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 /* bool */
 679
 680 static int get_bool(QEMUFile *f, void *pv, size_t size)
 681 {
 682     bool *v = pv;
 683     *v = qemu_get_byte(f);
 684     return 0;
 685 }
 686
 687 static void put_bool(QEMUFile *f, void *pv, size_t size)
 688 {
 689     bool *v = pv;
 690     qemu_put_byte(f, *v);
 691 }
 692
 693 const VMStateInfo vmstate_info_bool = {
 694     .name = "bool",
 695     .get  = get_bool,
 696     .put  = put_bool,
 697 };
 698
 699 /* 8 bit int */
 700
 701 static int get_int8(QEMUFile *f, void *pv, size_t size)
 702 {
 703     int8_t *v = pv;
 704     qemu_get_s8s(f, v);
 705     return 0;
 706 }
 707
 708 static void put_int8(QEMUFile *f, void *pv, size_t size)
 709 {
 710     int8_t *v = pv;
 711     qemu_put_s8s(f, v);
 712 }
 713
 714 const VMStateInfo vmstate_info_int8 = {
 715     .name = "int8",
 716     .get  = get_int8,
 717     .put  = put_int8,
 718 };
 719
 720 /* 16 bit int */
 721
 722 static int get_int16(QEMUFile *f, void *pv, size_t size)
 723 {
 724     int16_t *v = pv;
 725     qemu_get_sbe16s(f, v);
 726     return 0;
 727 }
 728
 729 static void put_int16(QEMUFile *f, void *pv, size_t size)
 730 {
 731     int16_t *v = pv;
 732     qemu_put_sbe16s(f, v);
 733 }
 734
 735 const VMStateInfo vmstate_info_int16 = {
 736     .name = "int16",
 737     .get  = get_int16,
 738     .put  = put_int16,
 739 };
 740
 741 /* 32 bit int */
 742
 743 static int get_int32(QEMUFile *f, void *pv, size_t size)
 744 {
 745     int32_t *v = pv;
 746     qemu_get_sbe32s(f, v);
 747     return 0;
 748 }
 749
 750 static void put_int32(QEMUFile *f, void *pv, size_t size)
 751 {
 752     int32_t *v = pv;
 753     qemu_put_sbe32s(f, v);
 754 }
 755
 756 const VMStateInfo vmstate_info_int32 = {
 757     .name = "int32",
 758     .get  = get_int32,
 759     .put  = put_int32,
 760 };
 761
 762 /* 32 bit int. See that the received value is the same than the one
 763    in the field */
 764
 765 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 766 {
 767     int32_t *v = pv;
 768     int32_t v2;
 769     qemu_get_sbe32s(f, &v2);
 770
 771     if (*v == v2)
 772         return 0;
 773     return -EINVAL;
 774 }
 775
 776 const VMStateInfo vmstate_info_int32_equal = {
 777     .name = "int32 equal",
 778     .get  = get_int32_equal,
 779     .put  = put_int32,
 780 };
 781
 782 /* 32 bit int. See that the received value is the less or the same
 783    than the one in the field */
 784
 785 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 786 {
 787     int32_t *old = pv;
 788     int32_t new;
 789     qemu_get_sbe32s(f, &new);
 790
 791     if (*old <= new)
 792         return 0;
 793     return -EINVAL;
 794 }
 795
 796 const VMStateInfo vmstate_info_int32_le = {
 797     .name = "int32 equal",
 798     .get  = get_int32_le,
 799     .put  = put_int32,
 800 };
 801
 802 /* 64 bit int */
 803
 804 static int get_int64(QEMUFile *f, void *pv, size_t size)
 805 {
 806     int64_t *v = pv;
 807     qemu_get_sbe64s(f, v);
 808     return 0;
 809 }
 810
 811 static void put_int64(QEMUFile *f, void *pv, size_t size)
 812 {
 813     int64_t *v = pv;
 814     qemu_put_sbe64s(f, v);
 815 }
 816
 817 const VMStateInfo vmstate_info_int64 = {
 818     .name = "int64",
 819     .get  = get_int64,
 820     .put  = put_int64,
 821 };
 822
 823 /* 8 bit unsigned int */
 824
 825 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 826 {
 827     uint8_t *v = pv;
 828     qemu_get_8s(f, v);
 829     return 0;
 830 }
 831
 832 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 833 {
 834     uint8_t *v = pv;
 835     qemu_put_8s(f, v);
 836 }
 837
 838 const VMStateInfo vmstate_info_uint8 = {
 839     .name = "uint8",
 840     .get  = get_uint8,
 841     .put  = put_uint8,
 842 };
 843
 844 /* 16 bit unsigned int */
 845
 846 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 847 {
 848     uint16_t *v = pv;
 849     qemu_get_be16s(f, v);
 850     return 0;
 851 }
 852
 853 static void put_uint16(QEMUFile *f, void *pv, size_t size)
 854 {
 855     uint16_t *v = pv;
 856     qemu_put_be16s(f, v);
 857 }
 858
 859 const VMStateInfo vmstate_info_uint16 = {
 860     .name = "uint16",
 861     .get  = get_uint16,
 862     .put  = put_uint16,
 863 };
 864
 865 /* 32 bit unsigned int */
 866
 867 static int get_uint32(QEMUFile *f, void *pv, size_t size)
 868 {
 869     uint32_t *v = pv;
 870     qemu_get_be32s(f, v);
 871     return 0;
 872 }
 873
 874 static void put_uint32(QEMUFile *f, void *pv, size_t size)
 875 {
 876     uint32_t *v = pv;
 877     qemu_put_be32s(f, v);
 878 }
 879
 880 const VMStateInfo vmstate_info_uint32 = {
 881     .name = "uint32",
 882     .get  = get_uint32,
 883     .put  = put_uint32,
 884 };
 885
 886 /* 64 bit unsigned int */
 887
 888 static int get_uint64(QEMUFile *f, void *pv, size_t size)
 889 {
 890     uint64_t *v = pv;
 891     qemu_get_be64s(f, v);
 892     return 0;
 893 }
 894
 895 static void put_uint64(QEMUFile *f, void *pv, size_t size)
