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 CONFIG_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 CONFIG_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     /* any failed or completed migration keeps its state to allow probing of
 560      * migration data, but has no associated file anymore */
 561     if (f && f->set_rate_limit)
 562         return f->set_rate_limit(f->opaque, new_rate);
 563
 564     return 0;
 565 }
 566
 567 void qemu_put_be16(QEMUFile *f, unsigned int v)
 568 {
 569     qemu_put_byte(f, v >> 8);
 570     qemu_put_byte(f, v);
 571 }
 572
 573 void qemu_put_be32(QEMUFile *f, unsigned int v)
 574 {
 575     qemu_put_byte(f, v >> 24);
 576     qemu_put_byte(f, v >> 16);
 577     qemu_put_byte(f, v >> 8);
 578     qemu_put_byte(f, v);
 579 }
 580
 581 void qemu_put_be64(QEMUFile *f, uint64_t v)
 582 {
 583     qemu_put_be32(f, v >> 32);
 584     qemu_put_be32(f, v);
 585 }
 586
 587 unsigned int qemu_get_be16(QEMUFile *f)
 588 {
 589     unsigned int v;
 590     v = qemu_get_byte(f) << 8;
 591     v |= qemu_get_byte(f);
 592     return v;
 593 }
 594
 595 unsigned int qemu_get_be32(QEMUFile *f)
 596 {
 597     unsigned int v;
 598     v = qemu_get_byte(f) << 24;
 599     v |= qemu_get_byte(f) << 16;
 600     v |= qemu_get_byte(f) << 8;
 601     v |= qemu_get_byte(f);
 602     return v;
 603 }
 604
 605 uint64_t qemu_get_be64(QEMUFile *f)
 606 {
 607     uint64_t v;
 608     v = (uint64_t)qemu_get_be32(f) << 32;
 609     v |= qemu_get_be32(f);
 610     return v;
 611 }
 612
 613 typedef struct SaveStateEntry {
 614     char idstr[256];
 615     int instance_id;
 616     int version_id;
 617     int section_id;
 618     SaveLiveStateHandler *save_live_state;
 619     SaveStateHandler *save_state;
 620     LoadStateHandler *load_state;
 621     void *opaque;
 622     struct SaveStateEntry *next;
 623 } SaveStateEntry;
 624
 625 static SaveStateEntry *first_se;
 626
 627 /* TODO: Individual devices generally have very little idea about the rest
 628    of the system, so instance_id should be removed/replaced.
 629    Meanwhile pass -1 as instance_id if you do not already have a clearly
 630    distinguishing id for all instances of your device class. */
 631 int register_savevm_live(const char *idstr,
 632                          int instance_id,
 633                          int version_id,
 634                          SaveLiveStateHandler *save_live_state,
 635                          SaveStateHandler *save_state,
 636                          LoadStateHandler *load_state,
 637                          void *opaque)
 638 {
 639     SaveStateEntry *se, **pse;
 640     static int global_section_id;
 641
 642     se = qemu_malloc(sizeof(SaveStateEntry));
 643     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 644     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 645     se->version_id = version_id;
 646     se->section_id = global_section_id++;
 647     se->save_live_state = save_live_state;
 648     se->save_state = save_state;
 649     se->load_state = load_state;
 650     se->opaque = opaque;
 651     se->next = NULL;
 652
 653     /* add at the end of list */
 654     pse = &first_se;
 655     while (*pse != NULL) {
 656         if (instance_id == -1
 657                 && strcmp(se->idstr, (*pse)->idstr) == 0
 658                 && se->instance_id <= (*pse)->instance_id)
 659             se->instance_id = (*pse)->instance_id + 1;
 660         pse = &(*pse)->next;
 661     }
 662     *pse = se;
 663     return 0;
 664 }
 665
 666 int register_savevm(const char *idstr,
 667                     int instance_id,
 668                     int version_id,
 669                     SaveStateHandler *save_state,
 670                     LoadStateHandler *load_state,
 671                     void *opaque)
 672 {
 673     return register_savevm_live(idstr, instance_id, version_id,
 674                                 NULL, save_state, load_state, opaque);
 675 }
 676
 677 void unregister_savevm(const char *idstr, void *opaque)
 678 {
 679     SaveStateEntry **pse;
 680
 681     pse = &first_se;
 682     while (*pse != NULL) {
 683         if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
 684             SaveStateEntry *next = (*pse)->next;
 685             qemu_free(*pse);
 686             *pse = next;
 687             continue;
 688         }
 689         pse = &(*pse)->next;
 690     }
 691 }
 692
 693 #define QEMU_VM_FILE_MAGIC           0x5145564d
 694 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 695 #define QEMU_VM_FILE_VERSION         0x00000003
