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, (void *)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     FILE *fp = s->popen_file;
 219     int bytes;
 220
 221     do {
 222         clearerr(fp);
 223         bytes = fread(buf, 1, size, fp);
 224     } while ((bytes == 0) && ferror(fp) && (errno == EINTR));
 225     return bytes;
 226 }
 227
 228 static int popen_close(void *opaque)
 229 {
 230     QEMUFilePopen *s = opaque;
 231     pclose(s->popen_file);
 232     qemu_free(s);
 233     return 0;
 234 }
 235
 236 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 237 {
 238     QEMUFilePopen *s;
 239
 240     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 241         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 242         return NULL;
 243     }
 244
 245     s = qemu_mallocz(sizeof(QEMUFilePopen));
 246
 247     s->popen_file = popen_file;
 248
 249     if(mode[0] == 'r') {
 250         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL, NULL);
 251     } else {
 252         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL, NULL);
 253     }
 254     return s->file;
 255 }
 256
 257 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 258 {
 259     FILE *popen_file;
 260
 261     popen_file = popen(command, mode);
 262     if(popen_file == NULL) {
 263         return NULL;
 264     }
 265
 266     return qemu_popen(popen_file, mode);
 267 }
 268
 269 int qemu_popen_fd(QEMUFile *f)
 270 {
 271     QEMUFilePopen *p;
 272     int fd;
 273
 274     p = (QEMUFilePopen *)f->opaque;
 275     fd = fileno(p->popen_file);
 276
 277     return fd;
 278 }
 279
 280 QEMUFile *qemu_fopen_socket(int fd)
 281 {
 282     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 283
 284     s->fd = fd;
 285     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL, NULL);
 286     return s->file;
 287 }
 288
 289 typedef struct QEMUFileStdio
 290 {
 291     FILE *outfile;
 292 } QEMUFileStdio;
 293
 294 static int file_put_buffer(void *opaque, const uint8_t *buf,
 295                             int64_t pos, int size)
 296 {
 297     QEMUFileStdio *s = opaque;
 298     fseek(s->outfile, pos, SEEK_SET);
 299     fwrite(buf, 1, size, s->outfile);
 300     return size;
 301 }
 302
 303 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 304 {
 305     QEMUFileStdio *s = opaque;
 306     fseek(s->outfile, pos, SEEK_SET);
 307     return fread(buf, 1, size, s->outfile);
 308 }
 309
 310 static int file_close(void *opaque)
 311 {
 312     QEMUFileStdio *s = opaque;
 313     fclose(s->outfile);
 314     qemu_free(s);
 315     return 0;
 316 }
 317
 318 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 319 {
 320     QEMUFileStdio *s;
 321
 322     s = qemu_mallocz(sizeof(QEMUFileStdio));
 323
 324     s->outfile = fopen(filename, mode);
 325     if (!s->outfile)
 326         goto fail;
 327
 328     if (!strcmp(mode, "wb"))
 329         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL, NULL);
 330     else if (!strcmp(mode, "rb"))
 331         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL, NULL);
 332
 333 fail:
 334     if (s->outfile)
 335         fclose(s->outfile);
 336     qemu_free(s);
 337     return NULL;
 338 }
 339
 340 typedef struct QEMUFileBdrv
 341 {
 342     BlockDriverState *bs;
 343     int64_t base_offset;
 344 } QEMUFileBdrv;
 345
 346 static int block_put_buffer(void *opaque, const uint8_t *buf,
 347                            int64_t pos, int size)
 348 {
 349     QEMUFileBdrv *s = opaque;
 350     bdrv_put_buffer(s->bs, buf, s->base_offset + pos, size);
 351     return size;
 352 }
 353
 354 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 355 {
 356     QEMUFileBdrv *s = opaque;
