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 void qemu_announce_self(void)
 120 {
 121     int i, j, len;
 122     VLANState *vlan;
 123     VLANClientState *vc;
 124     uint8_t buf[256];
 125
 126     for (i = 0; i < MAX_NICS; i++) {
 127         if (!nd_table[i].used)
 128             continue;
 129         len = announce_self_create(buf, nd_table[i].macaddr);
 130         vlan = nd_table[i].vlan;
 131         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 132             for (j=0; j < SELF_ANNOUNCE_ROUNDS; j++)
 133                 vc->fd_read(vc->opaque, buf, len);
 134         }
 135     }
 136 }
 137
 138 /***********************************************************/
 139 /* savevm/loadvm support */
 140
 141 #define IO_BUF_SIZE 32768
 142
 143 struct QEMUFile {
 144     QEMUFilePutBufferFunc *put_buffer;
 145     QEMUFileGetBufferFunc *get_buffer;
 146     QEMUFileCloseFunc *close;
 147     QEMUFileRateLimit *rate_limit;
 148     void *opaque;
 149     int is_write;
 150
 151     int64_t buf_offset; /* start of buffer when writing, end of buffer
 152                            when reading */
 153     int buf_index;
 154     int buf_size; /* 0 when writing */
 155     uint8_t buf[IO_BUF_SIZE];
 156
 157     int has_error;
 158 };
 159
 160 typedef struct QEMUFilePopen
 161 {
 162     FILE *popen_file;
 163     QEMUFile *file;
 164 } QEMUFilePopen;
 165
 166 typedef struct QEMUFileSocket
 167 {
 168     int fd;
 169     QEMUFile *file;
 170 } QEMUFileSocket;
 171
 172 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 173 {
 174     QEMUFileSocket *s = opaque;
 175     ssize_t len;
 176
 177     do {
 178         len = recv(s->fd, buf, size, 0);
 179     } while (len == -1 && socket_error() == EINTR);
 180
 181     if (len == -1)
 182         len = -socket_error();
 183
 184     return len;
 185 }
 186
 187 static int socket_close(void *opaque)
 188 {
 189     QEMUFileSocket *s = opaque;
 190     qemu_free(s);
 191     return 0;
 192 }
 193
 194 static int popen_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 195 {
 196     QEMUFilePopen *s = opaque;
 197     return fwrite(buf, 1, size, s->popen_file);
 198 }
 199
 200 static int popen_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 201 {
 202     QEMUFilePopen *s = opaque;
 203     return fread(buf, 1, size, s->popen_file);
 204 }
 205
 206 static int popen_close(void *opaque)
 207 {
 208     QEMUFilePopen *s = opaque;
 209     pclose(s->popen_file);
 210     qemu_free(s);
 211     return 0;
 212 }
 213
 214 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 215 {
 216     QEMUFilePopen *s;
 217
 218     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 219         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 220         return NULL;
 221     }
 222
 223     s = qemu_mallocz(sizeof(QEMUFilePopen));
 224
 225     s->popen_file = popen_file;
 226
 227     if(mode[0] == 'r') {
 228         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL);
 229     } else {
 230         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL);
 231     }
 232     fprintf(stderr, "qemu_popen: returning result of qemu_fopen_ops\n");
 233     return s->file;
 234 }
 235
 236 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 237 {
 238     FILE *popen_file;
 239
 240     popen_file = popen(command, mode);
 241     if(popen_file == NULL) {
 242         return NULL;
 243     }
 244
 245     return qemu_popen(popen_file, mode);
 246 }
 247
 248 QEMUFile *qemu_fopen_socket(int fd)
 249 {
 250     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 251
 252     s->fd = fd;
 253     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL);
 254     return s->file;
 255 }
 256
 257 typedef struct QEMUFileStdio
 258 {
 259     FILE *outfile;
 260 } QEMUFileStdio;
 261
 262 static int file_put_buffer(void *opaque, const uint8_t *buf,
 263                             int64_t pos, int size)
 264 {
 265     QEMUFileStdio *s = opaque;
 266     fseek(s->outfile, pos, SEEK_SET);
 267     fwrite(buf, 1, size, s->outfile);
 268     return size;
 269 }
 270
 271 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 272 {
 273     QEMUFileStdio *s = opaque;
 274     fseek(s->outfile, pos, SEEK_SET);
