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, 0xff, 6);         /* h_dst */
 107     memcpy(buf + 6, mac_addr, 6); /* h_src */
 108     memcpy(buf + 12, &proto, 2);  /* h_proto */
 109     memcpy(buf + 14, &magic, 4);  /* magic */
 110
 111     return 18; /* len */
 112 }
 113
 114 void qemu_announce_self(void)
 115 {
 116     int i, j, len;
 117     VLANState *vlan;
 118     VLANClientState *vc;
 119     uint8_t buf[256];
 120
 121     for (i = 0; i < MAX_NICS; i++) {
 122         if (!nd_table[i].used)
 123             continue;
 124         len = announce_self_create(buf, nd_table[i].macaddr);
 125         vlan = nd_table[i].vlan;
 126         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 127             for (j=0; j < SELF_ANNOUNCE_ROUNDS; j++)
 128                 vc->fd_read(vc->opaque, buf, len);
 129         }
 130     }
 131 }
 132
 133 /***********************************************************/
 134 /* savevm/loadvm support */
 135
 136 #define IO_BUF_SIZE 32768
 137
 138 struct QEMUFile {
 139     QEMUFilePutBufferFunc *put_buffer;
 140     QEMUFileGetBufferFunc *get_buffer;
 141     QEMUFileCloseFunc *close;
 142     QEMUFileRateLimit *rate_limit;
 143     void *opaque;
 144     int is_write;
 145
 146     int64_t buf_offset; /* start of buffer when writing, end of buffer
 147                            when reading */
 148     int buf_index;
 149     int buf_size; /* 0 when writing */
 150     uint8_t buf[IO_BUF_SIZE];
 151
 152     int has_error;
 153 };
 154
 155 typedef struct QEMUFilePopen
 156 {
 157     FILE *popen_file;
 158     QEMUFile *file;
 159 } QEMUFilePopen;
 160
 161 typedef struct QEMUFileSocket
 162 {
 163     int fd;
 164     QEMUFile *file;
 165 } QEMUFileSocket;
 166
 167 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 168 {
 169     QEMUFileSocket *s = opaque;
 170     ssize_t len;
 171
 172     do {
 173         len = recv(s->fd, buf, size, 0);
 174     } while (len == -1 && socket_error() == EINTR);
 175
 176     if (len == -1)
 177         len = -socket_error();
 178
 179     return len;
 180 }
 181
 182 static int socket_close(void *opaque)
 183 {
 184     QEMUFileSocket *s = opaque;
 185     qemu_free(s);
 186     return 0;
 187 }
 188
 189 static int popen_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 190 {
 191     QEMUFilePopen *s = opaque;
 192     return fwrite(buf, 1, size, s->popen_file);
 193 }
 194
 195 static int popen_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 196 {
 197     QEMUFilePopen *s = opaque;
 198     return fread(buf, 1, size, s->popen_file);
 199 }
 200
 201 static int popen_close(void *opaque)
 202 {
 203     QEMUFilePopen *s = opaque;
 204     pclose(s->popen_file);
 205     qemu_free(s);
 206     return 0;
 207 }
 208
 209 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 210 {
 211     QEMUFilePopen *s;
 212
 213     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 214         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 215         return NULL;
 216     }
 217
 218     s = qemu_mallocz(sizeof(QEMUFilePopen));
 219
 220     s->popen_file = popen_file;
 221
 222     if(mode[0] == 'r') {
 223         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL);
 224     } else {
 225         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL);
 226     }
 227     fprintf(stderr, "qemu_popen: returning result of qemu_fopen_ops\n");
 228     return s->file;
 229 }
 230
 231 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 232 {
 233     FILE *popen_file;
 234
 235     popen_file = popen(command, mode);
 236     if(popen_file == NULL) {
 237         return NULL;
 238     }
 239
 240     return qemu_popen(popen_file, mode);
 241 }
 242
 243 QEMUFile *qemu_fopen_socket(int fd)
 244 {
 245     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 246
 247     s->fd = fd;
 248     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL);
 249     return s->file;
 250 }
 251
 252 typedef struct QEMUFileStdio
 253 {
 254     FILE *outfile;
 255 } QEMUFileStdio;
 256
 257 static int file_put_buffer(void *opaque, const uint8_t *buf,
