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 "qemu-common.h"
  25 #include "hw/hw.h"
  26 #include "net.h"
  27 #include "monitor.h"
  28 #include "sysemu.h"
  29 #include "qemu-timer.h"
  30 #include "qemu-char.h"
  31 #include "block.h"
  32 #include "audio/audio.h"
  33 #include "migration.h"
  34 #include "qemu_socket.h"
  35
  36 #include <unistd.h>
  37 #include <fcntl.h>
  38 #include <signal.h>
  39 #include <time.h>
  40 #include <errno.h>
  41 #include <sys/time.h>
  42 #include <zlib.h>
  43
  44 #ifndef _WIN32
  45 #include <sys/times.h>
  46 #include <sys/wait.h>
  47 #include <termios.h>
  48 #include <sys/mman.h>
  49 #include <sys/ioctl.h>
  50 #include <sys/resource.h>
  51 #include <sys/socket.h>
  52 #include <netinet/in.h>
  53 #include <net/if.h>
  54 #if defined(__NetBSD__)
  55 #include <net/if_tap.h>
  56 #endif
  57 #ifdef __linux__
  58 #include <linux/if_tun.h>
  59 #endif
  60 #include <arpa/inet.h>
  61 #include <dirent.h>
  62 #include <netdb.h>
  63 #include <sys/select.h>
  64 #ifdef _BSD
  65 #include <sys/stat.h>
  66 #ifdef __FreeBSD__
  67 #include <libutil.h>
  68 #else
  69 #include <util.h>
  70 #endif
  71 #elif defined (__GLIBC__) && defined (__FreeBSD_kernel__)
  72 #include <freebsd/stdlib.h>
  73 #else
  74 #ifdef __linux__
  75 #include <pty.h>
  76 #include <malloc.h>
  77 #include <linux/rtc.h>
  78 #endif
  79 #endif
  80 #endif
  81
  82 #ifdef _WIN32
  83 #include <malloc.h>
  84 #include <sys/timeb.h>
  85 #include <mmsystem.h>
  86 #define getopt_long_only getopt_long
  87 #define memalign(align, size) malloc(size)
  88 #endif
  89
  90 /* point to the block driver where the snapshots are managed */
  91 static BlockDriverState *bs_snapshots;
  92
  93 #define SELF_ANNOUNCE_ROUNDS 5
  94 #define ETH_P_EXPERIMENTAL 0x01F1 /* just a number */
  95 //#define ETH_P_EXPERIMENTAL 0x0012 /* make it the size of the packet */
  96 #define EXPERIMENTAL_MAGIC 0xf1f23f4f
  97
  98 static int announce_self_create(uint8_t *buf,
  99                                 uint8_t *mac_addr)
 100 {
 101     uint32_t magic = EXPERIMENTAL_MAGIC;
 102     uint16_t proto = htons(ETH_P_EXPERIMENTAL);
 103
 104     /* FIXME: should we send a different packet (arp/rarp/ping)? */
 105
 106     memset(buf, 0xff, 6);         /* h_dst */
 107     memcpy(buf + 6, mac_addr, 6); /* h_src */
 108     memcpy(buf + 12, &proto, 2);  /* h_proto */
 109     memcpy(buf + 14, &magic, 4);  /* magic */
 110
 111     return 18; /* len */
 112 }
 113
 114 void qemu_announce_self(void)
 115 {
 116     int i, j, len;
 117     VLANState *vlan;
 118     VLANClientState *vc;
 119     uint8_t buf[256];
 120
 121     for (i = 0; i < nb_nics; i++) {
 122         len = announce_self_create(buf, nd_table[i].macaddr);
 123         vlan = nd_table[i].vlan;
 124         for(vc = vlan->first_client; vc != NULL; vc = vc->next) {
 125             for (j=0; j < SELF_ANNOUNCE_ROUNDS; j++)
 126                 vc->fd_read(vc->opaque, buf, len);
 127         }
 128     }
 129 }
 130
 131 /***********************************************************/
 132 /* savevm/loadvm support */
 133
 134 #define IO_BUF_SIZE 32768
 135
 136 struct QEMUFile {
 137     QEMUFilePutBufferFunc *put_buffer;
 138     QEMUFileGetBufferFunc *get_buffer;
 139     QEMUFileCloseFunc *close;
 140     QEMUFileRateLimit *rate_limit;
 141     void *opaque;
 142     int is_write;
 143
 144     int64_t buf_offset; /* start of buffer when writing, end of buffer
 145                            when reading */
 146     int buf_index;
 147     int buf_size; /* 0 when writing */
 148     uint8_t buf[IO_BUF_SIZE];
 149
 150     int has_error;
 151 };
 152
 153 typedef struct QEMUFilePopen
 154 {
 155     FILE *popen_file;
 156     QEMUFile *file;
 157 } QEMUFilePopen;
 158
 159 typedef struct QEMUFileSocket
 160 {
 161     int fd;
 162     QEMUFile *file;
 163 } QEMUFileSocket;
 164
 165 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 166 {
 167     QEMUFileSocket *s = opaque;
 168     ssize_t len;
 169
 170     do {
 171         len = recv(s->fd, buf, size, 0);
 172     } while (len == -1 && socket_error() == EINTR);
 173
 174     if (len == -1)
 175         len = -socket_error();
 176
 177     return len;
 178 }
 179
 180 static int socket_close(void *opaque)
 181 {
 182     QEMUFileSocket *s = opaque;
 183     qemu_free(s);
 184     return 0;
 185 }
 186
 187 static int popen_put_buffer(void *opaque, const uint8_t *buf, int64_t pos, int size)
 188 {
 189     QEMUFilePopen *s = opaque;
