Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging
[qemu/ar7.git] / migration / qemu-file.c
blob26fb25ddc1341af62b80571f8699b5e9740dd8d3
1 /*
2 * QEMU System Emulator
4 * Copyright (c) 2003-2008 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
24 #include "qemu/osdep.h"
25 #include <zlib.h>
26 #include "qemu/error-report.h"
27 #include "qemu/iov.h"
28 #include "migration.h"
29 #include "qemu-file.h"
30 #include "trace.h"
31 #include "qapi/error.h"
33 #define IO_BUF_SIZE 32768
34 #define MAX_IOV_SIZE MIN(IOV_MAX, 64)
36 struct QEMUFile {
37 const QEMUFileOps *ops;
38 const QEMUFileHooks *hooks;
39 void *opaque;
41 int64_t bytes_xfer;
42 int64_t xfer_limit;
44 int64_t pos; /* start of buffer when writing, end of buffer
45 when reading */
46 int buf_index;
47 int buf_size; /* 0 when writing */
48 uint8_t buf[IO_BUF_SIZE];
50 DECLARE_BITMAP(may_free, MAX_IOV_SIZE);
51 struct iovec iov[MAX_IOV_SIZE];
52 unsigned int iovcnt;
54 int last_error;
55 Error *last_error_obj;
59 * Stop a file from being read/written - not all backing files can do this
60 * typically only sockets can.
62 int qemu_file_shutdown(QEMUFile *f)
64 if (!f->ops->shut_down) {
65 return -ENOSYS;
67 return f->ops->shut_down(f->opaque, true, true, NULL);
71 * Result: QEMUFile* for a 'return path' for comms in the opposite direction
72 * NULL if not available
74 QEMUFile *qemu_file_get_return_path(QEMUFile *f)
76 if (!f->ops->get_return_path) {
77 return NULL;
79 return f->ops->get_return_path(f->opaque);
82 bool qemu_file_mode_is_not_valid(const char *mode)
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;
91 return false;
94 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
96 QEMUFile *f;
98 f = g_new0(QEMUFile, 1);
100 f->opaque = opaque;
101 f->ops = ops;
102 return f;
106 void qemu_file_set_hooks(QEMUFile *f, const QEMUFileHooks *hooks)
108 f->hooks = hooks;
112 * Get last error for stream f with optional Error*
114 * Return negative error value if there has been an error on previous
115 * operations, return 0 if no error happened.
116 * Optional, it returns Error* in errp, but it may be NULL even if return value
117 * is not 0.
120 int qemu_file_get_error_obj(QEMUFile *f, Error **errp)
122 if (errp) {
123 *errp = f->last_error_obj ? error_copy(f->last_error_obj) : NULL;
125 return f->last_error;
129 * Set the last error for stream f with optional Error*
131 void qemu_file_set_error_obj(QEMUFile *f, int ret, Error *err)
133 if (f->last_error == 0 && ret) {
134 f->last_error = ret;
135 error_propagate(&f->last_error_obj, err);
136 } else if (err) {
137 error_report_err(err);
142 * Get last error for stream f
144 * Return negative error value if there has been an error on previous
145 * operations, return 0 if no error happened.
148 int qemu_file_get_error(QEMUFile *f)
150 return qemu_file_get_error_obj(f, NULL);
154 * Set the last error for stream f
156 void qemu_file_set_error(QEMUFile *f, int ret)
158 qemu_file_set_error_obj(f, ret, NULL);
161 bool qemu_file_is_writable(QEMUFile *f)
163 return f->ops->writev_buffer;
166 static void qemu_iovec_release_ram(QEMUFile *f)
168 struct iovec iov;
169 unsigned long idx;
171 /* Find and release all the contiguous memory ranges marked as may_free. */
172 idx = find_next_bit(f->may_free, f->iovcnt, 0);
173 if (idx >= f->iovcnt) {
174 return;
176 iov = f->iov[idx];
178 /* The madvise() in the loop is called for iov within a continuous range and
179 * then reinitialize the iov. And in the end, madvise() is called for the
180 * last iov.
