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->fd_read(vc->opaque, 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, 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     return fread(buf, 1, size, s->popen_file);
 219 }
 220
 221 static int popen_close(void *opaque)
 222 {
 223     QEMUFilePopen *s = opaque;
 224     pclose(s->popen_file);
 225     qemu_free(s);
 226     return 0;
 227 }
 228
 229 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 230 {
 231     QEMUFilePopen *s;
 232
 233     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 234         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 235         return NULL;
 236     }
 237
 238     s = qemu_mallocz(sizeof(QEMUFilePopen));
 239
 240     s->popen_file = popen_file;
 241
 242     if(mode[0] == 'r') {
 243         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL, NULL);
 244     } else {
 245         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL, NULL);
 246     }
 247     fprintf(stderr, "qemu_popen: returning result of qemu_fopen_ops\n");
 248     return s->file;
 249 }
 250
 251 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 252 {
 253     FILE *popen_file;
 254
 255     popen_file = popen(command, mode);
 256     if(popen_file == NULL) {
 257         return NULL;
 258     }
 259
 260     return qemu_popen(popen_file, mode);
 261 }
 262
 263 QEMUFile *qemu_fopen_socket(int fd)
 264 {
 265     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 266
 267     s->fd = fd;
 268     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL, NULL);
 269     return s->file;
 270 }
 271
 272 typedef struct QEMUFileStdio
 273 {
 274     FILE *outfile;
 275 } QEMUFileStdio;
 276
 277 static int file_put_buffer(void *opaque, const uint8_t *buf,
 278                             int64_t pos, int size)
 279 {
 280     QEMUFileStdio *s = opaque;
 281     fseek(s->outfile, pos, SEEK_SET);
 282     fwrite(buf, 1, size, s->outfile);
 283     return size;
 284 }
 285
 286 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 287 {
 288     QEMUFileStdio *s = opaque;
 289     fseek(s->outfile, pos, SEEK_SET);
 290     return fread(buf, 1, size, s->outfile);
 291 }
 292
 293 static int file_close(void *opaque)
 294 {
 295     QEMUFileStdio *s = opaque;
 296     fclose(s->outfile);
 297     qemu_free(s);
 298     return 0;
 299 }
 300
 301 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 302 {
 303     QEMUFileStdio *s;
 304
 305     s = qemu_mallocz(sizeof(QEMUFileStdio));
 306
 307     s->outfile = fopen(filename, mode);
 308     if (!s->outfile)
 309         goto fail;
 310
 311     if (!strcmp(mode, "wb"))
 312         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL, NULL);
 313     else if (!strcmp(mode, "rb"))
 314         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL, NULL);
 315
 316 fail:
 317     if (s->outfile)
 318         fclose(s->outfile);
 319     qemu_free(s);
 320     return NULL;
 321 }
 322
 323 typedef struct QEMUFileBdrv
 324 {
 325     BlockDriverState *bs;
 326     int64_t base_offset;
 327 } QEMUFileBdrv;
 328
 329 static int block_put_buffer(void *opaque, const uint8_t *buf,
 330                            int64_t pos, int size)
 331 {
 332     QEMUFileBdrv *s = opaque;
 333     bdrv_put_buffer(s->bs, buf, s->base_offset + pos, size);
 334     return size;
 335 }
 336
 337 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 338 {
 339     QEMUFileBdrv *s = opaque;
 340     return bdrv_get_buffer(s->bs, buf, s->base_offset + pos, size);
 341 }
 342
 343 static int bdrv_fclose(void *opaque)
 344 {
 345     QEMUFileBdrv *s = opaque;
