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 < MAX_NICS; i++) {
 126         if (!nd_table[i].used)
 127             continue;
 128         len = announce_self_create(buf, nd_table[i].macaddr);
 129         vlan = nd_table[i].vlan;
 130         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 131             for (j=0; j < SELF_ANNOUNCE_ROUNDS; j++)
 132                 vc->fd_read(vc->opaque, buf, len);
 133         }
 134     }
 135 }
 136
 137 /***********************************************************/
 138 /* savevm/loadvm support */
 139
 140 #define IO_BUF_SIZE 32768
 141
 142 struct QEMUFile {
 143     QEMUFilePutBufferFunc *put_buffer;
 144     QEMUFileGetBufferFunc *get_buffer;
 145     QEMUFileCloseFunc *close;
 146     QEMUFileRateLimit *rate_limit;
 147     void *opaque;
 148     int is_write;
 149
 150     int64_t buf_offset; /* start of buffer when writing, end of buffer
 151                            when reading */
 152     int buf_index;
 153     int buf_size; /* 0 when writing */
 154     uint8_t buf[IO_BUF_SIZE];
 155
 156     int has_error;
 157 };
 158
 159 typedef struct QEMUFilePopen
 160 {
 161     FILE *popen_file;
 162     QEMUFile *file;
 163 } QEMUFilePopen;
 164
 165 typedef struct QEMUFileSocket
 166 {
 167     int fd;
 168     QEMUFile *file;
 169 } QEMUFileSocket;
 170
 171 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 172 {
 173     QEMUFileSocket *s = opaque;
 174     ssize_t len;
 175
 176     do {
 177         len = recv(s->fd, buf, size, 0);
 178     } while (len == -1 && socket_error() == EINTR);
 179
 180     if (len == -1)
 181         len = -socket_error();
 182
 183     return len;
 184 }
 185
 186 static int socket_close(void *opaque)
 187 {
 188     QEMUFileSocket *s = opaque;
 189     qemu_free(s);
 190     return 0;
 191 }
 192
 193 static int popen_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 194 {
 195     QEMUFilePopen *s = opaque;
 196     return fwrite(buf, 1, size, s->popen_file);
 197 }
 198
 199 static int popen_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 200 {
 201     QEMUFilePopen *s = opaque;
 202     return fread(buf, 1, size, s->popen_file);
 203 }
 204
 205 static int popen_close(void *opaque)
 206 {
 207     QEMUFilePopen *s = opaque;
 208     pclose(s->popen_file);
 209     qemu_free(s);
 210     return 0;
 211 }
 212
 213 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 214 {
 215     QEMUFilePopen *s;
 216
 217     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 218         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 219         return NULL;
 220     }
 221
 222     s = qemu_mallocz(sizeof(QEMUFilePopen));
 223
 224     s->popen_file = popen_file;
 225
 226     if(mode[0] == 'r') {
 227         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL);
 228     } else {
 229         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL);
 230     }
 231     fprintf(stderr, "qemu_popen: returning result of qemu_fopen_ops\n");
 232     return s->file;
 233 }
 234
 235 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 236 {
 237     FILE *popen_file;
 238
 239     popen_file = popen(command, mode);
 240     if(popen_file == NULL) {
 241         return NULL;
 242     }
 243
 244     return qemu_popen(popen_file, mode);
 245 }
 246
 247 QEMUFile *qemu_fopen_socket(int fd)
 248 {
 249     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 250
 251     s->fd = fd;
 252     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL);
 253     return s->file;
 254 }
 255
 256 typedef struct QEMUFileStdio
 257 {
 258     FILE *outfile;
 259 } QEMUFileStdio;
 260
 261 static int file_put_buffer(void *opaque, const uint8_t *buf,
 262                             int64_t pos, int size)
 263 {
 264     QEMUFileStdio *s = opaque;
 265     fseek(s->outfile, pos, SEEK_SET);
 266     fwrite(buf, 1, size, s->outfile);
 267     return size;
 268 }
 269
 270 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 271 {
 272     QEMUFileStdio *s = opaque;
 273     fseek(s->outfile, pos, SEEK_SET);
 274     return fread(buf, 1, size, s->outfile);
 275 }
 276
 277 static int file_close(void *opaque)
