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 void qemu_file_set_hooks(QEMUFile *f, const QEMUFileHooks *hooks)
  85 {
  86     f->hooks = hooks;
  87 }
  88
  89 /*
  90  * Get last error for stream f
  91  *
  92  * Return negative error value if there has been an error on previous
  93  * operations, return 0 if no error happened.
  94  *
  95  */
  96 int qemu_file_get_error(QEMUFile *f)
  97 {
  98     return f->last_error;
  99 }
 100
 101 void qemu_file_set_error(QEMUFile *f, int ret)
 102 {
 103     if (f->last_error == 0) {
 104         f->last_error = ret;
 105     }
 106 }
 107
 108 bool qemu_file_is_writable(QEMUFile *f)
 109 {
 110     return f->ops->writev_buffer || f->ops->put_buffer;
 111 }
 112
 113 /**
 114  * Flushes QEMUFile buffer
 115  *
 116  * If there is writev_buffer QEMUFileOps it uses it otherwise uses
 117  * put_buffer ops. This will flush all pending data. If data was
 118  * only partially flushed, it will set an error state.
 119  */
 120 void qemu_fflush(QEMUFile *f)
 121 {
 122     ssize_t ret = 0;
 123     ssize_t expect = 0;
 124
 125     if (!qemu_file_is_writable(f)) {
 126         return;
 127     }
 128
 129     if (f->ops->writev_buffer) {
 130         if (f->iovcnt > 0) {
 131             expect = iov_size(f->iov, f->iovcnt);
 132             ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt, f->pos);
 133         }
 134     } else {
 135         if (f->buf_index > 0) {
 136             expect = f->buf_index;
 137             ret = f->ops->put_buffer(f->opaque, f->buf, f->pos, f->buf_index);
 138         }
 139     }
 140
 141     if (ret >= 0) {
 142         f->pos += ret;
 143     }
 144     /* We expect the QEMUFile write impl to send the full
 145      * data set we requested, so sanity check that.
 146      */
 147     if (ret != expect) {
 148         qemu_file_set_error(f, ret < 0 ? ret : -EIO);
 149     }
 150     f->buf_index = 0;
 151     f->iovcnt = 0;
 152 }
 153
 154 void ram_control_before_iterate(QEMUFile *f, uint64_t flags)
 155 {
 156     int ret = 0;
 157
 158     if (f->hooks && f->hooks->before_ram_iterate) {
 159         ret = f->hooks->before_ram_iterate(f, f->opaque, flags, NULL);
 160         if (ret < 0) {
 161             qemu_file_set_error(f, ret);
 162         }
 163     }
 164 }
 165
 166 void ram_control_after_iterate(QEMUFile *f, uint64_t flags)
 167 {
 168     int ret = 0;
 169
 170     if (f->hooks && f->hooks->after_ram_iterate) {
 171         ret = f->hooks->after_ram_iterate(f, f->opaque, flags, NULL);
 172         if (ret < 0) {
 173             qemu_file_set_error(f, ret);
 174         }
 175     }
 176 }
 177
 178 void ram_control_load_hook(QEMUFile *f, uint64_t flags, void *data)
 179 {
 180     int ret = -EINVAL;
 181
 182     if (f->hooks && f->hooks->hook_ram_load) {
 183         ret = f->hooks->hook_ram_load(f, f->opaque, flags, data);
 184         if (ret < 0) {
 185             qemu_file_set_error(f, ret);
 186         }
 187     } else {
 188         /*
 189          * Hook is a hook specifically requested by the source sending a flag
 190          * that expects there to be a hook on the destination.