 696
 697 #define QEMU_VM_EOF                  0x00
 698 #define QEMU_VM_SECTION_START        0x01
 699 #define QEMU_VM_SECTION_PART         0x02
 700 #define QEMU_VM_SECTION_END          0x03
 701 #define QEMU_VM_SECTION_FULL         0x04
 702
 703 int qemu_savevm_state_begin(QEMUFile *f)
 704 {
 705     SaveStateEntry *se;
 706
 707     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 708     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 709
 710     for (se = first_se; se != NULL; se = se->next) {
 711         int len;
 712
 713         if (se->save_live_state == NULL)
 714             continue;
 715
 716         /* Section type */
 717         qemu_put_byte(f, QEMU_VM_SECTION_START);
 718         qemu_put_be32(f, se->section_id);
 719
 720         /* ID string */
 721         len = strlen(se->idstr);
 722         qemu_put_byte(f, len);
 723         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 724
 725         qemu_put_be32(f, se->instance_id);
 726         qemu_put_be32(f, se->version_id);
 727
 728         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 729     }
 730
 731     if (qemu_file_has_error(f))
 732         return -EIO;
 733
 734     return 0;
 735 }
 736
 737 int qemu_savevm_state_iterate(QEMUFile *f)
 738 {
 739     SaveStateEntry *se;
 740     int ret = 1;
 741
 742     for (se = first_se; se != NULL; se = se->next) {
 743         if (se->save_live_state == NULL)
 744             continue;
 745
 746         /* Section type */
 747         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 748         qemu_put_be32(f, se->section_id);
 749
 750         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 751     }
 752
 753     if (ret)
 754         return 1;
 755
 756     if (qemu_file_has_error(f))
 757         return -EIO;
 758
 759     return 0;
 760 }
 761
 762 int qemu_savevm_state_complete(QEMUFile *f)
 763 {
 764     SaveStateEntry *se;
 765
 766     for (se = first_se; se != NULL; se = se->next) {
 767         if (se->save_live_state == NULL)
 768             continue;
 769
 770         /* Section type */
 771         qemu_put_byte(f, QEMU_VM_SECTION_END);
 772         qemu_put_be32(f, se->section_id);
 773
 774         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 775     }
 776
 777     for(se = first_se; se != NULL; se = se->next) {
 778         int len;
 779
 780         if (se->save_state == NULL)
 781             continue;
 782
 783         /* Section type */
 784         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 785         qemu_put_be32(f, se->section_id);
 786
 787         /* ID string */
 788         len = strlen(se->idstr);
 789         qemu_put_byte(f, len);
 790         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 791
 792         qemu_put_be32(f, se->instance_id);
 793         qemu_put_be32(f, se->version_id);
 794
 795         se->save_state(f, se->opaque);
 796     }
 797
 798     qemu_put_byte(f, QEMU_VM_EOF);
 799
 800     if (qemu_file_has_error(f))
 801         return -EIO;
 802
 803     return 0;
 804 }
 805
 806 int qemu_savevm_state(QEMUFile *f)
 807 {
 808     int saved_vm_running;
 809     int ret;
 810
 811     saved_vm_running = vm_running;
 812     vm_stop(0);
 813
 814     bdrv_flush_all();
 815
 816     ret = qemu_savevm_state_begin(f);
 817     if (ret < 0)
 818         goto out;
 819
 820     do {
 821         ret = qemu_savevm_state_iterate(f);
 822         if (ret < 0)
 823             goto out;
 824     } while (ret == 0);
 825
 826     ret = qemu_savevm_state_complete(f);
 827
 828 out:
 829     if (qemu_file_has_error(f))
 830         ret = -EIO;
 831
 832     if (!ret && saved_vm_running)
 833         vm_start();
 834
 835     return ret;
 836 }
 837
 838 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 839 {
 840     SaveStateEntry *se;
 841
 842     for(se = first_se; se != NULL; se = se->next) {
 843         if (!strcmp(se->idstr, idstr) &&
 844             instance_id == se->instance_id)
 845             return se;
 846     }
 847     return NULL;
 848 }
 849
 850 typedef struct LoadStateEntry {
 851     SaveStateEntry *se;
 852     int section_id;
 853     int version_id;
 854     struct LoadStateEntry *next;
 855 } LoadStateEntry;
 856
 857 static int qemu_loadvm_state_v2(QEMUFile *f)
 858 {
 859     SaveStateEntry *se;
 860     int len, ret, instance_id, record_len, version_id;
 861     int64_t total_len, end_pos, cur_pos;
 862     char idstr[256];
 863
 864     total_len = qemu_get_be64(f);
