n",
2099                                    bdrv_get_device_name(bs1));
2100                 else
2101                     monitor_printf(mon, "Error %d while deleting snapshot on "
2102                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2103             }
2104         }
2105     }
2106 }
2107
2108 void do_info_snapshots(Monitor *mon)
2109 {
2110     BlockDriverState *bs, *bs1;
2111     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2112     int nb_sns, i, ret, available;
2113     int total;
2114     int *available_snapshots;
2115     char buf[256];
2116
2117     bs = bdrv_snapshots();
2118     if (!bs) {
2119         monitor_printf(mon, "No available block device supports snapshots\n");
2120         return;
2121     }
2122
2123     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2124     if (nb_sns < 0) {
2125         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2126         return;
2127     }
2128
2129     if (nb_sns == 0) {
2130         monitor_printf(mon, "There is no snapshot available.\n");
2131         return;
2132     }
2133
2134     available_snapshots = qemu_mallocz(sizeof(int) * nb_sns);
2135     total = 0;
2136     for (i = 0; i < nb_sns; i++) {
2137         sn = &sn_tab[i];
2138         available = 1;
2139         bs1 = NULL;
2140
2141         while ((bs1 = bdrv_next(bs1))) {
2142             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2143                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2144                 if (ret < 0) {
2145                     available = 0;
2146                     break;
2147                 }
2148             }
2149         }
2150
2151         if (available) {
2152             available_snapshots[total] = i;
2153             total++;
2154         }
2155     }
2156
2157     if (total > 0) {
2158         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2159         for (i = 0; i < total; i++) {
2160             sn = &sn_tab[available_snapshots[i]];
2161             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2162         }
2163     } else {
2164         monitor_printf(mon, "There is no suitable snapshot available\n");
2165     }
2166
2167     qemu_free(sn_tab);
2168     qemu_free(available_snapshots);
2169
2170 }