 346     qemu_free(s);
 347     return 0;
 348 }
 349
 350 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable)
 351 {
 352     QEMUFileBdrv *s;
 353
 354     s = qemu_mallocz(sizeof(QEMUFileBdrv));
 355
 356     s->bs = bs;
 357     s->base_offset = offset;
 358
 359     if (is_writable)
 360         return qemu_fopen_ops(s, block_put_buffer, NULL, bdrv_fclose, NULL, NULL);
 361
 362     return qemu_fopen_ops(s, NULL, block_get_buffer, bdrv_fclose, NULL, NULL);
 363 }
 364
 365 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 366                          QEMUFileGetBufferFunc *get_buffer,
 367                          QEMUFileCloseFunc *close,
 368                          QEMUFileRateLimit *rate_limit,
 369                          QEMUFileSetRateLimit *set_rate_limit)
 370 {
 371     QEMUFile *f;
 372
 373     f = qemu_mallocz(sizeof(QEMUFile));
 374
 375     f->opaque = opaque;
 376     f->put_buffer = put_buffer;
 377     f->get_buffer = get_buffer;
 378     f->close = close;
 379     f->rate_limit = rate_limit;
 380     f->set_rate_limit = set_rate_limit;
 381     f->is_write = 0;
 382
 383     return f;
 384 }
 385
 386 int qemu_file_has_error(QEMUFile *f)
 387 {
 388     return f->has_error;
 389 }
 390
 391 void qemu_file_set_error(QEMUFile *f)
 392 {
 393     f->has_error = 1;
 394 }
 395
 396 void qemu_fflush(QEMUFile *f)
 397 {
 398     if (!f->put_buffer)
 399         return;
 400
 401     if (f->is_write && f->buf_index > 0) {
 402         int len;
 403
 404         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 405         if (len > 0)
 406             f->buf_offset += f->buf_index;
 407         else
 408             f->has_error = 1;
 409         f->buf_index = 0;
 410     }
 411 }
 412
 413 static void qemu_fill_buffer(QEMUFile *f)
 414 {
 415     int len;
 416
 417     if (!f->get_buffer)
 418         return;
 419
 420     if (f->is_write)
 421         abort();
 422
 423     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 424     if (len > 0) {
 425         f->buf_index = 0;
 426         f->buf_size = len;
 427         f->buf_offset += len;
 428     } else if (len != -EAGAIN)
 429         f->has_error = 1;
 430 }
 431
 432 int qemu_fclose(QEMUFile *f)
 433 {
 434     int ret = 0;
 435     qemu_fflush(f);
 436     if (f->close)
 437         ret = f->close(f->opaque);
 438     qemu_free(f);
 439     return ret;
 440 }
 441
 442 void qemu_file_put_notify(QEMUFile *f)
 443 {
 444     f->put_buffer(f->opaque, NULL, 0, 0);
 445 }
 446
 447 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 448 {
 449     int l;
 450
 451     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 452         fprintf(stderr,
 453                 "Attempted to write to buffer while read buffer is not empty\n");
 454         abort();
 455     }
 456
 457     while (!f->has_error && size > 0) {
 458         l = IO_BUF_SIZE - f->buf_index;
 459         if (l > size)
 460             l = size;
 461         memcpy(f->buf + f->buf_index, buf, l);
 462         f->is_write = 1;
 463         f->buf_index += l;
 464         buf += l;
 465         size -= l;
 466         if (f->buf_index >= IO_BUF_SIZE)
 467             qemu_fflush(f);
 468     }
 469 }
 470
 471 void qemu_put_byte(QEMUFile *f, int v)
 472 {
 473     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 474         fprintf(stderr,
 475                 "Attempted to write to buffer while read buffer is not empty\n");
 476         abort();
 477     }
 478
 479     f->buf[f->buf_index++] = v;
 480     f->is_write = 1;
 481     if (f->buf_index >= IO_BUF_SIZE)
 482         qemu_fflush(f);
 483 }
 484