 275     return fread(buf, 1, size, s->outfile);
 276 }
 277
 278 static int file_close(void *opaque)
 279 {
 280     QEMUFileStdio *s = opaque;
 281     fclose(s->outfile);
 282     qemu_free(s);
 283     return 0;
 284 }
 285
 286 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 287 {
 288     QEMUFileStdio *s;
 289
 290     s = qemu_mallocz(sizeof(QEMUFileStdio));
 291
 292     s->outfile = fopen(filename, mode);
 293     if (!s->outfile)
 294         goto fail;
 295
 296     if (!strcmp(mode, "wb"))
 297         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL);
 298     else if (!strcmp(mode, "rb"))
 299         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL);
 300
 301 fail:
 302     if (s->outfile)
 303         fclose(s->outfile);
 304     qemu_free(s);
 305     return NULL;
 306 }
 307
 308 typedef struct QEMUFileBdrv
 309 {
 310     BlockDriverState *bs;
 311     int64_t base_offset;
 312 } QEMUFileBdrv;
 313
 314 static int block_put_buffer(void *opaque, const uint8_t *buf,
 315                            int64_t pos, int size)
 316 {
 317     QEMUFileBdrv *s = opaque;
 318     bdrv_put_buffer(s->bs, buf, s->base_offset + pos, size);
 319     return size;
 320 }
 321
 322 static int block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 323 {
 324     QEMUFileBdrv *s = opaque;
 325     return bdrv_get_buffer(s->bs, buf, s->base_offset + pos, size);
 326 }
 327
 328 static int bdrv_fclose(void *opaque)
 329 {
 330     QEMUFileBdrv *s = opaque;
 331     qemu_free(s);
 332     return 0;
 333 }
 334
 335 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable)
 336 {
 337     QEMUFileBdrv *s;
 338
 339     s = qemu_mallocz(sizeof(QEMUFileBdrv));
 340
 341     s->bs = bs;
 342     s->base_offset = offset;
 343
 344     if (is_writable)
 345         return qemu_fopen_ops(s, block_put_buffer, NULL, bdrv_fclose, NULL);
 346
 347     return qemu_fopen_ops(s, NULL, block_get_buffer, bdrv_fclose, NULL);
 348 }
 349
 350 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 351                          QEMUFileGetBufferFunc *get_buffer,
 352                          QEMUFileCloseFunc *close,
 353                          QEMUFileRateLimit *rate_limit)
 354 {
 355     QEMUFile *f;
 356
 357     f = qemu_mallocz(sizeof(QEMUFile));
 358
 359     f->opaque = opaque;
 360     f->put_buffer = put_buffer;
 361     f->get_buffer = get_buffer;
 362     f->close = close;
 363     f->rate_limit = rate_limit;
 364     f->is_write = 0;
 365
 366     return f;
 367 }
 368
 369 int qemu_file_has_error(QEMUFile *f)
 370 {
 371     return f->has_error;
 372 }
 373
 374 void qemu_file_set_error(QEMUFile *f)
 375 {
 376     f->has_error = 1;
 377 }
 378
 379 void qemu_fflush(QEMUFile *f)
 380 {
 381     if (!f->put_buffer)
 382         return;
 383
 384     if (f->is_write && f->buf_index > 0) {
 385         int len;
 386
 387         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 388         if (len > 0)
 389             f->buf_offset += f->buf_index;
 390         else
 391             f->has_error = 1;
 392         f->buf_index = 0;
 393     }
 394 }
 395
 396 static void qemu_fill_buffer(QEMUFile *f)
 397 {
 398     int len;
 399
 400     if (!f->get_buffer)
 401         return;
 402
 403     if (f->is_write)
 404         abort();
 405
 406     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 407     if (len > 0) {
 408         f->buf_index = 0;
 409         f->buf_size = len;
 410         f->buf_offset += len;
 411     } else if (len != -EAGAIN)
 412         f->has_error = 1;
 413 }
 414
 415 int qemu_fclose(QEMUFile *f)
 416 {
 417     int ret = 0;
 418     qemu_fflush(f);
 419     if (f->close)
 420         ret = f->close(f->opaque);
 421     qemu_free(f);
 422     return ret;
 423 }
 424
 425 void qemu_file_put_notify(QEMUFile *f)
 426 {
 427     f->put_buffer(f->opaque, NULL, 0, 0);
 428 }
 429
 430 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 431 {
 432     int l;
 433
 434     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 435         fprintf(stderr,
 436                 "Attempted to write to buffer while read buffer is not empty\n");
 437         abort();
 438     }
 439
 440     while (!f->has_error && size > 0) {
