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 "qemu-common.h"
  25 #include "hw/hw.h"
  26 #include "net.h"
  27 #include "console.h"
  28 #include "sysemu.h"
  29 #include "qemu-timer.h"
  30 #include "qemu-char.h"
  31 #include "block.h"
  32 #include "audio/audio.h"
  33 #include "migration.h"
  34 #include "qemu_socket.h"
  35
  36 #include <unistd.h>
  37 #include <fcntl.h>
  38 #include <signal.h>
  39 #include <time.h>
  40 #include <errno.h>
  41 #include <sys/time.h>
  42 #include <zlib.h>
  43
  44 #ifndef _WIN32
  45 #include <sys/times.h>
  46 #include <sys/wait.h>
  47 #include <termios.h>
  48 #include <sys/mman.h>
  49 #include <sys/ioctl.h>
  50 #include <sys/resource.h>
  51 #include <sys/socket.h>
  52 #include <netinet/in.h>
  53 #include <net/if.h>
  54 #if defined(__NetBSD__)
  55 #include <net/if_tap.h>
  56 #endif
  57 #ifdef __linux__
  58 #include <linux/if_tun.h>
  59 #endif
  60 #include <arpa/inet.h>
  61 #include <dirent.h>
  62 #include <netdb.h>
  63 #include <sys/select.h>
  64 #ifdef _BSD
  65 #include <sys/stat.h>
  66 #ifdef __FreeBSD__
  67 #include <libutil.h>
  68 #else
  69 #include <util.h>
  70 #endif
  71 #elif defined (__GLIBC__) && defined (__FreeBSD_kernel__)
  72 #include <freebsd/stdlib.h>
  73 #else
  74 #ifdef __linux__
  75 #include <pty.h>
  76 #include <malloc.h>
  77 #include <linux/rtc.h>
  78 #endif
  79 #endif
  80 #endif
  81
  82 #ifdef _WIN32
  83 #include <malloc.h>
  84 #include <sys/timeb.h>
  85 #include <mmsystem.h>
  86 #define getopt_long_only getopt_long
  87 #define memalign(align, size) malloc(size)
  88 #endif
  89
  90 /* point to the block driver where the snapshots are managed */
  91 static BlockDriverState *bs_snapshots;
  92
  93 #define SELF_ANNOUNCE_ROUNDS 5
  94 #define ETH_P_EXPERIMENTAL 0x01F1 /* just a number */
  95 //#define ETH_P_EXPERIMENTAL 0x0012 /* make it the size of the packet */
  96 #define EXPERIMENTAL_MAGIC 0xf1f23f4f
  97
  98 static int announce_self_create(uint8_t *buf,
  99                                 uint8_t *mac_addr)
 100 {
 101     uint32_t magic = EXPERIMENTAL_MAGIC;
 102     uint16_t proto = htons(ETH_P_EXPERIMENTAL);
 103
 104     /* FIXME: should we send a different packet (arp/rarp/ping)? */
 105
 106     memset(buf, 0, 64);
 107     memset(buf, 0xff, 6);         /* h_dst */
 108     memcpy(buf + 6, mac_addr, 6); /* h_src */
 109     memcpy(buf + 12, &proto, 2);  /* h_proto */
 110     memcpy(buf + 14, &magic, 4);  /* magic */
 111
 112     return 64; /* len */
 113 }
 114
 115 static void qemu_announce_self_once(void *opaque)
 116 {
 117     int i, len;
 118     VLANState *vlan;
 119     VLANClientState *vc;
 120     uint8_t buf[256];
 121     static int count = SELF_ANNOUNCE_ROUNDS;
 122     QEMUTimer *timer = *(QEMUTimer **)opaque;
 123
 124     for (i = 0; i < MAX_NICS; i++) {
 125         if (!nd_table[i].used)
 126             continue;
 127         len = announce_self_create(buf, nd_table[i].macaddr);
 128         vlan = nd_table[i].vlan;
 129         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 130             vc->fd_read(vc->opaque, buf, len);
 131         }
 132     }
 133     if (count--) {
 134             qemu_mod_timer(timer, qemu_get_clock(rt_clock) + 100);
 135     } else {
 136             qemu_del_timer(timer);
 137             qemu_free_timer(timer);
 138     }
 139 }
 140
 141 void qemu_announce_self(void)
 142 {
 143         static QEMUTimer *timer;
 144         timer = qemu_new_timer(rt_clock, qemu_announce_self_once, &timer);
 145         qemu_announce_self_once(&timer);
 146 }
 147
 148 /***********************************************************/
 149 /* savevm/loadvm support */
 150
 151 #define IO_BUF_SIZE 32768
 152
 153 struct QEMUFile {
 154     QEMUFilePutBufferFunc *put_buffer;
 155     QEMUFileGetBufferFunc *get_buffer;
 156     QEMUFileCloseFunc *close;