 865     end_pos = total_len + qemu_ftell(f);
 866     for(;;) {
 867         if (qemu_ftell(f) >= end_pos)
 868             break;
 869         len = qemu_get_byte(f);
 870         qemu_get_buffer(f, (uint8_t *)idstr, len);
 871         idstr[len] = '\0';
 872         instance_id = qemu_get_be32(f);
 873         version_id = qemu_get_be32(f);
 874         record_len = qemu_get_be32(f);
 875         cur_pos = qemu_ftell(f);
 876         se = find_se(idstr, instance_id);
 877         if (!se) {
 878             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 879                     instance_id, idstr);
 880         } else {
 881             ret = se->load_state(f, se->opaque, version_id);
 882             if (ret < 0) {
 883                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 884                         instance_id, idstr);
 885                 return ret;
 886             }
 887         }
 888         /* always seek to exact end of record */
 889         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 890     }
 891
 892     if (qemu_file_has_error(f))
 893         return -EIO;
 894
 895     return 0;
 896 }
 897
 898 int qemu_loadvm_state(QEMUFile *f)
 899 {
 900     LoadStateEntry *first_le = NULL;
 901     uint8_t section_type;
 902     unsigned int v;
 903     int ret;
 904
 905     v = qemu_get_be32(f);
 906     if (v != QEMU_VM_FILE_MAGIC)
 907         return -EINVAL;
 908
 909     v = qemu_get_be32(f);
 910     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 911         return qemu_loadvm_state_v2(f);
 912     if (v != QEMU_VM_FILE_VERSION)
 913         return -ENOTSUP;
 914
 915     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 916         uint32_t instance_id, version_id, section_id;
 917         LoadStateEntry *le;
 918         SaveStateEntry *se;
 919         char idstr[257];
 920         int len;
 921
 922         switch (section_type) {
 923         case QEMU_VM_SECTION_START:
 924         case QEMU_VM_SECTION_FULL:
 925             /* Read section start */
 926             section_id = qemu_get_be32(f);
 927             len = qemu_get_byte(f);
 928             qemu_get_buffer(f, (uint8_t *)idstr, len);
 929             idstr[len] = 0;
 930             instance_id = qemu_get_be32(f);
 931             version_id = qemu_get_be32(f);
 932
 933             /* Find savevm section */
 934             se = find_se(idstr, instance_id);
 935             if (se == NULL) {
 936                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 937                 ret = -EINVAL;
 938                 goto out;
 939             }
 940
 941             /* Validate version */
 942             if (version_id > se->version_id) {
 943                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 944                         version_id, idstr, se->version_id);
 945                 ret = -EINVAL;
 946                 goto out;
 947             }
 948
 949             /* Add entry */
 950             le = qemu_mallocz(sizeof(*le));
 951
 952             le->se = se;
 953             le->section_id = section_id;
 954             le->version_id = version_id;
 955             le->next = first_le;
 956             first_le = le;
 957
 958             le->se->load_state(f, le->se->opaque, le->version_id);
 959             break;
 960         case QEMU_VM_SECTION_PART:
 961         case QEMU_VM_SECTION_END:
 962             section_id = qemu_get_be32(f);
 963
 964             for (le = first_le; le && le->section_id != section_id; le = le->next);
 965             if (le == NULL) {
 966                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 967                 ret = -EINVAL;
 968                 goto out;
 969             }
 970
 971             le->se->load_state(f, le->se->opaque, le->version_id);
 972             break;
 973         default:
 974             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 975             ret = -EINVAL;
 976             goto out;
 977         }
 978     }
 979
 980     ret = 0;
 981
 982 out:
 983     while (first_le) {
 984         LoadStateEntry *le = first_le;
 985         first_le = first_le->next;
 986         qemu_free(le);
 987     }
 988
 989     if (qemu_file_has_error(f))
 990         ret = -EIO;
 991
 992     return ret;
 993 }
 994
 995 /* device can contain snapshots */
 996 static int bdrv_can_snapshot(BlockDriverState *bs)
 997 {
 998     return (bs &&
 999             !bdrv_is_removable(bs) &&
1000             !bdrv_is_read_only(bs));
1001 }
1002
1003 /* device must be snapshots in order to have a reliable snapshot */