 441         l = IO_BUF_SIZE - f->buf_index;
 442         if (l > size)
 443             l = size;
 444         memcpy(f->buf + f->buf_index, buf, l);
 445         f->is_write = 1;
 446         f->buf_index += l;
 447         buf += l;
 448         size -= l;
 449         if (f->buf_index >= IO_BUF_SIZE)
 450             qemu_fflush(f);
 451     }
 452 }
 453
 454 void qemu_put_byte(QEMUFile *f, int v)
 455 {
 456     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 457         fprintf(stderr,
 458                 "Attempted to write to buffer while read buffer is not empty\n");
 459         abort();
 460     }
 461
 462     f->buf[f->buf_index++] = v;
 463     f->is_write = 1;
 464     if (f->buf_index >= IO_BUF_SIZE)
 465         qemu_fflush(f);
 466 }
 467
 468 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 469 {
 470     int size, l;
 471
 472     if (f->is_write)
 473         abort();
 474
 475     size = size1;
 476     while (size > 0) {
 477         l = f->buf_size - f->buf_index;
 478         if (l == 0) {
 479             qemu_fill_buffer(f);
 480             l = f->buf_size - f->buf_index;
 481             if (l == 0)
 482                 break;
 483         }
 484         if (l > size)
 485             l = size;
 486         memcpy(buf, f->buf + f->buf_index, l);
 487         f->buf_index += l;
 488         buf += l;
 489         size -= l;
 490     }
 491     return size1 - size;
 492 }
 493
 494 int qemu_get_byte(QEMUFile *f)
 495 {
 496     if (f->is_write)
 497         abort();
 498
 499     if (f->buf_index >= f->buf_size) {
 500         qemu_fill_buffer(f);
 501         if (f->buf_index >= f->buf_size)
 502             return 0;
 503     }
 504     return f->buf[f->buf_index++];
 505 }
 506
 507 int64_t qemu_ftell(QEMUFile *f)
 508 {
 509     return f->buf_offset - f->buf_size + f->buf_index;
 510 }
 511
 512 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 513 {
 514     if (whence == SEEK_SET) {
 515         /* nothing to do */
 516     } else if (whence == SEEK_CUR) {
 517         pos += qemu_ftell(f);
 518     } else {
 519         /* SEEK_END not supported */
 520         return -1;
 521     }
 522     if (f->put_buffer) {
 523         qemu_fflush(f);
 524         f->buf_offset = pos;
 525     } else {
 526         f->buf_offset = pos;
 527         f->buf_index = 0;
 528         f->buf_size = 0;
 529     }
 530     return pos;
 531 }
 532
 533 int qemu_file_rate_limit(QEMUFile *f)
 534 {
 535     if (f->rate_limit)
 536         return f->rate_limit(f->opaque);
 537
 538     return 0;
 539 }
 540
 541 void qemu_put_be16(QEMUFile *f, unsigned int v)
 542 {
 543     qemu_put_byte(f, v >> 8);
 544     qemu_put_byte(f, v);
 545 }
 546
 547 void qemu_put_be32(QEMUFile *f, unsigned int v)
 548 {
 549     qemu_put_byte(f, v >> 24);
 550     qemu_put_byte(f, v >> 16);
 551     qemu_put_byte(f, v >> 8);
 552     qemu_put_byte(f, v);
 553 }
 554
 555 void qemu_put_be64(QEMUFile *f, uint64_t v)
 556 {
 557     qemu_put_be32(f, v >> 32);
 558     qemu_put_be32(f, v);
 559 }
 560
 561 unsigned int qemu_get_be16(QEMUFile *f)
 562 {
 563     unsigned int v;
 564     v = qemu_get_byte(f) << 8;
 565     v |= qemu_get_byte(f);
 566     return v;
 567 }
 568
 569 unsigned int qemu_get_be32(QEMUFile *f)
 570 {
 571     unsigned int v;
 572     v = qemu_get_byte(f) << 24;
 573     v |= qemu_get_byte(f) << 16;
 574     v |= qemu_get_byte(f) << 8;
 575     v |= qemu_get_byte(f);
 576     return v;
 577 }
 578
 579 uint64_t qemu_get_be64(QEMUFile *f)
 580 {
 581     uint64_t v;
 582     v = (uint64_t)qemu_get_be32(f) << 32;
 583     v |= qemu_get_be32(f);
 584     return v;
 585 }
 586
 587 typedef struct SaveStateEntry {
 588     char idstr[256];
 589     int instance_id;
 590     int version_id;
 591     int section_id;
 592     SaveLiveStateHandler *save_live_state;
 593     SaveStateHandler *save_state;
 594     LoadStateHandler *load_state;
 595     void *opaque;
 596     struct SaveStateEntry *next;
 597 } SaveStateEntry;
 598
 599 static SaveStateEntry *first_se;
 600
 601 /* TODO: Individual devices generally have very little idea about the rest
 602    of the system, so instance_id should be removed/replaced.