 896 {
 897     uint64_t *v = pv;
 898     qemu_put_be64s(f, v);
 899 }
 900
 901 const VMStateInfo vmstate_info_uint64 = {
 902     .name = "uint64",
 903     .get  = get_uint64,
 904     .put  = put_uint64,
 905 };
 906
 907 /* 8 bit int. See that the received value is the same than the one
 908    in the field */
 909
 910 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
 911 {
 912     uint8_t *v = pv;
 913     uint8_t v2;
 914     qemu_get_8s(f, &v2);
 915
 916     if (*v == v2)
 917         return 0;
 918     return -EINVAL;
 919 }
 920
 921 const VMStateInfo vmstate_info_uint8_equal = {
 922     .name = "uint8 equal",
 923     .get  = get_uint8_equal,
 924     .put  = put_uint8,
 925 };
 926
 927 /* 16 bit unsigned int int. See that the received value is the same than the one
 928    in the field */
 929
 930 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
 931 {
 932     uint16_t *v = pv;
 933     uint16_t v2;
 934     qemu_get_be16s(f, &v2);
 935
 936     if (*v == v2)
 937         return 0;
 938     return -EINVAL;
 939 }
 940
 941 const VMStateInfo vmstate_info_uint16_equal = {
 942     .name = "uint16 equal",
 943     .get  = get_uint16_equal,
 944     .put  = put_uint16,
 945 };
 946
 947 /* timers  */
 948
 949 static int get_timer(QEMUFile *f, void *pv, size_t size)
 950 {
 951     QEMUTimer *v = pv;
 952     qemu_get_timer(f, v);
 953     return 0;
 954 }
 955
 956 static void put_timer(QEMUFile *f, void *pv, size_t size)
 957 {
 958     QEMUTimer *v = pv;
 959     qemu_put_timer(f, v);
 960 }
 961
 962 const VMStateInfo vmstate_info_timer = {
 963     .name = "timer",
 964     .get  = get_timer,
 965     .put  = put_timer,
 966 };
 967
 968 /* uint8_t buffers */
 969
 970 static int get_buffer(QEMUFile *f, void *pv, size_t size)
 971 {
 972     uint8_t *v = pv;
 973     qemu_get_buffer(f, v, size);
 974     return 0;
 975 }
 976
 977 static void put_buffer(QEMUFile *f, void *pv, size_t size)
 978 {
 979     uint8_t *v = pv;
 980     qemu_put_buffer(f, v, size);
 981 }
 982
 983 const VMStateInfo vmstate_info_buffer = {
 984     .name = "buffer",
 985     .get  = get_buffer,
 986     .put  = put_buffer,
 987 };
 988
 989 /* unused buffers: space that was used for some fields that are
 990    not usefull anymore */
 991
 992 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
 993 {
 994     uint8_t buf[1024];
 995     int block_len;
 996
 997     while (size > 0) {
 998         block_len = MIN(sizeof(buf), size);
 999         size -= block_len;
1000         qemu_get_buffer(f, buf, block_len);
1001     }
1002    return 0;
1003 }
1004
1005 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1006 {
1007     static const uint8_t buf[1024];
1008     int block_len;
1009
1010     while (size > 0) {
1011         block_len = MIN(sizeof(buf), size);
1012         size -= block_len;
1013         qemu_put_buffer(f, buf, block_len);
1014     }
1015 }
1016
1017 const VMStateInfo vmstate_info_unused_buffer = {
1018     .name = "unused_buffer",
1019     .get  = get_unused_buffer,
1020     .put  = put_unused_buffer,
1021 };
1022
1023 typedef struct CompatEntry {
1024     char idstr[256];
1025     int instance_id;
1026 } CompatEntry;
1027
1028 typedef struct SaveStateEntry {
1029     QTAILQ_ENTRY(SaveStateEntry) entry;
1030     char idstr[256];
1031     int instance_id;
1032     int alias_id;
1033     int version_id;
1034     int section_id;
1035     SaveSetParamsHandler *set_params;
1036     SaveLiveStateHandler *save_live_state;
1037     SaveStateHandler *save_state;
1038     LoadStateHandler *load_state;
1039     const VMStateDescription *vmsd;
1040     void *opaque;
1041     CompatEntry *compat;
1042     int no_migrate;
1043 } SaveStateEntry;
1044
1045
1046 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1047     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1048 static int global_section_id;
1049
1050 static int calculate_new_instance_id(const char *idstr)
1051 {
1052     SaveStateEntry *se;
1053     int instance_id = 0;
1054
1055     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1056         if (strcmp(idstr, se->idstr) == 0
1057             && instance_id <= se->instance_id) {
1058             instance_id = se->instance_id + 1;
1059         }
1060     }
1061     return instance_id;
1062 }
1063
1064 static int calculate_compat_instance_id(const char *idstr)
1065 {
1066     SaveStateEntry *se;
1067     int instance_id = 0;
1068
1069     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1070         if (!se->compat)
1071             continue;
1072
1073         if (strcmp(idstr, se->compat->idstr) == 0
1074             && instance_id <= se->compat->instance_id) {
1075             instance_id = se->compat->instance_id + 1;
1076         }
1077     }
1078     return instance_id;
1079 }
1080
1081 /* TODO: Individual devices generally have very little idea about the rest
1082    of the system, so instance_id should be removed/replaced.
