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, 0, 64);
 111     memset(buf, 0xff, 6);         /* h_dst */
 112     memcpy(buf + 6, mac_addr, 6); /* h_src */
 113     memcpy(buf + 12, &proto, 2);  /* h_proto */
 114     memcpy(buf + 14, &magic, 4);  /* magic */
 115
 116     return 64; /* len */
 117 }
 118
 119 static void qemu_announce_self_once(void *opaque)
 120 {
 121     int i, len;
 122     VLANState *vlan;
 123     VLANClientState *vc;
 124     uint8_t buf[256];
 125     static int count = SELF_ANNOUNCE_ROUNDS;
 126     QEMUTimer *timer = *(QEMUTimer **)opaque;
 127
 128     for (i = 0; i < MAX_NICS; i++) {
 129         if (!nd_table[i].used)
 130             continue;
 131         len = announce_self_create(buf, nd_table[i].macaddr);
 132         vlan = nd_table[i].vlan;
 133         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 134             vc->receive(vc, buf, len);
 135         }
 136     }
 137     if (count--) {
 138             qemu_mod_timer(timer, qemu_get_clock(rt_clock) + 100);
 139     } else {
 140             qemu_del_timer(timer);
 141             qemu_free_timer(timer);
 142     }
 143 }
 144
 145 void qemu_announce_self(void)
 146 {
 147         static QEMUTimer *timer;
 148         timer = qemu_new_timer(rt_clock, qemu_announce_self_once, &timer);
 149         qemu_announce_self_once(&timer);
 150 }
 151
 152 /***********************************************************/
 153 /* savevm/loadvm support */
 154
 155 #define IO_BUF_SIZE 32768
 156
 157 struct QEMUFile {
 158     QEMUFilePutBufferFunc *put_buffer;
 159     QEMUFileGetBufferFunc *get_buffer;
 160     QEMUFileCloseFunc *close;
 161     QEMUFileRateLimit *rate_limit;
 162     QEMUFileSetRateLimit *set_rate_limit;
 163     void *opaque;
 164     int is_write;
 165
 166     int64_t buf_offset; /* start of buffer when writing, end of buffer
 167                            when reading */
 168     int buf_index;
 169     int buf_size; /* 0 when writing */
 170     uint8_t buf[IO_BUF_SIZE];
 171
 172     int has_error;
 173 };
 174
 175 typedef struct QEMUFilePopen
 176 {
 177     FILE *popen_file;
 178     QEMUFile *file;
 179 } QEMUFilePopen;
 180
 181 typedef struct QEMUFileSocket
 182 {
 183     int fd;
 184     QEMUFile *file;
 185 } QEMUFileSocket;
 186
 187 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 188 {
 189     QEMUFileSocket *s = opaque;
 190     ssize_t len;
 191
 192     do {
 193         len = recv(s->fd, (void *)buf, size, 0);
 194     } while (len == -1 && socket_error() == EINTR);
 195
 196     if (len == -1)
 197         len = -socket_error();
 198
 199     return len;
 200 }
 201
 202 static int socket_close(void *opaque)
 203 {
 204     QEMUFileSocket *s = opaque;
 205     qemu_free(s);
 206     return 0;
 207 }
 208
 209 static int popen_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 210 {
 211     QEMUFilePopen *s = opaque;
 212     return fwrite(buf, 1, size, s->popen_file);
 213 }
 214
 215 static int popen_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 216 {
 217     QEMUFilePopen *s = opaque;
 218     FILE *fp = s->popen_file;
 219     int bytes;
 220
 221     do {
 222         clearerr(fp);
 223         bytes = fread(buf, 1, size, fp);
 224     } while ((bytes == 0) && ferror(fp) && (errno == EINTR));
 225     return bytes;
 226 }
 227
 228 static int popen_close(void *opaque)
 229 {
 230     QEMUFilePopen *s = opaque;
 231     pclose(s->popen_file);
 232     qemu_free(s);
 233     return 0;
 234 }
 235
 236 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 237 {
 238     QEMUFilePopen *s;
 239
 240     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 241         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 242         return NULL;
 243     }
 244
 245     s = qemu_mallocz(sizeof(QEMUFilePopen));
 246
 247     s->popen_file = popen_file;
 248
 249     if(mode[0] == 'r') {
 250         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL, NULL);
 251     } else {
 252         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL, NULL);
 253     }
 254     return s->file;
 255 }
 256
 257 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 258 {
 259     FILE *popen_file;
 260
 261     popen_file = popen(command, mode);
 262     if(popen_file == NULL) {