 485 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 486 {
 487     int size, l;
 488
 489     if (f->is_write)
 490         abort();
 491
 492     size = size1;
 493     while (size > 0) {
 494         l = f->buf_size - f->buf_index;
 495         if (l == 0) {
 496             qemu_fill_buffer(f);
 497             l = f->buf_size - f->buf_index;
 498             if (l == 0)
 499                 break;
 500         }
 501         if (l > size)
 502             l = size;
 503         memcpy(buf, f->buf + f->buf_index, l);
 504         f->buf_index += l;
 505         buf += l;
 506         size -= l;
 507     }
 508     return size1 - size;
 509 }
 510
 511 int qemu_get_byte(QEMUFile *f)
 512 {
 513     if (f->is_write)
 514         abort();
 515
 516     if (f->buf_index >= f->buf_size) {
 517         qemu_fill_buffer(f);
 518         if (f->buf_index >= f->buf_size)
 519             return 0;
 520     }
 521     return f->buf[f->buf_index++];
 522 }
 523
 524 int64_t qemu_ftell(QEMUFile *f)
 525 {
 526     return f->buf_offset - f->buf_size + f->buf_index;
 527 }
 528
 529 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 530 {
 531     if (whence == SEEK_SET) {
 532         /* nothing to do */
 533     } else if (whence == SEEK_CUR) {
 534         pos += qemu_ftell(f);
 535     } else {
 536         /* SEEK_END not supported */
 537         return -1;
 538     }
 539     if (f->put_buffer) {
 540         qemu_fflush(f);
 541         f->buf_offset = pos;
 542     } else {
 543         f->buf_offset = pos;
 544         f->buf_index = 0;
 545         f->buf_size = 0;
 546     }
 547     return pos;
 548 }
 549
 550 int qemu_file_rate_limit(QEMUFile *f)
 551 {
 552     if (f->rate_limit)
 553         return f->rate_limit(f->opaque);
 554
 555     return 0;
 556 }
 557
 558 size_t qemu_file_set_rate_limit(QEMUFile *f, size_t new_rate)
 559 {
 560     if (f->set_rate_limit)
 561         return f->set_rate_limit(f->opaque, new_rate);
 562
 563     return 0;
 564 }
 565
 566 void qemu_put_be16(QEMUFile *f, unsigned int v)
 567 {
 568     qemu_put_byte(f, v >> 8);
 569     qemu_put_byte(f, v);
 570 }
 571
 572 void qemu_put_be32(QEMUFile *f, unsigned int v)
 573 {
 574     qemu_put_byte(f, v >> 24);
 575     qemu_put_byte(f, v >> 16);
 576     qemu_put_byte(f, v >> 8);
 577     qemu_put_byte(f, v);
 578 }
 579
 580 void qemu_put_be64(QEMUFile *f, uint64_t v)
 581 {
 582     qemu_put_be32(f, v >> 32);
 583     qemu_put_be32(f, v);
 584 }
 585
 586 unsigned int qemu_get_be16(QEMUFile *f)
 587 {
 588     unsigned int v;
 589     v = qemu_get_byte(f) << 8;
 590     v |= qemu_get_byte(f);
 591     return v;
 592 }
 593
 594 unsigned int qemu_get_be32(QEMUFile *f)
 595 {
 596     unsigned int v;
 597     v = qemu_get_byte(f) << 24;
 598     v |= qemu_get_byte(f) << 16;
 599     v |= qemu_get_byte(f) << 8;
 600     v |= qemu_get_byte(f);
 601     return v;
 602 }
 603
 604 uint64_t qemu_get_be64(QEMUFile *f)
 605 {
 606     uint64_t v;
 607     v = (uint64_t)qemu_get_be32(f) << 32;
 608     v |= qemu_get_be32(f);
 609     return v;
 610 }
 611
 612 typedef struct SaveStateEntry {
 613     char idstr[256];
 614     int instance_id;
 615     int version_id;
 616     int section_id;
 617     SaveLiveStateHandler *save_live_state;
 618     SaveStateHandler *save_state;
 619     LoadStateHandler *load_state;
 620     void *opaque;
 621     struct SaveStateEntry *next;
 622 } SaveStateEntry;
 623
 624 static SaveStateEntry *first_se;
 625
 626 /* TODO: Individual devices generally have very little idea about the rest
 627    of the system, so instance_id should be removed/replaced.
