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 HOST_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 #if defined(__NetBSD__)
  46 #include <net/if_tap.h>
  47 #endif
  48 #ifdef __linux__
  49 #include <linux/if_tun.h>
  50 #endif
  51 #include <arpa/inet.h>
  52 #include <dirent.h>
  53 #include <netdb.h>
  54 #include <sys/select.h>
  55 #ifdef HOST_BSD
  56 #include <sys/stat.h>
  57 #if defined(__FreeBSD__) || defined(__DragonFly__)
  58 #include <libutil.h>
  59 #else
  60 #include <util.h>
  61 #endif
  62 #elif defined (__GLIBC__) && defined (__FreeBSD_kernel__)
  63 #include <freebsd/stdlib.h>
  64 #else
  65 #ifdef __linux__
  66 #include <pty.h>
  67 #include <malloc.h>
  68 #include <linux/rtc.h>
  69 #endif
  70 #endif
  71 #endif
  72
  73 #ifdef _WIN32
  74 #include <windows.h>
  75 #include <malloc.h>
  76 #include <sys/timeb.h>
  77 #include <mmsystem.h>
  78 #define getopt_long_only getopt_long
  79 #define memalign(align, size) malloc(size)
  80 #endif
  81
  82 #include "qemu-common.h"
  83 #include "hw/hw.h"
  84 #include "net.h"
  85 #include "monitor.h"
  86 #include "sysemu.h"
  87 #include "qemu-timer.h"
  88 #include "qemu-char.h"
  89 #include "block.h"
  90 #include "audio/audio.h"
  91 #include "migration.h"
  92 #include "qemu_socket.h"
  93
  94 /* point to the block driver where the snapshots are managed */
  95 static BlockDriverState *bs_snapshots;
  96
  97 #define SELF_ANNOUNCE_ROUNDS 5
  98 #define ETH_P_EXPERIMENTAL 0x01F1 /* just a number */
  99 //#define ETH_P_EXPERIMENTAL 0x0012 /* make it the size of the packet */
 100 #define EXPERIMENTAL_MAGIC 0xf1f23f4f
 101
 102 static int announce_self_create(uint8_t *buf,
 103                                 uint8_t *mac_addr)
 104 {
 105     uint32_t magic = EXPERIMENTAL_MAGIC;
 106     uint16_t proto = htons(ETH_P_EXPERIMENTAL);
 107
 108     /* FIXME: should we send a different packet (arp/rarp/ping)? */
 109
 110     memset(buf, 0, 64);
 111     memset(buf, 0xff, 6);         /* h_dst */
 112     memcpy(buf + 6, mac_addr, 6); /* h_src */
 113     memcpy(buf + 12, &proto, 2);  /* h_proto */
 114     memcpy(buf + 14, &magic, 4);  /* magic */
 115
 116     return 64; /* len */
 117 }
 118
 119 static void qemu_announce_self_once(void *opaque)
 120 {
 121     int i, len;
 122     VLANState *vlan;
 123     VLANClientState *vc;
 124     uint8_t buf[256];
 125     static int count = SELF_ANNOUNCE_ROUNDS;
 126     QEMUTimer *timer = *(QEMUTimer **)opaque;
 127
 128     for (i = 0; i < MAX_NICS; i++) {
 129         if (!nd_table[i].used)
 130             continue;
 131         len = announce_self_create(buf, nd_table[i].macaddr);
 132         vlan = nd_table[i].vlan;
 133         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 134             vc->receive(vc, buf, len);
 135         }
 136     }
 137     if (count--) {
 138             qemu_mod_timer(timer, qemu_get_clock(rt_clock) + 100);
 139     } else {
 140             qemu_del_timer(timer);
 141             qemu_free_timer(timer);
 142     }
 143 }
 144
 145 void qemu_announce_self(void)
 146 {
 147         static QEMUTimer *timer;
 148         timer = qemu_new_timer(rt_clock, qemu_announce_self_once, &timer);
 149         qemu_announce_self_once(&timer);
 150 }
 151
 152 /***********************************************************/
 153 /* savevm/loadvm support */
 154
 155 #define IO_BUF_SIZE 32768
 156
 157 struct QEMUFile {
 158     QEMUFilePutBufferFunc *put_buffer;
 159     QEMUFileGetBufferFunc *get_buffer;
 160     QEMUFileCloseFunc *close;
 161     QEMUFileRateLimit *rate_limit;
 162     QEMUFileSetRateLimit *set_rate_limit;
 163     void *opaque;
 164     int is_write;
 165
 166     int64_t buf_offset; /* start of buffer when writing, end of buffer
 167                            when reading */
 168     int buf_index;
 169     int buf_size; /* 0 when writing */
 170     uint8_t buf[IO_BUF_SIZE];
 171
 172     int has_error;
 173 };