 258                             int64_t pos, int size)
 259 {
 260     QEMUFileStdio *s = opaque;
 261     fseek(s->outfile, pos, SEEK_SET);
 262     fwrite(buf, 1, size, s->outfile);
 263     return size;
 264 }
 265
 266 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 267 {
 268     QEMUFileStdio *s = opaque;
 269     fseek(s->outfile, pos, SEEK_SET);
 270     return fread(buf, 1, size, s->outfile);
 271 }
 272
 273 static int file_close(void *opaque)
 274 {
 275     QEMUFileStdio *s = opaque;
 276     fclose(s->outfile);
 277     qemu_free(s);
 278     return 0;
 279 }
 280
 281 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 282 {
 283     QEMUFileStdio *s;
 284
 285     s = qemu_mallocz(sizeof(QEMUFileStdio));
 286
 287     s->outfile = fopen(filename, mode);
 288     if (!s->outfile)
 289         goto fail;
 290
 291     if (!strcmp(mode, "wb"))
 292         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL);
 293     else if (!strcmp(mode, "rb"))
 294         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL);
 295
 296 fail:
 297     if (s->outfile)
 298         fclose(s->outfile);
 299     qemu_free(s);
 300     return NULL;
 301 }
 302
 303 typedef struct QEMUFileBdrv
 304 {
 305     BlockDriverState *bs;
 306     int64_t base_offset;
 307 } QEMUFileBdrv;
 308
 309 static int block_put_buffer(void *opaque, const uint8_t *buf,
 310                            int64_t pos, int size)
 311 {
 312     QEMUFileBdrv *s = opaque;
 313     bdrv_put_buffer(s->bs, buf, s->base_offset + pos, size);
 314     return size;
 315 }
 316
 317 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 318 {
 319     QEMUFileBdrv *s = opaque;
 320     return bdrv_get_buffer(s->bs, buf, s->base_offset + pos, size);
 321 }
 322
 323 static int bdrv_fclose(void *opaque)
 324 {
 325     QEMUFileBdrv *s = opaque;
 326     qemu_free(s);
 327     return 0;
 328 }
 329
 330 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable)
 331 {
 332     QEMUFileBdrv *s;
 333
 334     s = qemu_mallocz(sizeof(QEMUFileBdrv));
 335
 336     s->bs = bs;
 337     s->base_offset = offset;
 338
 339     if (is_writable)
 340         return qemu_fopen_ops(s, block_put_buffer, NULL, bdrv_fclose, NULL);
 341
 342     return qemu_fopen_ops(s, NULL, block_get_buffer, bdrv_fclose, NULL);
 343 }
 344
 345 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 346                          QEMUFileGetBufferFunc *get_buffer,
 347                          QEMUFileCloseFunc *close,
 348                          QEMUFileRateLimit *rate_limit)
 349 {
 350     QEMUFile *f;
 351
 352     f = qemu_mallocz(sizeof(QEMUFile));
 353
 354     f->opaque = opaque;
 355     f->put_buffer = put_buffer;
 356     f->get_buffer = get_buffer;
 357     f->close = close;
 358     f->rate_limit = rate_limit;
 359     f->is_write = 0;
 360
 361     return f;
 362 }
 363
 364 int qemu_file_has_error(QEMUFile *f)
 365 {
 366     return f->has_error;
 367 }
 368
 369 void qemu_file_set_error(QEMUFile *f)
 370 {
 371     f->has_error = 1;
 372 }
 373
 374 void qemu_fflush(QEMUFile *f)
 375 {
 376     if (!f->put_buffer)
 377         return;
 378
 379     if (f->is_write && f->buf_index > 0) {
 380         int len;
 381
 382         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 383         if (len > 0)
 384             f->buf_offset += f->buf_index;
 385         else
 386             f->has_error = 1;
 387         f->buf_index = 0;
 388     }
 389 }
 390
 391 static void qemu_fill_buffer(QEMUFile *f)
 392 {
 393     int len;
 394
 395     if (!f->get_buffer)
 396         return;
 397
 398     if (f->is_write)
 399         abort();
 400
 401     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 402     if (len > 0) {
 403         f->buf_index = 0;
 404         f->buf_size = len;
 405         f->buf_offset += len;
 406     } else if (len != -EAGAIN)
 407         f->has_error = 1;
 408 }
 409
 410 int qemu_fclose(QEMUFile *f)
 411 {
 412     int ret = 0;
 413     qemu_fflush(f);
 414     if (f->close)
 415         ret = f->close(f->opaque);
 416     qemu_free(f);
 417     return ret;
 418 }
 419
 420 void qemu_file_put_notify(QEMUFile *f)