 357     return bdrv_get_buffer(s->bs, buf, s->base_offset + pos, size);
 358 }
 359
 360 static int bdrv_fclose(void *opaque)
 361 {
 362     QEMUFileBdrv *s = opaque;
 363     qemu_free(s);
 364     return 0;
 365 }
 366
 367 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable)
 368 {
 369     QEMUFileBdrv *s;
 370
 371     s = qemu_mallocz(sizeof(QEMUFileBdrv));
 372
 373     s->bs = bs;
 374     s->base_offset = offset;
 375
 376     if (is_writable)
 377         return qemu_fopen_ops(s, block_put_buffer, NULL, bdrv_fclose, NULL, NULL);
 378
 379     return qemu_fopen_ops(s, NULL, block_get_buffer, bdrv_fclose, NULL, NULL);
 380 }
 381
 382 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 383                          QEMUFileGetBufferFunc *get_buffer,
 384                          QEMUFileCloseFunc *close,
 385                          QEMUFileRateLimit *rate_limit,
 386                          QEMUFileSetRateLimit *set_rate_limit)
 387 {
 388     QEMUFile *f;
 389
 390     f = qemu_mallocz(sizeof(QEMUFile));
 391
 392     f->opaque = opaque;
 393     f->put_buffer = put_buffer;
 394     f->get_buffer = get_buffer;
 395     f->close = close;
 396     f->rate_limit = rate_limit;
 397     f->set_rate_limit = set_rate_limit;
 398     f->is_write = 0;
 399
 400     return f;
 401 }
 402
 403 int qemu_file_has_error(QEMUFile *f)
 404 {
 405     return f->has_error;
 406 }
 407
 408 void qemu_file_set_error(QEMUFile *f)
 409 {
 410     f->has_error = 1;
 411 }
 412
 413 void qemu_fflush(QEMUFile *f)
 414 {
 415     if (!f->put_buffer)
 416         return;
 417
 418     if (f->is_write && f->buf_index > 0) {
 419         int len;
 420
 421         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 422         if (len > 0)
 423             f->buf_offset += f->buf_index;
 424         else
 425             f->has_error = 1;
 426         f->buf_index = 0;
 427     }
 428 }
 429
 430 static void qemu_fill_buffer(QEMUFile *f)
 431 {
 432     int len;
 433
 434     if (!f->get_buffer)
 435         return;
 436
 437     if (f->is_write)
 438         abort();
 439
 440     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 441     if (len > 0) {
 442         f->buf_index = 0;
 443         f->buf_size = len;
 444         f->buf_offset += len;
 445     } else if (len != -EAGAIN)
 446         f->has_error = 1;
 447 }
 448
 449 int qemu_fclose(QEMUFile *f)
 450 {
 451     int ret = 0;
 452     qemu_fflush(f);
 453     if (f->close)
 454         ret = f->close(f->opaque);
 455     qemu_free(f);
 456     return ret;
 457 }
 458
 459 void qemu_file_put_notify(QEMUFile *f)
 460 {
 461     f->put_buffer(f->opaque, NULL, 0, 0);
 462 }
 463
 464 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 465 {
 466     int l;
 467
 468     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 469         fprintf(stderr,
 470                 "Attempted to write to buffer while read buffer is not empty\n");
 471         abort();
 472     }
 473
 474     while (!f->has_error && size > 0) {
 475         l = IO_BUF_SIZE - f->buf_index;
 476         if (l > size)
 477             l = size;
 478         memcpy(f->buf + f->buf_index, buf, l);
 479         f->is_write = 1;
 480         f->buf_index += l;
 481         buf += l;
 482         size -= l;
 483         if (f->buf_index >= IO_BUF_SIZE)
 484             qemu_fflush(f);
 485     }
 486 }
 487
 488 void qemu_put_byte(QEMUFile *f, int v)
 489 {
 490     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 491         fprintf(stderr,
 492                 "Attempted to write to buffer while read buffer is not empty\n");
 493         abort();
 494     }
 495
 496     f->buf[f->buf_index++] = v;
 497     f->is_write = 1;