182 while ((idx = find_next_bit(f->may_free, f->iovcnt, idx + 1)) < f->iovcnt) {
183 /* check for adjacent buffer and coalesce them */
184 if (iov.iov_base + iov.iov_len == f->iov[idx].iov_base) {
185 iov.iov_len += f->iov[idx].iov_len;
186 continue;
188 if (qemu_madvise(iov.iov_base, iov.iov_len, QEMU_MADV_DONTNEED) < 0) {
189 error_report("migrate: madvise DONTNEED failed %p %zd: %s",
190 iov.iov_base, iov.iov_len, strerror(errno));
192 iov = f->iov[idx];
194 if (qemu_madvise(iov.iov_base, iov.iov_len, QEMU_MADV_DONTNEED) < 0) {
195 error_report("migrate: madvise DONTNEED failed %p %zd: %s",
196 iov.iov_base, iov.iov_len, strerror(errno));
198 memset(f->may_free, 0, sizeof(f->may_free));
202 * Flushes QEMUFile buffer
204 * This will flush all pending data. If data was only partially flushed, it
205 * will set an error state.
207 void qemu_fflush(QEMUFile *f)
209 ssize_t ret = 0;
210 ssize_t expect = 0;
211 Error *local_error = NULL;
213 if (!qemu_file_is_writable(f)) {
214 return;
217 if (f->iovcnt > 0) {
218 expect = iov_size(f->iov, f->iovcnt);
219 ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt, f->pos,
220 &local_error);
222 qemu_iovec_release_ram(f);
225 if (ret >= 0) {
226 f->pos += ret;
228 /* We expect the QEMUFile write impl to send the full
229 * data set we requested, so sanity check that.
231 if (ret != expect) {
232 qemu_file_set_error_obj(f, ret < 0 ? ret : -EIO, local_error);
234 f->buf_index = 0;
235 f->iovcnt = 0;
238 void ram_control_before_iterate(QEMUFile *f, uint64_t flags)
240 int ret = 0;
242 if (f->hooks && f->hooks->before_ram_iterate) {
243 ret = f->hooks->before_ram_iterate(f, f->opaque, flags, NULL);
244 if (ret < 0) {
245 qemu_file_set_error(f, ret);
250 void ram_control_after_iterate(QEMUFile *f, uint64_t flags)
252 int ret = 0;
254 if (f->hooks && f->hooks->after_ram_iterate) {
255 ret = f->hooks->after_ram_iterate(f, f->opaque, flags, NULL);
256 if (ret < 0) {
257 qemu_file_set_error(f, ret);
262 void ram_control_load_hook(QEMUFile *f, uint64_t flags, void *data)
264 int ret = -EINVAL;
266 if (f->hooks && f->hooks->hook_ram_load) {
267 ret = f->hooks->hook_ram_load(f, f->opaque, flags, data);
268 if (ret < 0) {
269 qemu_file_set_error(f, ret);
271 } else {
273 * Hook is a hook specifically requested by the source sending a flag
274 * that expects there to be a hook on the destination.
276 if (flags == RAM_CONTROL_HOOK) {
277 qemu_file_set_error(f, ret);
282 size_t ram_control_save_page(QEMUFile *f, ram_addr_t block_offset,
283 ram_addr_t offset, size_t size,
284 uint64_t *bytes_sent)
286 if (f->hooks && f->hooks->save_page) {
287 int ret = f->hooks->save_page(f, f->opaque, block_offset,
288 offset, size, bytes_sent);
289 if (ret != RAM_SAVE_CONTROL_NOT_SUPP) {
290 f->bytes_xfer += size;
293 if (ret != RAM_SAVE_CONTROL_DELAYED &&
294 ret != RAM_SAVE_CONTROL_NOT_SUPP) {
295 if (bytes_sent && *bytes_sent > 0) {
296 qemu_update_position(f, *bytes_sent);
297 } else if (ret < 0) {
298 qemu_file_set_error(f, ret);
302 return ret;
305 return RAM_SAVE_CONTROL_NOT_SUPP;
309 * Attempt to fill the buffer from the underlying file
310 * Returns the number of bytes read, or negative value for an error.
312 * Note that it can return a partially full buffer even in a not error/not EOF
313 * case if the underlying file descriptor gives a short read, and that can
314 * happen even on a blocking fd.
316 static ssize_t qemu_fill_buffer(QEMUFile *f)
318 int len;
319 int pending;
320 Error *local_error = NULL;
322 assert(!qemu_file_is_writable(f));
324 pending = f->buf_size - f->buf_index;
325 if (pending > 0) {
326 memmove(f->buf, f->buf + f->buf_index, pending);
328 f->buf_index = 0;
329 f->buf_size = pending;
331 len = f->ops->get_buffer(f->opaque, f->buf + pending, f->pos,
332 IO_BUF_SIZE - pending, &local_error);
333 if (len > 0) {
334 f->buf_size += len;
335 f->pos += len;
336 } else if (len == 0) {
337 qemu_file_set_error_obj(f, -EIO, local_error);
338 } else if (len != -EAGAIN) {
339 qemu_file_set_error_obj(f, len, local_error);
340 } else {
341 error_free(local_error);
344 return len;
347 void qemu_update_position(QEMUFile *f, size_t size)
349 f->pos += size;
352 /** Closes the file
354 * Returns negative error value if any error happened on previous operations or
355 * while closing the file. Returns 0 or positive number on success.