 603    Meanwhile pass -1 as instance_id if you do not already have a clearly
 604    distinguishing id for all instances of your device class. */
 605 int register_savevm_live(const char *idstr,
 606                          int instance_id,
 607                          int version_id,
 608                          SaveLiveStateHandler *save_live_state,
 609                          SaveStateHandler *save_state,
 610                          LoadStateHandler *load_state,
 611                          void *opaque)
 612 {
 613     SaveStateEntry *se, **pse;
 614     static int global_section_id;
 615
 616     se = qemu_malloc(sizeof(SaveStateEntry));
 617     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 618     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 619     se->version_id = version_id;
 620     se->section_id = global_section_id++;
 621     se->save_live_state = save_live_state;
 622     se->save_state = save_state;
 623     se->load_state = load_state;
 624     se->opaque = opaque;
 625     se->next = NULL;
 626
 627     /* add at the end of list */
 628     pse = &first_se;
 629     while (*pse != NULL) {
 630         if (instance_id == -1
 631                 && strcmp(se->idstr, (*pse)->idstr) == 0
 632                 && se->instance_id <= (*pse)->instance_id)
 633             se->instance_id = (*pse)->instance_id + 1;
 634         pse = &(*pse)->next;
 635     }
 636     *pse = se;
 637     return 0;
 638 }
 639
 640 int register_savevm(const char *idstr,
 641                     int instance_id,
 642                     int version_id,
 643                     SaveStateHandler *save_state,
 644                     LoadStateHandler *load_state,
 645                     void *opaque)
 646 {
 647     return register_savevm_live(idstr, instance_id, version_id,
 648                                 NULL, save_state, load_state, opaque);
 649 }
 650
 651 void unregister_savevm(const char *idstr, void *opaque)
 652 {
 653     SaveStateEntry **pse;
 654
 655     pse = &first_se;
 656     while (*pse != NULL) {
 657         if (strcmp((*pse)->idstr, idstr) == 0 && (*pse)->opaque == opaque) {
 658             SaveStateEntry *next = (*pse)->next;
 659             qemu_free(*pse);
 660             *pse = next;
 661             continue;
 662         }
 663         pse = &(*pse)->next;
 664     }
 665 }
 666
 667 #define QEMU_VM_FILE_MAGIC           0x5145564d
 668 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 669 #define QEMU_VM_FILE_VERSION         0x00000003
 670
 671 #define QEMU_VM_EOF                  0x00
 672 #define QEMU_VM_SECTION_START        0x01
 673 #define QEMU_VM_SECTION_PART         0x02
 674 #define QEMU_VM_SECTION_END          0x03
 675 #define QEMU_VM_SECTION_FULL         0x04
 676
 677 int qemu_savevm_state_begin(QEMUFile *f)
 678 {
 679     SaveStateEntry *se;
 680
 681     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 682     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 683
 684     for (se = first_se; se != NULL; se = se->next) {
 685         int len;
 686
 687         if (se->save_live_state == NULL)
 688             continue;
 689
 690         /* Section type */
 691         qemu_put_byte(f, QEMU_VM_SECTION_START);
 692         qemu_put_be32(f, se->section_id);
 693
 694         /* ID string */
 695         len = strlen(se->idstr);
 696         qemu_put_byte(f, len);
 697         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 698
 699         qemu_put_be32(f, se->instance_id);
 700         qemu_put_be32(f, se->version_id);
 701
 702         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 703     }
 704
 705     if (qemu_file_has_error(f))
 706         return -EIO;
 707
 708     return 0;
 709 }
 710
 711 int qemu_savevm_state_iterate(QEMUFile *f)
 712 {
 713     SaveStateEntry *se;
 714     int ret = 1;
 715
 716     for (se = first_se; se != NULL; se = se->next) {
 717         if (se->save_live_state == NULL)
 718             continue;
 719
 720         /* Section type */
 721         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 722         qemu_put_be32(f, se->section_id);
 723
 724         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 725     }
 726
 727     if (ret)