 263         return NULL;
 264     }
 265
 266     return qemu_popen(popen_file, mode);
 267 }
 268
 269 int qemu_popen_fd(QEMUFile *f)
 270 {
 271     QEMUFilePopen *p;
 272     int fd;
 273
 274     p = (QEMUFilePopen *)f->opaque;
 275     fd = fileno(p->popen_file);
 276
 277     return fd;
 278 }
 279
 280 QEMUFile *qemu_fopen_socket(int fd)
 281 {
 282     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 283
 284     s->fd = fd;
 285     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL, NULL);
 286     return s->file;
 287 }
 288
 289 typedef struct QEMUFileStdio
 290 {
 291     FILE *outfile;
 292 } QEMUFileStdio;
 293
 294 static int file_put_buffer(void *opaque, const uint8_t *buf,
 295                             int64_t pos, int size)
 296 {
 297     QEMUFileStdio *s = opaque;
 298     fseek(s->outfile, pos, SEEK_SET);
 299     fwrite(buf, 1, size, s->outfile);
 300     return size;
 301 }
 302
 303 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 304 {
 305     QEMUFileStdio *s = opaque;
 306     fseek(s->outfile, pos, SEEK_SET);
 307     return fread(buf, 1, size, s->outfile);
 308 }
 309
 310 static int file_close(void *opaque)
 311 {
 312     QEMUFileStdio *s = opaque;
 313     fclose(s->outfile);
 314     qemu_free(s);
 315     return 0;
 316 }
 317
 318 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 319 {
 320     QEMUFileStdio *s;
 321
 322     s = qemu_mallocz(sizeof(QEMUFileStdio));
 323
 324     s->outfile = fopen(filename, mode);
 325     if (!s->outfile)
 326         goto fail;
 327
 328     if (!strcmp(mode, "wb"))
 329         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL, NULL);
 330     else if (!strcmp(mode, "rb"))
 331         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL, NULL);
 332
 333 fail:
 334     if (s->outfile)
 335         fclose(s->outfile);
 336     qemu_free(s);
 337     return NULL;
 338 }
 339
 340 static int block_put_buffer(void *opaque, const uint8_t *buf,
 341                            int64_t pos, int size)
 342 {
 343     bdrv_save_vmstate(opaque, buf, pos, size);
 344     return size;
 345 }
 346
 347 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 348 {
 349     return bdrv_load_vmstate(opaque, buf, pos, size);
 350 }
 351
 352 static int bdrv_fclose(void *opaque)
 353 {
 354     return 0;
 355 }
 356
 357 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 358 {
 359     if (is_writable)
 360         return qemu_fopen_ops(bs, block_put_buffer, NULL, bdrv_fclose, NULL, NULL);
 361     return qemu_fopen_ops(bs, NULL, block_get_buffer, bdrv_fclose, NULL, NULL);
 362 }
 363
 364 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 365                          QEMUFileGetBufferFunc *get_buffer,
 366                          QEMUFileCloseFunc *close,
 367                          QEMUFileRateLimit *rate_limit,
 368                          QEMUFileSetRateLimit *set_rate_limit)
 369 {
 370     QEMUFile *f;
 371
 372     f = qemu_mallocz(sizeof(QEMUFile));
 373
 374     f->opaque = opaque;
 375     f->put_buffer = put_buffer;
 376     f->get_buffer = get_buffer;
 377     f->close = close;
 378     f->rate_limit = rate_limit;
 379     f->set_rate_limit = set_rate_limit;
 380     f->is_write = 0;
 381
 382     return f;
 383 }
 384
 385 int qemu_file_has_error(QEMUFile *f)
 386 {
 387     return f->has_error;
 388 }
 389
 390 void qemu_file_set_error(QEMUFile *f)
 391 {
 392     f->has_error = 1;
 393 }
 394
 395 void qemu_fflush(QEMUFile *f)
 396 {
 397     if (!f->put_buffer)
 398         return;
 399
 400     if (f->is_write && f->buf_index > 0) {
 401         int len;
 402
 403         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 404         if (len > 0)
 405             f->buf_offset += f->buf_index;
 406         else
 407             f->has_error = 1;
 408         f->buf_index = 0;
 409     }
 410 }
 411
 412 static void qemu_fill_buffer(QEMUFile *f)
 413 {
 414     int len;
 415
 416     if (!f->get_buffer)
 417         return;
 418
 419     if (f->is_write)
 420         abort();
 421
 422     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 423     if (len > 0) {
 424         f->buf_index = 0;
 425         f->buf_size = len;
 426         f->buf_offset += len;
 427     } else if (len != -EAGAIN)