 278 {
 279     QEMUFileStdio *s = opaque;
 280     fclose(s->outfile);
 281     qemu_free(s);
 282     return 0;
 283 }
 284
 285 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 286 {
 287     QEMUFileStdio *s;
 288
 289     s = qemu_mallocz(sizeof(QEMUFileStdio));
 290
 291     s->outfile = fopen(filename, mode);
 292     if (!s->outfile)
 293         goto fail;
 294
 295     if (!strcmp(mode, "wb"))
 296         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL);
 297     else if (!strcmp(mode, "rb"))
 298         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL);
 299
 300 fail:
 301     if (s->outfile)
 302         fclose(s->outfile);
 303     qemu_free(s);
 304     return NULL;
 305 }
 306
 307 typedef struct QEMUFileBdrv
 308 {
 309     BlockDriverState *bs;
 310     int64_t base_offset;
 311 } QEMUFileBdrv;
 312
 313 static int block_put_buffer(void *opaque, const uint8_t *buf,
 314                            int64_t pos, int size)
 315 {
 316     QEMUFileBdrv *s = opaque;
 317     bdrv_put_buffer(s->bs, buf, s->base_offset + pos, size);
 318     return size;
 319 }
 320
 321 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 322 {
 323     QEMUFileBdrv *s = opaque;
 324     return bdrv_get_buffer(s->bs, buf, s->base_offset + pos, size);
 325 }
 326
 327 static int bdrv_fclose(void *opaque)
 328 {
 329     QEMUFileBdrv *s = opaque;
 330     qemu_free(s);
 331     return 0;
 332 }
 333
 334 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable)
 335 {
 336     QEMUFileBdrv *s;
 337
 338     s = qemu_mallocz(sizeof(QEMUFileBdrv));
 339
 340     s->bs = bs;
 341     s->base_offset = offset;
 342
 343     if (is_writable)
 344         return qemu_fopen_ops(s, block_put_buffer, NULL, bdrv_fclose, NULL);
 345
 346     return qemu_fopen_ops(s, NULL, block_get_buffer, bdrv_fclose, NULL);
 347 }
 348
 349 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 350                          QEMUFileGetBufferFunc *get_buffer,
 351                          QEMUFileCloseFunc *close,
 352                          QEMUFileRateLimit *rate_limit)
 353 {
 354     QEMUFile *f;
 355
 356     f = qemu_mallocz(sizeof(QEMUFile));
 357
 358     f->opaque = opaque;
 359     f->put_buffer = put_buffer;
 360     f->get_buffer = get_buffer;
 361     f->close = close;
 362     f->rate_limit = rate_limit;
 363     f->is_write = 0;
 364
 365     return f;
 366 }
 367
 368 int qemu_file_has_error(QEMUFile *f)
 369 {
 370     return f->has_error;
 371 }
 372
 373 void qemu_file_set_error(QEMUFile *f)
 374 {
 375     f->has_error = 1;
 376 }
 377
 378 void qemu_fflush(QEMUFile *f)
 379 {
 380     if (!f->put_buffer)
 381         return;
 382
 383     if (f->is_write && f->buf_index > 0) {
 384         int len;
 385
 386         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 387         if (len > 0)
 388             f->buf_offset += f->buf_index;
 389         else
 390             f->has_error = 1;
 391         f->buf_index = 0;
 392     }
 393 }
 394
 395 static void qemu_fill_buffer(QEMUFile *f)
 396 {
 397     int len;
 398
 399     if (!f->get_buffer)
 400         return;
 401
 402     if (f->is_write)
 403         abort();
 404
 405     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 406     if (len > 0) {
 407         f->buf_index = 0;
 408         f->buf_size = len;
 409         f->buf_offset += len;
 410     } else if (len != -EAGAIN)
 411         f->has_error = 1;
 412 }
 413
 414 int qemu_fclose(QEMUFile *f)
 415 {
 416     int ret = 0;
 417     qemu_fflush(f);
 418     if (f->close)
 419         ret = f->close(f->opaque);
 420     qemu_free(f);
 421     return ret;
 422 }
 423
 424 void qemu_file_put_notify(QEMUFile *f)
 425 {
 426     f->put_buffer(f->opaque, NULL, 0, 0);
 427 }
 428
 429 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 430 {
 431     int l;
 432
 433     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 434         fprintf(stderr,
 435                 "Attempted to write to buffer while read buffer is not empty\n");
 436         abort();
 437     }
 438
 439     while (!f->has_error && size > 0) {
 440         l = IO_BUF_SIZE - f->buf_index;