 157     QEMUFileRateLimit *rate_limit;
 158     void *opaque;
 159     int is_write;
 160
 161     int64_t buf_offset; /* start of buffer when writing, end of buffer
 162                            when reading */
 163     int buf_index;
 164     int buf_size; /* 0 when writing */
 165     uint8_t buf[IO_BUF_SIZE];
 166
 167     int has_error;
 168 };
 169
 170 typedef struct QEMUFilePopen
 171 {
 172     FILE *popen_file;
 173     QEMUFile *file;
 174 } QEMUFilePopen;
 175
 176 typedef struct QEMUFileSocket
 177 {
 178     int fd;
 179     QEMUFile *file;
 180 } QEMUFileSocket;
 181
 182 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 183 {
 184     QEMUFileSocket *s = opaque;
 185     ssize_t len;
 186
 187     do {
 188         len = recv(s->fd, buf, size, 0);
 189     } while (len == -1 && socket_error() == EINTR);
 190
 191     if (len == -1)
 192         len = -socket_error();
 193
 194     return len;
 195 }
 196
 197 static int socket_close(void *opaque)
 198 {
 199     QEMUFileSocket *s = opaque;
 200     qemu_free(s);
 201     return 0;
 202 }
 203
 204 static int popen_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 205 {
 206     QEMUFilePopen *s = opaque;
 207     return fwrite(buf, 1, size, s->popen_file);
 208 }
 209
 210 static int popen_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 211 {
 212     QEMUFilePopen *s = opaque;
 213     FILE *fp = s->popen_file;
 214     int bytes;
 215
 216     do {
 217         clearerr(fp);
 218         bytes = fread(buf, 1, size, fp);
 219     } while ((bytes == 0) && ferror(fp) && (errno == EINTR));
 220     return bytes;
 221 }
 222
 223 static int popen_close(void *opaque)
 224 {
 225     QEMUFilePopen *s = opaque;
 226     pclose(s->popen_file);
 227     qemu_free(s);
 228     return 0;
 229 }
 230
 231 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 232 {
 233     QEMUFilePopen *s;
 234
 235     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 236         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 237         return NULL;
 238     }
 239
 240     s = qemu_mallocz(sizeof(QEMUFilePopen));
 241
 242     s->popen_file = popen_file;
 243
 244     if(mode[0] == 'r') {
 245         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL);
 246     } else {
 247         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL);
 248     }
 249     return s->file;
 250 }
 251
 252 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 253 {
 254     FILE *popen_file;
 255
 256     popen_file = popen(command, mode);
 257     if(popen_file == NULL) {
 258         return NULL;
 259     }
 260
 261     return qemu_popen(popen_file, mode);
 262 }
 263
 264 int qemu_popen_fd(QEMUFile *f)
 265 {
 266     QEMUFilePopen *p;
 267     int fd;
 268
 269     p = (QEMUFilePopen *)f->opaque;
 270     fd = fileno(p->popen_file);
 271
 272     return fd;
 273 }
 274
 275 QEMUFile *qemu_fopen_socket(int fd)
 276 {
 277     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 278
 279     s->fd = fd;
 280     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL);
 281     return s->file;
 282 }
 283
 284 typedef struct QEMUFileStdio
 285 {
 286     FILE *outfile;
 287 } QEMUFileStdio;
 288
 289 static int file_put_buffer(void *opaque, const uint8_t *buf,
 290                             int64_t pos, int size)
 291 {
 292     QEMUFileStdio *s = opaque;
 293     fseek(s->outfile, pos, SEEK_SET);
 294     fwrite(buf, 1, size, s->outfile);
 295     return size;
 296 }
 297
 298 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 299 {
 300     QEMUFileStdio *s = opaque;
 301     fseek(s->outfile, pos, SEEK_SET);
 302     return fread(buf, 1, size, s->outfile);
 303 }
 304
 305 static int file_close(void *opaque)
 306 {
 307     QEMUFileStdio *s = opaque;
 308     fclose(s->outfile);
 309     qemu_free(s);
 310     return 0;
 311 }
 312
