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 /* 64 bit linux kernel unsigned int */
 908
 909 #ifdef __linux__
 910 static int get_u64(QEMUFile *f, void *pv, size_t size)
 911 {
 912     uint64_t *v = pv;
 913     qemu_get_be64s(f, v);
 914     return 0;
 915 }
 916
 917 static void put_u64(QEMUFile *f, void *pv, size_t size)
 918 {
 919     uint64_t *v = pv;
 920     qemu_put_be64s(f, v);
 921 }
 922
 923 const VMStateInfo vmstate_info_u64 = {
 924     .name = "__u64",
 925     .get  = get_u64,
 926     .put  = put_u64,
 927 };
 928 #endif /* __linux__ */
 929
 930 /* 8 bit int. See that the received value is the same than the one
 931    in the field */
 932
 933 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
 934 {
 935     uint8_t *v = pv;
 936     uint8_t v2;
 937     qemu_get_8s(f, &v2);
 938
 939     if (*v == v2)
 940         return 0;
 941     return -EINVAL;
 942 }
 943
 944 const VMStateInfo vmstate_info_uint8_equal = {
 945     .name = "uint8 equal",
 946     .get  = get_uint8_equal,
 947     .put  = put_uint8,
 948 };
 949
 950 /* 16 bit unsigned int int. See that the received value is the same than the one
 951    in the field */
 952
 953 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
 954 {
 955     uint16_t *v = pv;
 956     uint16_t v2;
 957     qemu_get_be16s(f, &v2);
 958
 959     if (*v == v2)
 960         return 0;
 961     return -EINVAL;
 962 }
 963
 964 const VMStateInfo vmstate_info_uint16_equal = {
 965     .name = "uint16 equal",
 966     .get  = get_uint16_equal,
 967     .put  = put_uint16,
 968 };
 969
 970 /* timers  */
 971
 972 static int get_timer(QEMUFile *f, void *pv, size_t size)
 973 {
 974     QEMUTimer *v = pv;
 975     qemu_get_timer(f, v);
 976     return 0;
 977 }
 978
 979 static void put_timer(QEMUFile *f, void *pv, size_t size)
 980 {
 981     QEMUTimer *v = pv;
 982     qemu_put_timer(f, v);
 983 }
 984
 985 const VMStateInfo vmstate_info_timer = {
 986     .name = "timer",
 987     .get  = get_timer,
 988     .put  = put_timer,
 989 };
 990
 991 /* uint8_t buffers */
 992
 993 static int get_buffer(QEMUFile *f, void *pv, size_t size)
 994 {
 995     uint8_t *v = pv;
 996     qemu_get_buffer(f, v, size);
 997     return 0;
 998 }
 999
1000 static void put_buffer(QEMUFile *f, void *pv, size_t size)
1001 {
1002     uint8_t *v = pv;
1003     qemu_put_buffer(f, v, size);
1004 }
1005
1006 const VMStateInfo vmstate_info_buffer = {
1007     .name = "buffer",
1008     .get  = get_buffer,
1009     .put  = put_buffer,
1010 };
1011
1012 /* unused buffers: space that was used for some fields that are
1013    not usefull anymore */
1014
1015 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
1016 {
1017     uint8_t buf[1024];
1018     int block_len;
1019
1020     while (size > 0) {
1021         block_len = MIN(sizeof(buf), size);
1022         size -= block_len;
1023         qemu_get_buffer(f, buf, block_len);
1024     }
1025    return 0;
1026 }
1027
1028 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1029 {
1030     static const uint8_t buf[1024];
1031     int block_len;
1032
1033     while (size > 0) {
1034         block_len = MIN(sizeof(buf), size);
1035         size -= block_len;
1036         qemu_put_buffer(f, buf, block_len);
1037     }
1038 }
1039
1040 const VMStateInfo vmstate_info_unused_buffer = {
1041     .name = "unused_buffer",
1042     .get  = get_unused_buffer,
1043     .put  = put_unused_buffer,
1044 };
1045
1046 typedef struct CompatEntry {
1047     char idstr[256];
1048     int instance_id;
1049 } CompatEntry;
1050
1051 typedef struct SaveStateEntry {
1052     QTAILQ_ENTRY(SaveStateEntry) entry;
1053     char idstr[256];
1054     int instance_id;
1055     int alias_id;
1056     int version_id;
1057     int section_id;
1058     SaveSetParamsHandler *set_params;
1059     SaveLiveStateHandler *save_live_state;
1060     SaveStateHandler *save_state;
1061     LoadStateHandler *load_state;
1062     const VMStateDescription *vmsd;
1063     void *opaque;
1064     CompatEntry *compat;
1065     int no_migrate;
1066 } SaveStateEntry;
1067
1068
1069 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1070     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1071 static int global_section_id;
1072
1073 static int calculate_new_instance_id(const char *idstr)
1074 {
1075     SaveStateEntry *se;
1076     int instance_id = 0;
1077
1078     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1079         if (strcmp(idstr, se->idstr) == 0
1080             && instance_id <= se->instance_id) {
1081             instance_id = se->instance_id + 1;
1082         }
1083     }
1084     return instance_id;
1085 }
1086
1087 static int calculate_compat_instance_id(const char *idstr)
1088 {
1089     SaveStateEntry *se;
1090     int instance_id = 0;
1091
1092     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1093         if (!se->compat)
1094             continue;
1095
1096         if (strcmp(idstr, se->compat->idstr) == 0
1097             && instance_id <= se->compat->instance_id) {
1098             instance_id = se->compat->instance_id + 1;
1099         }
1100     }
1101     return instance_id;
1102 }
1103
1104 /* TODO: Individual devices generally have very little idea about the rest
1105    of the system, so instance_id should be removed/replaced.
