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_fdopen(int fd, const char *mode)
 289 {
 290     QEMUFileStdio *s;
 291
 292     if (mode == NULL ||
 293         (mode[0] != 'r' && mode[0] != 'w') ||
 294         mode[1] != 'b' || mode[2] != 0) {
 295         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 296         return NULL;
 297     }
 298
 299     s = qemu_mallocz(sizeof(QEMUFileStdio));
 300     s->stdio_file = fdopen(fd, mode);
 301     if (!s->stdio_file)
 302         goto fail;
 303
 304     if(mode[0] == 'r') {
 305         s->file = qemu_fopen_ops(s, NULL, stdio_get_buffer, stdio_fclose, NULL, NULL);
 306     } else {
 307         s->file = qemu_fopen_ops(s, stdio_put_buffer, NULL, stdio_fclose, NULL, NULL);
 308     }
 309     return s->file;
 310
 311 fail:
 312     qemu_free(s);
 313     return NULL;
 314 }
 315
 316 QEMUFile *qemu_fopen_socket(int fd)
 317 {
 318     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 319
 320     s->fd = fd;
 321     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL, NULL);
 322     return s->file;
 323 }
 324
 325 static int file_put_buffer(void *opaque, const uint8_t *buf,
 326                             int64_t pos, int size)
 327 {
 328     QEMUFileStdio *s = opaque;
 329     fseek(s->stdio_file, pos, SEEK_SET);
 330     fwrite(buf, 1, size, s->stdio_file);
 331     return size;
 332 }
 333
 334 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 335 {
 336     QEMUFileStdio *s = opaque;
 337     fseek(s->stdio_file, pos, SEEK_SET);
 338     return fread(buf, 1, size, s->stdio_file);
 339 }
 340
 341 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 342 {
 343     QEMUFileStdio *s;
 344
 345     if (mode == NULL ||
 346         (mode[0] != 'r' && mode[0] != 'w') ||
 347         mode[1] != 'b' || mode[2] != 0) {
 348         fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
 349         return NULL;
 350     }
 351
 352     s = qemu_mallocz(sizeof(QEMUFileStdio));
 353
 354     s->stdio_file = fopen(filename, mode);
 355     if (!s->stdio_file)
 356         goto fail;
 357
 358     if(mode[0] == 'w') {
 359         s->file = qemu_fopen_ops(s, file_put_buffer, NULL, stdio_fclose, NULL, NULL);
 360     } else {
 361         s->file = qemu_fopen_ops(s, NULL, file_get_buffer, stdio_fclose, NULL, NULL);
 362     }
 363     return s->file;
 364 fail:
 365     qemu_free(s);
 366     return NULL;
 367 }
 368
 369 static int block_put_buffer(void *opaque, const uint8_t *buf,
 370                            int64_t pos, int size)
 371 {
 372     bdrv_save_vmstate(opaque, buf, pos, size);
 373     return size;
 374 }
 375
 376 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 377 {
 378     return bdrv_load_vmstate(opaque, buf, pos, size);
 379 }
 380
 381 static int bdrv_fclose(void *opaque)
 382 {
 383     return 0;
 384 }
 385
 386 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable)
 387 {
 388     if (is_writable)
 389         return qemu_fopen_ops(bs, block_put_buffer, NULL, bdrv_fclose, NULL, NULL);
 390     return qemu_fopen_ops(bs, NULL, block_get_buffer, bdrv_fclose, NULL, NULL);
 391 }
 392
 393 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 394                          QEMUFileGetBufferFunc *get_buffer,
 395                          QEMUFileCloseFunc *close,
 396                          QEMUFileRateLimit *rate_limit,
