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 QEMUFileStdio
 176 {
 177     FILE *stdio_file;
 178     QEMUFile *file;
 179 } QEMUFileStdio;
 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 stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 210 {
 211     QEMUFileStdio *s = opaque;
 212     return fwrite(buf, 1, size, s->stdio_file);
 213 }
 214
 215 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 216 {
 217     QEMUFileStdio *s = opaque;
 218     FILE *fp = s->stdio_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 stdio_pclose(void *opaque)
 229 {
 230     QEMUFileStdio *s = opaque;
 231     pclose(s->stdio_file);
 232     qemu_free(s);
 233     return 0;
 234 }
 235
 236 static int stdio_fclose(void *opaque)
 237 {
 238     QEMUFileStdio *s = opaque;
 239     fclose(s->stdio_file);
 240     qemu_free(s);
 241     return 0;
 242 }
 243
 244 QEMUFile *qemu_popen(FILE *stdio_file, const char *mode)
 245 {
 246     QEMUFileStdio *s;
 247
 248     if (stdio_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 249         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 250         return NULL;
 251     }
 252
 253     s = qemu_mallocz(sizeof(QEMUFileStdio));
 254
 255     s->stdio_file = stdio_file;
 256
 257     if(mode[0] == 'r') {
 258         s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_pclose, NULL, NULL);
 259     } else {
 260         s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_pclose, NULL, NULL);
 261     }
 262     return s->file;
 263 }
 264
 265 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 266 {
 267     FILE *popen_file;
 268
 269     popen_file = popen(command, mode);
 270     if(popen_file == NULL) {
 271         return NULL;
 272     }
 273
 274     return qemu_popen(popen_file, mode);
 275 }
 276
 277 int qemu_stdio_fd(QEMUFile *f)
 278 {
 279     QEMUFileStdio *p;
 280     int fd;
 281
 282     p = (QEMUFileStdio *)f->opaque;
 283     fd = fileno(p->stdio_file);
 284
 285     return fd;
 286 }
 287
 288 QEMUFile *qemu_fopen_socket(int fd)
 289 {
 290     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 291
 292     s->fd = fd;
 293     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL, NULL);
 294     return s->file;
 295 }
 296
 297 static int file_put_buffer(void *opaque, const uint8_t *buf,
 298                             int64_t pos, int size)
 299 {
 300     QEMUFileStdio *s = opaque;
 301     fseek(s->stdio_file, pos, SEEK_SET);
 302     fwrite(buf, 1, size, s->stdio_file);
 303     return size;
 304 }
 305
 306 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 307 {
 308     QEMUFileStdio *s = opaque;
 309     fseek(s->stdio_file, pos, SEEK_SET);
 310     return fread(buf, 1, size, s->stdio_file);
 311 }
 312
 313 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 314 {
 315     QEMUFileStdio *s;
 316
 317     if (mode == NULL ||
 318         (mode[0] != 'r' && mode[0] != 'w') ||
 319         mode[1] != 'b' || mode[2] != 0) {
 320         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 321         return NULL;
 322     }
 323
 324     s = qemu_mallocz(sizeof(QEMUFileStdio));
 325
 326     s->stdio_file = fopen(filename, mode);
 327     if (!s->stdio_file)
 328         goto fail;
 329
 330     if(mode[0] == 'w') {
 331         s->file = qemu_fopen_ops(s, file_put_buffer, NULL, stdio_fclose, NULL, NULL);
 332     } else {
 333         s->file = qemu_fopen_ops(s, NULL, file_get_buffer, stdio_fclose, NULL, NULL);
 334     }
 335     return s->file;
 336 fail:
 337     qemu_free(s);
 338     return NULL;
 339 }
 340
 341 static int block_put_buffer(void *opaque, const uint8_t *buf,
 342                            int64_t pos, int size)
 343 {
 344     bdrv_save_vmstate(opaque, buf, pos, size);
 345     return size;
 346 }
 347
 348 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 349 {
 350     return bdrv_load_vmstate(opaque, buf, pos, size);
 351 }
 352
 353 static int bdrv_fclose(void *opaque)
 354 {
 355     return 0;
 356 }
 357
 358 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 359 {
 360     if (is_writable)
 361         return qemu_fopen_ops(bs, block_put_buffer, NULL, bdrv_fclose, NULL, NULL);
 362     return qemu_fopen_ops(bs, 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     /* any failed or completed migration keeps its state to allow probing of
 561      * migration data, but has no associated file anymore */
 562     if (f && f->set_rate_limit)