1106    Meanwhile pass -1 as instance_id if you do not already have a clearly
1107    distinguishing id for all instances of your device class. */
1108 int register_savevm_live(DeviceState *dev,
1109                          const char *idstr,
1110                          int instance_id,
1111                          int version_id,
1112                          SaveSetParamsHandler *set_params,
1113                          SaveLiveStateHandler *save_live_state,
1114                          SaveStateHandler *save_state,
1115                          LoadStateHandler *load_state,
1116                          void *opaque)
1117 {
1118     SaveStateEntry *se;
1119
1120     se = qemu_mallocz(sizeof(SaveStateEntry));
1121     se->version_id = version_id;
1122     se->section_id = global_section_id++;
1123     se->set_params = set_params;
1124     se->save_live_state = save_live_state;
1125     se->save_state = save_state;
1126     se->load_state = load_state;
1127     se->opaque = opaque;
1128     se->vmsd = NULL;
1129     se->no_migrate = 0;
1130
1131     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1132         char *id = dev->parent_bus->info->get_dev_path(dev);
1133         if (id) {
1134             pstrcpy(se->idstr, sizeof(se->idstr), id);
1135             pstrcat(se->idstr, sizeof(se->idstr), "/");
1136             qemu_free(id);
1137
1138             se->compat = qemu_mallocz(sizeof(CompatEntry));
1139             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1140             se->compat->instance_id = instance_id == -1 ?
1141                          calculate_compat_instance_id(idstr) : instance_id;
1142             instance_id = -1;
1143         }
1144     }
1145     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1146
1147     if (instance_id == -1) {
1148         se->instance_id = calculate_new_instance_id(se->idstr);
1149     } else {
1150         se->instance_id = instance_id;
1151     }
1152     assert(!se->compat || se->instance_id == 0);
1153     /* add at the end of list */
1154     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1155     return 0;
1156 }
1157
1158 int register_savevm(DeviceState *dev,
1159                     const char *idstr,
1160                     int instance_id,
1161                     int version_id,
1162                     SaveStateHandler *save_state,
1163                     LoadStateHandler *load_state,
1164                     void *opaque)
1165 {
1166     return register_savevm_live(dev, idstr, instance_id, version_id,
1167                                 NULL, NULL, save_state, load_state, opaque);
1168 }
1169
1170 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1171 {
1172     SaveStateEntry *se, *new_se;
1173     char id[256] = "";
1174
1175     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1176         char *path = dev->parent_bus->info->get_dev_path(dev);
1177         if (path) {
1178             pstrcpy(id, sizeof(id), path);
1179             pstrcat(id, sizeof(id), "/");
1180             qemu_free(path);
1181         }
1182     }
1183     pstrcat(id, sizeof(id), idstr);
1184
1185     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1186         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1187             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1188             if (se->compat) {
1189                 qemu_free(se->compat);
1190             }
1191             qemu_free(se);
1192         }
1193     }
1194 }
1195
1196 /* mark a device as not to be migrated, that is the device should be
1197    unplugged before migration */
1198 void register_device_unmigratable(DeviceState *dev, const char *idstr,
1199                                                             void *opaque)
1200 {
1201     SaveStateEntry *se;
1202     char id[256] = "";
1203
1204     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1205         char *path = dev->parent_bus->info->get_dev_path(dev);
1206         if (path) {
1207             pstrcpy(id, sizeof(id), path);
1208             pstrcat(id, sizeof(id), "/");
1209             qemu_free(path);
1210         }
1211     }
1212     pstrcat(id, sizeof(id), idstr);
1213
1214     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1215         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1216             se->no_migrate = 1;
1217         }
1218     }
1219 }
1220
1221 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1222                                    const VMStateDescription *vmsd,
1223                                    void *opaque, int alias_id,
1224                                    int required_for_version)
1225 {
1226     SaveStateEntry *se;
1227
1228     /* If this triggers, alias support can be dropped for the vmsd. */
1229     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1230
1231     se = qemu_mallocz(sizeof(SaveStateEntry));
1232     se->version_id = vmsd->version_id;
1233     se->section_id = global_section_id++;
1234     se->save_live_state = NULL;
1235     se->save_state = NULL;
1236     se->load_state = NULL;
1237     se->opaque = opaque;
1238     se->vmsd = vmsd;
1239     se->alias_id = alias_id;
1240
1241     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1242         char *id = dev->parent_bus->info->get_dev_path(dev);
1243         if (id) {
1244             pstrcpy(se->idstr, sizeof(se->idstr), id);
1245             pstrcat(se->idstr, sizeof(se->idstr), "/");
1246             qemu_free(id);
1247
1248             se->compat = qemu_mallocz(sizeof(CompatEntry));
1249             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1250             se->compat->instance_id = instance_id == -1 ?
