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
  25 #include "config-host.h"
  26 #include "qemu-common.h"
  27 #include "hw/hw.h"
  28 #include "hw/qdev.h"
  29 #include "net/net.h"
  30 #include "monitor/monitor.h"
  31 #include "sysemu/sysemu.h"
  32 #include "qemu/timer.h"
  33 #include "audio/audio.h"
  34 #include "migration/migration.h"
  35 #include "qemu/sockets.h"
  36 #include "qemu/queue.h"
  37 #include "sysemu/cpus.h"
  38 #include "exec/memory.h"
  39 #include "qmp-commands.h"
  40 #include "trace.h"
  41 #include "qemu/bitops.h"
  42
  43 #define SELF_ANNOUNCE_ROUNDS 5
  44
  45 #ifndef ETH_P_RARP
  46 #define ETH_P_RARP 0x8035
  47 #endif
  48 #define ARP_HTYPE_ETH 0x0001
  49 #define ARP_PTYPE_IP 0x0800
  50 #define ARP_OP_REQUEST_REV 0x3
  51
  52 static int announce_self_create(uint8_t *buf,
  53                                 uint8_t *mac_addr)
  54 {
  55     /* Ethernet header. */
  56     memset(buf, 0xff, 6);         /* destination MAC addr */
  57     memcpy(buf + 6, mac_addr, 6); /* source MAC addr */
  58     *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */
  59
  60     /* RARP header. */
  61     *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */
  62     *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */
  63     *(buf + 18) = 6; /* hardware addr length (ethernet) */
  64     *(buf + 19) = 4; /* protocol addr length (IPv4) */
  65     *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */
  66     memcpy(buf + 22, mac_addr, 6); /* source hw addr */
  67     memset(buf + 28, 0x00, 4);     /* source protocol addr */
  68     memcpy(buf + 32, mac_addr, 6); /* target hw addr */
  69     memset(buf + 38, 0x00, 4);     /* target protocol addr */
  70
  71     /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */
  72     memset(buf + 42, 0x00, 18);
  73
  74     return 60; /* len (FCS will be added by hardware) */
  75 }
  76
  77 static void qemu_announce_self_iter(NICState *nic, void *opaque)
  78 {
  79     uint8_t buf[60];
  80     int len;
  81
  82     len = announce_self_create(buf, nic->conf->macaddr.a);
  83
  84     qemu_send_packet_raw(qemu_get_queue(nic), buf, len);
  85 }
  86
  87
  88 static void qemu_announce_self_once(void *opaque)
  89 {
  90     static int count = SELF_ANNOUNCE_ROUNDS;
  91     QEMUTimer *timer = *(QEMUTimer **)opaque;
  92
  93     qemu_foreach_nic(qemu_announce_self_iter, NULL);
  94
  95     if (--count) {
  96         /* delay 50ms, 150ms, 250ms, ... */
  97         qemu_mod_timer(timer, qemu_get_clock_ms(rt_clock) +
  98                        50 + (SELF_ANNOUNCE_ROUNDS - count - 1) * 100);
  99     } else {
 100             qemu_del_timer(timer);
 101             qemu_free_timer(timer);
 102     }
 103 }
 104
 105 void qemu_announce_self(void)
 106 {
 107         static QEMUTimer *timer;
 108         timer = qemu_new_timer_ms(rt_clock, qemu_announce_self_once, &timer);
 109         qemu_announce_self_once(&timer);
 110 }
 111
 112 /***********************************************************/
 113 /* savevm/loadvm support */
 114
 115 #define IO_BUF_SIZE 32768
 116
 117 struct QEMUFile {
 118     const QEMUFileOps *ops;
 119     void *opaque;
 120     int is_write;
 121
 122     int64_t bytes_xfer;
 123     int64_t xfer_limit;
 124
 125     int64_t pos; /* start of buffer when writing, end of buffer
 126                     when reading */
 127     int buf_index;
 128     int buf_size; /* 0 when writing */
 129     uint8_t buf[IO_BUF_SIZE];
 130
 131     int last_error;
 132 };
 133
 134 typedef struct QEMUFileStdio
 135 {
 136     FILE *stdio_file;
 137     QEMUFile *file;
 138 } QEMUFileStdio;
 139
 140 typedef struct QEMUFileSocket
 141 {
 142     int fd;
 143     QEMUFile *file;
 144 } QEMUFileSocket;
 145
 146 typedef struct {
 147     Coroutine *co;
 148     int fd;
 149 } FDYieldUntilData;
 150
 151 static void fd_coroutine_enter(void *opaque)
 152 {
 153     FDYieldUntilData *data = opaque;
 154     qemu_set_fd_handler(data->fd, NULL, NULL, NULL);
 155     qemu_coroutine_enter(data->co, NULL);
 156 }
 157
 158 /**
 159  * Yield until a file descriptor becomes readable
 160  *
 161  * Note that this function clobbers the handlers for the file descriptor.
 162  */
 163 static void coroutine_fn yield_until_fd_readable(int fd)
 164 {
 165     FDYieldUntilData data;
 166
 167     assert(qemu_in_coroutine());
 168     data.co = qemu_coroutine_self();
 169     data.fd = fd;
 170     qemu_set_fd_handler(fd, fd_coroutine_enter, NULL, &data);
 171     qemu_coroutine_yield();
 172 }
 173
 174 static int socket_get_fd(void *opaque)
 175 {
 176     QEMUFileSocket *s = opaque;
 177
 178     return s->fd;
 179 }
 180
 181 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 182 {
 183     QEMUFileSocket *s = opaque;
 184     ssize_t len;
 185
 186     for (;;) {
 187         len = qemu_recv(s->fd, buf, size, 0);
 188         if (len != -1) {
 189             break;
 190         }
 191         if (socket_error() == EAGAIN) {
 192             yield_until_fd_readable(s->fd);
 193         } else if (socket_error() != EINTR) {
 194             break;
 195         }
 196     }
 197
 198     if (len == -1) {
 199         len = -socket_error();
 200     }
 201     return len;
 202 }
 203
 204 static int socket_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 205 {
 206     QEMUFileSocket *s = opaque;
 207     ssize_t len;
 208
 209     len = qemu_send_full(s->fd, buf, size, 0);
 210     if (len < size) {
 211         len = -socket_error();
 212     }
 213     return len;
 214 }
 215
 216 static int socket_close(void *opaque)
 217 {
 218     QEMUFileSocket *s = opaque;
 219     closesocket(s->fd);
 220     g_free(s);
 221     return 0;
 222 }
 223
 224 static int stdio_get_fd(void *opaque)
 225 {
 226     QEMUFileStdio *s = opaque;
 227
 228     return fileno(s->stdio_file);
 229 }
 230
 231 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 232 {
 233     QEMUFileStdio *s = opaque;
 234     return fwrite(buf, 1, size, s->stdio_file);
 235 }
 236
 237 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 238 {
 239     QEMUFileStdio *s = opaque;
 240     FILE *fp = s->stdio_file;
 241     int bytes;
 242
 243     for (;;) {
 244         clearerr(fp);
 245         bytes = fread(buf, 1, size, fp);
 246         if (bytes != 0 || !ferror(fp)) {
 247             break;
 248         }
 249         if (errno == EAGAIN) {
 250             yield_until_fd_readable(fileno(fp));
 251         } else if (errno != EINTR) {
 252             break;
 253         }
 254     }
 255     return bytes;
 256 }
 257
 258 static int stdio_pclose(void *opaque)
 259 {
 260     QEMUFileStdio *s = opaque;
 261     int ret;
 262     ret = pclose(s->stdio_file);
 263     if (ret == -1) {
 264         ret = -errno;
 265     } else if (!WIFEXITED(ret) || WEXITSTATUS(ret) != 0) {
 266         /* close succeeded, but non-zero exit code: */
 267         ret = -EIO; /* fake errno value */
 268     }
 269     g_free(s);
 270     return ret;
 271 }
 272
 273 static int stdio_fclose(void *opaque)
 274 {
 275     QEMUFileStdio *s = opaque;
 276     int ret = 0;
 277
 278     if (s->file->ops->put_buffer) {
 279         int fd = fileno(s->stdio_file);
 280         struct stat st;
 281
 282         ret = fstat(fd, &st);
 283         if (ret == 0 && S_ISREG(st.st_mode)) {
 284             /*
 285              * If the file handle is a regular file make sure the
 286              * data is flushed to disk before signaling success.