 421 {
 422     f->put_buffer(f->opaque, NULL, 0, 0);
 423 }
 424
 425 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 426 {
 427     int l;
 428
 429     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 430         fprintf(stderr,
 431                 "Attempted to write to buffer while read buffer is not empty\n");
 432         abort();
 433     }
 434
 435     while (!f->has_error && size > 0) {
 436         l = IO_BUF_SIZE - f->buf_index;
 437         if (l > size)
 438             l = size;
 439         memcpy(f->buf + f->buf_index, buf, l);
 440         f->is_write = 1;
 441         f->buf_index += l;
 442         buf += l;
 443         size -= l;
 444         if (f->buf_index >= IO_BUF_SIZE)
 445             qemu_fflush(f);
 446     }
 447 }
 448
 449 void qemu_put_byte(QEMUFile *f, int v)
 450 {
 451     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 452         fprintf(stderr,
 453                 "Attempted to write to buffer while read buffer is not empty\n");
 454         abort();
 455     }
 456
 457     f->buf[f->buf_index++] = v;
 458     f->is_write = 1;
 459     if (f->buf_index >= IO_BUF_SIZE)
 460         qemu_fflush(f);
 461 }
 462
 463 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 464 {
 465     int size, l;
 466
 467     if (f->is_write)
 468         abort();
 469
 470     size = size1;
 471     while (size > 0) {
 472         l = f->buf_size - f->buf_index;
 473         if (l == 0) {
 474             qemu_fill_buffer(f);
 475             l = f->buf_size - f->buf_index;
 476             if (l == 0)
 477                 break;
 478         }
 479         if (l > size)
 480             l = size;
 481         memcpy(buf, f->buf + f->buf_index, l);
 482         f->buf_index += l;
 483         buf += l;
 484         size -= l;
 485     }
 486     return size1 - size;
 487 }
 488
 489 int qemu_get_byte(QEMUFile *f)
 490 {
 491     if (f->is_write)
 492         abort();
 493
 494     if (f->buf_index >= f->buf_size) {
 495         qemu_fill_buffer(f);
 496         if (f->buf_index >= f->buf_size)
 497             return 0;
 498     }
 499     return f->buf[f->buf_index++];
 500 }
 501
 502 int64_t qemu_ftell(QEMUFile *f)
 503 {
 504     return f->buf_offset - f->buf_size + f->buf_index;
 505 }
 506
 507 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 508 {
 509     if (whence == SEEK_SET) {
 510         /* nothing to do */
 511     } else if (whence == SEEK_CUR) {
 512         pos += qemu_ftell(f);
 513     } else {
 514         /* SEEK_END not supported */
 515         return -1;
 516     }
 517     if (f->put_buffer) {
 518         qemu_fflush(f);
 519         f->buf_offset = pos;
 520     } else {
 521         f->buf_offset = pos;
 522         f->buf_index = 0;
 523         f->buf_size = 0;
 524     }
 525     return pos;
 526 }
 527
 528 int qemu_file_rate_limit(QEMUFile *f)
 529 {
 530     if (f->rate_limit)
 531         return f->rate_limit(f->opaque);
 532
 533     return 0;
 534 }
 535
 536 void qemu_put_be16(QEMUFile *f, unsigned int v)
 537 {
 538     qemu_put_byte(f, v >> 8);
 539     qemu_put_byte(f, v);
 540 }
 541
 542 void qemu_put_be32(QEMUFile *f, unsigned int v)
 543 {
 544     qemu_put_byte(f, v >> 24);
 545     qemu_put_byte(f, v >> 16);
 546     qemu_put_byte(f, v >> 8);
 547     qemu_put_byte(f, v);
 548 }
 549
 550 void qemu_put_be64(QEMUFile *f, uint64_t v)
 551 {
 552     qemu_put_be32(f, v >> 32);
 553     qemu_put_be32(f, v);
 554 }
 555
 556 unsigned int qemu_get_be16(QEMUFile *f)
 557 {
 558     unsigned int v;
 559     v = qemu_get_byte(f) << 8;
 560     v |= qemu_get_byte(f);
 561     return v;
 562 }
 563
 564 unsigned int qemu_get_be32(QEMUFile *f)
 565 {
 566     unsigned int v;
 567     v = qemu_get_byte(f) << 24;
 568     v |= qemu_get_byte(f) << 16;
 569     v |= qemu_get_byte(f) << 8;
 570     v |= qemu_get_byte(f);
 571     return v;
 572 }
 573
 574 uint64_t qemu_get_be64(QEMUFile *f)
 575 {
 576     uint64_t v;
 577     v = (uint64_t)qemu_get_be32(f) << 32;
 578     v |= qemu_get_be32(f);
 579     return v;
 580 }
 581
 582 typedef struct SaveStateEntry {