 498     if (f->buf_index >= IO_BUF_SIZE)
 499         qemu_fflush(f);
 500 }
 501
 502 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 503 {
 504     int size, l;
 505
 506     if (f->is_write)
 507         abort();
 508
 509     size = size1;
 510     while (size > 0) {
 511         l = f->buf_size - f->buf_index;
 512         if (l == 0) {
 513             qemu_fill_buffer(f);
 514             l = f->buf_size - f->buf_index;
 515             if (l == 0)
 516                 break;
 517         }
 518         if (l > size)
 519             l = size;
 520         memcpy(buf, f->buf + f->buf_index, l);
 521         f->buf_index += l;
 522         buf += l;
 523         size -= l;
 524     }
 525     return size1 - size;
 526 }
 527
 528 int qemu_get_byte(QEMUFile *f)
 529 {
 530     if (f->is_write)
 531         abort();
 532
 533     if (f->buf_index >= f->buf_size) {
 534         qemu_fill_buffer(f);
 535         if (f->buf_index >= f->buf_size)
 536             return 0;
 537     }
 538     return f->buf[f->buf_index++];
 539 }
 540
 541 int64_t qemu_ftell(QEMUFile *f)
 542 {
 543     return f->buf_offset - f->buf_size + f->buf_index;
 544 }
 545
 546 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 547 {
 548     if (whence == SEEK_SET) {
 549         /* nothing to do */
 550     } else if (whence == SEEK_CUR) {
 551         pos += qemu_ftell(f);
 552     } else {
 553         /* SEEK_END not supported */
 554         return -1;
 555     }
 556     if (f->put_buffer) {
 557         qemu_fflush(f);
 558         f->buf_offset = pos;
 559     } else {
 560         f->buf_offset = pos;
 561         f->buf_index = 0;
 562         f->buf_size = 0;
 563     }
 564     return pos;
 565 }
 566
 567 int qemu_file_rate_limit(QEMUFile *f)
 568 {
 569     if (f->rate_limit)
 570         return f->rate_limit(f->opaque);
 571
 572     return 0;
 573 }
 574
 575 size_t qemu_file_set_rate_limit(QEMUFile *f, size_t new_rate)
 576 {
 577     if (f->set_rate_limit)
 578         return f->set_rate_limit(f->opaque, new_rate);
 579
 580     return 0;
 581 }
 582
 583 void qemu_put_be16(QEMUFile *f, unsigned int v)
 584 {
 585     qemu_put_byte(f, v >> 8);
 586     qemu_put_byte(f, v);
 587 }
 588
 589 void qemu_put_be32(QEMUFile *f, unsigned int v)
 590 {
 591     qemu_put_byte(f, v >> 24);
 592     qemu_put_byte(f, v >> 16);
 593     qemu_put_byte(f, v >> 8);
 594     qemu_put_byte(f, v);
 595 }
 596
 597 void qemu_put_be64(QEMUFile *f, uint64_t v)
 598 {
 599     qemu_put_be32(f, v >> 32);
 600     qemu_put_be32(f, v);
 601 }
 602
 603 unsigned int qemu_get_be16(QEMUFile *f)
 604 {
 605     unsigned int v;
 606     v = qemu_get_byte(f) << 8;
 607     v |= qemu_get_byte(f);
 608     return v;
 609 }
 610
 611 unsigned int qemu_get_be32(QEMUFile *f)
 612 {
 613     unsigned int v;
 614     v = qemu_get_byte(f) << 24;
 615     v |= qemu_get_byte(f) << 16;
 616     v |= qemu_get_byte(f) << 8;
 617     v |= qemu_get_byte(f);
 618     return v;
 619 }
 620
 621 uint64_t qemu_get_be64(QEMUFile *f)
 622 {
 623     uint64_t v;
 624     v = (uint64_t)qemu_get_be32(f) << 32;
 625     v |= qemu_get_be32(f);
 626     return v;
 627 }
 628
 629 typedef struct SaveStateEntry {
 630     char idstr[256];
 631     int instance_id;
 632     int version_id;
 633     int section_id;
 634     SaveLiveStateHandler *save_live_state;
 635     SaveStateHandler *save_state;
 636     LoadStateHandler *load_state;
 637     void *opaque;
 638     struct SaveStateEntry *next;
 639 } SaveStateEntry;
 640
 641 static SaveStateEntry *first_se;
 642
 643 /* TODO: Individual devices generally have very little idea about the rest
 644    of the system, so instance_id should be removed/replaced.
