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 "trace.h"
  34
  35 #define IO_BUF_SIZE 32768
  36 #define MAX_IOV_SIZE MIN(IOV_MAX, 64)
  37
  38 struct QEMUFile {
  39     const QEMUFileOps *ops;
  40     const QEMUFileHooks *hooks;
  41     void *opaque;
  42
  43     int64_t bytes_xfer;
  44     int64_t xfer_limit;
  45
  46     int64_t pos; /* start of buffer when writing, end of buffer
  47                     when reading */
  48     int buf_index;
  49     int buf_size; /* 0 when writing */
  50     uint8_t buf[IO_BUF_SIZE];
  51
  52     struct iovec iov[MAX_IOV_SIZE];
  53     unsigned int iovcnt;
  54
  55     int last_error;
  56 };
  57
  58 /*
  59  * Stop a file from being read/written - not all backing files can do this
  60  * typically only sockets can.
  61  */
  62 int qemu_file_shutdown(QEMUFile *f)
  63 {
  64     if (!f->ops->shut_down) {
  65         return -ENOSYS;
  66     }
  67     return f->ops->shut_down(f->opaque, true, true);
  68 }
  69
  70 /*
  71  * Result: QEMUFile* for a 'return path' for comms in the opposite direction
  72  *         NULL if not available
  73  */
  74 QEMUFile *qemu_file_get_return_path(QEMUFile *f)
  75 {
  76     if (!f->ops->get_return_path) {
  77         return NULL;
  78     }
  79     return f->ops->get_return_path(f->opaque);
  80 }
  81
  82 bool qemu_file_mode_is_not_valid(const char *mode)
  83 {
  84     if (mode == NULL ||
  85         (mode[0] != 'r' && mode[0] != 'w') ||
  86         mode[1] != 'b' || mode[2] != 0) {
  87         fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
  88         return true;
  89     }
  90
  91     return false;
  92 }
  93
  94 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
  95 {
  96     QEMUFile *f;
  97
  98     f = g_new0(QEMUFile, 1);
  99
 100     f->opaque = opaque;
 101     f->ops = ops;
 102     return f;
 103 }
 104
 105
 106 void qemu_file_set_hooks(QEMUFile *f, const QEMUFileHooks *hooks)
 107 {
 108     f->hooks = hooks;
 109 }
 110
 111 /*
 112  * Get last error for stream f
 113  *
 114  * Return negative error value if there has been an error on previous
 115  * operations, return 0 if no error happened.
 116  *
 117  */
 118 int qemu_file_get_error(QEMUFile *f)
 119 {
 120     return f->last_error;
 121 }
 122
 123 void qemu_file_set_error(QEMUFile *f, int ret)
 124 {
 125     if (f->last_error == 0) {
 126         f->last_error = ret;
 127     }
 128 }
 129
 130 bool qemu_file_is_writable(QEMUFile *f)
 131 {
 132     return f->ops->writev_buffer;
 133 }
 134
 135 /**
 136  * Flushes QEMUFile buffer
 137  *
 138  * If there is writev_buffer QEMUFileOps it uses it otherwise uses
 139  * put_buffer ops. This will flush all pending data. If data was
 140  * only partially flushed, it will set an error state.
 141  */
 142 void qemu_fflush(QEMUFile *f)
 143 {
 144     ssize_t ret = 0;
 145     ssize_t expect = 0;
 146
 147     if (!qemu_file_is_writable(f)) {
 148         return;
 149     }
 150
 151     if (f->iovcnt > 0) {
 152         expect = iov_size(f->iov, f->iovcnt);
 153         ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt, f->pos);
 154     }
 155
 156     if (ret >= 0) {
 157         f->pos += ret;
 158     }
 159     /* We expect the QEMUFile write impl to send the full
 160      * data set we requested, so sanity check that.