 287              */
 288             ret = fsync(fd);
 289             if (ret != 0) {
 290                 ret = -errno;
 291                 return ret;
 292             }
 293         }
 294     }
 295     if (fclose(s->stdio_file) == EOF) {
 296         ret = -errno;
 297     }
 298     g_free(s);
 299     return ret;
 300 }
 301
 302 static const QEMUFileOps stdio_pipe_read_ops = {
 303     .get_fd =     stdio_get_fd,
 304     .get_buffer = stdio_get_buffer,
 305     .close =      stdio_pclose
 306 };
 307
 308 static const QEMUFileOps stdio_pipe_write_ops = {
 309     .get_fd =     stdio_get_fd,
 310     .put_buffer = stdio_put_buffer,
 311     .close =      stdio_pclose
 312 };
 313
 314 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 315 {
 316     FILE *stdio_file;
 317     QEMUFileStdio *s;
 318
 319     stdio_file = popen(command, mode);
 320     if (stdio_file == NULL) {
 321         return NULL;
 322     }
 323
 324     if (mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 325         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 326         return NULL;
 327     }
 328
 329     s = g_malloc0(sizeof(QEMUFileStdio));
 330
 331     s->stdio_file = stdio_file;
 332
 333     if(mode[0] == 'r') {
 334         s->file = qemu_fopen_ops(s, &stdio_pipe_read_ops);
 335     } else {
 336         s->file = qemu_fopen_ops(s, &stdio_pipe_write_ops);
 337     }
 338     return s->file;
 339 }
 340
 341 static const QEMUFileOps stdio_file_read_ops = {
 342     .get_fd =     stdio_get_fd,
 343     .get_buffer = stdio_get_buffer,
 344     .close =      stdio_fclose
 345 };
 346
 347 static const QEMUFileOps stdio_file_write_ops = {
 348     .get_fd =     stdio_get_fd,
 349     .put_buffer = stdio_put_buffer,
 350     .close =      stdio_fclose
 351 };
 352
 353 QEMUFile *qemu_fdopen(int fd, const char *mode)
 354 {
 355     QEMUFileStdio *s;
 356
 357     if (mode == NULL ||
 358         (mode[0] != 'r' && mode[0] != 'w') ||
 359         mode[1] != 'b' || mode[2] != 0) {
 360         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 361         return NULL;
 362     }
 363
 364     s = g_malloc0(sizeof(QEMUFileStdio));
 365     s->stdio_file = fdopen(fd, mode);
 366     if (!s->stdio_file)
 367         goto fail;
 368
 369     if(mode[0] == 'r') {
 370         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 371     } else {
 372         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 373     }
 374     return s->file;
 375
 376 fail:
 377     g_free(s);
 378     return NULL;
 379 }
 380
 381 static const QEMUFileOps socket_read_ops = {
 382     .get_fd =     socket_get_fd,
 383     .get_buffer = socket_get_buffer,
 384     .close =      socket_close
 385 };
 386
 387 static const QEMUFileOps socket_write_ops = {
 388     .get_fd =     socket_get_fd,
 389     .put_buffer = socket_put_buffer,
 390     .close =      socket_close
 391 };
 392
 393 QEMUFile *qemu_fopen_socket(int fd, const char *mode)
 394 {
 395     QEMUFileSocket *s = g_malloc0(sizeof(QEMUFileSocket));
 396
 397     if (mode == NULL ||
 398         (mode[0] != 'r' && mode[0] != 'w') ||
 399         mode[1] != 'b' || mode[2] != 0) {
 400         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 401         return NULL;
 402     }
 403
 404     s->fd = fd;
 405     if (mode[0] == 'w') {
 406         socket_set_block(s->fd);
 407         s->file = qemu_fopen_ops(s, &socket_write_ops);
 408     } else {
 409         s->file = qemu_fopen_ops(s, &socket_read_ops);
 410     }
 411     return s->file;
 412 }
 413
 414 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 415 {
 416     QEMUFileStdio *s;
 417
 418     if (mode == NULL ||
 419         (mode[0] != 'r' && mode[0] != 'w') ||
 420         mode[1] != 'b' || mode[2] != 0) {
 421         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
 422         return NULL;
 423     }
 424
 425     s = g_malloc0(sizeof(QEMUFileStdio));
 426
 427     s->stdio_file = fopen(filename, mode);
 428     if (!s->stdio_file)
 429         goto fail;
 430
 431     if(mode[0] == 'w') {
 432         s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
 433     } else {
 434         s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
 435     }
 436     return s->file;
 437 fail:
 438     g_free(s);
 439     return NULL;
 440 }
 441
 442 static int block_put_buffer(void *opaque, const uint8_t *buf,
 443                            int64_t pos, int size)
 444 {
 445     bdrv_save_vmstate(opaque, buf, pos, size);
 446     return size;
 447 }
 448
 449 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 450 {
 451     return bdrv_load_vmstate(opaque, buf, pos, size);
 452 }
 453
 454 static int bdrv_fclose(void *opaque)
 455 {
 456     return bdrv_flush(opaque);
 457 }
 458
 459 static const QEMUFileOps bdrv_read_ops = {
 460     .get_buffer = block_get_buffer,
 461     .close =      bdrv_fclose
 462 };
 463
 464 static const QEMUFileOps bdrv_write_ops = {
 465     .put_buffer = block_put_buffer,
 466     .close =      bdrv_fclose
 467 };
 468
 469 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 470 {
 471     if (is_writable)
 472         return qemu_fopen_ops(bs, &bdrv_write_ops);
 473     return qemu_fopen_ops(bs, &bdrv_read_ops);
 474 }
 475
 476 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
 477 {
 478     QEMUFile *f;
 479
 480     f = g_malloc0(sizeof(QEMUFile));
 481
 482     f->opaque = opaque;
 483     f->ops = ops;
 484     f->is_write = 0;
 485     return f;
 486 }
 487
 488 int qemu_file_get_error(QEMUFile *f)
 489 {
 490     return f->last_error;
 491 }
 492
 493 static void qemu_file_set_error(QEMUFile *f, int ret)
 494 {
 495     if (f->last_error == 0) {
 496         f->last_error = ret;
 497     }
 498 }
 499
 500 /** Flushes QEMUFile buffer
 501  *
 502  */
 503 static void qemu_fflush(QEMUFile *f)
 504 {
 505     int ret = 0;
 506
 507     if (!f->ops->put_buffer) {
 508         return;
 509     }
 510     if (f->is_write && f->buf_index > 0) {
 511         ret = f->ops->put_buffer(f->opaque, f->buf, f->pos, f->buf_index);
 512         if (ret >= 0) {
 513             f->pos += f->buf_index;
 514         }
 515         f->buf_index = 0;
 516     }
 517     if (ret < 0) {
 518         qemu_file_set_error(f, ret);
 519     }
 520 }
 521
 522 static void qemu_fill_buffer(QEMUFile *f)
 523 {
 524     int len;
 525     int pending;
 526
 527     if (!f->ops->get_buffer)
 528         return;
 529
 530     if (f->is_write)
 531         abort();
 532
 533     pending = f->buf_size - f->buf_index;
 534     if (pending > 0) {
 535         memmove(f->buf, f->buf + f->buf_index, pending);
 536     }
 537     f->buf_index = 0;
 538     f->buf_size = pending;
 539
 540     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->pos,
 541                         IO_BUF_SIZE - pending);
 542     if (len > 0) {
 543         f->buf_size += len;
 544         f->pos += len;
 545     } else if (len == 0) {
 546         qemu_file_set_error(f, -EIO);
 547     } else if (len != -EAGAIN)
 548         qemu_file_set_error(f, len);
 549 }
 550
 551 int qemu_get_fd(QEMUFile *f)
 552 {
 553     if (f->ops->get_fd) {
 554         return f->ops->get_fd(f->opaque);
 555     }
 556     return -1;
 557 }
 558
 559 /** Closes the file
 560  *
 561  * Returns negative error value if any error happened on previous operations or
 562  * while closing the file. Returns 0 or positive number on success.
 563  *
 564  * The meaning of return value on success depends on the specific backend
 565  * being used.
 566  */
 567 int qemu_fclose(QEMUFile *f)
 568 {
 569     int ret;
 570     qemu_fflush(f);
 571     ret = qemu_file_get_error(f);
 572
 573     if (f->ops->close) {
 574         int ret2 = f->ops->close(f->opaque);
 575         if (ret >= 0) {
 576             ret = ret2;
 577         }
 578     }
 579     /* If any error was spotted before closing, we should report it
 580      * instead of the close() return value.