 174
 175 typedef struct QEMUFilePopen
 176 {
 177     FILE *popen_file;
 178     QEMUFile *file;
 179 } QEMUFilePopen;
 180
 181 typedef struct QEMUFileSocket
 182 {
 183     int fd;
 184     QEMUFile *file;
 185 } QEMUFileSocket;
 186
 187 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 188 {
 189     QEMUFileSocket *s = opaque;
 190     ssize_t len;
 191
 192     do {
 193         len = recv(s->fd, buf, size, 0);
 194     } while (len == -1 && socket_error() == EINTR);
 195
 196     if (len == -1)
 197         len = -socket_error();
 198
 199     return len;
 200 }
 201
 202 static int socket_close(void *opaque)
 203 {
 204     QEMUFileSocket *s = opaque;
 205     qemu_free(s);
 206     return 0;
 207 }
 208
 209 static int popen_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 210 {
 211     QEMUFilePopen *s = opaque;
 212     return fwrite(buf, 1, size, s->popen_file);
 213 }
 214
 215 static int popen_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 216 {
 217     QEMUFilePopen *s = opaque;
 218     return fread(buf, 1, size, s->popen_file);
 219 }
 220
 221 static int popen_close(void *opaque)
 222 {
 223     QEMUFilePopen *s = opaque;
 224     pclose(s->popen_file);
 225     qemu_free(s);
 226     return 0;
 227 }
 228
 229 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 230 {
 231     QEMUFilePopen *s;
 232
 233     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 234         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 235         return NULL;
 236     }
 237
 238     s = qemu_mallocz(sizeof(QEMUFilePopen));
 239
 240     s->popen_file = popen_file;
 241
 242     if(mode[0] == 'r') {
 243         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL, NULL);
 244     } else {
 245         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL, NULL);
 246     }
 247     return s->file;
 248 }
 249
 250 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 251 {
 252     FILE *popen_file;
 253
 254     popen_file = popen(command, mode);
 255     if(popen_file == NULL) {
 256         return NULL;
 257     }
 258
 259     return qemu_popen(popen_file, mode);
 260 }
 261
 262 int qemu_popen_fd(QEMUFile *f)
 263 {
 264     QEMUFilePopen *p;
 265     int fd;
 266
 267     p = (QEMUFilePopen *)f->opaque;
 268     fd = fileno(p->popen_file);
 269
 270     return fd;
 271 }
 272
 273 QEMUFile *qemu_fopen_socket(int fd)
 274 {
 275     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 276
 277     s->fd = fd;
 278     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL, NULL);
 279     return s->file;
 280 }
 281
 282 typedef struct QEMUFileStdio
 283 {
 284     FILE *outfile;
 285 } QEMUFileStdio;
 286
 287 static int file_put_buffer(void *opaque, const uint8_t *buf,
 288                             int64_t pos, int size)
 289 {
 290     QEMUFileStdio *s = opaque;
 291     fseek(s->outfile, pos, SEEK_SET);
 292     fwrite(buf, 1, size, s->outfile);
 293     return size;
 294 }
 295
 296 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 297 {
 298     QEMUFileStdio *s = opaque;
 299     fseek(s->outfile, pos, SEEK_SET);
 300     return fread(buf, 1, size, s->outfile);
 301 }
 302
 303 static int file_close(void *opaque)
 304 {
 305     QEMUFileStdio *s = opaque;
 306     fclose(s->outfile);
 307     qemu_free(s);
 308     return 0;
 309 }
 310
 311 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 312 {
 313     QEMUFileStdio *s;
 314
 315     s = qemu_mallocz(sizeof(QEMUFileStdio));
 316
 317     s->outfile = fopen(filename, mode);
 318     if (!s->outfile)
 319         goto fail;
 320
 321     if (!strcmp(mode, "wb"))
 322         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL, NULL);
 323     else if (!strcmp(mode, "rb"))
 324         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL, NULL);
 325
 326 fail:
 327     if (s->outfile)
 328         fclose(s->outfile);
 329     qemu_free(s);
 330     return NULL;
 331 }
 332