 628    Meanwhile pass -1 as instance_id if you do not already have a clearly
 629    distinguishing id for all instances of your device class. */
 630 int register_savevm_live(const char *idstr,
 631                          int instance_id,
 632                          int version_id,
 633                          SaveLiveStateHandler *save_live_state,
 634                          SaveStateHandler *save_state,
 635                          LoadStateHandler *load_state,
 636                          void *opaque)
 637 {
 638     SaveStateEntry *se, **pse;
 639     static int global_section_id;
 640
 641     se = qemu_malloc(sizeof(SaveStateEntry));
 642     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 643     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 644     se->version_id = version_id;
 645     se->section_id = global_section_id++;
 646     se->save_live_state = save_live_state;
 647     se->save_state = save_state;
 648     se->load_state = load_state;
 649     se->opaque = opaque;
 650     se->next = NULL;
 651
 652     /* add at the end of list */
 653     pse = &first_se;
 654     while (*pse != NULL) {
 655         if (instance_id == -1
 656                 && strcmp(se->idstr, (*pse)->idstr) == 0
 657                 && se->instance_id <= (*pse)->instance_id)
 658             se->instance_id = (*pse)->instance_id + 1;
 659         pse = &(*pse)->next;
 660     }
 661     *pse = se;
 662     return 0;
 663 }
 664
 665 int register_savevm(const char *idstr,
 666                     int instance_id,
 667                     int version_id,
 668                     SaveStateHandler *save_state,
 669                     LoadStateHandler *load_state,
 670                     void *opaque)
 671 {
 672     return register_savevm_live(idstr, instance_id, version_id,
 673                                 NULL, save_state, load_state, opaque);
 674 }
 675
 676 void unregister_savevm(const char *idstr, void *opaque)
 677 {
 678     SaveStateEntry **pse;
 679
 680     pse = &first_se;
 681     while (*pse != NULL) {
 682         if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
 683             SaveStateEntry *next = (*pse)->next;
 684             qemu_free(*pse);
 685             *pse = next;
 686             continue;
 687         }
 688         pse = &(*pse)->next;
 689     }
 690 }
 691
 692 #define QEMU_VM_FILE_MAGIC           0x5145564d
 693 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 694 #define QEMU_VM_FILE_VERSION         0x00000003
 695
 696 #define QEMU_VM_EOF                  0x00
 697 #define QEMU_VM_SECTION_START        0x01
 698 #define QEMU_VM_SECTION_PART         0x02
 699 #define QEMU_VM_SECTION_END          0x03
 700 #define QEMU_VM_SECTION_FULL         0x04
 701
 702 int qemu_savevm_state_begin(QEMUFile *f)
 703 {
 704     SaveStateEntry *se;
 705
 706     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 707     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 708
 709     for (se = first_se; se != NULL; se = se->next) {
 710         int len;
 711
 712         if (se->save_live_state == NULL)
 713             continue;
 714
 715         /* Section type */
 716         qemu_put_byte(f, QEMU_VM_SECTION_START);
 717         qemu_put_be32(f, se->section_id);
 718
 719         /* ID string */
 720         len = strlen(se->idstr);
 721         qemu_put_byte(f, len);
 722         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 723
 724         qemu_put_be32(f, se->instance_id);
 725         qemu_put_be32(f, se->version_id);
 726
 727         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 728     }
 729
 730     if (qemu_file_has_error(f))
 731         return -EIO;
 732
 733     return 0;
 734 }
 735
 736 int qemu_savevm_state_iterate(QEMUFile *f)
 737 {
 738     SaveStateEntry *se;
 739     int ret = 1;
 740
 741     for (se = first_se; se != NULL; se = se->next) {
 742         if (se->save_live_state == NULL)
 743             continue;
 744
 745         /* Section type */
 746         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 747         qemu_put_be32(f, se->section_id);
 748
 749         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 750     }
 751
 752     if (ret)
 753         return 1;