 441         if (l > size)
 442             l = size;
 443         memcpy(f->buf + f->buf_index, buf, l);
 444         f->is_write = 1;
 445         f->buf_index += l;
 446         buf += l;
 447         size -= l;
 448         if (f->buf_index >= IO_BUF_SIZE)
 449             qemu_fflush(f);
 450     }
 451 }
 452
 453 void qemu_put_byte(QEMUFile *f, int v)
 454 {
 455     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 456         fprintf(stderr,
 457                 "Attempted to write to buffer while read buffer is not empty\n");
 458         abort();
 459     }
 460
 461     f->buf[f->buf_index++] = v;
 462     f->is_write = 1;
 463     if (f->buf_index >= IO_BUF_SIZE)
 464         qemu_fflush(f);
 465 }
 466
 467 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 468 {
 469     int size, l;
 470
 471     if (f->is_write)
 472         abort();
 473
 474     size = size1;
 475     while (size > 0) {
 476         l = f->buf_size - f->buf_index;
 477         if (l == 0) {
 478             qemu_fill_buffer(f);
 479             l = f->buf_size - f->buf_index;
 480             if (l == 0)
 481                 break;
 482         }
 483         if (l > size)
 484             l = size;
 485         memcpy(buf, f->buf + f->buf_index, l);
 486         f->buf_index += l;
 487         buf += l;
 488         size -= l;
 489     }
 490     return size1 - size;
 491 }
 492
 493 int qemu_get_byte(QEMUFile *f)
 494 {
 495     if (f->is_write)
 496         abort();
 497
 498     if (f->buf_index >= f->buf_size) {
 499         qemu_fill_buffer(f);
 500         if (f->buf_index >= f->buf_size)
 501             return 0;
 502     }
 503     return f->buf[f->buf_index++];
 504 }
 505
 506 int64_t qemu_ftell(QEMUFile *f)
 507 {
 508     return f->buf_offset - f->buf_size + f->buf_index;
 509 }
 510
 511 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 512 {
 513     if (whence == SEEK_SET) {
 514         /* nothing to do */
 515     } else if (whence == SEEK_CUR) {
 516         pos += qemu_ftell(f);
 517     } else {
 518         /* SEEK_END not supported */
 519         return -1;
 520     }
 521     if (f->put_buffer) {
 522         qemu_fflush(f);
 523         f->buf_offset = pos;
 524     } else {
 525         f->buf_offset = pos;
 526         f->buf_index = 0;
 527         f->buf_size = 0;
 528     }
 529     return pos;
 530 }
 531
 532 int qemu_file_rate_limit(QEMUFile *f)
 533 {
 534     if (f->rate_limit)
 535         return f->rate_limit(f->opaque);
 536
 537     return 0;
 538 }
 539
 540 void qemu_put_be16(QEMUFile *f, unsigned int v)
 541 {
 542     qemu_put_byte(f, v >> 8);
 543     qemu_put_byte(f, v);
 544 }
 545
 546 void qemu_put_be32(QEMUFile *f, unsigned int v)
 547 {
 548     qemu_put_byte(f, v >> 24);
 549     qemu_put_byte(f, v >> 16);
 550     qemu_put_byte(f, v >> 8);
 551     qemu_put_byte(f, v);
 552 }
 553
 554 void qemu_put_be64(QEMUFile *f, uint64_t v)
 555 {
 556     qemu_put_be32(f, v >> 32);
 557     qemu_put_be32(f, v);
 558 }
 559
 560 unsigned int qemu_get_be16(QEMUFile *f)
 561 {
 562     unsigned int v;
 563     v = qemu_get_byte(f) << 8;
 564     v |= qemu_get_byte(f);
 565     return v;
 566 }
 567
 568 unsigned int qemu_get_be32(QEMUFile *f)
 569 {
 570     unsigned int v;
 571     v = qemu_get_byte(f) << 24;
 572     v |= qemu_get_byte(f) << 16;
 573     v |= qemu_get_byte(f) << 8;
 574     v |= qemu_get_byte(f);
 575     return v;
 576 }
 577
 578 uint64_t qemu_get_be64(QEMUFile *f)
 579 {
 580     uint64_t v;
 581     v = (uint64_t)qemu_get_be32(f) << 32;
 582     v |= qemu_get_be32(f);
 583     return v;
 584 }
 585
 586 typedef struct SaveStateEntry {
 587     char idstr[256];
 588     int instance_id;
 589     int version_id;
 590     int section_id;
 591     SaveLiveStateHandler *save_live_state;
 592     SaveStateHandler *save_state;
 593     LoadStateHandler *load_state;
 594     void *opaque;
 595     struct SaveStateEntry *next;
 596 } SaveStateEntry;
 597
 598 static SaveStateEntry *first_se;
 599
 600 /* TODO: Individual devices generally have very little idea about the rest
 601    of the system, so instance_id should be removed/replaced.