 563         return f->set_rate_limit(f->opaque, new_rate);
 564
 565     return 0;
 566 }
 567
 568 void qemu_put_be16(QEMUFile *f, unsigned int v)
 569 {
 570     qemu_put_byte(f, v >> 8);
 571     qemu_put_byte(f, v);
 572 }
 573
 574 void qemu_put_be32(QEMUFile *f, unsigned int v)
 575 {
 576     qemu_put_byte(f, v >> 24);
 577     qemu_put_byte(f, v >> 16);
 578     qemu_put_byte(f, v >> 8);
 579     qemu_put_byte(f, v);
 580 }
 581
 582 void qemu_put_be64(QEMUFile *f, uint64_t v)
 583 {
 584     qemu_put_be32(f, v >> 32);
 585     qemu_put_be32(f, v);
 586 }
 587
 588 unsigned int qemu_get_be16(QEMUFile *f)
 589 {
 590     unsigned int v;
 591     v = qemu_get_byte(f) << 8;
 592     v |= qemu_get_byte(f);
 593     return v;
 594 }
 595
 596 unsigned int qemu_get_be32(QEMUFile *f)
 597 {
 598     unsigned int v;
 599     v = qemu_get_byte(f) << 24;
 600     v |= qemu_get_byte(f) << 16;
 601     v |= qemu_get_byte(f) << 8;
 602     v |= qemu_get_byte(f);
 603     return v;
 604 }
 605
 606 uint64_t qemu_get_be64(QEMUFile *f)
 607 {
 608     uint64_t v;
 609     v = (uint64_t)qemu_get_be32(f) << 32;
 610     v |= qemu_get_be32(f);
 611     return v;
 612 }
 613
 614 typedef struct SaveStateEntry {
 615     char idstr[256];
 616     int instance_id;
 617     int version_id;
 618     int section_id;
 619     SaveLiveStateHandler *save_live_state;
 620     SaveStateHandler *save_state;
 621     LoadStateHandler *load_state;
 622     void *opaque;
 623     struct SaveStateEntry *next;
 624 } SaveStateEntry;
 625
 626 static SaveStateEntry *first_se;
 627
 628 /* TODO: Individual devices generally have very little idea about the rest
 629    of the system, so instance_id should be removed/replaced.
 630    Meanwhile pass -1 as instance_id if you do not already have a clearly
 631    distinguishing id for all instances of your device class. */
 632 int register_savevm_live(const char *idstr,
 633                          int instance_id,
 634                          int version_id,
 635                          SaveLiveStateHandler *save_live_state,
 636                          SaveStateHandler *save_state,
 637                          LoadStateHandler *load_state,
 638                          void *opaque)
 639 {
 640     SaveStateEntry *se, **pse;
 641     static int global_section_id;
 642
 643     se = qemu_malloc(sizeof(SaveStateEntry));
 644     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 645     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 646     se->version_id = version_id;
 647     se->section_id = global_section_id++;
 648     se->save_live_state = save_live_state;
 649     se->save_state = save_state;
 650     se->load_state = load_state;
 651     se->opaque = opaque;
 652     se->next = NULL;
 653
 654     /* add at the end of list */
 655     pse = &first_se;
 656     while (*pse != NULL) {
 657         if (instance_id == -1
 658                 && strcmp(se->idstr, (*pse)->idstr) == 0
 659                 && se->instance_id <= (*pse)->instance_id)
 660             se->instance_id = (*pse)->instance_id + 1;
 661         pse = &(*pse)->next;
 662     }
 663     *pse = se;
 664     return 0;
 665 }
 666
 667 int register_savevm(const char *idstr,
 668                     int instance_id,
 669                     int version_id,
 670                     SaveStateHandler *save_state,
 671                     LoadStateHandler *load_state,
 672                     void *opaque)
 673 {
 674     return register_savevm_live(idstr, instance_id, version_id,
 675                                 NULL, save_state, load_state, opaque);
 676 }
 677
 678 void unregister_savevm(const char *idstr, void *opaque)
 679 {
 680     SaveStateEntry **pse;
 681
 682     pse = &first_se;
 683     while (*pse != NULL) {
 684         if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
 685             SaveStateEntry *next = (*pse)->next;
 686             qemu_free(*pse);
 687             *pse = next;
 688             continue;
 689         }
 690         pse = &(*pse)->next;
 691     }
 692 }
 693
 694 #define QEMU_VM_FILE_MAGIC           0x5145564d
 695 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 696 #define QEMU_VM_FILE_VERSION         0x00000003
 697
 698 #define QEMU_VM_EOF                  0x00
 699 #define QEMU_VM_SECTION_START        0x01
 700 #define QEMU_VM_SECTION_PART         0x02
 701 #define QEMU_VM_SECTION_END          0x03
 702 #define QEMU_VM_SECTION_FULL         0x04
 703