 754
 755     if (qemu_file_has_error(f))
 756         return -EIO;
 757
 758     return 0;
 759 }
 760
 761 int qemu_savevm_state_complete(QEMUFile *f)
 762 {
 763     SaveStateEntry *se;
 764
 765     for (se = first_se; se != NULL; se = se->next) {
 766         if (se->save_live_state == NULL)
 767             continue;
 768
 769         /* Section type */
 770         qemu_put_byte(f, QEMU_VM_SECTION_END);
 771         qemu_put_be32(f, se->section_id);
 772
 773         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 774     }
 775
 776     for(se = first_se; se != NULL; se = se->next) {
 777         int len;
 778
 779         if (se->save_state == NULL)
 780             continue;
 781
 782         /* Section type */
 783         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 784         qemu_put_be32(f, se->section_id);
 785
 786         /* ID string */
 787         len = strlen(se->idstr);
 788         qemu_put_byte(f, len);
 789         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 790
 791         qemu_put_be32(f, se->instance_id);
 792         qemu_put_be32(f, se->version_id);
 793
 794         se->save_state(f, se->opaque);
 795     }
 796
 797     qemu_put_byte(f, QEMU_VM_EOF);
 798
 799     if (qemu_file_has_error(f))
 800         return -EIO;
 801
 802     return 0;
 803 }
 804
 805 int qemu_savevm_state(QEMUFile *f)
 806 {
 807     int saved_vm_running;
 808     int ret;
 809
 810     saved_vm_running = vm_running;
 811     vm_stop(0);
 812
 813     bdrv_flush_all();
 814
 815     ret = qemu_savevm_state_begin(f);
 816     if (ret < 0)
 817         goto out;
 818
 819     do {
 820         ret = qemu_savevm_state_iterate(f);
 821         if (ret < 0)
 822             goto out;
 823     } while (ret == 0);
 824
 825     ret = qemu_savevm_state_complete(f);
 826
 827 out:
 828     if (qemu_file_has_error(f))
 829         ret = -EIO;
 830
 831     if (!ret && saved_vm_running)
 832         vm_start();
 833
 834     return ret;
 835 }
 836
 837 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 838 {
 839     SaveStateEntry *se;
 840
 841     for(se = first_se; se != NULL; se = se->next) {
 842         if (!strcmp(se->idstr, idstr) &&
 843             instance_id == se->instance_id)
 844             return se;
 845     }
 846     return NULL;
 847 }
 848
 849 typedef struct LoadStateEntry {
 850     SaveStateEntry *se;
 851     int section_id;
 852     int version_id;
 853     struct LoadStateEntry *next;
 854 } LoadStateEntry;
 855
 856 static int qemu_loadvm_state_v2(QEMUFile *f)
 857 {
 858     SaveStateEntry *se;
 859     int len, ret, instance_id, record_len, version_id;
 860     int64_t total_len, end_pos, cur_pos;
 861     char idstr[256];
 862
 863     total_len = qemu_get_be64(f);
 864     end_pos = total_len + qemu_ftell(f);
 865     for(;;) {
 866         if (qemu_ftell(f) >= end_pos)
 867             break;
 868         len = qemu_get_byte(f);
 869         qemu_get_buffer(f, (uint8_t *)idstr, len);
 870         idstr[len] = '\0';
 871         instance_id = qemu_get_be32(f);
 872         version_id = qemu_get_be32(f);
 873         record_len = qemu_get_be32(f);
 874         cur_pos = qemu_ftell(f);
 875         se = find_se(idstr, instance_id);
 876         if (!se) {
 877             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 878                     instance_id, idstr);
 879         } else {
 880             ret = se->load_state(f, se->opaque, version_id);
 881             if (ret < 0) {
 882                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 883                         instance_id, idstr);
 884                 return ret;
 885             }
 886         }
 887         /* always seek to exact end of record */
 888         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 889     }
 890
 891     if (qemu_file_has_error(f))
 892         return -EIO;
 893
 894     return 0;
 895 }
 896
 897 int qemu_loadvm_state(QEMUFile *f)
 898 {
 899     LoadStateEntry *first_le = NULL;
 900     uint8_t section_type;
 901     unsigned int v;
 902     int ret;
 903
 904     v = qemu_get_be32(f);
 905     if (v != QEMU_VM_FILE_MAGIC)
 906         return -EINVAL;
 907
 908     v = qemu_get_be32(f);
