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 /* 64 bit unsigned int */
 886
 887 static int get_uint64(QEMUFile *f, void *pv, size_t size)
 888 {
 889     uint64_t *v = pv;
 890     qemu_get_be64s(f, v);
 891     return 0;
 892 }
 893
 894 static void put_uint64(QEMUFile *f, void *pv, size_t size)
 895 {
 896     uint64_t *v = pv;
 897     qemu_put_be64s(f, v);
 898 }
 899
 900 const VMStateInfo vmstate_info_uint64 = {
 901     .name = "uint64",
 902     .get  = get_uint64,
 903     .put  = put_uint64,
 904 };
 905
 906 /* 8 bit int. See that the received value is the same than the one
 907    in the field */
 908
 909 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
 910 {
 911     uint8_t *v = pv;
 912     uint8_t v2;
 913     qemu_get_8s(f, &v2);
 914
 915     if (*v == v2)
 916         return 0;
 917     return -EINVAL;
 918 }
 919
 920 const VMStateInfo vmstate_info_uint8_equal = {
 921     .name = "uint8 equal",
 922     .get  = get_uint8_equal,
 923     .put  = put_uint8,
 924 };
 925
 926 /* 16 bit unsigned int int. See that the received value is the same than the one
 927    in the field */
 928
 929 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
 930 {
 931     uint16_t *v = pv;
 932     uint16_t v2;
 933     qemu_get_be16s(f, &v2);
 934
 935     if (*v == v2)
 936         return 0;
 937     return -EINVAL;
 938 }
 939
 940 const VMStateInfo vmstate_info_uint16_equal = {
 941     .name = "uint16 equal",
 942     .get  = get_uint16_equal,
 943     .put  = put_uint16,
 944 };
 945
 946 /* timers  */
 947
 948 static int get_timer(QEMUFile *f, void *pv, size_t size)
 949 {
 950     QEMUTimer *v = pv;
 951     qemu_get_timer(f, v);
 952     return 0;
 953 }
 954
 955 static void put_timer(QEMUFile *f, void *pv, size_t size)
 956 {
 957     QEMUTimer *v = pv;
 958     qemu_put_timer(f, v);
 959 }
 960
 961 const VMStateInfo vmstate_info_timer = {
 962     .name = "timer",
 963     .get  = get_timer,
 964     .put  = put_timer,
 965 };
 966
 967 /* uint8_t buffers */
 968
 969 static int get_buffer(QEMUFile *f, void *pv, size_t size)
 970 {
 971     uint8_t *v = pv;
 972     qemu_get_buffer(f, v, size);
 973     return 0;
 974 }
 975
 976 static void put_buffer(QEMUFile *f, void *pv, size_t size)
 977 {
 978     uint8_t *v = pv;
 979     qemu_put_buffer(f, v, size);
 980 }
 981
 982 const VMStateInfo vmstate_info_buffer = {
 983     .name = "buffer",
 984     .get  = get_buffer,
 985     .put  = put_buffer,
 986 };
 987
 988 /* unused buffers: space that was used for some fields that are
 989    not usefull anymore */
 990
 991 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
 992 {
 993     uint8_t buf[1024];
 994     int block_len;
 995
 996     while (size > 0) {
 997         block_len = MIN(sizeof(buf), size);
 998         size -= block_len;
 999         qemu_get_buffer(f, buf, block_len);
1000     }
1001    return 0;
1002 }
1003
1004 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1005 {
1006     static const uint8_t buf[1024];
1007     int block_len;
1008
1009     while (size > 0) {
1010         block_len = MIN(sizeof(buf), size);
1011         size -= block_len;
1012         qemu_put_buffer(f, buf, block_len);
1013     }
1014 }
1015
1016 const VMStateInfo vmstate_info_unused_buffer = {
1017     .name = "unused_buffer",
1018     .get  = get_unused_buffer,
1019     .put  = put_unused_buffer,
1020 };
1021
1022 typedef struct CompatEntry {
1023     char idstr[256];
1024     int instance_id;
1025 } CompatEntry;
1026
1027 typedef struct SaveStateEntry {
1028     QTAILQ_ENTRY(SaveStateEntry) entry;
1029     char idstr[256];
1030     int instance_id;
1031     int alias_id;
1032     int version_id;
1033     int section_id;
1034     SaveSetParamsHandler *set_params;
1035     SaveLiveStateHandler *save_live_state;
1036     SaveStateHandler *save_state;
1037     LoadStateHandler *load_state;
1038     const VMStateDescription *vmsd;
1039     void *opaque;
1040     CompatEntry *compat;
1041     int no_migrate;
1042 } SaveStateEntry;
1043
1044
1045 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1046     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1047 static int global_section_id;
1048
1049 static int calculate_new_instance_id(const char *idstr)
1050 {
1051     SaveStateEntry *se;
1052     int instance_id = 0;
1053
1054     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1055         if (strcmp(idstr, se->idstr) == 0
1056             && instance_id <= se->instance_id) {
1057             instance_id = se->instance_id + 1;
1058         }
1059     }
1060     return instance_id;
1061 }
1062
1063 static int calculate_compat_instance_id(const char *idstr)
1064 {
1065     SaveStateEntry *se;
1066     int instance_id = 0;
1067
1068     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1069         if (!se->compat)
1070             continue;
1071
1072         if (strcmp(idstr, se->compat->idstr) == 0
1073             && instance_id <= se->compat->instance_id) {
1074             instance_id = se->compat->instance_id + 1;
1075         }
1076     }
1077     return instance_id;
1078 }
1079
1080 /* TODO: Individual devices generally have very little idea about the rest
1081    of the system, so instance_id should be removed/replaced.