 602    Meanwhile pass -1 as instance_id if you do not already have a clearly
 603    distinguishing id for all instances of your device class. */
 604 int register_savevm_live(const char *idstr,
 605                          int instance_id,
 606                          int version_id,
 607                          SaveLiveStateHandler *save_live_state,
 608                          SaveStateHandler *save_state,
 609                          LoadStateHandler *load_state,
 610                          void *opaque)
 611 {
 612     SaveStateEntry *se, **pse;
 613     static int global_section_id;
 614
 615     se = qemu_malloc(sizeof(SaveStateEntry));
 616     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 617     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 618     se->version_id = version_id;
 619     se->section_id = global_section_id++;
 620     se->save_live_state = save_live_state;
 621     se->save_state = save_state;
 622     se->load_state = load_state;
 623     se->opaque = opaque;
 624     se->next = NULL;
 625
 626     /* add at the end of list */
 627     pse = &first_se;
 628     while (*pse != NULL) {
 629         if (instance_id == -1
 630                 && strcmp(se->idstr, (*pse)->idstr) == 0
 631                 && se->instance_id <= (*pse)->instance_id)
 632             se->instance_id = (*pse)->instance_id + 1;
 633         pse = &(*pse)->next;
 634     }
 635     *pse = se;
 636     return 0;
 637 }
 638
 639 int register_savevm(const char *idstr,
 640                     int instance_id,
 641                     int version_id,
 642                     SaveStateHandler *save_state,
 643                     LoadStateHandler *load_state,
 644                     void *opaque)
 645 {
 646     return register_savevm_live(idstr, instance_id, version_id,
 647                                 NULL, save_state, load_state, opaque);
 648 }
 649
 650 #define QEMU_VM_FILE_MAGIC           0x5145564d
 651 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 652 #define QEMU_VM_FILE_VERSION         0x00000003
 653
 654 #define QEMU_VM_EOF                  0x00
 655 #define QEMU_VM_SECTION_START        0x01
 656 #define QEMU_VM_SECTION_PART         0x02
 657 #define QEMU_VM_SECTION_END          0x03
 658 #define QEMU_VM_SECTION_FULL         0x04
 659
 660 int qemu_savevm_state_begin(QEMUFile *f)
 661 {
 662     SaveStateEntry *se;
 663
 664     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 665     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 666
 667     for (se = first_se; se != NULL; se = se->next) {
 668         int len;
 669
 670         if (se->save_live_state == NULL)
 671             continue;
 672
 673         /* Section type */
 674         qemu_put_byte(f, QEMU_VM_SECTION_START);
 675         qemu_put_be32(f, se->section_id);
 676
 677         /* ID string */
 678         len = strlen(se->idstr);
 679         qemu_put_byte(f, len);
 680         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 681
 682         qemu_put_be32(f, se->instance_id);
 683         qemu_put_be32(f, se->version_id);
 684
 685         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 686     }
 687
 688     if (qemu_file_has_error(f))
 689         return -EIO;
 690
 691     return 0;
 692 }
 693
 694 int qemu_savevm_state_iterate(QEMUFile *f)
 695 {
 696     SaveStateEntry *se;
 697     int ret = 1;
 698
 699     for (se = first_se; se != NULL; se = se->next) {
 700         if (se->save_live_state == NULL)
 701             continue;
 702
 703         /* Section type */
 704         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 705         qemu_put_be32(f, se->section_id);
 706
 707         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 708     }
 709
 710     if (ret)
 711         return 1;
 712
 713     if (qemu_file_has_error(f))
 714         return -EIO;
 715
 716     return 0;
 717 }
 718
 719 int qemu_savevm_state_complete(QEMUFile *f)
 720 {
 721     SaveStateEntry *se;
 722
 723     for (se = first_se; se != NULL; se = se->next) {
 724         if (se->save_live_state == NULL)
 725             continue;
 726
 727         /* Section type */