 190     return fwrite(buf, 1, size, s->popen_file);
 191 }
 192
 193 static int popen_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 194 {
 195     QEMUFilePopen *s = opaque;
 196     return fread(buf, 1, size, s->popen_file);
 197 }
 198
 199 static int popen_close(void *opaque)
 200 {
 201     QEMUFilePopen *s = opaque;
 202     pclose(s->popen_file);
 203     qemu_free(s);
 204     return 0;
 205 }
 206
 207 QEMUFile *qemu_popen(FILE *popen_file, const char *mode)
 208 {
 209     QEMUFilePopen *s;
 210
 211     if (popen_file == NULL || mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
 212         fprintf(stderr, "qemu_popen: Argument validity check failed\n");
 213         return NULL;
 214     }
 215
 216     s = qemu_mallocz(sizeof(QEMUFilePopen));
 217
 218     s->popen_file = popen_file;
 219
 220     if(mode[0] == 'r') {
 221         s->file = qemu_fopen_ops(s, NULL, popen_get_buffer, popen_close, NULL);
 222     } else {
 223         s->file = qemu_fopen_ops(s, popen_put_buffer, NULL, popen_close, NULL);
 224     }
 225     fprintf(stderr, "qemu_popen: returning result of qemu_fopen_ops\n");
 226     return s->file;
 227 }
 228
 229 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
 230 {
 231     FILE *popen_file;
 232
 233     popen_file = popen(command, mode);
 234     if(popen_file == NULL) {
 235         return NULL;
 236     }
 237
 238     return qemu_popen(popen_file, mode);
 239 }
 240
 241 QEMUFile *qemu_fopen_socket(int fd)
 242 {
 243     QEMUFileSocket *s = qemu_mallocz(sizeof(QEMUFileSocket));
 244
 245     s->fd = fd;
 246     s->file = qemu_fopen_ops(s, NULL, socket_get_buffer, socket_close, NULL);
 247     return s->file;
 248 }
 249
 250 typedef struct QEMUFileStdio
 251 {
 252     FILE *outfile;
 253 } QEMUFileStdio;
 254
 255 static int file_put_buffer(void *opaque, const uint8_t *buf,
 256                             int64_t pos, int size)
 257 {
 258     QEMUFileStdio *s = opaque;
 259     fseek(s->outfile, pos, SEEK_SET);
 260     fwrite(buf, 1, size, s->outfile);
 261     return size;
 262 }
 263
 264 static int file_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 265 {
 266     QEMUFileStdio *s = opaque;
 267     fseek(s->outfile, pos, SEEK_SET);
 268     return fread(buf, 1, size, s->outfile);
 269 }
 270
 271 static int file_close(void *opaque)
 272 {
 273     QEMUFileStdio *s = opaque;
 274     fclose(s->outfile);
 275     qemu_free(s);
 276     return 0;
 277 }
 278
 279 QEMUFile *qemu_fopen(const char *filename, const char *mode)
 280 {
 281     QEMUFileStdio *s;
 282
 283     s = qemu_mallocz(sizeof(QEMUFileStdio));
 284
 285     s->outfile = fopen(filename, mode);
 286     if (!s->outfile)
 287         goto fail;
 288
 289     if (!strcmp(mode, "wb"))
 290         return qemu_fopen_ops(s, file_put_buffer, NULL, file_close, NULL);
 291     else if (!strcmp(mode, "rb"))
 292         return qemu_fopen_ops(s, NULL, file_get_buffer, file_close, NULL);
 293
 294 fail:
 295     if (s->outfile)
 296         fclose(s->outfile);
 297     qemu_free(s);
 298     return NULL;
 299 }
 300
 301 typedef struct QEMUFileBdrv
 302 {
 303     BlockDriverState *bs;
 304     int64_t base_offset;
 305 } QEMUFileBdrv;
 306
 307 static int bdrv_put_buffer(void *opaque, const uint8_t *buf,
 308                            int64_t pos, int size)
 309 {
 310     QEMUFileBdrv *s = opaque;
 311     bdrv_pwrite(s->bs, s->base_offset + pos, buf, size);
 312     return size;
 313 }
 314
 315 static int bdrv_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
 316 {
 317     QEMUFileBdrv *s = opaque;
 318     return bdrv_pread(s->bs, s->base_offset + pos, buf, size);
 319 }
 320
 321 static int bdrv_fclose(void *opaque)
 322 {
 323     QEMUFileBdrv *s = opaque;
 324     qemu_free(s);
 325     return 0;
 326 }
 327
 328 static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int64_t offset, int is_writable)
 329 {
 330     QEMUFileBdrv *s;
 331
 332     s = qemu_mallocz(sizeof(QEMUFileBdrv));
 333
 334     s->bs = bs;
 335     s->base_offset = offset;
 336
 337     if (is_writable)
 338         return qemu_fopen_ops(s, bdrv_put_buffer, NULL, bdrv_fclose, NULL);
 339
 340     return qemu_fopen_ops(s, NULL, bdrv_get_buffer, bdrv_fclose, NULL);
 341 }
 342
 343 QEMUFile *qemu_fopen_ops(void *opaque, QEMUFilePutBufferFunc *put_buffer,
 344                          QEMUFileGetBufferFunc *get_buffer,
 345                          QEMUFileCloseFunc *close,
 346                          QEMUFileRateLimit *rate_limit)