 428         f->has_error = 1;
 429 }
 430
 431 int qemu_fclose(QEMUFile *f)
 432 {
 433     int ret = 0;
 434     qemu_fflush(f);
 435     if (f->close)
 436         ret = f->close(f->opaque);
 437     qemu_free(f);
 438     return ret;
 439 }
 440
 441 void qemu_file_put_notify(QEMUFile *f)
 442 {
 443     f->put_buffer(f->opaque, NULL, 0, 0);
 444 }
 445
 446 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 447 {
 448     int l;
 449
 450     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 451         fprintf(stderr,
 452                 "Attempted to write to buffer while read buffer is not empty\n");
 453         abort();
 454     }
 455
 456     while (!f->has_error && size > 0) {
 457         l = IO_BUF_SIZE - f->buf_index;
 458         if (l > size)
 459             l = size;
 460         memcpy(f->buf + f->buf_index, buf, l);
 461         f->is_write = 1;
 462         f->buf_index += l;
 463         buf += l;
 464         size -= l;
 465         if (f->buf_index >= IO_BUF_SIZE)
 466             qemu_fflush(f);
 467     }
 468 }
 469
 470 void qemu_put_byte(QEMUFile *f, int v)
 471 {
 472     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 473         fprintf(stderr,
 474                 "Attempted to write to buffer while read buffer is not empty\n");
 475         abort();
 476     }
 477
 478     f->buf[f->buf_index++] = v;
 479     f->is_write = 1;
 480     if (f->buf_index >= IO_BUF_SIZE)
 481         qemu_fflush(f);
 482 }
 483
 484 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 485 {
 486     int size, l;
 487
 488     if (f->is_write)
 489         abort();
 490
 491     size = size1;
 492     while (size > 0) {
 493         l = f->buf_size - f->buf_index;
 494         if (l == 0) {
 495             qemu_fill_buffer(f);
 496             l = f->buf_size - f->buf_index;
 497             if (l == 0)
 498                 break;
 499         }
 500         if (l > size)
 501             l = size;
 502         memcpy(buf, f->buf + f->buf_index, l);
 503         f->buf_index += l;
 504         buf += l;
 505         size -= l;
 506     }
 507     return size1 - size;
 508 }
 509
 510 int qemu_get_byte(QEMUFile *f)
 511 {
 512     if (f->is_write)
 513         abort();
 514
 515     if (f->buf_index >= f->buf_size) {
 516         qemu_fill_buffer(f);
 517         if (f->buf_index >= f->buf_size)
 518             return 0;
 519     }
 520     return f->buf[f->buf_index++];
 521 }
 522
 523 int64_t qemu_ftell(QEMUFile *f)
 524 {
 525     return f->buf_offset - f->buf_size + f->buf_index;
 526 }
 527
 528 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 529 {
 530     if (whence == SEEK_SET) {
 531         /* nothing to do */
 532     } else if (whence == SEEK_CUR) {
 533         pos += qemu_ftell(f);
 534     } else {
 535         /* SEEK_END not supported */
 536         return -1;
 537     }
 538     if (f->put_buffer) {
 539         qemu_fflush(f);
 540         f->buf_offset = pos;
 541     } else {
 542         f->buf_offset = pos;
 543         f->buf_index = 0;
 544         f->buf_size = 0;
 545     }
 546     return pos;
 547 }
 548
 549 int qemu_file_rate_limit(QEMUFile *f)
 550 {
 551     if (f->rate_limit)
 552         return f->rate_limit(f->opaque);
 553
 554     return 0;
 555 }
 556
 557 size_t qemu_file_set_rate_limit(QEMUFile *f, size_t new_rate)
 558 {
 559     if (f->set_rate_limit)
 560         return f->set_rate_limit(f->opaque, new_rate);
 561
 562     return 0;
 563 }
 564
 565 void qemu_put_be16(QEMUFile *f, unsigned int v)
 566 {
 567     qemu_put_byte(f, v >> 8);
 568     qemu_put_byte(f, v);
 569 }
 570
 571 void qemu_put_be32(QEMUFile *f, unsigned int v)
 572 {
 573     qemu_put_byte(f, v >> 24);
 574     qemu_put_byte(f, v >> 16);
 575     qemu_put_byte(f, v >> 8);
 576     qemu_put_byte(f, v);
 577 }
 578
 579 void qemu_put_be64(QEMUFile *f, uint64_t v)
 580 {
 581     qemu_put_be32(f, v >> 32);
 582     qemu_put_be32(f, v);
 583 }
 584
 585 unsigned int qemu_get_be16(QEMUFile *f)
 586 {
 587     unsigned int v;
 588     v = qemu_get_byte(f) << 8;
 589     v |= qemu_get_byte(f);
 590     return v;
 591 }
 592
 593 unsigned int qemu_get_be32(QEMUFile *f)
 594 {
 595     unsigned int v;
 596     v = qemu_get_byte(f) << 24;
 597     v |= qemu_get_byte(f) << 16;