1083    Meanwhile pass -1 as instance_id if you do not already have a clearly
1084    distinguishing id for all instances of your device class. */
1085 int register_savevm_live(DeviceState *dev,
1086                          const char *idstr,
1087                          int instance_id,
1088                          int version_id,
1089                          SaveSetParamsHandler *set_params,
1090                          SaveLiveStateHandler *save_live_state,
1091                          SaveStateHandler *save_state,
1092                          LoadStateHandler *load_state,
1093                          void *opaque)
1094 {
1095     SaveStateEntry *se;
1096
1097     se = qemu_mallocz(sizeof(SaveStateEntry));
1098     se->version_id = version_id;
1099     se->section_id = global_section_id++;
1100     se->set_params = set_params;
1101     se->save_live_state = save_live_state;
1102     se->save_state = save_state;
1103     se->load_state = load_state;
1104     se->opaque = opaque;
1105     se->vmsd = NULL;
1106     se->no_migrate = 0;
1107
1108     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1109         char *id = dev->parent_bus->info->get_dev_path(dev);
1110         if (id) {
1111             pstrcpy(se->idstr, sizeof(se->idstr), id);
1112             pstrcat(se->idstr, sizeof(se->idstr), "/");
1113             qemu_free(id);
1114
1115             se->compat = qemu_mallocz(sizeof(CompatEntry));
1116             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1117             se->compat->instance_id = instance_id == -1 ?
1118                          calculate_compat_instance_id(idstr) : instance_id;
1119             instance_id = -1;
1120         }
1121     }
1122     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1123
1124     if (instance_id == -1) {
1125         se->instance_id = calculate_new_instance_id(se->idstr);
1126     } else {
1127         se->instance_id = instance_id;
1128     }
1129     assert(!se->compat || se->instance_id == 0);
1130     /* add at the end of list */
1131     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1132     return 0;
1133 }
1134
1135 int register_savevm(DeviceState *dev,
1136                     const char *idstr,
1137                     int instance_id,
1138                     int version_id,
1139                     SaveStateHandler *save_state,
1140                     LoadStateHandler *load_state,
1141                     void *opaque)
1142 {
1143     return register_savevm_live(dev, idstr, instance_id, version_id,
1144                                 NULL, NULL, save_state, load_state, opaque);
1145 }
1146
1147 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1148 {
1149     SaveStateEntry *se, *new_se;
1150     char id[256] = "";
1151
1152     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1153         char *path = dev->parent_bus->info->get_dev_path(dev);
1154         if (path) {
1155             pstrcpy(id, sizeof(id), path);
1156             pstrcat(id, sizeof(id), "/");
1157             qemu_free(path);
1158         }
1159     }
1160     pstrcat(id, sizeof(id), idstr);
1161
1162     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1163         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1164             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1165             if (se->compat) {
1166                 qemu_free(se->compat);
1167             }
1168             qemu_free(se);
1169         }
1170     }
1171 }
1172
1173 /* mark a device as not to be migrated, that is the device should be
1174    unplugged before migration */
1175 void register_device_unmigratable(DeviceState *dev, const char *idstr,
1176                                                             void *opaque)
1177 {
1178     SaveStateEntry *se;
1179     char id[256] = "";
1180
1181     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1182         char *path = dev->parent_bus->info->get_dev_path(dev);
1183         if (path) {
1184             pstrcpy(id, sizeof(id), path);
1185             pstrcat(id, sizeof(id), "/");
1186             qemu_free(path);
1187         }
1188     }
1189     pstrcat(id, sizeof(id), idstr);
1190
1191     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1192         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1193             se->no_migrate = 1;
1194         }
1195     }
1196 }
1197
1198 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1199                                    const VMStateDescription *vmsd,
1200                                    void *opaque, int alias_id,
1201                                    int required_for_version)
1202 {
1203     SaveStateEntry *se;
1204
1205     /* If this triggers, alias support can be dropped for the vmsd. */
1206     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1207
1208     se = qemu_mallocz(sizeof(SaveStateEntry));
1209     se->version_id = vmsd->version_id;
1210     se->section_id = global_section_id++;
1211     se->save_live_state = NULL;
1212     se->save_state = NULL;
1213     se->load_state = NULL;
1214     se->opaque = opaque;
1215     se->vmsd = vmsd;
1216     se->alias_id = alias_id;
1217
1218     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1219         char *id = dev->parent_bus->info->get_dev_path(dev);
1220         if (id) {
1221             pstrcpy(se->idstr, sizeof(se->idstr), id);
1222             pstrcat(se->idstr, sizeof(se->idstr), "/");
1223             qemu_free(id);
1224
1225             se->compat = qemu_mallocz(sizeof(CompatEntry));
1226             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1227             se->compat->instance_id = instance_id == -1 ?