 728         qemu_put_byte(f, QEMU_VM_SECTION_END);
 729         qemu_put_be32(f, se->section_id);
 730
 731         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 732     }
 733
 734     for(se = first_se; se != NULL; se = se->next) {
 735         int len;
 736
 737         if (se->save_state == NULL)
 738             continue;
 739
 740         /* Section type */
 741         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 742         qemu_put_be32(f, se->section_id);
 743
 744         /* ID string */
 745         len = strlen(se->idstr);
 746         qemu_put_byte(f, len);
 747         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 748
 749         qemu_put_be32(f, se->instance_id);
 750         qemu_put_be32(f, se->version_id);
 751
 752         se->save_state(f, se->opaque);
 753     }
 754
 755     qemu_put_byte(f, QEMU_VM_EOF);
 756
 757     if (qemu_file_has_error(f))
 758         return -EIO;
 759
 760     return 0;
 761 }
 762
 763 int qemu_savevm_state(QEMUFile *f)
 764 {
 765     int saved_vm_running;
 766     int ret;
 767
 768     saved_vm_running = vm_running;
 769     vm_stop(0);
 770
 771     bdrv_flush_all();
 772
 773     ret = qemu_savevm_state_begin(f);
 774     if (ret < 0)
 775         goto out;
 776
 777     do {
 778         ret = qemu_savevm_state_iterate(f);
 779         if (ret < 0)
 780             goto out;
 781     } while (ret == 0);
 782
 783     ret = qemu_savevm_state_complete(f);
 784
 785 out:
 786     if (qemu_file_has_error(f))
 787         ret = -EIO;
 788
 789     if (!ret && saved_vm_running)
 790         vm_start();
 791
 792     return ret;
 793 }
 794
 795 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 796 {
 797     SaveStateEntry *se;
 798
 799     for(se = first_se; se != NULL; se = se->next) {
 800         if (!strcmp(se->idstr, idstr) &&
 801             instance_id == se->instance_id)
 802             return se;
 803     }
 804     return NULL;
 805 }
 806
 807 typedef struct LoadStateEntry {
 808     SaveStateEntry *se;
 809     int section_id;
 810     int version_id;
 811     struct LoadStateEntry *next;
 812 } LoadStateEntry;
 813
 814 static int qemu_loadvm_state_v2(QEMUFile *f)
 815 {
 816     SaveStateEntry *se;
 817     int len, ret, instance_id, record_len, version_id;
 818     int64_t total_len, end_pos, cur_pos;
 819     char idstr[256];
 820
 821     total_len = qemu_get_be64(f);
 822     end_pos = total_len + qemu_ftell(f);
 823     for(;;) {
 824         if (qemu_ftell(f) >= end_pos)
 825             break;
 826         len = qemu_get_byte(f);
 827         qemu_get_buffer(f, (uint8_t *)idstr, len);
 828         idstr[len] = '\0';
 829         instance_id = qemu_get_be32(f);
 830         version_id = qemu_get_be32(f);
 831         record_len = qemu_get_be32(f);
 832         cur_pos = qemu_ftell(f);
 833         se = find_se(idstr, instance_id);
 834         if (!se) {
 835             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 836                     instance_id, idstr);
 837         } else {
 838             ret = se->load_state(f, se->opaque, version_id);
 839             if (ret < 0) {
 840                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 841                         instance_id, idstr);
 842                 return ret;
 843             }
 844         }
 845         /* always seek to exact end of record */
 846         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 847     }
 848
 849     if (qemu_file_has_error(f))
 850         return -EIO;
 851
 852     return 0;
 853 }
 854
 855 int qemu_loadvm_state(QEMUFile *f)
 856 {
 857     LoadStateEntry *first_le = NULL;
 858     uint8_t section_type;
 859     unsigned int v;
 860     int ret;
 861
 862     v = qemu_get_be32(f);
 863     if (v != QEMU_VM_FILE_MAGIC)
 864         return -EINVAL;
 865
 866     v = qemu_get_be32(f);
 867     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 868         return qemu_loadvm_state_v2(f);
 869     if (v != QEMU_VM_FILE_VERSION)
 870         return -ENOTSUP;
 871
 872     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 873         uint32_t instance_id, version_id, section_id;