1082    Meanwhile pass -1 as instance_id if you do not already have a clearly
1083    distinguishing id for all instances of your device class. */
1084 int register_savevm_live(DeviceState *dev,
1085                          const char *idstr,
1086                          int instance_id,
1087                          int version_id,
1088                          SaveSetParamsHandler *set_params,
1089                          SaveLiveStateHandler *save_live_state,
1090                          SaveStateHandler *save_state,
1091                          LoadStateHandler *load_state,
1092                          void *opaque)
1093 {
1094     SaveStateEntry *se;
1095
1096     se = qemu_mallocz(sizeof(SaveStateEntry));
1097     se->version_id = version_id;
1098     se->section_id = global_section_id++;
1099     se->set_params = set_params;
1100     se->save_live_state = save_live_state;
1101     se->save_state = save_state;
1102     se->load_state = load_state;
1103     se->opaque = opaque;
1104     se->vmsd = NULL;
1105     se->no_migrate = 0;
1106
1107     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1108         char *id = dev->parent_bus->info->get_dev_path(dev);
1109         if (id) {
1110             pstrcpy(se->idstr, sizeof(se->idstr), id);
1111             pstrcat(se->idstr, sizeof(se->idstr), "/");
1112             qemu_free(id);
1113
1114             se->compat = qemu_mallocz(sizeof(CompatEntry));
1115             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1116             se->compat->instance_id = instance_id == -1 ?
1117                          calculate_compat_instance_id(idstr) : instance_id;
1118             instance_id = -1;
1119         }
1120     }
1121     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1122
1123     if (instance_id == -1) {
1124         se->instance_id = calculate_new_instance_id(se->idstr);
1125     } else {
1126         se->instance_id = instance_id;
1127     }
1128     assert(!se->compat || se->instance_id == 0);
1129     /* add at the end of list */
1130     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1131     return 0;
1132 }
1133
1134 int register_savevm(DeviceState *dev,
1135                     const char *idstr,
1136                     int instance_id,
1137                     int version_id,
1138                     SaveStateHandler *save_state,
1139                     LoadStateHandler *load_state,
1140                     void *opaque)
1141 {
1142     return register_savevm_live(dev, idstr, instance_id, version_id,
1143                                 NULL, NULL, save_state, load_state, opaque);
1144 }
1145
1146 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1147 {
1148     SaveStateEntry *se, *new_se;
1149     char id[256] = "";
1150
1151     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1152         char *path = dev->parent_bus->info->get_dev_path(dev);
1153         if (path) {
1154             pstrcpy(id, sizeof(id), path);
1155             pstrcat(id, sizeof(id), "/");
1156             qemu_free(path);
1157         }
1158     }
1159     pstrcat(id, sizeof(id), idstr);
1160
1161     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1162         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1163             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1164             if (se->compat) {
1165                 qemu_free(se->compat);
1166             }
1167             qemu_free(se);
1168         }
1169     }
1170 }
1171
1172 /* mark a device as not to be migrated, that is the device should be
1173    unplugged before migration */
1174 void register_device_unmigratable(DeviceState *dev, const char *idstr,
1175                                                             void *opaque)
1176 {
1177     SaveStateEntry *se;
1178     char id[256] = "";
1179
1180     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1181         char *path = dev->parent_bus->info->get_dev_path(dev);
1182         if (path) {
1183             pstrcpy(id, sizeof(id), path);
1184             pstrcat(id, sizeof(id), "/");
1185             qemu_free(path);
1186         }
1187     }
1188     pstrcat(id, sizeof(id), idstr);
1189
1190     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1191         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1192             se->no_migrate = 1;
1193         }
1194     }
1195 }
1196
1197 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1198                                    const VMStateDescription *vmsd,
1199                                    void *opaque, int alias_id,
1200                                    int required_for_version)
1201 {
1202     SaveStateEntry *se;
1203
1204     /* If this triggers, alias support can be dropped for the vmsd. */
1205     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1206
1207     se = qemu_mallocz(sizeof(SaveStateEntry));
1208     se->version_id = vmsd->version_id;
1209     se->section_id = global_section_id++;
1210     se->save_live_state = NULL;
1211     se->save_state = NULL;
1212     se->load_state = NULL;
1213     se->opaque = opaque;
1214     se->vmsd = vmsd;
1215     se->alias_id = alias_id;
1216
1217     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1218         char *id = dev->parent_bus->info->get_dev_path(dev);
1219         if (id) {
1220             pstrcpy(se->idstr, sizeof(se->idstr), id);
1221             pstrcat(se->idstr, sizeof(se->idstr), "/");
1222             qemu_free(id);
1223
1224             se->compat = qemu_mallocz(sizeof(CompatEntry));
1225             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1226             se->compat->instance_id = instance_id == -1 ?