 598     v |= qemu_get_byte(f) << 8;
 599     v |= qemu_get_byte(f);
 600     return v;
 601 }
 602
 603 uint64_t qemu_get_be64(QEMUFile *f)
 604 {
 605     uint64_t v;
 606     v = (uint64_t)qemu_get_be32(f) << 32;
 607     v |= qemu_get_be32(f);
 608     return v;
 609 }
 610
 611 typedef struct SaveStateEntry {
 612     char idstr[256];
 613     int instance_id;
 614     int version_id;
 615     int section_id;
 616     SaveLiveStateHandler *save_live_state;
 617     SaveStateHandler *save_state;
 618     LoadStateHandler *load_state;
 619     void *opaque;
 620     struct SaveStateEntry *next;
 621 } SaveStateEntry;
 622
 623 static SaveStateEntry *first_se;
 624
 625 /* TODO: Individual devices generally have very little idea about the rest
 626    of the system, so instance_id should be removed/replaced.
 627    Meanwhile pass -1 as instance_id if you do not already have a clearly
 628    distinguishing id for all instances of your device class. */
 629 int register_savevm_live(const char *idstr,
 630                          int instance_id,
 631                          int version_id,
 632                          SaveLiveStateHandler *save_live_state,
 633                          SaveStateHandler *save_state,
 634                          LoadStateHandler *load_state,
 635                          void *opaque)
 636 {
 637     SaveStateEntry *se, **pse;
 638     static int global_section_id;
 639
 640     se = qemu_malloc(sizeof(SaveStateEntry));
 641     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 642     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 643     se->version_id = version_id;
 644     se->section_id = global_section_id++;
 645     se->save_live_state = save_live_state;
 646     se->save_state = save_state;
 647     se->load_state = load_state;
 648     se->opaque = opaque;
 649     se->next = NULL;
 650
 651     /* add at the end of list */
 652     pse = &first_se;
 653     while (*pse != NULL) {
 654         if (instance_id == -1
 655                 && strcmp(se->idstr, (*pse)->idstr) == 0
 656                 && se->instance_id <= (*pse)->instance_id)
 657             se->instance_id = (*pse)->instance_id + 1;
 658         pse = &(*pse)->next;
 659     }
 660     *pse = se;
 661     return 0;
 662 }
 663
 664 int register_savevm(const char *idstr,
 665                     int instance_id,
 666                     int version_id,
 667                     SaveStateHandler *save_state,
 668                     LoadStateHandler *load_state,
 669                     void *opaque)
 670 {
 671     return register_savevm_live(idstr, instance_id, version_id,
 672                                 NULL, save_state, load_state, opaque);
 673 }
 674
 675 void unregister_savevm(const char *idstr, void *opaque)
 676 {
 677     SaveStateEntry **pse;
 678
 679     pse = &first_se;
 680     while (*pse != NULL) {
 681         if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
 682             SaveStateEntry *next = (*pse)->next;
 683             qemu_free(*pse);
 684             *pse = next;
 685             continue;
 686         }
 687         pse = &(*pse)->next;
 688     }
 689 }
 690
 691 #define QEMU_VM_FILE_MAGIC           0x5145564d
 692 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 693 #define QEMU_VM_FILE_VERSION         0x00000003
 694
 695 #define QEMU_VM_EOF                  0x00
 696 #define QEMU_VM_SECTION_START        0x01
 697 #define QEMU_VM_SECTION_PART         0x02
 698 #define QEMU_VM_SECTION_END          0x03
 699 #define QEMU_VM_SECTION_FULL         0x04
 700
 701 int qemu_savevm_state_begin(QEMUFile *f)
 702 {
 703     SaveStateEntry *se;
 704
 705     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 706     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 707
 708     for (se = first_se; se != NULL; se = se->next) {
 709         int len;
 710
 711         if (se->save_live_state == NULL)
 712             continue;
 713
 714         /* Section type */
 715         qemu_put_byte(f, QEMU_VM_SECTION_START);
 716         qemu_put_be32(f, se->section_id);
 717
 718         /* ID string */
 719         len = strlen(se->idstr);
 720         qemu_put_byte(f, len);
 721         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 722