 704 int qemu_savevm_state_begin(QEMUFile *f)
 705 {
 706     SaveStateEntry *se;
 707
 708     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 709     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 710
 711     for (se = first_se; se != NULL; se = se->next) {
 712         int len;
 713
 714         if (se->save_live_state == NULL)
 715             continue;
 716
 717         /* Section type */
 718         qemu_put_byte(f, QEMU_VM_SECTION_START);
 719         qemu_put_be32(f, se->section_id);
 720
 721         /* ID string */
 722         len = strlen(se->idstr);
 723         qemu_put_byte(f, len);
 724         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 725
 726         qemu_put_be32(f, se->instance_id);
 727         qemu_put_be32(f, se->version_id);
 728
 729         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 730     }
 731
 732     if (qemu_file_has_error(f))
 733         return -EIO;
 734
 735     return 0;
 736 }
 737
 738 int qemu_savevm_state_iterate(QEMUFile *f)
 739 {
 740     SaveStateEntry *se;
 741     int ret = 1;
 742
 743     for (se = first_se; se != NULL; se = se->next) {
 744         if (se->save_live_state == NULL)
 745             continue;
 746
 747         /* Section type */
 748         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 749         qemu_put_be32(f, se->section_id);
 750
 751         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 752     }
 753
 754     if (ret)
 755         return 1;
 756
 757     if (qemu_file_has_error(f))
 758         return -EIO;
 759
 760     return 0;
 761 }
 762
 763 int qemu_savevm_state_complete(QEMUFile *f)
 764 {
 765     SaveStateEntry *se;
 766
 767     for (se = first_se; se != NULL; se = se->next) {
 768         if (se->save_live_state == NULL)
 769             continue;
 770
 771         /* Section type */
 772         qemu_put_byte(f, QEMU_VM_SECTION_END);
 773         qemu_put_be32(f, se->section_id);
 774
 775         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 776     }
 777
 778     for(se = first_se; se != NULL; se = se->next) {
 779         int len;
 780
 781         if (se->save_state == NULL)
 782             continue;
 783
 784         /* Section type */
 785         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 786         qemu_put_be32(f, se->section_id);
 787
 788         /* ID string */
 789         len = strlen(se->idstr);
 790         qemu_put_byte(f, len);
 791         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 792
 793         qemu_put_be32(f, se->instance_id);
 794         qemu_put_be32(f, se->version_id);
 795
 796         se->save_state(f, se->opaque);
 797     }
 798
 799     qemu_put_byte(f, QEMU_VM_EOF);
 800
 801     if (qemu_file_has_error(f))
 802         return -EIO;
 803
 804     return 0;
 805 }
 806
 807 int qemu_savevm_state(QEMUFile *f)
 808 {
 809     int saved_vm_running;
 810     int ret;
 811
 812     saved_vm_running = vm_running;
 813     vm_stop(0);
 814
 815     bdrv_flush_all();
 816
 817     ret = qemu_savevm_state_begin(f);
 818     if (ret < 0)
 819         goto out;
 820
 821     do {
 822         ret = qemu_savevm_state_iterate(f);
 823         if (ret < 0)
 824             goto out;
 825     } while (ret == 0);
 826
 827     ret = qemu_savevm_state_complete(f);
 828
 829 out:
 830     if (qemu_file_has_error(f))
 831         ret = -EIO;
 832
 833     if (!ret && saved_vm_running)
 834         vm_start();
 835
 836     return ret;
 837 }
 838
 839 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 840 {
 841     SaveStateEntry *se;
 842
 843     for(se = first_se; se != NULL; se = se->next) {
 844         if (!strcmp(se->idstr, idstr) &&
 845             instance_id == se->instance_id)
 846             return se;
 847     }
 848     return NULL;
 849 }
 850
 851 typedef struct LoadStateEntry {
 852     SaveStateEntry *se;
 853     int section_id;
 854     int version_id;
 855     struct LoadStateEntry *next;
 856 } LoadStateEntry;
 857
 858 static int qemu_loadvm_state_v2(QEMUFile *f)
 859 {
 860     SaveStateEntry *se;
 861     int len, ret, instance_id, record_len, version_id;
 862     int64_t total_len, end_pos, cur_pos;
 863     char idstr[256];
 864
 865     total_len = qemu_get_be64(f);
 866     end_pos = total_len + qemu_ftell(f);
 867     for(;;) {
 868         if (qemu_ftell(f) >= end_pos)
 869             break;
 870         len = qemu_get_byte(f);
 871         qemu_get_buffer(f, (uint8_t *)idstr, len);
 872         idstr[len] = '\0';
