savevm.c

   1 /*
   2  * QEMU System Emulator
   3  *
   4  * Copyright (c) 2003-2008 Fabrice Bellard
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a copy
   7  * of this software and associated documentation files (the "Software"), to deal
   8  * in the Software without restriction, including without limitation the rights
   9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  10  * copies of the Software, and to permit persons to whom the Software is
  11  * furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice shall be included in
  14  * all copies or substantial portions of the Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  22  * THE SOFTWARE.
  23  */
  24 #include <unistd.h>
  25 #include <fcntl.h>
  26 #include <signal.h>
  27 #include <time.h>
  28 #include <errno.h>
  29 #include <sys/time.h>
  30 #include <zlib.h>
  31
  32 /* Needed early for CONFIG_BSD etc. */
  33 #include "config-host.h"
  34
  35 #ifndef _WIN32
  36 #include <sys/times.h>
  37 #include <sys/wait.h>
  38 #include <termios.h>
  39 #include <sys/mman.h>
  40 #include <sys/ioctl.h>
  41 #include <sys/resource.h>
  42 #include <sys/socket.h>
  43 #include <netinet/in.h>
  44 #include <net/if.h>
  45 #include <arpa/inet.h>
  46 #include <dirent.h>
  47 #include <netdb.h>
  48 #include <sys/select.h>
  49 #ifdef CONFIG_BSD
  50 #include <sys/stat.h>
  51 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) || defined(__DragonFly__)
  52 #include <libutil.h>
  53 #else
  54 #include <util.h>
  55 #endif
  56 #ifdef __linux__
  57 #include <pty.h>
  58 #include <malloc.h>
  59 #include <linux/rtc.h>
  60 #endif
  61 #endif
  62 #endif
  63
  64 #ifdef _WIN32
  65 #include <windows.h>
  66 #include <malloc.h>
  67 #include <sys/timeb.h>
  68 #include <mmsystem.h>
  69 #define getopt_long_only getopt_long
  70 #define memalign(align, size) malloc(size)
  71 #endif
  72
  73 #include "qemu-common.h"
  74 #include "hw/hw.h"
  75 #include "hw/qdev.h"
  76 #include "net.h"
  77 #include "monitor.h"
  78 #include "sysemu.h"
  79 #include "qemu-timer.h"
  80 #include "qemu-char.h"
  81 #include "blockdev.h"
  82 #include "audio/audio.h"
  83 #include "migration.h"
  84 #include "qemu_socket.h"
  85 #include "qemu-queue.h"
  86
  87 #define SELF_ANNOUNCE_ROUNDS 5
  88
  89 #ifndef ETH_P_RARP
  90 #define ETH_P_RARP 0x8035
  91 #endif
  92 #define ARP_HTYPE_ETH 0x0001
  93 #define ARP_PTYPE_IP 0x0800
  94 #define ARP_OP_REQUEST_REV 0x3
  95
  96 static int announce_self_create(uint8_t *buf,
  97                                 uint8_t *mac_addr)
  98 {
  99     /* Ethernet header. */
 100     memset(buf, 0xff, 6);         /* destination MAC addr */
 101     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
 102     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
 103
 104     /* RARP header. */
 105     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
 106     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
 107     *(buf + 18) = 6; /* hardware addr length (ethernet) */
 108     *(buf + 19) = 4; /* protocol addr length (IPv4) */
 109     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
 110     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
 111     memset(buf + 28, 0x00, 4);     /* source protocol addr */
 112     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
 113     memset(buf + 38, 0x00, 4);     /* target protocol addr */
 114
 115     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
 116     memset(buf + 42, 0x00, 18);
 117
 118     return 60; /* len (FCS will be added by hardware) */
 119 }
 120
 121 static void qemu_announce_self_iter(NICState *nic, void *opaque)
 122 {
 123     uint8_t buf[60];
 124     int len;
 125
 126     len = announce_self_create(buf, nic->conf->macaddr.a);
 127
 128     qemu_send_packet_raw(&nic->nc, buf, len);
 129 }
 130
 131
 132 static void qemu_announce_self_once(void *opaque)
 133 {
 134     static int count = SELF_ANNOUNCE_ROUNDS;
 135     QEMUTimer *timer = *(QEMUTimer **)opaque;
 136
 137     qemu_foreach_nic(qemu_announce_self_iter, NULL);
 138
 139     if (--count) {
 140         /* delay 50ms, 150ms, 250ms, ... */
 141         qemu_mod_timer(timer, qemu_get_clock(rt_clock) +
 142                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
 143     } else {
 144             qemu_del_timer(timer);
 145             qemu_free_timer(timer);
 146     }
 147 }
 148
 149 void qemu_announce_self(void)
 150 {
 151         static QEMUTimer *timer;
 152         timer = qemu_new_timer(rt_clock, qemu_announce_self_once, &timer);
 153         qemu_announce_self_once(&timer);
 154 }
 155
 156 /***********************************************************/
 157 /* savevm/loadvm support */
 158
 159 #define IO_BUF_SIZE 32768
 160
 161 struct QEMUFile {
 162     QEMUFilePutBufferFunc *put_buffer;
 163     QEMUFileGetBufferFunc *get_buffer;
 164     QEMUFileCloseFunc *close;
 165     QEMUFileRateLimit *rate_limit;
 166     QEMUFileSetRateLimit *set_rate_limit;
 167     QEMUFileGetRateLimit *get_rate_limit;
 168     void *opaque;
 169     int is_write;
 170
 171     int64_t buf_offset; /* start of buffer when writing, end of buffer
 172                            when reading */
 173     int buf_index;
 174     int buf_size; /* 0 when writing */
 175     uint8_t buf[IO_BUF_SIZE];
 176
 177     int has_error;
 178 };
 179
 180 typedef struct QEMUFileStdio
 181 {
 182     FILE *stdio_file;
 183     QEMUFile *file;
 184 } QEMUFileStdio;
 185
 186 typedef struct QEMUFileSocket
 187 {
 188     int fd;
 189     QEMUFile *file;
 190 } QEMUFileSocket;
 191
 192 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 193 {
 194     QEMUFileSocket *s = opaque;
 195     ssize_t len;
 196
 197     do {
 198         len = recv(s->fd, (void *)buf, size, 0);
 199     } while (len == -1 && socket_error() == EINTR);
 200
 201     if (len == -1)
 202         len = -socket_error();
 203
 204     return len;
 205 }
 206
 207 static int socket_close(void *opaque)
 208 {
 209     QEMUFileSocket *s = opaque;
 210     qemu_free(s);
 211     return 0;
 212 }
 213
 214 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 215 {
 216     QEMUFileStdio *s = opaque;
 217     return fwrite(buf, 1, size, s->stdio_file);
 218 }
 219
 220 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 221 {
 222     QEMUFileStdio *s = opaque;
 223     FILE *fp = s->stdio_file;
 224     int bytes;
 225
 226     do {
 227         clearerr(fp);
 228         bytes = fread(buf, 1, size, fp);
 229     } while ((bytes == 0) && ferror(fp) && (errno == EINTR));
 230     return bytes;
 231 }
 232
 233 static int stdio_pclose(void *opaque)
 234 {
 235     QEMUFileStdio *s = opaque;
 236     int ret;
 237     ret = pclose(s->stdio_file);
 238     qemu_free(s);
 239     return ret;
 240 }
 241
 242 static int stdio_fclose(void *opaque)
 243 {
 244     QEMUFileStdio *s = opaque;
 245     fclose(s->stdio_file);
 246     qemu_free(s);
 247     return 0;
 248 }
 249
 250 QEMUFile *qemu_popen(FILE *stdio_file, const char *mode)
 251 {
 252     QEMUFileStdio *s;
 253
