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, 0xff, 6);         /* h_dst */
 111     memcpy(buf + 6, mac_addr, 6); /* h_src */
 112     memcpy(buf + 12, &proto, 2);  /* h_proto */
 113     memcpy(buf + 14, &magic, 4);  /* magic */
 114
 115     return 18; /* len */
 116 }
 117
 118 void qemu_announce_self(void)
 119 {
 120     int i, j, len;
 121     VLANState *vlan;
 122     VLANClientState *vc;
 123     uint8_t buf[256];
 124
 125     for (i = 0; i < nb_nics; i++) {
 126         len = announce_self_create(buf, nd_table[i].macaddr);
 127         vlan = nd_table[i].vlan;
 128         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 129             for (j=0; j < SELF_ANNOUNCE_ROUNDS; j++)
 130                 vc->fd_read(vc->opaque, buf, len);
 131         }
 132     }
 133 }
 134
 135 /***********************************************************/
 136 /* savevm/loadvm support */
 137
 138 #define IO_BUF_SIZE 32768
 139
 140 struct QEMUFile {
 141     QEMUFilePutBufferFunc *put_buffer;
 142     QEMUFileGetBufferFunc *get_buffer;
 143     QEMUFileCloseFunc *close;
 144     QEMUFileRateLimit *rate_limit;
 145     void *opaque;
 146     int is_write;
 147
 148     int64_t buf_offset; /* start of buffer when writing, end of buffer
 149                            when reading */
 150     int buf_index;
 151     int buf_size; /* 0 when writing */
 152     uint8_t buf[IO_BUF_SIZE];
 153
 154     int has_error;
 155 };
 156
 157 typedef struct QEMUFilePopen
 158 {
 159     FILE *popen_file;
 160     QEMUFile *file;
 161 } QEMUFilePopen;
 162
 163 typedef struct QEMUFileSocket
 164 {
 165     int fd;
 166     QEMUFile *file;
 167 } QEMUFileSocket;
 168
 169 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 170 {
 171     QEMUFileSocket *s = opaque;
 172     ssize_t len;
 173
 174     do {
 175         len = recv(s->fd, buf, size, 0);
 176     } while (len == -1 && socket_error() == EINTR);
 177
 178     if (len == -1)
 179         len = -socket_error();
 180
 181     return len;
 182 }
 183
 184 static int socket_close(void *opaque)
 185 {
 186     QEMUFileSocket *s = opaque;
 187     qemu_free(s);
 188     return 0;
 189 }
 190
 191 static int popen_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 192 {
 193     QEMUFilePopen *s = opaque;
 194     return fwrite(buf, 1, size, s->popen_file);
 195 }
 196
 197 static int popen_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 198 {
 199     QEMUFilePopen *s = opaque;
 200     return fread(buf, 1, size, s->popen_file);
 201 }
 202
 203 static int popen_close(void *opaque)
 204 {
 205     QEMUFilePopen *s = opaque;
 206     pclose(s->popen_file);
 207     qemu_free(s);
 208     return 0;
 209 }
 210
 211 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 212 {
 213     QEMUFilePopen *s;
 214
 215     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 216         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 217         return NULL;
 218     }
 219
 220     s = qemu_mallocz(sizeof(QEMUFilePopen));
 221
 222     s->popen_file = popen_file;
 223
 224     if(mode[0] == 'r') {
 225         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL);
 226     } else {
 227         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL);
 228     }
 229     fprintf(stderr, "qemu_popen: returning result of qemu_fopen_ops\n");
 230     return s->file;
 231 }
 232
 233 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 234 {
 235     FILE *popen_file;
 236
 237     popen_file = popen(command, mode);
 238     if(popen_file == NULL) {
 239         return NULL;
 240     }
 241
 242     return qemu_popen(popen_file, mode);
 243 }
 244
 245 QEMUFile *qemu_fopen_socket(int fd)
 246 {
 247     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 248
 249     s->fd = fd;
 250     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL);
 251     return s->file;
 252 }
 253
 254 typedef struct QEMUFileStdio
 255 {
 256     FILE *outfile;
 257 } QEMUFileStdio;
 258
 259 static int file_put_buffer(void *opaque, const uint8_t *buf,
 260                             int64_t pos, int size)
 261 {
 262     QEMUFileStdio *s = opaque;
 263     fseek(s->outfile, pos, SEEK_SET);