 333 typedef struct QEMUFileBdrv
 334 {
 335     BlockDriverState *bs;
 336     int64_t base_offset;
 337 } QEMUFileBdrv;
 338
 339 static int block_put_buffer(void *opaque, const uint8_t *buf,
 340                            int64_t pos, int size)
 341 {
 342     QEMUFileBdrv *s = opaque;
 343     bdrv_put_buffer(s->bs, buf, s->base_offset + pos, size);
 344     return size;
 345 }
 346
 347 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 348 {
 349     QEMUFileBdrv *s = opaque;
 350     return bdrv_get_buffer(s->bs, buf, s->base_offset + pos, size);
 351 }
 352
 353 static int bdrv_fclose(void *opaque)
 354 {
 355     QEMUFileBdrv *s = opaque;
 356     qemu_free(s);
 357     return 0;
 358 }
 359
 360 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable)
 361 {
 362     QEMUFileBdrv *s;
 363
 364     s = qemu_mallocz(sizeof(QEMUFileBdrv));
 365
 366     s->bs = bs;
 367     s->base_offset = offset;
 368
 369     if (is_writable)
 370         return qemu_fopen_ops(s, block_put_buffer, NULL, bdrv_fclose, NULL, NULL);
 371
 372     return qemu_fopen_ops(s, NULL, block_get_buffer, bdrv_fclose, NULL, NULL);
 373 }
 374
 375 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 376                          QEMUFileGetBufferFunc *get_buffer,
 377                          QEMUFileCloseFunc *close,
 378                          QEMUFileRateLimit *rate_limit,
 379                          QEMUFileSetRateLimit *set_rate_limit)
 380 {
 381     QEMUFile *f;
 382
 383     f = qemu_mallocz(sizeof(QEMUFile));
 384
 385     f->opaque = opaque;
 386     f->put_buffer = put_buffer;
 387     f->get_buffer = get_buffer;
 388     f->close = close;
 389     f->rate_limit = rate_limit;
 390     f->set_rate_limit = set_rate_limit;
 391     f->is_write = 0;
 392
 393     return f;
 394 }
 395
 396 int qemu_file_has_error(QEMUFile *f)
 397 {
 398     return f->has_error;
 399 }
 400
 401 void qemu_file_set_error(QEMUFile *f)
 402 {
 403     f->has_error = 1;
 404 }
 405
 406 void qemu_fflush(QEMUFile *f)
 407 {
 408     if (!f->put_buffer)
 409         return;
 410
 411     if (f->is_write && f->buf_index > 0) {
 412         int len;
 413
 414         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 415         if (len > 0)
 416             f->buf_offset += f->buf_index;
 417         else
 418             f->has_error = 1;
 419         f->buf_index = 0;
 420     }
 421 }
 422
 423 static void qemu_fill_buffer(QEMUFile *f)
 424 {
 425     int len;
 426
 427     if (!f->get_buffer)
 428         return;
 429
 430     if (f->is_write)
 431         abort();
 432
 433     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 434     if (len > 0) {
 435         f->buf_index = 0;
 436         f->buf_size = len;
 437         f->buf_offset += len;
 438     } else if (len != -EAGAIN)
 439         f->has_error = 1;
 440 }
 441
 442 int qemu_fclose(QEMUFile *f)
 443 {
 444     int ret = 0;
 445     qemu_fflush(f);
 446     if (f->close)
 447         ret = f->close(f->opaque);
 448     qemu_free(f);
 449     return ret;
 450 }
 451
 452 void qemu_file_put_notify(QEMUFile *f)
 453 {
 454     f->put_buffer(f->opaque, NULL, 0, 0);
 455 }
 456
 457 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 458 {
 459     int l;
 460
 461     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 462         fprintf(stderr,
 463                 "Attempted to write to buffer while read buffer is not empty\n");
 464         abort();
 465     }
 466
 467     while (!f->has_error && size > 0) {
 468         l = IO_BUF_SIZE - f->buf_index;
 469         if (l > size)
 470             l = size;
 471         memcpy(f->buf + f->buf_index, buf, l);
 472         f->is_write = 1;
 473         f->buf_index += l;
 474         buf += l;
 475         size -= l;
 476         if (f->buf_index >= IO_BUF_SIZE)
 477             qemu_fflush(f);
 478     }
 479 }
 480
 481 void qemu_put_byte(QEMUFile *f, int v)
 482 {