 581      */
 582     if (f->last_error) {
 583         ret = f->last_error;
 584     }
 585     g_free(f);
 586     return ret;
 587 }
 588
 589 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 590 {
 591     int l;
 592
 593     if (f->last_error) {
 594         return;
 595     }
 596
 597     if (f->is_write == 0 && f->buf_index > 0) {
 598         fprintf(stderr,
 599                 "Attempted to write to buffer while read buffer is not empty\n");
 600         abort();
 601     }
 602
 603     while (size > 0) {
 604         l = IO_BUF_SIZE - f->buf_index;
 605         if (l > size)
 606             l = size;
 607         memcpy(f->buf + f->buf_index, buf, l);
 608         f->is_write = 1;
 609         f->buf_index += l;
 610         f->bytes_xfer += l;
 611         buf += l;
 612         size -= l;
 613         if (f->buf_index >= IO_BUF_SIZE) {
 614             qemu_fflush(f);
 615             if (qemu_file_get_error(f)) {
 616                 break;
 617             }
 618         }
 619     }
 620 }
 621
 622 void qemu_put_byte(QEMUFile *f, int v)
 623 {
 624     if (f->last_error) {
 625         return;
 626     }
 627
 628     if (f->is_write == 0 && f->buf_index > 0) {
 629         fprintf(stderr,
 630                 "Attempted to write to buffer while read buffer is not empty\n");
 631         abort();
 632     }
 633
 634     f->buf[f->buf_index++] = v;
 635     f->is_write = 1;
 636     if (f->buf_index >= IO_BUF_SIZE) {
 637         qemu_fflush(f);
 638     }
 639 }
 640
 641 static void qemu_file_skip(QEMUFile *f, int size)
 642 {
 643     if (f->buf_index + size <= f->buf_size) {
 644         f->buf_index += size;
 645     }
 646 }
 647
 648 static int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset)
 649 {
 650     int pending;
 651     int index;
 652
 653     if (f->is_write) {
 654         abort();
 655     }
 656
 657     index = f->buf_index + offset;
 658     pending = f->buf_size - index;
 659     if (pending < size) {
 660         qemu_fill_buffer(f);
 661         index = f->buf_index + offset;
 662         pending = f->buf_size - index;
 663     }
 664
 665     if (pending <= 0) {
 666         return 0;
 667     }
 668     if (size > pending) {
 669         size = pending;
 670     }
 671
 672     memcpy(buf, f->buf + index, size);
 673     return size;
 674 }
 675
 676 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size)
 677 {
 678     int pending = size;
 679     int done = 0;
 680
 681     while (pending > 0) {
 682         int res;
 683
 684         res = qemu_peek_buffer(f, buf, pending, 0);
 685         if (res == 0) {
 686             return done;
 687         }
 688         qemu_file_skip(f, res);
 689         buf += res;
 690         pending -= res;
 691         done += res;
 692     }
 693     return done;
 694 }
 695
 696 static int qemu_peek_byte(QEMUFile *f, int offset)
 697 {
 698     int index = f->buf_index + offset;
 699
 700     if (f->is_write) {
 701         abort();
 702     }
 703
 704     if (index >= f->buf_size) {
 705         qemu_fill_buffer(f);
 706         index = f->buf_index + offset;
 707         if (index >= f->buf_size) {
 708             return 0;
 709         }
 710     }
 711     return f->buf[index];
 712 }
 713
 714 int qemu_get_byte(QEMUFile *f)
 715 {
 716     int result;
 717
 718     result = qemu_peek_byte(f, 0);
 719     qemu_file_skip(f, 1);
 720     return result;
 721 }
 722
 723 int64_t qemu_ftell(QEMUFile *f)
 724 {
 725     qemu_fflush(f);
 726     return f->pos;
 727 }
 728
 729 int qemu_file_rate_limit(QEMUFile *f)
 730 {
 731     if (qemu_file_get_error(f)) {
 732         return 1;
 733     }
 734     if (f->xfer_limit > 0 && f->bytes_xfer > f->xfer_limit) {
 735         return 1;
 736     }
 737     return 0;
 738 }
 739
 740 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 741 {
 742     return f->xfer_limit;
 743 }
 744
 745 void qemu_file_set_rate_limit(QEMUFile *f, int64_t limit)
 746 {
 747     f->xfer_limit = limit;
 748 }
 749
 750 void qemu_file_reset_rate_limit(QEMUFile *f)
 751 {
 752     f->bytes_xfer = 0;
 753 }
 754
 755 void qemu_put_be16(QEMUFile *f, unsigned int v)
 756 {
 757     qemu_put_byte(f, v >> 8);
 758     qemu_put_byte(f, v);
 759 }
 760
 761 void qemu_put_be32(QEMUFile *f, unsigned int v)
 762 {
 763     qemu_put_byte(f, v >> 24);
 764     qemu_put_byte(f, v >> 16);
 765     qemu_put_byte(f, v >> 8);
 766     qemu_put_byte(f, v);
 767 }
 768
 769 void qemu_put_be64(QEMUFile *f, uint64_t v)
 770 {
 771     qemu_put_be32(f, v >> 32);
 772     qemu_put_be32(f, v);
 773 }
 774
 775 unsigned int qemu_get_be16(QEMUFile *f)
 776 {
 777     unsigned int v;
 778     v = qemu_get_byte(f) << 8;
 779     v |= qemu_get_byte(f);
 780     return v;
 781 }
 782
 783 unsigned int qemu_get_be32(QEMUFile *f)
 784 {
 785     unsigned int v;
 786     v = qemu_get_byte(f) << 24;
 787     v |= qemu_get_byte(f) << 16;
 788     v |= qemu_get_byte(f) << 8;
 789     v |= qemu_get_byte(f);
 790     return v;
 791 }
 792
 793 uint64_t qemu_get_be64(QEMUFile *f)
 794 {
 795     uint64_t v;
 796     v = (uint64_t)qemu_get_be32(f) << 32;
 797     v |= qemu_get_be32(f);
 798     return v;
 799 }
 800
 801
 802 /* timer */
 803
 804 void qemu_put_timer(QEMUFile *f, QEMUTimer *ts)
 805 {
 806     uint64_t expire_time;
 807
 808     expire_time = qemu_timer_expire_time_ns(ts);
 809     qemu_put_be64(f, expire_time);
 810 }
 811
 812 void qemu_get_timer(QEMUFile *f, QEMUTimer *ts)
 813 {
 814     uint64_t expire_time;
 815
 816     expire_time = qemu_get_be64(f);
 817     if (expire_time != -1) {
 818         qemu_mod_timer_ns(ts, expire_time);
 819     } else {
 820         qemu_del_timer(ts);
 821     }
 822 }
 823
 824
 825 /* bool */
 826
 827 static int get_bool(QEMUFile *f, void *pv, size_t size)
 828 {
 829     bool *v = pv;
 830     *v = qemu_get_byte(f);
 831     return 0;
 832 }
 833
 834 static void put_bool(QEMUFile *f, void *pv, size_t size)
 835 {
 836     bool *v = pv;
 837     qemu_put_byte(f, *v);
 838 }
 839
 840 const VMStateInfo vmstate_info_bool = {
 841     .name = "bool",
 842     .get  = get_bool,
 843     .put  = put_bool,
 844 };
 845
 846 /* 8 bit int */
 847
 848 static int get_int8(QEMUFile *f, void *pv, size_t size)
 849 {
 850     int8_t *v = pv;
 851     qemu_get_s8s(f, v);
 852     return 0;
 853 }
 854
 855 static void put_int8(QEMUFile *f, void *pv, size_t size)
 856 {
 857     int8_t *v = pv;
 858     qemu_put_s8s(f, v);
 859 }
 860
 861 const VMStateInfo vmstate_info_int8 = {
 862     .name = "int8",
 863     .get  = get_int8,
 864     .put  = put_int8,
 865 };
 866
 867 /* 16 bit int */
 868
 869 static int get_int16(QEMUFile *f, void *pv, size_t size)
 870 {
 871     int16_t *v = pv;
 872     qemu_get_sbe16s(f, v);
 873     return 0;
 874 }
 875
 876 static void put_int16(QEMUFile *f, void *pv, size_t size)
 877 {
 878     int16_t *v = pv;
 879     qemu_put_sbe16s(f, v);
 880 }
 881
 882 const VMStateInfo vmstate_info_int16 = {
 883     .name = "int16",
 884     .get  = get_int16,
 885     .put  = put_int16,
 886 };
 887
 888 /* 32 bit int */
 889
 890 static int get_int32(QEMUFile *f, void *pv, size_t size)
 891 {
 892     int32_t *v = pv;
 893     qemu_get_sbe32s(f, v);
 894     return 0;
 895 }
 896
 897 static void put_int32(QEMUFile *f, void *pv, size_t size)
 898 {
 899     int32_t *v = pv;
 900     qemu_put_sbe32s(f, v);
 901 }
 902
 903 const VMStateInfo vmstate_info_int32 = {
 904     .name = "int32",
 905     .get  = get_int32,
 906     .put  = put_int32,
 907 };
 908
 909 /* 32 bit int. See that the received value is the same than the one
 910    in the field */
 911
 912 static int get_int32_equal(QEMUFile *f, void *pv, size_t size)
 913 {
 914     int32_t *v = pv;
 915     int32_t v2;
 916     qemu_get_sbe32s(f, &v2);
 917
 918     if (*v == v2)
 919         return 0;
 920     return -EINVAL;
 921 }
 922
 923 const VMStateInfo vmstate_info_int32_equal = {
 924     .name = "int32 equal",
 925     .get  = get_int32_equal,
 926     .put  = put_int32,
 927 };
 928
 929 /* 32 bit int. See that the received value is the less or the same
 930    than the one in the field */
 931
 932 static int get_int32_le(QEMUFile *f, void *pv, size_t size)
 933 {
 934     int32_t *old = pv;
 935     int32_t new;
 936     qemu_get_sbe32s(f, &new);
 937
 938     if (*old <= new)
 939         return 0;
 940     return -EINVAL;
 941 }
 942
 943 const VMStateInfo vmstate_info_int32_le = {
 944     .name = "int32 equal",
 945     .get  = get_int32_le,
 946     .put  = put_int32,
 947 };
 948
 949 /* 64 bit int */
 950
 951 static int get_int64(QEMUFile *f, void *pv, size_t size)
 952 {
 953     int64_t *v = pv;
 954     qemu_get_sbe64s(f, v);
 955     return 0;
 956 }
 957