 161      */
 162     if (ret != expect) {
 163         qemu_file_set_error(f, ret < 0 ? ret : -EIO);
 164     }
 165     f->buf_index = 0;
 166     f->iovcnt = 0;
 167 }
 168
 169 void ram_control_before_iterate(QEMUFile *f, uint64_t flags)
 170 {
 171     int ret = 0;
 172
 173     if (f->hooks && f->hooks->before_ram_iterate) {
 174         ret = f->hooks->before_ram_iterate(f, f->opaque, flags, NULL);
 175         if (ret < 0) {
 176             qemu_file_set_error(f, ret);
 177         }
 178     }
 179 }
 180
 181 void ram_control_after_iterate(QEMUFile *f, uint64_t flags)
 182 {
 183     int ret = 0;
 184
 185     if (f->hooks && f->hooks->after_ram_iterate) {
 186         ret = f->hooks->after_ram_iterate(f, f->opaque, flags, NULL);
 187         if (ret < 0) {
 188             qemu_file_set_error(f, ret);
 189         }
 190     }
 191 }
 192
 193 void ram_control_load_hook(QEMUFile *f, uint64_t flags, void *data)
 194 {
 195     int ret = -EINVAL;
 196
 197     if (f->hooks && f->hooks->hook_ram_load) {
 198         ret = f->hooks->hook_ram_load(f, f->opaque, flags, data);
 199         if (ret < 0) {
 200             qemu_file_set_error(f, ret);
 201         }
 202     } else {
 203         /*
 204          * Hook is a hook specifically requested by the source sending a flag
 205          * that expects there to be a hook on the destination.
 206          */
 207         if (flags == RAM_CONTROL_HOOK) {
 208             qemu_file_set_error(f, ret);
 209         }
 210     }
 211 }
 212
 213 size_t ram_control_save_page(QEMUFile *f, ram_addr_t block_offset,
 214                              ram_addr_t offset, size_t size,
 215                              uint64_t *bytes_sent)
 216 {
 217     if (f->hooks && f->hooks->save_page) {
 218         int ret = f->hooks->save_page(f, f->opaque, block_offset,
 219                                       offset, size, bytes_sent);
 220
 221         if (ret != RAM_SAVE_CONTROL_DELAYED) {
 222             if (bytes_sent && *bytes_sent > 0) {
 223                 qemu_update_position(f, *bytes_sent);
 224             } else if (ret < 0) {
 225                 qemu_file_set_error(f, ret);
 226             }
 227         }
 228
 229         return ret;
 230     }
 231
 232     return RAM_SAVE_CONTROL_NOT_SUPP;
 233 }
 234
 235 /*
 236  * Attempt to fill the buffer from the underlying file
 237  * Returns the number of bytes read, or negative value for an error.
 238  *
 239  * Note that it can return a partially full buffer even in a not error/not EOF
 240  * case if the underlying file descriptor gives a short read, and that can
 241  * happen even on a blocking fd.
 242  */
 243 static ssize_t qemu_fill_buffer(QEMUFile *f)
 244 {
 245     int len;
 246     int pending;
 247
 248     assert(!qemu_file_is_writable(f));
 249
 250     pending = f->buf_size - f->buf_index;
 251     if (pending > 0) {
 252         memmove(f->buf, f->buf + f->buf_index, pending);
 253     }
 254     f->buf_index = 0;
 255     f->buf_size = pending;
 256
 257     len = f->ops->get_buffer(f->opaque, f->buf + pending, f->pos,
 258                         IO_BUF_SIZE - pending);
 259     if (len > 0) {
 260         f->buf_size += len;
 261         f->pos += len;
 262     } else if (len == 0) {
 263         qemu_file_set_error(f, -EIO);
 264     } else if (len != -EAGAIN) {
 265         qemu_file_set_error(f, len);
 266     }
 267
 268     return len;
 269 }
 270
 271 int qemu_get_fd(QEMUFile *f)
 272 {
 273     if (f->ops->get_fd) {
 274         return f->ops->get_fd(f->opaque);
 275     }
 276     return -1;
 277 }
 278
 279 void qemu_update_position(QEMUFile *f, size_t size)
 280 {
 281     f->pos += size;
 282 }
 283
 284 /** Closes the file
 285  *
 286  * Returns negative error value if any error happened on previous operations or
 287  * while closing the file. Returns 0 or positive number on success.
 288  *
 289  * The meaning of return value on success depends on the specific backend
 290  * being used.
 291  */
 292 int qemu_fclose(QEMUFile *f)
 293 {
 294     int ret;
 295     qemu_fflush(f);
 296     ret = qemu_file_get_error(f);
 297
 298     if (f->ops->close) {
 299         int ret2 = f->ops->close(f->opaque);
 300         if (ret >= 0) {
 301             ret = ret2;
 302         }
 303     }
 304     /* If any error was spotted before closing, we should report it
 305      * instead of the close() return value.