 313 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 314 {
 315     QEMUFileStdio *s;
 316
 317     s = qemu_mallocz(sizeof(QEMUFileStdio));
 318
 319     s->outfile = fopen(filename, mode);
 320     if (!s->outfile)
 321         goto fail;
 322
 323     if (!strcmp(mode, "wb"))
 324         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL);
 325     else if (!strcmp(mode, "rb"))
 326         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL);
 327
 328 fail:
 329     if (s->outfile)
 330         fclose(s->outfile);
 331     qemu_free(s);
 332     return NULL;
 333 }
 334
 335 typedef struct QEMUFileBdrv
 336 {
 337     BlockDriverState *bs;
 338     int64_t base_offset;
 339 } QEMUFileBdrv;
 340
 341 static int block_put_buffer(void *opaque, const uint8_t *buf,
 342                            int64_t pos, int size)
 343 {
 344     QEMUFileBdrv *s = opaque;
 345     bdrv_put_buffer(s->bs, buf, s->base_offset + pos, size);
 346     return size;
 347 }
 348
 349 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 350 {
 351     QEMUFileBdrv *s = opaque;
 352     return bdrv_get_buffer(s->bs, buf, s->base_offset + pos, size);
 353 }
 354
 355 static int bdrv_fclose(void *opaque)
 356 {
 357     QEMUFileBdrv *s = opaque;
 358     qemu_free(s);
 359     return 0;
 360 }
 361
 362 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable)
 363 {
 364     QEMUFileBdrv *s;
 365
 366     s = qemu_mallocz(sizeof(QEMUFileBdrv));
 367
 368     s->bs = bs;
 369     s->base_offset = offset;
 370
 371     if (is_writable)
 372         return qemu_fopen_ops(s, block_put_buffer, NULL, bdrv_fclose, NULL);
 373
 374     return qemu_fopen_ops(s, NULL, block_get_buffer, bdrv_fclose, NULL);
 375 }
 376
 377 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 378                          QEMUFileGetBufferFunc *get_buffer,
 379                          QEMUFileCloseFunc *close,
 380                          QEMUFileRateLimit *rate_limit)
 381 {
 382     QEMUFile *f;
 383
 384     f = qemu_mallocz(sizeof(QEMUFile));
 385
 386     f->opaque = opaque;
 387     f->put_buffer = put_buffer;
 388     f->get_buffer = get_buffer;
 389     f->close = close;
 390     f->rate_limit = 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 void qemu_put_be16(QEMUFile *f, unsigned int v)
 569 {
 570     qemu_put_byte(f, v >> 8);
 571     qemu_put_byte(f, v);
 572 }
 573
 574 void qemu_put_be32(QEMUFile *f, unsigned int v)
 575 {
 576     qemu_put_byte(f, v >> 24);
 577     qemu_put_byte(f, v >> 16);
 578     qemu_put_byte(f, v >> 8);
 579     qemu_put_byte(f, v);
 580 }
 581
 582 void qemu_put_be64(QEMUFile *f, uint64_t v)
 583 {
 584     qemu_put_be32(f, v >> 32);
 585     qemu_put_be32(f, v);
 586 }
 587
 588 unsigned int qemu_get_be16(QEMUFile *f)
 589 {
 590     unsigned int v;
 591     v = qemu_get_byte(f) << 8;
 592     v |= qemu_get_byte(f);
 593     return v;
 594 }
 595
 596 unsigned int qemu_get_be32(QEMUFile *f)
 597 {
 598     unsigned int v;
 599     v = qemu_get_byte(f) << 24;
 600     v |= qemu_get_byte(f) << 16;
 601     v |= qemu_get_byte(f) << 8;
 602     v |= qemu_get_byte(f);
 603     return v;
 604 }
 605
 606 uint64_t qemu_get_be64(QEMUFile *f)
 607 {
 608     uint64_t v;
 609     v = (uint64_t)qemu_get_be32(f) << 32;
 610     v |= qemu_get_be32(f);
 611     return v;
 612 }
 613
 614 typedef struct SaveStateEntry {
 615     char idstr[256];
 616     int instance_id;
 617     int version_id;
 618     int section_id;
 619     SaveLiveStateHandler *save_live_state;
 620     SaveStateHandler *save_state;
 621     LoadStateHandler *load_state;
 622     void *opaque;
 623     struct SaveStateEntry *next;
 624 } SaveStateEntry;
 625
 626 static SaveStateEntry *first_se;
 627
 628 /* TODO: Individual devices generally have very little idea about the rest
 629    of the system, so instance_id should be removed/replaced.