 254     if (stdio_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 255         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 256         return NULL;
 257     }
 258
 259     s = qemu_mallocz(sizeof(QEMUFileStdio));
 260
 261     s->stdio_file = stdio_file;
 262
 263     if(mode[0] == 'r') {
 264         s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_pclose,
 265                                  NULL, NULL, NULL);
 266     } else {
 267         s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_pclose,
 268                                  NULL, NULL, NULL);
 269     }
 270     return s->file;
 271 }
 272
 273 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 274 {
 275     FILE *popen_file;
 276
 277     popen_file = popen(command, mode);
 278     if(popen_file == NULL) {
 279         return NULL;
 280     }
 281
 282     return qemu_popen(popen_file, mode);
 283 }
 284
 285 int qemu_stdio_fd(QEMUFile *f)
 286 {
 287     QEMUFileStdio *p;
 288     int fd;
 289
 290     p = (QEMUFileStdio *)f->opaque;
 291     fd = fileno(p->stdio_file);
 292
 293     return fd;
 294 }
 295
 296 QEMUFile *qemu_fdopen(int fd, const char *mode)
 297 {
 298     QEMUFileStdio *s;
 299
 300     if (mode == NULL ||
 301         (mode[0] != 'r' && mode[0] != 'w') ||
 302         mode[1] != 'b' || mode[2] != 0) {
 303         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 304         return NULL;
 305     }
 306
 307     s = qemu_mallocz(sizeof(QEMUFileStdio));
 308     s->stdio_file = fdopen(fd, mode);
 309     if (!s->stdio_file)
 310         goto fail;
 311
 312     if(mode[0] == 'r') {
 313         s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_fclose,
 314                                  NULL, NULL, NULL);
 315     } else {
 316         s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_fclose,
 317                                  NULL, NULL, NULL);
 318     }
 319     return s->file;
 320
 321 fail:
 322     qemu_free(s);
 323     return NULL;
 324 }
 325
 326 QEMUFile *qemu_fopen_socket(int fd)
 327 {
 328     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 329
 330     s->fd = fd;
 331     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close,
 332                              NULL, NULL, NULL);
 333     return s->file;
 334 }
 335
 336 static int file_put_buffer(void *opaque, const uint8_t *buf,
 337                             int64_t pos, int size)
 338 {
 339     QEMUFileStdio *s = opaque;
 340     fseek(s->stdio_file, pos, SEEK_SET);
 341     return fwrite(buf, 1, size, s->stdio_file);
 342 }
 343
 344 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 345 {
 346     QEMUFileStdio *s = opaque;
 347     fseek(s->stdio_file, pos, SEEK_SET);
 348     return fread(buf, 1, size, s->stdio_file);
 349 }
 350
 351 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 352 {
 353     QEMUFileStdio *s;
 354
 355     if (mode == NULL ||
 356         (mode[0] != 'r' && mode[0] != 'w') ||
 357         mode[1] != 'b' || mode[2] != 0) {
 358         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 359         return NULL;
 360     }
 361
 362     s = qemu_mallocz(sizeof(QEMUFileStdio));
 363
 364     s->stdio_file = fopen(filename, mode);
 365     if (!s->stdio_file)
 366         goto fail;
 367
 368     if(mode[0] == 'w') {
 369         s->file = qemu_fopen_ops(s, file_put_buffer, NULL, stdio_fclose,
 370                                  NULL, NULL, NULL);
 371     } else {
 372         s->file = qemu_fopen_ops(s, NULL, file_get_buffer, stdio_fclose,
 373                                NULL, NULL, NULL);
 374     }
 375     return s->file;
 376 fail:
 377     qemu_free(s);
 378     return NULL;
 379 }
 380
 381 static int block_put_buffer(void *opaque, const uint8_t *buf,
 382                            int64_t pos, int size)
 383 {
 384     bdrv_save_vmstate(opaque, buf, pos, size);
 385     return size;
 386 }
 387
 388 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 389 {
 390     return bdrv_load_vmstate(opaque, buf, pos, size);
 391 }
 392
 393 static int bdrv_fclose(void *opaque)
 394 {
 395     return 0;
 396 }
 397
 398 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 399 {
 400     if (is_writable)
 401         return qemu_fopen_ops(bs, block_put_buffer, NULL, bdrv_fclose,
 402                               NULL, NULL, NULL);
 403     return qemu_fopen_ops(bs, NULL, block_get_buffer, bdrv_fclose, NULL, NULL, NULL);
 404 }
 405
 406 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 407                          QEMUFileGetBufferFunc *get_buffer,
 408                          QEMUFileCloseFunc *close,
 409                          QEMUFileRateLimit *rate_limit,
 410                          QEMUFileSetRateLimit *set_rate_limit,
 411                          QEMUFileGetRateLimit *get_rate_limit)
 412 {
 413     QEMUFile *f;
 414
 415     f = qemu_mallocz(sizeof(QEMUFile));
 416
 417     f->opaque = opaque;
 418     f->put_buffer = put_buffer;
 419     f->get_buffer = get_buffer;
 420     f->close = close;
 421     f->rate_limit = rate_limit;
 422     f->set_rate_limit = set_rate_limit;
 423     f->get_rate_limit = get_rate_limit;
 424     f->is_write = 0;
 425
 426     return f;
 427 }
 428
 429 int qemu_file_has_error(QEMUFile *f)
 430 {
 431     return f->has_error;
 432 }
 433
 434 void qemu_file_set_error(QEMUFile *f)
 435 {
 436     f->has_error = 1;
 437 }
 438
 439 void qemu_fflush(QEMUFile *f)
 440 {
 441     if (!f->put_buffer)
 442         return;
 443
 444     if (f->is_write && f->buf_index > 0) {
 445         int len;
 446
 447         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 448         if (len > 0)
 449             f->buf_offset += f->buf_index;
 450         else
 451             f->has_error = 1;
 452         f->buf_index = 0;
 453     }
 454 }
 455
 456 static void qemu_fill_buffer(QEMUFile *f)
 457 {
 458     int len;
 459
 460     if (!f->get_buffer)
 461         return;
 462
 463     if (f->is_write)
 464         abort();
 465
 466     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 467     if (len > 0) {
 468         f->buf_index = 0;
 469         f->buf_size = len;
 470         f->buf_offset += len;
 471     } else if (len != -EAGAIN)
 472         f->has_error = 1;
 473 }
 474
 475 int qemu_fclose(QEMUFile *f)
 476 {
 477     int ret = 0;
 478     qemu_fflush(f);
 479     if (f->close)
 480         ret = f->close(f->opaque);
 481     qemu_free(f);
 482     return ret;
 483 }
 484
 485 void qemu_file_put_notify(QEMUFile *f)
 486 {
 487     f->put_buffer(f->opaque, NULL, 0, 0);
 488 }
 489
 490 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 491 {
 492     int l;
 493
 494     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 495         fprintf(stderr,
 496                 "Attempted to write to buffer while read buffer is not empty\n");
 497         abort();
 498     }
 499
 500     while (!f->has_error && size > 0) {
 501         l = IO_BUF_SIZE - f->buf_index;
 502         if (l > size)
 503             l = size;
 504         memcpy(f->buf + f->buf_index, buf, l);
 505         f->is_write = 1;
 506         f->buf_index += l;
 507         buf += l;