 583     char idstr[256];
 584     int instance_id;
 585     int version_id;
 586     int section_id;
 587     SaveLiveStateHandler *save_live_state;
 588     SaveStateHandler *save_state;
 589     LoadStateHandler *load_state;
 590     void *opaque;
 591     struct SaveStateEntry *next;
 592 } SaveStateEntry;
 593
 594 static SaveStateEntry *first_se;
 595
 596 /* TODO: Individual devices generally have very little idea about the rest
 597    of the system, so instance_id should be removed/replaced.
 598    Meanwhile pass -1 as instance_id if you do not already have a clearly
 599    distinguishing id for all instances of your device class. */
 600 int register_savevm_live(const char *idstr,
 601                          int instance_id,
 602                          int version_id,
 603                          SaveLiveStateHandler *save_live_state,
 604                          SaveStateHandler *save_state,
 605                          LoadStateHandler *load_state,
 606                          void *opaque)
 607 {
 608     SaveStateEntry *se, **pse;
 609     static int global_section_id;
 610
 611     se = qemu_malloc(sizeof(SaveStateEntry));
 612     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 613     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 614     se->version_id = version_id;
 615     se->section_id = global_section_id++;
 616     se->save_live_state = save_live_state;
 617     se->save_state = save_state;
 618     se->load_state = load_state;
 619     se->opaque = opaque;
 620     se->next = NULL;
 621
 622     /* add at the end of list */
 623     pse = &first_se;
 624     while (*pse != NULL) {
 625         if (instance_id == -1
 626                 && strcmp(se->idstr, (*pse)->idstr) == 0
 627                 && se->instance_id <= (*pse)->instance_id)
 628             se->instance_id = (*pse)->instance_id + 1;
 629         pse = &(*pse)->next;
 630     }
 631     *pse = se;
 632     return 0;
 633 }
 634
 635 int register_savevm(const char *idstr,
 636                     int instance_id,
 637                     int version_id,
 638                     SaveStateHandler *save_state,
 639                     LoadStateHandler *load_state,
 640                     void *opaque)
 641 {
 642     return register_savevm_live(idstr, instance_id, version_id,
 643                                 NULL, save_state, load_state, opaque);
 644 }
 645
 646 #define QEMU_VM_FILE_MAGIC           0x5145564d
 647 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 648 #define QEMU_VM_FILE_VERSION         0x00000003
 649
 650 #define QEMU_VM_EOF                  0x00
 651 #define QEMU_VM_SECTION_START        0x01
 652 #define QEMU_VM_SECTION_PART         0x02
 653 #define QEMU_VM_SECTION_END          0x03
 654 #define QEMU_VM_SECTION_FULL         0x04
 655
 656 int qemu_savevm_state_begin(QEMUFile *f)
 657 {
 658     SaveStateEntry *se;
 659
 660     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 661     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 662
 663     for (se = first_se; se != NULL; se = se->next) {
 664         int len;
 665
 666         if (se->save_live_state == NULL)
 667             continue;
 668
 669         /* Section type */
 670         qemu_put_byte(f, QEMU_VM_SECTION_START);
 671         qemu_put_be32(f, se->section_id);
 672
 673         /* ID string */
 674         len = strlen(se->idstr);
 675         qemu_put_byte(f, len);
 676         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 677
 678         qemu_put_be32(f, se->instance_id);
 679         qemu_put_be32(f, se->version_id);
 680
 681         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 682     }
 683
 684     if (qemu_file_has_error(f))
 685         return -EIO;
 686
 687     return 0;
 688 }
 689
 690 int qemu_savevm_state_iterate(QEMUFile *f)
 691 {
 692     SaveStateEntry *se;
 693     int ret = 1;
 694
 695     for (se = first_se; se != NULL; se = se->next) {
 696         if (se->save_live_state == NULL)
 697             continue;
 698
 699         /* Section type */
 700         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 701         qemu_put_be32(f, se->section_id);