 347 {
 348     QEMUFile *f;
 349
 350     f = qemu_mallocz(sizeof(QEMUFile));
 351
 352     f->opaque = opaque;
 353     f->put_buffer = put_buffer;
 354     f->get_buffer = get_buffer;
 355     f->close = close;
 356     f->rate_limit = rate_limit;
 357     f->is_write = 0;
 358
 359     return f;
 360 }
 361
 362 int qemu_file_has_error(QEMUFile *f)
 363 {
 364     return f->has_error;
 365 }
 366
 367 void qemu_fflush(QEMUFile *f)
 368 {
 369     if (!f->put_buffer)
 370         return;
 371
 372     if (f->is_write && f->buf_index > 0) {
 373         int len;
 374
 375         len = f->put_buffer(f->opaque, f->buf, f->buf_offset, f->buf_index);
 376         if (len > 0)
 377             f->buf_offset += f->buf_index;
 378         else
 379             f->has_error = 1;
 380         f->buf_index = 0;
 381     }
 382 }
 383
 384 static void qemu_fill_buffer(QEMUFile *f)
 385 {
 386     int len;
 387
 388     if (!f->get_buffer)
 389         return;
 390
 391     if (f->is_write)
 392         abort();
 393
 394     len = f->get_buffer(f->opaque, f->buf, f->buf_offset, IO_BUF_SIZE);
 395     if (len > 0) {
 396         f->buf_index = 0;
 397         f->buf_size = len;
 398         f->buf_offset += len;
 399     } else if (len != -EAGAIN)
 400         f->has_error = 1;
 401 }
 402
 403 int qemu_fclose(QEMUFile *f)
 404 {
 405     int ret = 0;
 406     qemu_fflush(f);
 407     if (f->close)
 408         ret = f->close(f->opaque);
 409     qemu_free(f);
 410     return ret;
 411 }
 412
 413 void qemu_file_put_notify(QEMUFile *f)
 414 {
 415     f->put_buffer(f->opaque, NULL, 0, 0);
 416 }
 417
 418 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
 419 {
 420     int l;
 421
 422     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 423         fprintf(stderr,
 424                 "Attempted to write to buffer while read buffer is not empty\n");
 425         abort();
 426     }
 427
 428     while (!f->has_error && size > 0) {
 429         l = IO_BUF_SIZE - f->buf_index;
 430         if (l > size)
 431             l = size;
 432         memcpy(f->buf + f->buf_index, buf, l);
 433         f->is_write = 1;
 434         f->buf_index += l;
 435         buf += l;
 436         size -= l;
 437         if (f->buf_index >= IO_BUF_SIZE)
 438             qemu_fflush(f);
 439     }
 440 }
 441
 442 void qemu_put_byte(QEMUFile *f, int v)
 443 {
 444     if (!f->has_error && f->is_write == 0 && f->buf_index > 0) {
 445         fprintf(stderr,
 446                 "Attempted to write to buffer while read buffer is not empty\n");
 447         abort();
 448     }
 449
 450     f->buf[f->buf_index++] = v;
 451     f->is_write = 1;
 452     if (f->buf_index >= IO_BUF_SIZE)
 453         qemu_fflush(f);
 454 }
 455
 456 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size1)
 457 {
 458     int size, l;
 459
 460     if (f->is_write)
 461         abort();
 462
 463     size = size1;
 464     while (size > 0) {
 465         l = f->buf_size - f->buf_index;
 466         if (l == 0) {
 467             qemu_fill_buffer(f);
 468             l = f->buf_size - f->buf_index;
 469             if (l == 0)
 470                 break;
 471         }
 472         if (l > size)
 473             l = size;
 474         memcpy(buf, f->buf + f->buf_index, l);
 475         f->buf_index += l;
 476         buf += l;
 477         size -= l;
 478     }
 479     return size1 - size;
 480 }
 481
 482 int qemu_get_byte(QEMUFile *f)
 483 {
 484     if (f->is_write)
 485         abort();
 486
 487     if (f->buf_index >= f->buf_size) {
 488         qemu_fill_buffer(f);
 489         if (f->buf_index >= f->buf_size)
 490             return 0;
 491     }
 492     return f->buf[f->buf_index++];
 493 }
 494
 495 int64_t qemu_ftell(QEMUFile *f)
 496 {
 497     return f->buf_offset - f->buf_size + f->buf_index;
 498 }
 499
 500 int64_t qemu_fseek(QEMUFile *f, int64_t pos, int whence)
 501 {
 502     if (whence == SEEK_SET) {
 503         /* nothing to do */
 504     } else if (whence == SEEK_CUR) {
 505         pos += qemu_ftell(f);
 506     } else {
 507         /* SEEK_END not supported */
 508         return -1;
 509     }
 510     if (f->put_buffer) {
 511         qemu_fflush(f);
 512         f->buf_offset = pos;
 513     } else {
 514         f->buf_offset = pos;
 515         f->buf_index = 0;
 516         f->buf_size = 0;
 517     }
 518     return pos;
 519 }
 520
 521 int qemu_file_rate_limit(QEMUFile *f)
 522 {
 523     if (f->rate_limit)
 524         return f->rate_limit(f->opaque);