1228                          calculate_compat_instance_id(vmsd->name) : instance_id;
1229             instance_id = -1;
1230         }
1231     }
1232     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1233
1234     if (instance_id == -1) {
1235         se->instance_id = calculate_new_instance_id(se->idstr);
1236     } else {
1237         se->instance_id = instance_id;
1238     }
1239     assert(!se->compat || se->instance_id == 0);
1240     /* add at the end of list */
1241     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1242     return 0;
1243 }
1244
1245 int vmstate_register(DeviceState *dev, int instance_id,
1246                      const VMStateDescription *vmsd, void *opaque)
1247 {
1248     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1249                                           opaque, -1, 0);
1250 }
1251
1252 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1253                         void *opaque)
1254 {
1255     SaveStateEntry *se, *new_se;
1256
1257     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1258         if (se->vmsd == vmsd && se->opaque == opaque) {
1259             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1260             if (se->compat) {
1261                 qemu_free(se->compat);
1262             }
1263             qemu_free(se);
1264         }
1265     }
1266 }
1267
1268 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1269                                     void *opaque);
1270 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1271                                    void *opaque);
1272
1273 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1274                        void *opaque, int version_id)
1275 {
1276     VMStateField *field = vmsd->fields;
1277     int ret;
1278
1279     if (version_id > vmsd->version_id) {
1280         return -EINVAL;
1281     }
1282     if (version_id < vmsd->minimum_version_id_old) {
1283         return -EINVAL;
1284     }
1285     if  (version_id < vmsd->minimum_version_id) {
1286         return vmsd->load_state_old(f, opaque, version_id);
1287     }
1288     if (vmsd->pre_load) {
1289         int ret = vmsd->pre_load(opaque);
1290         if (ret)
1291             return ret;
1292     }
1293     while(field->name) {
1294         if ((field->field_exists &&
1295              field->field_exists(opaque, version_id)) ||
1296             (!field->field_exists &&
1297              field->version_id <= version_id)) {
1298             void *base_addr = opaque + field->offset;
1299             int i, n_elems = 1;
1300             int size = field->size;
1301
1302             if (field->flags & VMS_VBUFFER) {
1303                 size = *(int32_t *)(opaque+field->size_offset);
1304                 if (field->flags & VMS_MULTIPLY) {
1305                     size *= field->size;
1306                 }
1307             }
1308             if (field->flags & VMS_ARRAY) {
1309                 n_elems = field->num;
1310             } else if (field->flags & VMS_VARRAY_INT32) {
1311                 n_elems = *(int32_t *)(opaque+field->num_offset);
1312             } else if (field->flags & VMS_VARRAY_UINT16) {
1313                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1314             }
1315             if (field->flags & VMS_POINTER) {
1316                 base_addr = *(void **)base_addr + field->start;
1317             }
1318             for (i = 0; i < n_elems; i++) {
1319                 void *addr = base_addr + size * i;
1320
1321                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1322                     addr = *(void **)addr;
1323                 }
1324                 if (field->flags & VMS_STRUCT) {
1325                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1326                 } else {
1327                     ret = field->info->get(f, addr, size);
1328
1329                 }
1330                 if (ret < 0) {
1331                     return ret;
1332                 }
1333             }
1334         }
1335         field++;
1336     }
1337     ret = vmstate_subsection_load(f, vmsd, opaque);
1338     if (ret != 0) {
1339         return ret;
1340     }
1341     if (vmsd->post_load) {
1342         return vmsd->post_load(opaque, version_id);
1343     }
1344     return 0;
1345 }
1346
1347 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1348                         void *opaque)
1349 {
1350     VMStateField *field = vmsd->fields;
1351
1352     if (vmsd->pre_save) {
1353         vmsd->pre_save(opaque);
1354     }
1355     while(field->name) {
1356         if (!field->field_exists ||
1357             field->field_exists(opaque, vmsd->version_id)) {
1358             void *base_addr = opaque + field->offset;
1359             int i, n_elems = 1;
1360             int size = field->size;
1361
1362             if (field->flags & VMS_VBUFFER) {
1363                 size = *(int32_t *)(opaque+field->size_offset);
1364                 if (field->flags & VMS_MULTIPLY) {
1365                     size *= field->size;
1366                 }
1367             }
1368             if (field->flags & VMS_ARRAY) {
1369                 n_elems = field->num;
1370             } else if (field->flags & VMS_VARRAY_INT32) {
1371                 n_elems = *(int32_t *)(opaque+field->num_offset);
1372             } else if (field->flags & VMS_VARRAY_UINT16) {
1373                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1374             }
1375             if (field->flags & VMS_POINTER) {
1376                 base_addr = *(void **)base_addr + field->start;
1377             }
1378             for (i = 0; i < n_elems; i++) {
1379                 void *addr = base_addr + size * i;
1380
1381                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1382                     addr = *(void **)addr;
1383                 }
1384                 if (field->flags & VMS_STRUCT) {
1385                     vmstate_save_state(f, field->vmsd, addr);
1386                 } else {
1387                     field->info->put(f, addr, size);
1388                 }
1389             }
1390         }
1391         field++;
1392     }
1393     vmstate_subsection_save(f, vmsd, opaque);
1394 }
1395
1396 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1397 {
1398     if (!se->vmsd) {         /* Old style */
1399         return se->load_state(f, se->opaque, version_id);
1400     }
1401     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1402 }
1403