 645    Meanwhile pass -1 as instance_id if you do not already have a clearly
 646    distinguishing id for all instances of your device class. */
 647 int register_savevm_live(const char *idstr,
 648                          int instance_id,
 649                          int version_id,
 650                          SaveLiveStateHandler *save_live_state,
 651                          SaveStateHandler *save_state,
 652                          LoadStateHandler *load_state,
 653                          void *opaque)
 654 {
 655     SaveStateEntry *se, **pse;
 656     static int global_section_id;
 657
 658     se = qemu_malloc(sizeof(SaveStateEntry));
 659     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 660     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 661     se->version_id = version_id;
 662     se->section_id = global_section_id++;
 663     se->save_live_state = save_live_state;
 664     se->save_state = save_state;
 665     se->load_state = load_state;
 666     se->opaque = opaque;
 667     se->next = NULL;
 668
 669     /* add at the end of list */
 670     pse = &first_se;
 671     while (*pse != NULL) {
 672         if (instance_id == -1
 673                 && strcmp(se->idstr, (*pse)->idstr) == 0
 674                 && se->instance_id <= (*pse)->instance_id)
 675             se->instance_id = (*pse)->instance_id + 1;
 676         pse = &(*pse)->next;
 677     }
 678     *pse = se;
 679     return 0;
 680 }
 681
 682 int register_savevm(const char *idstr,
 683                     int instance_id,
 684                     int version_id,
 685                     SaveStateHandler *save_state,
 686                     LoadStateHandler *load_state,
 687                     void *opaque)
 688 {
 689     return register_savevm_live(idstr, instance_id, version_id,
 690                                 NULL, save_state, load_state, opaque);
 691 }
 692
 693 void unregister_savevm(const char *idstr, void *opaque)
 694 {
 695     SaveStateEntry **pse;
 696
 697     pse = &first_se;
 698     while (*pse != NULL) {
 699         if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
 700             SaveStateEntry *next = (*pse)->next;
 701             qemu_free(*pse);
 702             *pse = next;
 703             continue;
 704         }
 705         pse = &(*pse)->next;
 706     }
 707 }
 708
 709 #define QEMU_VM_FILE_MAGIC           0x5145564d
 710 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 711 #define QEMU_VM_FILE_VERSION         0x00000003
 712
 713 #define QEMU_VM_EOF                  0x00
 714 #define QEMU_VM_SECTION_START        0x01
 715 #define QEMU_VM_SECTION_PART         0x02
 716 #define QEMU_VM_SECTION_END          0x03
 717 #define QEMU_VM_SECTION_FULL         0x04
 718
 719 int qemu_savevm_state_begin(QEMUFile *f)
 720 {
 721     SaveStateEntry *se;
 722
 723     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 724     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 725
 726     for (se = first_se; se != NULL; se = se->next) {
 727         int len;
 728
 729         if (se->save_live_state == NULL)
 730             continue;
 731
 732         /* Section type */
 733         qemu_put_byte(f, QEMU_VM_SECTION_START);
 734         qemu_put_be32(f, se->section_id);
 735
 736         /* ID string */
 737         len = strlen(se->idstr);
 738         qemu_put_byte(f, len);
 739         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 740
 741         qemu_put_be32(f, se->instance_id);
 742         qemu_put_be32(f, se->version_id);
 743
 744         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 745     }
 746
 747     if (qemu_file_has_error(f))
 748         return -EIO;
 749
 750     return 0;
 751 }
 752
 753 int qemu_savevm_state_iterate(QEMUFile *f)
 754 {
 755     SaveStateEntry *se;
 756     int ret = 1;
 757
 758     for (se = first_se; se != NULL; se = se->next) {
 759         if (se->save_live_state == NULL)
 760             continue;
 761
 762         /* Section type */
 763         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 764         qemu_put_be32(f, se->section_id);
 765
 766         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 767     }
 768
 769     if (ret)
 770         return 1;