 702
 703         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 704     }
 705
 706     if (ret)
 707         return 1;
 708
 709     if (qemu_file_has_error(f))
 710         return -EIO;
 711
 712     return 0;
 713 }
 714
 715 int qemu_savevm_state_complete(QEMUFile *f)
 716 {
 717     SaveStateEntry *se;
 718
 719     for (se = first_se; se != NULL; se = se->next) {
 720         if (se->save_live_state == NULL)
 721             continue;
 722
 723         /* Section type */
 724         qemu_put_byte(f, QEMU_VM_SECTION_END);
 725         qemu_put_be32(f, se->section_id);
 726
 727         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 728     }
 729
 730     for(se = first_se; se != NULL; se = se->next) {
 731         int len;
 732
 733         if (se->save_state == NULL)
 734             continue;
 735
 736         /* Section type */
 737         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 738         qemu_put_be32(f, se->section_id);
 739
 740         /* ID string */
 741         len = strlen(se->idstr);
 742         qemu_put_byte(f, len);
 743         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 744
 745         qemu_put_be32(f, se->instance_id);
 746         qemu_put_be32(f, se->version_id);
 747
 748         se->save_state(f, se->opaque);
 749     }
 750
 751     qemu_put_byte(f, QEMU_VM_EOF);
 752
 753     if (qemu_file_has_error(f))
 754         return -EIO;
 755
 756     return 0;
 757 }
 758
 759 int qemu_savevm_state(QEMUFile *f)
 760 {
 761     int saved_vm_running;
 762     int ret;
 763
 764     saved_vm_running = vm_running;
 765     vm_stop(0);
 766
 767     bdrv_flush_all();
 768
 769     ret = qemu_savevm_state_begin(f);
 770     if (ret < 0)
 771         goto out;
 772
 773     do {
 774         ret = qemu_savevm_state_iterate(f);
 775         if (ret < 0)
 776             goto out;
 777     } while (ret == 0);
 778
 779     ret = qemu_savevm_state_complete(f);
 780
 781 out:
 782     if (qemu_file_has_error(f))
 783         ret = -EIO;
 784
 785     if (!ret && saved_vm_running)
 786         vm_start();
 787
 788     return ret;
 789 }
 790
 791 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 792 {
 793     SaveStateEntry *se;
 794
 795     for(se = first_se; se != NULL; se = se->next) {
 796         if (!strcmp(se->idstr, idstr) &&
 797             instance_id == se->instance_id)
 798             return se;
 799     }
 800     return NULL;
 801 }
 802
 803 typedef struct LoadStateEntry {
 804     SaveStateEntry *se;
 805     int section_id;
 806     int version_id;
 807     struct LoadStateEntry *next;
 808 } LoadStateEntry;
 809
 810 static int qemu_loadvm_state_v2(QEMUFile *f)
 811 {
 812     SaveStateEntry *se;
 813     int len, ret, instance_id, record_len, version_id;
 814     int64_t total_len, end_pos, cur_pos;
 815     char idstr[256];
 816
 817     total_len = qemu_get_be64(f);
 818     end_pos = total_len + qemu_ftell(f);
 819     for(;;) {
 820         if (qemu_ftell(f) >= end_pos)
 821             break;
 822         len = qemu_get_byte(f);
 823         qemu_get_buffer(f, (uint8_t *)idstr, len);
 824         idstr[len] = '\0';
 825         instance_id = qemu_get_be32(f);
 826         version_id = qemu_get_be32(f);
 827         record_len = qemu_get_be32(f);
 828         cur_pos = qemu_ftell(f);
 829         se = find_se(idstr, instance_id);
 830         if (!se) {
 831             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 832                     instance_id, idstr);
 833         } else {
 834             ret = se->load_state(f, se->opaque, version_id);
 835             if (ret < 0) {
 836                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 837                         instance_id, idstr);
 838                 return ret;
 839             }
 840         }
 841         /* always seek to exact end of record */
 842         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 843     }
 844
 845     if (qemu_file_has_error(f))
 846         return -EIO;
 847
 848     return 0;
 849 }
 850
 851 int qemu_loadvm_state(QEMUFile *f)