 397                          QEMUFileSetRateLimit *set_rate_limit)
 398 {
 399     QEMUFile *f;
 400
 401     f = qemu_mallocz(sizeof(QEMUFile));
 402
 403     f->opaque = opaque;
 404     f->put_buffer = put_buffer;
 405     f->get_buffer = get_buffer;
 406     f->close = close;
 407     f->rate_limit = rate_limit;
 408     f->set_rate_limit = set_rate_limit;
 409     f->is_write = 0;
 410
 411     return f;
 412 }
 413
 414 int qemu_file_has_error(QEMUFile *f)
 415 {
 416     return f->has_error;
 417 }
 418
 419 void qemu_file_set_error(QEMUFile *f)
 420 {
 421     f->has_error = 1;
 422 }
 423
 424 void qemu_fflush(QEMUFile *f)
 425 {
 426     if (!f->put_buffer)
 427         return;
 428
 429     if (f->is_write && f->buf_index > 0) {
 430         int len;
 431
 432         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 433         if (len > 0)
 434             f->buf_offset += f->buf_index;
 435         else
 436             f->has_error = 1;
 437         f->buf_index = 0;
 438     }
 439 }
 440
 441 static void qemu_fill_buffer(QEMUFile *f)
 442 {
 443     int len;
 444
 445     if (!f->get_buffer)
 446         return;
 447
 448     if (f->is_write)
 449         abort();
 450
 451     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 452     if (len > 0) {
 453         f->buf_index = 0;
 454         f->buf_size = len;
 455         f->buf_offset += len;
 456     } else if (len != -EAGAIN)
 457         f->has_error = 1;
 458 }
 459
 460 int qemu_fclose(QEMUFile *f)
 461 {
 462     int ret = 0;
 463     qemu_fflush(f);
 464     if (f->close)
 465         ret = f->close(f->opaque);
 466     qemu_free(f);
 467     return ret;
 468 }
 469
 470 void qemu_file_put_notify(QEMUFile *f)
 471 {
 472     f->put_buffer(f->opaque, NULL, 0, 0);
 473 }
 474
 475 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 476 {
 477     int l;
 478
 479     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 480         fprintf(stderr,
 481                 "Attempted to write to buffer while read buffer is not empty\n");
 482         abort();
 483     }
 484
 485     while (!f->has_error && size > 0) {
 486         l = IO_BUF_SIZE - f->buf_index;
 487         if (l > size)
 488             l = size;
 489         memcpy(f->buf + f->buf_index, buf, l);
 490         f->is_write = 1;
 491         f->buf_index += l;
 492         buf += l;
 493         size -= l;
 494         if (f->buf_index >= IO_BUF_SIZE)
 495             qemu_fflush(f);
 496     }
 497 }
 498
 499 void qemu_put_byte(QEMUFile *f, int v)
 500 {
 501     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 502         fprintf(stderr,
 503                 "Attempted to write to buffer while read buffer is not empty\n");
 504         abort();
 505     }
 506
 507     f->buf[f->buf_index++] = v;
 508     f->is_write = 1;
 509     if (f->buf_index >= IO_BUF_SIZE)
 510         qemu_fflush(f);
 511 }
 512
 513 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 514 {
 515     int size, l;
 516
 517     if (f->is_write)
 518         abort();
 519
 520     size = size1;
 521     while (size > 0) {
 522         l = f->buf_size - f->buf_index;
 523         if (l == 0) {
 524             qemu_fill_buffer(f);
 525             l = f->buf_size - f->buf_index;
 526             if (l == 0)
 527                 break;
 528         }