 306      */
 307     if (f->last_error) {
 308         ret = f->last_error;
 309     }
 310     g_free(f);
 311     trace_qemu_file_fclose();
 312     return ret;
 313 }
 314
 315 static void add_to_iovec(QEMUFile *f, const uint8_t *buf, size_t size)
 316 {
 317     /* check for adjacent buffer and coalesce them */
 318     if (f->iovcnt > 0 && buf == f->iov[f->iovcnt - 1].iov_base +
 319         f->iov[f->iovcnt - 1].iov_len) {
 320         f->iov[f->iovcnt - 1].iov_len += size;
 321     } else {
 322         f->iov[f->iovcnt].iov_base = (uint8_t *)buf;
 323         f->iov[f->iovcnt++].iov_len = size;
 324     }
 325
 326     if (f->iovcnt >= MAX_IOV_SIZE) {
 327         qemu_fflush(f);
 328     }
 329 }
 330
 331 void qemu_put_buffer_async(QEMUFile *f, const uint8_t *buf, size_t size)
 332 {
 333     if (f->last_error) {
 334         return;
 335     }
 336
 337     f->bytes_xfer += size;
 338     add_to_iovec(f, buf, size);
 339 }
 340
 341 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, size_t size)
 342 {
 343     size_t l;
 344
 345     if (f->last_error) {
 346         return;
 347     }
 348
 349     while (size > 0) {
 350         l = IO_BUF_SIZE - f->buf_index;
 351         if (l > size) {
 352             l = size;
 353         }
 354         memcpy(f->buf + f->buf_index, buf, l);
 355         f->bytes_xfer += l;
 356         add_to_iovec(f, f->buf + f->buf_index, l);
 357         f->buf_index += l;
 358         if (f->buf_index == IO_BUF_SIZE) {
 359             qemu_fflush(f);
 360         }
 361         if (qemu_file_get_error(f)) {
 362             break;
 363         }
 364         buf += l;
 365         size -= l;
 366     }
 367 }
 368
 369 void qemu_put_byte(QEMUFile *f, int v)
 370 {
 371     if (f->last_error) {
 372         return;
 373     }
 374
 375     f->buf[f->buf_index] = v;
 376     f->bytes_xfer++;
 377     add_to_iovec(f, f->buf + f->buf_index, 1);
 378     f->buf_index++;
 379     if (f->buf_index == IO_BUF_SIZE) {
 380         qemu_fflush(f);
 381     }
 382 }
 383
 384 void qemu_file_skip(QEMUFile *f, int size)
 385 {
 386     if (f->buf_index + size <= f->buf_size) {
 387         f->buf_index += size;
 388     }
 389 }
 390
 391 /*
 392  * Read 'size' bytes from file (at 'offset') without moving the
 393  * pointer and set 'buf' to point to that data.
 394  *
 395  * It will return size bytes unless there was an error, in which case it will
 396  * return as many as it managed to read (assuming blocking fd's which
 397  * all current QEMUFile are)
 398  */
 399 size_t qemu_peek_buffer(QEMUFile *f, uint8_t **buf, size_t size, size_t offset)
 400 {
 401     ssize_t pending;
 402     size_t index;
 403
 404     assert(!qemu_file_is_writable(f));
 405     assert(offset < IO_BUF_SIZE);
 406     assert(size <= IO_BUF_SIZE - offset);
 407
 408     /* The 1st byte to read from */
 409     index = f->buf_index + offset;
 410     /* The number of available bytes starting at index */
 411     pending = f->buf_size - index;
 412
 413     /*
 414      * qemu_fill_buffer might return just a few bytes, even when there isn't
 415      * an error, so loop collecting them until we get enough.
 416      */
 417     while (pending < size) {
 418         int received = qemu_fill_buffer(f);
 419
 420         if (received <= 0) {
 421             break;
 422         }
 423
 424         index = f->buf_index + offset;
 425         pending = f->buf_size - index;
 426     }
 427
 428     if (pending <= 0) {
 429         return 0;
 430     }
 431     if (size > pending) {
 432         size = pending;
 433     }
 434
 435     *buf = f->buf + index;
 436     return size;
 437 }
 438
 439 /*
 440  * Read 'size' bytes of data from the file into buf.