 264     fwrite(buf, 1, size, s->outfile);
 265     return size;
 266 }
 267
 268 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 269 {
 270     QEMUFileStdio *s = opaque;
 271     fseek(s->outfile, pos, SEEK_SET);
 272     return fread(buf, 1, size, s->outfile);
 273 }
 274
 275 static int file_close(void *opaque)
 276 {
 277     QEMUFileStdio *s = opaque;
 278     fclose(s->outfile);
 279     qemu_free(s);
 280     return 0;
 281 }
 282
 283 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 284 {
 285     QEMUFileStdio *s;
 286
 287     s = qemu_mallocz(sizeof(QEMUFileStdio));
 288
 289     s->outfile = fopen(filename, mode);
 290     if (!s->outfile)
 291         goto fail;
 292
 293     if (!strcmp(mode, "wb"))
 294         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL);
 295     else if (!strcmp(mode, "rb"))
 296         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL);
 297
 298 fail:
 299     if (s->outfile)
 300         fclose(s->outfile);
 301     qemu_free(s);
 302     return NULL;
 303 }
 304
 305 typedef struct QEMUFileBdrv
 306 {
 307     BlockDriverState *bs;
 308     int64_t base_offset;
 309 } QEMUFileBdrv;
 310
 311 static int bdrv_put_buffer(void *opaque, const uint8_t *buf,
 312                            int64_t pos, int size)
 313 {
 314     QEMUFileBdrv *s = opaque;
 315     bdrv_pwrite(s->bs, s->base_offset + pos, buf, size);
 316     return size;
 317 }
 318
 319 static int bdrv_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 320 {
 321     QEMUFileBdrv *s = opaque;
 322     return bdrv_pread(s->bs, s->base_offset + pos, buf, size);
 323 }
 324
 325 static int bdrv_fclose(void *opaque)
 326 {
 327     QEMUFileBdrv *s = opaque;
 328     qemu_free(s);
 329     return 0;
 330 }
 331
 332 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable)
 333 {
 334     QEMUFileBdrv *s;
 335
 336     s = qemu_mallocz(sizeof(QEMUFileBdrv));
 337
 338     s->bs = bs;
 339     s->base_offset = offset;
 340
 341     if (is_writable)
 342         return qemu_fopen_ops(s, bdrv_put_buffer, NULL, bdrv_fclose, NULL);
 343
 344     return qemu_fopen_ops(s, NULL, bdrv_get_buffer, bdrv_fclose, NULL);
 345 }
 346
 347 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 348                          QEMUFileGetBufferFunc *get_buffer,
 349                          QEMUFileCloseFunc *close,
 350                          QEMUFileRateLimit *rate_limit)
 351 {
 352     QEMUFile *f;
 353
 354     f = qemu_mallocz(sizeof(QEMUFile));
 355
 356     f->opaque = opaque;
 357     f->put_buffer = put_buffer;
 358     f->get_buffer = get_buffer;
 359     f->close = close;
 360     f->rate_limit = rate_limit;
 361     f->is_write = 0;
 362
 363     return f;
 364 }
 365
 366 int qemu_file_has_error(QEMUFile *f)
 367 {
 368     return f->has_error;
 369 }
 370
 371 void qemu_fflush(QEMUFile *f)
 372 {
 373     if (!f->put_buffer)
 374         return;
 375
 376     if (f->is_write && f->buf_index > 0) {
 377         int len;
 378
 379         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 380         if (len > 0)
 381             f->buf_offset += f->buf_index;
 382         else
 383             f->has_error = 1;
 384         f->buf_index = 0;
 385     }
 386 }
 387
 388 static void qemu_fill_buffer(QEMUFile *f)
 389 {
 390     int len;
 391
 392     if (!f->get_buffer)
 393         return;
 394
 395     if (f->is_write)
 396         abort();
 397
 398     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 399     if (len > 0) {
 400         f->buf_index = 0;
 401         f->buf_size = len;
 402         f->buf_offset += len;
 403     } else if (len != -EAGAIN)
 404         f->has_error = 1;
 405 }
 406
 407 int qemu_fclose(QEMUFile *f)
 408 {
 409     int ret = 0;
 410     qemu_fflush(f);
 411     if (f->close)
 412         ret = f->close(f->opaque);
 413     qemu_free(f);
 414     return ret;
 415 }
 416
 417 void qemu_file_put_notify(QEMUFile *f)
 418 {
 419     f->put_buffer(f->opaque, NULL, 0, 0);
 420 }
 421
 422 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 423 {
 424     int l;
 425
 426     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 427         fprintf(stderr,