 191          */
 192         if (flags == RAM_CONTROL_HOOK) {
 193             qemu_file_set_error(f, ret);
 194         }
 195     }
 196 }
 197
 198 size_t ram_control_save_page(QEMUFile *f, ram_addr_t block_offset,
 199                              ram_addr_t offset, size_t size,
 200                              uint64_t *bytes_sent)
 201 {
 202     if (f->hooks && f->hooks->save_page) {
 203         int ret = f->hooks->save_page(f, f->opaque, block_offset,
 204                                       offset, size, bytes_sent);
 205
 206         if (ret != RAM_SAVE_CONTROL_DELAYED) {
 207             if (bytes_sent && *bytes_sent > 0) {
 208                 qemu_update_position(f, *bytes_sent);
 209             } else if (ret < 0) {
 210                 qemu_file_set_error(f, ret);
 211             }
 212         }
 213
 214         return ret;
 215     }
 216
 217     return RAM_SAVE_CONTROL_NOT_SUPP;
 218 }
 219
 220 /*
 221  * Attempt to fill the buffer from the underlying file
 222  * Returns the number of bytes read, or negative value for an error.
 223  *
 224  * Note that it can return a partially full buffer even in a not error/not EOF
 225  * case if the underlying file descriptor gives a short read, and that can
 226  * happen even on a blocking fd.
 227  */
 228 static ssize_t qemu_fill_buffer(QEMUFile *f)
 229 {
 230     int len;
 231     int pending;
 232
 233     assert(!qemu_file_is_writable(f));
 234
 235     pending = f->buf_size - f->buf_index;
 236     if (pending > 0) {
 237         memmove(f->buf, f->buf + f->buf_index, pending);
 238     }
 239     f->buf_index = 0;
 240     f->buf_size = pending;
 241
 242     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->pos,
 243                         IO_BUF_SIZE - pending);
 244     if (len > 0) {
 245         f->buf_size += len;
 246         f->pos += len;
 247     } else if (len == 0) {
 248         qemu_file_set_error(f, -EIO);
 249     } else if (len != -EAGAIN) {
 250         qemu_file_set_error(f, len);
 251     }
 252
 253     return len;
 254 }
 255
 256 int qemu_get_fd(QEMUFile *f)
 257 {
 258     if (f->ops->get_fd) {
 259         return f->ops->get_fd(f->opaque);
 260     }
 261     return -1;
 262 }
 263
 264 void qemu_update_position(QEMUFile *f, size_t size)
 265 {
 266     f->pos += size;
 267 }
 268
 269 /** Closes the file
 270  *
 271  * Returns negative error value if any error happened on previous operations or
 272  * while closing the file. Returns 0 or positive number on success.
 273  *
 274  * The meaning of return value on success depends on the specific backend
 275  * being used.
 276  */
 277 int qemu_fclose(QEMUFile *f)
 278 {
 279     int ret;
 280     qemu_fflush(f);
 281     ret = qemu_file_get_error(f);
 282
 283     if (f->ops->close) {
 284         int ret2 = f->ops->close(f->opaque);
 285         if (ret >= 0) {
 286             ret = ret2;
 287         }
 288     }
 289     /* If any error was spotted before closing, we should report it
 290      * instead of the close() return value.