1404 static int vmstate_save(QEMUFile *f, SaveStateEntry *se)
1405 {
1406     if (se->no_migrate) {
1407         return -1;
1408     }
1409
1410     if (!se->vmsd) {         /* Old style */
1411         se->save_state(f, se->opaque);
1412         return 0;
1413     }
1414     vmstate_save_state(f,se->vmsd, se->opaque);
1415
1416     return 0;
1417 }
1418
1419 #define QEMU_VM_FILE_MAGIC           0x5145564d
1420 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1421 #define QEMU_VM_FILE_VERSION         0x00000003
1422
1423 #define QEMU_VM_EOF                  0x00
1424 #define QEMU_VM_SECTION_START        0x01
1425 #define QEMU_VM_SECTION_PART         0x02
1426 #define QEMU_VM_SECTION_END          0x03
1427 #define QEMU_VM_SECTION_FULL         0x04
1428 #define QEMU_VM_SUBSECTION           0x05
1429
1430 int qemu_savevm_state_begin(Monitor *mon, QEMUFile *f, int blk_enable,
1431                             int shared)
1432 {
1433     SaveStateEntry *se;
1434
1435     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1436         if(se->set_params == NULL) {
1437             continue;
1438         }
1439         se->set_params(blk_enable, shared, se->opaque);
1440     }
1441
1442     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1443     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1444
1445     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1446         int len;
1447
1448         if (se->save_live_state == NULL)
1449             continue;
1450
1451         /* Section type */
1452         qemu_put_byte(f, QEMU_VM_SECTION_START);
1453         qemu_put_be32(f, se->section_id);
1454
1455         /* ID string */
1456         len = strlen(se->idstr);
1457         qemu_put_byte(f, len);
1458         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1459
1460         qemu_put_be32(f, se->instance_id);
1461         qemu_put_be32(f, se->version_id);
1462
1463         se->save_live_state(mon, f, QEMU_VM_SECTION_START, se->opaque);
1464     }
1465
1466     if (qemu_file_has_error(f)) {
1467         qemu_savevm_state_cancel(mon, f);
1468         return -EIO;
1469     }
1470
1471     return 0;
1472 }
1473
1474 int qemu_savevm_state_iterate(Monitor *mon, QEMUFile *f)
1475 {
1476     SaveStateEntry *se;
1477     int ret = 1;
1478
1479     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1480         if (se->save_live_state == NULL)
1481             continue;
1482
1483         /* Section type */
1484         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1485         qemu_put_be32(f, se->section_id);
1486
1487         ret = se->save_live_state(mon, f, QEMU_VM_SECTION_PART, se->opaque);
1488         if (!ret) {
1489             /* Do not proceed to the next vmstate before this one reported
1490                completion of the current stage. This serializes the migration
1491                and reduces the probability that a faster changing state is
1492                synchronized over and over again. */
1493             break;
1494         }
1495     }
1496
1497     if (ret)
1498         return 1;
1499
1500     if (qemu_file_has_error(f)) {
1501         qemu_savevm_state_cancel(mon, f);
1502         return -EIO;
1503     }
1504
1505     return 0;
1506 }
1507
1508 int qemu_savevm_state_complete(Monitor *mon, QEMUFile *f)
1509 {
1510     SaveStateEntry *se;
1511     int r;
1512
1513     cpu_synchronize_all_states();
1514
1515     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1516         if (se->save_live_state == NULL)
1517             continue;
1518
1519         /* Section type */
1520         qemu_put_byte(f, QEMU_VM_SECTION_END);
1521         qemu_put_be32(f, se->section_id);
1522
1523         se->save_live_state(mon, f, QEMU_VM_SECTION_END, se->opaque);
1524     }
1525
1526     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1527         int len;
1528
1529         if (se->save_state == NULL && se->vmsd == NULL)
1530             continue;
1531
1532         /* Section type */
1533         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1534         qemu_put_be32(f, se->section_id);
1535
1536         /* ID string */
1537         len = strlen(se->idstr);
1538         qemu_put_byte(f, len);
1539         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1540
1541         qemu_put_be32(f, se->instance_id);
1542         qemu_put_be32(f, se->version_id);
1543
1544         r = vmstate_save(f, se);
1545         if (r < 0) {
1546             monitor_printf(mon, "cannot migrate with device '%s'\n", se->idstr);
1547             return r;
1548         }
1549     }
1550
1551     qemu_put_byte(f, QEMU_VM_EOF);
1552
1553     if (qemu_file_has_error(f))
1554         return -EIO;
1555
1556     return 0;
1557 }
1558
1559 void qemu_savevm_state_cancel(Monitor *mon, QEMUFile *f)
1560 {
1561     SaveStateEntry *se;
1562
1563     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1564         if (se->save_live_state) {
1565             se->save_live_state(mon, f, -1, se->opaque);
1566         }
1567     }
1568 }
1569
1570 static int qemu_savevm_state(Monitor *mon, QEMUFile *f)
1571 {
1572     int saved_vm_running;
1573     int ret;
1574
1575     saved_vm_running = vm_running;
1576     vm_stop(0);
1577
1578     bdrv_flush_all();
1579
1580     ret = qemu_savevm_state_begin(mon, f, 0, 0);
1581     if (ret < 0)
1582         goto out;
1583
1584     do {
1585         ret = qemu_savevm_state_iterate(mon, f);
1586         if (ret < 0)
1587             goto out;
1588     } while (ret == 0);
1589
1590     ret = qemu_savevm_state_complete(mon, f);
1591
1592 out:
1593     if (qemu_file_has_error(f))
1594         ret = -EIO;
1595
1596     if (!ret && saved_vm_running)
1597         vm_start();
1598
1599     return ret;
1600 }
1601
1602 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1603 {
1604     SaveStateEntry *se;
1605
1606     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1607         if (!strcmp(se->idstr, idstr) &&
1608             (instance_id == se->instance_id ||
1609              instance_id == se->alias_id))
1610             return se;
1611         /* Migrating from an older version? */
1612         if (strstr(se->idstr, idstr) && se->compat) {
1613             if (!strcmp(se->compat->idstr, idstr) &&
1614                 (instance_id == se->compat->instance_id ||
1615                  instance_id == se->alias_id))
1616                 return se;
1617         }
1618     }
1619     return NULL;
1620 }
1621
1622 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1623 {
1624     while(sub && sub->needed) {
