migration/qemu-file.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/osdep.h"
  25 #include <zlib.h>
  26 #include "qemu-common.h"
  27 #include "qemu/error-report.h"
  28 #include "qemu/iov.h"
  29 #include "qemu/sockets.h"
  30 #include "qemu/coroutine.h"
  31 #include "migration/migration.h"
  32 #include "migration/qemu-file.h"
  33 #include "migration/qemu-file-internal.h"
  34 #include "trace.h"
  35
  36 /*
  37  * Stop a file from being read/written - not all backing files can do this
  38  * typically only sockets can.
  39  */
  40 int qemu_file_shutdown(QEMUFile *f)
  41 {
  42     if (!f->ops->shut_down) {
  43         return -ENOSYS;
  44     }
  45     return f->ops->shut_down(f->opaque, true, true);
  46 }
  47
  48 /*
  49  * Result: QEMUFile* for a 'return path' for comms in the opposite direction
  50  *         NULL if not available
  51  */
  52 QEMUFile *qemu_file_get_return_path(QEMUFile *f)
  53 {
  54     if (!f->ops->get_return_path) {
  55         return NULL;
  56     }
  57     return f->ops->get_return_path(f->opaque);
  58 }
  59
  60 bool qemu_file_mode_is_not_valid(const char *mode)
  61 {
  62     if (mode == NULL ||
  63         (mode[0] != 'r' && mode[0] != 'w') ||
  64         mode[1] != 'b' || mode[2] != 0) {
  65         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
  66         return true;
  67     }
  68
  69     return false;
  70 }
  71
  72 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
  73 {
  74     QEMUFile *f;
  75
  76     f = g_new0(QEMUFile, 1);
  77
  78     f->opaque = opaque;
  79     f->ops = ops;
  80     return f;
  81 }
  82
  83 /*
  84  * Get last error for stream f
  85  *
  86  * Return negative error value if there has been an error on previous
  87  * operations, return 0 if no error happened.
  88  *
  89  */
  90 int qemu_file_get_error(QEMUFile *f)
  91 {
  92     return f->last_error;
  93 }
  94
  95 void qemu_file_set_error(QEMUFile *f, int ret)
  96 {
  97     if (f->last_error == 0) {
  98         f->last_error = ret;
  99     }
 100 }
 101
 102 bool qemu_file_is_writable(QEMUFile *f)
 103 {
 104     return f->ops->writev_buffer || f->ops->put_buffer;
 105 }
 106
 107 /**
 108  * Flushes QEMUFile buffer
 109  *
 110  * If there is writev_buffer QEMUFileOps it uses it otherwise uses
 111  * put_buffer ops.
 112  */
 113 void qemu_fflush(QEMUFile *f)
 114 {
 115     ssize_t ret = 0;
 116
 117     if (!qemu_file_is_writable(f)) {
 118         return;
 119     }
 120
 121     if (f->ops->writev_buffer) {
 122         if (f->iovcnt > 0) {
 123             ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt, f->pos);
 124         }
 125     } else {
 126         if (f->buf_index > 0) {
 127             ret = f->ops->put_buffer(f->opaque, f->buf, f->pos, f->buf_index);
 128         }
 129     }
 130     if (ret >= 0) {
 131         f->pos += ret;
 132     }
 133     f->buf_index = 0;
 134     f->iovcnt = 0;
 135     if (ret < 0) {
 136         qemu_file_set_error(f, ret);
 137     }
 138 }
 139
 140 void ram_control_before_iterate(QEMUFile *f, uint64_t flags)
 141 {
 142     int ret = 0;
 143
 144     if (f->ops->before_ram_iterate) {
 145         ret = f->ops->before_ram_iterate(f, f->opaque, flags, NULL);
 146         if (ret < 0) {
 147             qemu_file_set_error(f, ret);
 148         }
 149     }
 150 }
 151
 152 void ram_control_after_iterate(QEMUFile *f, uint64_t flags)
 153 {
 154     int ret = 0;
 155
 156     if (f->ops->after_ram_iterate) {
 157         ret = f->ops->after_ram_iterate(f, f->opaque, flags, NULL);
 158         if (ret < 0) {
 159             qemu_file_set_error(f, ret);
 160         }
 161     }
 162 }
 163
 164 void ram_control_load_hook(QEMUFile *f, uint64_t flags, void *data)
 165 {
 166     int ret = -EINVAL;
 167
 168     if (f->ops->hook_ram_load) {
 169         ret = f->ops->hook_ram_load(f, f->opaque, flags, data);
 170         if (ret < 0) {
 171             qemu_file_set_error(f, ret);
 172         }
 173     } else {
 174         /*
 175          * Hook is a hook specifically requested by the source sending a flag
 176          * that expects there to be a hook on the destination.