 771
 772     if (qemu_file_has_error(f))
 773         return -EIO;
 774
 775     return 0;
 776 }
 777
 778 int qemu_savevm_state_complete(QEMUFile *f)
 779 {
 780     SaveStateEntry *se;
 781
 782     for (se = first_se; se != NULL; se = se->next) {
 783         if (se->save_live_state == NULL)
 784             continue;
 785
 786         /* Section type */
 787         qemu_put_byte(f, QEMU_VM_SECTION_END);
 788         qemu_put_be32(f, se->section_id);
 789
 790         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 791     }
 792
 793     for(se = first_se; se != NULL; se = se->next) {
 794         int len;
 795
 796         if (se->save_state == NULL)
 797             continue;
 798
 799         /* Section type */
 800         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 801         qemu_put_be32(f, se->section_id);
 802
 803         /* ID string */
 804         len = strlen(se->idstr);
 805         qemu_put_byte(f, len);
 806         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 807
 808         qemu_put_be32(f, se->instance_id);
 809         qemu_put_be32(f, se->version_id);
 810
 811         se->save_state(f, se->opaque);
 812     }
 813
 814     qemu_put_byte(f, QEMU_VM_EOF);
 815
 816     if (qemu_file_has_error(f))
 817         return -EIO;
 818
 819     return 0;
 820 }
 821
 822 int qemu_savevm_state(QEMUFile *f)
 823 {
 824     int saved_vm_running;
 825     int ret;
 826
 827     saved_vm_running = vm_running;
 828     vm_stop(0);
 829
 830     bdrv_flush_all();
 831
 832     ret = qemu_savevm_state_begin(f);
 833     if (ret < 0)
 834         goto out;
 835
 836     do {
 837         ret = qemu_savevm_state_iterate(f);
 838         if (ret < 0)
 839             goto out;
 840     } while (ret == 0);
 841
 842     ret = qemu_savevm_state_complete(f);
 843
 844 out:
 845     if (qemu_file_has_error(f))
 846         ret = -EIO;
 847
 848     if (!ret && saved_vm_running)
 849         vm_start();
 850
 851     return ret;
 852 }
 853
 854 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 855 {
 856     SaveStateEntry *se;
 857
 858     for(se = first_se; se != NULL; se = se->next) {
 859         if (!strcmp(se->idstr, idstr) &&
 860             instance_id == se->instance_id)
 861             return se;
 862     }
 863     return NULL;
 864 }
 865
 866 typedef struct LoadStateEntry {
 867     SaveStateEntry *se;
 868     int section_id;
 869     int version_id;
 870     struct LoadStateEntry *next;
 871 } LoadStateEntry;
 872
 873 static int qemu_loadvm_state_v2(QEMUFile *f)
 874 {
 875     SaveStateEntry *se;
 876     int len, ret, instance_id, record_len, version_id;
 877     int64_t total_len, end_pos, cur_pos;
 878     char idstr[256];
 879
 880     total_len = qemu_get_be64(f);
 881     end_pos = total_len + qemu_ftell(f);
 882     for(;;) {
 883         if (qemu_ftell(f) >= end_pos)
 884             break;
 885         len = qemu_get_byte(f);
 886         qemu_get_buffer(f, (uint8_t *)idstr, len);
 887         idstr[len] = '\0';
 888         instance_id = qemu_get_be32(f);
 889         version_id = qemu_get_be32(f);
 890         record_len = qemu_get_be32(f);
 891         cur_pos = qemu_ftell(f);
 892         se = find_se(idstr, instance_id);
 893         if (!se) {
 894             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 895                     instance_id, idstr);
 896         } else {
 897             ret = se->load_state(f, se->opaque, version_id);
 898             if (ret < 0) {
 899                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 900                         instance_id, idstr);
 901                 return ret;
 902             }
 903         }
 904         /* always seek to exact end of record */
 905         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 906     }
 907
 908     if (qemu_file_has_error(f))
 909         return -EIO;
 910
 911     return 0;
 912 }
 913
 914 int qemu_loadvm_state(QEMUFile *f)
 915 {
 916     LoadStateEntry *first_le = NULL;
 917     uint8_t section_type;
 918     unsigned int v;
 919     int ret;
 920
 921     v = qemu_get_be32(f);
 922     if (v != QEMU_VM_FILE_MAGIC)
 923         return -EINVAL;
 924
 925     v = qemu_get_be32(f);