 958 static void put_int64(QEMUFile *f, void *pv, size_t size)
 959 {
 960     int64_t *v = pv;
 961     qemu_put_sbe64s(f, v);
 962 }
 963
 964 const VMStateInfo vmstate_info_int64 = {
 965     .name = "int64",
 966     .get  = get_int64,
 967     .put  = put_int64,
 968 };
 969
 970 /* 8 bit unsigned int */
 971
 972 static int get_uint8(QEMUFile *f, void *pv, size_t size)
 973 {
 974     uint8_t *v = pv;
 975     qemu_get_8s(f, v);
 976     return 0;
 977 }
 978
 979 static void put_uint8(QEMUFile *f, void *pv, size_t size)
 980 {
 981     uint8_t *v = pv;
 982     qemu_put_8s(f, v);
 983 }
 984
 985 const VMStateInfo vmstate_info_uint8 = {
 986     .name = "uint8",
 987     .get  = get_uint8,
 988     .put  = put_uint8,
 989 };
 990
 991 /* 16 bit unsigned int */
 992
 993 static int get_uint16(QEMUFile *f, void *pv, size_t size)
 994 {
 995     uint16_t *v = pv;
 996     qemu_get_be16s(f, v);
 997     return 0;
 998 }
 999
1000 static void put_uint16(QEMUFile *f, void *pv, size_t size)
1001 {
1002     uint16_t *v = pv;
1003     qemu_put_be16s(f, v);
1004 }
1005
1006 const VMStateInfo vmstate_info_uint16 = {
1007     .name = "uint16",
1008     .get  = get_uint16,
1009     .put  = put_uint16,
1010 };
1011
1012 /* 32 bit unsigned int */
1013
1014 static int get_uint32(QEMUFile *f, void *pv, size_t size)
1015 {
1016     uint32_t *v = pv;
1017     qemu_get_be32s(f, v);
1018     return 0;
1019 }
1020
1021 static void put_uint32(QEMUFile *f, void *pv, size_t size)
1022 {
1023     uint32_t *v = pv;
1024     qemu_put_be32s(f, v);
1025 }
1026
1027 const VMStateInfo vmstate_info_uint32 = {
1028     .name = "uint32",
1029     .get  = get_uint32,
1030     .put  = put_uint32,
1031 };
1032
1033 /* 32 bit uint. See that the received value is the same than the one
1034    in the field */
1035
1036 static int get_uint32_equal(QEMUFile *f, void *pv, size_t size)
1037 {
1038     uint32_t *v = pv;
1039     uint32_t v2;
1040     qemu_get_be32s(f, &v2);
1041
1042     if (*v == v2) {
1043         return 0;
1044     }
1045     return -EINVAL;
1046 }
1047
1048 const VMStateInfo vmstate_info_uint32_equal = {
1049     .name = "uint32 equal",
1050     .get  = get_uint32_equal,
1051     .put  = put_uint32,
1052 };
1053
1054 /* 64 bit unsigned int */
1055
1056 static int get_uint64(QEMUFile *f, void *pv, size_t size)
1057 {
1058     uint64_t *v = pv;
1059     qemu_get_be64s(f, v);
1060     return 0;
1061 }
1062
1063 static void put_uint64(QEMUFile *f, void *pv, size_t size)
1064 {
1065     uint64_t *v = pv;
1066     qemu_put_be64s(f, v);
1067 }
1068
1069 const VMStateInfo vmstate_info_uint64 = {
1070     .name = "uint64",
1071     .get  = get_uint64,
1072     .put  = put_uint64,
1073 };
1074
1075 /* 8 bit int. See that the received value is the same than the one
1076    in the field */
1077
1078 static int get_uint8_equal(QEMUFile *f, void *pv, size_t size)
1079 {
1080     uint8_t *v = pv;
1081     uint8_t v2;
1082     qemu_get_8s(f, &v2);
1083
1084     if (*v == v2)
1085         return 0;
1086     return -EINVAL;
1087 }
1088
1089 const VMStateInfo vmstate_info_uint8_equal = {
1090     .name = "uint8 equal",
1091     .get  = get_uint8_equal,
1092     .put  = put_uint8,
1093 };
1094
1095 /* 16 bit unsigned int int. See that the received value is the same than the one
1096    in the field */
1097
1098 static int get_uint16_equal(QEMUFile *f, void *pv, size_t size)
1099 {
1100     uint16_t *v = pv;
1101     uint16_t v2;
1102     qemu_get_be16s(f, &v2);
1103
1104     if (*v == v2)
1105         return 0;
1106     return -EINVAL;
1107 }
1108
1109 const VMStateInfo vmstate_info_uint16_equal = {
1110     .name = "uint16 equal",
1111     .get  = get_uint16_equal,
1112     .put  = put_uint16,
1113 };
1114
1115 /* timers  */
1116
1117 static int get_timer(QEMUFile *f, void *pv, size_t size)
1118 {
1119     QEMUTimer *v = pv;
1120     qemu_get_timer(f, v);
1121     return 0;
1122 }
1123
1124 static void put_timer(QEMUFile *f, void *pv, size_t size)
1125 {
1126     QEMUTimer *v = pv;
1127     qemu_put_timer(f, v);
1128 }
1129
1130 const VMStateInfo vmstate_info_timer = {
1131     .name = "timer",
1132     .get  = get_timer,
1133     .put  = put_timer,
1134 };
1135
1136 /* uint8_t buffers */
1137
1138 static int get_buffer(QEMUFile *f, void *pv, size_t size)
1139 {
1140     uint8_t *v = pv;
1141     qemu_get_buffer(f, v, size);
1142     return 0;
1143 }
1144
1145 static void put_buffer(QEMUFile *f, void *pv, size_t size)
1146 {
1147     uint8_t *v = pv;
1148     qemu_put_buffer(f, v, size);
1149 }
1150
1151 const VMStateInfo vmstate_info_buffer = {
1152     .name = "buffer",
1153     .get  = get_buffer,
1154     .put  = put_buffer,
1155 };
1156
1157 /* unused buffers: space that was used for some fields that are
1158    not useful anymore */
1159
1160 static int get_unused_buffer(QEMUFile *f, void *pv, size_t size)
1161 {
1162     uint8_t buf[1024];
1163     int block_len;
1164
1165     while (size > 0) {
1166         block_len = MIN(sizeof(buf), size);
1167         size -= block_len;
1168         qemu_get_buffer(f, buf, block_len);
1169     }
1170    return 0;
1171 }
1172
1173 static void put_unused_buffer(QEMUFile *f, void *pv, size_t size)
1174 {
1175     static const uint8_t buf[1024];
1176     int block_len;
1177
1178     while (size > 0) {
1179         block_len = MIN(sizeof(buf), size);
1180         size -= block_len;
1181         qemu_put_buffer(f, buf, block_len);
1182     }
1183 }
1184
1185 const VMStateInfo vmstate_info_unused_buffer = {
1186     .name = "unused_buffer",
1187     .get  = get_unused_buffer,
1188     .put  = put_unused_buffer,
1189 };
1190
1191 /* bitmaps (as defined by bitmap.h). Note that size here is the size
1192  * of the bitmap in bits. The on-the-wire format of a bitmap is 64
1193  * bit words with the bits in big endian order. The in-memory format
1194  * is an array of 'unsigned long', which may be either 32 or 64 bits.
1195  */
1196 /* This is the number of 64 bit words sent over the wire */
1197 #define BITS_TO_U64S(nr) DIV_ROUND_UP(nr, 64)
1198 static int get_bitmap(QEMUFile *f, void *pv, size_t size)
1199 {
1200     unsigned long *bmp = pv;
1201     int i, idx = 0;
1202     for (i = 0; i < BITS_TO_U64S(size); i++) {
1203         uint64_t w = qemu_get_be64(f);
1204         bmp[idx++] = w;
1205         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1206             bmp[idx++] = w >> 32;
1207         }
1208     }
1209     return 0;
1210 }
1211
1212 static void put_bitmap(QEMUFile *f, void *pv, size_t size)
1213 {
1214     unsigned long *bmp = pv;
1215     int i, idx = 0;
1216     for (i = 0; i < BITS_TO_U64S(size); i++) {
1217         uint64_t w = bmp[idx++];
1218         if (sizeof(unsigned long) == 4 && idx < BITS_TO_LONGS(size)) {
1219             w |= ((uint64_t)bmp[idx++]) << 32;
1220         }
1221         qemu_put_be64(f, w);
1222     }
1223 }
1224
1225 const VMStateInfo vmstate_info_bitmap = {
1226     .name = "bitmap",
1227     .get = get_bitmap,
1228     .put = put_bitmap,
1229 };
1230
1231 typedef struct CompatEntry {
1232     char idstr[256];
1233     int instance_id;
1234 } CompatEntry;
1235
1236 typedef struct SaveStateEntry {
1237     QTAILQ_ENTRY(SaveStateEntry) entry;
1238     char idstr[256];
1239     int instance_id;
1240     int alias_id;
1241     int version_id;
1242     int section_id;
1243     SaveVMHandlers *ops;
1244     const VMStateDescription *vmsd;
1245     void *opaque;
1246     CompatEntry *compat;
1247     int no_migrate;
1248     int is_ram;
1249 } SaveStateEntry;
1250
1251
1252 static QTAILQ_HEAD(savevm_handlers, SaveStateEntry) savevm_handlers =
1253     QTAILQ_HEAD_INITIALIZER(savevm_handlers);
1254 static int global_section_id;
1255
1256 static int calculate_new_instance_id(const char *idstr)
1257 {
1258     SaveStateEntry *se;
1259     int instance_id = 0;
1260
1261     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1262         if (strcmp(idstr, se->idstr) == 0
1263             && instance_id <= se->instance_id) {
1264             instance_id = se->instance_id + 1;
1265         }
1266     }
1267     return instance_id;
1268 }
1269
1270 static int calculate_compat_instance_id(const char *idstr)
1271 {
1272     SaveStateEntry *se;
1273     int instance_id = 0;
1274
1275     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1276         if (!se->compat)
1277             continue;
1278
1279         if (strcmp(idstr, se->compat->idstr) == 0
1280             && instance_id <= se->compat->instance_id) {
1281             instance_id = se->compat->instance_id + 1;
1282         }
1283     }
1284     return instance_id;
1285 }
1286
1287 /* TODO: Individual devices generally have very little idea about the rest
1288    of the system, so instance_id should be removed/replaced.