 873         instance_id = qemu_get_be32(f);
 874         version_id = qemu_get_be32(f);
 875         record_len = qemu_get_be32(f);
 876         cur_pos = qemu_ftell(f);
 877         se = find_se(idstr, instance_id);
 878         if (!se) {
 879             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 880                     instance_id, idstr);
 881         } else {
 882             ret = se->load_state(f, se->opaque, version_id);
 883             if (ret < 0) {
 884                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 885                         instance_id, idstr);
 886                 return ret;
 887             }
 888         }
 889         /* always seek to exact end of record */
 890         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 891     }
 892
 893     if (qemu_file_has_error(f))
 894         return -EIO;
 895
 896     return 0;
 897 }
 898
 899 int qemu_loadvm_state(QEMUFile *f)
 900 {
 901     LoadStateEntry *first_le = NULL;
 902     uint8_t section_type;
 903     unsigned int v;
 904     int ret;
 905
 906     v = qemu_get_be32(f);
 907     if (v != QEMU_VM_FILE_MAGIC)
 908         return -EINVAL;
 909
 910     v = qemu_get_be32(f);
 911     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 912         return qemu_loadvm_state_v2(f);
 913     if (v != QEMU_VM_FILE_VERSION)
 914         return -ENOTSUP;
 915
 916     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 917         uint32_t instance_id, version_id, section_id;
 918         LoadStateEntry *le;
 919         SaveStateEntry *se;
 920         char idstr[257];
 921         int len;
 922
 923         switch (section_type) {
 924         case QEMU_VM_SECTION_START:
 925         case QEMU_VM_SECTION_FULL:
 926             /* Read section start */
 927             section_id = qemu_get_be32(f);
 928             len = qemu_get_byte(f);
 929             qemu_get_buffer(f, (uint8_t *)idstr, len);
 930             idstr[len] = 0;
 931             instance_id = qemu_get_be32(f);
 932             version_id = qemu_get_be32(f);
 933
 934             /* Find savevm section */
 935             se = find_se(idstr, instance_id);
 936             if (se == NULL) {
 937                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 938                 ret = -EINVAL;
 939                 goto out;
 940             }
 941
 942             /* Validate version */
 943             if (version_id > se->version_id) {
 944                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 945                         version_id, idstr, se->version_id);
 946                 ret = -EINVAL;
 947                 goto out;
 948             }
 949
 950             /* Add entry */
 951             le = qemu_mallocz(sizeof(*le));
 952
 953             le->se = se;
 954             le->section_id = section_id;
 955             le->version_id = version_id;
 956             le->next = first_le;
 957             first_le = le;
 958
 959             le->se->load_state(f, le->se->opaque, le->version_id);
 960             break;
 961         case QEMU_VM_SECTION_PART:
 962         case QEMU_VM_SECTION_END:
 963             section_id = qemu_get_be32(f);
 964
 965             for (le = first_le; le && le->section_id != section_id; le = le->next);
 966             if (le == NULL) {
 967                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 968                 ret = -EINVAL;
 969                 goto out;
 970             }
 971
 972             le->se->load_state(f, le->se->opaque, le->version_id);
 973             break;
 974         default:
 975             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 976             ret = -EINVAL;
 977             goto out;
 978         }
 979     }
 980
 981     ret = 0;
 982
 983 out:
 984     while (first_le) {
 985         LoadStateEntry *le = first_le;
 986         first_le = first_le->next;
 987         qemu_free(le);
 988     }
 989
 990     if (qemu_file_has_error(f))
 991         ret = -EIO;
 992
 993     return ret;
 994 }
 995
 996 /* device can contain snapshots */
 997 static int bdrv_can_snapshot(BlockDriverState *bs)
 998 {
 999     return (bs &&
1000             !bdrv_is_removable(bs) &&
1001             !bdrv_is_read_only(bs));
1002 }
1003
1004 /* device must be snapshots in order to have a reliable snapshot */
1005 static int bdrv_has_snapshot(BlockDriverState *bs)
1006 {
1007     return (bs &&
1008             !bdrv_is_removable(bs) &&