 529         if (l > size)
 530             l = size;
 531         memcpy(buf, f->buf + f->buf_index, l);
 532         f->buf_index += l;
 533         buf += l;
 534         size -= l;
 535     }
 536     return size1 - size;
 537 }
 538
 539 int qemu_get_byte(QEMUFile *f)
 540 {
 541     if (f->is_write)
 542         abort();
 543
 544     if (f->buf_index >= f->buf_size) {
 545         qemu_fill_buffer(f);
 546         if (f->buf_index >= f->buf_size)
 547             return 0;
 548     }
 549     return f->buf[f->buf_index++];
 550 }
 551
 552 int64_t qemu_ftell(QEMUFile *f)
 553 {
 554     return f->buf_offset - f->buf_size + f->buf_index;
 555 }
 556
 557 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 558 {
 559     if (whence == SEEK_SET) {
 560         /* nothing to do */
 561     } else if (whence == SEEK_CUR) {
 562         pos += qemu_ftell(f);
 563     } else {
 564         /* SEEK_END not supported */
 565         return -1;
 566     }
 567     if (f->put_buffer) {
 568         qemu_fflush(f);
 569         f->buf_offset = pos;
 570     } else {
 571         f->buf_offset = pos;
 572         f->buf_index = 0;
 573         f->buf_size = 0;
 574     }
 575     return pos;
 576 }
 577
 578 int qemu_file_rate_limit(QEMUFile *f)
 579 {
 580     if (f->rate_limit)
 581         return f->rate_limit(f->opaque);
 582
 583     return 0;
 584 }
 585
 586 size_t qemu_file_set_rate_limit(QEMUFile *f, size_t new_rate)
 587 {
 588     /* any failed or completed migration keeps its state to allow probing of
 589      * migration data, but has no associated file anymore */
 590     if (f && f->set_rate_limit)
 591         return f->set_rate_limit(f->opaque, new_rate);
 592
 593     return 0;
 594 }
 595
 596 void qemu_put_be16(QEMUFile *f, unsigned int v)
 597 {
 598     qemu_put_byte(f, v >> 8);
 599     qemu_put_byte(f, v);
 600 }
 601
 602 void qemu_put_be32(QEMUFile *f, unsigned int v)
 603 {
 604     qemu_put_byte(f, v >> 24);
 605     qemu_put_byte(f, v >> 16);
 606     qemu_put_byte(f, v >> 8);
 607     qemu_put_byte(f, v);
 608 }
 609
 610 void qemu_put_be64(QEMUFile *f, uint64_t v)
 611 {
 612     qemu_put_be32(f, v >> 32);
 613     qemu_put_be32(f, v);
 614 }
 615
 616 unsigned int qemu_get_be16(QEMUFile *f)
 617 {
 618     unsigned int v;
 619     v = qemu_get_byte(f) << 8;
 620     v |= qemu_get_byte(f);
 621     return v;
 622 }
 623
 624 unsigned int qemu_get_be32(QEMUFile *f)
 625 {
 626     unsigned int v;
 627     v = qemu_get_byte(f) << 24;
 628     v |= qemu_get_byte(f) << 16;
 629     v |= qemu_get_byte(f) << 8;
 630     v |= qemu_get_byte(f);
 631     return v;
 632 }
 633
 634 uint64_t qemu_get_be64(QEMUFile *f)
 635 {
 636     uint64_t v;
 637     v = (uint64_t)qemu_get_be32(f) << 32;
 638     v |= qemu_get_be32(f);
 639     return v;
 640 }
 641
 642 typedef struct SaveStateEntry {
 643     char idstr[256];
 644     int instance_id;
 645     int version_id;
 646     int section_id;
 647     SaveLiveStateHandler *save_live_state;
 648     SaveStateHandler *save_state;
 649     LoadStateHandler *load_state;
 650     void *opaque;
 651     struct SaveStateEntry *next;
 652 } SaveStateEntry;
 653
 654 static SaveStateEntry *first_se;
 655
 656 /* TODO: Individual devices generally have very little idea about the rest
 657    of the system, so instance_id should be removed/replaced.