 874         LoadStateEntry *le;
 875         SaveStateEntry *se;
 876         char idstr[257];
 877         int len;
 878
 879         switch (section_type) {
 880         case QEMU_VM_SECTION_START:
 881         case QEMU_VM_SECTION_FULL:
 882             /* Read section start */
 883             section_id = qemu_get_be32(f);
 884             len = qemu_get_byte(f);
 885             qemu_get_buffer(f, (uint8_t *)idstr, len);
 886             idstr[len] = 0;
 887             instance_id = qemu_get_be32(f);
 888             version_id = qemu_get_be32(f);
 889
 890             /* Find savevm section */
 891             se = find_se(idstr, instance_id);
 892             if (se == NULL) {
 893                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 894                 ret = -EINVAL;
 895                 goto out;
 896             }
 897
 898             /* Validate version */
 899             if (version_id > se->version_id) {
 900                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 901                         version_id, idstr, se->version_id);
 902                 ret = -EINVAL;
 903                 goto out;
 904             }
 905
 906             /* Add entry */
 907             le = qemu_mallocz(sizeof(*le));
 908
 909             le->se = se;
 910             le->section_id = section_id;
 911             le->version_id = version_id;
 912             le->next = first_le;
 913             first_le = le;
 914
 915             le->se->load_state(f, le->se->opaque, le->version_id);
 916             break;
 917         case QEMU_VM_SECTION_PART:
 918         case QEMU_VM_SECTION_END:
 919             section_id = qemu_get_be32(f);
 920
 921             for (le = first_le; le && le->section_id != section_id; le = le->next);
 922             if (le == NULL) {
 923                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 924                 ret = -EINVAL;
 925                 goto out;
 926             }
 927
 928             le->se->load_state(f, le->se->opaque, le->version_id);
 929             break;
 930         default:
 931             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 932             ret = -EINVAL;
 933             goto out;
 934         }
 935     }
 936
 937     ret = 0;
 938
 939 out:
 940     while (first_le) {
 941         LoadStateEntry *le = first_le;
 942         first_le = first_le->next;
 943         qemu_free(le);
 944     }
 945
 946     if (qemu_file_has_error(f))
 947         ret = -EIO;
 948
 949     return ret;
 950 }
 951
 952 /* device can contain snapshots */
 953 static int bdrv_can_snapshot(BlockDriverState *bs)
 954 {
 955     return (bs &&
 956             !bdrv_is_removable(bs) &&
 957             !bdrv_is_read_only(bs));
 958 }
 959
 960 /* device must be snapshots in order to have a reliable snapshot */
 961 static int bdrv_has_snapshot(BlockDriverState *bs)
 962 {
 963     return (bs &&
 964             !bdrv_is_removable(bs) &&
 965             !bdrv_is_read_only(bs));
 966 }
 967
 968 static BlockDriverState *get_bs_snapshots(void)
 969 {
 970     BlockDriverState *bs;
 971     int i;
 972
 973     if (bs_snapshots)
 974         return bs_snapshots;
 975     for(i = 0; i <= nb_drives; i++) {
 976         bs = drives_table[i].bdrv;
 977         if (bdrv_can_snapshot(bs))
 978             goto ok;
 979     }
 980     return NULL;
 981  ok:
 982     bs_snapshots = bs;
 983     return bs;
 984 }
 985
 986 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
 987                               const char *name)
 988 {
 989     QEMUSnapshotInfo *sn_tab, *sn;
 990     int nb_sns, i, ret;
 991
 992     ret = -ENOENT;
 993     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
 994     if (nb_sns < 0)
 995         return ret;
 996     for(i = 0; i < nb_sns; i++) {
 997         sn = &sn_tab[i];
 998         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
 999             *sn_info = *sn;
1000             ret = 0;
1001             break;
1002         }
1003     }
1004     qemu_free(sn_tab);
1005     return ret;
1006 }
1007
1008 void do_savevm(Monitor *mon, const char *name)