1625         if (strcmp(idstr, sub->vmsd->name) == 0) {
1626             return sub->vmsd;
1627         }
1628         sub++;
1629     }
1630     return NULL;
1631 }
1632
1633 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1634                                    void *opaque)
1635 {
1636     while (qemu_peek_byte(f) == QEMU_VM_SUBSECTION) {
1637         char idstr[256];
1638         int ret;
1639         uint8_t version_id, len;
1640         const VMStateDescription *sub_vmsd;
1641
1642         qemu_get_byte(f); /* subsection */
1643         len = qemu_get_byte(f);
1644         qemu_get_buffer(f, (uint8_t *)idstr, len);
1645         idstr[len] = 0;
1646         version_id = qemu_get_be32(f);
1647
1648         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
1649         if (sub_vmsd == NULL) {
1650             return -ENOENT;
1651         }
1652         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1653         if (ret) {
1654             return ret;
1655         }
1656     }
1657     return 0;
1658 }
1659
1660 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1661                                     void *opaque)
1662 {
1663     const VMStateSubsection *sub = vmsd->subsections;
1664
1665     while (sub && sub->needed) {
1666         if (sub->needed(opaque)) {
1667             const VMStateDescription *vmsd = sub->vmsd;
1668             uint8_t len;
1669
1670             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1671             len = strlen(vmsd->name);
1672             qemu_put_byte(f, len);
1673             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1674             qemu_put_be32(f, vmsd->version_id);
1675             vmstate_save_state(f, vmsd, opaque);
1676         }
1677         sub++;
1678     }
1679 }
1680
1681 typedef struct LoadStateEntry {
1682     QLIST_ENTRY(LoadStateEntry) entry;
1683     SaveStateEntry *se;
1684     int section_id;
1685     int version_id;
1686 } LoadStateEntry;
1687
1688 int qemu_loadvm_state(QEMUFile *f)
1689 {
1690     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1691         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1692     LoadStateEntry *le, *new_le;
1693     uint8_t section_type;
1694     unsigned int v;
1695     int ret;
1696
1697     v = qemu_get_be32(f);
1698     if (v != QEMU_VM_FILE_MAGIC)
1699         return -EINVAL;
1700
1701     v = qemu_get_be32(f);
1702     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1703         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1704         return -ENOTSUP;
1705     }
1706     if (v != QEMU_VM_FILE_VERSION)
1707         return -ENOTSUP;
1708
1709     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1710         uint32_t instance_id, version_id, section_id;
1711         SaveStateEntry *se;
1712         char idstr[257];
1713         int len;
1714
1715         switch (section_type) {
1716         case QEMU_VM_SECTION_START:
1717         case QEMU_VM_SECTION_FULL:
1718             /* Read section start */
1719             section_id = qemu_get_be32(f);
1720             len = qemu_get_byte(f);
1721             qemu_get_buffer(f, (uint8_t *)idstr, len);
1722             idstr[len] = 0;
1723             instance_id = qemu_get_be32(f);
1724             version_id = qemu_get_be32(f);
1725
1726             /* Find savevm section */
1727             se = find_se(idstr, instance_id);
1728             if (se == NULL) {
1729                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1730                 ret = -EINVAL;
1731                 goto out;
1732             }
1733
1734             /* Validate version */
1735             if (version_id > se->version_id) {
1736                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
1737                         version_id, idstr, se->version_id);
1738                 ret = -EINVAL;
1739                 goto out;
1740             }
1741
1742             /* Add entry */
1743             le = qemu_mallocz(sizeof(*le));
1744
1745             le->se = se;
1746             le->section_id = section_id;
1747             le->version_id = version_id;
1748             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
1749
1750             ret = vmstate_load(f, le->se, le->version_id);
1751             if (ret < 0) {
1752                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
1753                         instance_id, idstr);
1754                 goto out;
1755             }
1756             break;
1757         case QEMU_VM_SECTION_PART:
1758         case QEMU_VM_SECTION_END:
1759             section_id = qemu_get_be32(f);
1760
1761             QLIST_FOREACH(le, &loadvm_handlers, entry) {
1762                 if (le->section_id == section_id) {
1763                     break;
1764                 }
1765             }
1766             if (le == NULL) {
1767                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
1768                 ret = -EINVAL;
1769                 goto out;
1770             }
1771
1772             ret = vmstate_load(f, le->se, le->version_id);
1773             if (ret < 0) {
1774                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
1775                         section_id);
1776                 goto out;
1777             }
1778             break;
1779         default:
1780             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
1781             ret = -EINVAL;
1782             goto out;
1783         }
1784     }
1785
1786     cpu_synchronize_all_post_init();
1787
1788     ret = 0;
1789
1790 out:
1791     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
1792         QLIST_REMOVE(le, entry);
1793         qemu_free(le);
1794     }
1795
1796     if (qemu_file_has_error(f))
1797         ret = -EIO;
1798
1799     return ret;
1800 }
1801
1802 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1803                               const char *name)
1804 {
1805     QEMUSnapshotInfo *sn_tab, *sn;
1806     int nb_sns, i, ret;
1807
1808     ret = -ENOENT;
1809     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1810     if (nb_sns < 0)
1811         return ret;
1812     for(i = 0; i < nb_sns; i++) {
1813         sn = &sn_tab[i];
1814         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1815             *sn_info = *sn;
1816             ret = 0;
1817             break;
1818         }
1819     }
1820     qemu_free(sn_tab);
1821     return ret;
1822 }
1823
1824 /*
1825  * Deletes snapshots of a given name in all opened images.