 658    Meanwhile pass -1 as instance_id if you do not already have a clearly
 659    distinguishing id for all instances of your device class. */
 660 int register_savevm_live(const char *idstr,
 661                          int instance_id,
 662                          int version_id,
 663                          SaveLiveStateHandler *save_live_state,
 664                          SaveStateHandler *save_state,
 665                          LoadStateHandler *load_state,
 666                          void *opaque)
 667 {
 668     SaveStateEntry *se, **pse;
 669     static int global_section_id;
 670
 671     se = qemu_malloc(sizeof(SaveStateEntry));
 672     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 673     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 674     se->version_id = version_id;
 675     se->section_id = global_section_id++;
 676     se->save_live_state = save_live_state;
 677     se->save_state = save_state;
 678     se->load_state = load_state;
 679     se->opaque = opaque;
 680     se->next = NULL;
 681
 682     /* add at the end of list */
 683     pse = &first_se;
 684     while (*pse != NULL) {
 685         if (instance_id == -1
 686                 && strcmp(se->idstr, (*pse)->idstr) == 0
 687                 && se->instance_id <= (*pse)->instance_id)
 688             se->instance_id = (*pse)->instance_id + 1;
 689         pse = &(*pse)->next;
 690     }
 691     *pse = se;
 692     return 0;
 693 }
 694
 695 int register_savevm(const char *idstr,
 696                     int instance_id,
 697                     int version_id,
 698                     SaveStateHandler *save_state,
 699                     LoadStateHandler *load_state,
 700                     void *opaque)
 701 {
 702     return register_savevm_live(idstr, instance_id, version_id,
 703                                 NULL, save_state, load_state, opaque);
 704 }
 705
 706 void unregister_savevm(const char *idstr, void *opaque)
 707 {
 708     SaveStateEntry **pse;
 709
 710     pse = &first_se;
 711     while (*pse != NULL) {
 712         if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
 713             SaveStateEntry *next = (*pse)->next;
 714             qemu_free(*pse);
 715             *pse = next;
 716             continue;
 717         }
 718         pse = &(*pse)->next;
 719     }
 720 }
 721
 722 #define QEMU_VM_FILE_MAGIC           0x5145564d
 723 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 724 #define QEMU_VM_FILE_VERSION         0x00000003
 725
 726 #define QEMU_VM_EOF                  0x00
 727 #define QEMU_VM_SECTION_START        0x01
 728 #define QEMU_VM_SECTION_PART         0x02
 729 #define QEMU_VM_SECTION_END          0x03
 730 #define QEMU_VM_SECTION_FULL         0x04
 731
 732 int qemu_savevm_state_begin(QEMUFile *f)
 733 {
 734     SaveStateEntry *se;
 735
 736     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 737     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 738
 739     for (se = first_se; se != NULL; se = se->next) {
 740         int len;
 741
 742         if (se->save_live_state == NULL)
 743             continue;
 744
 745         /* Section type */
 746         qemu_put_byte(f, QEMU_VM_SECTION_START);
 747         qemu_put_be32(f, se->section_id);
 748
 749         /* ID string */
 750         len = strlen(se->idstr);
 751         qemu_put_byte(f, len);
 752         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 753
 754         qemu_put_be32(f, se->instance_id);
 755         qemu_put_be32(f, se->version_id);
 756
 757         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 758     }
 759
 760     if (qemu_file_has_error(f))
 761         return -EIO;
 762
 763     return 0;
 764 }
 765
 766 int qemu_savevm_state_iterate(QEMUFile *f)
 767 {
 768     SaveStateEntry *se;
 769     int ret = 1;
 770
 771     for (se = first_se; se != NULL; se = se->next) {
 772         if (se->save_live_state == NULL)
 773             continue;
 774
 775         /* Section type */
 776         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 777         qemu_put_be32(f, se->section_id);
 778