1227                          calculate_compat_instance_id(vmsd->name) : instance_id;
1228             instance_id = -1;
1229         }
1230     }
1231     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1232
1233     if (instance_id == -1) {
1234         se->instance_id = calculate_new_instance_id(se->idstr);
1235     } else {
1236         se->instance_id = instance_id;
1237     }
1238     assert(!se->compat || se->instance_id == 0);
1239     /* add at the end of list */
1240     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1241     return 0;
1242 }
1243
1244 int vmstate_register(DeviceState *dev, int instance_id,
1245                      const VMStateDescription *vmsd, void *opaque)
1246 {
1247     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1248                                           opaque, -1, 0);
1249 }
1250
1251 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1252                         void *opaque)
1253 {
1254     SaveStateEntry *se, *new_se;
1255
1256     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1257         if (se->vmsd == vmsd && se->opaque == opaque) {
1258             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1259             if (se->compat) {
1260                 qemu_free(se->compat);
1261             }
1262             qemu_free(se);
1263         }
1264     }
1265 }
1266
1267 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1268                                     void *opaque);
1269 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1270                                    void *opaque);
1271
1272 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1273                        void *opaque, int version_id)
1274 {
1275     VMStateField *field = vmsd->fields;
1276     int ret;
1277
1278     if (version_id > vmsd->version_id) {
1279         return -EINVAL;
1280     }
1281     if (version_id < vmsd->minimum_version_id_old) {
1282         return -EINVAL;
1283     }
1284     if  (version_id < vmsd->minimum_version_id) {
1285         return vmsd->load_state_old(f, opaque, version_id);
1286     }
1287     if (vmsd->pre_load) {
1288         int ret = vmsd->pre_load(opaque);
1289         if (ret)
1290             return ret;
1291     }
1292     while(field->name) {
1293         if ((field->field_exists &&
1294              field->field_exists(opaque, version_id)) ||
1295             (!field->field_exists &&
1296              field->version_id <= version_id)) {
1297             void *base_addr = opaque + field->offset;
1298             int i, n_elems = 1;
1299             int size = field->size;
1300
1301             if (field->flags & VMS_VBUFFER) {
1302                 size = *(int32_t *)(opaque+field->size_offset);
1303                 if (field->flags & VMS_MULTIPLY) {
1304                     size *= field->size;
1305                 }
1306             }
1307             if (field->flags & VMS_ARRAY) {
1308                 n_elems = field->num;
1309             } else if (field->flags & VMS_VARRAY_INT32) {
1310                 n_elems = *(int32_t *)(opaque+field->num_offset);
1311             } else if (field->flags & VMS_VARRAY_UINT16) {
1312                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1313             }
1314             if (field->flags & VMS_POINTER) {
1315                 base_addr = *(void **)base_addr + field->start;
1316             }
1317             for (i = 0; i < n_elems; i++) {
1318                 void *addr = base_addr + size * i;
1319
1320                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1321                     addr = *(void **)addr;
1322                 }
1323                 if (field->flags & VMS_STRUCT) {
1324                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1325                 } else {
1326                     ret = field->info->get(f, addr, size);
1327
1328                 }
1329                 if (ret < 0) {
1330                     return ret;
1331                 }
1332             }
1333         }
1334         field++;
1335     }
1336     ret = vmstate_subsection_load(f, vmsd, opaque);
1337     if (ret != 0) {
1338         return ret;
1339     }
1340     if (vmsd->post_load) {
1341         return vmsd->post_load(opaque, version_id);
1342     }
1343     return 0;
1344 }
1345
1346 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1347                         void *opaque)
1348 {
1349     VMStateField *field = vmsd->fields;
1350
1351     if (vmsd->pre_save) {
1352         vmsd->pre_save(opaque);
1353     }
1354     while(field->name) {
1355         if (!field->field_exists ||
1356             field->field_exists(opaque, vmsd->version_id)) {
1357             void *base_addr = opaque + field->offset;
1358             int i, n_elems = 1;
1359             int size = field->size;
1360
1361             if (field->flags & VMS_VBUFFER) {
1362                 size = *(int32_t *)(opaque+field->size_offset);
1363                 if (field->flags & VMS_MULTIPLY) {