 428                 "Attempted to write to buffer while read buffer is not empty\n");
 429         abort();
 430     }
 431
 432     while (!f->has_error && size > 0) {
 433         l = IO_BUF_SIZE - f->buf_index;
 434         if (l > size)
 435             l = size;
 436         memcpy(f->buf + f->buf_index, buf, l);
 437         f->is_write = 1;
 438         f->buf_index += l;
 439         buf += l;
 440         size -= l;
 441         if (f->buf_index >= IO_BUF_SIZE)
 442             qemu_fflush(f);
 443     }
 444 }
 445
 446 void qemu_put_byte(QEMUFile *f, int v)
 447 {
 448     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 449         fprintf(stderr,
 450                 "Attempted to write to buffer while read buffer is not empty\n");
 451         abort();
 452     }
 453
 454     f->buf[f->buf_index++] = v;
 455     f->is_write = 1;
 456     if (f->buf_index >= IO_BUF_SIZE)
 457         qemu_fflush(f);
 458 }
 459
 460 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 461 {
 462     int size, l;
 463
 464     if (f->is_write)
 465         abort();
 466
 467     size = size1;
 468     while (size > 0) {
 469         l = f->buf_size - f->buf_index;
 470         if (l == 0) {
 471             qemu_fill_buffer(f);
 472             l = f->buf_size - f->buf_index;
 473             if (l == 0)
 474                 break;
 475         }
 476         if (l > size)
 477             l = size;
 478         memcpy(buf, f->buf + f->buf_index, l);
 479         f->buf_index += l;
 480         buf += l;
 481         size -= l;
 482     }
 483     return size1 - size;
 484 }
 485
 486 int qemu_get_byte(QEMUFile *f)
 487 {
 488     if (f->is_write)
 489         abort();
 490
 491     if (f->buf_index >= f->buf_size) {
 492         qemu_fill_buffer(f);
 493         if (f->buf_index >= f->buf_size)
 494             return 0;
 495     }
 496     return f->buf[f->buf_index++];
 497 }
 498
 499 int64_t qemu_ftell(QEMUFile *f)
 500 {
 501     return f->buf_offset - f->buf_size + f->buf_index;
 502 }
 503
 504 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 505 {
 506     if (whence == SEEK_SET) {
 507         /* nothing to do */
 508     } else if (whence == SEEK_CUR) {
 509         pos += qemu_ftell(f);
 510     } else {
 511         /* SEEK_END not supported */
 512         return -1;
 513     }
 514     if (f->put_buffer) {
 515         qemu_fflush(f);
 516         f->buf_offset = pos;
 517     } else {
 518         f->buf_offset = pos;
 519         f->buf_index = 0;
 520         f->buf_size = 0;
 521     }
 522     return pos;
 523 }
 524
 525 int qemu_file_rate_limit(QEMUFile *f)
 526 {
 527     if (f->rate_limit)
 528         return f->rate_limit(f->opaque);
 529
 530     return 0;
 531 }
 532
 533 void qemu_put_be16(QEMUFile *f, unsigned int v)
 534 {
 535     qemu_put_byte(f, v >> 8);
 536     qemu_put_byte(f, v);
 537 }
 538
 539 void qemu_put_be32(QEMUFile *f, unsigned int v)
 540 {
 541     qemu_put_byte(f, v >> 24);
 542     qemu_put_byte(f, v >> 16);
 543     qemu_put_byte(f, v >> 8);
 544     qemu_put_byte(f, v);
 545 }
 546
 547 void qemu_put_be64(QEMUFile *f, uint64_t v)
 548 {
 549     qemu_put_be32(f, v >> 32);
 550     qemu_put_be32(f, v);
 551 }
 552
 553 unsigned int qemu_get_be16(QEMUFile *f)
 554 {
 555     unsigned int v;
 556     v = qemu_get_byte(f) << 8;
 557     v |= qemu_get_byte(f);
 558     return v;
 559 }
 560
 561 unsigned int qemu_get_be32(QEMUFile *f)
 562 {
 563     unsigned int v;
 564     v = qemu_get_byte(f) << 24;
 565     v |= qemu_get_byte(f) << 16;
 566     v |= qemu_get_byte(f) << 8;
 567     v |= qemu_get_byte(f);
 568     return v;
 569 }
 570
 571 uint64_t qemu_get_be64(QEMUFile *f)
 572 {
 573     uint64_t v;
 574     v = (uint64_t)qemu_get_be32(f) << 32;
 575     v |= qemu_get_be32(f);
 576     return v;
 577 }
 578
 579 typedef struct SaveStateEntry {
 580     char idstr[256];
 581     int instance_id;
 582     int version_id;
 583     int section_id;
 584     SaveLiveStateHandler *save_live_state;
 585     SaveStateHandler *save_state;
 586     LoadStateHandler *load_state;
 587     void *opaque;
 588     struct SaveStateEntry *next;
 589 } SaveStateEntry;