1826  */
1827 static int del_existing_snapshots(Monitor *mon, const char *name)
1828 {
1829     BlockDriverState *bs;
1830     QEMUSnapshotInfo sn1, *snapshot = &sn1;
1831     int ret;
1832
1833     bs = NULL;
1834     while ((bs = bdrv_next(bs))) {
1835         if (bdrv_can_snapshot(bs) &&
1836             bdrv_snapshot_find(bs, snapshot, name) >= 0)
1837         {
1838             ret = bdrv_snapshot_delete(bs, name);
1839             if (ret < 0) {
1840                 monitor_printf(mon,
1841                                "Error while deleting snapshot on '%s'\n",
1842                                bdrv_get_device_name(bs));
1843                 return -1;
1844             }
1845         }
1846     }
1847
1848     return 0;
1849 }
1850
1851 void do_savevm(Monitor *mon, const QDict *qdict)
1852 {
1853     BlockDriverState *bs, *bs1;
1854     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1855     int ret;
1856     QEMUFile *f;
1857     int saved_vm_running;
1858     uint32_t vm_state_size;
1859 #ifdef _WIN32
1860     struct _timeb tb;
1861     struct tm *ptm;
1862 #else
1863     struct timeval tv;
1864     struct tm tm;
1865 #endif
1866     const char *name = qdict_get_try_str(qdict, "name");
1867
1868     /* Verify if there is a device that doesn't support snapshots and is writable */
1869     bs = NULL;
1870     while ((bs = bdrv_next(bs))) {
1871
1872         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1873             continue;
1874         }
1875
1876         if (!bdrv_can_snapshot(bs)) {
1877             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
1878                                bdrv_get_device_name(bs));
1879             return;
1880         }
1881     }
1882
1883     bs = bdrv_snapshots();
1884     if (!bs) {
1885         monitor_printf(mon, "No block device can accept snapshots\n");
1886         return;
1887     }
1888     /* ??? Should this occur after vm_stop?  */
1889     qemu_aio_flush();
1890
1891     saved_vm_running = vm_running;
1892     vm_stop(0);
1893
1894     memset(sn, 0, sizeof(*sn));
1895
1896     /* fill auxiliary fields */
1897 #ifdef _WIN32
1898     _ftime(&tb);
1899     sn->date_sec = tb.time;
1900     sn->date_nsec = tb.millitm * 1000000;
1901 #else
1902     gettimeofday(&tv, NULL);
1903     sn->date_sec = tv.tv_sec;
1904     sn->date_nsec = tv.tv_usec * 1000;
1905 #endif
1906     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1907
1908     if (name) {
1909         ret = bdrv_snapshot_find(bs, old_sn, name);
1910         if (ret >= 0) {
1911             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1912             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1913         } else {
1914             pstrcpy(sn->name, sizeof(sn->name), name);
1915         }
1916     } else {
1917 #ifdef _WIN32
1918         ptm = localtime(&tb.time);
1919         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", ptm);
1920 #else
1921         /* cast below needed for OpenBSD where tv_sec is still 'long' */
1922         localtime_r((const time_t *)&tv.tv_sec, &tm);
1923         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
1924 #endif
1925     }
1926
1927     /* Delete old snapshots of the same name */
1928     if (name && del_existing_snapshots(mon, name) < 0) {
1929         goto the_end;
1930     }
1931
1932     /* save the VM state */
1933     f = qemu_fopen_bdrv(bs, 1);
1934     if (!f) {
1935         monitor_printf(mon, "Could not open VM state file\n");
1936         goto the_end;
1937     }
1938     ret = qemu_savevm_state(mon, f);
1939     vm_state_size = qemu_ftell(f);
1940     qemu_fclose(f);
1941     if (ret < 0) {
1942         monitor_printf(mon, "Error %d while writing VM\n", ret);
1943         goto the_end;
1944     }
1945
1946     /* create the snapshots */
1947
1948     bs1 = NULL;
1949     while ((bs1 = bdrv_next(bs1))) {
1950         if (bdrv_can_snapshot(bs1)) {
1951             /* Write VM state size only to the image that contains the state */
1952             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1953             ret = bdrv_snapshot_create(bs1, sn);
1954             if (ret < 0) {
1955                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1956                                bdrv_get_device_name(bs1));
1957             }
1958         }
1959     }
1960
1961  the_end:
1962     if (saved_vm_running)
1963         vm_start();
1964 }
1965
1966 int load_vmstate(const char *name)
1967 {
1968     BlockDriverState *bs, *bs_vm_state;
1969     QEMUSnapshotInfo sn;
1970     QEMUFile *f;
1971     int ret;
1972
1973     bs_vm_state = bdrv_snapshots();
1974     if (!bs_vm_state) {
1975         error_report("No block device supports snapshots");
1976         return -ENOTSUP;
1977     }
1978
1979     /* Don't even try to load empty VM states */
1980     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
1981     if (ret < 0) {