1364                     size *= field->size;
1365                 }
1366             }
1367             if (field->flags & VMS_ARRAY) {
1368                 n_elems = field->num;
1369             } else if (field->flags & VMS_VARRAY_INT32) {
1370                 n_elems = *(int32_t *)(opaque+field->num_offset);
1371             } else if (field->flags & VMS_VARRAY_UINT16) {
1372                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1373             }
1374             if (field->flags & VMS_POINTER) {
1375                 base_addr = *(void **)base_addr + field->start;
1376             }
1377             for (i = 0; i < n_elems; i++) {
1378                 void *addr = base_addr + size * i;
1379
1380                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1381                     addr = *(void **)addr;
1382                 }
1383                 if (field->flags & VMS_STRUCT) {
1384                     vmstate_save_state(f, field->vmsd, addr);
1385                 } else {
1386                     field->info->put(f, addr, size);
1387                 }
1388             }
1389         }
1390         field++;
1391     }
1392     vmstate_subsection_save(f, vmsd, opaque);
1393 }
1394
1395 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1396 {
1397     if (!se->vmsd) {         /* Old style */
1398         return se->load_state(f, se->opaque, version_id);
1399     }
1400     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1401 }
1402
1403 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1404 {
1405     if (!se->vmsd) {         /* Old style */
1406         se->save_state(f, se->opaque);
1407         return;
1408     }
1409     vmstate_save_state(f,se->vmsd, se->opaque);
1410 }
1411
1412 #define QEMU_VM_FILE_MAGIC           0x5145564d
1413 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1414 #define QEMU_VM_FILE_VERSION         0x00000003
1415
1416 #define QEMU_VM_EOF                  0x00
1417 #define QEMU_VM_SECTION_START        0x01
1418 #define QEMU_VM_SECTION_PART         0x02
1419 #define QEMU_VM_SECTION_END          0x03
1420 #define QEMU_VM_SECTION_FULL         0x04
1421 #define QEMU_VM_SUBSECTION           0x05
1422
1423 bool qemu_savevm_state_blocked(Monitor *mon)
1424 {
1425     SaveStateEntry *se;
1426
1427     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1428         if (se->no_migrate) {
1429             monitor_printf(mon, "state blocked by non-migratable device '%s'\n",
1430                            se->idstr);
1431             return true;
1432         }
1433     }
1434     return false;
1435 }
1436
1437 int qemu_savevm_state_begin(Monitor *mon, QEMUFile *f, int blk_enable,
1438                             int shared)
1439 {
1440     SaveStateEntry *se;
1441
1442     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1443         if(se->set_params == NULL) {
1444             continue;
1445         }
1446         se->set_params(blk_enable, shared, se->opaque);
1447     }
1448
1449     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1450     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1451
1452     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1453         int len;
1454
1455         if (se->save_live_state == NULL)
1456             continue;
1457
1458         /* Section type */
1459         qemu_put_byte(f, QEMU_VM_SECTION_START);
1460         qemu_put_be32(f, se->section_id);
1461
1462         /* ID string */
1463         len = strlen(se->idstr);
1464         qemu_put_byte(f, len);
1465         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1466
1467         qemu_put_be32(f, se->instance_id);
1468         qemu_put_be32(f, se->version_id);
1469
1470         se->save_live_state(mon, f, QEMU_VM_SECTION_START, se->opaque);
1471     }
1472
1473     if (qemu_file_has_error(f)) {
1474         qemu_savevm_state_cancel(mon, f);
1475         return -EIO;
1476     }
1477
1478     return 0;
1479 }
1480
1481 int qemu_savevm_state_iterate(Monitor *mon, QEMUFile *f)
1482 {
1483     SaveStateEntry *se;
1484     int ret = 1;
1485
1486     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1487         if (se->save_live_state == NULL)
1488             continue;
1489
1490         /* Section type */
1491         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1492         qemu_put_be32(f, se->section_id);
1493
1494         ret = se->save_live_state(mon, f, QEMU_VM_SECTION_PART, se->opaque);
1495         if (!ret) {
1496             /* Do not proceed to the next vmstate before this one reported
1497                completion of the current stage. This serializes the migration
1498                and reduces the probability that a faster changing state is
1499                synchronized over and over again. */
1500             break;
1501         }
1502     }
1503
1504     if (ret)
1505         return 1;
1506
1507     if (qemu_file_has_error(f)) {
1508         qemu_savevm_state_cancel(mon, f);
1509         return -EIO;
1510     }
1511
1512     return 0;
1513 }
1514