 525
 526     return 0;
 527 }
 528
 529 void qemu_put_be16(QEMUFile *f, unsigned int v)
 530 {
 531     qemu_put_byte(f, v >> 8);
 532     qemu_put_byte(f, v);
 533 }
 534
 535 void qemu_put_be32(QEMUFile *f, unsigned int v)
 536 {
 537     qemu_put_byte(f, v >> 24);
 538     qemu_put_byte(f, v >> 16);
 539     qemu_put_byte(f, v >> 8);
 540     qemu_put_byte(f, v);
 541 }
 542
 543 void qemu_put_be64(QEMUFile *f, uint64_t v)
 544 {
 545     qemu_put_be32(f, v >> 32);
 546     qemu_put_be32(f, v);
 547 }
 548
 549 unsigned int qemu_get_be16(QEMUFile *f)
 550 {
 551     unsigned int v;
 552     v = qemu_get_byte(f) << 8;
 553     v |= qemu_get_byte(f);
 554     return v;
 555 }
 556
 557 unsigned int qemu_get_be32(QEMUFile *f)
 558 {
 559     unsigned int v;
 560     v = qemu_get_byte(f) << 24;
 561     v |= qemu_get_byte(f) << 16;
 562     v |= qemu_get_byte(f) << 8;
 563     v |= qemu_get_byte(f);
 564     return v;
 565 }
 566
 567 uint64_t qemu_get_be64(QEMUFile *f)
 568 {
 569     uint64_t v;
 570     v = (uint64_t)qemu_get_be32(f) << 32;
 571     v |= qemu_get_be32(f);
 572     return v;
 573 }
 574
 575 typedef struct SaveStateEntry {
 576     char idstr[256];
 577     int instance_id;
 578     int version_id;
 579     int section_id;
 580     SaveLiveStateHandler *save_live_state;
 581     SaveStateHandler *save_state;
 582     LoadStateHandler *load_state;
 583     void *opaque;
 584     struct SaveStateEntry *next;
 585 } SaveStateEntry;
 586
 587 static SaveStateEntry *first_se;
 588
 589 /* TODO: Individual devices generally have very little idea about the rest
 590    of the system, so instance_id should be removed/replaced.
 591    Meanwhile pass -1 as instance_id if you do not already have a clearly
 592    distinguishing id for all instances of your device class. */
 593 int register_savevm_live(const char *idstr,
 594                          int instance_id,
 595                          int version_id,
 596                          SaveLiveStateHandler *save_live_state,
 597                          SaveStateHandler *save_state,
 598                          LoadStateHandler *load_state,
 599                          void *opaque)
 600 {
 601     SaveStateEntry *se, **pse;
 602     static int global_section_id;
 603
 604     se = qemu_malloc(sizeof(SaveStateEntry));
 605     pstrcpy(se->idstr, sizeof(se->idstr), idstr);
 606     se->instance_id = (instance_id == -1) ? 0 : instance_id;
 607     se->version_id = version_id;
 608     se->section_id = global_section_id++;
 609     se->save_live_state = save_live_state;
 610     se->save_state = save_state;
 611     se->load_state = load_state;
 612     se->opaque = opaque;
 613     se->next = NULL;
 614
 615     /* add at the end of list */
 616     pse = &first_se;
 617     while (*pse != NULL) {
 618         if (instance_id == -1
 619                 && strcmp(se->idstr, (*pse)->idstr) == 0
 620                 && se->instance_id <= (*pse)->instance_id)
 621             se->instance_id = (*pse)->instance_id + 1;
 622         pse = &(*pse)->next;
 623     }
 624     *pse = se;
 625     return 0;
 626 }
 627
 628 int register_savevm(const char *idstr,
 629                     int instance_id,
 630                     int version_id,
 631                     SaveStateHandler *save_state,
 632                     LoadStateHandler *load_state,
 633                     void *opaque)
 634 {
 635     return register_savevm_live(idstr, instance_id, version_id,
 636                                 NULL, save_state, load_state, opaque);
 637 }
 638
 639 #define QEMU_VM_FILE_MAGIC           0x5145564d
 640 #define QEMU_VM_FILE_VERSION_COMPAT  0x00000002
 641 #define QEMU_VM_FILE_VERSION         0x00000003
 642
 643 #define QEMU_VM_EOF                  0x00
 644 #define QEMU_VM_SECTION_START        0x01
 645 #define QEMU_VM_SECTION_PART         0x02
 646 #define QEMU_VM_SECTION_END          0x03
 647 #define QEMU_VM_SECTION_FULL         0x04
 648
 649 int qemu_savevm_state_begin(QEMUFile *f)
 650 {
 651     SaveStateEntry *se;
 652
 653     qemu_put_be32(f, QEMU_VM_FILE_MAGIC);
 654     qemu_put_be32(f, QEMU_VM_FILE_VERSION);
 655
 656     for (se = first_se; se != NULL; se = se->next) {
 657         int len;
 658
 659         if (se->save_live_state == NULL)
 660             continue;
 661
 662         /* Section type */
 663         qemu_put_byte(f, QEMU_VM_SECTION_START);
 664         qemu_put_be32(f, se->section_id);
 665
 666         /* ID string */
 667         len = strlen(se->idstr);