 852 {
 853     LoadStateEntry *first_le = NULL;
 854     uint8_t section_type;
 855     unsigned int v;
 856     int ret;
 857
 858     v = qemu_get_be32(f);
 859     if (v != QEMU_VM_FILE_MAGIC)
 860         return -EINVAL;
 861
 862     v = qemu_get_be32(f);
 863     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 864         return qemu_loadvm_state_v2(f);
 865     if (v != QEMU_VM_FILE_VERSION)
 866         return -ENOTSUP;
 867
 868     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 869         uint32_t instance_id, version_id, section_id;
 870         LoadStateEntry *le;
 871         SaveStateEntry *se;
 872         char idstr[257];
 873         int len;
 874
 875         switch (section_type) {
 876         case QEMU_VM_SECTION_START:
 877         case QEMU_VM_SECTION_FULL:
 878             /* Read section start */
 879             section_id = qemu_get_be32(f);
 880             len = qemu_get_byte(f);
 881             qemu_get_buffer(f, (uint8_t *)idstr, len);
 882             idstr[len] = 0;
 883             instance_id = qemu_get_be32(f);
 884             version_id = qemu_get_be32(f);
 885
 886             /* Find savevm section */
 887             se = find_se(idstr, instance_id);
 888             if (se == NULL) {
 889                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 890                 ret = -EINVAL;
 891                 goto out;
 892             }
 893
 894             /* Validate version */
 895             if (version_id > se->version_id) {
 896                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 897                         version_id, idstr, se->version_id);
 898                 ret = -EINVAL;
 899                 goto out;
 900             }
 901
 902             /* Add entry */
 903             le = qemu_mallocz(sizeof(*le));
 904
 905             le->se = se;
 906             le->section_id = section_id;
 907             le->version_id = version_id;
 908             le->next = first_le;
 909             first_le = le;
 910
 911             le->se->load_state(f, le->se->opaque, le->version_id);
 912             break;
 913         case QEMU_VM_SECTION_PART:
 914         case QEMU_VM_SECTION_END:
 915             section_id = qemu_get_be32(f);
 916
 917             for (le = first_le; le && le->section_id != section_id; le = le->next);
 918             if (le == NULL) {
 919                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 920                 ret = -EINVAL;
 921                 goto out;
 922             }
 923
 924             le->se->load_state(f, le->se->opaque, le->version_id);
 925             break;
 926         default:
 927             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 928             ret = -EINVAL;
 929             goto out;
 930         }
 931     }
 932
 933     ret = 0;
 934
 935 out:
 936     while (first_le) {
 937         LoadStateEntry *le = first_le;
 938         first_le = first_le->next;
 939         qemu_free(le);
 940     }
 941
 942     if (qemu_file_has_error(f))
 943         ret = -EIO;
 944
 945     return ret;
 946 }
 947
 948 /* device can contain snapshots */
 949 static int bdrv_can_snapshot(BlockDriverState *bs)
 950 {
 951     return (bs &&
 952             !bdrv_is_removable(bs) &&
 953             !bdrv_is_read_only(bs));
 954 }
 955
 956 /* device must be snapshots in order to have a reliable snapshot */
 957 static int bdrv_has_snapshot(BlockDriverState *bs)
 958 {
 959     return (bs &&
 960             !bdrv_is_removable(bs) &&
 961             !bdrv_is_read_only(bs));
 962 }
 963
 964 static BlockDriverState *get_bs_snapshots(void)
 965 {
 966     BlockDriverState *bs;
 967     int i;
 968
 969     if (bs_snapshots)
 970         return bs_snapshots;
 971     for(i = 0; i <= nb_drives; i++) {
 972         bs = drives_table[i].bdrv;
 973         if (bdrv_can_snapshot(bs))
 974             goto ok;
 975     }
 976     return NULL;
 977  ok:
 978     bs_snapshots = bs;
 979     return bs;
 980 }
 981
 982 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
 983                               const char *name)
 984 {