 630    Meanwhile pass -1 as instance_id if you do not already have a clearly
 631    distinguishing id for all instances of your device class. */
 632 int register_savevm_live(const char *idstr,
 633                          int instance_id,
 634                          int version_id,
 635                          SaveLiveStateHandler *save_live_state,
 636                          SaveStateHandler *save_state,
 637                          LoadStateHandler *load_state,
 638                          void *opaque)
 639 {
 640     SaveStateEntry *se, **pse;
 641     static int global_section_id;
 642
 643     se = qemu_malloc(sizeof(SaveStateEntry));
 644     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 645     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 646     se->version_id = version_id;
 647     se->section_id = global_section_id++;
 648     se->save_live_state = save_live_state;
 649     se->save_state = save_state;
 650     se->load_state = load_state;
 651     se->opaque = opaque;
 652     se->next = NULL;
 653
 654     /* add at the end of list */
 655     pse = &first_se;
 656     while (*pse != NULL) {
 657         if (instance_id == -1
 658                 && strcmp(se->idstr, (*pse)->idstr) == 0
 659                 && se->instance_id <= (*pse)->instance_id)
 660             se->instance_id = (*pse)->instance_id + 1;
 661         pse = &(*pse)->next;
 662     }
 663     *pse = se;
 664     return 0;
 665 }
 666
 667 int register_savevm(const char *idstr,
 668                     int instance_id,
 669                     int version_id,
 670                     SaveStateHandler *save_state,
 671                     LoadStateHandler *load_state,
 672                     void *opaque)
 673 {
 674     return register_savevm_live(idstr, instance_id, version_id,
 675                                 NULL, save_state, load_state, opaque);
 676 }
 677
 678 void unregister_savevm(const char *idstr, void *opaque)
 679 {
 680     SaveStateEntry **pse;
 681
 682     pse = &first_se;
 683     while (*pse != NULL) {
 684         if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
 685             SaveStateEntry *next = (*pse)->next;
 686             qemu_free(*pse);
 687             *pse = next;
 688             continue;
 689         }
 690         pse = &(*pse)->next;
 691     }
 692 }
 693
 694 #define QEMU_VM_FILE_MAGIC           0x5145564d
 695 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 696 #define QEMU_VM_FILE_VERSION         0x00000003
 697
 698 #define QEMU_VM_EOF                  0x00
 699 #define QEMU_VM_SECTION_START        0x01
 700 #define QEMU_VM_SECTION_PART         0x02
 701 #define QEMU_VM_SECTION_END          0x03
 702 #define QEMU_VM_SECTION_FULL         0x04
 703
 704 int qemu_savevm_state_begin(QEMUFile *f)
 705 {
 706     SaveStateEntry *se;
 707
 708     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 709     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 710
 711     for (se = first_se; se != NULL; se = se->next) {
 712         int len;
 713
 714         if (se->save_live_state == NULL)
 715             continue;
 716
 717         /* Section type */
 718         qemu_put_byte(f, QEMU_VM_SECTION_START);
 719         qemu_put_be32(f, se->section_id);
 720
 721         /* ID string */
 722         len = strlen(se->idstr);
 723         qemu_put_byte(f, len);
 724         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 725
 726         qemu_put_be32(f, se->instance_id);
 727         qemu_put_be32(f, se->version_id);
 728
 729         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 730     }
 731
 732     if (qemu_file_has_error(f))
 733         return -EIO;
 734
 735     return 0;
 736 }
 737
 738 int qemu_savevm_state_iterate(QEMUFile *f)
 739 {
 740     SaveStateEntry *se;
 741     int ret = 1;
 742
 743     for (se = first_se; se != NULL; se = se->next) {
 744         if (se->save_live_state == NULL)
 745             continue;
 746
 747         /* Section type */
 748         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 749         qemu_put_be32(f, se->section_id);
 750
 751         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 752     }