1009 {
1010     BlockDriverState *bs, *bs1;
1011     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1012     int must_delete, ret, i;
1013     BlockDriverInfo bdi1, *bdi = &bdi1;
1014     QEMUFile *f;
1015     int saved_vm_running;
1016     uint32_t vm_state_size;
1017 #ifdef _WIN32
1018     struct _timeb tb;
1019 #else
1020     struct timeval tv;
1021 #endif
1022
1023     bs = get_bs_snapshots();
1024     if (!bs) {
1025         monitor_printf(mon, "No block device can accept snapshots\n");
1026         return;
1027     }
1028
1029     /* ??? Should this occur after vm_stop?  */
1030     qemu_aio_flush();
1031
1032     saved_vm_running = vm_running;
1033     vm_stop(0);
1034
1035     must_delete = 0;
1036     if (name) {
1037         ret = bdrv_snapshot_find(bs, old_sn, name);
1038         if (ret >= 0) {
1039             must_delete = 1;
1040         }
1041     }
1042     memset(sn, 0, sizeof(*sn));
1043     if (must_delete) {
1044         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1045         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1046     } else {
1047         if (name)
1048             pstrcpy(sn->name, sizeof(sn->name), name);
1049     }
1050
1051     /* fill auxiliary fields */
1052 #ifdef _WIN32
1053     _ftime(&tb);
1054     sn->date_sec = tb.time;
1055     sn->date_nsec = tb.millitm * 1000000;
1056 #else
1057     gettimeofday(&tv, NULL);
1058     sn->date_sec = tv.tv_sec;
1059     sn->date_nsec = tv.tv_usec * 1000;
1060 #endif
1061     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1062
1063     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1064         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1065                        bdrv_get_device_name(bs));
1066         goto the_end;
1067     }
1068
1069     /* save the VM state */
1070     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 1);
1071     if (!f) {
1072         monitor_printf(mon, "Could not open VM state file\n");
1073         goto the_end;
1074     }
1075     ret = qemu_savevm_state(f);
1076     vm_state_size = qemu_ftell(f);
1077     qemu_fclose(f);
1078     if (ret < 0) {
1079         monitor_printf(mon, "Error %d while writing VM\n", ret);
1080         goto the_end;
1081     }
1082
1083     /* create the snapshots */
1084
1085     for(i = 0; i < nb_drives; i++) {
1086         bs1 = drives_table[i].bdrv;
1087         if (bdrv_has_snapshot(bs1)) {
1088             if (must_delete) {
1089                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1090                 if (ret < 0) {
1091                     monitor_printf(mon,
1092                                    "Error while deleting snapshot on '%s'\n",
1093                                    bdrv_get_device_name(bs1));
1094                 }
1095             }
1096             /* Write VM state size only to the image that contains the state */
1097             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1098             ret = bdrv_snapshot_create(bs1, sn);
1099             if (ret < 0) {
1100                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1101                                bdrv_get_device_name(bs1));
1102             }
1103         }
1104     }
1105
1106  the_end:
1107     if (saved_vm_running)
1108         vm_start();
1109 }
1110
1111 void do_loadvm(Monitor *mon, const char *name)
1112 {
1113     BlockDriverState *bs, *bs1;
1114     BlockDriverInfo bdi1, *bdi = &bdi1;
1115     QEMUSnapshotInfo sn;
1116     QEMUFile *f;
1117     int i, ret;
1118     int saved_vm_running;
1119
1120     bs = get_bs_snapshots();
1121     if (!bs) {
1122         monitor_printf(mon, "No block device supports snapshots\n");
1123         return;
1124     }
1125
1126     /* Flush all IO requests so they don't interfere with the new state.  */
1127     qemu_aio_flush();
1128
1129     saved_vm_running = vm_running;
1130     vm_stop(0);
1131
1132     for(i = 0; i <= nb_drives; i++) {
1133         bs1 = drives_table[i].bdrv;
1134         if (bdrv_has_snapshot(bs1)) {
1135             ret = bdrv_snapshot_goto(bs1, name);
1136             if (ret < 0) {
1137                 if (bs != bs1)