 291      */
 292     if (f->last_error) {
 293         ret = f->last_error;
 294     }
 295     g_free(f);
 296     trace_qemu_file_fclose();
 297     return ret;
 298 }
 299
 300 static void add_to_iovec(QEMUFile *f, const uint8_t *buf, size_t size)
 301 {
 302     /* check for adjacent buffer and coalesce them */
 303     if (f->iovcnt > 0 && buf == f->iov[f->iovcnt - 1].iov_base +
 304         f->iov[f->iovcnt - 1].iov_len) {
 305         f->iov[f->iovcnt - 1].iov_len += size;
 306     } else {
 307         f->iov[f->iovcnt].iov_base = (uint8_t *)buf;
 308         f->iov[f->iovcnt++].iov_len = size;
 309     }
 310
 311     if (f->iovcnt >= MAX_IOV_SIZE) {
 312         qemu_fflush(f);
 313     }
 314 }
 315
 316 void qemu_put_buffer_async(QEMUFile *f, const uint8_t *buf, size_t size)
 317 {
 318     if (!f->ops->writev_buffer) {
 319         qemu_put_buffer(f, buf, size);
 320         return;
 321     }
 322
 323     if (f->last_error) {
 324         return;
 325     }
 326
 327     f->bytes_xfer += size;
 328     add_to_iovec(f, buf, size);
 329 }
 330
 331 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, size_t size)
 332 {
 333     size_t l;
 334
 335     if (f->last_error) {
 336         return;
 337     }
 338
 339     while (size > 0) {
 340         l = IO_BUF_SIZE - f->buf_index;
 341         if (l > size) {
 342             l = size;
 343         }
 344         memcpy(f->buf + f->buf_index, buf, l);
 345         f->bytes_xfer += l;
 346         if (f->ops->writev_buffer) {
 347             add_to_iovec(f, f->buf + f->buf_index, l);
 348         }
 349         f->buf_index += l;
 350         if (f->buf_index == IO_BUF_SIZE) {
 351             qemu_fflush(f);
 352         }
 353         if (qemu_file_get_error(f)) {
 354             break;
 355         }
 356         buf += l;
 357         size -= l;
 358     }
 359 }
 360
 361 void qemu_put_byte(QEMUFile *f, int v)
 362 {
 363     if (f->last_error) {
 364         return;
 365     }
 366
 367     f->buf[f->buf_index] = v;
 368     f->bytes_xfer++;
 369     if (f->ops->writev_buffer) {
 370         add_to_iovec(f, f->buf + f->buf_index, 1);
 371     }
 372     f->buf_index++;
 373     if (f->buf_index == IO_BUF_SIZE) {
 374         qemu_fflush(f);
 375     }
 376 }
 377
 378 void qemu_file_skip(QEMUFile *f, int size)
 379 {
 380     if (f->buf_index + size <= f->buf_size) {
 381         f->buf_index += size;
 382     }
 383 }
 384
 385 /*
 386  * Read 'size' bytes from file (at 'offset') without moving the
 387  * pointer and set 'buf' to point to that data.
 388  *
 389  * It will return size bytes unless there was an error, in which case it will
 390  * return as many as it managed to read (assuming blocking fd's which
 391  * all current QEMUFile are)
 392  */
 393 size_t qemu_peek_buffer(QEMUFile *f, uint8_t **buf, size_t size, size_t offset)
 394 {
 395     ssize_t pending;
 396     size_t index;
 397
 398     assert(!qemu_file_is_writable(f));
 399     assert(offset < IO_BUF_SIZE);
 400     assert(size <= IO_BUF_SIZE - offset);
 401
 402     /* The 1st byte to read from */
 403     index = f->buf_index + offset;
 404     /* The number of available bytes starting at index */
 405     pending = f->buf_size - index;
 406
 407     /*
 408      * qemu_fill_buffer might return just a few bytes, even when there isn't
 409      * an error, so loop collecting them until we get enough.
 410      */
 411     while (pending < size) {
 412         int received = qemu_fill_buffer(f);
 413
 414         if (received <= 0) {
 415             break;
 416         }
 417
 418         index = f->buf_index + offset;
 419         pending = f->buf_size - index;
 420     }
 421
 422     if (pending <= 0) {
 423         return 0;
 424     }
 425     if (size > pending) {
 426         size = pending;
 427     }
 428
 429     *buf = f->buf + index;
 430     return size;
 431 }
 432
 433 /*
 434  * Read 'size' bytes of data from the file into buf.
 435  * 'size' can be larger than the internal buffer.
 436  *
 437  * It will return size bytes unless there was an error, in which case it will
 438  * return as many as it managed to read (assuming blocking fd's which
 439  * all current QEMUFile are)
 440  */
 441 size_t qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size)
 442 {
 443     size_t pending = size;
 444     size_t done = 0;
 445
 446     while (pending > 0) {
 447         size_t res;
 448         uint8_t *src;
 449
 450         res = qemu_peek_buffer(f, &src, MIN(pending, IO_BUF_SIZE), 0);
 451         if (res == 0) {
 452             return done;
 453         }
 454         memcpy(buf, src, res);
 455         qemu_file_skip(f, res);
 456         buf += res;
 457         pending -= res;
 458         done += res;
 459     }
 460     return done;
 461 }
 462
 463 /*
 464  * Read 'size' bytes of data from the file.