 753
 754     if (ret)
 755         return 1;
 756
 757     if (qemu_file_has_error(f))
 758         return -EIO;
 759
 760     return 0;
 761 }
 762
 763 int qemu_savevm_state_complete(QEMUFile *f)
 764 {
 765     SaveStateEntry *se;
 766
 767     for (se = first_se; se != NULL; se = se->next) {
 768         if (se->save_live_state == NULL)
 769             continue;
 770
 771         /* Section type */
 772         qemu_put_byte(f, QEMU_VM_SECTION_END);
 773         qemu_put_be32(f, se->section_id);
 774
 775         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 776     }
 777
 778     for(se = first_se; se != NULL; se = se->next) {
 779         int len;
 780
 781         if (se->save_state == NULL)
 782             continue;
 783
 784         /* Section type */
 785         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 786         qemu_put_be32(f, se->section_id);
 787
 788         /* ID string */
 789         len = strlen(se->idstr);
 790         qemu_put_byte(f, len);
 791         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 792
 793         qemu_put_be32(f, se->instance_id);
 794         qemu_put_be32(f, se->version_id);
 795
 796         se->save_state(f, se->opaque);
 797     }
 798
 799     qemu_put_byte(f, QEMU_VM_EOF);
 800
 801     if (qemu_file_has_error(f))
 802         return -EIO;
 803
 804     return 0;
 805 }
 806
 807 int qemu_savevm_state(QEMUFile *f)
 808 {
 809     int saved_vm_running;
 810     int ret;
 811
 812     saved_vm_running = vm_running;
 813     vm_stop(0);
 814
 815     bdrv_flush_all();
 816
 817     ret = qemu_savevm_state_begin(f);
 818     if (ret < 0)
 819         goto out;
 820
 821     do {
 822         ret = qemu_savevm_state_iterate(f);
 823         if (ret < 0)
 824             goto out;
 825     } while (ret == 0);
 826
 827     ret = qemu_savevm_state_complete(f);
 828
 829 out:
 830     if (qemu_file_has_error(f))
 831         ret = -EIO;
 832
 833     if (!ret && saved_vm_running)
 834         vm_start();
 835
 836     return ret;
 837 }
 838
 839 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 840 {
 841     SaveStateEntry *se;
 842
 843     for(se = first_se; se != NULL; se = se->next) {
 844         if (!strcmp(se->idstr, idstr) &&
 845             instance_id == se->instance_id)
 846             return se;
 847     }
 848     return NULL;
 849 }
 850
 851 typedef struct LoadStateEntry {
 852     SaveStateEntry *se;
 853     int section_id;
 854     int version_id;
 855     struct LoadStateEntry *next;
 856 } LoadStateEntry;
 857
 858 static int qemu_loadvm_state_v2(QEMUFile *f)
 859 {
 860     SaveStateEntry *se;
 861     int len, ret, instance_id, record_len, version_id;
 862     int64_t total_len, end_pos, cur_pos;
 863     char idstr[256];
 864
 865     total_len = qemu_get_be64(f);
 866     end_pos = total_len + qemu_ftell(f);
 867     for(;;) {
 868         if (qemu_ftell(f) >= end_pos)
 869             break;
 870         len = qemu_get_byte(f);
 871         qemu_get_buffer(f, (uint8_t *)idstr, len);
 872         idstr[len] = '\0';
 873         instance_id = qemu_get_be32(f);
 874         version_id = qemu_get_be32(f);
 875         record_len = qemu_get_be32(f);
 876         cur_pos = qemu_ftell(f);
 877         se = find_se(idstr, instance_id);
 878         if (!se) {
 879             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 880                     instance_id, idstr);
 881         } else {
 882             ret = se->load_state(f, se->opaque, version_id);
 883             if (ret < 0) {
 884                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 885                         instance_id, idstr);
 886                 return ret;
 887             }
 888         }
 889         /* always seek to exact end of record */
 890         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 891     }
 892
 893     if (qemu_file_has_error(f))
 894         return -EIO;
 895
 896     return 0;
 897 }
 898
 899 int qemu_loadvm_state(QEMUFile *f)
 900 {
 901     LoadStateEntry *first_le = NULL;
 902     uint8_t section_type;
 903     unsigned int v;
 904     int ret;
 905