 483     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 484         fprintf(stderr,
 485                 "Attempted to write to buffer while read buffer is not empty\n");
 486         abort();
 487     }
 488
 489     f->buf[f->buf_index++] = v;
 490     f->is_write = 1;
 491     if (f->buf_index >= IO_BUF_SIZE)
 492         qemu_fflush(f);
 493 }
 494
 495 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 496 {
 497     int size, l;
 498
 499     if (f->is_write)
 500         abort();
 501
 502     size = size1;
 503     while (size > 0) {
 504         l = f->buf_size - f->buf_index;
 505         if (l == 0) {
 506             qemu_fill_buffer(f);
 507             l = f->buf_size - f->buf_index;
 508             if (l == 0)
 509                 break;
 510         }
 511         if (l > size)
 512             l = size;
 513         memcpy(buf, f->buf + f->buf_index, l);
 514         f->buf_index += l;
 515         buf += l;
 516         size -= l;
 517     }
 518     return size1 - size;
 519 }
 520
 521 int qemu_get_byte(QEMUFile *f)
 522 {
 523     if (f->is_write)
 524         abort();
 525
 526     if (f->buf_index >= f->buf_size) {
 527         qemu_fill_buffer(f);
 528         if (f->buf_index >= f->buf_size)
 529             return 0;
 530     }
 531     return f->buf[f->buf_index++];
 532 }
 533
 534 int64_t qemu_ftell(QEMUFile *f)
 535 {
 536     return f->buf_offset - f->buf_size + f->buf_index;
 537 }
 538
 539 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 540 {
 541     if (whence == SEEK_SET) {
 542         /* nothing to do */
 543     } else if (whence == SEEK_CUR) {
 544         pos += qemu_ftell(f);
 545     } else {
 546         /* SEEK_END not supported */
 547         return -1;
 548     }
 549     if (f->put_buffer) {
 550         qemu_fflush(f);
 551         f->buf_offset = pos;
 552     } else {
 553         f->buf_offset = pos;
 554         f->buf_index = 0;
 555         f->buf_size = 0;
 556     }
 557     return pos;
 558 }
 559
 560 int qemu_file_rate_limit(QEMUFile *f)
 561 {
 562     if (f->rate_limit)
 563         return f->rate_limit(f->opaque);
 564
 565     return 0;
 566 }
 567
 568 size_t qemu_file_set_rate_limit(QEMUFile *f, size_t new_rate)
 569 {
 570     if (f->set_rate_limit)
 571         return f->set_rate_limit(f->opaque, new_rate);
 572
 573     return 0;
 574 }
 575
 576 void qemu_put_be16(QEMUFile *f, unsigned int v)
 577 {
 578     qemu_put_byte(f, v >> 8);
 579     qemu_put_byte(f, v);
 580 }
 581
 582 void qemu_put_be32(QEMUFile *f, unsigned int v)
 583 {
 584     qemu_put_byte(f, v >> 24);
 585     qemu_put_byte(f, v >> 16);
 586     qemu_put_byte(f, v >> 8);
 587     qemu_put_byte(f, v);
 588 }
 589
 590 void qemu_put_be64(QEMUFile *f, uint64_t v)
 591 {
 592     qemu_put_be32(f, v >> 32);
 593     qemu_put_be32(f, v);
 594 }
 595
 596 unsigned int qemu_get_be16(QEMUFile *f)
 597 {
 598     unsigned int v;
 599     v = qemu_get_byte(f) << 8;
 600     v |= qemu_get_byte(f);
 601     return v;
 602 }
 603
 604 unsigned int qemu_get_be32(QEMUFile *f)
 605 {
 606     unsigned int v;
 607     v = qemu_get_byte(f) << 24;
 608     v |= qemu_get_byte(f) << 16;
 609     v |= qemu_get_byte(f) << 8;
 610     v |= qemu_get_byte(f);
 611     return v;
 612 }
 613
 614 uint64_t qemu_get_be64(QEMUFile *f)
 615 {
 616     uint64_t v;
 617     v = (uint64_t)qemu_get_be32(f) << 32;
 618     v |= qemu_get_be32(f);
 619     return v;
 620 }
 621
 622 typedef struct SaveStateEntry {
 623     char idstr[256];
 624     int instance_id;
 625     int version_id;
 626     int section_id;
 627     SaveLiveStateHandler *save_live_state;
 628     SaveStateHandler *save_state;
 629     LoadStateHandler *load_state;
 630     void *opaque;
 631     struct SaveStateEntry *next;
 632 } SaveStateEntry;
 633
 634 static SaveStateEntry *first_se;
 635
 636 /* TODO: Individual devices generally have very little idea about the rest
 637    of the system, so instance_id should be removed/replaced.