1251                          calculate_compat_instance_id(vmsd->name) : instance_id;
1252             instance_id = -1;
1253         }
1254     }
1255     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1256
1257     if (instance_id == -1) {
1258         se->instance_id = calculate_new_instance_id(se->idstr);
1259     } else {
1260         se->instance_id = instance_id;
1261     }
1262     assert(!se->compat || se->instance_id == 0);
1263     /* add at the end of list */
1264     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1265     return 0;
1266 }
1267
1268 int vmstate_register(DeviceState *dev, int instance_id,
1269                      const VMStateDescription *vmsd, void *opaque)
1270 {
1271     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1272                                           opaque, -1, 0);
1273 }
1274
1275 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1276                         void *opaque)
1277 {
1278     SaveStateEntry *se, *new_se;
1279
1280     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1281         if (se->vmsd == vmsd && se->opaque == opaque) {
1282             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1283             if (se->compat) {
1284                 qemu_free(se->compat);
1285             }
1286             qemu_free(se);
1287         }
1288     }
1289 }
1290
1291 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1292                                     void *opaque);
1293 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1294                                    void *opaque);
1295
1296 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1297                        void *opaque, int version_id)
1298 {
1299     VMStateField *field = vmsd->fields;
1300     int ret;
1301
1302     if (version_id > vmsd->version_id) {
1303         return -EINVAL;
1304     }
1305     if (version_id < vmsd->minimum_version_id_old) {
1306         return -EINVAL;
1307     }
1308     if  (version_id < vmsd->minimum_version_id) {
1309         return vmsd->load_state_old(f, opaque, version_id);
1310     }
1311     if (vmsd->pre_load) {
1312         int ret = vmsd->pre_load(opaque);
1313         if (ret)
1314             return ret;
1315     }
1316     while(field->name) {
1317         if ((field->field_exists &&
1318              field->field_exists(opaque, version_id)) ||
1319             (!field->field_exists &&
1320              field->version_id <= version_id)) {
1321             void *base_addr = opaque + field->offset;
1322             int i, n_elems = 1;
1323             int size = field->size;
1324
1325             if (field->flags & VMS_VBUFFER) {
1326                 size = *(int32_t *)(opaque+field->size_offset);
1327                 if (field->flags & VMS_MULTIPLY) {
1328                     size *= field->size;
1329                 }
1330             }
1331             if (field->flags & VMS_ARRAY) {
1332                 n_elems = field->num;
1333             } else if (field->flags & VMS_VARRAY_INT32) {
1334                 n_elems = *(int32_t *)(opaque+field->num_offset);
1335             } else if (field->flags & VMS_VARRAY_UINT16) {
1336                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1337             }
1338             if (field->flags & VMS_POINTER) {
1339                 base_addr = *(void **)base_addr + field->start;
1340             }
1341             for (i = 0; i < n_elems; i++) {
1342                 void *addr = base_addr + size * i;
1343
1344                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1345                     addr = *(void **)addr;
1346                 }
1347                 if (field->flags & VMS_STRUCT) {
1348                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1349                 } else {
1350                     ret = field->info->get(f, addr, size);
1351
1352                 }
1353                 if (ret < 0) {
1354                     return ret;
1355                 }
1356             }
1357         }
1358         field++;
1359     }
1360     ret = vmstate_subsection_load(f, vmsd, opaque);
1361     if (ret != 0) {
1362         return ret;
1363     }
1364     if (vmsd->post_load) {
1365         return vmsd->post_load(opaque, version_id);
1366     }
1367     return 0;
1368 }
1369
1370 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1371                         void *opaque)
1372 {
1373     VMStateField *field = vmsd->fields;
1374
1375     if (vmsd->pre_save) {
1376         vmsd->pre_save(opaque);
1377     }
1378     while(field->name) {
1379         if (!field->field_exists ||
1380             field->field_exists(opaque, vmsd->version_id)) {
1381             void *base_addr = opaque + field->offset;
1382             int i, n_elems = 1;
1383             int size = field->size;
1384
1385             if (field->flags & VMS_VBUFFER) {