1004 static int bdrv_has_snapshot(BlockDriverState *bs)
1005 {
1006     return (bs &&
1007             !bdrv_is_removable(bs) &&
1008             !bdrv_is_read_only(bs));
1009 }
1010
1011 static BlockDriverState *get_bs_snapshots(void)
1012 {
1013     BlockDriverState *bs;
1014     DriveInfo *dinfo;
1015
1016     if (bs_snapshots)
1017         return bs_snapshots;
1018     TAILQ_FOREACH(dinfo, &drives, next) {
1019         bs = dinfo->bdrv;
1020         if (bdrv_can_snapshot(bs))
1021             goto ok;
1022     }
1023     return NULL;
1024  ok:
1025     bs_snapshots = bs;
1026     return bs;
1027 }
1028
1029 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1030                               const char *name)
1031 {
1032     QEMUSnapshotInfo *sn_tab, *sn;
1033     int nb_sns, i, ret;
1034
1035     ret = -ENOENT;
1036     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1037     if (nb_sns < 0)
1038         return ret;
1039     for(i = 0; i < nb_sns; i++) {
1040         sn = &sn_tab[i];
1041         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1042             *sn_info = *sn;
1043             ret = 0;
1044             break;
1045         }
1046     }
1047     qemu_free(sn_tab);
1048     return ret;
1049 }
1050
1051 void do_savevm(Monitor *mon, const char *name)
1052 {
1053     DriveInfo *dinfo;
1054     BlockDriverState *bs, *bs1;
1055     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1056     int must_delete, ret;
1057     QEMUFile *f;
1058     int saved_vm_running;
1059     uint32_t vm_state_size;
1060 #ifdef _WIN32
1061     struct _timeb tb;
1062 #else
1063     struct timeval tv;
1064 #endif
1065
1066     bs = get_bs_snapshots();
1067     if (!bs) {
1068         monitor_printf(mon, "No block device can accept snapshots\n");
1069         return;
1070     }
1071
1072     /* ??? Should this occur after vm_stop?  */
1073     qemu_aio_flush();
1074
1075     saved_vm_running = vm_running;
1076     vm_stop(0);
1077
1078     must_delete = 0;
1079     if (name) {
1080         ret = bdrv_snapshot_find(bs, old_sn, name);
1081         if (ret >= 0) {
1082             must_delete = 1;
1083         }
1084     }
1085     memset(sn, 0, sizeof(*sn));
1086     if (must_delete) {
1087         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1088         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1089     } else {
1090         if (name)
1091             pstrcpy(sn->name, sizeof(sn->name), name);
1092     }
1093
1094     /* fill auxiliary fields */
1095 #ifdef _WIN32
1096     _ftime(&tb);
1097     sn->date_sec = tb.time;
1098     sn->date_nsec = tb.millitm * 1000000;
1099 #else
1100     gettimeofday(&tv, NULL);
1101     sn->date_sec = tv.tv_sec;
1102     sn->date_nsec = tv.tv_usec * 1000;
1103 #endif
1104     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1105
1106     /* save the VM state */
1107     f = qemu_fopen_bdrv(bs, 1);
1108     if (!f) {
1109         monitor_printf(mon, "Could not open VM state file\n");
1110         goto the_end;
1111     }
1112     ret = qemu_savevm_state(f);
1113     vm_state_size = qemu_ftell(f);
1114     qemu_fclose(f);
1115     if (ret < 0) {
1116         monitor_printf(mon, "Error %d while writing VM\n", ret);
1117         goto the_end;
1118     }
1119
1120     /* create the snapshots */
1121
1122     TAILQ_FOREACH(dinfo, &drives, next) {
1123         bs1 = dinfo->bdrv;
1124         if (bdrv_has_snapshot(bs1)) {
1125             if (must_delete) {
1126                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1127                 if (ret < 0) {
1128                     monitor_printf(mon,
1129                                    "Error while deleting snapshot on '%s'\n",
1130                                    bdrv_get_device_name(bs1));
1131                 }
1132             }
1133             /* Write VM state size only to the image that contains the state */
1134             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1135             ret = bdrv_snapshot_create(bs1, sn);
1136             if (ret < 0) {
1137                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1138                                bdrv_get_device_name(bs1));
1139             }
1140         }
1141     }
1142
1143  the_end:
1144     if (saved_vm_running)
1145         vm_start();
1146 }
1147
1148 void do_loadvm(Monitor *mon, const char *name)
1149 {
1150     DriveInfo *dinfo;
1151     BlockDriverState *bs, *bs1;
1152     QEMUSnapshotInfo sn;
1153     QEMUFile *f;
1154     int ret;