 638    Meanwhile pass -1 as instance_id if you do not already have a clearly
 639    distinguishing id for all instances of your device class. */
 640 int register_savevm_live(const char *idstr,
 641                          int instance_id,
 642                          int version_id,
 643                          SaveLiveStateHandler *save_live_state,
 644                          SaveStateHandler *save_state,
 645                          LoadStateHandler *load_state,
 646                          void *opaque)
 647 {
 648     SaveStateEntry *se, **pse;
 649     static int global_section_id;
 650
 651     se = qemu_malloc(sizeof(SaveStateEntry));
 652     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 653     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 654     se->version_id = version_id;
 655     se->section_id = global_section_id++;
 656     se->save_live_state = save_live_state;
 657     se->save_state = save_state;
 658     se->load_state = load_state;
 659     se->opaque = opaque;
 660     se->next = NULL;
 661
 662     /* add at the end of list */
 663     pse = &first_se;
 664     while (*pse != NULL) {
 665         if (instance_id == -1
 666                 && strcmp(se->idstr, (*pse)->idstr) == 0
 667                 && se->instance_id <= (*pse)->instance_id)
 668             se->instance_id = (*pse)->instance_id + 1;
 669         pse = &(*pse)->next;
 670     }
 671     *pse = se;
 672     return 0;
 673 }
 674
 675 int register_savevm(const char *idstr,
 676                     int instance_id,
 677                     int version_id,
 678                     SaveStateHandler *save_state,
 679                     LoadStateHandler *load_state,
 680                     void *opaque)
 681 {
 682     return register_savevm_live(idstr, instance_id, version_id,
 683                                 NULL, save_state, load_state, opaque);
 684 }
 685
 686 void unregister_savevm(const char *idstr, void *opaque)
 687 {
 688     SaveStateEntry **pse;
 689
 690     pse = &first_se;
 691     while (*pse != NULL) {
 692         if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
 693             SaveStateEntry *next = (*pse)->next;
 694             qemu_free(*pse);
 695             *pse = next;
 696             continue;
 697         }
 698         pse = &(*pse)->next;
 699     }
 700 }
 701
 702 #define QEMU_VM_FILE_MAGIC           0x5145564d
 703 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 704 #define QEMU_VM_FILE_VERSION         0x00000003
 705
 706 #define QEMU_VM_EOF                  0x00
 707 #define QEMU_VM_SECTION_START        0x01
 708 #define QEMU_VM_SECTION_PART         0x02
 709 #define QEMU_VM_SECTION_END          0x03
 710 #define QEMU_VM_SECTION_FULL         0x04
 711
 712 int qemu_savevm_state_begin(QEMUFile *f)
 713 {
 714     SaveStateEntry *se;
 715
 716     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 717     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 718
 719     for (se = first_se; se != NULL; se = se->next) {
 720         int len;
 721
 722         if (se->save_live_state == NULL)
 723             continue;
 724
 725         /* Section type */
 726         qemu_put_byte(f, QEMU_VM_SECTION_START);
 727         qemu_put_be32(f, se->section_id);
 728
 729         /* ID string */
 730         len = strlen(se->idstr);
 731         qemu_put_byte(f, len);
 732         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 733
 734         qemu_put_be32(f, se->instance_id);
 735         qemu_put_be32(f, se->version_id);
 736
 737         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 738     }
 739
 740     if (qemu_file_has_error(f))
 741         return -EIO;
 742
 743     return 0;
 744 }
 745
 746 int qemu_savevm_state_iterate(QEMUFile *f)
 747 {
 748     SaveStateEntry *se;
 749     int ret = 1;
 750
 751     for (se = first_se; se != NULL; se = se->next) {
 752         if (se->save_live_state == NULL)
 753             continue;
 754
 755         /* Section type */
 756         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 757         qemu_put_be32(f, se->section_id);
 758
 759         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 760     }
 761
 762     if (ret)
 763         return 1;
 764
 765     if (qemu_file_has_error(f))
 766         return -EIO;
 767
 768     return 0;
 769 }
 770
 771 int qemu_savevm_state_complete(QEMUFile *f)
 772 {
 773     SaveStateEntry *se;
 774
 775     for (se = first_se; se != NULL; se = se->next) {
 776         if (se->save_live_state == NULL)
 777             continue;
 778
 779         /* Section type */
 780         qemu_put_byte(f, QEMU_VM_SECTION_END);
 781         qemu_put_be32(f, se->section_id);
 782
 783         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 784     }
 785
 786     for(se = first_se; se != NULL; se = se->next) {
 787         int len;
 788
 789         if (se->save_state == NULL)
 790             continue;
 791
 792         /* Section type */
 793         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 794         qemu_put_be32(f, se->section_id);
 795
 796         /* ID string */
 797         len = strlen(se->idstr);
 798         qemu_put_byte(f, len);
 799         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 800
 801         qemu_put_be32(f, se->instance_id);