 985     QEMUSnapshotInfo *sn_tab, *sn;
 986     int nb_sns, i, ret;
 987
 988     ret = -ENOENT;
 989     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
 990     if (nb_sns < 0)
 991         return ret;
 992     for(i = 0; i < nb_sns; i++) {
 993         sn = &sn_tab[i];
 994         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
 995             *sn_info = *sn;
 996             ret = 0;
 997             break;
 998         }
 999     }
1000     qemu_free(sn_tab);
1001     return ret;
1002 }
1003
1004 void do_savevm(const char *name)
1005 {
1006     BlockDriverState *bs, *bs1;
1007     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1008     int must_delete, ret, i;
1009     BlockDriverInfo bdi1, *bdi = &bdi1;
1010     QEMUFile *f;
1011     int saved_vm_running;
1012     uint32_t vm_state_size;
1013 #ifdef _WIN32
1014     struct _timeb tb;
1015 #else
1016     struct timeval tv;
1017 #endif
1018
1019     bs = get_bs_snapshots();
1020     if (!bs) {
1021         term_printf("No block device can accept snapshots\n");
1022         return;
1023     }
1024
1025     /* ??? Should this occur after vm_stop?  */
1026     qemu_aio_flush();
1027
1028     saved_vm_running = vm_running;
1029     vm_stop(0);
1030
1031     must_delete = 0;
1032     if (name) {
1033         ret = bdrv_snapshot_find(bs, old_sn, name);
1034         if (ret >= 0) {
1035             must_delete = 1;
1036         }
1037     }
1038     memset(sn, 0, sizeof(*sn));
1039     if (must_delete) {
1040         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1041         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1042     } else {
1043         if (name)
1044             pstrcpy(sn->name, sizeof(sn->name), name);
1045     }
1046
1047     /* fill auxiliary fields */
1048 #ifdef _WIN32
1049     _ftime(&tb);
1050     sn->date_sec = tb.time;
1051     sn->date_nsec = tb.millitm * 1000000;
1052 #else
1053     gettimeofday(&tv, NULL);
1054     sn->date_sec = tv.tv_sec;
1055     sn->date_nsec = tv.tv_usec * 1000;
1056 #endif
1057     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1058
1059     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1060         term_printf("Device %s does not support VM state snapshots\n",
1061                     bdrv_get_device_name(bs));
1062         goto the_end;
1063     }
1064
1065     /* save the VM state */
1066     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 1);
1067     if (!f) {
1068         term_printf("Could not open VM state file\n");
1069         goto the_end;
1070     }
1071     ret = qemu_savevm_state(f);
1072     vm_state_size = qemu_ftell(f);
1073     qemu_fclose(f);
1074     if (ret < 0) {
1075         term_printf("Error %d while writing VM\n", ret);
1076         goto the_end;
1077     }
1078
1079     /* create the snapshots */
1080
1081     for(i = 0; i < nb_drives; i++) {
1082         bs1 = drives_table[i].bdrv;
1083         if (bdrv_has_snapshot(bs1)) {
1084             if (must_delete) {
1085                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1086                 if (ret < 0) {
1087                     term_printf("Error while deleting snapshot on '%s'\n",
1088                                 bdrv_get_device_name(bs1));
1089                 }
1090             }
1091             /* Write VM state size only to the image that contains the state */
1092             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1093             ret = bdrv_snapshot_create(bs1, sn);
1094             if (ret < 0) {
1095                 term_printf("Error while creating snapshot on '%s'\n",
1096                             bdrv_get_device_name(bs1));
1097             }
1098         }
1099     }
1100
1101  the_end:
1102     if (saved_vm_running)
1103         vm_start();
1104 }
1105
1106 void do_loadvm(const char *name)
1107 {
1108     BlockDriverState *bs, *bs1;
1109     BlockDriverInfo bdi1, *bdi = &bdi1;
1110     QEMUSnapshotInfo sn;
1111     QEMUFile *f;
1112     int i, ret;
1113     int saved_vm_running;
1114
1115     bs = get_bs_snapshots();
1116     if (!bs) {
1117         term_printf("No block device supports snapshots\n");