 668         qemu_put_byte(f, len);
 669         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 670
 671         qemu_put_be32(f, se->instance_id);
 672         qemu_put_be32(f, se->version_id);
 673
 674         se->save_live_state(f, QEMU_VM_SECTION_START, se->opaque);
 675     }
 676
 677     if (qemu_file_has_error(f))
 678         return -EIO;
 679
 680     return 0;
 681 }
 682
 683 int qemu_savevm_state_iterate(QEMUFile *f)
 684 {
 685     SaveStateEntry *se;
 686     int ret = 1;
 687
 688     for (se = first_se; se != NULL; se = se->next) {
 689         if (se->save_live_state == NULL)
 690             continue;
 691
 692         /* Section type */
 693         qemu_put_byte(f, QEMU_VM_SECTION_PART);
 694         qemu_put_be32(f, se->section_id);
 695
 696         ret &= !!se->save_live_state(f, QEMU_VM_SECTION_PART, se->opaque);
 697     }
 698
 699     if (ret)
 700         return 1;
 701
 702     if (qemu_file_has_error(f))
 703         return -EIO;
 704
 705     return 0;
 706 }
 707
 708 int qemu_savevm_state_complete(QEMUFile *f)
 709 {
 710     SaveStateEntry *se;
 711
 712     for (se = first_se; se != NULL; se = se->next) {
 713         if (se->save_live_state == NULL)
 714             continue;
 715
 716         /* Section type */
 717         qemu_put_byte(f, QEMU_VM_SECTION_END);
 718         qemu_put_be32(f, se->section_id);
 719
 720         se->save_live_state(f, QEMU_VM_SECTION_END, se->opaque);
 721     }
 722
 723     for(se = first_se; se != NULL; se = se->next) {
 724         int len;
 725
 726         if (se->save_state == NULL)
 727             continue;
 728
 729         /* Section type */
 730         qemu_put_byte(f, QEMU_VM_SECTION_FULL);
 731         qemu_put_be32(f, se->section_id);
 732
 733         /* ID string */
 734         len = strlen(se->idstr);
 735         qemu_put_byte(f, len);
 736         qemu_put_buffer(f, (uint8_t *)se->idstr, len);
 737
 738         qemu_put_be32(f, se->instance_id);
 739         qemu_put_be32(f, se->version_id);
 740
 741         se->save_state(f, se->opaque);
 742     }
 743
 744     qemu_put_byte(f, QEMU_VM_EOF);
 745
 746     if (qemu_file_has_error(f))
 747         return -EIO;
 748
 749     return 0;
 750 }
 751
 752 int qemu_savevm_state(QEMUFile *f)
 753 {
 754     int saved_vm_running;
 755     int ret;
 756
 757     saved_vm_running = vm_running;
 758     vm_stop(0);
 759
 760     bdrv_flush_all();
 761
 762     ret = qemu_savevm_state_begin(f);
 763     if (ret < 0)
 764         goto out;
 765
 766     do {
 767         ret = qemu_savevm_state_iterate(f);
 768         if (ret < 0)
 769             goto out;
 770     } while (ret == 0);
 771
 772     ret = qemu_savevm_state_complete(f);
 773
 774 out:
 775     if (qemu_file_has_error(f))
 776         ret = -EIO;
 777
 778     if (!ret && saved_vm_running)
 779         vm_start();
 780
 781     return ret;
 782 }
 783
 784 static SaveStateEntry *find_se(const char *idstr, int instance_id)
 785 {
 786     SaveStateEntry *se;
 787
 788     for(se = first_se; se != NULL; se = se->next) {
 789         if (!strcmp(se->idstr, idstr) &&
 790             instance_id == se->instance_id)
 791             return se;
 792     }
 793     return NULL;
 794 }
 795
 796 typedef struct LoadStateEntry {
 797     SaveStateEntry *se;
 798     int section_id;
 799     int version_id;
 800     struct LoadStateEntry *next;
 801 } LoadStateEntry;
 802
 803 static int qemu_loadvm_state_v2(QEMUFile *f)
 804 {
 805     SaveStateEntry *se;
 806     int len, ret, instance_id, record_len, version_id;
 807     int64_t total_len, end_pos, cur_pos;
 808     char idstr[256];
 809
 810     total_len = qemu_get_be64(f);
 811     end_pos = total_len + qemu_ftell(f);
 812     for(;;) {
 813         if (qemu_ftell(f) >= end_pos)
 814             break;
 815         len = qemu_get_byte(f);
 816         qemu_get_buffer(f, (uint8_t *)idstr, len);
 817         idstr[len] = '\0';
 818         instance_id = qemu_get_be32(f);
 819         version_id = qemu_get_be32(f);
 820         record_len = qemu_get_be32(f);
 821         cur_pos = qemu_ftell(f);
 822         se = find_se(idstr, instance_id);
 823         if (!se) {
 824             fprintf(stderr, "qemu: warning: instance 0x%x of device '%s' not present in current VM\n",
 825                     instance_id, idstr);
 826         } else {
 827             ret = se->load_state(f, se->opaque, version_id);
 828             if (ret < 0) {
 829                 fprintf(stderr, "qemu: warning: error while loading state for instance 0x%x of device '%s'\n",