1386                 size = *(int32_t *)(opaque+field->size_offset);
1387                 if (field->flags & VMS_MULTIPLY) {
1388                     size *= field->size;
1389                 }
1390             }
1391             if (field->flags & VMS_ARRAY) {
1392                 n_elems = field->num;
1393             } else if (field->flags & VMS_VARRAY_INT32) {
1394                 n_elems = *(int32_t *)(opaque+field->num_offset);
1395             } else if (field->flags & VMS_VARRAY_UINT16) {
1396                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1397             }
1398             if (field->flags & VMS_POINTER) {
1399                 base_addr = *(void **)base_addr + field->start;
1400             }
1401             for (i = 0; i < n_elems; i++) {
1402                 void *addr = base_addr + size * i;
1403
1404                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1405                     addr = *(void **)addr;
1406                 }
1407                 if (field->flags & VMS_STRUCT) {
1408                     vmstate_save_state(f, field->vmsd, addr);
1409                 } else {
1410                     field->info->put(f, addr, size);
1411                 }
1412             }
1413         }
1414         field++;
1415     }
1416     vmstate_subsection_save(f, vmsd, opaque);
1417 }
1418
1419 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1420 {
1421     if (!se->vmsd) {         /* Old style */
1422         return se->load_state(f, se->opaque, version_id);
1423     }
1424     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1425 }
1426
1427 static int vmstate_save(QEMUFile *f, SaveStateEntry *se)
1428 {
1429     if (se->no_migrate) {
1430         return -1;
1431     }
1432
1433     if (!se->vmsd) {         /* Old style */
1434         se->save_state(f, se->opaque);
1435         return 0;
1436     }
1437     vmstate_save_state(f,se->vmsd, se->opaque);
1438
1439     return 0;
1440 }
1441
1442 #define QEMU_VM_FILE_MAGIC           0x5145564d
1443 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1444 #define QEMU_VM_FILE_VERSION         0x00000003
1445
1446 #define QEMU_VM_EOF                  0x00
1447 #define QEMU_VM_SECTION_START        0x01
1448 #define QEMU_VM_SECTION_PART         0x02
1449 #define QEMU_VM_SECTION_END          0x03
1450 #define QEMU_VM_SECTION_FULL         0x04
1451 #define QEMU_VM_SUBSECTION           0x05
1452
1453 int qemu_savevm_state_begin(Monitor *mon, QEMUFile *f, int blk_enable,
1454                             int shared)
1455 {
1456     SaveStateEntry *se;
1457
1458     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1459         if(se->set_params == NULL) {
1460             continue;
1461         }
1462         se->set_params(blk_enable, shared, se->opaque);
1463     }
1464
1465     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1466     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1467
1468     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1469         int len;
1470
1471         if (se->save_live_state == NULL)
1472             continue;
1473
1474         /* Section type */
1475         qemu_put_byte(f, QEMU_VM_SECTION_START);
1476         qemu_put_be32(f, se->section_id);
1477
1478         /* ID string */
1479         len = strlen(se->idstr);
1480         qemu_put_byte(f, len);
1481         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1482
1483         qemu_put_be32(f, se->instance_id);
1484         qemu_put_be32(f, se->version_id);
1485
1486         se->save_live_state(mon, f, QEMU_VM_SECTION_START, se->opaque);
1487     }
1488
1489     if (qemu_file_has_error(f)) {
1490         qemu_savevm_state_cancel(mon, f);
1491         return -EIO;
1492     }
1493
1494     return 0;
1495 }
1496
1497 int qemu_savevm_state_iterate(Monitor *mon, QEMUFile *f)
1498 {
1499     SaveStateEntry *se;
1500     int ret = 1;
1501
1502     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1503         if (se->save_live_state == NULL)
1504             continue;
1505
1506         /* Section type */
1507         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1508         qemu_put_be32(f, se->section_id);
1509
1510         ret = se->save_live_state(mon, f, QEMU_VM_SECTION_PART, se->opaque);
1511         if (!ret) {
1512             /* Do not proceed to the next vmstate before this one reported
1513                completion of the current stage. This serializes the migration
1514                and reduces the probability that a faster changing state is
1515                synchronized over and over again. */
1516             break;
1517         }
1518     }
1519
1520     if (ret)
1521         return 1;
1522
1523     if (qemu_file_has_error(f)) {
1524         qemu_savevm_state_cancel(mon, f);
1525         return -EIO;
1526     }
1527
1528     return 0;
1529 }
1530
1531 int qemu_savevm_state_complete(Monitor *mon, QEMUFile *f)