 590
 591 static SaveStateEntry *first_se;
 592
 593 /* TODO: Individual devices generally have very little idea about the rest
 594    of the system, so instance_id should be removed/replaced.
 595    Meanwhile pass -1 as instance_id if you do not already have a clearly
 596    distinguishing id for all instances of your device class. */
 597 int register_savevm_live(const char *idstr,
 598                          int instance_id,
 599                          int version_id,
 600                          SaveLiveStateHandler *save_live_state,
 601                          SaveStateHandler *save_state,
 602                          LoadStateHandler *load_state,
 603                          void *opaque)
 604 {
 605     SaveStateEntry *se, **pse;
 606     static int global_section_id;
 607
 608     se = qemu_malloc(sizeof(SaveStateEntry));
 609     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 610     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 611     se->version_id = version_id;
 612     se->section_id = global_section_id++;
 613     se->save_live_state = save_live_state;
 614     se->save_state = save_state;
 615     se->load_state = load_state;
 616     se->opaque = opaque;
 617     se->next = NULL;
 618
 619     /* add at the end of list */
 620     pse = &first_se;
 621     while (*pse != NULL) {
 622         if (instance_id == -1
 623                 && strcmp(se->idstr, (*pse)->idstr) == 0
 624                 && se->instance_id <= (*pse)->instance_id)
 625             se->instance_id = (*pse)->instance_id + 1;
 626         pse = &(*pse)->next;
 627     }
 628     *pse = se;
 629     return 0;
 630 }
 631
 632 int register_savevm(const char *idstr,
 633                     int instance_id,
 634                     int version_id,
 635                     SaveStateHandler *save_state,
 636                     LoadStateHandler *load_state,
 637                     void *opaque)
 638 {
 639     return register_savevm_live(idstr, instance_id, version_id,
 640                                 NULL, save_state, load_state, opaque);
 641 }
 642
 643 #define QEMU_VM_FILE_MAGIC           0x5145564d
 644 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 645 #define QEMU_VM_FILE_VERSION         0x00000003
 646
 647 #define QEMU_VM_EOF                  0x00
 648 #define QEMU_VM_SECTION_START        0x01
 649 #define QEMU_VM_SECTION_PART         0x02
 650 #define QEMU_VM_SECTION_END          0x03
 651 #define QEMU_VM_SECTION_FULL         0x04
 652
 653 int qemu_savevm_state_begin(QEMUFile *f)
 654 {
 655     SaveStateEntry *se;
 656
 657     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 658     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 659
 660     for (se = first_se; se != NULL; se = se->next) {
 661         int len;
 662
 663         if (se->save_live_state == NULL)
 664             continue;
 665
 666         /* Section type */
 667         qemu_put_byte(f, QEMU_VM_SECTION_START);
 668         qemu_put_be32(f, se->section_id);
 669
 670         /* ID string */
 671         len = strlen(se->idstr);
 672         qemu_put_byte(f, len);
 673         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 674
 675         qemu_put_be32(f, se->instance_id);
 676         qemu_put_be32(f, se->version_id);
 677
 678         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 679     }
 680
 681     if (qemu_file_has_error(f))
 682         return -EIO;
 683
 684     return 0;
 685 }
 686
 687 int qemu_savevm_state_iterate(QEMUFile *f)
 688 {
 689     SaveStateEntry *se;
 690     int ret = 1;
 691
 692     for (se = first_se; se != NULL; se = se->next) {
 693         if (se->save_live_state == NULL)
 694             continue;
 695
 696         /* Section type */
 697         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 698         qemu_put_be32(f, se->section_id);
 699
 700         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 701     }
 702
 703     if (ret)
 704         return 1;
 705
 706     if (qemu_file_has_error(f))
 707         return -EIO;
 708
 709     return 0;
 710 }
 711
 712 int qemu_savevm_state_complete(QEMUFile *f)
 713 {
 714     SaveStateEntry *se;
 715