1138                     monitor_printf(mon, "Warning: ");
1139                 switch(ret) {
1140                 case -ENOTSUP:
1141                     monitor_printf(mon,
1142                                    "Snapshots not supported on device '%s'\n",
1143                                    bdrv_get_device_name(bs1));
1144                     break;
1145                 case -ENOENT:
1146                     monitor_printf(mon, "Could not find snapshot '%s' on "
1147                                    "device '%s'\n",
1148                                    name, bdrv_get_device_name(bs1));
1149                     break;
1150                 default:
1151                     monitor_printf(mon, "Error %d while activating snapshot on"
1152                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1153                     break;
1154                 }
1155                 /* fatal on snapshot block device */
1156                 if (bs == bs1)
1157                     goto the_end;
1158             }
1159         }
1160     }
1161
1162     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1163         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1164                        bdrv_get_device_name(bs));
1165         return;
1166     }
1167
1168     /* Don't even try to load empty VM states */
1169     ret = bdrv_snapshot_find(bs, &sn, name);
1170     if ((ret >= 0) && (sn.vm_state_size == 0))
1171         goto the_end;
1172
1173     /* restore the VM state */
1174     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 0);
1175     if (!f) {
1176         monitor_printf(mon, "Could not open VM state file\n");
1177         goto the_end;
1178     }
1179     ret = qemu_loadvm_state(f);
1180     qemu_fclose(f);
1181     if (ret < 0) {
1182         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1183     }
1184  the_end:
1185     if (saved_vm_running)
1186         vm_start();
1187 }
1188
1189 void do_delvm(Monitor *mon, const char *name)
1190 {
1191     BlockDriverState *bs, *bs1;
1192     int i, ret;
1193
1194     bs = get_bs_snapshots();
1195     if (!bs) {
1196         monitor_printf(mon, "No block device supports snapshots\n");
1197         return;
1198     }
1199
1200     for(i = 0; i <= nb_drives; i++) {
1201         bs1 = drives_table[i].bdrv;
1202         if (bdrv_has_snapshot(bs1)) {
1203             ret = bdrv_snapshot_delete(bs1, name);
1204             if (ret < 0) {
1205                 if (ret == -ENOTSUP)
1206                     monitor_printf(mon,
1207                                    "Snapshots not supported on device '%s'\n",
1208                                    bdrv_get_device_name(bs1));
1209                 else
1210                     monitor_printf(mon, "Error %d while deleting snapshot on "
1211                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1212             }
1213         }
1214     }
1215 }
1216
1217 void do_info_snapshots(Monitor *mon)
1218 {
1219     BlockDriverState *bs, *bs1;
1220     QEMUSnapshotInfo *sn_tab, *sn;
1221     int nb_sns, i;
1222     char buf[256];
1223
1224     bs = get_bs_snapshots();
1225     if (!bs) {
1226         monitor_printf(mon, "No available block device supports snapshots\n");
1227         return;
1228     }
1229     monitor_printf(mon, "Snapshot devices:");
1230     for(i = 0; i <= nb_drives; i++) {
1231         bs1 = drives_table[i].bdrv;
1232         if (bdrv_has_snapshot(bs1)) {
1233             if (bs == bs1)
1234                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1235         }
1236     }
1237     monitor_printf(mon, "\n");
1238
1239     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1240     if (nb_sns < 0) {
1241         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1242         return;
1243     }
1244     monitor_printf(mon, "Snapshot list (from %s):\n",
1245                    bdrv_get_device_name(bs));
1246     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1247     for(i = 0; i < nb_sns; i++) {
1248         sn = &sn_tab[i];
1249         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1250     }
1251     qemu_free(sn_tab);
1252 }