 508         size -= l;
 509         if (f->buf_index >= IO_BUF_SIZE)
 510             qemu_fflush(f);
 511     }
 512 }
 513
 514 void qemu_put_byte(QEMUFile *f, int v)
 515 {
 516     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 517         fprintf(stderr,
 518                 "Attempted to write to buffer while read buffer is not empty\n");
 519         abort();
 520     }
 521
 522     f->buf[f->buf_index++] = v;
 523     f->is_write = 1;
 524     if (f->buf_index >= IO_BUF_SIZE)
 525         qemu_fflush(f);
 526 }
 527
 528 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 529 {
 530     int size, l;
 531
 532     if (f->is_write)
 533         abort();
 534
 535     size = size1;
 536     while (size > 0) {
 537         l = f->buf_size - f->buf_index;
 538         if (l == 0) {
 539             qemu_fill_buffer(f);
 540             l = f->buf_size - f->buf_index;
 541             if (l == 0)
 542                 break;
 543         }
 544         if (l > size)
 545             l = size;
 546         memcpy(buf, f->buf + f->buf_index, l);
 547         f->buf_index += l;
 548         buf += l;
 549         size -= l;
 550     }
 551     return size1 - size;
 552 }
 553
 554 static int qemu_peek_byte(QEMUFile *f)
 555 {
 556     if (f->is_write)
 557         abort();
 558
 559     if (f->buf_index >= f->buf_size) {
 560         qemu_fill_buffer(f);
 561         if (f->buf_index >= f->buf_size)
 562             return 0;
 563     }
 564     return f->buf[f->buf_index];
 565 }
 566
 567 int qemu_get_byte(QEMUFile *f)
 568 {
 569     if (f->is_write)
 570         abort();
 571
 572     if (f->buf_index >= f->buf_size) {
 573         qemu_fill_buffer(f);
 574         if (f->buf_index >= f->buf_size)
 575             return 0;
 576     }
 577     return f->buf[f->buf_index++];
 578 }
 579
 580 int64_t qemu_ftell(QEMUFile *f)
 581 {
 582     return f->buf_offset - f->buf_size + f->buf_index;
 583 }
 584
 585 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 586 {
 587     if (whence == SEEK_SET) {
 588         /* nothing to do */
 589     } else if (whence == SEEK_CUR) {
 590         pos += qemu_ftell(f);
 591     } else {
 592         /* SEEK_END not supported */
 593         return -1;
 594     }
 595     if (f->put_buffer) {
 596         qemu_fflush(f);
 597         f->buf_offset = pos;
 598     } else {
 599         f->buf_offset = pos;
 600         f->buf_index = 0;
 601         f->buf_size = 0;
 602     }
 603     return pos;
 604 }
 605
 606 int qemu_file_rate_limit(QEMUFile *f)
 607 {
 608     if (f->rate_limit)
 609         return f->rate_limit(f->opaque);
 610
 611     return 0;
 612 }
 613
 614 size_t qemu_file_get_rate_limit(QEMUFile *f)
 615 {
 616     if (f->get_rate_limit)
 617         return f->get_rate_limit(f->opaque);
 618
 619     return 0;
 620 }
 621
 622 size_t qemu_file_set_rate_limit(QEMUFile *f, size_t new_rate)
 623 {
 624     /* any failed or completed migration keeps its state to allow probing of
 625      * migration data, but has no associated file anymore */
 626     if (f && f->set_rate_limit)
 627         return f->set_rate_limit(f->opaque, new_rate);
 628
 629     return 0;
 630 }
 631
 632 void qemu_put_be16(QEMUFile *f, unsigned int v)
 633 {
 634     qemu_put_byte(f, v >> 8);
 635     qemu_put_byte(f, v);
 636 }
 637
 638 void qemu_put_be32(QEMUFile *f, unsigned int v)
 639 {
 640     qemu_put_byte(f, v >> 24);
 641     qemu_put_byte(f, v >> 16);
 642     qemu_put_byte(f, v >> 8);
 643     qemu_put_byte(f, v);
 644 }
 645
 646 void qemu_put_be64(QEMUFile *f, uint64_t v)
 647 {
 648     qemu_put_be32(f, v >> 32);
 649     qemu_put_be32(f, v);
 650 }
 651
 652 unsigned int qemu_get_be16(QEMUFile *f)
 653 {
 654     unsigned int v;
 655     v = qemu_get_byte(f) << 8;
 656     v |= qemu_get_byte(f);
 657     return v;
 658 }
 659
 660 unsigned int qemu_get_be32(QEMUFile *f)
 661 {
 662     unsigned int v;
 663     v = qemu_get_byte(f) << 24;
 664     v |= qemu_get_byte(f) << 16;
 665     v |= qemu_get_byte(f) << 8;
 666     v |= qemu_get_byte(f);
 667     return v;
 668 }
 669
 670 uint64_t qemu_get_be64(QEMUFile *f)
 671 {
 672     uint64_t v;
 673     v = (uint64_t)qemu_get_be32(f) << 32;
 674     v |= qemu_get_be32(f);
 675     return v;
 676 }
 677
 678 /* 8 bit int */
 679
 680 static int get_int8(QEMUFile *f, void *pv, size_t size)
 681 {
 682     int8_t *v = pv;
 683     qemu_get_s8s(f, v);
 684     return 0;
 685 }
 686
 687 static void put_int8(QEMUFile *f, void *pv, size_t size)
 688 {
 689     int8_t *v = pv;
 690     qemu_put_s8s(f, v);
 691 }
 692
 693 const VMStateInfo vmstate_info_int8 = {
 694     .name = "int8",
 695     .get  = get_int8,
 696     .put  = put_int8,
 697 };
 698
 699 /* 16 bit int */
 700
 701 static int get_int16(QEMUFile *f, void *pv, size_t size)
 702 {
 703     int16_t *v = pv;
 704     qemu_get_sbe16s(f, v);
 705     return 0;
 706 }
 707
 708 static void put_int16(QEMUFile *f, void *pv, size_t size)
 709 {
 710     int16_t *v = pv;
 711     qemu_put_sbe16s(f, v);
 712 }
 713
 714 const VMStateInfo vmstate_info_int16 = {
 715     .name = "int16",
 716     .get  = get_int16,
 717     .put  = put_int16,
 718 };
 719
 720 /* 32 bit int */
 721
 722 static int get_int32(QEMUFile *f, void *pv, size_t size)
 723 {
 724     int32_t *v = pv;
 725     qemu_get_sbe32s(f, v);
 726     return 0;
 727 }
 728
 729 static void put_int32(QEMUFile *f, void *pv, size_t size)
 730 {
 731     int32_t *v = pv;
 732     qemu_put_sbe32s(f, v);
 733 }
 734
 735 const VMStateInfo vmstate_info_int32 = {
 736     .name = "int32",
 737     .get  = get_int32,
 738     .put  = put_int32,
 739 };
 740
 741 /* 32 bit int. See that the received value is the same than the one
 742    in the field */
 743
 744 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 745 {
 746     int32_t *v = pv;
 747     int32_t v2;
 748     qemu_get_sbe32s(f, &v2);
 749
 750     if (*v == v2)
 751         return 0;
 752     return -EINVAL;
 753 }
 754
 755 const VMStateInfo vmstate_info_int32_equal = {
 756     .name = "int32 equal",
 757     .get  = get_int32_equal,
 758     .put  = put_int32,
 759 };
 760
 761 /* 32 bit int. See that the received value is the less or the same
 762    than the one in the field */
 763
 764 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 765 {
 766     int32_t *old = pv;
 767     int32_t new;
 768     qemu_get_sbe32s(f, &new);
 769
 770     if (*old <= new)
 771         return 0;
 772     return -EINVAL;
 773 }
 774
 775 const VMStateInfo vmstate_info_int32_le = {
 776     .name = "int32 equal",
 777     .get  = get_int32_le,
 778     .put  = put_int32,
 779 };
 780
 781 /* 64 bit int */
 782
 783 static int get_int64(QEMUFile *f, void *pv, size_t size)
 784 {
 785     int64_t *v = pv;
 786     qemu_get_sbe64s(f, v);
 787     return 0;
 788 }
 789
 790 static void put_int64(QEMUFile *f, void *pv, size_t size)
 791 {
 792     int64_t *v = pv;
 793     qemu_put_sbe64s(f, v);
 794 }
 795
 796 const VMStateInfo vmstate_info_int64 = {
 797     .name = "int64",
 798     .get  = get_int64,
 799     .put  = put_int64,
 800 };
 801