1289    Meanwhile pass -1 as instance_id if you do not already have a clearly
1290    distinguishing id for all instances of your device class. */
1291 int register_savevm_live(DeviceState *dev,
1292                          const char *idstr,
1293                          int instance_id,
1294                          int version_id,
1295                          SaveVMHandlers *ops,
1296                          void *opaque)
1297 {
1298     SaveStateEntry *se;
1299
1300     se = g_malloc0(sizeof(SaveStateEntry));
1301     se->version_id = version_id;
1302     se->section_id = global_section_id++;
1303     se->ops = ops;
1304     se->opaque = opaque;
1305     se->vmsd = NULL;
1306     se->no_migrate = 0;
1307     /* if this is a live_savem then set is_ram */
1308     if (ops->save_live_setup != NULL) {
1309         se->is_ram = 1;
1310     }
1311
1312     if (dev) {
1313         char *id = qdev_get_dev_path(dev);
1314         if (id) {
1315             pstrcpy(se->idstr, sizeof(se->idstr), id);
1316             pstrcat(se->idstr, sizeof(se->idstr), "/");
1317             g_free(id);
1318
1319             se->compat = g_malloc0(sizeof(CompatEntry));
1320             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr);
1321             se->compat->instance_id = instance_id == -1 ?
1322                          calculate_compat_instance_id(idstr) : instance_id;
1323             instance_id = -1;
1324         }
1325     }
1326     pstrcat(se->idstr, sizeof(se->idstr), idstr);
1327
1328     if (instance_id == -1) {
1329         se->instance_id = calculate_new_instance_id(se->idstr);
1330     } else {
1331         se->instance_id = instance_id;
1332     }
1333     assert(!se->compat || se->instance_id == 0);
1334     /* add at the end of list */
1335     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1336     return 0;
1337 }
1338
1339 int register_savevm(DeviceState *dev,
1340                     const char *idstr,
1341                     int instance_id,
1342                     int version_id,
1343                     SaveStateHandler *save_state,
1344                     LoadStateHandler *load_state,
1345                     void *opaque)
1346 {
1347     SaveVMHandlers *ops = g_malloc0(sizeof(SaveVMHandlers));
1348     ops->save_state = save_state;
1349     ops->load_state = load_state;
1350     return register_savevm_live(dev, idstr, instance_id, version_id,
1351                                 ops, opaque);
1352 }
1353
1354 void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque)
1355 {
1356     SaveStateEntry *se, *new_se;
1357     char id[256] = "";
1358
1359     if (dev) {
1360         char *path = qdev_get_dev_path(dev);
1361         if (path) {
1362             pstrcpy(id, sizeof(id), path);
1363             pstrcat(id, sizeof(id), "/");
1364             g_free(path);
1365         }
1366     }
1367     pstrcat(id, sizeof(id), idstr);
1368
1369     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1370         if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) {
1371             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1372             if (se->compat) {
1373                 g_free(se->compat);
1374             }
1375             g_free(se->ops);
1376             g_free(se);
1377         }
1378     }
1379 }
1380
1381 int vmstate_register_with_alias_id(DeviceState *dev, int instance_id,
1382                                    const VMStateDescription *vmsd,
1383                                    void *opaque, int alias_id,
1384                                    int required_for_version)
1385 {
1386     SaveStateEntry *se;
1387
1388     /* If this triggers, alias support can be dropped for the vmsd. */
1389     assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id);
1390
1391     se = g_malloc0(sizeof(SaveStateEntry));
1392     se->version_id = vmsd->version_id;
1393     se->section_id = global_section_id++;
1394     se->opaque = opaque;
1395     se->vmsd = vmsd;
1396     se->alias_id = alias_id;
1397     se->no_migrate = vmsd->unmigratable;
1398
1399     if (dev) {
1400         char *id = qdev_get_dev_path(dev);
1401         if (id) {
1402             pstrcpy(se->idstr, sizeof(se->idstr), id);
1403             pstrcat(se->idstr, sizeof(se->idstr), "/");
1404             g_free(id);
1405
1406             se->compat = g_malloc0(sizeof(CompatEntry));
1407             pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name);
1408             se->compat->instance_id = instance_id == -1 ?
1409                          calculate_compat_instance_id(vmsd->name) : instance_id;
1410             instance_id = -1;
1411         }
1412     }
1413     pstrcat(se->idstr, sizeof(se->idstr), vmsd->name);
1414
1415     if (instance_id == -1) {
1416         se->instance_id = calculate_new_instance_id(se->idstr);
1417     } else {
1418         se->instance_id = instance_id;
1419     }
1420     assert(!se->compat || se->instance_id == 0);
1421     /* add at the end of list */
1422     QTAILQ_INSERT_TAIL(&savevm_handlers, se, entry);
1423     return 0;
1424 }
1425
1426 void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd,
1427                         void *opaque)
1428 {
1429     SaveStateEntry *se, *new_se;
1430
1431     QTAILQ_FOREACH_SAFE(se, &savevm_handlers, entry, new_se) {
1432         if (se->vmsd == vmsd && se->opaque == opaque) {
1433             QTAILQ_REMOVE(&savevm_handlers, se, entry);
1434             if (se->compat) {
1435                 g_free(se->compat);
1436             }
1437             g_free(se);
1438         }
1439     }
1440 }
1441
1442 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1443                                     void *opaque);
1444 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1445                                    void *opaque);
1446
1447 int vmstate_load_state(QEMUFile *f, const VMStateDescription *vmsd,
1448                        void *opaque, int version_id)
1449 {
1450     VMStateField *field = vmsd->fields;
1451     int ret;
1452
1453     if (version_id > vmsd->version_id) {
1454         return -EINVAL;
1455     }
1456     if (version_id < vmsd->minimum_version_id_old) {
1457         return -EINVAL;
1458     }
1459     if  (version_id < vmsd->minimum_version_id) {
1460         return vmsd->load_state_old(f, opaque, version_id);
1461     }
1462     if (vmsd->pre_load) {
1463         int ret = vmsd->pre_load(opaque);
1464         if (ret)
1465             return ret;
1466     }
1467     while(field->name) {
1468         if ((field->field_exists &&
1469              field->field_exists(opaque, version_id)) ||
1470             (!field->field_exists &&
1471              field->version_id <= version_id)) {
1472             void *base_addr = opaque + field->offset;
1473             int i, n_elems = 1;
1474             int size = field->size;
1475
1476             if (field->flags & VMS_VBUFFER) {
1477                 size = *(int32_t *)(opaque+field->size_offset);
1478                 if (field->flags & VMS_MULTIPLY) {
1479                     size *= field->size;
1480                 }
1481             }
1482             if (field->flags & VMS_ARRAY) {
1483                 n_elems = field->num;
1484             } else if (field->flags & VMS_VARRAY_INT32) {
1485                 n_elems = *(int32_t *)(opaque+field->num_offset);
1486             } else if (field->flags & VMS_VARRAY_UINT32) {
1487                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1488             } else if (field->flags & VMS_VARRAY_UINT16) {
1489                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1490             } else if (field->flags & VMS_VARRAY_UINT8) {
1491                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1492             }
1493             if (field->flags & VMS_POINTER) {
1494                 base_addr = *(void **)base_addr + field->start;
1495             }
1496             for (i = 0; i < n_elems; i++) {
1497                 void *addr = base_addr + size * i;
1498
1499                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1500                     addr = *(void **)addr;
1501                 }
1502                 if (field->flags & VMS_STRUCT) {
1503                     ret = vmstate_load_state(f, field->vmsd, addr, field->vmsd->version_id);
1504                 } else {
1505                     ret = field->info->get(f, addr, size);
1506
1507                 }
1508                 if (ret < 0) {
1509                     return ret;
1510                 }
1511             }
1512         }
1513         field++;
1514     }
1515     ret = vmstate_subsection_load(f, vmsd, opaque);
1516     if (ret != 0) {
1517         return ret;
1518     }
1519     if (vmsd->post_load) {
1520         return vmsd->post_load(opaque, version_id);
1521     }
1522     return 0;
1523 }
1524
1525 void vmstate_save_state(QEMUFile *f, const VMStateDescription *vmsd,
1526                         void *opaque)
1527 {
1528     VMStateField *field = vmsd->fields;
1529
1530     if (vmsd->pre_save) {
1531         vmsd->pre_save(opaque);
1532     }
1533     while(field->name) {
1534         if (!field->field_exists ||
1535             field->field_exists(opaque, vmsd->version_id)) {
1536             void *base_addr = opaque + field->offset;
1537             int i, n_elems = 1;
1538             int size = field->size;