 716     for (se = first_se; se != NULL; se = se->next) {
 717         if (se->save_live_state == NULL)
 718             continue;
 719
 720         /* Section type */
 721         qemu_put_byte(f, QEMU_VM_SECTION_END);
 722         qemu_put_be32(f, se->section_id);
 723
 724         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 725     }
 726
 727     for(se = first_se; se != NULL; se = se->next) {
 728         int len;
 729
 730         if (se->save_state == NULL)
 731             continue;
 732
 733         /* Section type */
 734         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 735         qemu_put_be32(f, se->section_id);
 736
 737         /* ID string */
 738         len = strlen(se->idstr);
 739         qemu_put_byte(f, len);
 740         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 741
 742         qemu_put_be32(f, se->instance_id);
 743         qemu_put_be32(f, se->version_id);
 744
 745         se->save_state(f, se->opaque);
 746     }
 747
 748     qemu_put_byte(f, QEMU_VM_EOF);
 749
 750     if (qemu_file_has_error(f))
 751         return -EIO;
 752
 753     return 0;
 754 }
 755
 756 int qemu_savevm_state(QEMUFile *f)
 757 {
 758     int saved_vm_running;
 759     int ret;
 760
 761     saved_vm_running = vm_running;
 762     vm_stop(0);
 763
 764     bdrv_flush_all();
 765
 766     ret = qemu_savevm_state_begin(f);
 767     if (ret < 0)
 768         goto out;
 769
 770     do {
 771         ret = qemu_savevm_state_iterate(f);
 772         if (ret < 0)
 773             goto out;
 774     } while (ret == 0);
 775
 776     ret = qemu_savevm_state_complete(f);
 777
 778 out:
 779     if (qemu_file_has_error(f))
 780         ret = -EIO;
 781
 782     if (!ret && saved_vm_running)
 783         vm_start();
 784
 785     return ret;
 786 }
 787
 788 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 789 {
 790     SaveStateEntry *se;
 791
 792     for(se = first_se; se != NULL; se = se->next) {
 793         if (!strcmp(se->idstr, idstr) &&
 794             instance_id == se->instance_id)
 795             return se;
 796     }
 797     return NULL;
 798 }
 799
 800 typedef struct LoadStateEntry {
 801     SaveStateEntry *se;
 802     int section_id;
 803     int version_id;
 804     struct LoadStateEntry *next;
 805 } LoadStateEntry;
 806
 807 static int qemu_loadvm_state_v2(QEMUFile *f)
 808 {
 809     SaveStateEntry *se;
 810     int len, ret, instance_id, record_len, version_id;
 811     int64_t total_len, end_pos, cur_pos;
 812     char idstr[256];
 813
 814     total_len = qemu_get_be64(f);
 815     end_pos = total_len + qemu_ftell(f);
 816     for(;;) {
 817         if (qemu_ftell(f) >= end_pos)
 818             break;
 819         len = qemu_get_byte(f);
 820         qemu_get_buffer(f, (uint8_t *)idstr, len);
 821         idstr[len] = '\0';
 822         instance_id = qemu_get_be32(f);
 823         version_id = qemu_get_be32(f);
 824         record_len = qemu_get_be32(f);
 825         cur_pos = qemu_ftell(f);
 826         se = find_se(idstr, instance_id);
 827         if (!se) {
 828             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 829                     instance_id, idstr);
 830         } else {
 831             ret = se->load_state(f, se->opaque, version_id);
 832             if (ret < 0) {
 833                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 834                         instance_id, idstr);
 835             }
 836         }
 837         /* always seek to exact end of record */
 838         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 839     }
 840
 841     if (qemu_file_has_error(f))
 842         return -EIO;
 843
 844     return 0;
 845 }
 846
 847 int qemu_loadvm_state(QEMUFile *f)
 848 {
 849     LoadStateEntry *first_le = NULL;
 850     uint8_t section_type;
 851     unsigned int v;
 852     int ret;
 853
 854     v = qemu_get_be32(f);
 855     if (v != QEMU_VM_FILE_MAGIC)
 856         return -EINVAL;
 857
 858     v = qemu_get_be32(f);
 859     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 860         return qemu_loadvm_state_v2(f);
 861     if (v != QEMU_VM_FILE_VERSION)
 862         return -ENOTSUP;
 863