 802 /* 8 bit unsigned int */
 803
 804 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 805 {
 806     uint8_t *v = pv;
 807     qemu_get_8s(f, v);
 808     return 0;
 809 }
 810
 811 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 812 {
 813     uint8_t *v = pv;
 814     qemu_put_8s(f, v);
 815 }
 816
 817 const VMStateInfo vmstate_info_uint8 = {
 818     .name = "uint8",
 819     .get  = get_uint8,
 820     .put  = put_uint8,
 821 };
 822
 823 /* 16 bit unsigned int */
 824
 825 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 826 {
 827     uint16_t *v = pv;
 828     qemu_get_be16s(f, v);
 829     return 0;
 830 }
 831
 832 static void put_uint16(QEMUFile *f, void *pv, size_t size)
 833 {
 834     uint16_t *v = pv;
 835     qemu_put_be16s(f, v);
 836 }
 837
 838 const VMStateInfo vmstate_info_uint16 = {
 839     .name = "uint16",
 840     .get  = get_uint16,
 841     .put  = put_uint16,
 842 };
 843
 844 /* 32 bit unsigned int */
 845
 846 static int get_uint32(QEMUFile *f, void *pv, size_t size)
 847 {
 848     uint32_t *v = pv;
 849     qemu_get_be32s(f, v);
 850     return 0;
 851 }
 852
 853 static void put_uint32(QEMUFile *f, void *pv, size_t size)
 854 {
 855     uint32_t *v = pv;
 856     qemu_put_be32s(f, v);
 857 }
 858
 859 const VMStateInfo vmstate_info_uint32 = {
 860     .name = "uint32",
 861     .get  = get_uint32,
 862     .put  = put_uint32,
 863 };
 864
 865 /* 64 bit unsigned int */
 866
 867 static int get_uint64(QEMUFile *f, void *pv, size_t size)
 868 {
 869     uint64_t *v = pv;
 870     qemu_get_be64s(f, v);
 871     return 0;
 872 }
 873
 874 static void put_uint64(QEMUFile *f, void *pv, size_t size)
 875 {
 876     uint64_t *v = pv;
 877     qemu_put_be64s(f, v);
 878 }
 879
 880 const VMStateInfo vmstate_info_uint64 = {
 881     .name = "uint64",
 882     .get  = get_uint64,
 883     .put  = put_uint64,
 884 };
 885
 886 /* 64 bit linux kernel unsigned int */
 887
 888 #ifdef __linux__
 889 static int get_u64(QEMUFile *f, void *pv, size_t size)
 890 {
 891     uint64_t *v = pv;
 892     qemu_get_be64s(f, v);
 893     return 0;
 894 }
 895
 896 static void put_u64(QEMUFile *f, void *pv, size_t size)
 897 {
 898     uint64_t *v = pv;
 899     qemu_put_be64s(f, v);
 900 }
 901
 902 const VMStateInfo vmstate_info_u64 = {
 903     .name = "__u64",
 904     .get  = get_u64,
 905     .put  = put_u64,
 906 };
 907 #endif /* __linux__ */
 908
 909 /* 8 bit int. See that the received value is the same than the one
 910    in the field */
 911
 912 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
 913 {
 914     uint8_t *v = pv;
 915     uint8_t v2;
 916     qemu_get_8s(f, &v2);
 917
 918     if (*v == v2)
 919         return 0;
 920     return -EINVAL;
 921 }
 922
 923 const VMStateInfo vmstate_info_uint8_equal = {
 924     .name = "uint8 equal",
 925     .get  = get_uint8_equal,
 926     .put  = put_uint8,
 927 };
 928
 929 /* 16 bit unsigned int int. See that the received value is the same than the one
 930    in the field */
 931
 932 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
 933 {
 934     uint16_t *v = pv;
 935     uint16_t v2;
 936     qemu_get_be16s(f, &v2);
 937
 938     if (*v == v2)
 939         return 0;
 940     return -EINVAL;
 941 }
 942
 943 const VMStateInfo vmstate_info_uint16_equal = {
 944     .name = "uint16 equal",
 945     .get  = get_uint16_equal,
 946     .put  = put_uint16,
 947 };
 948
 949 /* timers  */
 950
 951 static int get_timer(QEMUFile *f, void *pv, size_t size)
 952 {
 953     QEMUTimer *v = pv;
 954     qemu_get_timer(f, v);
 955     return 0;
 956 }
 957
 958 static void put_timer(QEMUFile *f, void *pv, size_t size)
 959 {
 960     QEMUTimer *v = pv;
 961     qemu_put_timer(f, v);
 962 }
 963
 964 const VMStateInfo vmstate_info_timer = {
 965     .name = "timer",
 966     .get  = get_timer,
 967     .put  = put_timer,
 968 };
 969
 970 /* uint8_t buffers */
 971
 972 static int get_buffer(QEMUFile *f, void *pv, size_t size)
 973 {
 974     uint8_t *v = pv;
 975     qemu_get_buffer(f, v, size);
 976     return 0;
 977 }
 978
 979 static void put_buffer(QEMUFile *f, void *pv, size_t size)
 980 {
 981     uint8_t *v = pv;
 982     qemu_put_buffer(f, v, size);
 983 }
 984
 985 const VMStateInfo vmstate_info_buffer = {
 986     .name = "buffer",
 987     .get  = get_buffer,
 988     .put  = put_buffer,
 989 };
 990
 991 /* unused buffers: space that was used for some fields that are
 992    not usefull anymore */
 993
 994 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
 995 {
 996     uint8_t buf[1024];
 997     int block_len;
 998
 999     while (size > 0) {
1000         block_len = MIN(sizeof(buf), size);
1001         size -= block_len;
1002         qemu_get_buffer(f, buf, block_len);
1003     }
1004    return 0;
1005 }
1006
1007 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1008 {
1009     static const uint8_t buf[1024];
1010     int block_len;
1011
1012     while (size > 0) {
1013         block_len = MIN(sizeof(buf), size);
1014         size -= block_len;
1015         qemu_put_buffer(f, buf, block_len);
1016     }
1017 }
1018
1019 const VMStateInfo vmstate_info_unused_buffer = {
1020     .name = "unused_buffer",
1021     .get  = get_unused_buffer,
1022     .put  = put_unused_buffer,
1023 };
1024
1025 typedef struct CompatEntry {
1026     char idstr[256];
1027     int instance_id;
1028 } CompatEntry;
1029
1030 typedef struct SaveStateEntry {
1031     QTAILQ_ENTRY(SaveStateEntry) entry;
1032     char idstr[256];
1033     int instance_id;
1034     int alias_id;
1035     int version_id;
1036     int section_id;
1037     SaveSetParamsHandler *set_params;
1038     SaveLiveStateHandler *save_live_state;
1039     SaveStateHandler *save_state;
1040     LoadStateHandler *load_state;
1041     const VMStateDescription *vmsd;
1042     void *opaque;
1043     CompatEntry *compat;
1044     int no_migrate;
1045 } SaveStateEntry;
1046
1047
1048 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1049     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1050 static int global_section_id;
1051
1052 static int calculate_new_instance_id(const char *idstr)
1053 {
1054     SaveStateEntry *se;
1055     int instance_id = 0;
1056
1057     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1058         if (strcmp(idstr, se->idstr) == 0
1059             && instance_id <= se->instance_id) {
1060             instance_id = se->instance_id + 1;
1061         }
1062     }
1063     return instance_id;
1064 }
1065
1066 static int calculate_compat_instance_id(const char *idstr)
1067 {
1068     SaveStateEntry *se;
1069     int instance_id = 0;
1070
1071     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1072         if (!se->compat)
1073             continue;
1074
1075         if (strcmp(idstr, se->compat->idstr) == 0
1076             && instance_id <= se->compat->instance_id) {
1077             instance_id = se->compat->instance_id + 1;
1078         }
1079     }
1080     return instance_id;
1081 }
1082
1083 /* TODO: Individual devices generally have very little idea about the rest
1084    of the system, so instance_id should be removed/replaced.