 779         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 780     }
 781
 782     if (ret)
 783         return 1;
 784
 785     if (qemu_file_has_error(f))
 786         return -EIO;
 787
 788     return 0;
 789 }
 790
 791 int qemu_savevm_state_complete(QEMUFile *f)
 792 {
 793     SaveStateEntry *se;
 794
 795     for (se = first_se; se != NULL; se = se->next) {
 796         if (se->save_live_state == NULL)
 797             continue;
 798
 799         /* Section type */
 800         qemu_put_byte(f, QEMU_VM_SECTION_END);
 801         qemu_put_be32(f, se->section_id);
 802
 803         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 804     }
 805
 806     for(se = first_se; se != NULL; se = se->next) {
 807         int len;
 808
 809         if (se->save_state == NULL)
 810             continue;
 811
 812         /* Section type */
 813         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 814         qemu_put_be32(f, se->section_id);
 815
 816         /* ID string */
 817         len = strlen(se->idstr);
 818         qemu_put_byte(f, len);
 819         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 820
 821         qemu_put_be32(f, se->instance_id);
 822         qemu_put_be32(f, se->version_id);
 823
 824         se->save_state(f, se->opaque);
 825     }
 826
 827     qemu_put_byte(f, QEMU_VM_EOF);
 828
 829     if (qemu_file_has_error(f))
 830         return -EIO;
 831
 832     return 0;
 833 }
 834
 835 int qemu_savevm_state(QEMUFile *f)
 836 {
 837     int saved_vm_running;
 838     int ret;
 839
 840     saved_vm_running = vm_running;
 841     vm_stop(0);
 842
 843     bdrv_flush_all();
 844
 845     ret = qemu_savevm_state_begin(f);
 846     if (ret < 0)
 847         goto out;
 848
 849     do {
 850         ret = qemu_savevm_state_iterate(f);
 851         if (ret < 0)
 852             goto out;
 853     } while (ret == 0);
 854
 855     ret = qemu_savevm_state_complete(f);
 856
 857 out:
 858     if (qemu_file_has_error(f))
 859         ret = -EIO;
 860
 861     if (!ret && saved_vm_running)
 862         vm_start();
 863
 864     return ret;
 865 }
 866
 867 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 868 {
 869     SaveStateEntry *se;
 870
 871     for(se = first_se; se != NULL; se = se->next) {
 872         if (!strcmp(se->idstr, idstr) &&
 873             instance_id == se->instance_id)
 874             return se;
 875     }
 876     return NULL;
 877 }
 878
 879 typedef struct LoadStateEntry {
 880     SaveStateEntry *se;
 881     int section_id;
 882     int version_id;
 883     struct LoadStateEntry *next;
 884 } LoadStateEntry;
 885
 886 static int qemu_loadvm_state_v2(QEMUFile *f)
 887 {
 888     SaveStateEntry *se;
 889     int len, ret, instance_id, record_len, version_id;
 890     int64_t total_len, end_pos, cur_pos;
 891     char idstr[256];
 892
 893     total_len = qemu_get_be64(f);
 894     end_pos = total_len + qemu_ftell(f);
 895     for(;;) {
 896         if (qemu_ftell(f) >= end_pos)
 897             break;
 898         len = qemu_get_byte(f);
 899         qemu_get_buffer(f, (uint8_t *)idstr, len);
 900         idstr[len] = '\0';
 901         instance_id = qemu_get_be32(f);
 902         version_id = qemu_get_be32(f);
 903         record_len = qemu_get_be32(f);
 904         cur_pos = qemu_ftell(f);
 905         se = find_se(idstr, instance_id);
 906         if (!se) {
 907             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 908                     instance_id, idstr);
 909         } else {
 910             ret = se->load_state(f, se->opaque, version_id);
 911             if (ret < 0) {
 912                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 913                         instance_id, idstr);
 914                 return ret;
 915             }
 916         }
 917         /* always seek to exact end of record */
 918         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 919     }
 920
 921     if (qemu_file_has_error(f))
 922         return -EIO;
 923
 924     return 0;
 925 }
 926
 927 int qemu_loadvm_state(QEMUFile *f)
 928 {
 929     LoadStateEntry *first_le = NULL;
 930     uint8_t section_type;