 909     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 910         return qemu_loadvm_state_v2(f);
 911     if (v != QEMU_VM_FILE_VERSION)
 912         return -ENOTSUP;
 913
 914     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 915         uint32_t instance_id, version_id, section_id;
 916         LoadStateEntry *le;
 917         SaveStateEntry *se;
 918         char idstr[257];
 919         int len;
 920
 921         switch (section_type) {
 922         case QEMU_VM_SECTION_START:
 923         case QEMU_VM_SECTION_FULL:
 924             /* Read section start */
 925             section_id = qemu_get_be32(f);
 926             len = qemu_get_byte(f);
 927             qemu_get_buffer(f, (uint8_t *)idstr, len);
 928             idstr[len] = 0;
 929             instance_id = qemu_get_be32(f);
 930             version_id = qemu_get_be32(f);
 931
 932             /* Find savevm section */
 933             se = find_se(idstr, instance_id);
 934             if (se == NULL) {
 935                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 936                 ret = -EINVAL;
 937                 goto out;
 938             }
 939
 940             /* Validate version */
 941             if (version_id > se->version_id) {
 942                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 943                         version_id, idstr, se->version_id);
 944                 ret = -EINVAL;
 945                 goto out;
 946             }
 947
 948             /* Add entry */
 949             le = qemu_mallocz(sizeof(*le));
 950
 951             le->se = se;
 952             le->section_id = section_id;
 953             le->version_id = version_id;
 954             le->next = first_le;
 955             first_le = le;
 956
 957             le->se->load_state(f, le->se->opaque, le->version_id);
 958             break;
 959         case QEMU_VM_SECTION_PART:
 960         case QEMU_VM_SECTION_END:
 961             section_id = qemu_get_be32(f);
 962
 963             for (le = first_le; le && le->section_id != section_id; le = le->next);
 964             if (le == NULL) {
 965                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 966                 ret = -EINVAL;
 967                 goto out;
 968             }
 969
 970             le->se->load_state(f, le->se->opaque, le->version_id);
 971             break;
 972         default:
 973             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 974             ret = -EINVAL;
 975             goto out;
 976         }
 977     }
 978
 979     ret = 0;
 980
 981 out:
 982     while (first_le) {
 983         LoadStateEntry *le = first_le;
 984         first_le = first_le->next;
 985         qemu_free(le);
 986     }
 987
 988     if (qemu_file_has_error(f))
 989         ret = -EIO;
 990
 991     return ret;
 992 }
 993
 994 /* device can contain snapshots */
 995 static int bdrv_can_snapshot(BlockDriverState *bs)
 996 {
 997     return (bs &&
 998             !bdrv_is_removable(bs) &&
 999             !bdrv_is_read_only(bs));
1000 }
1001
1002 /* device must be snapshots in order to have a reliable snapshot */
1003 static int bdrv_has_snapshot(BlockDriverState *bs)
1004 {
1005     return (bs &&
1006             !bdrv_is_removable(bs) &&
1007             !bdrv_is_read_only(bs));
1008 }
1009
1010 static BlockDriverState *get_bs_snapshots(void)
1011 {
1012     BlockDriverState *bs;
1013     int i;
1014
1015     if (bs_snapshots)
1016         return bs_snapshots;
1017     for(i = 0; i <= nb_drives; i++) {
1018         bs = drives_table[i].bdrv;
1019         if (bdrv_can_snapshot(bs))
1020             goto ok;
1021     }
1022     return NULL;
1023  ok:
1024     bs_snapshots = bs;
1025     return bs;
1026 }
1027
1028 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1029                               const char *name)
1030 {
1031     QEMUSnapshotInfo *sn_tab, *sn;
1032     int nb_sns, i, ret;
1033
1034     ret = -ENOENT;
1035     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1036     if (nb_sns < 0)
1037         return ret;
1038     for(i = 0; i < nb_sns; i++) {
1039         sn = &sn_tab[i];
1040         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1041             *sn_info = *sn;