1085    Meanwhile pass -1 as instance_id if you do not already have a clearly
1086    distinguishing id for all instances of your device class. */
1087 int register_savevm_live(DeviceState *dev,
1088                          const char *idstr,
1089                          int instance_id,
1090                          int version_id,
1091                          SaveSetParamsHandler *set_params,
1092                          SaveLiveStateHandler *save_live_state,
1093                          SaveStateHandler *save_state,
1094                          LoadStateHandler *load_state,
1095                          void *opaque)
1096 {
1097     SaveStateEntry *se;
1098
1099     se = qemu_mallocz(sizeof(SaveStateEntry));
1100     se->version_id = version_id;
1101     se->section_id = global_section_id++;
1102     se->set_params = set_params;
1103     se->save_live_state = save_live_state;
1104     se->save_state = save_state;
1105     se->load_state = load_state;
1106     se->opaque = opaque;
1107     se->vmsd = NULL;
1108     se->no_migrate = 0;
1109
1110     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1111         char *id = dev->parent_bus->info->get_dev_path(dev);
1112         if (id) {
1113             pstrcpy(se->idstr, sizeof(se->idstr), id);
1114             pstrcat(se->idstr, sizeof(se->idstr), "/");
1115             qemu_free(id);
1116
1117             se->compat = qemu_mallocz(sizeof(CompatEntry));
1118             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1119             se->compat->instance_id = instance_id == -1 ?
1120                          calculate_compat_instance_id(idstr) : instance_id;
1121             instance_id = -1;
1122         }
1123     }
1124     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1125
1126     if (instance_id == -1) {
1127         se->instance_id = calculate_new_instance_id(se->idstr);
1128     } else {
1129         se->instance_id = instance_id;
1130     }
1131     assert(!se->compat || se->instance_id == 0);
1132     /* add at the end of list */
1133     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1134     return 0;
1135 }
1136
1137 int register_savevm(DeviceState *dev,
1138                     const char *idstr,
1139                     int instance_id,
1140                     int version_id,
1141                     SaveStateHandler *save_state,
1142                     LoadStateHandler *load_state,
1143                     void *opaque)
1144 {
1145     return register_savevm_live(dev, idstr, instance_id, version_id,
1146                                 NULL, NULL, save_state, load_state, opaque);
1147 }
1148
1149 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1150 {
1151     SaveStateEntry *se, *new_se;
1152     char id[256] = "";
1153
1154     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1155         char *path = dev->parent_bus->info->get_dev_path(dev);
1156         if (path) {
1157             pstrcpy(id, sizeof(id), path);
1158             pstrcat(id, sizeof(id), "/");
1159             qemu_free(path);
1160         }
1161     }
1162     pstrcat(id, sizeof(id), idstr);
1163
1164     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1165         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1166             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1167             if (se->compat) {
1168                 qemu_free(se->compat);
1169             }
1170             qemu_free(se);
1171         }
1172     }
1173 }
1174
1175 /* mark a device as not to be migrated, that is the device should be
1176    unplugged before migration */
1177 void register_device_unmigratable(DeviceState *dev, const char *idstr,
1178                                                             void *opaque)
1179 {
1180     SaveStateEntry *se;
1181     char id[256] = "";
1182
1183     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1184         char *path = dev->parent_bus->info->get_dev_path(dev);
1185         if (path) {
1186             pstrcpy(id, sizeof(id), path);
1187             pstrcat(id, sizeof(id), "/");
1188             qemu_free(path);
1189         }
1190     }
1191     pstrcat(id, sizeof(id), idstr);
1192
1193     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1194         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1195             se->no_migrate = 1;
1196         }
1197     }
1198 }
1199
1200 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1201                                    const VMStateDescription *vmsd,
1202                                    void *opaque, int alias_id,
1203                                    int required_for_version)
1204 {
1205     SaveStateEntry *se;
1206
1207     /* If this triggers, alias support can be dropped for the vmsd. */
1208     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1209
1210     se = qemu_mallocz(sizeof(SaveStateEntry));
1211     se->version_id = vmsd->version_id;
1212     se->section_id = global_section_id++;
1213     se->save_live_state = NULL;
1214     se->save_state = NULL;
1215     se->load_state = NULL;
1216     se->opaque = opaque;
1217     se->vmsd = vmsd;
1218     se->alias_id = alias_id;
1219
1220     if (dev && dev->parent_bus && dev->parent_bus->info->get_dev_path) {
1221         char *id = dev->parent_bus->info->get_dev_path(dev);
1222         if (id) {
1223             pstrcpy(se->idstr, sizeof(se->idstr), id);
1224             pstrcat(se->idstr, sizeof(se->idstr), "/");
1225             qemu_free(id);
1226
1227             se->compat = qemu_mallocz(sizeof(CompatEntry));
1228             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1229             se->compat->instance_id = instance_id == -1 ?
1230                          calculate_compat_instance_id(vmsd->name) : instance_id;
1231             instance_id = -1;
1232         }
1233     }
1234     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1235
1236     if (instance_id == -1) {
1237         se->instance_id = calculate_new_instance_id(se->idstr);
1238     } else {
1239         se->instance_id = instance_id;
1240     }
1241     assert(!se->compat || se->instance_id == 0);
1242     /* add at the end of list */
1243     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1244     return 0;
1245 }
1246
1247 int vmstate_register(DeviceState *dev, int instance_id,
1248                      const VMStateDescription *vmsd, void *opaque)
1249 {
1250     return vmstate_register_with_alias_id(dev, instance_id, vmsd,
1251                                           opaque, -1, 0);
1252 }
1253
1254 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1255                         void *opaque)
1256 {
1257     SaveStateEntry *se, *new_se;
1258
1259     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1260         if (se->vmsd == vmsd && se->opaque == opaque) {
1261             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1262             if (se->compat) {
1263                 qemu_free(se->compat);
1264             }
1265             qemu_free(se);
1266         }
1267     }
1268 }
1269
1270 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1271                                     void *opaque);
1272 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1273                                    void *opaque);
1274
1275 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1276                        void *opaque, int version_id)
1277 {
1278     VMStateField *field = vmsd->fields;
1279     int ret;
1280
1281     if (version_id > vmsd->version_id) {
1282         return -EINVAL;
1283     }
1284     if (version_id < vmsd->minimum_version_id_old) {
1285         return -EINVAL;
1286     }
1287     if  (version_id < vmsd->minimum_version_id) {
1288         return vmsd->load_state_old(f, opaque, version_id);
1289     }
1290     if (vmsd->pre_load) {
1291         int ret = vmsd->pre_load(opaque);
1292         if (ret)
1293             return ret;
1294     }
1295     while(field->name) {
1296         if ((field->field_exists &&
1297              field->field_exists(opaque, version_id)) ||
1298             (!field->field_exists &&
1299              field->version_id <= version_id)) {
1300             void *base_addr = opaque + field->offset;
1301             int i, n_elems = 1;
1302             int size = field->size;
1303
1304             if (field->flags & VMS_VBUFFER) {
1305                 size = *(int32_t *)(opaque+field->size_offset);