 906     v = qemu_get_be32(f);
 907     if (v != QEMU_VM_FILE_MAGIC)
 908         return -EINVAL;
 909
 910     v = qemu_get_be32(f);
 911     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 912         return qemu_loadvm_state_v2(f);
 913     if (v != QEMU_VM_FILE_VERSION)
 914         return -ENOTSUP;
 915
 916     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 917         uint32_t instance_id, version_id, section_id;
 918         LoadStateEntry *le;
 919         SaveStateEntry *se;
 920         char idstr[257];
 921         int len;
 922
 923         switch (section_type) {
 924         case QEMU_VM_SECTION_START:
 925         case QEMU_VM_SECTION_FULL:
 926             /* Read section start */
 927             section_id = qemu_get_be32(f);
 928             len = qemu_get_byte(f);
 929             qemu_get_buffer(f, (uint8_t *)idstr, len);
 930             idstr[len] = 0;
 931             instance_id = qemu_get_be32(f);
 932             version_id = qemu_get_be32(f);
 933
 934             /* Find savevm section */
 935             se = find_se(idstr, instance_id);
 936             if (se == NULL) {
 937                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 938                 ret = -EINVAL;
 939                 goto out;
 940             }
 941
 942             /* Validate version */
 943             if (version_id > se->version_id) {
 944                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 945                         version_id, idstr, se->version_id);
 946                 ret = -EINVAL;
 947                 goto out;
 948             }
 949
 950             /* Add entry */
 951             le = qemu_mallocz(sizeof(*le));
 952
 953             le->se = se;
 954             le->section_id = section_id;
 955             le->version_id = version_id;
 956             le->next = first_le;
 957             first_le = le;
 958
 959             le->se->load_state(f, le->se->opaque, le->version_id);
 960             break;
 961         case QEMU_VM_SECTION_PART:
 962         case QEMU_VM_SECTION_END:
 963             section_id = qemu_get_be32(f);
 964
 965             for (le = first_le; le && le->section_id != section_id; le = le->next);
 966             if (le == NULL) {
 967                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 968                 ret = -EINVAL;
 969                 goto out;
 970             }
 971
 972             le->se->load_state(f, le->se->opaque, le->version_id);
 973             break;
 974         default:
 975             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 976             ret = -EINVAL;
 977             goto out;
 978         }
 979     }
 980
 981     ret = 0;
 982
 983 out:
 984     while (first_le) {
 985         LoadStateEntry *le = first_le;
 986         first_le = first_le->next;
 987         qemu_free(le);
 988     }
 989
 990     if (qemu_file_has_error(f))
 991         ret = -EIO;
 992
 993     return ret;
 994 }
 995
 996 /* device can contain snapshots */
 997 static int bdrv_can_snapshot(BlockDriverState *bs)
 998 {
 999     return (bs &&
1000             !bdrv_is_removable(bs) &&
1001             !bdrv_is_read_only(bs));
1002 }
1003
1004 /* device must be snapshots in order to have a reliable snapshot */
1005 static int bdrv_has_snapshot(BlockDriverState *bs)
1006 {
1007     return (bs &&
1008             !bdrv_is_removable(bs) &&
1009             !bdrv_is_read_only(bs));
1010 }
1011
1012 static BlockDriverState *get_bs_snapshots(void)
1013 {
1014     BlockDriverState *bs;
1015     int i;
1016
1017     if (bs_snapshots)
1018         return bs_snapshots;
1019     for(i = 0; i <= nb_drives; i++) {
1020         bs = drives_table[i].bdrv;
1021         if (bdrv_can_snapshot(bs))
1022             goto ok;
1023     }
1024     return NULL;
1025  ok:
1026     bs_snapshots = bs;
1027     return bs;
1028 }
1029
1030 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1031                               const char *name)
1032 {
1033     QEMUSnapshotInfo *sn_tab, *sn;
1034     int nb_sns, i, ret;
1035
1036     ret = -ENOENT;
1037     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1038     if (nb_sns < 0)