 864     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 865         uint32_t instance_id, version_id, section_id;
 866         LoadStateEntry *le;
 867         SaveStateEntry *se;
 868         char idstr[257];
 869         int len;
 870
 871         switch (section_type) {
 872         case QEMU_VM_SECTION_START:
 873         case QEMU_VM_SECTION_FULL:
 874             /* Read section start */
 875             section_id = qemu_get_be32(f);
 876             len = qemu_get_byte(f);
 877             qemu_get_buffer(f, (uint8_t *)idstr, len);
 878             idstr[len] = 0;
 879             instance_id = qemu_get_be32(f);
 880             version_id = qemu_get_be32(f);
 881
 882             /* Find savevm section */
 883             se = find_se(idstr, instance_id);
 884             if (se == NULL) {
 885                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 886                 ret = -EINVAL;
 887                 goto out;
 888             }
 889
 890             /* Validate version */
 891             if (version_id > se->version_id) {
 892                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 893                         version_id, idstr, se->version_id);
 894                 ret = -EINVAL;
 895                 goto out;
 896             }
 897
 898             /* Add entry */
 899             le = qemu_mallocz(sizeof(*le));
 900
 901             le->se = se;
 902             le->section_id = section_id;
 903             le->version_id = version_id;
 904             le->next = first_le;
 905             first_le = le;
 906
 907             le->se->load_state(f, le->se->opaque, le->version_id);
 908             break;
 909         case QEMU_VM_SECTION_PART:
 910         case QEMU_VM_SECTION_END:
 911             section_id = qemu_get_be32(f);
 912
 913             for (le = first_le; le && le->section_id != section_id; le = le->next);
 914             if (le == NULL) {
 915                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 916                 ret = -EINVAL;
 917                 goto out;
 918             }
 919
 920             le->se->load_state(f, le->se->opaque, le->version_id);
 921             break;
 922         default:
 923             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 924             ret = -EINVAL;
 925             goto out;
 926         }
 927     }
 928
 929     ret = 0;
 930
 931 out:
 932     while (first_le) {
 933         LoadStateEntry *le = first_le;
 934         first_le = first_le->next;
 935         qemu_free(le);
 936     }
 937
 938     if (qemu_file_has_error(f))
 939         ret = -EIO;
 940
 941     return ret;
 942 }
 943
 944 /* device can contain snapshots */
 945 static int bdrv_can_snapshot(BlockDriverState *bs)
 946 {
 947     return (bs &&
 948             !bdrv_is_removable(bs) &&
 949             !bdrv_is_read_only(bs));
 950 }
 951
 952 /* device must be snapshots in order to have a reliable snapshot */
 953 static int bdrv_has_snapshot(BlockDriverState *bs)
 954 {
 955     return (bs &&
 956             !bdrv_is_removable(bs) &&
 957             !bdrv_is_read_only(bs));
 958 }
 959
 960 static BlockDriverState *get_bs_snapshots(void)
 961 {
 962     BlockDriverState *bs;
 963     int i;
 964
 965     if (bs_snapshots)
 966         return bs_snapshots;
 967     for(i = 0; i <= nb_drives; i++) {
 968         bs = drives_table[i].bdrv;
 969         if (bdrv_can_snapshot(bs))
 970             goto ok;
 971     }
 972     return NULL;
 973  ok:
 974     bs_snapshots = bs;
 975     return bs;
 976 }
 977
 978 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
 979                               const char *name)
 980 {
 981     QEMUSnapshotInfo *sn_tab, *sn;
 982     int nb_sns, i, ret;
 983
 984     ret = -ENOENT;
 985     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
 986     if (nb_sns < 0)
 987         return ret;
 988     for(i = 0; i < nb_sns; i++) {
 989         sn = &sn_tab[i];
 990         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
 991             *sn_info = *sn;
 992             ret = 0;
 993             break;
 994         }
 995     }
 996     qemu_free(sn_tab);
 997     return ret;