 723         qemu_put_be32(f, se->instance_id);
 724         qemu_put_be32(f, se->version_id);
 725
 726         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 727     }
 728
 729     if (qemu_file_has_error(f))
 730         return -EIO;
 731
 732     return 0;
 733 }
 734
 735 int qemu_savevm_state_iterate(QEMUFile *f)
 736 {
 737     SaveStateEntry *se;
 738     int ret = 1;
 739
 740     for (se = first_se; se != NULL; se = se->next) {
 741         if (se->save_live_state == NULL)
 742             continue;
 743
 744         /* Section type */
 745         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 746         qemu_put_be32(f, se->section_id);
 747
 748         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 749     }
 750
 751     if (ret)
 752         return 1;
 753
 754     if (qemu_file_has_error(f))
 755         return -EIO;
 756
 757     return 0;
 758 }
 759
 760 int qemu_savevm_state_complete(QEMUFile *f)
 761 {
 762     SaveStateEntry *se;
 763
 764     for (se = first_se; se != NULL; se = se->next) {
 765         if (se->save_live_state == NULL)
 766             continue;
 767
 768         /* Section type */
 769         qemu_put_byte(f, QEMU_VM_SECTION_END);
 770         qemu_put_be32(f, se->section_id);
 771
 772         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 773     }
 774
 775     for(se = first_se; se != NULL; se = se->next) {
 776         int len;
 777
 778         if (se->save_state == NULL)
 779             continue;
 780
 781         /* Section type */
 782         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 783         qemu_put_be32(f, se->section_id);
 784
 785         /* ID string */
 786         len = strlen(se->idstr);
 787         qemu_put_byte(f, len);
 788         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 789
 790         qemu_put_be32(f, se->instance_id);
 791         qemu_put_be32(f, se->version_id);
 792
 793         se->save_state(f, se->opaque);
 794     }
 795
 796     qemu_put_byte(f, QEMU_VM_EOF);
 797
 798     if (qemu_file_has_error(f))
 799         return -EIO;
 800
 801     return 0;
 802 }
 803
 804 int qemu_savevm_state(QEMUFile *f)
 805 {
 806     int saved_vm_running;
 807     int ret;
 808
 809     saved_vm_running = vm_running;
 810     vm_stop(0);
 811
 812     bdrv_flush_all();
 813
 814     ret = qemu_savevm_state_begin(f);
 815     if (ret < 0)
 816         goto out;
 817
 818     do {
 819         ret = qemu_savevm_state_iterate(f);
 820         if (ret < 0)
 821             goto out;
 822     } while (ret == 0);
 823
 824     ret = qemu_savevm_state_complete(f);
 825
 826 out:
 827     if (qemu_file_has_error(f))
 828         ret = -EIO;
 829
 830     if (!ret && saved_vm_running)
 831         vm_start();
 832
 833     return ret;
 834 }
 835
 836 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 837 {
 838     SaveStateEntry *se;
 839
 840     for(se = first_se; se != NULL; se = se->next) {
 841         if (!strcmp(se->idstr, idstr) &&
 842             instance_id == se->instance_id)
 843             return se;
 844     }
 845     return NULL;
 846 }
 847
 848 typedef struct LoadStateEntry {
 849     SaveStateEntry *se;
 850     int section_id;
 851     int version_id;
 852     struct LoadStateEntry *next;
 853 } LoadStateEntry;
 854
 855 static int qemu_loadvm_state_v2(QEMUFile *f)
 856 {
 857     SaveStateEntry *se;
 858     int len, ret, instance_id, record_len, version_id;
 859     int64_t total_len, end_pos, cur_pos;
 860     char idstr[256];
 861
 862     total_len = qemu_get_be64(f);
 863     end_pos = total_len + qemu_ftell(f);
 864     for(;;) {
 865         if (qemu_ftell(f) >= end_pos)
 866             break;
 867         len = qemu_get_byte(f);
 868         qemu_get_buffer(f, (uint8_t *)idstr, len);
 869         idstr[len] = '\0';
 870         instance_id = qemu_get_be32(f);
 871         version_id = qemu_get_be32(f);
 872         record_len = qemu_get_be32(f);
 873         cur_pos = qemu_ftell(f);
 874         se = find_se(idstr, instance_id);
 875         if (!se) {
 876             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 877                     instance_id, idstr);
 878         } else {
 879             ret = se->load_state(f, se->opaque, version_id);
 880             if (ret < 0) {