 441  * 'size' can be larger than the internal buffer.
 442  *
 443  * It will return size bytes unless there was an error, in which case it will
 444  * return as many as it managed to read (assuming blocking fd's which
 445  * all current QEMUFile are)
 446  */
 447 size_t qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size)
 448 {
 449     size_t pending = size;
 450     size_t done = 0;
 451
 452     while (pending > 0) {
 453         size_t res;
 454         uint8_t *src;
 455
 456         res = qemu_peek_buffer(f, &src, MIN(pending, IO_BUF_SIZE), 0);
 457         if (res == 0) {
 458             return done;
 459         }
 460         memcpy(buf, src, res);
 461         qemu_file_skip(f, res);
 462         buf += res;
 463         pending -= res;
 464         done += res;
 465     }
 466     return done;
 467 }
 468
 469 /*
 470  * Read 'size' bytes of data from the file.
 471  * 'size' can be larger than the internal buffer.
 472  *
 473  * The data:
 474  *   may be held on an internal buffer (in which case *buf is updated
 475  *     to point to it) that is valid until the next qemu_file operation.
 476  * OR
 477  *   will be copied to the *buf that was passed in.
 478  *
 479  * The code tries to avoid the copy if possible.
 480  *
 481  * It will return size bytes unless there was an error, in which case it will
 482  * return as many as it managed to read (assuming blocking fd's which
 483  * all current QEMUFile are)
 484  *
 485  * Note: Since **buf may get changed, the caller should take care to
 486  *       keep a pointer to the original buffer if it needs to deallocate it.
 487  */
 488 size_t qemu_get_buffer_in_place(QEMUFile *f, uint8_t **buf, size_t size)
 489 {
 490     if (size < IO_BUF_SIZE) {
 491         size_t res;
 492         uint8_t *src;
 493
 494         res = qemu_peek_buffer(f, &src, size, 0);
 495
 496         if (res == size) {
 497             qemu_file_skip(f, res);
 498             *buf = src;
 499             return res;
 500         }
 501     }
 502
 503     return qemu_get_buffer(f, *buf, size);
 504 }
 505
 506 /*
 507  * Peeks a single byte from the buffer; this isn't guaranteed to work if
 508  * offset leaves a gap after the previous read/peeked data.
 509  */
 510 int qemu_peek_byte(QEMUFile *f, int offset)
 511 {
 512     int index = f->buf_index + offset;
 513
 514     assert(!qemu_file_is_writable(f));
 515     assert(offset < IO_BUF_SIZE);
 516
 517     if (index >= f->buf_size) {
 518         qemu_fill_buffer(f);
 519         index = f->buf_index + offset;
 520         if (index >= f->buf_size) {
 521             return 0;
 522         }
 523     }
 524     return f->buf[index];
 525 }
 526
 527 int qemu_get_byte(QEMUFile *f)
 528 {
 529     int result;
 530
 531     result = qemu_peek_byte(f, 0);
 532     qemu_file_skip(f, 1);
 533     return result;
 534 }
 535
 536 int64_t qemu_ftell_fast(QEMUFile *f)
 537 {
 538     int64_t ret = f->pos;
 539     int i;
 540
 541     for (i = 0; i < f->iovcnt; i++) {
 542         ret += f->iov[i].iov_len;
 543     }
 544
 545     return ret;
 546 }
 547
 548 int64_t qemu_ftell(QEMUFile *f)
 549 {
 550     qemu_fflush(f);
 551     return f->pos;
 552 }
 553
 554 int qemu_file_rate_limit(QEMUFile *f)
 555 {
 556     if (qemu_file_get_error(f)) {
 557         return 1;
 558     }
 559     if (f->xfer_limit > 0 && f->bytes_xfer > f->xfer_limit) {
 560         return 1;
 561     }
 562     return 0;
 563 }
 564
 565 int64_t qemu_file_get_rate_limit(QEMUFile *f)
 566 {
 567     return f->xfer_limit;
 568 }
 569
 570 void qemu_file_set_rate_limit(QEMUFile *f, int64_t limit)
 571 {