1515 int qemu_savevm_state_complete(Monitor *mon, QEMUFile *f)
1516 {
1517     SaveStateEntry *se;
1518
1519     cpu_synchronize_all_states();
1520
1521     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1522         if (se->save_live_state == NULL)
1523             continue;
1524
1525         /* Section type */
1526         qemu_put_byte(f, QEMU_VM_SECTION_END);
1527         qemu_put_be32(f, se->section_id);
1528
1529         se->save_live_state(mon, f, QEMU_VM_SECTION_END, se->opaque);
1530     }
1531
1532     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1533         int len;
1534
1535         if (se->save_state == NULL && se->vmsd == NULL)
1536             continue;
1537
1538         /* Section type */
1539         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1540         qemu_put_be32(f, se->section_id);
1541
1542         /* ID string */
1543         len = strlen(se->idstr);
1544         qemu_put_byte(f, len);
1545         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1546
1547         qemu_put_be32(f, se->instance_id);
1548         qemu_put_be32(f, se->version_id);
1549
1550         vmstate_save(f, se);
1551     }
1552
1553     qemu_put_byte(f, QEMU_VM_EOF);
1554
1555     if (qemu_file_has_error(f))
1556         return -EIO;
1557
1558     return 0;
1559 }
1560
1561 void qemu_savevm_state_cancel(Monitor *mon, QEMUFile *f)
1562 {
1563     SaveStateEntry *se;
1564
1565     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1566         if (se->save_live_state) {
1567             se->save_live_state(mon, f, -1, se->opaque);
1568         }
1569     }
1570 }
1571
1572 static int qemu_savevm_state(Monitor *mon, QEMUFile *f)
1573 {
1574     int saved_vm_running;
1575     int ret;
1576
1577     saved_vm_running = vm_running;
1578     vm_stop(0);
1579
1580     if (qemu_savevm_state_blocked(mon)) {
1581         ret = -EINVAL;
1582         goto out;
1583     }
1584
1585     ret = qemu_savevm_state_begin(mon, f, 0, 0);
1586     if (ret < 0)
1587         goto out;
1588
1589     do {
1590         ret = qemu_savevm_state_iterate(mon, f);
1591         if (ret < 0)
1592             goto out;
1593     } while (ret == 0);
1594
1595     ret = qemu_savevm_state_complete(mon, f);
1596
1597 out:
1598     if (qemu_file_has_error(f))
1599         ret = -EIO;
1600
1601     if (!ret && saved_vm_running)
1602         vm_start();
1603
1604     return ret;
1605 }
1606
1607 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1608 {
1609     SaveStateEntry *se;
1610
1611     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1612         if (!strcmp(se->idstr, idstr) &&
1613             (instance_id == se->instance_id ||
1614              instance_id == se->alias_id))
1615             return se;
1616         /* Migrating from an older version? */
1617         if (strstr(se->idstr, idstr) && se->compat) {
1618             if (!strcmp(se->compat->idstr, idstr) &&
1619                 (instance_id == se->compat->instance_id ||
1620                  instance_id == se->alias_id))
1621                 return se;
1622         }
1623     }
1624     return NULL;
1625 }
1626
1627 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1628 {
1629     while(sub && sub->needed) {
1630         if (strcmp(idstr, sub->vmsd->name) == 0) {
1631             return sub->vmsd;
1632         }
1633         sub++;
1634     }
1635     return NULL;
1636 }
1637
1638 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1639                                    void *opaque)
1640 {
1641     const VMStateSubsection *sub = vmsd->subsections;
1642
1643     if (!sub || !sub->needed) {
1644         return 0;
1645     }
1646
1647     while (qemu_peek_byte(f) == QEMU_VM_SUBSECTION) {
1648         char idstr[256];
1649         int ret;
1650         uint8_t version_id, len;
1651         const VMStateDescription *sub_vmsd;
1652
1653         qemu_get_byte(f); /* subsection */
1654         len = qemu_get_byte(f);
1655         qemu_get_buffer(f, (uint8_t *)idstr, len);
1656         idstr[len] = 0;
1657         version_id = qemu_get_be32(f);
1658
1659         sub_vmsd = vmstate_get_subsection(sub, idstr);
1660         if (sub_vmsd == NULL) {
1661             return -ENOENT;
1662         }
1663         assert(!sub_vmsd->subsections);
1664         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1665         if (ret) {
1666             return ret;
1667         }
1668     }
1669     return 0;
1670 }
1671
1672 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1673                                     void *opaque)
1674 {
1675     const VMStateSubsection *sub = vmsd->subsections;
1676
1677     while (sub && sub->needed) {
1678         if (sub->needed(opaque)) {
1679             const VMStateDescription *vmsd = sub->vmsd;
1680             uint8_t len;
1681
1682             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1683             len = strlen(vmsd->name);
1684             qemu_put_byte(f, len);
1685             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1686             qemu_put_be32(f, vmsd->version_id);