 802         qemu_put_be32(f, se->version_id);
 803
 804         se->save_state(f, se->opaque);
 805     }
 806
 807     qemu_put_byte(f, QEMU_VM_EOF);
 808
 809     if (qemu_file_has_error(f))
 810         return -EIO;
 811
 812     return 0;
 813 }
 814
 815 int qemu_savevm_state(QEMUFile *f)
 816 {
 817     int saved_vm_running;
 818     int ret;
 819
 820     saved_vm_running = vm_running;
 821     vm_stop(0);
 822
 823     bdrv_flush_all();
 824
 825     ret = qemu_savevm_state_begin(f);
 826     if (ret < 0)
 827         goto out;
 828
 829     do {
 830         ret = qemu_savevm_state_iterate(f);
 831         if (ret < 0)
 832             goto out;
 833     } while (ret == 0);
 834
 835     ret = qemu_savevm_state_complete(f);
 836
 837 out:
 838     if (qemu_file_has_error(f))
 839         ret = -EIO;
 840
 841     if (!ret && saved_vm_running)
 842         vm_start();
 843
 844     return ret;
 845 }
 846
 847 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 848 {
 849     SaveStateEntry *se;
 850
 851     for(se = first_se; se != NULL; se = se->next) {
 852         if (!strcmp(se->idstr, idstr) &&
 853             instance_id == se->instance_id)
 854             return se;
 855     }
 856     return NULL;
 857 }
 858
 859 typedef struct LoadStateEntry {
 860     SaveStateEntry *se;
 861     int section_id;
 862     int version_id;
 863     struct LoadStateEntry *next;
 864 } LoadStateEntry;
 865
 866 static int qemu_loadvm_state_v2(QEMUFile *f)
 867 {
 868     SaveStateEntry *se;
 869     int len, ret, instance_id, record_len, version_id;
 870     int64_t total_len, end_pos, cur_pos;
 871     char idstr[256];
 872
 873     total_len = qemu_get_be64(f);
 874     end_pos = total_len + qemu_ftell(f);
 875     for(;;) {
 876         if (qemu_ftell(f) >= end_pos)
 877             break;
 878         len = qemu_get_byte(f);
 879         qemu_get_buffer(f, (uint8_t *)idstr, len);
 880         idstr[len] = '\0';
 881         instance_id = qemu_get_be32(f);
 882         version_id = qemu_get_be32(f);
 883         record_len = qemu_get_be32(f);
 884         cur_pos = qemu_ftell(f);
 885         se = find_se(idstr, instance_id);
 886         if (!se) {
 887             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 888                     instance_id, idstr);
 889         } else {
 890             ret = se->load_state(f, se->opaque, version_id);
 891             if (ret < 0) {
 892                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 893                         instance_id, idstr);
 894                 return ret;
 895             }
 896         }
 897         /* always seek to exact end of record */
 898         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 899     }
 900
 901     if (qemu_file_has_error(f))
 902         return -EIO;
 903
 904     return 0;
 905 }
 906
 907 int qemu_loadvm_state(QEMUFile *f)
 908 {
 909     LoadStateEntry *first_le = NULL;
 910     uint8_t section_type;
 911     unsigned int v;
 912     int ret;
 913
 914     v = qemu_get_be32(f);
 915     if (v != QEMU_VM_FILE_MAGIC)
 916         return -EINVAL;
 917
 918     v = qemu_get_be32(f);
 919     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 920         return qemu_loadvm_state_v2(f);
 921     if (v != QEMU_VM_FILE_VERSION)
 922         return -ENOTSUP;
 923
 924     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 925         uint32_t instance_id, version_id, section_id;
 926         LoadStateEntry *le;
 927         SaveStateEntry *se;
 928         char idstr[257];
 929         int len;
 930
 931         switch (section_type) {
 932         case QEMU_VM_SECTION_START:
 933         case QEMU_VM_SECTION_FULL:
 934             /* Read section start */
 935             section_id = qemu_get_be32(f);
 936             len = qemu_get_byte(f);
 937             qemu_get_buffer(f, (uint8_t *)idstr, len);
 938             idstr[len] = 0;
 939             instance_id = qemu_get_be32(f);
 940             version_id = qemu_get_be32(f);
 941
 942             /* Find savevm section */
 943             se = find_se(idstr, instance_id);
 944             if (se == NULL) {
 945                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 946                 ret = -EINVAL;
 947                 goto out;
 948             }
 949
 950             /* Validate version */
 951             if (version_id > se->version_id) {
 952                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 953                         version_id, idstr, se->version_id);
 954                 ret = -EINVAL;
 955                 goto out;
 956             }
 957
 958             /* Add entry */
 959             le = qemu_mallocz(sizeof(*le));
 960
 961             le->se = se;
 962             le->section_id = section_id;
 963             le->version_id = version_id;
 964             le->next = first_le;
 965             first_le = le;
 966
 967             le->se->load_state(f, le->se->opaque, le->version_id);
 968             break;
 969         case QEMU_VM_SECTION_PART:
 970         case QEMU_VM_SECTION_END:
 971             section_id = qemu_get_be32(f);
 972
 973             for (le = first_le; le && le->section_id != section_id; le = le->next);
 974             if (le == NULL) {