 830                         instance_id, idstr);
 831             }
 832         }
 833         /* always seek to exact end of record */
 834         qemu_fseek(f, cur_pos + record_len, SEEK_SET);
 835     }
 836
 837     if (qemu_file_has_error(f))
 838         return -EIO;
 839
 840     return 0;
 841 }
 842
 843 int qemu_loadvm_state(QEMUFile *f)
 844 {
 845     LoadStateEntry *first_le = NULL;
 846     uint8_t section_type;
 847     unsigned int v;
 848     int ret;
 849
 850     v = qemu_get_be32(f);
 851     if (v != QEMU_VM_FILE_MAGIC)
 852         return -EINVAL;
 853
 854     v = qemu_get_be32(f);
 855     if (v == QEMU_VM_FILE_VERSION_COMPAT)
 856         return qemu_loadvm_state_v2(f);
 857     if (v != QEMU_VM_FILE_VERSION)
 858         return -ENOTSUP;
 859
 860     while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) {
 861         uint32_t instance_id, version_id, section_id;
 862         LoadStateEntry *le;
 863         SaveStateEntry *se;
 864         char idstr[257];
 865         int len;
 866
 867         switch (section_type) {
 868         case QEMU_VM_SECTION_START:
 869         case QEMU_VM_SECTION_FULL:
 870             /* Read section start */
 871             section_id = qemu_get_be32(f);
 872             len = qemu_get_byte(f);
 873             qemu_get_buffer(f, (uint8_t *)idstr, len);
 874             idstr[len] = 0;
 875             instance_id = qemu_get_be32(f);
 876             version_id = qemu_get_be32(f);
 877
 878             /* Find savevm section */
 879             se = find_se(idstr, instance_id);
 880             if (se == NULL) {
 881                 fprintf(stderr, "Unknown savevm section or instance '%s' %d\n", idstr, instance_id);
 882                 ret = -EINVAL;
 883                 goto out;
 884             }
 885
 886             /* Validate version */
 887             if (version_id > se->version_id) {
 888                 fprintf(stderr, "savevm: unsupported version %d for '%s' v%d\n",
 889                         version_id, idstr, se->version_id);
 890                 ret = -EINVAL;
 891                 goto out;
 892             }
 893
 894             /* Add entry */
 895             le = qemu_mallocz(sizeof(*le));
 896
 897             le->se = se;
 898             le->section_id = section_id;
 899             le->version_id = version_id;
 900             le->next = first_le;
 901             first_le = le;
 902
 903             le->se->load_state(f, le->se->opaque, le->version_id);
 904             break;
 905         case QEMU_VM_SECTION_PART:
 906         case QEMU_VM_SECTION_END:
 907             section_id = qemu_get_be32(f);
 908
 909             for (le = first_le; le && le->section_id != section_id; le = le->next);
 910             if (le == NULL) {
 911                 fprintf(stderr, "Unknown savevm section %d\n", section_id);
 912                 ret = -EINVAL;
 913                 goto out;
 914             }
 915
 916             le->se->load_state(f, le->se->opaque, le->version_id);
 917             break;
 918         default:
 919             fprintf(stderr, "Unknown savevm section type %d\n", section_type);
 920             ret = -EINVAL;
 921             goto out;
 922         }
 923     }
 924
 925     ret = 0;
 926
 927 out:
 928     while (first_le) {
 929         LoadStateEntry *le = first_le;
 930         first_le = first_le->next;
 931         qemu_free(le);
 932     }
 933
 934     if (qemu_file_has_error(f))
 935         ret = -EIO;
 936
 937     return ret;
 938 }
 939
 940 /* device can contain snapshots */
 941 static int bdrv_can_snapshot(BlockDriverState *bs)
 942 {
 943     return (bs &&
 944             !bdrv_is_removable(bs) &&
 945             !bdrv_is_read_only(bs));
 946 }
 947
 948 /* device must be snapshots in order to have a reliable snapshot */
 949 static int bdrv_has_snapshot(BlockDriverState *bs)
 950 {
 951     return (bs &&
 952             !bdrv_is_removable(bs) &&
 953             !bdrv_is_read_only(bs));
 954 }
 955
 956 static BlockDriverState *get_bs_snapshots(void)
 957 {
 958     BlockDriverState *bs;
 959     int i;
 960
 961     if (bs_snapshots)
 962         return bs_snapshots;
 963     for(i = 0; i <= nb_drives; i++) {
 964         bs = drives_table[i].bdrv;
 965         if (bdrv_can_snapshot(bs))
 966             goto ok;
 967     }
 968     return NULL;
 969  ok:
 970     bs_snapshots = bs;
 971     return bs;
 972 }
 973
 974 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