 728         return 1;
 729
 730     if (qemu_file_has_error(f))
 731         return -EIO;
 732
 733     return 0;
 734 }
 735
 736 int qemu_savevm_state_complete(QEMUFile *f)
 737 {
 738     SaveStateEntry *se;
 739
 740     for (se = first_se; se != NULL; se = se->next) {
 741         if (se->save_live_state == NULL)
 742             continue;
 743
 744         /* Section type */
 745         qemu_put_byte(f, QEMU_VM_SECTION_END);
 746         qemu_put_be32(f, se->section_id);
 747
 748         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 749     }
 750
 751     for(se = first_se; se != NULL; se = se->next) {
 752         int len;
 753
 754         if (se->save_state == NULL)
 755             continue;
 756
 757         /* Section type */
 758         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 759         qemu_put_be32(f, se->section_id);
 760
 761         /* ID string */
 762         len = strlen(se->idstr);
 763         qemu_put_byte(f, len);
 764         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 765
 766         qemu_put_be32(f, se->instance_id);
 767         qemu_put_be32(f, se->version_id);
 768
 769         se->save_state(f, se->opaque);
 770     }
 771
 772     qemu_put_byte(f, QEMU_VM_EOF);
 773
 774     if (qemu_file_has_error(f))
 775         return -EIO;
 776
 777     return 0;
 778 }
 779
 780 int qemu_savevm_state(QEMUFile *f)
 781 {
 782     int saved_vm_running;
 783     int ret;
 784
 785     saved_vm_running = vm_running;
 786     vm_stop(0);
 787
 788     bdrv_flush_all();
 789
 790     ret = qemu_savevm_state_begin(f);
 791     if (ret < 0)
 792         goto out;
 793
 794     do {
 795         ret = qemu_savevm_state_iterate(f);
 796         if (ret < 0)
 797             goto out;
 798     } while (ret == 0);
 799
 800     ret = qemu_savevm_state_complete(f);
 801
 802 out:
 803     if (qemu_file_has_error(f))
 804         ret = -EIO;
 805
 806     if (!ret && saved_vm_running)
 807         vm_start();
 808
 809     return ret;
 810 }
 811
 812 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 813 {
 814     SaveStateEntry *se;
 815
 816     for(se = first_se; se != NULL; se = se->next) {
 817         if (!strcmp(se->idstr, idstr) &&
 818             instance_id == se->instance_id)
 819             return se;
 820     }
 821     return NULL;
 822 }
 823
 824 typedef struct LoadStateEntry {
 825     SaveStateEntry *se;
 826     int section_id;
 827     int version_id;
 828     struct LoadStateEntry *next;
 829 } LoadStateEntry;
 830
 831 static int qemu_loadvm_state_v2(QEMUFile *f)
 832 {
 833     SaveStateEntry *se;
 834     int len, ret, instance_id, record_len, version_id;
 835     int64_t total_len, end_pos, cur_pos;
 836     char idstr[256];
 837
 838     total_len = qemu_get_be64(f);
 839     end_pos = total_len + qemu_ftell(f);
 840     for(;;) {
 841         if (qemu_ftell(f) >= end_pos)
 842             break;
 843         len = qemu_get_byte(f);
 844         qemu_get_buffer(f, (uint8_t *)idstr, len);
 845         idstr[len] = '\0';
 846         instance_id = qemu_get_be32(f);
 847         version_id = qemu_get_be32(f);
 848         record_len = qemu_get_be32(f);
 849         cur_pos = qemu_ftell(f);
 850         se = find_se(idstr, instance_id);
 851         if (!se) {
 852             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 853                     instance_id, idstr);
 854         } else {
 855             ret = se->load_state(f, se->opaque, version_id);
 856             if (ret < 0) {
 857                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 858                         instance_id, idstr);
 859                 return ret;
 860             }
 861         }
 862         /* always seek to exact end of record */
 863         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 864     }
 865
 866     if (qemu_file_has_error(f))
 867         return -EIO;
 868
 869     return 0;
 870 }
 871
 872 int qemu_loadvm_state(QEMUFile *f)
 873 {
 874     LoadStateEntry *first_le = NULL;
 875     uint8_t section_type;
 876     unsigned int v;
 877     int ret;
 878
 879     v = qemu_get_be32(f);
 880     if (v != QEMU_VM_FILE_MAGIC)
 881         return -EINVAL;
 882
 883     v = qemu_get_be32(f);