 975                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 976                 ret = -EINVAL;
 977                 goto out;
 978             }
 979
 980             le->se->load_state(f, le->se->opaque, le->version_id);
 981             break;
 982         default:
 983             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 984             ret = -EINVAL;
 985             goto out;
 986         }
 987     }
 988
 989     ret = 0;
 990
 991 out:
 992     while (first_le) {
 993         LoadStateEntry *le = first_le;
 994         first_le = first_le->next;
 995         qemu_free(le);
 996     }
 997
 998     if (qemu_file_has_error(f))
 999         ret = -EIO;
1000
1001     return ret;
1002 }
1003
1004 /* device can contain snapshots */
1005 static int bdrv_can_snapshot(BlockDriverState *bs)
1006 {
1007     return (bs &&
1008             !bdrv_is_removable(bs) &&
1009             !bdrv_is_read_only(bs));
1010 }
1011
1012 /* device must be snapshots in order to have a reliable snapshot */
1013 static int bdrv_has_snapshot(BlockDriverState *bs)
1014 {
1015     return (bs &&
1016             !bdrv_is_removable(bs) &&
1017             !bdrv_is_read_only(bs));
1018 }
1019
1020 static BlockDriverState *get_bs_snapshots(void)
1021 {
1022     BlockDriverState *bs;
1023     int i;
1024
1025     if (bs_snapshots)
1026         return bs_snapshots;
1027     for(i = 0; i <= nb_drives; i++) {
1028         bs = drives_table[i].bdrv;
1029         if (bdrv_can_snapshot(bs))
1030             goto ok;
1031     }
1032     return NULL;
1033  ok:
1034     bs_snapshots = bs;
1035     return bs;
1036 }
1037
1038 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1039                               const char *name)
1040 {
1041     QEMUSnapshotInfo *sn_tab, *sn;
1042     int nb_sns, i, ret;
1043
1044     ret = -ENOENT;
1045     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1046     if (nb_sns < 0)
1047         return ret;
1048     for(i = 0; i < nb_sns; i++) {
1049         sn = &sn_tab[i];
1050         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1051             *sn_info = *sn;
1052             ret = 0;
1053             break;
1054         }
1055     }
1056     qemu_free(sn_tab);
1057     return ret;
1058 }
1059
1060 void do_savevm(Monitor *mon, const char *name)
1061 {
1062     BlockDriverState *bs, *bs1;
1063     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1064     int must_delete, ret, i;
1065     BlockDriverInfo bdi1, *bdi = &bdi1;
1066     QEMUFile *f;
1067     int saved_vm_running;
1068     uint32_t vm_state_size;
1069 #ifdef _WIN32
1070     struct _timeb tb;
1071 #else
1072     struct timeval tv;
1073 #endif
1074
1075     bs = get_bs_snapshots();
1076     if (!bs) {
1077         monitor_printf(mon, "No block device can accept snapshots\n");
1078         return;
1079     }
1080
1081     /* ??? Should this occur after vm_stop?  */
1082     qemu_aio_flush();
1083
1084     saved_vm_running = vm_running;
1085     vm_stop(0);
1086
1087     must_delete = 0;
1088     if (name) {
1089         ret = bdrv_snapshot_find(bs, old_sn, name);
1090         if (ret >= 0) {
1091             must_delete = 1;
1092         }
1093     }
1094     memset(sn, 0, sizeof(*sn));
1095     if (must_delete) {
1096         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1097         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1098     } else {
1099         if (name)
1100             pstrcpy(sn->name, sizeof(sn->name), name);
1101     }
1102
1103     /* fill auxiliary fields */
1104 #ifdef _WIN32
1105     _ftime(&tb);
1106     sn->date_sec = tb.time;
1107     sn->date_nsec = tb.millitm * 1000000;
1108 #else
1109     gettimeofday(&tv, NULL);
1110     sn->date_sec = tv.tv_sec;
1111     sn->date_nsec = tv.tv_usec * 1000;
1112 #endif
1113     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1114
1115     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1116         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1117                        bdrv_get_device_name(bs));
1118         goto the_end;
1119     }
1120
1121     /* save the VM state */
1122     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 1);
1123     if (!f) {
1124         monitor_printf(mon, "Could not open VM state file\n");
1125         goto the_end;
1126     }
1127     ret = qemu_savevm_state(f);
1128     vm_state_size = qemu_ftell(f);
1129     qemu_fclose(f);
1130     if (ret < 0) {
1131         monitor_printf(mon, "Error %d while writing VM\n", ret);
1132         goto the_end;
1133     }
1134
1135     /* create the snapshots */
1136
1137     for(i = 0; i < nb_drives; i++) {
1138         bs1 = drives_table[i].bdrv;
1139         if (bdrv_has_snapshot(bs1)) {
1140             if (must_delete) {
1141                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1142                 if (ret < 0) {
1143                     monitor_printf(mon,
1144                                    "Error while deleting snapshot on '%s'\n",
1145                                    bdrv_get_device_name(bs1));
1146                 }
1147             }
1148             /* Write VM state size only to the image that contains the state */
1149             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1150             ret = bdrv_snapshot_create(bs1, sn);