 572     f->xfer_limit = limit;
 573 }
 574
 575 void qemu_file_reset_rate_limit(QEMUFile *f)
 576 {
 577     f->bytes_xfer = 0;
 578 }
 579
 580 void qemu_put_be16(QEMUFile *f, unsigned int v)
 581 {
 582     qemu_put_byte(f, v >> 8);
 583     qemu_put_byte(f, v);
 584 }
 585
 586 void qemu_put_be32(QEMUFile *f, unsigned int v)
 587 {
 588     qemu_put_byte(f, v >> 24);
 589     qemu_put_byte(f, v >> 16);
 590     qemu_put_byte(f, v >> 8);
 591     qemu_put_byte(f, v);
 592 }
 593
 594 void qemu_put_be64(QEMUFile *f, uint64_t v)
 595 {
 596     qemu_put_be32(f, v >> 32);
 597     qemu_put_be32(f, v);
 598 }
 599
 600 unsigned int qemu_get_be16(QEMUFile *f)
 601 {
 602     unsigned int v;
 603     v = qemu_get_byte(f) << 8;
 604     v |= qemu_get_byte(f);
 605     return v;
 606 }
 607
 608 unsigned int qemu_get_be32(QEMUFile *f)
 609 {
 610     unsigned int v;
 611     v = (unsigned int)qemu_get_byte(f) << 24;
 612     v |= qemu_get_byte(f) << 16;
 613     v |= qemu_get_byte(f) << 8;
 614     v |= qemu_get_byte(f);
 615     return v;
 616 }
 617
 618 uint64_t qemu_get_be64(QEMUFile *f)
 619 {
 620     uint64_t v;
 621     v = (uint64_t)qemu_get_be32(f) << 32;
 622     v |= qemu_get_be32(f);
 623     return v;
 624 }
 625
 626 /* compress size bytes of data start at p with specific compression
 627  * level and store the compressed data to the buffer of f.
 628  */
 629
 630 ssize_t qemu_put_compression_data(QEMUFile *f, const uint8_t *p, size_t size,
 631                                   int level)
 632 {
 633     ssize_t blen = IO_BUF_SIZE - f->buf_index - sizeof(int32_t);
 634
 635     if (blen < compressBound(size)) {
 636         return 0;
 637     }
 638     if (compress2(f->buf + f->buf_index + sizeof(int32_t), (uLongf *)&blen,
 639                   (Bytef *)p, size, level) != Z_OK) {
 640         error_report("Compress Failed!");
 641         return 0;
 642     }
 643     qemu_put_be32(f, blen);
 644     f->buf_index += blen;
 645     return blen + sizeof(int32_t);
 646 }
 647
 648 /* Put the data in the buffer of f_src to the buffer of f_des, and
 649  * then reset the buf_index of f_src to 0.
 650  */
 651
 652 int qemu_put_qemu_file(QEMUFile *f_des, QEMUFile *f_src)
 653 {
 654     int len = 0;
 655
 656     if (f_src->buf_index > 0) {
 657         len = f_src->buf_index;
 658         qemu_put_buffer(f_des, f_src->buf, f_src->buf_index);
 659         f_src->buf_index = 0;
 660     }
 661     return len;
 662 }
 663
 664 /*
 665  * Get a string whose length is determined by a single preceding byte
 666  * A preallocated 256 byte buffer must be passed in.
 667  * Returns: len on success and a 0 terminated string in the buffer
 668  *          else 0
 669  *          (Note a 0 length string will return 0 either way)
 670  */
 671 size_t qemu_get_counted_string(QEMUFile *f, char buf[256])
 672 {
 673     size_t len = qemu_get_byte(f);
 674     size_t res = qemu_get_buffer(f, (uint8_t *)buf, len);
 675
 676     buf[res] = 0;
 677
 678     return res == len ? res : 0;
 679 }
 680
 681 /*
 682  * Set the blocking state of the QEMUFile.
 683  * Note: On some transports the OS only keeps a single blocking state for
 684  *       both directions, and thus changing the blocking on the main
 685  *       QEMUFile can also affect the return path.
 686  */
 687 void qemu_file_set_blocking(QEMUFile *f, bool block)
 688 {
 689     if (f->ops->set_blocking) {
 690         f->ops->set_blocking(f->opaque, block);
 691     } else {
 692         if (block) {
 693             qemu_set_block(qemu_get_fd(f));
 694         } else {
 695             qemu_set_nonblock(qemu_get_fd(f));
 696         }
 697     }
 698 }