 884     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 885         return qemu_loadvm_state_v2(f);
 886     if (v != QEMU_VM_FILE_VERSION)
 887         return -ENOTSUP;
 888
 889     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 890         uint32_t instance_id, version_id, section_id;
 891         LoadStateEntry *le;
 892         SaveStateEntry *se;
 893         char idstr[257];
 894         int len;
 895
 896         switch (section_type) {
 897         case QEMU_VM_SECTION_START:
 898         case QEMU_VM_SECTION_FULL:
 899             /* Read section start */
 900             section_id = qemu_get_be32(f);
 901             len = qemu_get_byte(f);
 902             qemu_get_buffer(f, (uint8_t *)idstr, len);
 903             idstr[len] = 0;
 904             instance_id = qemu_get_be32(f);
 905             version_id = qemu_get_be32(f);
 906
 907             /* Find savevm section */
 908             se = find_se(idstr, instance_id);
 909             if (se == NULL) {
 910                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 911                 ret = -EINVAL;
 912                 goto out;
 913             }
 914
 915             /* Validate version */
 916             if (version_id > se->version_id) {
 917                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 918                         version_id, idstr, se->version_id);
 919                 ret = -EINVAL;
 920                 goto out;
 921             }
 922
 923             /* Add entry */
 924             le = qemu_mallocz(sizeof(*le));
 925
 926             le->se = se;
 927             le->section_id = section_id;
 928             le->version_id = version_id;
 929             le->next = first_le;
 930             first_le = le;
 931
 932             le->se->load_state(f, le->se->opaque, le->version_id);
 933             break;
 934         case QEMU_VM_SECTION_PART:
 935         case QEMU_VM_SECTION_END:
 936             section_id = qemu_get_be32(f);
 937
 938             for (le = first_le; le && le->section_id != section_id; le = le->next);
 939             if (le == NULL) {
 940                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 941                 ret = -EINVAL;
 942                 goto out;
 943             }
 944
 945             le->se->load_state(f, le->se->opaque, le->version_id);
 946             break;
 947         default:
 948             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 949             ret = -EINVAL;
 950             goto out;
 951         }
 952     }
 953
 954     ret = 0;
 955
 956 out:
 957     while (first_le) {
 958         LoadStateEntry *le = first_le;
 959         first_le = first_le->next;
 960         qemu_free(le);
 961     }
 962
 963     if (qemu_file_has_error(f))
 964         ret = -EIO;
 965
 966     return ret;
 967 }
 968
 969 /* device can contain snapshots */
 970 static int bdrv_can_snapshot(BlockDriverState *bs)
 971 {
 972     return (bs &&
 973             !bdrv_is_removable(bs) &&
 974             !bdrv_is_read_only(bs));
 975 }
 976
 977 /* device must be snapshots in order to have a reliable snapshot */
 978 static int bdrv_has_snapshot(BlockDriverState *bs)
 979 {
 980     return (bs &&
 981             !bdrv_is_removable(bs) &&
 982             !bdrv_is_read_only(bs));
 983 }
 984
 985 static BlockDriverState *get_bs_snapshots(void)
 986 {
 987     BlockDriverState *bs;
 988     int i;
 989
 990     if (bs_snapshots)
 991         return bs_snapshots;
 992     for(i = 0; i <= nb_drives; i++) {
 993         bs = drives_table[i].bdrv;
 994         if (bdrv_can_snapshot(bs))
 995             goto ok;
 996     }
 997     return NULL;
 998  ok:
 999     bs_snapshots = bs;
1000     return bs;
1001 }
1002
1003 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
1004                               const char *name)
1005 {
1006     QEMUSnapshotInfo *sn_tab, *sn;
1007     int nb_sns, i, ret;
1008
1009     ret = -ENOENT;
1010     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1011     if (nb_sns < 0)
1012         return ret;
1013     for(i = 0; i < nb_sns; i++) {
1014         sn = &sn_tab[i];
1015         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
1016             *sn_info = *sn;