1151             if (ret < 0) {
1152                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1153                                bdrv_get_device_name(bs1));
1154             }
1155         }
1156     }
1157
1158  the_end:
1159     if (saved_vm_running)
1160         vm_start();
1161 }
1162
1163 void do_loadvm(Monitor *mon, const char *name)
1164 {
1165     BlockDriverState *bs, *bs1;
1166     BlockDriverInfo bdi1, *bdi = &bdi1;
1167     QEMUSnapshotInfo sn;
1168     QEMUFile *f;
1169     int i, ret;
1170     int saved_vm_running;
1171
1172     bs = get_bs_snapshots();
1173     if (!bs) {
1174         monitor_printf(mon, "No block device supports snapshots\n");
1175         return;
1176     }
1177
1178     /* Flush all IO requests so they don't interfere with the new state.  */
1179     qemu_aio_flush();
1180
1181     saved_vm_running = vm_running;
1182     vm_stop(0);
1183
1184     for(i = 0; i <= nb_drives; i++) {
1185         bs1 = drives_table[i].bdrv;
1186         if (bdrv_has_snapshot(bs1)) {
1187             ret = bdrv_snapshot_goto(bs1, name);
1188             if (ret < 0) {
1189                 if (bs != bs1)
1190                     monitor_printf(mon, "Warning: ");
1191                 switch(ret) {
1192                 case -ENOTSUP:
1193                     monitor_printf(mon,
1194                                    "Snapshots not supported on device '%s'\n",
1195                                    bdrv_get_device_name(bs1));
1196                     break;
1197                 case -ENOENT:
1198                     monitor_printf(mon, "Could not find snapshot '%s' on "
1199                                    "device '%s'\n",
1200                                    name, bdrv_get_device_name(bs1));
1201                     break;
1202                 default:
1203                     monitor_printf(mon, "Error %d while activating snapshot on"
1204                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1205                     break;
1206                 }
1207                 /* fatal on snapshot block device */
1208                 if (bs == bs1)
1209                     goto the_end;
1210             }
1211         }
1212     }
1213
1214     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1215         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1216                        bdrv_get_device_name(bs));
1217         return;
1218     }
1219
1220     /* Don't even try to load empty VM states */
1221     ret = bdrv_snapshot_find(bs, &sn, name);
1222     if ((ret >= 0) && (sn.vm_state_size == 0))
1223         goto the_end;
1224
1225     /* restore the VM state */
1226     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 0);
1227     if (!f) {
1228         monitor_printf(mon, "Could not open VM state file\n");
1229         goto the_end;
1230     }
1231     ret = qemu_loadvm_state(f);
1232     qemu_fclose(f);
1233     if (ret < 0) {
1234         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1235     }
1236  the_end:
1237     if (saved_vm_running)
1238         vm_start();
1239 }
1240
1241 void do_delvm(Monitor *mon, const char *name)
1242 {
1243     BlockDriverState *bs, *bs1;
1244     int i, ret;
1245
1246     bs = get_bs_snapshots();
1247     if (!bs) {
1248         monitor_printf(mon, "No block device supports snapshots\n");
1249         return;
1250     }
1251
1252     for(i = 0; i <= nb_drives; i++) {
1253         bs1 = drives_table[i].bdrv;
1254         if (bdrv_has_snapshot(bs1)) {
1255             ret = bdrv_snapshot_delete(bs1, name);
1256             if (ret < 0) {
1257                 if (ret == -ENOTSUP)
1258                     monitor_printf(mon,
1259                                    "Snapshots not supported on device '%s'\n",
1260                                    bdrv_get_device_name(bs1));
1261                 else
1262                     monitor_printf(mon, "Error %d while deleting snapshot on "
1263                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1264             }
1265         }
1266     }
1267 }
1268
1269 void do_info_snapshots(Monitor *mon)
1270 {
1271     BlockDriverState *bs, *bs1;
1272     QEMUSnapshotInfo *sn_tab, *sn;
1273     int nb_sns, i;
1274     char buf[256];
1275
1276     bs = get_bs_snapshots();
1277     if (!bs) {
1278         monitor_printf(mon, "No available block device supports snapshots\n");
1279         return;
1280     }
1281     monitor_printf(mon, "Snapshot devices:");
1282     for(i = 0; i <= nb_drives; i++) {
1283         bs1 = drives_table[i].bdrv;
1284         if (bdrv_has_snapshot(bs1)) {
1285             if (bs == bs1)
1286                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1287         }
1288     }
1289     monitor_printf(mon, "\n");
1290
1291     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1292     if (nb_sns < 0) {
1293         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1294         return;
1295     }
1296     monitor_printf(mon, "Snapshot list (from %s):\n",
1297                    bdrv_get_device_name(bs));
1298     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1299     for(i = 0; i < nb_sns; i++) {
1300         sn = &sn_tab[i];
1301         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1302     }
1303     qemu_free(sn_tab);
1304 }