1539
1540             if (field->flags & VMS_VBUFFER) {
1541                 size = *(int32_t *)(opaque+field->size_offset);
1542                 if (field->flags & VMS_MULTIPLY) {
1543                     size *= field->size;
1544                 }
1545             }
1546             if (field->flags & VMS_ARRAY) {
1547                 n_elems = field->num;
1548             } else if (field->flags & VMS_VARRAY_INT32) {
1549                 n_elems = *(int32_t *)(opaque+field->num_offset);
1550             } else if (field->flags & VMS_VARRAY_UINT32) {
1551                 n_elems = *(uint32_t *)(opaque+field->num_offset);
1552             } else if (field->flags & VMS_VARRAY_UINT16) {
1553                 n_elems = *(uint16_t *)(opaque+field->num_offset);
1554             } else if (field->flags & VMS_VARRAY_UINT8) {
1555                 n_elems = *(uint8_t *)(opaque+field->num_offset);
1556             }
1557             if (field->flags & VMS_POINTER) {
1558                 base_addr = *(void **)base_addr + field->start;
1559             }
1560             for (i = 0; i < n_elems; i++) {
1561                 void *addr = base_addr + size * i;
1562
1563                 if (field->flags & VMS_ARRAY_OF_POINTER) {
1564                     addr = *(void **)addr;
1565                 }
1566                 if (field->flags & VMS_STRUCT) {
1567                     vmstate_save_state(f, field->vmsd, addr);
1568                 } else {
1569                     field->info->put(f, addr, size);
1570                 }
1571             }
1572         }
1573         field++;
1574     }
1575     vmstate_subsection_save(f, vmsd, opaque);
1576 }
1577
1578 static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id)
1579 {
1580     if (!se->vmsd) {         /* Old style */
1581         return se->ops->load_state(f, se->opaque, version_id);
1582     }
1583     return vmstate_load_state(f, se->vmsd, se->opaque, version_id);
1584 }
1585
1586 static void vmstate_save(QEMUFile *f, SaveStateEntry *se)
1587 {
1588     if (!se->vmsd) {         /* Old style */
1589         se->ops->save_state(f, se->opaque);
1590         return;
1591     }
1592     vmstate_save_state(f,se->vmsd, se->opaque);
1593 }
1594
1595 #define QEMU_VM_FILE_MAGIC           0x5145564d
1596 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
1597 #define QEMU_VM_FILE_VERSION         0x00000003
1598
1599 #define QEMU_VM_EOF                  0x00
1600 #define QEMU_VM_SECTION_START        0x01
1601 #define QEMU_VM_SECTION_PART         0x02
1602 #define QEMU_VM_SECTION_END          0x03
1603 #define QEMU_VM_SECTION_FULL         0x04
1604 #define QEMU_VM_SUBSECTION           0x05
1605
1606 bool qemu_savevm_state_blocked(Error **errp)
1607 {
1608     SaveStateEntry *se;
1609
1610     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1611         if (se->no_migrate) {
1612             error_set(errp, QERR_MIGRATION_NOT_SUPPORTED, se->idstr);
1613             return true;
1614         }
1615     }
1616     return false;
1617 }
1618
1619 void qemu_savevm_state_begin(QEMUFile *f,
1620                              const MigrationParams *params)
1621 {
1622     SaveStateEntry *se;
1623     int ret;
1624
1625     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1626         if (!se->ops || !se->ops->set_params) {
1627             continue;
1628         }
1629         se->ops->set_params(params, se->opaque);
1630     }
1631
1632     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1633     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1634
1635     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1636         int len;
1637
1638         if (!se->ops || !se->ops->save_live_setup) {
1639             continue;
1640         }
1641         if (se->ops && se->ops->is_active) {
1642             if (!se->ops->is_active(se->opaque)) {
1643                 continue;
1644             }
1645         }
1646         /* Section type */
1647         qemu_put_byte(f, QEMU_VM_SECTION_START);
1648         qemu_put_be32(f, se->section_id);
1649
1650         /* ID string */
1651         len = strlen(se->idstr);
1652         qemu_put_byte(f, len);
1653         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1654
1655         qemu_put_be32(f, se->instance_id);
1656         qemu_put_be32(f, se->version_id);
1657
1658         ret = se->ops->save_live_setup(f, se->opaque);
1659         if (ret < 0) {
1660             qemu_file_set_error(f, ret);
1661             break;
1662         }
1663     }
1664 }
1665
1666 /*
1667  * this function has three return values:
1668  *   negative: there was one error, and we have -errno.
1669  *   0 : We haven't finished, caller have to go again
1670  *   1 : We have finished, we can go to complete phase
1671  */
1672 int qemu_savevm_state_iterate(QEMUFile *f)
1673 {
1674     SaveStateEntry *se;
1675     int ret = 1;
1676
1677     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1678         if (!se->ops || !se->ops->save_live_iterate) {
1679             continue;
1680         }
1681         if (se->ops && se->ops->is_active) {
1682             if (!se->ops->is_active(se->opaque)) {
1683                 continue;
1684             }
1685         }
1686         if (qemu_file_rate_limit(f)) {
1687             return 0;
1688         }
1689         trace_savevm_section_start();
1690         /* Section type */
1691         qemu_put_byte(f, QEMU_VM_SECTION_PART);
1692         qemu_put_be32(f, se->section_id);
1693
1694         ret = se->ops->save_live_iterate(f, se->opaque);
1695         trace_savevm_section_end(se->section_id);
1696
1697         if (ret < 0) {
1698             qemu_file_set_error(f, ret);
1699         }
1700         if (ret <= 0) {
1701             /* Do not proceed to the next vmstate before this one reported
1702                completion of the current stage. This serializes the migration
1703                and reduces the probability that a faster changing state is
1704                synchronized over and over again. */
1705             break;
1706         }
1707     }
1708     return ret;
1709 }
1710
1711 void qemu_savevm_state_complete(QEMUFile *f)
1712 {
1713     SaveStateEntry *se;
1714     int ret;
1715
1716     cpu_synchronize_all_states();
1717
1718     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1719         if (!se->ops || !se->ops->save_live_complete) {
1720             continue;
1721         }
1722         if (se->ops && se->ops->is_active) {
1723             if (!se->ops->is_active(se->opaque)) {
1724                 continue;
1725             }
1726         }
1727         trace_savevm_section_start();
1728         /* Section type */
1729         qemu_put_byte(f, QEMU_VM_SECTION_END);
1730         qemu_put_be32(f, se->section_id);
1731
1732         ret = se->ops->save_live_complete(f, se->opaque);
1733         trace_savevm_section_end(se->section_id);
1734         if (ret < 0) {
1735             qemu_file_set_error(f, ret);
1736             return;
1737         }
1738     }
1739
1740     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1741         int len;
1742
1743         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1744             continue;
1745         }
1746         trace_savevm_section_start();
1747         /* Section type */
1748         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1749         qemu_put_be32(f, se->section_id);
1750
1751         /* ID string */
1752         len = strlen(se->idstr);
1753         qemu_put_byte(f, len);
1754         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1755
1756         qemu_put_be32(f, se->instance_id);
1757         qemu_put_be32(f, se->version_id);
1758
1759         vmstate_save(f, se);
1760         trace_savevm_section_end(se->section_id);
1761     }
1762
1763     qemu_put_byte(f, QEMU_VM_EOF);
1764     qemu_fflush(f);
1765 }
1766
1767 uint64_t qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size)
1768 {
1769     SaveStateEntry *se;
1770     uint64_t ret = 0;
1771
1772     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1773         if (!se->ops || !se->ops->save_live_pending) {
1774             continue;
1775         }
1776         if (se->ops && se->ops->is_active) {
1777             if (!se->ops->is_active(se->opaque)) {
1778                 continue;
1779             }
1780         }
1781         ret += se->ops->save_live_pending(f, se->opaque, max_size);
1782     }
1783     return ret;
1784 }
1785
1786 void qemu_savevm_state_cancel(void)
1787 {
1788     SaveStateEntry *se;
1789
1790     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1791         if (se->ops && se->ops->cancel) {
1792             se->ops->cancel(se->opaque);
1793         }
1794     }
1795 }
1796
1797 static int qemu_savevm_state(QEMUFile *f)
1798 {
1799     int ret;
1800     MigrationParams params = {
1801         .blk = 0,
1802         .shared = 0
1803     };
1804
1805     if (qemu_savevm_state_blocked(NULL)) {
1806         return -EINVAL;
1807     }
1808
1809     qemu_mutex_unlock_iothread();
1810     qemu_savevm_state_begin(f, &params);
1811     qemu_mutex_lock_iothread();
1812
1813     while (qemu_file_get_error(f) == 0) {
1814         if (qemu_savevm_state_iterate(f) > 0) {
1815             break;
1816         }
1817     }
1818
1819     ret = qemu_file_get_error(f);
1820     if (ret == 0) {
1821         qemu_savevm_state_complete(f);