1017             ret = 0;
1018             break;
1019         }
1020     }
1021     qemu_free(sn_tab);
1022     return ret;
1023 }
1024
1025 void do_savevm(Monitor *mon, const char *name)
1026 {
1027     BlockDriverState *bs, *bs1;
1028     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1029     int must_delete, ret, i;
1030     BlockDriverInfo bdi1, *bdi = &bdi1;
1031     QEMUFile *f;
1032     int saved_vm_running;
1033     uint32_t vm_state_size;
1034 #ifdef _WIN32
1035     struct _timeb tb;
1036 #else
1037     struct timeval tv;
1038 #endif
1039
1040     bs = get_bs_snapshots();
1041     if (!bs) {
1042         monitor_printf(mon, "No block device can accept snapshots\n");
1043         return;
1044     }
1045
1046     /* ??? Should this occur after vm_stop?  */
1047     qemu_aio_flush();
1048
1049     saved_vm_running = vm_running;
1050     vm_stop(0);
1051
1052     must_delete = 0;
1053     if (name) {
1054         ret = bdrv_snapshot_find(bs, old_sn, name);
1055         if (ret >= 0) {
1056             must_delete = 1;
1057         }
1058     }
1059     memset(sn, 0, sizeof(*sn));
1060     if (must_delete) {
1061         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1062         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1063     } else {
1064         if (name)
1065             pstrcpy(sn->name, sizeof(sn->name), name);
1066     }
1067
1068     /* fill auxiliary fields */
1069 #ifdef _WIN32
1070     _ftime(&tb);
1071     sn->date_sec = tb.time;
1072     sn->date_nsec = tb.millitm * 1000000;
1073 #else
1074     gettimeofday(&tv, NULL);
1075     sn->date_sec = tv.tv_sec;
1076     sn->date_nsec = tv.tv_usec * 1000;
1077 #endif
1078     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1079
1080     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1081         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1082                        bdrv_get_device_name(bs));
1083         goto the_end;
1084     }
1085
1086     /* save the VM state */
1087     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 1);
1088     if (!f) {
1089         monitor_printf(mon, "Could not open VM state file\n");
1090         goto the_end;
1091     }
1092     ret = qemu_savevm_state(f);
1093     vm_state_size = qemu_ftell(f);
1094     qemu_fclose(f);
1095     if (ret < 0) {
1096         monitor_printf(mon, "Error %d while writing VM\n", ret);
1097         goto the_end;
1098     }
1099
1100     /* create the snapshots */
1101
1102     for(i = 0; i < nb_drives; i++) {
1103         bs1 = drives_table[i].bdrv;
1104         if (bdrv_has_snapshot(bs1)) {
1105             if (must_delete) {
1106                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1107                 if (ret < 0) {
1108                     monitor_printf(mon,
1109                                    "Error while deleting snapshot on '%s'\n",
1110                                    bdrv_get_device_name(bs1));
1111                 }
1112             }
1113             /* Write VM state size only to the image that contains the state */
1114             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1115             ret = bdrv_snapshot_create(bs1, sn);
1116             if (ret < 0) {
1117                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1118                                bdrv_get_device_name(bs1));
1119             }
1120         }
1121     }
1122
1123  the_end:
1124     if (saved_vm_running)
1125         vm_start();
1126 }
1127
1128 void do_loadvm(Monitor *mon, const char *name)
1129 {
1130     BlockDriverState *bs, *bs1;
1131     BlockDriverInfo bdi1, *bdi = &bdi1;
1132     QEMUSnapshotInfo sn;
1133     QEMUFile *f;
1134     int i, ret;
1135     int saved_vm_running;
1136
1137     bs = get_bs_snapshots();
1138     if (!bs) {
1139         monitor_printf(mon, "No block device supports snapshots\n");
1140         return;
1141     }
1142
1143     /* Flush all IO requests so they don't interfere with the new state.  */
1144     qemu_aio_flush();
1145
1146     saved_vm_running = vm_running;
1147     vm_stop(0);
1148
1149     for(i = 0; i <= nb_drives; i++) {
1150         bs1 = drives_table[i].bdrv;
1151         if (bdrv_has_snapshot(bs1)) {
1152             ret = bdrv_snapshot_goto(bs1, name);
