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