 975                               const char *name)
 976 {
 977     QEMUSnapshotInfo *sn_tab, *sn;
 978     int nb_sns, i, ret;
 979
 980     ret = -ENOENT;
 981     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
 982     if (nb_sns < 0)
 983         return ret;
 984     for(i = 0; i < nb_sns; i++) {
 985         sn = &sn_tab[i];
 986         if (!strcmp(sn->id_str, name) || !strcmp(sn->name, name)) {
 987             *sn_info = *sn;
 988             ret = 0;
 989             break;
 990         }
 991     }
 992     qemu_free(sn_tab);
 993     return ret;
 994 }
 995
 996 void do_savevm(Monitor *mon, const char *name)
 997 {
 998     BlockDriverState *bs, *bs1;
 999     QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1;
1000     int must_delete, ret, i;
1001     BlockDriverInfo bdi1, *bdi = &bdi1;
1002     QEMUFile *f;
1003     int saved_vm_running;
1004     uint32_t vm_state_size;
1005 #ifdef _WIN32
1006     struct _timeb tb;
1007 #else
1008     struct timeval tv;
1009 #endif
1010
1011     bs = get_bs_snapshots();
1012     if (!bs) {
1013         monitor_printf(mon, "No block device can accept snapshots\n");
1014         return;
1015     }
1016
1017     /* ??? Should this occur after vm_stop?  */
1018     qemu_aio_flush();
1019
1020     saved_vm_running = vm_running;
1021     vm_stop(0);
1022
1023     must_delete = 0;
1024     if (name) {
1025         ret = bdrv_snapshot_find(bs, old_sn, name);
1026         if (ret >= 0) {
1027             must_delete = 1;
1028         }
1029     }
1030     memset(sn, 0, sizeof(*sn));
1031     if (must_delete) {
1032         pstrcpy(sn->name, sizeof(sn->name), old_sn->name);
1033         pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str);
1034     } else {
1035         if (name)
1036             pstrcpy(sn->name, sizeof(sn->name), name);
1037     }
1038
1039     /* fill auxiliary fields */
1040 #ifdef _WIN32
1041     _ftime(&tb);
1042     sn->date_sec = tb.time;
1043     sn->date_nsec = tb.millitm * 1000000;
1044 #else
1045     gettimeofday(&tv, NULL);
1046     sn->date_sec = tv.tv_sec;
1047     sn->date_nsec = tv.tv_usec * 1000;
1048 #endif
1049     sn->vm_clock_nsec = qemu_get_clock(vm_clock);
1050
1051     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1052         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1053                        bdrv_get_device_name(bs));
1054         goto the_end;
1055     }
1056
1057     /* save the VM state */
1058     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 1);
1059     if (!f) {
1060         monitor_printf(mon, "Could not open VM state file\n");
1061         goto the_end;
1062     }
1063     ret = qemu_savevm_state(f);
1064     vm_state_size = qemu_ftell(f);
1065     qemu_fclose(f);
1066     if (ret < 0) {
1067         monitor_printf(mon, "Error %d while writing VM\n", ret);
1068         goto the_end;
1069     }
1070
1071     /* create the snapshots */
1072
1073     for(i = 0; i < nb_drives; i++) {
1074         bs1 = drives_table[i].bdrv;
1075         if (bdrv_has_snapshot(bs1)) {
1076             if (must_delete) {
1077                 ret = bdrv_snapshot_delete(bs1, old_sn->id_str);
1078                 if (ret < 0) {
1079                     monitor_printf(mon,
1080                                    "Error while deleting snapshot on '%s'\n",
1081                                    bdrv_get_device_name(bs1));
1082                 }
1083             }
1084             /* Write VM state size only to the image that contains the state */
1085             sn->vm_state_size = (bs == bs1 ? vm_state_size : 0);
1086             ret = bdrv_snapshot_create(bs1, sn);
1087             if (ret < 0) {
1088                 monitor_printf(mon, "Error while creating snapshot on '%s'\n",
1089                                bdrv_get_device_name(bs1));
1090             }
1091         }
1092     }
1093
1094  the_end:
1095     if (saved_vm_running)
1096         vm_start();
1097 }
1098
1099 void do_loadvm(Monitor *mon, const char *name)
1100 {
1101     BlockDriverState *bs, *bs1;
1102     BlockDriverInfo bdi1, *bdi = &bdi1;
1103     QEMUSnapshotInfo sn;
1104     QEMUFile *f;
1105     int i, ret;
1106     int saved_vm_running;
1107
1108     bs = get_bs_snapshots();
1109     if (!bs) {
1110         monitor_printf(mon, "No block device supports snapshots\n");
1111         return;
1112     }
1113
1114     /* Flush all IO requests so they don't interfere with the new state.  */
1115     qemu_aio_flush();
1116
1117     saved_vm_running = vm_running;