1532 {
1533     SaveStateEntry *se;
1534     int r;
1535
1536     cpu_synchronize_all_states();
1537
1538     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1539         if (se->save_live_state == NULL)
1540             continue;
1541
1542         /* Section type */
1543         qemu_put_byte(f, QEMU_VM_SECTION_END);
1544         qemu_put_be32(f, se->section_id);
1545
1546         se->save_live_state(mon, f, QEMU_VM_SECTION_END, se->opaque);
1547     }
1548
1549     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1550         int len;
1551
1552         if (se->save_state == NULL && se->vmsd == NULL)
1553             continue;
1554
1555         /* Section type */
1556         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1557         qemu_put_be32(f, se->section_id);
1558
1559         /* ID string */
1560         len = strlen(se->idstr);
1561         qemu_put_byte(f, len);
1562         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1563
1564         qemu_put_be32(f, se->instance_id);
1565         qemu_put_be32(f, se->version_id);
1566
1567         r = vmstate_save(f, se);
1568         if (r < 0) {
1569             monitor_printf(mon, "cannot migrate with device '%s'\n", se->idstr);
1570             return r;
1571         }
1572     }
1573
1574     qemu_put_byte(f, QEMU_VM_EOF);
1575
1576     if (qemu_file_has_error(f))
1577         return -EIO;
1578
1579     return 0;
1580 }
1581
1582 void qemu_savevm_state_cancel(Monitor *mon, QEMUFile *f)
1583 {
1584     SaveStateEntry *se;
1585
1586     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1587         if (se->save_live_state) {
1588             se->save_live_state(mon, f, -1, se->opaque);
1589         }
1590     }
1591 }
1592
1593 static int qemu_savevm_state(Monitor *mon, QEMUFile *f)
1594 {
1595     int saved_vm_running;
1596     int ret;
1597
1598     saved_vm_running = vm_running;
1599     vm_stop(0);
1600
1601     bdrv_flush_all();
1602
1603     ret = qemu_savevm_state_begin(mon, f, 0, 0);
1604     if (ret < 0)
1605         goto out;
1606
1607     do {
1608         ret = qemu_savevm_state_iterate(mon, f);
1609         if (ret < 0)
1610             goto out;
1611     } while (ret == 0);
1612
1613     ret = qemu_savevm_state_complete(mon, f);
1614
1615 out:
1616     if (qemu_file_has_error(f))
1617         ret = -EIO;
1618
1619     if (!ret && saved_vm_running)
1620         vm_start();
1621
1622     return ret;
1623 }
1624
1625 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1626 {
1627     SaveStateEntry *se;
1628
1629     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1630         if (!strcmp(se->idstr, idstr) &&
1631             (instance_id == se->instance_id ||
1632              instance_id == se->alias_id))
1633             return se;
1634         /* Migrating from an older version? */
1635         if (strstr(se->idstr, idstr) && se->compat) {
1636             if (!strcmp(se->compat->idstr, idstr) &&
1637                 (instance_id == se->compat->instance_id ||
1638                  instance_id == se->alias_id))
1639                 return se;
1640         }
1641     }
1642     return NULL;
1643 }
1644
1645 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1646 {
1647     while(sub && sub->needed) {
1648         if (strcmp(idstr, sub->vmsd->name) == 0) {
1649             return sub->vmsd;
1650         }
1651         sub++;
1652     }
1653     return NULL;
1654 }
1655
1656 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1657                                    void *opaque)
1658 {
1659     while (qemu_peek_byte(f) == QEMU_VM_SUBSECTION) {
1660         char idstr[256];
1661         int ret;
1662         uint8_t version_id, len;
1663         const VMStateDescription *sub_vmsd;
1664
1665         qemu_get_byte(f); /* subsection */
1666         len = qemu_get_byte(f);
1667         qemu_get_buffer(f, (uint8_t *)idstr, len);
1668         idstr[len] = 0;
1669         version_id = qemu_get_be32(f);
1670
1671         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
1672         if (sub_vmsd == NULL) {
1673             return -ENOENT;
1674         }
1675         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1676         if (ret) {
1677             return ret;
1678         }
1679     }
1680     return 0;
1681 }
1682
1683 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1684                                     void *opaque)
1685 {
1686     const VMStateSubsection *sub = vmsd->subsections;
1687
1688     while (sub && sub->needed) {
1689         if (sub->needed(opaque)) {
1690             const VMStateDescription *vmsd = sub->vmsd;
1691             uint8_t len;
1692
1693             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1694             len = strlen(vmsd->name);
1695             qemu_put_byte(f, len);
1696             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1697             qemu_put_be32(f, vmsd->version_id);