357 * The meaning of return value on success depends on the specific backend
358 * being used.
360 int qemu_fclose(QEMUFile *f)
362 int ret;
363 qemu_fflush(f);
364 ret = qemu_file_get_error(f);
366 if (f->ops->close) {
367 int ret2 = f->ops->close(f->opaque, NULL);
368 if (ret >= 0) {
369 ret = ret2;
372 /* If any error was spotted before closing, we should report it
373 * instead of the close() return value.
375 if (f->last_error) {
376 ret = f->last_error;
378 error_free(f->last_error_obj);
379 g_free(f);
380 trace_qemu_file_fclose();
381 return ret;
385 * Add buf to iovec. Do flush if iovec is full.
387 * Return values:
388 * 1 iovec is full and flushed
389 * 0 iovec is not flushed
392 static int add_to_iovec(QEMUFile *f, const uint8_t *buf, size_t size,
393 bool may_free)
395 /* check for adjacent buffer and coalesce them */
396 if (f->iovcnt > 0 && buf == f->iov[f->iovcnt - 1].iov_base +
397 f->iov[f->iovcnt - 1].iov_len &&
398 may_free == test_bit(f->iovcnt - 1, f->may_free))
400 f->iov[f->iovcnt - 1].iov_len += size;
401 } else {
402 if (may_free) {
403 set_bit(f->iovcnt, f->may_free);
405 f->iov[f->iovcnt].iov_base = (uint8_t *)buf;
406 f->iov[f->iovcnt++].iov_len = size;
409 if (f->iovcnt >= MAX_IOV_SIZE) {
410 qemu_fflush(f);
411 return 1;
414 return 0;
417 static void add_buf_to_iovec(QEMUFile *f, size_t len)
419 if (!add_to_iovec(f, f->buf + f->buf_index, len, false)) {
420 f->buf_index += len;
421 if (f->buf_index == IO_BUF_SIZE) {
422 qemu_fflush(f);
427 void qemu_put_buffer_async(QEMUFile *f, const uint8_t *buf, size_t size,
428 bool may_free)
430 if (f->last_error) {
431 return;
434 f->bytes_xfer += size;
435 add_to_iovec(f, buf, size, may_free);
438 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, size_t size)
440 size_t l;
442 if (f->last_error) {
443 return;
446 while (size > 0) {
447 l = IO_BUF_SIZE - f->buf_index;
448 if (l > size) {
449 l = size;
451 memcpy(f->buf + f->buf_index, buf, l);
452 f->bytes_xfer += l;
453 add_buf_to_iovec(f, l);
454 if (qemu_file_get_error(f)) {
455 break;
457 buf += l;
458 size -= l;
462 void qemu_put_byte(QEMUFile *f, int v)
464 if (f->last_error) {
465 return;
468 f->buf[f->buf_index] = v;
469 f->bytes_xfer++;
470 add_buf_to_iovec(f, 1);
473 void qemu_file_skip(QEMUFile *f, int size)
475 if (f->buf_index + size <= f->buf_size) {
476 f->buf_index += size;
481 * Read 'size' bytes from file (at 'offset') without moving the
482 * pointer and set 'buf' to point to that data.
484 * It will return size bytes unless there was an error, in which case it will
485 * return as many as it managed to read (assuming blocking fd's which
486 * all current QEMUFile are)
488 size_t qemu_peek_buffer(QEMUFile *f, uint8_t **buf, size_t size, size_t offset)
490 ssize_t pending;
491 size_t index;
493 assert(!qemu_file_is_writable(f));
494 assert(offset < IO_BUF_SIZE);
495 assert(size <= IO_BUF_SIZE - offset);
497 /* The 1st byte to read from */
498 index = f->buf_index + offset;
499 /* The number of available bytes starting at index */
500 pending = f->buf_size - index;
503 * qemu_fill_buffer might return just a few bytes, even when there isn't
504 * an error, so loop collecting them until we get enough.
