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