1698             vmstate_save_state(f, vmsd, opaque);
1699         }
1700         sub++;
1701     }
1702 }
1703
1704 typedef struct LoadStateEntry {
1705     QLIST_ENTRY(LoadStateEntry) entry;
1706     SaveStateEntry *se;
1707     int section_id;
1708     int version_id;
1709 } LoadStateEntry;
1710
1711 int qemu_loadvm_state(QEMUFile *f)
1712 {
1713     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1714         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1715     LoadStateEntry *le, *new_le;
1716     uint8_t section_type;
1717     unsigned int v;
1718     int ret;
1719
1720     v = qemu_get_be32(f);
1721     if (v != QEMU_VM_FILE_MAGIC)
1722         return -EINVAL;
1723
1724     v = qemu_get_be32(f);
1725     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1726         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1727         return -ENOTSUP;
1728     }
1729     if (v != QEMU_VM_FILE_VERSION)
1730         return -ENOTSUP;
1731
1732     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1733         uint32_t instance_id, version_id, section_id;
1734         SaveStateEntry *se;
1735         char idstr[257];
1736         int len;
1737
1738         switch (section_type) {
1739         case QEMU_VM_SECTION_START:
1740         case QEMU_VM_SECTION_FULL:
1741             /* Read section start */
1742             section_id = qemu_get_be32(f);
1743             len = qemu_get_byte(f);
1744             qemu_get_buffer(f, (uint8_t *)idstr, len);
1745             idstr[len] = 0;
1746             instance_id = qemu_get_be32(f);
1747             version_id = qemu_get_be32(f);
1748
1749             /* Find savevm section */
1750             se = find_se(idstr, instance_id);
1751             if (se == NULL) {
1752                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1753                 ret = -EINVAL;
1754                 goto out;
1755             }
1756
1757             /* Validate version */
1758             if (version_id > se->version_id) {
1759                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
1760                         version_id, idstr, se->version_id);
1761                 ret = -EINVAL;
1762                 goto out;
1763             }
1764
1765             /* Add entry */
1766             le = qemu_mallocz(sizeof(*le));
1767
1768             le->se = se;
1769             le->section_id = section_id;
1770             le->version_id = version_id;
1771             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
1772
1773             ret = vmstate_load(f, le->se, le->version_id);
1774             if (ret < 0) {
1775                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
1776                         instance_id, idstr);
1777                 goto out;
1778             }
1779             break;
1780         case QEMU_VM_SECTION_PART:
1781         case QEMU_VM_SECTION_END:
1782             section_id = qemu_get_be32(f);
1783
1784             QLIST_FOREACH(le, &loadvm_handlers, entry) {
1785                 if (le->section_id == section_id) {
1786                     break;
1787                 }
1788             }
1789             if (le == NULL) {
1790                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
1791                 ret = -EINVAL;
1792                 goto out;
1793             }
1794
1795             ret = vmstate_load(f, le->se, le->version_id);
1796             if (ret < 0) {
1797                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
1798                         section_id);
1799                 goto out;
1800             }
1801             break;
1802         default:
1803             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
1804             ret = -EINVAL;
1805             goto out;
1806         }
1807     }
1808
1809     cpu_synchronize_all_post_init();
1810
1811     ret = 0;
1812
1813 out:
1814     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
1815         QLIST_REMOVE(le, entry);
1816         qemu_free(le);
1817     }
1818
1819     if (qemu_file_has_error(f))
1820         ret = -EIO;
1821
1822     return ret;
1823 }
1824
1825 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1826                               const char *name)
1827 {
1828     QEMUSnapshotInfo *sn_tab, *sn;
1829     int nb_sns, i, ret;
1830
1831     ret = -ENOENT;
1832     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1833     if (nb_sns < 0)
1834         return ret;
1835     for(i = 0; i < nb_sns; i++) {
1836         sn = &sn_tab[i];
1837         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1838             *sn_info = *sn;
1839             ret = 0;
1840             break;
1841         }
1842     }
1843     qemu_free(sn_tab);
1844     return ret;
1845 }
1846
1847 /*
1848  * Deletes snapshots of a given name in all opened images.