 926     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 927         return qemu_loadvm_state_v2(f);
 928     if (v != QEMU_VM_FILE_VERSION)
 929         return -ENOTSUP;
 930
 931     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 932         uint32_t instance_id, version_id, section_id;
 933         LoadStateEntry *le;
 934         SaveStateEntry *se;
 935         char idstr[257];
 936         int len;
 937
 938         switch (section_type) {
 939         case QEMU_VM_SECTION_START:
 940         case QEMU_VM_SECTION_FULL:
 941             /* Read section start */
 942             section_id = qemu_get_be32(f);
 943             len = qemu_get_byte(f);
 944             qemu_get_buffer(f, (uint8_t *)idstr, len);
 945             idstr[len] = 0;
 946             instance_id = qemu_get_be32(f);
 947             version_id = qemu_get_be32(f);
 948
 949             /* Find savevm section */
 950             se = find_se(idstr, instance_id);
 951             if (se == NULL) {
 952                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 953                 ret = -EINVAL;
 954                 goto out;
 955             }
 956
 957             /* Validate version */
 958             if (version_id > se->version_id) {
 959                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 960                         version_id, idstr, se->version_id);
 961                 ret = -EINVAL;
 962                 goto out;
 963             }
 964
 965             /* Add entry */
 966             le = qemu_mallocz(sizeof(*le));
 967
 968             le->se = se;
 969             le->section_id = section_id;
 970             le->version_id = version_id;
 971             le->next = first_le;
 972             first_le = le;
 973
 974             le->se->load_state(f, le->se->opaque, le->version_id);
 975             break;
 976         case QEMU_VM_SECTION_PART:
 977         case QEMU_VM_SECTION_END:
 978             section_id = qemu_get_be32(f);
 979
 980             for (le = first_le; le && le->section_id != section_id; le = le->next);
 981             if (le == NULL) {
 982                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 983                 ret = -EINVAL;
 984                 goto out;
 985             }
 986
 987             le->se->load_state(f, le->se->opaque, le->version_id);
 988             break;
 989         default:
 990             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 991             ret = -EINVAL;
 992             goto out;
 993         }
 994     }
 995
 996     ret = 0;
 997
 998 out:
 999     while (first_le) {
1000         LoadStateEntry *le = first_le;
1001         first_le = first_le->next;
1002         qemu_free(le);
1003     }
1004
1005     if (qemu_file_has_error(f))
1006         ret = -EIO;
1007
1008     return ret;
1009 }
1010
1011 /* device can contain snapshots */
1012 static int bdrv_can_snapshot(BlockDriverState *bs)
1013 {
1014     return (bs &&
1015             !bdrv_is_removable(bs) &&
1016             !bdrv_is_read_only(bs));
1017 }
1018
1019 /* device must be snapshots in order to have a reliable snapshot */
1020 static int bdrv_has_snapshot(BlockDriverState *bs)
1021 {
1022     return (bs &&
1023             !bdrv_is_removable(bs) &&
1024             !bdrv_is_read_only(bs));
1025 }
1026
1027 static BlockDriverState *get_bs_snapshots(void)
1028 {
1029     BlockDriverState *bs;
1030     int i;
1031
1032     if (bs_snapshots)
1033         return bs_snapshots;
1034     for(i = 0; i <= nb_drives; i++) {
1035         bs = drives_table[i].bdrv;
1036         if (bdrv_can_snapshot(bs))
1037             goto ok;
1038     }
1039     return NULL;
1040  ok:
1041     bs_snapshots = bs;
1042     return bs;
1043 }
1044
1045 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1046                               const char *name)
1047 {
1048     QEMUSnapshotInfo *sn_tab, *sn;
1049     int nb_sns, i, ret;
1050
1051     ret = -ENOENT;
1052     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1053     if (nb_sns < 0)
1054         return ret;
1055     for(i = 0; i < nb_sns; i++) {
1056         sn = &sn_tab[i];
1057         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1058             *sn_info = *sn;