1822         ret = qemu_file_get_error(f);
1823     }
1824     if (ret != 0) {
1825         qemu_savevm_state_cancel();
1826     }
1827     return ret;
1828 }
1829
1830 static int qemu_save_device_state(QEMUFile *f)
1831 {
1832     SaveStateEntry *se;
1833
1834     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
1835     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
1836
1837     cpu_synchronize_all_states();
1838
1839     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1840         int len;
1841
1842         if (se->is_ram) {
1843             continue;
1844         }
1845         if ((!se->ops || !se->ops->save_state) && !se->vmsd) {
1846             continue;
1847         }
1848
1849         /* Section type */
1850         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
1851         qemu_put_be32(f, se->section_id);
1852
1853         /* ID string */
1854         len = strlen(se->idstr);
1855         qemu_put_byte(f, len);
1856         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
1857
1858         qemu_put_be32(f, se->instance_id);
1859         qemu_put_be32(f, se->version_id);
1860
1861         vmstate_save(f, se);
1862     }
1863
1864     qemu_put_byte(f, QEMU_VM_EOF);
1865
1866     return qemu_file_get_error(f);
1867 }
1868
1869 static SaveStateEntry *find_se(const char *idstr, int instance_id)
1870 {
1871     SaveStateEntry *se;
1872
1873     QTAILQ_FOREACH(se, &savevm_handlers, entry) {
1874         if (!strcmp(se->idstr, idstr) &&
1875             (instance_id == se->instance_id ||
1876              instance_id == se->alias_id))
1877             return se;
1878         /* Migrating from an older version? */
1879         if (strstr(se->idstr, idstr) && se->compat) {
1880             if (!strcmp(se->compat->idstr, idstr) &&
1881                 (instance_id == se->compat->instance_id ||
1882                  instance_id == se->alias_id))
1883                 return se;
1884         }
1885     }
1886     return NULL;
1887 }
1888
1889 static const VMStateDescription *vmstate_get_subsection(const VMStateSubsection *sub, char *idstr)
1890 {
1891     while(sub && sub->needed) {
1892         if (strcmp(idstr, sub->vmsd->name) == 0) {
1893             return sub->vmsd;
1894         }
1895         sub++;
1896     }
1897     return NULL;
1898 }
1899
1900 static int vmstate_subsection_load(QEMUFile *f, const VMStateDescription *vmsd,
1901                                    void *opaque)
1902 {
1903     while (qemu_peek_byte(f, 0) == QEMU_VM_SUBSECTION) {
1904         char idstr[256];
1905         int ret;
1906         uint8_t version_id, len, size;
1907         const VMStateDescription *sub_vmsd;
1908
1909         len = qemu_peek_byte(f, 1);
1910         if (len < strlen(vmsd->name) + 1) {
1911             /* subsection name has be be "section_name/a" */
1912             return 0;
1913         }
1914         size = qemu_peek_buffer(f, (uint8_t *)idstr, len, 2);
1915         if (size != len) {
1916             return 0;
1917         }
1918         idstr[size] = 0;
1919
1920         if (strncmp(vmsd->name, idstr, strlen(vmsd->name)) != 0) {
1921             /* it don't have a valid subsection name */
1922             return 0;
1923         }
1924         sub_vmsd = vmstate_get_subsection(vmsd->subsections, idstr);
1925         if (sub_vmsd == NULL) {
1926             return -ENOENT;
1927         }
1928         qemu_file_skip(f, 1); /* subsection */
1929         qemu_file_skip(f, 1); /* len */
1930         qemu_file_skip(f, len); /* idstr */
1931         version_id = qemu_get_be32(f);
1932
1933         ret = vmstate_load_state(f, sub_vmsd, opaque, version_id);
1934         if (ret) {
1935             return ret;
1936         }
1937     }
1938     return 0;
1939 }
1940
1941 static void vmstate_subsection_save(QEMUFile *f, const VMStateDescription *vmsd,
1942                                     void *opaque)
1943 {
1944     const VMStateSubsection *sub = vmsd->subsections;
1945
1946     while (sub && sub->needed) {
1947         if (sub->needed(opaque)) {
1948             const VMStateDescription *vmsd = sub->vmsd;
1949             uint8_t len;
1950
1951             qemu_put_byte(f, QEMU_VM_SUBSECTION);
1952             len = strlen(vmsd->name);
1953             qemu_put_byte(f, len);
1954             qemu_put_buffer(f, (uint8_t *)vmsd->name, len);
1955             qemu_put_be32(f, vmsd->version_id);
1956             vmstate_save_state(f, vmsd, opaque);
1957         }
1958         sub++;
1959     }
1960 }
1961
1962 typedef struct LoadStateEntry {
1963     QLIST_ENTRY(LoadStateEntry) entry;
1964     SaveStateEntry *se;
1965     int section_id;
1966     int version_id;
1967 } LoadStateEntry;
1968
1969 int qemu_loadvm_state(QEMUFile *f)
1970 {
1971     QLIST_HEAD(, LoadStateEntry) loadvm_handlers =
1972         QLIST_HEAD_INITIALIZER(loadvm_handlers);
1973     LoadStateEntry *le, *new_le;
1974     uint8_t section_type;
1975     unsigned int v;
1976     int ret;
1977
1978     if (qemu_savevm_state_blocked(NULL)) {
1979         return -EINVAL;
1980     }
1981
1982     v = qemu_get_be32(f);
1983     if (v != QEMU_VM_FILE_MAGIC)
1984         return -EINVAL;
1985
1986     v = qemu_get_be32(f);
1987     if (v == QEMU_VM_FILE_VERSION_COMPAT) {
1988         fprintf(stderr, "SaveVM v2 format is obsolete and don't work anymore\n");
1989         return -ENOTSUP;
1990     }
1991     if (v != QEMU_VM_FILE_VERSION)
1992         return -ENOTSUP;
1993
1994     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
1995         uint32_t instance_id, version_id, section_id;
1996         SaveStateEntry *se;
1997         char idstr[257];
1998         int len;
1999
2000         switch (section_type) {
2001         case QEMU_VM_SECTION_START:
2002         case QEMU_VM_SECTION_FULL:
2003             /* Read section start */
2004             section_id = qemu_get_be32(f);
2005             len = qemu_get_byte(f);
2006             qemu_get_buffer(f, (uint8_t *)idstr, len);
2007             idstr[len] = 0;
2008             instance_id = qemu_get_be32(f);
2009             version_id = qemu_get_be32(f);
2010
2011             /* Find savevm section */
2012             se = find_se(idstr, instance_id);
2013             if (se == NULL) {
2014                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
2015                 ret = -EINVAL;
2016                 goto out;
2017             }
2018
2019             /* Validate version */
2020             if (version_id > se->version_id) {
2021                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
2022                         version_id, idstr, se->version_id);
2023                 ret = -EINVAL;
2024                 goto out;
2025             }
2026
2027             /* Add entry */
2028             le = g_malloc0(sizeof(*le));
2029
2030             le->se = se;
2031             le->section_id = section_id;
2032             le->version_id = version_id;
2033             QLIST_INSERT_HEAD(&loadvm_handlers, le, entry);
2034
2035             ret = vmstate_load(f, le->se, le->version_id);
2036             if (ret < 0) {
2037                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
2038                         instance_id, idstr);
2039                 goto out;
2040             }
2041             break;
2042         case QEMU_VM_SECTION_PART:
2043         case QEMU_VM_SECTION_END:
2044             section_id = qemu_get_be32(f);
2045
2046             QLIST_FOREACH(le, &loadvm_handlers, entry) {
2047                 if (le->section_id == section_id) {
2048                     break;
2049                 }
2050             }
2051             if (le == NULL) {
2052                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
2053                 ret = -EINVAL;
2054                 goto out;
2055             }
2056
2057             ret = vmstate_load(f, le->se, le->version_id);
2058             if (ret < 0) {
2059                 fprintf(stderr, "qemu: warning: error while loading state section id %d\n",
2060                         section_id);
2061                 goto out;
2062             }
2063             break;
2064         default:
2065             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
2066             ret = -EINVAL;
2067             goto out;
2068         }
2069     }
2070
2071     cpu_synchronize_all_post_init();
2072
2073     ret = 0;
2074
2075 out:
2076     QLIST_FOREACH_SAFE(le, &loadvm_handlers, entry, new_le) {
2077         QLIST_REMOVE(le, entry);
2078         g_free(le);
2079     }
2080
2081     if (ret == 0) {
2082         ret = qemu_file_get_error(f);
2083     }
2084
2085     return ret;
2086 }
2087
2088 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
2089                               const char *name)
2090 {
2091     QEMUSnapshotInfo *sn_tab, *sn;
2092     int nb_sns, i, ret;
2093
2094     ret = -ENOENT;
2095     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2096     if (nb_sns < 0)
2097         return ret;
2098     for(i = 0; i < nb_sns; i++) {
2099         sn = &sn_tab[i];
2100         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
2101             *sn_info = *sn;
2102             ret = 0;
2103             break;
2104         }
2105     }
2106     g_free(sn_tab);
2107     return ret;
2108 }
2109
2110 /*
2111  * Deletes snapshots of a given name in all opened images.