1849  */
1850 static int del_existing_snapshots(Monitor *mon, const char *name)
1851 {
1852     BlockDriverState *bs;
1853     QEMUSnapshotInfo sn1, *snapshot = &sn1;
1854     int ret;
1855
1856     bs = NULL;
1857     while ((bs = bdrv_next(bs))) {
1858         if (bdrv_can_snapshot(bs) &&
1859             bdrv_snapshot_find(bs, snapshot, name) >= 0)
1860         {
1861             ret = bdrv_snapshot_delete(bs, name);
1862             if (ret < 0) {
1863                 monitor_printf(mon,
1864                                "Error while deleting snapshot on '%s'\n",
1865                                bdrv_get_device_name(bs));
1866                 return -1;
1867             }
1868         }
1869     }
1870
1871     return 0;
1872 }
1873
1874 void do_savevm(Monitor *mon, const QDict *qdict)
1875 {
1876     BlockDriverState *bs, *bs1;
1877     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1878     int ret;
1879     QEMUFile *f;
1880     int saved_vm_running;
1881     uint32_t vm_state_size;
1882 #ifdef _WIN32
1883     struct _timeb tb;
1884     struct tm *ptm;
1885 #else
1886     struct timeval tv;
1887     struct tm tm;
1888 #endif
1889     const char *name = qdict_get_try_str(qdict, "name");
1890
1891     /* Verify if there is a device that doesn't support snapshots and is writable */
1892     bs = NULL;
1893     while ((bs = bdrv_next(bs))) {
1894
1895         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1896             continue;
1897         }
1898
1899         if (!bdrv_can_snapshot(bs)) {
1900             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
1901                                bdrv_get_device_name(bs));
1902             return;
1903         }
1904     }
1905
1906     bs = bdrv_snapshots();
1907     if (!bs) {
1908         monitor_printf(mon, "No block device can accept snapshots\n");
1909         return;
1910     }
1911     /* ??? Should this occur after vm_stop?  */
1912     qemu_aio_flush();
1913
1914     saved_vm_running = vm_running;
1915     vm_stop(0);
1916
1917     memset(sn, 0, sizeof(*sn));
1918
1919     /* fill auxiliary fields */
1920 #ifdef _WIN32
1921     _ftime(&tb);
1922     sn->date_sec = tb.time;
1923     sn->date_nsec = tb.millitm * 1000000;
1924 #else
1925     gettimeofday(&tv, NULL);
1926     sn->date_sec = tv.tv_sec;
1927     sn->date_nsec = tv.tv_usec * 1000;
1928 #endif
1929     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1930
1931     if (name) {
1932         ret = bdrv_snapshot_find(bs, old_sn, name);
1933         if (ret >= 0) {
1934             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1935             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1936         } else {
1937             pstrcpy(sn->name, sizeof(sn->name), name);
1938         }
1939     } else {
1940 #ifdef _WIN32
1941         ptm = localtime(&tb.time);
1942         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", ptm);
1943 #else
1944         /* cast below needed for OpenBSD where tv_sec is still 'long' */
1945         localtime_r((const time_t *)&tv.tv_sec, &tm);
1946         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
1947 #endif
1948     }
1949
1950     /* Delete old snapshots of the same name */
1951     if (name && del_existing_snapshots(mon, name) < 0) {
1952         goto the_end;
1953     }
1954
1955     /* save the VM state */
1956     f = qemu_fopen_bdrv(bs, 1);
1957     if (!f) {
1958         monitor_printf(mon, "Could not open VM state file\n");
1959         goto the_end;
1960     }
1961     ret = qemu_savevm_state(mon, f);
1962     vm_state_size = qemu_ftell(f);
1963     qemu_fclose(f);
1964     if (ret < 0) {
1965         monitor_printf(mon, "Error %d while writing VM\n", ret);
1966         goto the_end;
1967     }
1968
1969     /* create the snapshots */
1970
1971     bs1 = NULL;
1972     while ((bs1 = bdrv_next(bs1))) {
1973         if (bdrv_can_snapshot(bs1)) {
1974             /* Write VM state size only to the image that contains the state */
1975             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1976             ret = bdrv_snapshot_create(bs1, sn);
1977             if (ret < 0) {
1978                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1979                                bdrv_get_device_name(bs1));
1980             }
1981         }
1982     }
1983
1984  the_end:
1985     if (saved_vm_running)
1986         vm_start();
1987 }
1988
1989 int load_vmstate(const char *name)
1990 {
1991     BlockDriverState *bs, *bs_vm_state;
1992     QEMUSnapshotInfo sn;
1993     QEMUFile *f;
1994     int ret;
1995
1996     bs_vm_state = bdrv_snapshots();
1997     if (!bs_vm_state) {
1998         error_report("No block device supports snapshots");
1999         return -ENOTSUP;
2000     }
2001
2002     /* Don't even try to load empty VM states */
2003     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
2004     if (ret < 0) {