1039         return ret;
1040     for(i = 0; i < nb_sns; i++) {
1041         sn = &sn_tab[i];
1042         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1043             *sn_info = *sn;
1044             ret = 0;
1045             break;
1046         }
1047     }
1048     qemu_free(sn_tab);
1049     return ret;
1050 }
1051
1052 void do_savevm(const char *name)
1053 {
1054     BlockDriverState *bs, *bs1;
1055     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1056     int must_delete, ret, i;
1057     BlockDriverInfo bdi1, *bdi = &bdi1;
1058     QEMUFile *f;
1059     int saved_vm_running;
1060     uint32_t vm_state_size;
1061 #ifdef _WIN32
1062     struct _timeb tb;
1063 #else
1064     struct timeval tv;
1065 #endif
1066
1067     bs = get_bs_snapshots();
1068     if (!bs) {
1069         term_printf("No block device can accept snapshots\n");
1070         return;
1071     }
1072
1073     /* ??? Should this occur after vm_stop?  */
1074     qemu_aio_flush();
1075
1076     saved_vm_running = vm_running;
1077     vm_stop(0);
1078
1079     must_delete = 0;
1080     if (name) {
1081         ret = bdrv_snapshot_find(bs, old_sn, name);
1082         if (ret >= 0) {
1083             must_delete = 1;
1084         }
1085     }
1086     memset(sn, 0, sizeof(*sn));
1087     if (must_delete) {
1088         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1089         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1090     } else {
1091         if (name)
1092             pstrcpy(sn->name, sizeof(sn->name), name);
1093     }
1094
1095     /* fill auxiliary fields */
1096 #ifdef _WIN32
1097     _ftime(&tb);
1098     sn->date_sec = tb.time;
1099     sn->date_nsec = tb.millitm * 1000000;
1100 #else
1101     gettimeofday(&tv, NULL);
1102     sn->date_sec = tv.tv_sec;
1103     sn->date_nsec = tv.tv_usec * 1000;
1104 #endif
1105     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1106
1107     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1108         term_printf("Device %s does not support VM state snapshots\n",
1109                     bdrv_get_device_name(bs));
1110         goto the_end;
1111     }
1112
1113     /* save the VM state */
1114     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 1);
1115     if (!f) {
1116         term_printf("Could not open VM state file\n");
1117         goto the_end;
1118     }
1119     ret = qemu_savevm_state(f);
1120     vm_state_size = qemu_ftell(f);
1121     qemu_fclose(f);
1122     if (ret < 0) {
1123         term_printf("Error %d while writing VM\n", ret);
1124         goto the_end;
1125     }
1126
1127     /* create the snapshots */
1128
1129     for(i = 0; i < nb_drives; i++) {
1130         bs1 = drives_table[i].bdrv;
1131         if (bdrv_has_snapshot(bs1)) {
1132             if (must_delete) {
1133                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1134                 if (ret < 0) {
1135                     term_printf("Error while deleting snapshot on '%s'\n",
1136                                 bdrv_get_device_name(bs1));
1137                 }
1138             }
1139             /* Write VM state size only to the image that contains the state */
1140             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1141             ret = bdrv_snapshot_create(bs1, sn);
1142             if (ret < 0) {
1143                 term_printf("Error while creating snapshot on '%s'\n",
1144                             bdrv_get_device_name(bs1));
1145             }
1146         }
1147     }
1148
1149  the_end:
1150     if (saved_vm_running)
1151         vm_start();
1152 }
1153
1154 void do_loadvm(const char *name)
1155 {
1156     BlockDriverState *bs, *bs1;
1157     BlockDriverInfo bdi1, *bdi = &bdi1;
1158     QEMUSnapshotInfo sn;
1159     QEMUFile *f;
1160     int i, ret;
1161     int saved_vm_running;
1162
1163     bs = get_bs_snapshots();
1164     if (!bs) {
1165         term_printf("No block device supports snapshots\n");
1166         return;
1167     }
1168
1169     /* Flush all IO requests so they don't interfere with the new state.  */
1170     qemu_aio_flush();
1171
1172     saved_vm_running = vm_running;
1173     vm_stop(0);
1174