1687             assert(!vmsd->subsections);
1688             vmstate_save_state(f, vmsd, opaque);
1689         }
1690         sub++;
1691     }
1692 }
1693
1694 typedef struct LoadStateEntry {
1695     QLIST_ENTRY(LoadStateEntry) entry;
1696     SaveStateEntry *se;
1697     int section_id;
1698     int version_id;
1699 } LoadStateEntry;
1700
1701 int qemu_loadvm_state(QEMUFile *f)
1702 {
1703     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1704         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1705     LoadStateEntry *le, *new_le;
1706     uint8_t section_type;
1707     unsigned int v;
1708     int ret;
1709
1710     if (qemu_savevm_state_blocked(default_mon)) {
1711         return -EINVAL;
1712     }
1713
1714     v = qemu_get_be32(f);
1715     if (v != QEMU_VM_FILE_MAGIC)
1716         return -EINVAL;
1717
1718     v = qemu_get_be32(f);
1719     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1720         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1721         return -ENOTSUP;
1722     }
1723     if (v != QEMU_VM_FILE_VERSION)
1724         return -ENOTSUP;
1725
1726     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1727         uint32_t instance_id, version_id, section_id;
1728         SaveStateEntry *se;
1729         char idstr[257];
1730         int len;
1731
1732         switch (section_type) {
1733         case QEMU_VM_SECTION_START:
1734         case QEMU_VM_SECTION_FULL:
1735             /* Read section start */
1736             section_id = qemu_get_be32(f);
1737             len = qemu_get_byte(f);
1738             qemu_get_buffer(f, (uint8_t *)idstr, len);
1739             idstr[len] = 0;
1740             instance_id = qemu_get_be32(f);
1741             version_id = qemu_get_be32(f);
1742
1743             /* Find savevm section */
1744             se = find_se(idstr, instance_id);
1745             if (se == NULL) {
1746                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1747                 ret = -EINVAL;
1748                 goto out;
1749             }
1750
1751             /* Validate version */
1752             if (version_id > se->version_id) {
1753                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
1754                         version_id, idstr, se->version_id);
1755                 ret = -EINVAL;
1756                 goto out;
1757             }
1758
1759             /* Add entry */
1760             le = qemu_mallocz(sizeof(*le));
1761
1762             le->se = se;
1763             le->section_id = section_id;
1764             le->version_id = version_id;
1765             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
1766
1767             ret = vmstate_load(f, le->se, le->version_id);
1768             if (ret < 0) {
1769                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
1770                         instance_id, idstr);
1771                 goto out;
1772             }
1773             break;
1774         case QEMU_VM_SECTION_PART:
1775         case QEMU_VM_SECTION_END:
1776             section_id = qemu_get_be32(f);
1777
1778             QLIST_FOREACH(le, &loadvm_handlers, entry) {
1779                 if (le->section_id == section_id) {
1780                     break;
1781                 }
1782             }
1783             if (le == NULL) {
1784                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
1785                 ret = -EINVAL;
1786                 goto out;
1787             }
1788
1789             ret = vmstate_load(f, le->se, le->version_id);
1790             if (ret < 0) {
1791                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
1792                         section_id);
1793                 goto out;
1794             }
1795             break;
1796         default:
1797             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
1798             ret = -EINVAL;
1799             goto out;
1800         }
1801     }
1802
1803     cpu_synchronize_all_post_init();
1804
1805     ret = 0;
1806
1807 out:
1808     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
1809         QLIST_REMOVE(le, entry);
1810         qemu_free(le);
1811     }
1812
1813     if (qemu_file_has_error(f))
1814         ret = -EIO;
1815
1816     return ret;
1817 }
1818
1819 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1820                               const char *name)
1821 {
1822     QEMUSnapshotInfo *sn_tab, *sn;
1823     int nb_sns, i, ret;
1824
1825     ret = -ENOENT;
1826     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1827     if (nb_sns < 0)
1828         return ret;
1829     for(i = 0; i < nb_sns; i++) {
1830         sn = &sn_tab[i];
1831         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1832             *sn_info = *sn;
1833             ret = 0;
1834             break;
1835         }
1836     }
1837     qemu_free(sn_tab);
1838     return ret;
1839 }
1840
1841 /*
1842  * Deletes snapshots of a given name in all opened images.
