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