1175     for(i = 0; i <= nb_drives; i++) {
1176         bs1 = drives_table[i].bdrv;
1177         if (bdrv_has_snapshot(bs1)) {
1178             ret = bdrv_snapshot_goto(bs1, name);
1179             if (ret < 0) {
1180                 if (bs != bs1)
1181                     term_printf("Warning: ");
1182                 switch(ret) {
1183                 case -ENOTSUP:
1184                     term_printf("Snapshots not supported on device '%s'\n",
1185                                 bdrv_get_device_name(bs1));
1186                     break;
1187                 case -ENOENT:
1188                     term_printf("Could not find snapshot '%s' on device '%s'\n",
1189                                 name, bdrv_get_device_name(bs1));
1190                     break;
1191                 default:
1192                     term_printf("Error %d while activating snapshot on '%s'\n",
1193                                 ret, bdrv_get_device_name(bs1));
1194                     break;
1195                 }
1196                 /* fatal on snapshot block device */
1197                 if (bs == bs1)
1198                     goto the_end;
1199             }
1200         }
1201     }
1202
1203     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1204         term_printf("Device %s does not support VM state snapshots\n",
1205                     bdrv_get_device_name(bs));
1206         return;
1207     }
1208
1209     /* Don't even try to load empty VM states */
1210     ret = bdrv_snapshot_find(bs, &sn, name);
1211     if ((ret >= 0) && (sn.vm_state_size == 0))
1212         goto the_end;
1213
1214     /* restore the VM state */
1215     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 0);
1216     if (!f) {
1217         term_printf("Could not open VM state file\n");
1218         goto the_end;
1219     }
1220     ret = qemu_loadvm_state(f);
1221     qemu_fclose(f);
1222     if (ret < 0) {
1223         term_printf("Error %d while loading VM state\n", ret);
1224     }
1225  the_end:
1226     if (saved_vm_running)
1227         vm_start();
1228 }
1229
1230 void do_delvm(const char *name)
1231 {
1232     BlockDriverState *bs, *bs1;
1233     int i, ret;
1234
1235     bs = get_bs_snapshots();
1236     if (!bs) {
1237         term_printf("No block device supports snapshots\n");
1238         return;
1239     }
1240
1241     for(i = 0; i <= nb_drives; i++) {
1242         bs1 = drives_table[i].bdrv;
1243         if (bdrv_has_snapshot(bs1)) {
1244             ret = bdrv_snapshot_delete(bs1, name);
1245             if (ret < 0) {
1246                 if (ret == -ENOTSUP)
1247                     term_printf("Snapshots not supported on device '%s'\n",
1248                                 bdrv_get_device_name(bs1));
1249                 else
1250                     term_printf("Error %d while deleting snapshot on '%s'\n",
1251                                 ret, bdrv_get_device_name(bs1));
1252             }
1253         }
1254     }
1255 }
1256
1257 void do_info_snapshots(void)
1258 {
1259     BlockDriverState *bs, *bs1;
1260     QEMUSnapshotInfo *sn_tab, *sn;
1261     int nb_sns, i;
1262     char buf[256];
1263
1264     bs = get_bs_snapshots();
1265     if (!bs) {
1266         term_printf("No available block device supports snapshots\n");
1267         return;
1268     }
1269     term_printf("Snapshot devices:");
1270     for(i = 0; i <= nb_drives; i++) {
1271         bs1 = drives_table[i].bdrv;
1272         if (bdrv_has_snapshot(bs1)) {
1273             if (bs == bs1)
1274                 term_printf(" %s", bdrv_get_device_name(bs1));
1275         }
1276     }
1277     term_printf("\n");
1278
1279     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1280     if (nb_sns < 0) {
1281         term_printf("bdrv_snapshot_list: error %d\n", nb_sns);
1282         return;
1283     }
1284     term_printf("Snapshot list (from %s):\n", bdrv_get_device_name(bs));
1285     term_printf("%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1286     for(i = 0; i < nb_sns; i++) {
1287         sn = &sn_tab[i];
1288         term_printf("%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1289     }
1290     qemu_free(sn_tab);
1291 }