1118         return;
1119     }
1120
1121     /* Flush all IO requests so they don't interfere with the new state.  */
1122     qemu_aio_flush();
1123
1124     saved_vm_running = vm_running;
1125     vm_stop(0);
1126
1127     for(i = 0; i <= nb_drives; i++) {
1128         bs1 = drives_table[i].bdrv;
1129         if (bdrv_has_snapshot(bs1)) {
1130             ret = bdrv_snapshot_goto(bs1, name);
1131             if (ret < 0) {
1132                 if (bs != bs1)
1133                     term_printf("Warning: ");
1134                 switch(ret) {
1135                 case -ENOTSUP:
1136                     term_printf("Snapshots not supported on device '%s'\n",
1137                                 bdrv_get_device_name(bs1));
1138                     break;
1139                 case -ENOENT:
1140                     term_printf("Could not find snapshot '%s' on device '%s'\n",
1141                                 name, bdrv_get_device_name(bs1));
1142                     break;
1143                 default:
1144                     term_printf("Error %d while activating snapshot on '%s'\n",
1145                                 ret, bdrv_get_device_name(bs1));
1146                     break;
1147                 }
1148                 /* fatal on snapshot block device */
1149                 if (bs == bs1)
1150                     goto the_end;
1151             }
1152         }
1153     }
1154
1155     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1156         term_printf("Device %s does not support VM state snapshots\n",
1157                     bdrv_get_device_name(bs));
1158         return;
1159     }
1160
1161     /* Don't even try to load empty VM states */
1162     ret = bdrv_snapshot_find(bs, &sn, name);
1163     if ((ret >= 0) && (sn.vm_state_size == 0))
1164         goto the_end;
1165
1166     /* restore the VM state */
1167     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 0);
1168     if (!f) {
1169         term_printf("Could not open VM state file\n");
1170         goto the_end;
1171     }
1172     ret = qemu_loadvm_state(f);
1173     qemu_fclose(f);
1174     if (ret < 0) {
1175         term_printf("Error %d while loading VM state\n", ret);
1176     }
1177  the_end:
1178     if (saved_vm_running)
1179         vm_start();
1180 }
1181
1182 void do_delvm(const char *name)
1183 {
1184     BlockDriverState *bs, *bs1;
1185     int i, ret;
1186
1187     bs = get_bs_snapshots();
1188     if (!bs) {
1189         term_printf("No block device supports snapshots\n");
1190         return;
1191     }
1192
1193     for(i = 0; i <= nb_drives; i++) {
1194         bs1 = drives_table[i].bdrv;
1195         if (bdrv_has_snapshot(bs1)) {
1196             ret = bdrv_snapshot_delete(bs1, name);
1197             if (ret < 0) {
1198                 if (ret == -ENOTSUP)
1199                     term_printf("Snapshots not supported on device '%s'\n",
1200                                 bdrv_get_device_name(bs1));
1201                 else
1202                     term_printf("Error %d while deleting snapshot on '%s'\n",
1203                                 ret, bdrv_get_device_name(bs1));
1204             }
1205         }
1206     }
1207 }
1208
1209 void do_info_snapshots(void)
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         term_printf("No available block device supports snapshots\n");
1219         return;
1220     }
1221     term_printf("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                 term_printf(" %s", bdrv_get_device_name(bs1));
1227         }
1228     }
1229     term_printf("\n");
1230
1231     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1232     if (nb_sns < 0) {
1233         term_printf("bdrv_snapshot_list: error %d\n", nb_sns);
1234         return;
1235     }
1236     term_printf("Snapshot list (from %s):\n", bdrv_get_device_name(bs));
1237     term_printf("%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1238     for(i = 0; i < nb_sns; i++) {
1239         sn = &sn_tab[i];
1240         term_printf("%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1241     }
1242     qemu_free(sn_tab);
1243 }