 931     unsigned int v;
 932     int ret;
 933
 934     v = qemu_get_be32(f);
 935     if (v != QEMU_VM_FILE_MAGIC)
 936         return -EINVAL;
 937
 938     v = qemu_get_be32(f);
 939     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 940         return qemu_loadvm_state_v2(f);
 941     if (v != QEMU_VM_FILE_VERSION)
 942         return -ENOTSUP;
 943
 944     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 945         uint32_t instance_id, version_id, section_id;
 946         LoadStateEntry *le;
 947         SaveStateEntry *se;
 948         char idstr[257];
 949         int len;
 950
 951         switch (section_type) {
 952         case QEMU_VM_SECTION_START:
 953         case QEMU_VM_SECTION_FULL:
 954             /* Read section start */
 955             section_id = qemu_get_be32(f);
 956             len = qemu_get_byte(f);
 957             qemu_get_buffer(f, (uint8_t *)idstr, len);
 958             idstr[len] = 0;
 959             instance_id = qemu_get_be32(f);
 960             version_id = qemu_get_be32(f);
 961
 962             /* Find savevm section */
 963             se = find_se(idstr, instance_id);
 964             if (se == NULL) {
 965                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 966                 ret = -EINVAL;
 967                 goto out;
 968             }
 969
 970             /* Validate version */
 971             if (version_id > se->version_id) {
 972                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 973                         version_id, idstr, se->version_id);
 974                 ret = -EINVAL;
 975                 goto out;
 976             }
 977
 978             /* Add entry */
 979             le = qemu_mallocz(sizeof(*le));
 980
 981             le->se = se;
 982             le->section_id = section_id;
 983             le->version_id = version_id;
 984             le->next = first_le;
 985             first_le = le;
 986
 987             le->se->load_state(f, le->se->opaque, le->version_id);
 988             break;
 989         case QEMU_VM_SECTION_PART:
 990         case QEMU_VM_SECTION_END:
 991             section_id = qemu_get_be32(f);
 992
 993             for (le = first_le; le && le->section_id != section_id; le = le->next);
 994             if (le == NULL) {
 995                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 996                 ret = -EINVAL;
 997                 goto out;
 998             }
 999
1000             le->se->load_state(f, le->se->opaque, le->version_id);
1001             break;
1002         default:
1003             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
1004             ret = -EINVAL;
1005             goto out;
1006         }
1007     }
1008
1009     ret = 0;
1010
1011 out:
1012     while (first_le) {
1013         LoadStateEntry *le = first_le;
1014         first_le = first_le->next;
1015         qemu_free(le);
1016     }
1017
1018     if (qemu_file_has_error(f))
1019         ret = -EIO;
1020
1021     return ret;
1022 }
1023
1024 /* device can contain snapshots */
1025 static int bdrv_can_snapshot(BlockDriverState *bs)
1026 {
1027     return (bs &&
1028             !bdrv_is_removable(bs) &&
1029             !bdrv_is_read_only(bs));
1030 }
1031
1032 /* device must be snapshots in order to have a reliable snapshot */
1033 static int bdrv_has_snapshot(BlockDriverState *bs)
1034 {
1035     return (bs &&
1036             !bdrv_is_removable(bs) &&
1037             !bdrv_is_read_only(bs));
1038 }
1039
1040 static BlockDriverState *get_bs_snapshots(void)
1041 {
1042     BlockDriverState *bs;
1043     DriveInfo *dinfo;
1044
1045     if (bs_snapshots)
1046         return bs_snapshots;
1047     TAILQ_FOREACH(dinfo, &drives, next) {
1048         bs = dinfo->bdrv;
1049         if (bdrv_can_snapshot(bs))
1050             goto ok;
1051     }
1052     return NULL;
1053  ok:
1054     bs_snapshots = bs;
1055     return bs;
1056 }
1057
1058 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1059                               const char *name)
1060 {
1061     QEMUSnapshotInfo *sn_tab, *sn;
1062     int nb_sns, i, ret;
1063