1306                 if (field->flags & VMS_MULTIPLY) {
1307                     size *= field->size;
1308                 }
1309             }
1310             if (field->flags & VMS_ARRAY) {
1311                 n_elems = field->num;
1312             } else if (field->flags & VMS_VARRAY_INT32) {
1313                 n_elems = *(int32_t *)(opaque+field->num_offset);
1314             } else if (field->flags & VMS_VARRAY_UINT16) {
1315                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1316             }
1317             if (field->flags & VMS_POINTER) {
1318                 base_addr = *(void **)base_addr + field->start;
1319             }
1320             for (i = 0; i < n_elems; i++) {
1321                 void *addr = base_addr + size * i;
1322
1323                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1324                     addr = *(void **)addr;
1325                 }
1326                 if (field->flags & VMS_STRUCT) {
1327                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1328                 } else {
1329                     ret = field->info->get(f, addr, size);
1330
1331                 }
1332                 if (ret < 0) {
1333                     return ret;
1334                 }
1335             }
1336         }
1337         field++;
1338     }
1339     ret = vmstate_subsection_load(f, vmsd, opaque);
1340     if (ret != 0) {
1341         return ret;
1342     }
1343     if (vmsd->post_load) {
1344         return vmsd->post_load(opaque, version_id);
1345     }
1346     return 0;
1347 }
1348
1349 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1350                         void *opaque)
1351 {
1352     VMStateField *field = vmsd->fields;
1353
1354     if (vmsd->pre_save) {
1355         vmsd->pre_save(opaque);
1356     }
1357     while(field->name) {
1358         if (!field->field_exists ||
1359             field->field_exists(opaque, vmsd->version_id)) {
1360             void *base_addr = opaque + field->offset;
1361             int i, n_elems = 1;
1362             int size = field->size;
1363
1364             if (field->flags & VMS_VBUFFER) {
1365                 size = *(int32_t *)(opaque+field->size_offset);
1366                 if (field->flags & VMS_MULTIPLY) {
1367                     size *= field->size;
1368                 }
1369             }
1370             if (field->flags & VMS_ARRAY) {
1371                 n_elems = field->num;
1372             } else if (field->flags & VMS_VARRAY_INT32) {
1373                 n_elems = *(int32_t *)(opaque+field->num_offset);
1374             } else if (field->flags & VMS_VARRAY_UINT16) {
1375                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1376             }
1377             if (field->flags & VMS_POINTER) {
1378                 base_addr = *(void **)base_addr + field->start;
1379             }
1380             for (i = 0; i < n_elems; i++) {
1381                 void *addr = base_addr + size * i;
1382
1383                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1384                     addr = *(void **)addr;
1385                 }
1386                 if (field->flags & VMS_STRUCT) {
1387                     vmstate_save_state(f, field->vmsd, addr);
1388                 } else {
1389                     field->info->put(f, addr, size);
1390                 }
1391             }
1392         }
1393         field++;
1394     }
1395     vmstate_subsection_save(f, vmsd, opaque);
1396 }
1397
1398 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1399 {
1400     if (!se->vmsd) {         /* Old style */
1401         return se->load_state(f, se->opaque, version_id);
1402     }
1403     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1404 }
1405
1406 static int vmstate_save(QEMUFile *f, SaveStateEntry *se)
1407 {
1408     if (se->no_migrate) {
1409         return -1;
1410     }
1411
1412     if (!se->vmsd) {         /* Old style */
1413         se->save_state(f, se->opaque);
1414         return 0;
1415     }
1416     vmstate_save_state(f,se->vmsd, se->opaque);
1417
1418     return 0;
1419 }
1420
1421 #define QEMU_VM_FILE_MAGIC           0x5145564d
1422 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1423 #define QEMU_VM_FILE_VERSION         0x00000003
1424
1425 #define QEMU_VM_EOF                  0x00
1426 #define QEMU_VM_SECTION_START        0x01
1427 #define QEMU_VM_SECTION_PART         0x02
1428 #define QEMU_VM_SECTION_END          0x03
1429 #define QEMU_VM_SECTION_FULL         0x04
1430 #define QEMU_VM_SUBSECTION           0x05
1431
1432 int qemu_savevm_state_begin(Monitor *mon, QEMUFile *f, int blk_enable,
1433                             int shared)
1434 {
1435     SaveStateEntry *se;
1436
1437     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1438         if(se->set_params == NULL) {
1439             continue;
1440         }
1441         se->set_params(blk_enable, shared, se->opaque);
1442     }
1443
1444     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1445     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1446
1447     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1448         int len;
1449
1450         if (se->save_live_state == NULL)
1451             continue;
1452
1453         /* Section type */
1454         qemu_put_byte(f, QEMU_VM_SECTION_START);
1455         qemu_put_be32(f, se->section_id);
1456
1457         /* ID string */
1458         len = strlen(se->idstr);
1459         qemu_put_byte(f, len);
1460         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1461
1462         qemu_put_be32(f, se->instance_id);
1463         qemu_put_be32(f, se->version_id);
1464
1465         se->save_live_state(mon, f, QEMU_VM_SECTION_START, se->opaque);
1466     }
1467
1468     if (qemu_file_has_error(f)) {
1469         qemu_savevm_state_cancel(mon, f);
1470         return -EIO;
1471     }
1472
1473     return 0;
1474 }
1475
1476 int qemu_savevm_state_iterate(Monitor *mon, QEMUFile *f)
1477 {
1478     SaveStateEntry *se;
1479     int ret = 1;
1480
1481     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1482         if (se->save_live_state == NULL)
1483             continue;
1484
1485         /* Section type */
1486         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1487         qemu_put_be32(f, se->section_id);
1488
1489         ret = se->save_live_state(mon, f, QEMU_VM_SECTION_PART, se->opaque);
1490         if (!ret) {
1491             /* Do not proceed to the next vmstate before this one reported
1492                completion of the current stage. This serializes the migration
1493                and reduces the probability that a faster changing state is
1494                synchronized over and over again. */
1495             break;
1496         }
1497     }
1498
1499     if (ret)
1500         return 1;
1501
1502     if (qemu_file_has_error(f)) {
1503         qemu_savevm_state_cancel(mon, f);
1504         return -EIO;
1505     }
1506
1507     return 0;
1508 }
1509
1510 int qemu_savevm_state_complete(Monitor *mon, QEMUFile *f)
1511 {
1512     SaveStateEntry *se;
1513     int r;
1514
1515     cpu_synchronize_all_states();
1516
1517     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1518         if (se->save_live_state == NULL)
1519             continue;
1520
1521         /* Section type */
1522         qemu_put_byte(f, QEMU_VM_SECTION_END);
1523         qemu_put_be32(f, se->section_id);
1524
1525         se->save_live_state(mon, f, QEMU_VM_SECTION_END, se->opaque);
1526     }
1527
1528     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1529         int len;
1530
1531         if (se->save_state == NULL && se->vmsd == NULL)
1532             continue;
1533
1534         /* Section type */
1535         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1536         qemu_put_be32(f, se->section_id);
1537
1538         /* ID string */
1539         len = strlen(se->idstr);
1540         qemu_put_byte(f, len);
1541         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1542
1543         qemu_put_be32(f, se->instance_id);
1544         qemu_put_be32(f, se->version_id);
1545
1546         r = vmstate_save(f, se);
1547         if (r < 0) {
1548             monitor_printf(mon, "cannot migrate with device '%s'\n", se->idstr);
1549             return r;
1550         }
1551     }
1552
1553     qemu_put_byte(f, QEMU_VM_EOF);
1554
1555     if (qemu_file_has_error(f))