1009             !bdrv_is_read_only(bs));
1010 }
1011
1012 static BlockDriverState *get_bs_snapshots(void)
1013 {
1014     BlockDriverState *bs;
1015     DriveInfo *dinfo;
1016
1017     if (bs_snapshots)
1018         return bs_snapshots;
1019     TAILQ_FOREACH(dinfo, &drives, next) {
1020         bs = dinfo->bdrv;
1021         if (bdrv_can_snapshot(bs))
1022             goto ok;
1023     }
1024     return NULL;
1025  ok:
1026     bs_snapshots = bs;
1027     return bs;
1028 }
1029
1030 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1031                               const char *name)
1032 {
1033     QEMUSnapshotInfo *sn_tab, *sn;
1034     int nb_sns, i, ret;
1035
1036     ret = -ENOENT;
1037     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1038     if (nb_sns < 0)
1039         return ret;
1040     for(i = 0; i < nb_sns; i++) {
1041         sn = &sn_tab[i];
1042         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1043             *sn_info = *sn;
1044             ret = 0;
1045             break;
1046         }
1047     }
1048     qemu_free(sn_tab);
1049     return ret;
1050 }
1051
1052 void do_savevm(Monitor *mon, const char *name)
1053 {
1054     DriveInfo *dinfo;
1055     BlockDriverState *bs, *bs1;
1056     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1057     int must_delete, ret;
1058     QEMUFile *f;
1059     int saved_vm_running;
1060     uint32_t vm_state_size;
1061 #ifdef _WIN32
1062     struct _timeb tb;
1063 #else
1064     struct timeval tv;
1065 #endif
1066
1067     bs = get_bs_snapshots();
1068     if (!bs) {
1069         monitor_printf(mon, "No block device can accept snapshots\n");
1070         return;
1071     }
1072
1073     /* ??? Should this occur after vm_stop?  */
1074     qemu_aio_flush();
1075
1076     saved_vm_running = vm_running;
1077     vm_stop(0);
1078
1079     must_delete = 0;
1080     if (name) {
1081         ret = bdrv_snapshot_find(bs, old_sn, name);
1082         if (ret >= 0) {
1083             must_delete = 1;
1084         }
1085     }
1086     memset(sn, 0, sizeof(*sn));
1087     if (must_delete) {
1088         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1089         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1090     } else {
1091         if (name)
1092             pstrcpy(sn->name, sizeof(sn->name), name);
1093     }
1094
1095     /* fill auxiliary fields */
1096 #ifdef _WIN32
1097     _ftime(&tb);
1098     sn->date_sec = tb.time;
1099     sn->date_nsec = tb.millitm * 1000000;
1100 #else
1101     gettimeofday(&tv, NULL);
1102     sn->date_sec = tv.tv_sec;
1103     sn->date_nsec = tv.tv_usec * 1000;
1104 #endif
1105     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1106
1107     /* save the VM state */
1108     f = qemu_fopen_bdrv(bs, 1);
1109     if (!f) {
1110         monitor_printf(mon, "Could not open VM state file\n");
1111         goto the_end;
1112     }
1113     ret = qemu_savevm_state(f);
1114     vm_state_size = qemu_ftell(f);
1115     qemu_fclose(f);
1116     if (ret < 0) {
1117         monitor_printf(mon, "Error %d while writing VM\n", ret);
1118         goto the_end;
1119     }
1120
1121     /* create the snapshots */
1122
1123     TAILQ_FOREACH(dinfo, &drives, next) {
1124         bs1 = dinfo->bdrv;
1125         if (bdrv_has_snapshot(bs1)) {
1126             if (must_delete) {
1127                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1128                 if (ret < 0) {
1129                     monitor_printf(mon,
1130                                    "Error while deleting snapshot on '%s'\n",
1131                                    bdrv_get_device_name(bs1));
1132                 }
1133             }
1134             /* Write VM state size only to the image that contains the state */
1135             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1136             ret = bdrv_snapshot_create(bs1, sn);
1137             if (ret < 0) {
1138                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1139                                bdrv_get_device_name(bs1));
1140             }
1141         }
1142     }
1143
1144  the_end:
1145     if (saved_vm_running)
1146         vm_start();
1147 }
1148
1149 void do_loadvm(Monitor *mon, const char *name)
1150 {
1151     DriveInfo *dinfo;
1152     BlockDriverState *bs, *bs1;
1153     QEMUSnapshotInfo sn;
1154     QEMUFile *f;
1155     int ret;
1156     int saved_vm_running;
1157
1158     bs = get_bs_snapshots();
1159     if (!bs) {