1064     ret = -ENOENT;
1065     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1066     if (nb_sns < 0)
1067         return ret;
1068     for(i = 0; i < nb_sns; i++) {
1069         sn = &sn_tab[i];
1070         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1071             *sn_info = *sn;
1072             ret = 0;
1073             break;
1074         }
1075     }
1076     qemu_free(sn_tab);
1077     return ret;
1078 }
1079
1080 void do_savevm(Monitor *mon, const char *name)
1081 {
1082     DriveInfo *dinfo;
1083     BlockDriverState *bs, *bs1;
1084     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1085     int must_delete, ret;
1086     QEMUFile *f;
1087     int saved_vm_running;
1088     uint32_t vm_state_size;
1089 #ifdef _WIN32
1090     struct _timeb tb;
1091 #else
1092     struct timeval tv;
1093 #endif
1094
1095     bs = get_bs_snapshots();
1096     if (!bs) {
1097         monitor_printf(mon, "No block device can accept snapshots\n");
1098         return;
1099     }
1100
1101     /* ??? Should this occur after vm_stop?  */
1102     qemu_aio_flush();
1103
1104     saved_vm_running = vm_running;
1105     vm_stop(0);
1106
1107     must_delete = 0;
1108     if (name) {
1109         ret = bdrv_snapshot_find(bs, old_sn, name);
1110         if (ret >= 0) {
1111             must_delete = 1;
1112         }
1113     }
1114     memset(sn, 0, sizeof(*sn));
1115     if (must_delete) {
1116         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1117         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1118     } else {
1119         if (name)
1120             pstrcpy(sn->name, sizeof(sn->name), name);
1121     }
1122
1123     /* fill auxiliary fields */
1124 #ifdef _WIN32
1125     _ftime(&tb);
1126     sn->date_sec = tb.time;
1127     sn->date_nsec = tb.millitm * 1000000;
1128 #else
1129     gettimeofday(&tv, NULL);
1130     sn->date_sec = tv.tv_sec;
1131     sn->date_nsec = tv.tv_usec * 1000;
1132 #endif
1133     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1134
1135     /* save the VM state */
1136     f = qemu_fopen_bdrv(bs, 1);
1137     if (!f) {
1138         monitor_printf(mon, "Could not open VM state file\n");
1139         goto the_end;
1140     }
1141     ret = qemu_savevm_state(f);
1142     vm_state_size = qemu_ftell(f);
1143     qemu_fclose(f);
1144     if (ret < 0) {
1145         monitor_printf(mon, "Error %d while writing VM\n", ret);
1146         goto the_end;
1147     }
1148
1149     /* create the snapshots */
1150
1151     TAILQ_FOREACH(dinfo, &drives, next) {
1152         bs1 = dinfo->bdrv;
1153         if (bdrv_has_snapshot(bs1)) {
1154             if (must_delete) {
1155                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1156                 if (ret < 0) {
1157                     monitor_printf(mon,
1158                                    "Error while deleting snapshot on '%s'\n",
1159                                    bdrv_get_device_name(bs1));
1160                 }
1161             }
1162             /* Write VM state size only to the image that contains the state */
1163             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1164             ret = bdrv_snapshot_create(bs1, sn);
1165             if (ret < 0) {
1166                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1167                                bdrv_get_device_name(bs1));
1168             }
1169         }
1170     }
1171
1172  the_end:
1173     if (saved_vm_running)
1174         vm_start();
1175 }
1176
1177 void do_loadvm(Monitor *mon, const char *name)
1178 {
1179     DriveInfo *dinfo;
1180     BlockDriverState *bs, *bs1;
1181     QEMUSnapshotInfo sn;
1182     QEMUFile *f;
1183     int ret;
1184     int saved_vm_running;
1185
1186     bs = get_bs_snapshots();
1187     if (!bs) {
1188         monitor_printf(mon, "No block device supports snapshots\n");
1189         return;
1190     }
1191
1192     /* Flush all IO requests so they don't interfere with the new state.  */
1193     qemu_aio_flush();
1194
1195     saved_vm_running = vm_running;
1196     vm_stop(0);
1197
1198     TAILQ_FOREACH(dinfo, &drives, next) {