1556         return -EIO;
1557
1558     return 0;
1559 }
1560
1561 void qemu_savevm_state_cancel(Monitor *mon, QEMUFile *f)
1562 {
1563     SaveStateEntry *se;
1564
1565     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1566         if (se->save_live_state) {
1567             se->save_live_state(mon, f, -1, se->opaque);
1568         }
1569     }
1570 }
1571
1572 static int qemu_savevm_state(Monitor *mon, QEMUFile *f)
1573 {
1574     int saved_vm_running;
1575     int ret;
1576
1577     saved_vm_running = vm_running;
1578     vm_stop(0);
1579
1580     bdrv_flush_all();
1581
1582     ret = qemu_savevm_state_begin(mon, f, 0, 0);
1583     if (ret < 0)
1584         goto out;
1585
1586     do {
1587         ret = qemu_savevm_state_iterate(mon, f);
1588         if (ret < 0)
1589             goto out;
1590     } while (ret == 0);
1591
1592     ret = qemu_savevm_state_complete(mon, f);
1593
1594 out:
1595     if (qemu_file_has_error(f))
1596         ret = -EIO;
1597
1598     if (!ret && saved_vm_running)
1599         vm_start();
1600
1601     return ret;
1602 }
1603
1604 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1605 {
1606     SaveStateEntry *se;
1607
1608     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1609         if (!strcmp(se->idstr, idstr) &&
1610             (instance_id == se->instance_id ||
1611              instance_id == se->alias_id))
1612             return se;
1613         /* Migrating from an older version? */
1614         if (strstr(se->idstr, idstr) && se->compat) {
1615             if (!strcmp(se->compat->idstr, idstr) &&
1616                 (instance_id == se->compat->instance_id ||
1617                  instance_id == se->alias_id))
1618                 return se;
1619         }
1620     }
1621     return NULL;
1622 }
1623
1624 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1625 {
1626     while(sub && sub->needed) {
1627         if (strcmp(idstr, sub->vmsd->name) == 0) {
1628             return sub->vmsd;
1629         }
1630         sub++;
1631     }
1632     return NULL;
1633 }
1634
1635 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1636                                    void *opaque)
1637 {
1638     while (qemu_peek_byte(f) == QEMU_VM_SUBSECTION) {
1639         char idstr[256];
1640         int ret;
1641         uint8_t version_id, len;
1642         const VMStateDescription *sub_vmsd;
1643
1644         qemu_get_byte(f); /* subsection */
1645         len = qemu_get_byte(f);
1646         qemu_get_buffer(f, (uint8_t *)idstr, len);
1647         idstr[len] = 0;
1648         version_id = qemu_get_be32(f);
1649
1650         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
1651         if (sub_vmsd == NULL) {
1652             return -ENOENT;
1653         }
1654         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1655         if (ret) {
1656             return ret;
1657         }
1658     }
1659     return 0;
1660 }
1661
1662 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1663                                     void *opaque)
1664 {
1665     const VMStateSubsection *sub = vmsd->subsections;
1666
1667     while (sub && sub->needed) {
1668         if (sub->needed(opaque)) {
1669             const VMStateDescription *vmsd = sub->vmsd;
1670             uint8_t len;
1671
1672             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1673             len = strlen(vmsd->name);
1674             qemu_put_byte(f, len);
1675             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1676             qemu_put_be32(f, vmsd->version_id);
1677             vmstate_save_state(f, vmsd, opaque);
1678         }
1679         sub++;
1680     }
1681 }
1682
1683 typedef struct LoadStateEntry {
1684     QLIST_ENTRY(LoadStateEntry) entry;
1685     SaveStateEntry *se;
1686     int section_id;
1687     int version_id;
1688 } LoadStateEntry;
1689
1690 int qemu_loadvm_state(QEMUFile *f)
1691 {
1692     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1693         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1694     LoadStateEntry *le, *new_le;
1695     uint8_t section_type;
1696     unsigned int v;
1697     int ret;
1698
1699     v = qemu_get_be32(f);
1700     if (v != QEMU_VM_FILE_MAGIC)
1701         return -EINVAL;
1702
1703     v = qemu_get_be32(f);
1704     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1705         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1706         return -ENOTSUP;
1707     }
1708     if (v != QEMU_VM_FILE_VERSION)
1709         return -ENOTSUP;
1710
1711     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1712         uint32_t instance_id, version_id, section_id;
1713         SaveStateEntry *se;
1714         char idstr[257];
1715         int len;
1716
1717         switch (section_type) {
1718         case QEMU_VM_SECTION_START:
1719         case QEMU_VM_SECTION_FULL:
1720             /* Read section start */
1721             section_id = qemu_get_be32(f);
1722             len = qemu_get_byte(f);
1723             qemu_get_buffer(f, (uint8_t *)idstr, len);
1724             idstr[len] = 0;
1725             instance_id = qemu_get_be32(f);
1726             version_id = qemu_get_be32(f);
1727
1728             /* Find savevm section */
1729             se = find_se(idstr, instance_id);
1730             if (se == NULL) {
1731                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
1732                 ret = -EINVAL;
1733                 goto out;
1734             }
1735
1736             /* Validate version */
1737             if (version_id > se->version_id) {
1738                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
1739                         version_id, idstr, se->version_id);
1740                 ret = -EINVAL;
1741                 goto out;
1742             }
1743
1744             /* Add entry */
1745             le = qemu_mallocz(sizeof(*le));
1746
1747             le->se = se;
1748             le->section_id = section_id;
1749             le->version_id = version_id;
1750             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
1751
1752             ret = vmstate_load(f, le->se, le->version_id);
1753             if (ret < 0) {
1754                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
1755                         instance_id, idstr);
1756                 goto out;
1757             }
1758             break;
1759         case QEMU_VM_SECTION_PART:
1760         case QEMU_VM_SECTION_END:
1761             section_id = qemu_get_be32(f);
1762
1763             QLIST_FOREACH(le, &loadvm_handlers, entry) {
1764                 if (le->section_id == section_id) {
1765                     break;
1766                 }
1767             }
1768             if (le == NULL) {
1769                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
1770                 ret = -EINVAL;
1771                 goto out;
1772             }
1773
1774             ret = vmstate_load(f, le->se, le->version_id);
1775             if (ret < 0) {
1776                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
1777                         section_id);
1778                 goto out;
1779             }
1780             break;
1781         default:
1782             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
1783             ret = -EINVAL;
1784             goto out;
1785         }
1786     }
1787
1788     cpu_synchronize_all_post_init();
1789
1790     ret = 0;
1791
1792 out:
1793     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
1794         QLIST_REMOVE(le, entry);
1795         qemu_free(le);
1796     }
1797
1798     if (qemu_file_has_error(f))
1799         ret = -EIO;
1800
1801     return ret;
1802 }
1803
1804 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1805                               const char *name)
1806 {
1807     QEMUSnapshotInfo *sn_tab, *sn;
1808     int nb_sns, i, ret;
1809
1810     ret = -ENOENT;
1811     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1812     if (nb_sns < 0)
1813         return ret;
1814     for(i = 0; i < nb_sns; i++) {
1815         sn = &sn_tab[i];
1816         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1817             *sn_info = *sn;
1818             ret = 0;
1819             break;
1820         }
1821     }
1822     qemu_free(sn_tab);
1823     return ret;
1824 }
1825
1826 /*
1827  * Deletes snapshots of a given name in all opened images.