 998 }
 999
1000 void do_savevm(Monitor *mon, const char *name)
1001 {
1002     BlockDriverState *bs, *bs1;
1003     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1004     int must_delete, ret, i;
1005     BlockDriverInfo bdi1, *bdi = &bdi1;
1006     QEMUFile *f;
1007     int saved_vm_running;
1008     uint32_t vm_state_size;
1009 #ifdef _WIN32
1010     struct _timeb tb;
1011 #else
1012     struct timeval tv;
1013 #endif
1014
1015     bs = get_bs_snapshots();
1016     if (!bs) {
1017         monitor_printf(mon, "No block device can accept snapshots\n");
1018         return;
1019     }
1020
1021     /* ??? Should this occur after vm_stop?  */
1022     qemu_aio_flush();
1023
1024     saved_vm_running = vm_running;
1025     vm_stop(0);
1026
1027     must_delete = 0;
1028     if (name) {
1029         ret = bdrv_snapshot_find(bs, old_sn, name);
1030         if (ret >= 0) {
1031             must_delete = 1;
1032         }
1033     }
1034     memset(sn, 0, sizeof(*sn));
1035     if (must_delete) {
1036         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1037         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1038     } else {
1039         if (name)
1040             pstrcpy(sn->name, sizeof(sn->name), name);
1041     }
1042
1043     /* fill auxiliary fields */
1044 #ifdef _WIN32
1045     _ftime(&tb);
1046     sn->date_sec = tb.time;
1047     sn->date_nsec = tb.millitm * 1000000;
1048 #else
1049     gettimeofday(&tv, NULL);
1050     sn->date_sec = tv.tv_sec;
1051     sn->date_nsec = tv.tv_usec * 1000;
1052 #endif
1053     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1054
1055     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1056         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1057                        bdrv_get_device_name(bs));
1058         goto the_end;
1059     }
1060
1061     /* save the VM state */
1062     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 1);
1063     if (!f) {
1064         monitor_printf(mon, "Could not open VM state file\n");
1065         goto the_end;
1066     }
1067     ret = qemu_savevm_state(f);
1068     vm_state_size = qemu_ftell(f);
1069     qemu_fclose(f);
1070     if (ret < 0) {
1071         monitor_printf(mon, "Error %d while writing VM\n", ret);
1072         goto the_end;
1073     }
1074
1075     /* create the snapshots */
1076
1077     for(i = 0; i < nb_drives; i++) {
1078         bs1 = drives_table[i].bdrv;
1079         if (bdrv_has_snapshot(bs1)) {
1080             if (must_delete) {
1081                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1082                 if (ret < 0) {
1083                     monitor_printf(mon,
1084                                    "Error while deleting snapshot on '%s'\n",
1085                                    bdrv_get_device_name(bs1));
1086                 }
1087             }
1088             /* Write VM state size only to the image that contains the state */
1089             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1090             ret = bdrv_snapshot_create(bs1, sn);
1091             if (ret < 0) {
1092                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1093                                bdrv_get_device_name(bs1));
1094             }
1095         }
1096     }
1097
1098  the_end:
1099     if (saved_vm_running)
1100         vm_start();
1101 }
1102
1103 void do_loadvm(Monitor *mon, const char *name)
1104 {
1105     BlockDriverState *bs, *bs1;
1106     BlockDriverInfo bdi1, *bdi = &bdi1;
1107     QEMUSnapshotInfo sn;
1108     QEMUFile *f;
1109     int i, ret;
1110     int saved_vm_running;
1111
1112     bs = get_bs_snapshots();
1113     if (!bs) {
1114         monitor_printf(mon, "No block device supports snapshots\n");
1115         return;
1116     }
1117
1118     /* Flush all IO requests so they don't interfere with the new state.  */
1119     qemu_aio_flush();
1120
1121     saved_vm_running = vm_running;
1122     vm_stop(0);
1123
1124     for(i = 0; i <= nb_drives; i++) {
1125         bs1 = drives_table[i].bdrv;
1126         if (bdrv_has_snapshot(bs1)) {
1127             ret = bdrv_snapshot_goto(bs1, name);
1128             if (ret < 0) {