1199         bs1 = dinfo->bdrv;
1200         if (bdrv_has_snapshot(bs1)) {
1201             ret = bdrv_snapshot_goto(bs1, name);
1202             if (ret < 0) {
1203                 if (bs != bs1)
1204                     monitor_printf(mon, "Warning: ");
1205                 switch(ret) {
1206                 case -ENOTSUP:
1207                     monitor_printf(mon,
1208                                    "Snapshots not supported on device '%s'\n",
1209                                    bdrv_get_device_name(bs1));
1210                     break;
1211                 case -ENOENT:
1212                     monitor_printf(mon, "Could not find snapshot '%s' on "
1213                                    "device '%s'\n",
1214                                    name, bdrv_get_device_name(bs1));
1215                     break;
1216                 default:
1217                     monitor_printf(mon, "Error %d while activating snapshot on"
1218                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1219                     break;
1220                 }
1221                 /* fatal on snapshot block device */
1222                 if (bs == bs1)
1223                     goto the_end;
1224             }
1225         }
1226     }
1227
1228     /* Don't even try to load empty VM states */
1229     ret = bdrv_snapshot_find(bs, &sn, name);
1230     if ((ret >= 0) && (sn.vm_state_size == 0))
1231         goto the_end;
1232
1233     /* restore the VM state */
1234     f = qemu_fopen_bdrv(bs, 0);
1235     if (!f) {
1236         monitor_printf(mon, "Could not open VM state file\n");
1237         goto the_end;
1238     }
1239     ret = qemu_loadvm_state(f);
1240     qemu_fclose(f);
1241     if (ret < 0) {
1242         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1243     }
1244  the_end:
1245     if (saved_vm_running)
1246         vm_start();
1247 }
1248
1249 void do_delvm(Monitor *mon, const char *name)
1250 {
1251     DriveInfo *dinfo;
1252     BlockDriverState *bs, *bs1;
1253     int ret;
1254
1255     bs = get_bs_snapshots();
1256     if (!bs) {
1257         monitor_printf(mon, "No block device supports snapshots\n");
1258         return;
1259     }
1260
1261     TAILQ_FOREACH(dinfo, &drives, next) {
1262         bs1 = dinfo->bdrv;
1263         if (bdrv_has_snapshot(bs1)) {
1264             ret = bdrv_snapshot_delete(bs1, name);
1265             if (ret < 0) {
1266                 if (ret == -ENOTSUP)
1267                     monitor_printf(mon,
1268                                    "Snapshots not supported on device '%s'\n",
1269                                    bdrv_get_device_name(bs1));
1270                 else
1271                     monitor_printf(mon, "Error %d while deleting snapshot on "
1272                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1273             }
1274         }
1275     }
1276 }
1277
1278 void do_info_snapshots(Monitor *mon)
1279 {
1280     DriveInfo *dinfo;
1281     BlockDriverState *bs, *bs1;
1282     QEMUSnapshotInfo *sn_tab, *sn;
1283     int nb_sns, i;
1284     char buf[256];
1285
1286     bs = get_bs_snapshots();
1287     if (!bs) {
1288         monitor_printf(mon, "No available block device supports snapshots\n");
1289         return;
1290     }
1291     monitor_printf(mon, "Snapshot devices:");
1292     TAILQ_FOREACH(dinfo, &drives, next) {
1293         bs1 = dinfo->bdrv;
1294         if (bdrv_has_snapshot(bs1)) {
1295             if (bs == bs1)
1296                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1297         }
1298     }
1299     monitor_printf(mon, "\n");
1300
1301     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1302     if (nb_sns < 0) {
1303         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1304         return;
1305     }
1306     monitor_printf(mon, "Snapshot list (from %s):\n",
1307                    bdrv_get_device_name(bs));
1308     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1309     for(i = 0; i < nb_sns; i++) {
1310         sn = &sn_tab[i];
1311         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1312     }
1313     qemu_free(sn_tab);
1314 }