 177          */
 178         if (flags == RAM_CONTROL_HOOK) {
 179             qemu_file_set_error(f, ret);
 180         }
 181     }
 182 }
 183
 184 size_t ram_control_save_page(QEMUFile *f, ram_addr_t block_offset,
 185                              ram_addr_t offset, size_t size,
 186                              uint64_t *bytes_sent)
 187 {
 188     if (f->ops->save_page) {
 189         int ret = f->ops->save_page(f, f->opaque, block_offset,
 190                                     offset, size, bytes_sent);
 191
 192         if (ret != RAM_SAVE_CONTROL_DELAYED) {
 193             if (bytes_sent && *bytes_sent > 0) {
 194                 qemu_update_position(f, *bytes_sent);
 195             } else if (ret < 0) {
 196                 qemu_file_set_error(f, ret);
 197             }
 198         }
 199
 200         return ret;
 201     }
 202
 203     return RAM_SAVE_CONTROL_NOT_SUPP;
 204 }
 205
 206 /*
 207  * Attempt to fill the buffer from the underlying file
 208  * Returns the number of bytes read, or negative value for an error.
 209  *
 210  * Note that it can return a partially full buffer even in a not error/not EOF
 211  * case if the underlying file descriptor gives a short read, and that can
 212  * happen even on a blocking fd.
 213  */
 214 static ssize_t qemu_fill_buffer(QEMUFile *f)
 215 {
 216     int len;
 217     int pending;
 218
 219     assert(!qemu_file_is_writable(f));
 220
 221     pending = f->buf_size - f->buf_index;
 222     if (pending > 0) {
 223         memmove(f->buf, f->buf + f->buf_index, pending);
 224     }
 225     f->buf_index = 0;
 226     f->buf_size = pending;
 227
 228     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->pos,
 229                         IO_BUF_SIZE - pending);
 230     if (len > 0) {
 231         f->buf_size += len;
 232         f->pos += len;
 233     } else if (len == 0) {
 234         qemu_file_set_error(f, -EIO);
 235     } else if (len != -EAGAIN) {
 236         qemu_file_set_error(f, len);
 237     }
 238
 239     return len;
 240 }
 241
 242 int qemu_get_fd(QEMUFile *f)
 243 {
 244     if (f->ops->get_fd) {
 245         return f->ops->get_fd(f->opaque);
 246     }
 247     return -1;
 248 }
 249
 250 void qemu_update_position(QEMUFile *f, size_t size)
 251 {
 252     f->pos += size;
 253 }
 254
 255 /** Closes the file
 256  *
 257  * Returns negative error value if any error happened on previous operations or
 258  * while closing the file. Returns 0 or positive number on success.
 259  *
 260  * The meaning of return value on success depends on the specific backend
 261  * being used.
 262  */
 263 int qemu_fclose(QEMUFile *f)
 264 {
 265     int ret;
 266     qemu_fflush(f);
 267     ret = qemu_file_get_error(f);
 268
 269     if (f->ops->close) {
 270         int ret2 = f->ops->close(f->opaque);
 271         if (ret >= 0) {
 272             ret = ret2;
 273         }
 274     }
 275     /* If any error was spotted before closing, we should report it
 276      * instead of the close() return value.