 881                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 882                         instance_id, idstr);
 883                 return ret;
 884             }
 885         }
 886         /* always seek to exact end of record */
 887         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 888     }
 889
 890     if (qemu_file_has_error(f))
 891         return -EIO;
 892
 893     return 0;
 894 }
 895
 896 int qemu_loadvm_state(QEMUFile *f)
 897 {
 898     LoadStateEntry *first_le = NULL;
 899     uint8_t section_type;
 900     unsigned int v;
 901     int ret;
 902
 903     v = qemu_get_be32(f);
 904     if (v != QEMU_VM_FILE_MAGIC)
 905         return -EINVAL;
 906
 907     v = qemu_get_be32(f);
 908     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 909         return qemu_loadvm_state_v2(f);
 910     if (v != QEMU_VM_FILE_VERSION)
 911         return -ENOTSUP;
 912
 913     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 914         uint32_t instance_id, version_id, section_id;
 915         LoadStateEntry *le;
 916         SaveStateEntry *se;
 917         char idstr[257];
 918         int len;
 919
 920         switch (section_type) {
 921         case QEMU_VM_SECTION_START:
 922         case QEMU_VM_SECTION_FULL:
 923             /* Read section start */
 924             section_id = qemu_get_be32(f);
 925             len = qemu_get_byte(f);
 926             qemu_get_buffer(f, (uint8_t *)idstr, len);
 927             idstr[len] = 0;
 928             instance_id = qemu_get_be32(f);
 929             version_id = qemu_get_be32(f);
 930
 931             /* Find savevm section */
 932             se = find_se(idstr, instance_id);
 933             if (se == NULL) {
 934                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 935                 ret = -EINVAL;
 936                 goto out;
 937             }
 938
 939             /* Validate version */
 940             if (version_id > se->version_id) {
 941                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 942                         version_id, idstr, se->version_id);
 943                 ret = -EINVAL;
 944                 goto out;
 945             }
 946
 947             /* Add entry */
 948             le = qemu_mallocz(sizeof(*le));
 949
 950             le->se = se;
 951             le->section_id = section_id;
 952             le->version_id = version_id;
 953             le->next = first_le;
 954             first_le = le;
 955
 956             le->se->load_state(f, le->se->opaque, le->version_id);
 957             break;
 958         case QEMU_VM_SECTION_PART:
 959         case QEMU_VM_SECTION_END:
 960             section_id = qemu_get_be32(f);
 961
 962             for (le = first_le; le && le->section_id != section_id; le = le->next);
 963             if (le == NULL) {
 964                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 965                 ret = -EINVAL;
 966                 goto out;
 967             }
 968
 969             le->se->load_state(f, le->se->opaque, le->version_id);
 970             break;
 971         default:
 972             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 973             ret = -EINVAL;
 974             goto out;
 975         }
 976     }
 977
 978     ret = 0;
 979
 980 out:
 981     while (first_le) {
 982         LoadStateEntry *le = first_le;
 983         first_le = first_le->next;
 984         qemu_free(le);
 985     }
 986
 987     if (qemu_file_has_error(f))
 988         ret = -EIO;
 989
 990     return ret;
 991 }
 992
 993 /* device can contain snapshots */
 994 static int bdrv_can_snapshot(BlockDriverState *bs)
 995 {
 996     return (bs &&
 997             !bdrv_is_removable(bs) &&
 998             !bdrv_is_read_only(bs));
 999 }
1000
1001 /* device must be snapshots in order to have a reliable snapshot */
1002 static int bdrv_has_snapshot(BlockDriverState *bs)
1003 {
1004     return (bs &&
1005             !bdrv_is_removable(bs) &&
1006             !bdrv_is_read_only(bs));
1007 }
1008
1009 static BlockDriverState *get_bs_snapshots(void)
1010 {
1011     BlockDriverState *bs;
1012     int i;
1013
1014     if (bs_snapshots)
1015         return bs_snapshots;
1016     for(i = 0; i <= nb_drives; i++) {
1017         bs = drives_table[i].bdrv;