1153             if (ret < 0) {
1154                 if (bs != bs1)
1155                     monitor_printf(mon, "Warning: ");
1156                 switch(ret) {
1157                 case -ENOTSUP:
1158                     monitor_printf(mon,
1159                                    "Snapshots not supported on device '%s'\n",
1160                                    bdrv_get_device_name(bs1));
1161                     break;
1162                 case -ENOENT:
1163                     monitor_printf(mon, "Could not find snapshot '%s' on "
1164                                    "device '%s'\n",
1165                                    name, bdrv_get_device_name(bs1));
1166                     break;
1167                 default:
1168                     monitor_printf(mon, "Error %d while activating snapshot on"
1169                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1170                     break;
1171                 }
1172                 /* fatal on snapshot block device */
1173                 if (bs == bs1)
1174                     goto the_end;
1175             }
1176         }
1177     }
1178
1179     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1180         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1181                        bdrv_get_device_name(bs));
1182         return;
1183     }
1184
1185     /* Don't even try to load empty VM states */
1186     ret = bdrv_snapshot_find(bs, &sn, name);
1187     if ((ret >= 0) && (sn.vm_state_size == 0))
1188         goto the_end;
1189
1190     /* restore the VM state */
1191     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 0);
1192     if (!f) {
1193         monitor_printf(mon, "Could not open VM state file\n");
1194         goto the_end;
1195     }
1196     ret = qemu_loadvm_state(f);
1197     qemu_fclose(f);
1198     if (ret < 0) {
1199         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1200     }
1201  the_end:
1202     if (saved_vm_running)
1203         vm_start();
1204 }
1205
1206 void do_delvm(Monitor *mon, const char *name)
1207 {
1208     BlockDriverState *bs, *bs1;
1209     int i, ret;
1210
1211     bs = get_bs_snapshots();
1212     if (!bs) {
1213         monitor_printf(mon, "No block device supports snapshots\n");
1214         return;
1215     }
1216
1217     for(i = 0; i <= nb_drives; i++) {
1218         bs1 = drives_table[i].bdrv;
1219         if (bdrv_has_snapshot(bs1)) {
1220             ret = bdrv_snapshot_delete(bs1, name);
1221             if (ret < 0) {
1222                 if (ret == -ENOTSUP)
1223                     monitor_printf(mon,
1224                                    "Snapshots not supported on device '%s'\n",
1225                                    bdrv_get_device_name(bs1));
1226                 else
1227                     monitor_printf(mon, "Error %d while deleting snapshot on "
1228                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1229             }
1230         }
1231     }
1232 }
1233
1234 void do_info_snapshots(Monitor *mon)
1235 {
1236     BlockDriverState *bs, *bs1;
1237     QEMUSnapshotInfo *sn_tab, *sn;
1238     int nb_sns, i;
1239     char buf[256];
1240
1241     bs = get_bs_snapshots();
1242     if (!bs) {
1243         monitor_printf(mon, "No available block device supports snapshots\n");
1244         return;
1245     }
1246     monitor_printf(mon, "Snapshot devices:");
1247     for(i = 0; i <= nb_drives; i++) {
1248         bs1 = drives_table[i].bdrv;
1249         if (bdrv_has_snapshot(bs1)) {
1250             if (bs == bs1)
1251                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1252         }
1253     }
1254     monitor_printf(mon, "\n");
1255
1256     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1257     if (nb_sns < 0) {
1258         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1259         return;
1260     }
1261     monitor_printf(mon, "Snapshot list (from %s):\n",
1262                    bdrv_get_device_name(bs));
1263     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1264     for(i = 0; i < nb_sns; i++) {
1265         sn = &sn_tab[i];
1266         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1267     }
1268     qemu_free(sn_tab);
1269 }