 277      */
 278     if (f->last_error) {
 279         ret = f->last_error;
 280     }
 281     g_free(f);
 282     trace_qemu_file_fclose();
 283     return ret;
 284 }
 285
 286 static void add_to_iovec(QEMUFile *f, const uint8_t *buf, size_t size)
 287 {
 288     /* check for adjacent buffer and coalesce them */
 289     if (f->iovcnt > 0 && buf == f->iov[f->iovcnt - 1].iov_base +
 290         f->iov[f->iovcnt - 1].iov_len) {
 291         f->iov[f->iovcnt - 1].iov_len += size;
 292     } else {
 293         f->iov[f->iovcnt].iov_base = (uint8_t *)buf;
 294         f->iov[f->iovcnt++].iov_len = size;
 295     }
 296
 297     if (f->iovcnt >= MAX_IOV_SIZE) {
 298         qemu_fflush(f);
 299     }
 300 }
 301
 302 void qemu_put_buffer_async(QEMUFile *f, const uint8_t *buf, size_t size)
 303 {
 304     if (!f->ops->writev_buffer) {
 305         qemu_put_buffer(f, buf, size);
 306         return;
 307     }
 308
 309     if (f->last_error) {
 310         return;
 311     }
 312
 313     f->bytes_xfer += size;
 314     add_to_iovec(f, buf, size);
 315 }
 316
 317 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, size_t size)
 318 {
 319     size_t l;
 320
 321     if (f->last_error) {
 322         return;
 323     }
 324
 325     while (size > 0) {
 326         l = IO_BUF_SIZE - f->buf_index;
 327         if (l > size) {
 328             l = size;
 329         }
 330         memcpy(f->buf + f->buf_index, buf, l);
 331         f->bytes_xfer += l;
 332         if (f->ops->writev_buffer) {
 333             add_to_iovec(f, f->buf + f->buf_index, l);
 334         }
 335         f->buf_index += l;
 336         if (f->buf_index == IO_BUF_SIZE) {
 337             qemu_fflush(f);
 338         }
 339         if (qemu_file_get_error(f)) {
 340             break;
 341         }
 342         buf += l;
 343         size -= l;
 344     }
 345 }
 346
 347 void qemu_put_byte(QEMUFile *f, int v)
 348 {
 349     if (f->last_error) {
 350         return;
 351     }
 352
 353     f->buf[f->buf_index] = v;
 354     f->bytes_xfer++;
 355     if (f->ops->writev_buffer) {
 356         add_to_iovec(f, f->buf + f->buf_index, 1);
 357     }
 358     f->buf_index++;
 359     if (f->buf_index == IO_BUF_SIZE) {
 360         qemu_fflush(f);
 361     }
 362 }
 363
 364 void qemu_file_skip(QEMUFile *f, int size)
 365 {
 366     if (f->buf_index + size <= f->buf_size) {
 367         f->buf_index += size;
 368     }
 369 }
 370
 371 /*
 372  * Read 'size' bytes from file (at 'offset') without moving the
 373  * pointer and set 'buf' to point to that data.
 374  *
 375  * It will return size bytes unless there was an error, in which case it will
 376  * return as many as it managed to read (assuming blocking fd's which
 377  * all current QEMUFile are)
 378  */
 379 size_t qemu_peek_buffer(QEMUFile *f, uint8_t **buf, size_t size, size_t offset)
 380 {
 381     ssize_t pending;
 382     size_t index;
 383
 384     assert(!qemu_file_is_writable(f));
 385     assert(offset < IO_BUF_SIZE);
 386     assert(size <= IO_BUF_SIZE - offset);
 387
 388     /* The 1st byte to read from */
 389     index = f->buf_index + offset;
 390     /* The number of available bytes starting at index */
 391     pending = f->buf_size - index;
 392
 393     /*
 394      * qemu_fill_buffer might return just a few bytes, even when there isn't
 395      * an error, so loop collecting them until we get enough.
 396      */
 397     while (pending < size) {
 398         int received = qemu_fill_buffer(f);
 399
 400         if (received <= 0) {
 401             break;
 402         }
 403
 404         index = f->buf_index + offset;
 405         pending = f->buf_size - index;
 406     }
 407
 408     if (pending <= 0) {
 409         return 0;
 410     }
 411     if (size > pending) {
 412         size = pending;
 413     }
 414
 415     *buf = f->buf + index;
 416     return size;
 417 }
 418
 419 /*
 420  * Read 'size' bytes of data from the file into buf.
 421  * 'size' can be larger than the internal buffer.