2005         return ret;
2006     } else if (sn.vm_state_size == 0) {
2007         return -EINVAL;
2008     }
2009
2010     /* Verify if there is any device that doesn't support snapshots and is
2011     writable and check if the requested snapshot is available too. */
2012     bs = NULL;
2013     while ((bs = bdrv_next(bs))) {
2014
2015         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
2016             continue;
2017         }
2018
2019         if (!bdrv_can_snapshot(bs)) {
2020             error_report("Device '%s' is writable but does not support snapshots.",
2021                                bdrv_get_device_name(bs));
2022             return -ENOTSUP;
2023         }
2024
2025         ret = bdrv_snapshot_find(bs, &sn, name);
2026         if (ret < 0) {
2027             error_report("Device '%s' does not have the requested snapshot '%s'",
2028                            bdrv_get_device_name(bs), name);
2029             return ret;
2030         }
2031     }
2032
2033     /* Flush all IO requests so they don't interfere with the new state.  */
2034     qemu_aio_flush();
2035
2036     bs = NULL;
2037     while ((bs = bdrv_next(bs))) {
2038         if (bdrv_can_snapshot(bs)) {
2039             ret = bdrv_snapshot_goto(bs, name);
2040             if (ret < 0) {
2041                 error_report("Error %d while activating snapshot '%s' on '%s'",
2042                              ret, name, bdrv_get_device_name(bs));
2043                 return ret;
2044             }
2045         }
2046     }
2047
2048     /* restore the VM state */
2049     f = qemu_fopen_bdrv(bs_vm_state, 0);
2050     if (!f) {
2051         error_report("Could not open VM state file");
2052         return -EINVAL;
2053     }
2054
2055     ret = qemu_loadvm_state(f);
2056
2057     qemu_fclose(f);
2058     if (ret < 0) {
2059         error_report("Error %d while loading VM state", ret);
2060         return ret;
2061     }
2062
2063     return 0;
2064 }
2065
2066 void do_delvm(Monitor *mon, const QDict *qdict)
2067 {
2068     BlockDriverState *bs, *bs1;
2069     int ret;
2070     const char *name = qdict_get_str(qdict, "name");
2071
2072     bs = bdrv_snapshots();
2073     if (!bs) {
2074         monitor_printf(mon, "No block device supports snapshots\n");
2075         return;
2076     }
2077
2078     bs1 = NULL;
2079     while ((bs1 = bdrv_next(bs1))) {
2080         if (bdrv_can_snapshot(bs1)) {
2081             ret = bdrv_snapshot_delete(bs1, name);
2082             if (ret < 0) {
2083                 if (ret == -ENOTSUP)
2084                     monitor_printf(mon,
2085                                    "Snapshots not supported on device '%s'\n",
2086                                    bdrv_get_device_name(bs1));
2087                 else
2088                     monitor_printf(mon, "Error %d while deleting snapshot on "
2089                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2090             }
2091         }
2092     }
2093 }
2094
2095 void do_info_snapshots(Monitor *mon)
2096 {
2097     BlockDriverState *bs, *bs1;
2098     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2099     int nb_sns, i, ret, available;
2100     int total;
2101     int *available_snapshots;
2102     char buf[256];
2103
2104     bs = bdrv_snapshots();
2105     if (!bs) {
2106         monitor_printf(mon, "No available block device supports snapshots\n");
2107         return;
2108     }
2109
2110     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2111     if (nb_sns < 0) {
2112         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2113         return;
2114     }
2115
2116     if (nb_sns == 0) {
2117         monitor_printf(mon, "There is no snapshot available.\n");
2118         return;
2119     }
2120
2121     available_snapshots = qemu_mallocz(sizeof(int) * nb_sns);
2122     total = 0;
2123     for (i = 0; i < nb_sns; i++) {
2124         sn = &sn_tab[i];
2125         available = 1;
2126         bs1 = NULL;
2127
2128         while ((bs1 = bdrv_next(bs1))) {
2129             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2130                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2131                 if (ret < 0) {
2132                     available = 0;
2133                     break;
2134                 }
2135             }
2136         }
2137
2138         if (available) {
2139             available_snapshots[total] = i;
2140             total++;
2141         }
2142     }
2143
2144     if (total > 0) {
2145         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2146         for (i = 0; i < total; i++) {
2147             sn = &sn_tab[available_snapshots[i]];
2148             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2149         }
2150     } else {
2151         monitor_printf(mon, "There is no suitable snapshot available\n");
2152     }
2153
2154     qemu_free(sn_tab);
2155     qemu_free(available_snapshots);
2156
2157 }