1059             ret = 0;
1060             break;
1061         }
1062     }
1063     qemu_free(sn_tab);
1064     return ret;
1065 }
1066
1067 void do_savevm(Monitor *mon, const char *name)
1068 {
1069     BlockDriverState *bs, *bs1;
1070     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1071     int must_delete, ret, i;
1072     BlockDriverInfo bdi1, *bdi = &bdi1;
1073     QEMUFile *f;
1074     int saved_vm_running;
1075     uint32_t vm_state_size;
1076 #ifdef _WIN32
1077     struct _timeb tb;
1078 #else
1079     struct timeval tv;
1080 #endif
1081
1082     bs = get_bs_snapshots();
1083     if (!bs) {
1084         monitor_printf(mon, "No block device can accept snapshots\n");
1085         return;
1086     }
1087
1088     /* ??? Should this occur after vm_stop?  */
1089     qemu_aio_flush();
1090
1091     saved_vm_running = vm_running;
1092     vm_stop(0);
1093
1094     must_delete = 0;
1095     if (name) {
1096         ret = bdrv_snapshot_find(bs, old_sn, name);
1097         if (ret >= 0) {
1098             must_delete = 1;
1099         }
1100     }
1101     memset(sn, 0, sizeof(*sn));
1102     if (must_delete) {
1103         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1104         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1105     } else {
1106         if (name)
1107             pstrcpy(sn->name, sizeof(sn->name), name);
1108     }
1109
1110     /* fill auxiliary fields */
1111 #ifdef _WIN32
1112     _ftime(&tb);
1113     sn->date_sec = tb.time;
1114     sn->date_nsec = tb.millitm * 1000000;
1115 #else
1116     gettimeofday(&tv, NULL);
1117     sn->date_sec = tv.tv_sec;
1118     sn->date_nsec = tv.tv_usec * 1000;
1119 #endif
1120     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1121
1122     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1123         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1124                        bdrv_get_device_name(bs));
1125         goto the_end;
1126     }
1127
1128     /* save the VM state */
1129     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 1);
1130     if (!f) {
1131         monitor_printf(mon, "Could not open VM state file\n");
1132         goto the_end;
1133     }
1134     ret = qemu_savevm_state(f);
1135     vm_state_size = qemu_ftell(f);
1136     qemu_fclose(f);
1137     if (ret < 0) {
1138         monitor_printf(mon, "Error %d while writing VM\n", ret);
1139         goto the_end;
1140     }
1141
1142     /* create the snapshots */
1143
1144     for(i = 0; i < nb_drives; i++) {
1145         bs1 = drives_table[i].bdrv;
1146         if (bdrv_has_snapshot(bs1)) {
1147             if (must_delete) {
1148                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1149                 if (ret < 0) {
1150                     monitor_printf(mon,
1151                                    "Error while deleting snapshot on '%s'\n",
1152                                    bdrv_get_device_name(bs1));
1153                 }
1154             }
1155             /* Write VM state size only to the image that contains the state */
1156             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1157             ret = bdrv_snapshot_create(bs1, sn);
1158             if (ret < 0) {
1159                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1160                                bdrv_get_device_name(bs1));
1161             }
1162         }
1163     }
1164
1165  the_end:
1166     if (saved_vm_running)
1167         vm_start();
1168 }
1169
1170 void do_loadvm(Monitor *mon, const char *name)
1171 {
1172     BlockDriverState *bs, *bs1;
1173     BlockDriverInfo bdi1, *bdi = &bdi1;
1174     QEMUSnapshotInfo sn;
1175     QEMUFile *f;
1176     int i, ret;
1177     int saved_vm_running;
1178
1179     bs = get_bs_snapshots();
1180     if (!bs) {
1181         monitor_printf(mon, "No block device supports snapshots\n");
1182         return;
1183     }
1184
1185     /* Flush all IO requests so they don't interfere with the new state.  */
1186     qemu_aio_flush();
1187
1188     saved_vm_running = vm_running;
1189     vm_stop(0);
1190
1191     for(i = 0; i <= nb_drives; i++) {
1192         bs1 = drives_table[i].bdrv;
1193         if (bdrv_has_snapshot(bs1)) {
1194             ret = bdrv_snapshot_goto(bs1, name);