1155     int saved_vm_running;
1156
1157     bs = get_bs_snapshots();
1158     if (!bs) {
1159         monitor_printf(mon, "No block device supports snapshots\n");
1160         return;
1161     }
1162
1163     /* Flush all IO requests so they don't interfere with the new state.  */
1164     qemu_aio_flush();
1165
1166     saved_vm_running = vm_running;
1167     vm_stop(0);
1168
1169     TAILQ_FOREACH(dinfo, &drives, next) {
1170         bs1 = dinfo->bdrv;
1171         if (bdrv_has_snapshot(bs1)) {
1172             ret = bdrv_snapshot_goto(bs1, name);
1173             if (ret < 0) {
1174                 if (bs != bs1)
1175                     monitor_printf(mon, "Warning: ");
1176                 switch(ret) {
1177                 case -ENOTSUP:
1178                     monitor_printf(mon,
1179                                    "Snapshots not supported on device '%s'\n",
1180                                    bdrv_get_device_name(bs1));
1181                     break;
1182                 case -ENOENT:
1183                     monitor_printf(mon, "Could not find snapshot '%s' on "
1184                                    "device '%s'\n",
1185                                    name, bdrv_get_device_name(bs1));
1186                     break;
1187                 default:
1188                     monitor_printf(mon, "Error %d while activating snapshot on"
1189                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1190                     break;
1191                 }
1192                 /* fatal on snapshot block device */
1193                 if (bs == bs1)
1194                     goto the_end;
1195             }
1196         }
1197     }
1198
1199     /* Don't even try to load empty VM states */
1200     ret = bdrv_snapshot_find(bs, &sn, name);
1201     if ((ret >= 0) && (sn.vm_state_size == 0))
1202         goto the_end;
1203
1204     /* restore the VM state */
1205     f = qemu_fopen_bdrv(bs, 0);
1206     if (!f) {
1207         monitor_printf(mon, "Could not open VM state file\n");
1208         goto the_end;
1209     }
1210     ret = qemu_loadvm_state(f);
1211     qemu_fclose(f);
1212     if (ret < 0) {
1213         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1214     }
1215  the_end:
1216     if (saved_vm_running)
1217         vm_start();
1218 }
1219
1220 void do_delvm(Monitor *mon, const char *name)
1221 {
1222     DriveInfo *dinfo;
1223     BlockDriverState *bs, *bs1;
1224     int ret;
1225
1226     bs = get_bs_snapshots();
1227     if (!bs) {
1228         monitor_printf(mon, "No block device supports snapshots\n");
1229         return;
1230     }
1231
1232     TAILQ_FOREACH(dinfo, &drives, next) {
1233         bs1 = dinfo->bdrv;
1234         if (bdrv_has_snapshot(bs1)) {
1235             ret = bdrv_snapshot_delete(bs1, name);
1236             if (ret < 0) {
1237                 if (ret == -ENOTSUP)
1238                     monitor_printf(mon,
1239                                    "Snapshots not supported on device '%s'\n",
1240                                    bdrv_get_device_name(bs1));
1241                 else
1242                     monitor_printf(mon, "Error %d while deleting snapshot on "
1243                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1244             }
1245         }
1246     }
1247 }
1248
1249 void do_info_snapshots(Monitor *mon)
1250 {
1251     DriveInfo *dinfo;
1252     BlockDriverState *bs, *bs1;
1253     QEMUSnapshotInfo *sn_tab, *sn;
1254     int nb_sns, i;
1255     char buf[256];
1256
1257     bs = get_bs_snapshots();
1258     if (!bs) {
1259         monitor_printf(mon, "No available block device supports snapshots\n");
1260         return;
1261     }
1262     monitor_printf(mon, "Snapshot devices:");
1263     TAILQ_FOREACH(dinfo, &drives, next) {
1264         bs1 = dinfo->bdrv;
1265         if (bdrv_has_snapshot(bs1)) {
1266             if (bs == bs1)
1267                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1268         }
1269     }
1270     monitor_printf(mon, "\n");
1271
1272     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1273     if (nb_sns < 0) {
1274         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1275         return;
1276     }
1277     monitor_printf(mon, "Snapshot list (from %s):\n",
1278                    bdrv_get_device_name(bs));
1279     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1280     for(i = 0; i < nb_sns; i++) {
1281         sn = &sn_tab[i];
1282         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1283     }
1284     qemu_free(sn_tab);
1285 }