1160         monitor_printf(mon, "No block device supports snapshots\n");
1161         return;
1162     }
1163
1164     /* Flush all IO requests so they don't interfere with the new state.  */
1165     qemu_aio_flush();
1166
1167     saved_vm_running = vm_running;
1168     vm_stop(0);
1169
1170     TAILQ_FOREACH(dinfo, &drives, next) {
1171         bs1 = dinfo->bdrv;
1172         if (bdrv_has_snapshot(bs1)) {
1173             ret = bdrv_snapshot_goto(bs1, name);
1174             if (ret < 0) {
1175                 if (bs != bs1)
1176                     monitor_printf(mon, "Warning: ");
1177                 switch(ret) {
1178                 case -ENOTSUP:
1179                     monitor_printf(mon,
1180                                    "Snapshots not supported on device '%s'\n",
1181                                    bdrv_get_device_name(bs1));
1182                     break;
1183                 case -ENOENT:
1184                     monitor_printf(mon, "Could not find snapshot '%s' on "
1185                                    "device '%s'\n",
1186                                    name, bdrv_get_device_name(bs1));
1187                     break;
1188                 default:
1189                     monitor_printf(mon, "Error %d while activating snapshot on"
1190                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1191                     break;
1192                 }
1193                 /* fatal on snapshot block device */
1194                 if (bs == bs1)
1195                     goto the_end;
1196             }
1197         }
1198     }
1199
1200     /* Don't even try to load empty VM states */
1201     ret = bdrv_snapshot_find(bs, &sn, name);
1202     if ((ret >= 0) && (sn.vm_state_size == 0))
1203         goto the_end;
1204
1205     /* restore the VM state */
1206     f = qemu_fopen_bdrv(bs, 0);
1207     if (!f) {
1208         monitor_printf(mon, "Could not open VM state file\n");
1209         goto the_end;
1210     }
1211     ret = qemu_loadvm_state(f);
1212     qemu_fclose(f);
1213     if (ret < 0) {
1214         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1215     }
1216  the_end:
1217     if (saved_vm_running)
1218         vm_start();
1219 }
1220
1221 void do_delvm(Monitor *mon, const char *name)
1222 {
1223     DriveInfo *dinfo;
1224     BlockDriverState *bs, *bs1;
1225     int ret;
1226
1227     bs = get_bs_snapshots();
1228     if (!bs) {
1229         monitor_printf(mon, "No block device supports snapshots\n");
1230         return;
1231     }
1232
1233     TAILQ_FOREACH(dinfo, &drives, next) {
1234         bs1 = dinfo->bdrv;
1235         if (bdrv_has_snapshot(bs1)) {
1236             ret = bdrv_snapshot_delete(bs1, name);
1237             if (ret < 0) {
1238                 if (ret == -ENOTSUP)
1239                     monitor_printf(mon,
1240                                    "Snapshots not supported on device '%s'\n",
1241                                    bdrv_get_device_name(bs1));
1242                 else
1243                     monitor_printf(mon, "Error %d while deleting snapshot on "
1244                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1245             }
1246         }
1247     }
1248 }
1249
1250 void do_info_snapshots(Monitor *mon)
1251 {
1252     DriveInfo *dinfo;
1253     BlockDriverState *bs, *bs1;
1254     QEMUSnapshotInfo *sn_tab, *sn;
1255     int nb_sns, i;
1256     char buf[256];
1257
1258     bs = get_bs_snapshots();
1259     if (!bs) {
1260         monitor_printf(mon, "No available block device supports snapshots\n");
1261         return;
1262     }
1263     monitor_printf(mon, "Snapshot devices:");
1264     TAILQ_FOREACH(dinfo, &drives, next) {
1265         bs1 = dinfo->bdrv;
1266         if (bdrv_has_snapshot(bs1)) {
1267             if (bs == bs1)
1268                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1269         }
1270     }
1271     monitor_printf(mon, "\n");
1272
1273     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1274     if (nb_sns < 0) {
1275         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1276         return;
1277     }
1278     monitor_printf(mon, "Snapshot list (from %s):\n",
1279                    bdrv_get_device_name(bs));
1280     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1281     for(i = 0; i < nb_sns; i++) {
1282         sn = &sn_tab[i];
1283         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1284     }
1285     qemu_free(sn_tab);
1286 }