1118     vm_stop(0);
1119
1120     for(i = 0; i <= nb_drives; i++) {
1121         bs1 = drives_table[i].bdrv;
1122         if (bdrv_has_snapshot(bs1)) {
1123             ret = bdrv_snapshot_goto(bs1, name);
1124             if (ret < 0) {
1125                 if (bs != bs1)
1126                     monitor_printf(mon, "Warning: ");
1127                 switch(ret) {
1128                 case -ENOTSUP:
1129                     monitor_printf(mon,
1130                                    "Snapshots not supported on device '%s'\n",
1131                                    bdrv_get_device_name(bs1));
1132                     break;
1133                 case -ENOENT:
1134                     monitor_printf(mon, "Could not find snapshot '%s' on "
1135                                    "device '%s'\n",
1136                                    name, bdrv_get_device_name(bs1));
1137                     break;
1138                 default:
1139                     monitor_printf(mon, "Error %d while activating snapshot on"
1140                                    " '%s'\n", ret, bdrv_get_device_name(bs1));
1141                     break;
1142                 }
1143                 /* fatal on snapshot block device */
1144                 if (bs == bs1)
1145                     goto the_end;
1146             }
1147         }
1148     }
1149
1150     if (bdrv_get_info(bs, bdi) < 0 || bdi->vm_state_offset <= 0) {
1151         monitor_printf(mon, "Device %s does not support VM state snapshots\n",
1152                        bdrv_get_device_name(bs));
1153         return;
1154     }
1155
1156     /* Don't even try to load empty VM states */
1157     ret = bdrv_snapshot_find(bs, &sn, name);
1158     if ((ret >= 0) && (sn.vm_state_size == 0))
1159         goto the_end;
1160
1161     /* restore the VM state */
1162     f = qemu_fopen_bdrv(bs, bdi->vm_state_offset, 0);
1163     if (!f) {
1164         monitor_printf(mon, "Could not open VM state file\n");
1165         goto the_end;
1166     }
1167     ret = qemu_loadvm_state(f);
1168     qemu_fclose(f);
1169     if (ret < 0) {
1170         monitor_printf(mon, "Error %d while loading VM state\n", ret);
1171     }
1172  the_end:
1173     if (saved_vm_running)
1174         vm_start();
1175 }
1176
1177 void do_delvm(Monitor *mon, const char *name)
1178 {
1179     BlockDriverState *bs, *bs1;
1180     int i, ret;
1181
1182     bs = get_bs_snapshots();
1183     if (!bs) {
1184         monitor_printf(mon, "No block device supports snapshots\n");
1185         return;
1186     }
1187
1188     for(i = 0; i <= nb_drives; i++) {
1189         bs1 = drives_table[i].bdrv;
1190         if (bdrv_has_snapshot(bs1)) {
1191             ret = bdrv_snapshot_delete(bs1, name);
1192             if (ret < 0) {
1193                 if (ret == -ENOTSUP)
1194                     monitor_printf(mon,
1195                                    "Snapshots not supported on device '%s'\n",
1196                                    bdrv_get_device_name(bs1));
1197                 else
1198                     monitor_printf(mon, "Error %d while deleting snapshot on "
1199                                    "'%s'\n", ret, bdrv_get_device_name(bs1));
1200             }
1201         }
1202     }
1203 }
1204
1205 void do_info_snapshots(Monitor *mon)
1206 {
1207     BlockDriverState *bs, *bs1;
1208     QEMUSnapshotInfo *sn_tab, *sn;
1209     int nb_sns, i;
1210     char buf[256];
1211
1212     bs = get_bs_snapshots();
1213     if (!bs) {
1214         monitor_printf(mon, "No available block device supports snapshots\n");
1215         return;
1216     }
1217     monitor_printf(mon, "Snapshot devices:");
1218     for(i = 0; i <= nb_drives; i++) {
1219         bs1 = drives_table[i].bdrv;
1220         if (bdrv_has_snapshot(bs1)) {
1221             if (bs == bs1)
1222                 monitor_printf(mon, " %s", bdrv_get_device_name(bs1));
1223         }
1224     }
1225     monitor_printf(mon, "\n");
1226
1227     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1228     if (nb_sns < 0) {
1229         monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns);
1230         return;
1231     }
1232     monitor_printf(mon, "Snapshot list (from %s):\n",
1233                    bdrv_get_device_name(bs));
1234     monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), NULL));
1235     for(i = 0; i < nb_sns; i++) {
1236         sn = &sn_tab[i];
1237         monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf), sn));
1238     }
1239     qemu_free(sn_tab);
1240 }