1018         if (bdrv_can_snapshot(bs))
1019             goto ok;
1020     }
1021     return NULL;
1022  ok:
1023     bs_snapshots = bs;
1024     return bs;
1025 }
1026
1027 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1028                               const char *name)
1029 {
1030     QEMUSnapshotInfo *sn_tab, *sn;
1031     int nb_sns, i, ret;
1032
1033     ret = -ENOENT;
1034     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1035     if (nb_sns < 0)
1036         return ret;
1037     for(i = 0; i < nb_sns; i++) {
1038         sn = &sn_tab[i];
1039         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1040             *sn_info = *sn;
1041             ret = 0;
1042             break;
1043         }
1044     }
1045     qemu_free(sn_tab);
1046     return ret;
1047 }
1048
1049 void do_savevm(Monitor *mon, const char *name)
1050 {
1051     BlockDriverState *bs, *bs1;
1052     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1053     int must_delete, ret, i;
1054     QEMUFile *f;
1055     int saved_vm_running;
1056     uint32_t vm_state_size;
1057 #ifdef _WIN32
1058     struct _timeb tb;
1059 #else
1060     struct timeval tv;
1061 #endif
1062
1063     bs = get_bs_snapshots();
1064     if (!bs) {
1065         monitor_printf(mon, "No block device can accept snapshots\n");
1066         return;
1067     }
1068
1069     /* ??? Should this occur after vm_stop?  */
1070     qemu_aio_flush();
1071
1072     saved_vm_running = vm_running;
1073     vm_stop(0);
1074
1075     must_delete = 0;
1076     if (name) {
1077         ret = bdrv_snapshot_find(bs, old_sn, name);
1078         if (ret >= 0) {
1079             must_delete = 1;
1080         }
1081     }
1082     memset(sn, 0, sizeof(*sn));
1083     if (must_delete) {
1084         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1085         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1086     } else {
1087         if (name)
1088             pstrcpy(sn->name, sizeof(sn->name), name);
1089     }
1090
1091     /* fill auxiliary fields */
1092 #ifdef _WIN32
1093     _ftime(&tb);
1094     sn->date_sec = tb.time;
1095     sn->date_nsec = tb.millitm * 1000000;
1096 #else
1097     gettimeofday(&tv, NULL);
1098     sn->date_sec = tv.tv_sec;
1099     sn->date_nsec = tv.tv_usec * 1000;
1100 #endif
1101     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1102
1103     /* save the VM state */
1104     f = qemu_fopen_bdrv(bs, 1);
1105     if (!f) {
1106         monitor_printf(mon, "Could not open VM state file\n");
1107         goto the_end;
1108     }
1109     ret = qemu_savevm_state(f);
1110     vm_state_size = qemu_ftell(f);
1111     qemu_fclose(f);
1112     if (ret < 0) {
1113         monitor_printf(mon, "Error %d while writing VM\n", ret);
1114         goto the_end;
1115     }
1116
1117     /* create the snapshots */
1118
1119     for(i = 0; i < nb_drives; i++) {
1120         bs1 = drives_table[i].bdrv;
1121         if (bdrv_has_snapshot(bs1)) {
1122             if (must_delete) {
1123                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1124                 if (ret < 0) {
1125                     monitor_printf(mon,
1126                                    "Error while deleting snapshot on '%s'\n",
1127                                    bdrv_get_device_name(bs1));
1128                 }
1129             }
1130             /* Write VM state size only to the image that contains the state */
1131             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1132             ret = bdrv_snapshot_create(bs1, sn);
1133             if (ret < 0) {
1134                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1135                                bdrv_get_device_name(bs1));
1136             }
1137         }
1138     }
1139
1140  the_end:
1141     if (saved_vm_running)
1142         vm_start();
1143 }
1144
1145 void do_loadvm(Monitor *mon, const char *name)
1146 {
1147     BlockDriverState *bs, *bs1;
1148     QEMUSnapshotInfo sn;
1149     QEMUFile *f;
1150     int i, ret;
1151     int saved_vm_running;
1152
1153     bs = get_bs_snapshots();
1154     if (!bs) {
1155         monitor_printf(mon, "No block device supports snapshots\n");
1156         return;
1157     }
1158
1159     /* Flush all IO requests so they don't interfere with the new state.  */