 422  *
 423  * It will return size bytes unless there was an error, in which case it will
 424  * return as many as it managed to read (assuming blocking fd's which
 425  * all current QEMUFile are)
 426  */
 427 size_t qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size)
 428 {
 429     size_t pending = size;
 430     size_t done = 0;
 431
 432     while (pending > 0) {
 433         size_t res;
 434         uint8_t *src;
 435
 436         res = qemu_peek_buffer(f, &src, MIN(pending, IO_BUF_SIZE), 0);
 437         if (res == 0) {
 438             return done;
 439         }
 440         memcpy(buf, src, res);
 441         qemu_file_skip(f, res);
 442         buf += res;
 443         pending -= res;
 444         done += res;
 445     }
 446     return done;
 447 }
 448
 449 /*
 450  * Read 'size' bytes of data from the file.
 451  * 'size' can be larger than the internal buffer.
 452  *
 453  * The data:
 454  *   may be held on an internal buffer (in which case *buf is updated
 455  *     to point to it) that is valid until the next qemu_file operation.
 456  * OR
 457  *   will be copied to the *buf that was passed in.
 458  *
 459  * The code tries to avoid the copy if possible.
 460  *
 461  * It will return size bytes unless there was an error, in which case it will
 462  * return as many as it managed to read (assuming blocking fd's which
 463  * all current QEMUFile are)
 464  *
 465  * Note: Since **buf may get changed, the caller should take care to
 466  *       keep a pointer to the original buffer if it needs to deallocate it.
 467  */
 468 size_t qemu_get_buffer_in_place(QEMUFile *f, uint8_t **buf, size_t size)
 469 {
 470     if (size < IO_BUF_SIZE) {
 471         size_t res;
 472         uint8_t *src;
 473
 474         res = qemu_peek_buffer(f, &src, size, 0);
 475
 476         if (res == size) {
 477             qemu_file_skip(f, res);
 478             *buf = src;
 479             return res;
 480         }
 481     }
 482
 483     return qemu_get_buffer(f, *buf, size);
 484 }
 485
 486 /*
 487  * Peeks a single byte from the buffer; this isn't guaranteed to work if
 488  * offset leaves a gap after the previous read/peeked data.
 489  */
 490 int qemu_peek_byte(QEMUFile *f, int offset)
 491 {
 492     int index = f->buf_index + offset;
 493
 494     assert(!qemu_file_is_writable(f));
 495     assert(offset < IO_BUF_SIZE);
 496
 497     if (index >= f->buf_size) {
 498         qemu_fill_buffer(f);
 499         index = f->buf_index + offset;
 500         if (index >= f->buf_size) {
 501             return 0;
 502         }
 503     }
 504     return f->buf[index];
 505 }
 506
 507 int qemu_get_byte(QEMUFile *f)
 508 {
 509     int result;
 510
 511     result = qemu_peek_byte(f, 0);
 512     qemu_file_skip(f, 1);
 513     return result;
 514 }
 515
 516 int64_t qemu_ftell_fast(QEMUFile *f)
 517 {
 518     int64_t ret = f->pos;
 519     int i;
 520
 521     if (f->ops->writev_buffer) {
 522         for (i = 0; i < f->iovcnt; i++) {
 523             ret += f->iov[i].iov_len;
 524         }
 525     } else {
 526         ret += f->buf_index;
 527     }
 528
 529     return ret;
 530 }
 531
 532 int64_t qemu_ftell(QEMUFile *f)
 533 {
 534     qemu_fflush(f);
 535     return f->pos;
 536 }
 537
 538 int qemu_file_rate_limit(QEMUFile *f)
 539 {
 540     if (qemu_file_get_error(f)) {
 541         return 1;
 542     }
 543     if (f->xfer_limit > 0 && f->bytes_xfer > f->xfer_limit) {
 544         return 1;
 545     }
 546     return 0;
 547 }
 548
 549 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 550 {
 551     return f->xfer_limit;