 465  * 'size' can be larger than the internal buffer.
 466  *
 467  * The data:
 468  *   may be held on an internal buffer (in which case *buf is updated
 469  *     to point to it) that is valid until the next qemu_file operation.
 470  * OR
 471  *   will be copied to the *buf that was passed in.
 472  *
 473  * The code tries to avoid the copy if possible.
 474  *
 475  * It will return size bytes unless there was an error, in which case it will
 476  * return as many as it managed to read (assuming blocking fd's which
 477  * all current QEMUFile are)
 478  *
 479  * Note: Since **buf may get changed, the caller should take care to
 480  *       keep a pointer to the original buffer if it needs to deallocate it.
 481  */
 482 size_t qemu_get_buffer_in_place(QEMUFile *f, uint8_t **buf, size_t size)
 483 {
 484     if (size < IO_BUF_SIZE) {
 485         size_t res;
 486         uint8_t *src;
 487
 488         res = qemu_peek_buffer(f, &src, size, 0);
 489
 490         if (res == size) {
 491             qemu_file_skip(f, res);
 492             *buf = src;
 493             return res;
 494         }
 495     }
 496
 497     return qemu_get_buffer(f, *buf, size);
 498 }
 499
 500 /*
 501  * Peeks a single byte from the buffer; this isn't guaranteed to work if
 502  * offset leaves a gap after the previous read/peeked data.
 503  */
 504 int qemu_peek_byte(QEMUFile *f, int offset)
 505 {
 506     int index = f->buf_index + offset;
 507
 508     assert(!qemu_file_is_writable(f));
 509     assert(offset < IO_BUF_SIZE);
 510
 511     if (index >= f->buf_size) {
 512         qemu_fill_buffer(f);
 513         index = f->buf_index + offset;
 514         if (index >= f->buf_size) {
 515             return 0;
 516         }
 517     }
 518     return f->buf[index];
 519 }
 520
 521 int qemu_get_byte(QEMUFile *f)
 522 {
 523     int result;
 524
 525     result = qemu_peek_byte(f, 0);
 526     qemu_file_skip(f, 1);
 527     return result;
 528 }
 529
 530 int64_t qemu_ftell_fast(QEMUFile *f)
 531 {
 532     int64_t ret = f->pos;
 533     int i;
 534
 535     if (f->ops->writev_buffer) {
 536         for (i = 0; i < f->iovcnt; i++) {
 537             ret += f->iov[i].iov_len;
 538         }
 539     } else {
 540         ret += f->buf_index;
 541     }
 542
 543     return ret;
 544 }
 545
 546 int64_t qemu_ftell(QEMUFile *f)
 547 {
 548     qemu_fflush(f);
 549     return f->pos;
 550 }
 551
 552 int qemu_file_rate_limit(QEMUFile *f)
 553 {
 554     if (qemu_file_get_error(f)) {
 555         return 1;
 556     }
 557     if (f->xfer_limit > 0 && f->bytes_xfer > f->xfer_limit) {
 558         return 1;
 559     }
 560     return 0;
 561 }
 562
 563 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 564 {
 565     return f->xfer_limit;
 566 }
 567
 568 void qemu_file_set_rate_limit(QEMUFile *f, int64_t limit)
 569 {
 570     f->xfer_limit = limit;
 571 }
 572
 573 void qemu_file_reset_rate_limit(QEMUFile *f)
 574 {
 575     f->bytes_xfer = 0;
 576 }
 577
 578 void qemu_put_be16(QEMUFile *f, unsigned int v)
 579 {
 580     qemu_put_byte(f, v >> 8);
 581     qemu_put_byte(f, v);
 582 }
 583
 584 void qemu_put_be32(QEMUFile *f, unsigned int v)
 585 {
 586     qemu_put_byte(f, v >> 24);
 587     qemu_put_byte(f, v >> 16);