1828  */
1829 static int del_existing_snapshots(Monitor *mon, const char *name)
1830 {
1831     BlockDriverState *bs;
1832     QEMUSnapshotInfo sn1, *snapshot = &sn1;
1833     int ret;
1834
1835     bs = NULL;
1836     while ((bs = bdrv_next(bs))) {
1837         if (bdrv_can_snapshot(bs) &&
1838             bdrv_snapshot_find(bs, snapshot, name) >= 0)
1839         {
1840             ret = bdrv_snapshot_delete(bs, name);
1841             if (ret < 0) {
1842                 monitor_printf(mon,
1843                                "Error while deleting snapshot on '%s'\n",
1844                                bdrv_get_device_name(bs));
1845                 return -1;
1846             }
1847         }
1848     }
1849
1850     return 0;
1851 }
1852
1853 void do_savevm(Monitor *mon, const QDict *qdict)
1854 {
1855     BlockDriverState *bs, *bs1;
1856     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1857     int ret;
1858     QEMUFile *f;
1859     int saved_vm_running;
1860     uint32_t vm_state_size;
1861 #ifdef _WIN32
1862     struct _timeb tb;
1863     struct tm *ptm;
1864 #else
1865     struct timeval tv;
1866     struct tm tm;
1867 #endif
1868     const char *name = qdict_get_try_str(qdict, "name");
1869
1870     /* Verify if there is a device that doesn't support snapshots and is writable */
1871     bs = NULL;
1872     while ((bs = bdrv_next(bs))) {
1873
1874         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1875             continue;
1876         }
1877
1878         if (!bdrv_can_snapshot(bs)) {
1879             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
1880                                bdrv_get_device_name(bs));
1881             return;
1882         }
1883     }
1884
1885     bs = bdrv_snapshots();
1886     if (!bs) {
1887         monitor_printf(mon, "No block device can accept snapshots\n");
1888         return;
1889     }
1890     /* ??? Should this occur after vm_stop?  */
1891     qemu_aio_flush();
1892
1893     saved_vm_running = vm_running;
1894     vm_stop(0);
1895
1896     memset(sn, 0, sizeof(*sn));
1897
1898     /* fill auxiliary fields */
1899 #ifdef _WIN32
1900     _ftime(&tb);
1901     sn->date_sec = tb.time;
1902     sn->date_nsec = tb.millitm * 1000000;
1903 #else
1904     gettimeofday(&tv, NULL);
1905     sn->date_sec = tv.tv_sec;
1906     sn->date_nsec = tv.tv_usec * 1000;
1907 #endif
1908     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1909
1910     if (name) {
1911         ret = bdrv_snapshot_find(bs, old_sn, name);
1912         if (ret >= 0) {
1913             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1914             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1915         } else {
1916             pstrcpy(sn->name, sizeof(sn->name), name);
1917         }
1918     } else {
1919 #ifdef _WIN32
1920         ptm = localtime(&tb.time);
1921         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", ptm);
1922 #else
1923         /* cast below needed for OpenBSD where tv_sec is still 'long' */
1924         localtime_r((const time_t *)&tv.tv_sec, &tm);
1925         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
1926 #endif
1927     }
1928
1929     /* Delete old snapshots of the same name */
1930     if (name && del_existing_snapshots(mon, name) < 0) {
1931         goto the_end;
1932     }
1933
1934     /* save the VM state */
1935     f = qemu_fopen_bdrv(bs, 1);
1936     if (!f) {
1937         monitor_printf(mon, "Could not open VM state file\n");
1938         goto the_end;
1939     }
1940     ret = qemu_savevm_state(mon, f);
1941     vm_state_size = qemu_ftell(f);
1942     qemu_fclose(f);
1943     if (ret < 0) {
1944         monitor_printf(mon, "Error %d while writing VM\n", ret);
1945         goto the_end;
1946     }
1947
1948     /* create the snapshots */
1949
1950     bs1 = NULL;
1951     while ((bs1 = bdrv_next(bs1))) {
1952         if (bdrv_can_snapshot(bs1)) {
1953             /* Write VM state size only to the image that contains the state */
1954             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1955             ret = bdrv_snapshot_create(bs1, sn);
1956             if (ret < 0) {
1957                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1958                                bdrv_get_device_name(bs1));
1959             }
1960         }
1961     }
1962
1963  the_end:
1964     if (saved_vm_running)
1965         vm_start();
1966 }
1967
1968 int load_vmstate(const char *name)
1969 {
1970     BlockDriverState *bs, *bs_vm_state;
1971     QEMUSnapshotInfo sn;
1972     QEMUFile *f;
1973     int ret;
1974
1975     bs_vm_state = bdrv_snapshots();
1976     if (!bs_vm_state) {
1977         error_report("No block device supports snapshots");
1978         return -ENOTSUP;
1979     }
1980
1981     /* Don't even try to load empty VM states */
1982     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
1983     if (ret < 0) {
1984         return ret;
1985     } else if (sn.vm_state_size == 0) {
1986         return -EINVAL;
1987     }
1988
1989     /* Verify if there is any device that doesn't support snapshots and is
1990     writable and check if the requested snapshot is available too. */
1991     bs = NULL;
1992     while ((bs = bdrv_next(bs))) {
1993
1994         if (bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1995             continue;
1996         }
1997
1998         if (!bdrv_can_snapshot(bs)) {
1999             error_report("Device '%s' is writable but does not support snapshots.",
2000                                bdrv_get_device_name(bs));
2001             return -ENOTSUP;
2002         }
2003
2004         ret = bdrv_snapshot_find(bs, &sn, name);
2005         if (ret < 0) {
2006             error_report("Device '%s' does not have the requested snapshot '%s'",
2007                            bdrv_get_device_name(bs), name);
2008             return ret;
2009         }
2010     }
2011
2012     /* Flush all IO requests so they don't interfere with the new state.  */
2013     qemu_aio_flush();
2014
2015     bs = NULL;
2016     while ((bs = bdrv_next(bs))) {
2017         if (bdrv_can_snapshot(bs)) {
2018             ret = bdrv_snapshot_goto(bs, name);
2019             if (ret < 0) {
2020                 error_report("Error %d while activating snapshot '%s' on '%s'",
2021                              ret, name, bdrv_get_device_name(bs));
2022                 return ret;
2023             }
2024         }
2025     }
2026
2027     /* restore the VM state */
2028     f = qemu_fopen_bdrv(bs_vm_state, 0);
2029     if (!f) {
2030         error_report("Could not open VM state file");
2031         return -EINVAL;
2032     }
2033
2034     ret = qemu_loadvm_state(f);
2035
2036     qemu_fclose(f);
2037     if (ret < 0) {
2038         error_report("Error %d while loading VM state", ret);
2039         return ret;
2040     }
2041
2042     return 0;
2043 }
2044
2045 void do_delvm(Monitor *mon, const QDict *qdict)
2046 {
2047     BlockDriverState *bs, *bs1;
2048     int ret;
2049     const char *name = qdict_get_str(qdict, "name");
2050
2051     bs = bdrv_snapshots();
2052     if (!bs) {
2053         monitor_printf(mon, "No block device supports snapshots\n");
2054         return;
2055     }
2056
2057     bs1 = NULL;
2058     while ((bs1 = bdrv_next(bs1))) {
2059         if (bdrv_can_snapshot(bs1)) {
2060             ret = bdrv_snapshot_delete(bs1, name);
2061             if (ret < 0) {
2062                 if (ret == -ENOTSUP)
2063                     monitor_printf(mon,
2064                                    "Snapshots not supported on device '%s'\n",
2065                                    bdrv_get_device_name(bs1));
2066                 else
2067                     monitor_printf(mon, "Error %d while deleting snapshot on "
2068                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2069             }
2070         }
2071     }
2072 }
2073
2074 void do_info_snapshots(Monitor *mon)
2075 {
2076     BlockDriverState *bs, *bs1;
2077     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2078     int nb_sns, i, ret, available;
2079     int total;
2080     int *available_snapshots;
2081     char buf[256];
2082
2083     bs = bdrv_snapshots();
2084     if (!bs) {
2085         monitor_printf(mon, "No available block device supports snapshots\n");
2086         return;
2087     }
2088
2089     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2090     if (nb_sns < 0) {
2091         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2092         return;
2093     }
2094
2095     if (nb_sns == 0) {
2096         monitor_printf(mon, "There is no snapshot available.\n");
2097         return;
2098     }
2099
2100     available_snapshots = qemu_mallocz(sizeof(int) * nb_sns);
2101     total = 0;
2102     for (i = 0; i < nb_sns; i++) {
2103         sn = &sn_tab[i];
2104         available = 1;
2105         bs1 = NULL;
2106
2107         while ((bs1 = bdrv_next(bs1))) {
2108             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2109                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2110                 if (ret < 0) {
2111                     available = 0;
2112                     break;
2113                 }
2114             }
2115         }
2116
2117         if (available) {
2118             available_snapshots[total] = i;
2119             total++;
2120         }
2121     }
2122
2123     if (total > 0) {
2124         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2125         for (i = 0; i < total; i++) {
2126             sn = &sn_tab[available_snapshots[i]];
2127             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2128         }
2129     } else {
2130         monitor_printf(mon, "There is no suitable snapshot available\n");
2131     }
2132
2133     qemu_free(sn_tab);
2134     qemu_free(available_snapshots);
2135
2136 }