 552 }
 553
 554 void qemu_file_set_rate_limit(QEMUFile *f, int64_t limit)
 555 {
 556     f->xfer_limit = limit;
 557 }
 558
 559 void qemu_file_reset_rate_limit(QEMUFile *f)
 560 {
 561     f->bytes_xfer = 0;
 562 }
 563
 564 void qemu_put_be16(QEMUFile *f, unsigned int v)
 565 {
 566     qemu_put_byte(f, v >> 8);
 567     qemu_put_byte(f, v);
 568 }
 569
 570 void qemu_put_be32(QEMUFile *f, unsigned int v)
 571 {
 572     qemu_put_byte(f, v >> 24);
 573     qemu_put_byte(f, v >> 16);
 574     qemu_put_byte(f, v >> 8);
 575     qemu_put_byte(f, v);
 576 }
 577
 578 void qemu_put_be64(QEMUFile *f, uint64_t v)
 579 {
 580     qemu_put_be32(f, v >> 32);
 581     qemu_put_be32(f, v);
 582 }
 583
 584 unsigned int qemu_get_be16(QEMUFile *f)
 585 {
 586     unsigned int v;
 587     v = qemu_get_byte(f) << 8;
 588     v |= qemu_get_byte(f);
 589     return v;
 590 }
 591
 592 unsigned int qemu_get_be32(QEMUFile *f)
 593 {
 594     unsigned int v;
 595     v = (unsigned int)qemu_get_byte(f) << 24;
 596     v |= qemu_get_byte(f) << 16;
 597     v |= qemu_get_byte(f) << 8;
 598     v |= qemu_get_byte(f);
 599     return v;
 600 }
 601
 602 uint64_t qemu_get_be64(QEMUFile *f)
 603 {
 604     uint64_t v;
 605     v = (uint64_t)qemu_get_be32(f) << 32;
 606     v |= qemu_get_be32(f);
 607     return v;
 608 }
 609
 610 /* compress size bytes of data start at p with specific compression
 611  * level and store the compressed data to the buffer of f.
 612  */
 613
 614 ssize_t qemu_put_compression_data(QEMUFile *f, const uint8_t *p, size_t size,
 615                                   int level)
 616 {
 617     ssize_t blen = IO_BUF_SIZE - f->buf_index - sizeof(int32_t);
 618
 619     if (blen < compressBound(size)) {
 620         return 0;
 621     }
 622     if (compress2(f->buf + f->buf_index + sizeof(int32_t), (uLongf *)&blen,
 623                   (Bytef *)p, size, level) != Z_OK) {
 624         error_report("Compress Failed!");
 625         return 0;
 626     }
 627     qemu_put_be32(f, blen);
 628     f->buf_index += blen;
 629     return blen + sizeof(int32_t);
 630 }
 631
 632 /* Put the data in the buffer of f_src to the buffer of f_des, and
 633  * then reset the buf_index of f_src to 0.
 634  */
 635
 636 int qemu_put_qemu_file(QEMUFile *f_des, QEMUFile *f_src)
 637 {
 638     int len = 0;
 639
 640     if (f_src->buf_index > 0) {
 641         len = f_src->buf_index;
 642         qemu_put_buffer(f_des, f_src->buf, f_src->buf_index);
 643         f_src->buf_index = 0;
 644     }
 645     return len;
 646 }
 647
 648 /*
 649  * Get a string whose length is determined by a single preceding byte
 650  * A preallocated 256 byte buffer must be passed in.
 651  * Returns: len on success and a 0 terminated string in the buffer
 652  *          else 0
 653  *          (Note a 0 length string will return 0 either way)
 654  */
 655 size_t qemu_get_counted_string(QEMUFile *f, char buf[256])
 656 {
 657     size_t len = qemu_get_byte(f);
 658     size_t res = qemu_get_buffer(f, (uint8_t *)buf, len);
 659
 660     buf[res] = 0;
 661
 662     return res == len ? res : 0;
 663 }
 664
 665 /*
 666  * Set the blocking state of the QEMUFile.
 667  * Note: On some transports the OS only keeps a single blocking state for
 668  *       both directions, and thus changing the blocking on the main
 669  *       QEMUFile can also affect the return path.
 670  */
 671 void qemu_file_set_blocking(QEMUFile *f, bool block)
 672 {
 673     if (block) {
 674         qemu_set_block(qemu_get_fd(f));
 675     } else {
 676         qemu_set_nonblock(qemu_get_fd(f));
 677     }
 678 }