 588     qemu_put_byte(f, v >> 8);
 589     qemu_put_byte(f, v);
 590 }
 591
 592 void qemu_put_be64(QEMUFile *f, uint64_t v)
 593 {
 594     qemu_put_be32(f, v >> 32);
 595     qemu_put_be32(f, v);
 596 }
 597
 598 unsigned int qemu_get_be16(QEMUFile *f)
 599 {
 600     unsigned int v;
 601     v = qemu_get_byte(f) << 8;
 602     v |= qemu_get_byte(f);
 603     return v;
 604 }
 605
 606 unsigned int qemu_get_be32(QEMUFile *f)
 607 {
 608     unsigned int v;
 609     v = (unsigned int)qemu_get_byte(f) << 24;
 610     v |= qemu_get_byte(f) << 16;
 611     v |= qemu_get_byte(f) << 8;
 612     v |= qemu_get_byte(f);
 613     return v;
 614 }
 615
 616 uint64_t qemu_get_be64(QEMUFile *f)
 617 {
 618     uint64_t v;
 619     v = (uint64_t)qemu_get_be32(f) << 32;
 620     v |= qemu_get_be32(f);
 621     return v;
 622 }
 623
 624 /* compress size bytes of data start at p with specific compression
 625  * level and store the compressed data to the buffer of f.
 626  */
 627
 628 ssize_t qemu_put_compression_data(QEMUFile *f, const uint8_t *p, size_t size,
 629                                   int level)
 630 {
 631     ssize_t blen = IO_BUF_SIZE - f->buf_index - sizeof(int32_t);
 632
 633     if (blen < compressBound(size)) {
 634         return 0;
 635     }
 636     if (compress2(f->buf + f->buf_index + sizeof(int32_t), (uLongf *)&blen,
 637                   (Bytef *)p, size, level) != Z_OK) {
 638         error_report("Compress Failed!");
 639         return 0;
 640     }
 641     qemu_put_be32(f, blen);
 642     f->buf_index += blen;
 643     return blen + sizeof(int32_t);
 644 }
 645
 646 /* Put the data in the buffer of f_src to the buffer of f_des, and
 647  * then reset the buf_index of f_src to 0.
 648  */
 649
 650 int qemu_put_qemu_file(QEMUFile *f_des, QEMUFile *f_src)
 651 {
 652     int len = 0;
 653
 654     if (f_src->buf_index > 0) {
 655         len = f_src->buf_index;
 656         qemu_put_buffer(f_des, f_src->buf, f_src->buf_index);
 657         f_src->buf_index = 0;
 658     }
 659     return len;
 660 }
 661
 662 /*
 663  * Get a string whose length is determined by a single preceding byte
 664  * A preallocated 256 byte buffer must be passed in.
 665  * Returns: len on success and a 0 terminated string in the buffer
 666  *          else 0
 667  *          (Note a 0 length string will return 0 either way)
 668  */
 669 size_t qemu_get_counted_string(QEMUFile *f, char buf[256])
 670 {
 671     size_t len = qemu_get_byte(f);
 672     size_t res = qemu_get_buffer(f, (uint8_t *)buf, len);
 673
 674     buf[res] = 0;
 675
 676     return res == len ? res : 0;
 677 }
 678
 679 /*
 680  * Set the blocking state of the QEMUFile.
 681  * Note: On some transports the OS only keeps a single blocking state for
 682  *       both directions, and thus changing the blocking on the main
 683  *       QEMUFile can also affect the return path.
 684  */
 685 void qemu_file_set_blocking(QEMUFile *f, bool block)
 686 {
 687     if (f->ops->set_blocking) {
 688         f->ops->set_blocking(f->opaque, block);
 689     } else {
 690         if (block) {
 691             qemu_set_block(qemu_get_fd(f));
 692         } else {
 693             qemu_set_nonblock(qemu_get_fd(f));
 694         }
 695     }
 696 }