1129                 if (bs != bs1)
1130                     monitor_printf(mon, "Warning: ");
1131                 switch(ret) {
1132                 case -ENOTSUP:
1133                     monitor_printf(mon,
1134                                    "Snapshots not supported on device '%s'\n",
1135                                    bdrv_get_device_name(bs1));
1136                     break;
1137                 case -ENOENT:
1138                     monitor_printf(mon, "Could not find snapshot '%s' on "
1139                                    "device '%s'\n",
1140                                    name, bdrv_get_device_name(bs1));
1141                     break;
1142                 default:
1143                     monitor_printf(mon, "Error %d while activating snapshot on"
1144                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1145                     break;
1146                 }
1147                 /* fatal on snapshot block device */
1148                 if (bs == bs1)
1149                     goto the_end;
1150             }
1151         }
1152     }
1153
1154     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1155         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1156                        bdrv_get_device_name(bs));
1157         return;
1158     }
1159
1160     /* Don't even try to load empty VM states */
1161     ret = bdrv_snapshot_find(bs, &sn, name);
1162     if ((ret >= 0) && (sn.vm_state_size == 0))
1163         goto the_end;
1164
1165     /* restore the VM state */
1166     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 0);
1167     if (!f) {
1168         monitor_printf(mon, "Could not open VM state file\n");
1169         goto the_end;
1170     }
1171     ret = qemu_loadvm_state(f);
1172     qemu_fclose(f);
1173     if (ret < 0) {
1174         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1175     }
1176  the_end:
1177     if (saved_vm_running)
1178         vm_start();
1179 }
1180
1181 void do_delvm(Monitor *mon, const char *name)
1182 {
1183     BlockDriverState *bs, *bs1;
1184     int i, ret;
1185
1186     bs = get_bs_snapshots();
1187     if (!bs) {
1188         monitor_printf(mon, "No block device supports snapshots\n");
1189         return;
1190     }
1191
1192     for(i = 0; i <= nb_drives; i++) {
1193         bs1 = drives_table[i].bdrv;
1194         if (bdrv_has_snapshot(bs1)) {
1195             ret = bdrv_snapshot_delete(bs1, name);
1196             if (ret < 0) {
1197                 if (ret == -ENOTSUP)
1198                     monitor_printf(mon,
1199                                    "Snapshots not supported on device '%s'\n",
1200                                    bdrv_get_device_name(bs1));
1201                 else
1202                     monitor_printf(mon, "Error %d while deleting snapshot on "
1203                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1204             }
1205         }
1206     }
1207 }
1208
1209 void do_info_snapshots(Monitor *mon)
1210 {
1211     BlockDriverState *bs, *bs1;
1212     QEMUSnapshotInfo *sn_tab, *sn;
1213     int nb_sns, i;
1214     char buf[256];
1215
1216     bs = get_bs_snapshots();
1217     if (!bs) {
1218         monitor_printf(mon, "No available block device supports snapshots\n");
1219         return;
1220     }
1221     monitor_printf(mon, "Snapshot devices:");
1222     for(i = 0; i <= nb_drives; i++) {
1223         bs1 = drives_table[i].bdrv;
1224         if (bdrv_has_snapshot(bs1)) {
1225             if (bs == bs1)
1226                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1227         }
1228     }
1229     monitor_printf(mon, "\n");
1230
1231     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1232     if (nb_sns < 0) {
1233         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1234         return;
1235     }
1236     monitor_printf(mon, "Snapshot list (from %s):\n",
1237                    bdrv_get_device_name(bs));
1238     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1239     for(i = 0; i < nb_sns; i++) {
1240         sn = &sn_tab[i];
1241         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1242     }
1243     qemu_free(sn_tab);
1244 }