1195             if (ret < 0) {
1196                 if (bs != bs1)
1197                     monitor_printf(mon, "Warning: ");
1198                 switch(ret) {
1199                 case -ENOTSUP:
1200                     monitor_printf(mon,
1201                                    "Snapshots not supported on device '%s'\n",
1202                                    bdrv_get_device_name(bs1));
1203                     break;
1204                 case -ENOENT:
1205                     monitor_printf(mon, "Could not find snapshot '%s' on "
1206                                    "device '%s'\n",
1207                                    name, bdrv_get_device_name(bs1));
1208                     break;
1209                 default:
1210                     monitor_printf(mon, "Error %d while activating snapshot on"
1211                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1212                     break;
1213                 }
1214                 /* fatal on snapshot block device */
1215                 if (bs == bs1)
1216                     goto the_end;
1217             }
1218         }
1219     }
1220
1221     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1222         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1223                        bdrv_get_device_name(bs));
1224         return;
1225     }
1226
1227     /* Don't even try to load empty VM states */
1228     ret = bdrv_snapshot_find(bs, &sn, name);
1229     if ((ret >= 0) && (sn.vm_state_size == 0))
1230         goto the_end;
1231
1232     /* restore the VM state */
1233     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 0);
1234     if (!f) {
1235         monitor_printf(mon, "Could not open VM state file\n");
1236         goto the_end;
1237     }
1238     ret = qemu_loadvm_state(f);
1239     qemu_fclose(f);
1240     if (ret < 0) {
1241         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1242     }
1243  the_end:
1244     if (saved_vm_running)
1245         vm_start();
1246 }
1247
1248 void do_delvm(Monitor *mon, const char *name)
1249 {
1250     BlockDriverState *bs, *bs1;
1251     int i, ret;
1252
1253     bs = get_bs_snapshots();
1254     if (!bs) {
1255         monitor_printf(mon, "No block device supports snapshots\n");
1256         return;
1257     }
1258
1259     for(i = 0; i <= nb_drives; i++) {
1260         bs1 = drives_table[i].bdrv;
1261         if (bdrv_has_snapshot(bs1)) {
1262             ret = bdrv_snapshot_delete(bs1, name);
1263             if (ret < 0) {
1264                 if (ret == -ENOTSUP)
1265                     monitor_printf(mon,
1266                                    "Snapshots not supported on device '%s'\n",
1267                                    bdrv_get_device_name(bs1));
1268                 else
1269                     monitor_printf(mon, "Error %d while deleting snapshot on "
1270                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1271             }
1272         }
1273     }
1274 }
1275
1276 void do_info_snapshots(Monitor *mon)
1277 {
1278     BlockDriverState *bs, *bs1;
1279     QEMUSnapshotInfo *sn_tab, *sn;
1280     int nb_sns, i;
1281     char buf[256];
1282
1283     bs = get_bs_snapshots();
1284     if (!bs) {
1285         monitor_printf(mon, "No available block device supports snapshots\n");
1286         return;
1287     }
1288     monitor_printf(mon, "Snapshot devices:");
1289     for(i = 0; i <= nb_drives; i++) {
1290         bs1 = drives_table[i].bdrv;
1291         if (bdrv_has_snapshot(bs1)) {
1292             if (bs == bs1)
1293                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1294         }
1295     }
1296     monitor_printf(mon, "\n");
1297
1298     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1299     if (nb_sns < 0) {
1300         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1301         return;
1302     }
1303     monitor_printf(mon, "Snapshot list (from %s):\n",
1304                    bdrv_get_device_name(bs));
1305     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1306     for(i = 0; i < nb_sns; i++) {
1307         sn = &sn_tab[i];
1308         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1309     }
1310     qemu_free(sn_tab);
1311 }