1042             ret = 0;
1043             break;
1044         }
1045     }
1046     qemu_free(sn_tab);
1047     return ret;
1048 }
1049
1050 void do_savevm(Monitor *mon, const char *name)
1051 {
1052     BlockDriverState *bs, *bs1;
1053     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1054     int must_delete, ret, i;
1055     BlockDriverInfo bdi1, *bdi = &bdi1;
1056     QEMUFile *f;
1057     int saved_vm_running;
1058     uint32_t vm_state_size;
1059 #ifdef _WIN32
1060     struct _timeb tb;
1061 #else
1062     struct timeval tv;
1063 #endif
1064
1065     bs = get_bs_snapshots();
1066     if (!bs) {
1067         monitor_printf(mon, "No block device can accept snapshots\n");
1068         return;
1069     }
1070
1071     /* ??? Should this occur after vm_stop?  */
1072     qemu_aio_flush();
1073
1074     saved_vm_running = vm_running;
1075     vm_stop(0);
1076
1077     must_delete = 0;
1078     if (name) {
1079         ret = bdrv_snapshot_find(bs, old_sn, name);
1080         if (ret >= 0) {
1081             must_delete = 1;
1082         }
1083     }
1084     memset(sn, 0, sizeof(*sn));
1085     if (must_delete) {
1086         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1087         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1088     } else {
1089         if (name)
1090             pstrcpy(sn->name, sizeof(sn->name), name);
1091     }
1092
1093     /* fill auxiliary fields */
1094 #ifdef _WIN32
1095     _ftime(&tb);
1096     sn->date_sec = tb.time;
1097     sn->date_nsec = tb.millitm * 1000000;
1098 #else
1099     gettimeofday(&tv, NULL);
1100     sn->date_sec = tv.tv_sec;
1101     sn->date_nsec = tv.tv_usec * 1000;
1102 #endif
1103     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1104
1105     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1106         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1107                        bdrv_get_device_name(bs));
1108         goto the_end;
1109     }
1110
1111     /* save the VM state */
1112     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 1);
1113     if (!f) {
1114         monitor_printf(mon, "Could not open VM state file\n");
1115         goto the_end;
1116     }
1117     ret = qemu_savevm_state(f);
1118     vm_state_size = qemu_ftell(f);
1119     qemu_fclose(f);
1120     if (ret < 0) {
1121         monitor_printf(mon, "Error %d while writing VM\n", ret);
1122         goto the_end;
1123     }
1124
1125     /* create the snapshots */
1126
1127     for(i = 0; i < nb_drives; i++) {
1128         bs1 = drives_table[i].bdrv;
1129         if (bdrv_has_snapshot(bs1)) {
1130             if (must_delete) {
1131                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1132                 if (ret < 0) {
1133                     monitor_printf(mon,
1134                                    "Error while deleting snapshot on '%s'\n",
1135                                    bdrv_get_device_name(bs1));
1136                 }
1137             }
1138             /* Write VM state size only to the image that contains the state */
1139             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1140             ret = bdrv_snapshot_create(bs1, sn);
1141             if (ret < 0) {
1142                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1143                                bdrv_get_device_name(bs1));
1144             }
1145         }
1146     }
1147
1148  the_end:
1149     if (saved_vm_running)
1150         vm_start();
1151 }
1152
1153 void do_loadvm(Monitor *mon, const char *name)
1154 {
1155     BlockDriverState *bs, *bs1;
1156     BlockDriverInfo bdi1, *bdi = &bdi1;
1157     QEMUSnapshotInfo sn;
1158     QEMUFile *f;
1159     int i, ret;
1160     int saved_vm_running;
1161
1162     bs = get_bs_snapshots();
1163     if (!bs) {
1164         monitor_printf(mon, "No block device supports snapshots\n");
1165         return;
1166     }
1167
1168     /* Flush all IO requests so they don't interfere with the new state.  */
1169     qemu_aio_flush();
1170
1171     saved_vm_running = vm_running;
1172     vm_stop(0);
1173
1174     for(i = 0; i <= nb_drives; i++) {
1175         bs1 = drives_table[i].bdrv;
1176         if (bdrv_has_snapshot(bs1)) {
1177             ret = bdrv_snapshot_goto(bs1, name);