1843  */
1844 static int del_existing_snapshots(Monitor *mon, const char *name)
1845 {
1846     BlockDriverState *bs;
1847     QEMUSnapshotInfo sn1, *snapshot = &sn1;
1848     int ret;
1849
1850     bs = NULL;
1851     while ((bs = bdrv_next(bs))) {
1852         if (bdrv_can_snapshot(bs) &&
1853             bdrv_snapshot_find(bs, snapshot, name) >= 0)
1854         {
1855             ret = bdrv_snapshot_delete(bs, name);
1856             if (ret < 0) {
1857                 monitor_printf(mon,
1858                                "Error while deleting snapshot on '%s'\n",
1859                                bdrv_get_device_name(bs));
1860                 return -1;
1861             }
1862         }
1863     }
1864
1865     return 0;
1866 }
1867
1868 void do_savevm(Monitor *mon, const QDict *qdict)
1869 {
1870     BlockDriverState *bs, *bs1;
1871     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1872     int ret;
1873     QEMUFile *f;
1874     int saved_vm_running;
1875     uint32_t vm_state_size;
1876 #ifdef _WIN32
1877     struct _timeb tb;
1878     struct tm *ptm;
1879 #else
1880     struct timeval tv;
1881     struct tm tm;
1882 #endif
1883     const char *name = qdict_get_try_str(qdict, "name");
1884
1885     /* Verify if there is a device that doesn't support snapshots and is writable */
1886     bs = NULL;
1887     while ((bs = bdrv_next(bs))) {
1888
1889         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1890             continue;
1891         }
1892
1893         if (!bdrv_can_snapshot(bs)) {
1894             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
1895                                bdrv_get_device_name(bs));
1896             return;
1897         }
1898     }
1899
1900     bs = bdrv_snapshots();
1901     if (!bs) {
1902         monitor_printf(mon, "No block device can accept snapshots\n");
1903         return;
1904     }
1905
1906     saved_vm_running = vm_running;
1907     vm_stop(0);
1908
1909     memset(sn, 0, sizeof(*sn));
1910
1911     /* fill auxiliary fields */
1912 #ifdef _WIN32
1913     _ftime(&tb);
1914     sn->date_sec = tb.time;
1915     sn->date_nsec = tb.millitm * 1000000;
1916 #else
1917     gettimeofday(&tv, NULL);
1918     sn->date_sec = tv.tv_sec;
1919     sn->date_nsec = tv.tv_usec * 1000;
1920 #endif
1921     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1922
1923     if (name) {
1924         ret = bdrv_snapshot_find(bs, old_sn, name);
1925         if (ret >= 0) {
1926             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1927             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1928         } else {
1929             pstrcpy(sn->name, sizeof(sn->name), name);
1930         }
1931     } else {
1932 #ifdef _WIN32
1933         ptm = localtime(&tb.time);
1934         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", ptm);
1935 #else
1936         /* cast below needed for OpenBSD where tv_sec is still 'long' */
1937         localtime_r((const time_t *)&tv.tv_sec, &tm);
1938         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
1939 #endif
1940     }
1941
1942     /* Delete old snapshots of the same name */
1943     if (name && del_existing_snapshots(mon, name) < 0) {
1944         goto the_end;
1945     }
1946
1947     /* save the VM state */
1948     f = qemu_fopen_bdrv(bs, 1);
1949     if (!f) {
1950         monitor_printf(mon, "Could not open VM state file\n");
1951         goto the_end;
1952     }
1953     ret = qemu_savevm_state(mon, f);
1954     vm_state_size = qemu_ftell(f);
1955     qemu_fclose(f);
1956     if (ret < 0) {
1957         monitor_printf(mon, "Error %d while writing VM\n", ret);
1958         goto the_end;
1959     }
1960
1961     /* create the snapshots */
1962
1963     bs1 = NULL;
1964     while ((bs1 = bdrv_next(bs1))) {
1965         if (bdrv_can_snapshot(bs1)) {
1966             /* Write VM state size only to the image that contains the state */
1967             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1968             ret = bdrv_snapshot_create(bs1, sn);
1969             if (ret < 0) {
1970                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1971                                bdrv_get_device_name(bs1));
1972             }
1973         }
1974     }
1975
1976  the_end:
1977     if (saved_vm_running)
1978         vm_start();
1979 }
1980
1981 int load_vmstate(const char *name)
1982 {
1983     BlockDriverState *bs, *bs_vm_state;
1984     QEMUSnapshotInfo sn;
1985     QEMUFile *f;
1986     int ret;
1987
1988     bs_vm_state = bdrv_snapshots();
1989     if (!bs_vm_state) {
1990         error_report("No block device supports snapshots");
1991         return -ENOTSUP;
1992     }
1993
1994     /* Don't even try to load empty VM states */
1995     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
1996     if (ret < 0) {
1997         return ret;