2112  */
2113 static int del_existing_snapshots(Monitor *mon, const char *name)
2114 {
2115     BlockDriverState *bs;
2116     QEMUSnapshotInfo sn1, *snapshot = &sn1;
2117     int ret;
2118
2119     bs = NULL;
2120     while ((bs = bdrv_next(bs))) {
2121         if (bdrv_can_snapshot(bs) &&
2122             bdrv_snapshot_find(bs, snapshot, name) >= 0)
2123         {
2124             ret = bdrv_snapshot_delete(bs, name);
2125             if (ret < 0) {
2126                 monitor_printf(mon,
2127                                "Error while deleting snapshot on '%s'\n",
2128                                bdrv_get_device_name(bs));
2129                 return -1;
2130             }
2131         }
2132     }
2133
2134     return 0;
2135 }
2136
2137 void do_savevm(Monitor *mon, const QDict *qdict)
2138 {
2139     BlockDriverState *bs, *bs1;
2140     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
2141     int ret;
2142     QEMUFile *f;
2143     int saved_vm_running;
2144     uint64_t vm_state_size;
2145     qemu_timeval tv;
2146     struct tm tm;
2147     const char *name = qdict_get_try_str(qdict, "name");
2148
2149     /* Verify if there is a device that doesn't support snapshots and is writable */
2150     bs = NULL;
2151     while ((bs = bdrv_next(bs))) {
2152
2153         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2154             continue;
2155         }
2156
2157         if (!bdrv_can_snapshot(bs)) {
2158             monitor_printf(mon, "Device '%s' is writable but does not support snapshots.\n",
2159                                bdrv_get_device_name(bs));
2160             return;
2161         }
2162     }
2163
2164     bs = bdrv_snapshots();
2165     if (!bs) {
2166         monitor_printf(mon, "No block device can accept snapshots\n");
2167         return;
2168     }
2169
2170     saved_vm_running = runstate_is_running();
2171     vm_stop(RUN_STATE_SAVE_VM);
2172
2173     memset(sn, 0, sizeof(*sn));
2174
2175     /* fill auxiliary fields */
2176     qemu_gettimeofday(&tv);
2177     sn->date_sec = tv.tv_sec;
2178     sn->date_nsec = tv.tv_usec * 1000;
2179     sn->vm_clock_nsec = qemu_get_clock_ns(vm_clock);
2180
2181     if (name) {
2182         ret = bdrv_snapshot_find(bs, old_sn, name);
2183         if (ret >= 0) {
2184             pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
2185             pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
2186         } else {
2187             pstrcpy(sn->name, sizeof(sn->name), name);
2188         }
2189     } else {
2190         /* cast below needed for OpenBSD where tv_sec is still 'long' */
2191         localtime_r((const time_t *)&tv.tv_sec, &tm);
2192         strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm);
2193     }
2194
2195     /* Delete old snapshots of the same name */
2196     if (name && del_existing_snapshots(mon, name) < 0) {
2197         goto the_end;
2198     }
2199
2200     /* save the VM state */
2201     f = qemu_fopen_bdrv(bs, 1);
2202     if (!f) {
2203         monitor_printf(mon, "Could not open VM state file\n");
2204         goto the_end;
2205     }
2206     ret = qemu_savevm_state(f);
2207     vm_state_size = qemu_ftell(f);
2208     qemu_fclose(f);
2209     if (ret < 0) {
2210         monitor_printf(mon, "Error %d while writing VM\n", ret);
2211         goto the_end;
2212     }
2213
2214     /* create the snapshots */
2215
2216     bs1 = NULL;
2217     while ((bs1 = bdrv_next(bs1))) {
2218         if (bdrv_can_snapshot(bs1)) {
2219             /* Write VM state size only to the image that contains the state */
2220             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
2221             ret = bdrv_snapshot_create(bs1, sn);
2222             if (ret < 0) {
2223                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
2224                                bdrv_get_device_name(bs1));
2225             }
2226         }
2227     }
2228
2229  the_end:
2230     if (saved_vm_running)
2231         vm_start();
2232 }
2233
2234 void qmp_xen_save_devices_state(const char *filename, Error **errp)
2235 {
2236     QEMUFile *f;
2237     int saved_vm_running;
2238     int ret;
2239
2240     saved_vm_running = runstate_is_running();
2241     vm_stop(RUN_STATE_SAVE_VM);
2242
2243     f = qemu_fopen(filename, "wb");
2244     if (!f) {
2245         error_set(errp, QERR_OPEN_FILE_FAILED, filename);
2246         goto the_end;
2247     }
2248     ret = qemu_save_device_state(f);
2249     qemu_fclose(f);
2250     if (ret < 0) {
2251         error_set(errp, QERR_IO_ERROR);
2252     }
2253
2254  the_end:
2255     if (saved_vm_running)
2256         vm_start();
2257 }
2258
2259 int load_vmstate(const char *name)
2260 {
2261     BlockDriverState *bs, *bs_vm_state;
2262     QEMUSnapshotInfo sn;
2263     QEMUFile *f;
2264     int ret;
2265
2266     bs_vm_state = bdrv_snapshots();
2267     if (!bs_vm_state) {
2268         error_report("No block device supports snapshots");
2269         return -ENOTSUP;
2270     }
2271
2272     /* Don't even try to load empty VM states */
2273     ret = bdrv_snapshot_find(bs_vm_state, &sn, name);
2274     if (ret < 0) {
2275         return ret;
2276     } else if (sn.vm_state_size == 0) {
2277         error_report("This is a disk-only snapshot. Revert to it offline "
2278             "using qemu-img.");
2279         return -EINVAL;
2280     }
2281
2282     /* Verify if there is any device that doesn't support snapshots and is
2283     writable and check if the requested snapshot is available too. */
2284     bs = NULL;
2285     while ((bs = bdrv_next(bs))) {
2286
2287         if (!bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2288             continue;
2289         }
2290
2291         if (!bdrv_can_snapshot(bs)) {
2292             error_report("Device '%s' is writable but does not support snapshots.",
2293                                bdrv_get_device_name(bs));
2294             return -ENOTSUP;
2295         }
2296
2297         ret = bdrv_snapshot_find(bs, &sn, name);
2298         if (ret < 0) {
2299             error_report("Device '%s' does not have the requested snapshot '%s'",
2300                            bdrv_get_device_name(bs), name);
2301             return ret;
2302         }
2303     }
2304
2305     /* Flush all IO requests so they don't interfere with the new state.  */
2306     bdrv_drain_all();
2307
2308     bs = NULL;
2309     while ((bs = bdrv_next(bs))) {
2310         if (bdrv_can_snapshot(bs)) {
2311             ret = bdrv_snapshot_goto(bs, name);
2312             if (ret < 0) {
2313                 error_report("Error %d while activating snapshot '%s' on '%s'",
2314                              ret, name, bdrv_get_device_name(bs));
2315                 return ret;
2316             }
2317         }
2318     }
2319
2320     /* restore the VM state */
2321     f = qemu_fopen_bdrv(bs_vm_state, 0);
2322     if (!f) {
2323         error_report("Could not open VM state file");
2324         return -EINVAL;
2325     }
2326
2327     qemu_system_reset(VMRESET_SILENT);
2328     ret = qemu_loadvm_state(f);
2329
2330     qemu_fclose(f);
2331     if (ret < 0) {
2332         error_report("Error %d while loading VM state", ret);
2333         return ret;
2334     }
2335
2336     return 0;
2337 }
2338
2339 void do_delvm(Monitor *mon, const QDict *qdict)
2340 {
2341     BlockDriverState *bs, *bs1;
2342     int ret;
2343     const char *name = qdict_get_str(qdict, "name");
2344
2345     bs = bdrv_snapshots();
2346     if (!bs) {
2347         monitor_printf(mon, "No block device supports snapshots\n");
2348         return;
2349     }
2350
2351     bs1 = NULL;
2352     while ((bs1 = bdrv_next(bs1))) {
2353         if (bdrv_can_snapshot(bs1)) {
2354             ret = bdrv_snapshot_delete(bs1, name);
2355             if (ret < 0) {
2356                 if (ret == -ENOTSUP)
2357                     monitor_printf(mon,
2358                                    "Snapshots not supported on device '%s'\n",
2359                                    bdrv_get_device_name(bs1));
2360                 else
2361                     monitor_printf(mon, "Error %d while deleting snapshot on "
2362                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
2363             }
2364         }
2365     }
2366 }
2367
2368 void do_info_snapshots(Monitor *mon, const QDict *qdict)
2369 {
2370     BlockDriverState *bs, *bs1;
2371     QEMUSnapshotInfo *sn_tab, *sn, s, *sn_info = &s;
2372     int nb_sns, i, ret, available;
2373     int total;
2374     int *available_snapshots;
2375     char buf[256];
2376
2377     bs = bdrv_snapshots();
2378     if (!bs) {
2379         monitor_printf(mon, "No available block device supports snapshots\n");
2380         return;
2381     }
2382
2383     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2384     if (nb_sns < 0) {
2385         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
2386         return;
2387     }
2388
2389     if (nb_sns == 0) {
2390         monitor_printf(mon, "There is no snapshot available.\n");
2391         return;
2392     }
2393
2394     available_snapshots = g_malloc0(sizeof(int) * nb_sns);
2395     total = 0;
2396     for (i = 0; i < nb_sns; i++) {
2397         sn = &sn_tab[i];
2398         available = 1;
2399         bs1 = NULL;
2400
2401         while ((bs1 = bdrv_next(bs1))) {
2402             if (bdrv_can_snapshot(bs1) && bs1 != bs) {
2403                 ret = bdrv_snapshot_find(bs1, sn_info, sn->id_str);
2404                 if (ret < 0) {
2405                     available = 0;
2406                     break;
2407                 }
2408             }
2409         }
2410
2411         if (available) {
2412             available_snapshots[total] = i;
2413             total++;
2414         }
2415     }
2416
2417     if (total > 0) {
2418         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
2419         for (i = 0; i < total; i++) {
2420             sn = &sn_tab[available_snapshots[i]];
2421             monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
2422         }
2423     } else {
2424         monitor_printf(mon, "There is no suitable snapshot available\n");
2425     }
2426
2427     g_free(sn_tab);
2428     g_free(available_snapshots);
2429
2430 }
2431
2432 void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev)
2433 {
2434     qemu_ram_set_idstr(memory_region_get_ram_addr(mr) & TARGET_PAGE_MASK,
2435                        memory_region_name(mr), dev);
2436 }
2437
2438 void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev)
2439 {
2440     /* Nothing do to while the implementation is in RAMBlock */
2441 }
2442
2443 void vmstate_register_ram_global(MemoryRegion *mr)
2444 {
2445     vmstate_register_ram(mr, NULL);
2446 }