1982         return ret;
1983     } else if (sn.vm_state_size == 0) {
1984         return -EINVAL;
1985     }
1986
1987     /* Verify if there is any device that doesn't support snapshots and is
1988     writable and check if the requested snapshot is available too. */
1989     bs = NULL;
1990     while ((bs = bdrv_next(bs))) {
1991
1992         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1993             continue;
1994         }
1995
1996         if (!bdrv_can_snapshot(bs)) {
1997             error_report("Device '%s' is writable but does not support snapshots.",
1998                                bdrv_get_device_name(bs));
1999             return -ENOTSUP;
2000         }
2001
2002         ret = bdrv_snapshot_find(bs, &sn, name);
2003         if (ret < 0) {
2004             error_report("Device '%s' does not have the requested snapshot '%s'",
2005                            bdrv_get_device_name(bs), name);
2006             return ret;
2007         }
2008     }
2009
2010     /* Flush all IO requests so they don't interfere with the new state.  */
2011     qemu_aio_flush();
2012
2013     bs = NULL;
2014     while ((bs = bdrv_next(bs))) {
2015         if (bdrv_can_snapshot(bs)) {
2016             ret = bdrv_snapshot_goto(bs, name);
2017             if (ret < 0) {
2018                 error_report("Error %d while activating snapshot '%s' on '%s'",
2019                              ret, name, bdrv_get_device_name(bs));
2020                 return ret;
2021             }
2022         }
2023     }
2024
2025     /* restore the VM state */
2026     f = qemu_fopen_bdrv(bs_vm_state, 0);
2027     if (!f) {
2028         error_report("Could not open VM state file");
2029         return -EINVAL;
2030     }
2031
2032     ret = qemu_loadvm_state(f);
2033
2034     qemu_fclose(f);
2035     if (ret < 0) {
2036         error_report("Error %d while loading VM state", ret);
2037         return ret;
2038     }
2039
2040     return 0;
2041 }
2042
2043 void do_delvm(Monitor *mon, const QDict *qdict)
2044 {
2045     BlockDriverState *bs, *bs1;
2046     int ret;
2047     const char *name = qdict_get_str(qdict, "name");
2048
2049     bs = bdrv_snapshots();
2050     if (!bs) {
2051         monitor_printf(mon, "No block device supports snapshots\n");
2052         return;
2053     }
2054
2055     bs1 = NULL;
2056     while ((bs1 = bdrv_next(bs1))) {
2057         if (bdrv_can_snapshot(bs1)) {
2058             ret = bdrv_snapshot_delete(bs1, name);
2059             if (ret < 0) {
2060                 if (ret == -ENOTSUP)
2061                     monitor_printf(mon,
2062                                    "Snapshots not supported on device '%s'\n",
2063                                    bdrv_get_device_name(bs1));
2064                 else
2065                     monitor_printf(mon, "Error %d while deleting snapshot on "
2066                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2067             }
2068         }
2069     }
2070 }
2071
2072 void do_info_snapshots(Monitor *mon)
2073 {
2074     BlockDriverState *bs, *bs1;
2075     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2076     int nb_sns, i, ret, available;
2077     int total;
2078     int *available_snapshots;
2079     char buf[256];
2080
2081     bs = bdrv_snapshots();
2082     if (!bs) {
2083         monitor_printf(mon, "No available block device supports snapshots\n");
2084         return;
2085     }
2086
2087     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2088     if (nb_sns < 0) {
2089         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2090         return;
2091     }
2092
2093     if (nb_sns == 0) {
2094         monitor_printf(mon, "There is no snapshot available.\n");
2095         return;
2096     }
2097
2098     available_snapshots = qemu_mallocz(sizeof(int) * nb_sns);
2099     total = 0;
2100     for (i = 0; i < nb_sns; i++) {
2101         sn = &sn_tab[i];
2102         available = 1;
2103         bs1 = NULL;
2104
2105         while ((bs1 = bdrv_next(bs1))) {
2106             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2107                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2108                 if (ret < 0) {
2109                     available = 0;
2110                     break;
2111                 }
2112             }
2113         }
2114
2115         if (available) {
2116             available_snapshots[total] = i;
2117             total++;
2118         }
2119     }
2120
2121     if (total > 0) {
2122         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2123         for (i = 0; i < total; i++) {
2124             sn = &sn_tab[available_snapshots[i]];
2125             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2126         }
2127     } else {
2128         monitor_printf(mon, "There is no suitable snapshot available\n");
2129     }
2130
2131     qemu_free(sn_tab);
2132     qemu_free(available_snapshots);
2133
2134 }