1998     } else if (sn.vm_state_size == 0) {
1999         return -EINVAL;
2000     }
2001
2002     /* Verify if there is any device that doesn't support snapshots and is
2003     writable and check if the requested snapshot is available too. */
2004     bs = NULL;
2005     while ((bs = bdrv_next(bs))) {
2006
2007         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
2008             continue;
2009         }
2010
2011         if (!bdrv_can_snapshot(bs)) {
2012             error_report("Device '%s' is writable but does not support snapshots.",
2013                                bdrv_get_device_name(bs));
2014             return -ENOTSUP;
2015         }
2016
2017         ret = bdrv_snapshot_find(bs, &sn, name);
2018         if (ret < 0) {
2019             error_report("Device '%s' does not have the requested snapshot '%s'",
2020                            bdrv_get_device_name(bs), name);
2021             return ret;
2022         }
2023     }
2024
2025     /* Flush all IO requests so they don't interfere with the new state.  */
2026     qemu_aio_flush();
2027
2028     bs = NULL;
2029     while ((bs = bdrv_next(bs))) {
2030         if (bdrv_can_snapshot(bs)) {
2031             ret = bdrv_snapshot_goto(bs, name);
2032             if (ret < 0) {
2033                 error_report("Error %d while activating snapshot '%s' on '%s'",
2034                              ret, name, bdrv_get_device_name(bs));
2035                 return ret;
2036             }
2037         }
2038     }
2039
2040     /* restore the VM state */
2041     f = qemu_fopen_bdrv(bs_vm_state, 0);
2042     if (!f) {
2043         error_report("Could not open VM state file");
2044         return -EINVAL;
2045     }
2046
2047     ret = qemu_loadvm_state(f);
2048
2049     qemu_fclose(f);
2050     if (ret < 0) {
2051         error_report("Error %d while loading VM state", ret);
2052         return ret;
2053     }
2054
2055     return 0;
2056 }
2057
2058 void do_delvm(Monitor *mon, const QDict *qdict)
2059 {
2060     BlockDriverState *bs, *bs1;
2061     int ret;
2062     const char *name = qdict_get_str(qdict, "name");
2063
2064     bs = bdrv_snapshots();
2065     if (!bs) {
2066         monitor_printf(mon, "No block device supports snapshots\n");
2067         return;
2068     }
2069
2070     bs1 = NULL;
2071     while ((bs1 = bdrv_next(bs1))) {
2072         if (bdrv_can_snapshot(bs1)) {
2073             ret = bdrv_snapshot_delete(bs1, name);
2074             if (ret < 0) {
2075                 if (ret == -ENOTSUP)
2076                     monitor_printf(mon,
2077                                    "Snapshots not supported on device '%s'\n",
2078                                    bdrv_get_device_name(bs1));
2079                 else
2080                     monitor_printf(mon, "Error %d while deleting snapshot on "
2081                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2082             }
2083         }
2084     }
2085 }
2086
2087 void do_info_snapshots(Monitor *mon)
2088 {
2089     BlockDriverState *bs, *bs1;
2090     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2091     int nb_sns, i, ret, available;
2092     int total;
2093     int *available_snapshots;
2094     char buf[256];
2095
2096     bs = bdrv_snapshots();
2097     if (!bs) {
2098         monitor_printf(mon, "No available block device supports snapshots\n");
2099         return;
2100     }
2101
2102     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2103     if (nb_sns < 0) {
2104         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2105         return;
2106     }
2107
2108     if (nb_sns == 0) {
2109         monitor_printf(mon, "There is no snapshot available.\n");
2110         return;
2111     }
2112
2113     available_snapshots = qemu_mallocz(sizeof(int) * nb_sns);
2114     total = 0;
2115     for (i = 0; i < nb_sns; i++) {
2116         sn = &sn_tab[i];
2117         available = 1;
2118         bs1 = NULL;
2119
2120         while ((bs1 = bdrv_next(bs1))) {
2121             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2122                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2123                 if (ret < 0) {
2124                     available = 0;
2125                     break;
2126                 }
2127             }
2128         }
2129
2130         if (available) {
2131             available_snapshots[total] = i;
2132             total++;
2133         }
2134     }
2135
2136     if (total > 0) {
2137         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2138         for (i = 0; i < total; i++) {
2139             sn = &sn_tab[available_snapshots[i]];
2140             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2141         }
2142     } else {
2143         monitor_printf(mon, "There is no suitable snapshot available\n");
2144     }
2145
2146     qemu_free(sn_tab);
2147     qemu_free(available_snapshots);
2148
2149 }