1160     qemu_aio_flush();
1161
1162     saved_vm_running = vm_running;
1163     vm_stop(0);
1164
1165     for(i = 0; i <= nb_drives; i++) {
1166         bs1 = drives_table[i].bdrv;
1167         if (bdrv_has_snapshot(bs1)) {
1168             ret = bdrv_snapshot_goto(bs1, name);
1169             if (ret < 0) {
1170                 if (bs != bs1)
1171                     monitor_printf(mon, "Warning: ");
1172                 switch(ret) {
1173                 case -ENOTSUP:
1174                     monitor_printf(mon,
1175                                    "Snapshots not supported on device '%s'\n",
1176                                    bdrv_get_device_name(bs1));
1177                     break;
1178                 case -ENOENT:
1179                     monitor_printf(mon, "Could not find snapshot '%s' on "
1180                                    "device '%s'\n",
1181                                    name, bdrv_get_device_name(bs1));
1182                     break;
1183                 default:
1184                     monitor_printf(mon, "Error %d while activating snapshot on"
1185                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1186                     break;
1187                 }
1188                 /* fatal on snapshot block device */
1189                 if (bs == bs1)
1190                     goto the_end;
1191             }
1192         }
1193     }
1194
1195     /* Don't even try to load empty VM states */
1196     ret = bdrv_snapshot_find(bs, &sn, name);
1197     if ((ret >= 0) && (sn.vm_state_size == 0))
1198         goto the_end;
1199
1200     /* restore the VM state */
1201     f = qemu_fopen_bdrv(bs, 0);
1202     if (!f) {
1203         monitor_printf(mon, "Could not open VM state file\n");
1204         goto the_end;
1205     }
1206     ret = qemu_loadvm_state(f);
1207     qemu_fclose(f);
1208     if (ret < 0) {
1209         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1210     }
1211  the_end:
1212     if (saved_vm_running)
1213         vm_start();
1214 }
1215
1216 void do_delvm(Monitor *mon, const char *name)
1217 {
1218     BlockDriverState *bs, *bs1;
1219     int i, ret;
1220
1221     bs = get_bs_snapshots();
1222     if (!bs) {
1223         monitor_printf(mon, "No block device supports snapshots\n");
1224         return;
1225     }
1226
1227     for(i = 0; i <= nb_drives; i++) {
1228         bs1 = drives_table[i].bdrv;
1229         if (bdrv_has_snapshot(bs1)) {
1230             ret = bdrv_snapshot_delete(bs1, name);
1231             if (ret < 0) {
1232                 if (ret == -ENOTSUP)
1233                     monitor_printf(mon,
1234                                    "Snapshots not supported on device '%s'\n",
1235                                    bdrv_get_device_name(bs1));
1236                 else
1237                     monitor_printf(mon, "Error %d while deleting snapshot on "
1238                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1239             }
1240         }
1241     }
1242 }
1243
1244 void do_info_snapshots(Monitor *mon)
1245 {
1246     BlockDriverState *bs, *bs1;
1247     QEMUSnapshotInfo *sn_tab, *sn;
1248     int nb_sns, i;
1249     char buf[256];
1250
1251     bs = get_bs_snapshots();
1252     if (!bs) {
1253         monitor_printf(mon, "No available block device supports snapshots\n");
1254         return;
1255     }
1256     monitor_printf(mon, "Snapshot devices:");
1257     for(i = 0; i <= nb_drives; i++) {
1258         bs1 = drives_table[i].bdrv;
1259         if (bdrv_has_snapshot(bs1)) {
1260             if (bs == bs1)
1261                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1262         }
1263     }
1264     monitor_printf(mon, "\n");
1265
1266     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1267     if (nb_sns < 0) {
1268         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1269         return;
1270     }
1271     monitor_printf(mon, "Snapshot list (from %s):\n",
1272                    bdrv_get_device_name(bs));
1273     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1274     for(i = 0; i < nb_sns; i++) {
1275         sn = &sn_tab[i];
1276         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1277     }
1278     qemu_free(sn_tab);
1279 }