1178             if (ret < 0) {
1179                 if (bs != bs1)
1180                     monitor_printf(mon, "Warning: ");
1181                 switch(ret) {
1182                 case -ENOTSUP:
1183                     monitor_printf(mon,
1184                                    "Snapshots not supported on device '%s'\n",
1185                                    bdrv_get_device_name(bs1));
1186                     break;
1187                 case -ENOENT:
1188                     monitor_printf(mon, "Could not find snapshot '%s' on "
1189                                    "device '%s'\n",
1190                                    name, bdrv_get_device_name(bs1));
1191                     break;
1192                 default:
1193                     monitor_printf(mon, "Error %d while activating snapshot on"
1194                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1195                     break;
1196                 }
1197                 /* fatal on snapshot block device */
1198                 if (bs == bs1)
1199                     goto the_end;
1200             }
1201         }
1202     }
1203
1204     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1205         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1206                        bdrv_get_device_name(bs));
1207         return;
1208     }
1209
1210     /* Don't even try to load empty VM states */
1211     ret = bdrv_snapshot_find(bs, &sn, name);
1212     if ((ret >= 0) && (sn.vm_state_size == 0))
1213         goto the_end;
1214
1215     /* restore the VM state */
1216     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 0);
1217     if (!f) {
1218         monitor_printf(mon, "Could not open VM state file\n");
1219         goto the_end;
1220     }
1221     ret = qemu_loadvm_state(f);
1222     qemu_fclose(f);
1223     if (ret < 0) {
1224         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1225     }
1226  the_end:
1227     if (saved_vm_running)
1228         vm_start();
1229 }
1230
1231 void do_delvm(Monitor *mon, const char *name)
1232 {
1233     BlockDriverState *bs, *bs1;
1234     int i, ret;
1235
1236     bs = get_bs_snapshots();
1237     if (!bs) {
1238         monitor_printf(mon, "No block device supports snapshots\n");
1239         return;
1240     }
1241
1242     for(i = 0; i <= nb_drives; i++) {
1243         bs1 = drives_table[i].bdrv;
1244         if (bdrv_has_snapshot(bs1)) {
1245             ret = bdrv_snapshot_delete(bs1, name);
1246             if (ret < 0) {
1247                 if (ret == -ENOTSUP)
1248                     monitor_printf(mon,
1249                                    "Snapshots not supported on device '%s'\n",
1250                                    bdrv_get_device_name(bs1));
1251                 else
1252                     monitor_printf(mon, "Error %d while deleting snapshot on "
1253                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1254             }
1255         }
1256     }
1257 }
1258
1259 void do_info_snapshots(Monitor *mon)
1260 {
1261     BlockDriverState *bs, *bs1;
1262     QEMUSnapshotInfo *sn_tab, *sn;
1263     int nb_sns, i;
1264     char buf[256];
1265
1266     bs = get_bs_snapshots();
1267     if (!bs) {
1268         monitor_printf(mon, "No available block device supports snapshots\n");
1269         return;
1270     }
1271     monitor_printf(mon, "Snapshot devices:");
1272     for(i = 0; i <= nb_drives; i++) {
1273         bs1 = drives_table[i].bdrv;
1274         if (bdrv_has_snapshot(bs1)) {
1275             if (bs == bs1)
1276                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1277         }
1278     }
1279     monitor_printf(mon, "\n");
1280
1281     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1282     if (nb_sns < 0) {
1283         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1284         return;
1285     }
1286     monitor_printf(mon, "Snapshot list (from %s):\n",
1287                    bdrv_get_device_name(bs));
1288     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1289     for(i = 0; i < nb_sns; i++) {
1290         sn = &sn_tab[i];
1291         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1292     }
1293     qemu_free(sn_tab);
1294 }