506 while (pending < size) {
507 int received = qemu_fill_buffer(f);
509 if (received <= 0) {
510 break;
513 index = f->buf_index + offset;
514 pending = f->buf_size - index;
517 if (pending <= 0) {
518 return 0;
520 if (size > pending) {
521 size = pending;
524 *buf = f->buf + index;
525 return size;
529 * Read 'size' bytes of data from the file into buf.
530 * 'size' can be larger than the internal buffer.
532 * It will return size bytes unless there was an error, in which case it will
533 * return as many as it managed to read (assuming blocking fd's which
534 * all current QEMUFile are)
536 size_t qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size)
538 size_t pending = size;
539 size_t done = 0;
541 while (pending > 0) {
542 size_t res;
543 uint8_t *src;
545 res = qemu_peek_buffer(f, &src, MIN(pending, IO_BUF_SIZE), 0);
546 if (res == 0) {
547 return done;
549 memcpy(buf, src, res);
550 qemu_file_skip(f, res);
551 buf += res;
552 pending -= res;
553 done += res;
555 return done;
559 * Read 'size' bytes of data from the file.
560 * 'size' can be larger than the internal buffer.
562 * The data:
563 * may be held on an internal buffer (in which case *buf is updated
564 * to point to it) that is valid until the next qemu_file operation.
565 * OR
566 * will be copied to the *buf that was passed in.
568 * The code tries to avoid the copy if possible.
570 * It will return size bytes unless there was an error, in which case it will
571 * return as many as it managed to read (assuming blocking fd's which
572 * all current QEMUFile are)
574 * Note: Since **buf may get changed, the caller should take care to
575 * keep a pointer to the original buffer if it needs to deallocate it.
577 size_t qemu_get_buffer_in_place(QEMUFile *f, uint8_t **buf, size_t size)
579 if (size < IO_BUF_SIZE) {
580 size_t res;
581 uint8_t *src;
583 res = qemu_peek_buffer(f, &src, size, 0);
585 if (res == size) {
586 qemu_file_skip(f, res);
587 *buf = src;
588 return res;
592 return qemu_get_buffer(f, *buf, size);
596 * Peeks a single byte from the buffer; this isn't guaranteed to work if
597 * offset leaves a gap after the previous read/peeked data.
599 int qemu_peek_byte(QEMUFile *f, int offset)
601 int index = f->buf_index + offset;
603 assert(!qemu_file_is_writable(f));
604 assert(offset < IO_BUF_SIZE);
606 if (index >= f->buf_size) {
607 qemu_fill_buffer(f);
608 index = f->buf_index + offset;
609 if (index >= f->buf_size) {
610 return 0;
613 return f->buf[index];
616 int qemu_get_byte(QEMUFile *f)
618 int result;
620 result = qemu_peek_byte(f, 0);
621 qemu_file_skip(f, 1);
622 return result;
625 int64_t qemu_ftell_fast(QEMUFile *f)
627 int64_t ret = f->pos;
628 int i;
630 for (i = 0; i < f->iovcnt; i++) {
631 ret += f->iov[i].iov_len;
634 return ret;
637 int64_t qemu_ftell(QEMUFile *f)
639 qemu_fflush(f);
640 return f->pos;
643 int qemu_file_rate_limit(QEMUFile *f)
645 if (qemu_file_get_error(f)) {
646 return 1;
648 if (f->xfer_limit > 0 && f->bytes_xfer > f->xfer_limit) {
649 return 1;
651 return 0;
654 int64_t qemu_file_get_rate_limit(QEMUFile *f)
656 return f->xfer_limit;
659 void qemu_file_set_rate_limit(QEMUFile *f, int64_t limit)
661 f->xfer_limit = limit;
664 void qemu_file_reset_rate_limit(QEMUFile *f)
666 f->bytes_xfer = 0;
669 void qemu_file_update_transfer(QEMUFile *f, int64_t len)
671 f->bytes_xfer += len;
674 void qemu_put_be16(QEMUFile *f, unsigned int v)
676 qemu_put_byte(f, v >> 8);
677 qemu_put_byte(f, v);
680 void qemu_put_be32(QEMUFile *f, unsigned int v)
682 qemu_put_byte(f, v >> 24);
683 qemu_put_byte(f, v >> 16);
684 qemu_put_byte(f, v >> 8);
685 qemu_put_byte(f, v);
688 void qemu_put_be64(QEMUFile *f, uint64_t v)
690 qemu_put_be32(f, v >> 32);
691 qemu_put_be32(f, v);
694 unsigned int qemu_get_be16(QEMUFile *f)
696 unsigned int v;
697 v = qemu_get_byte(f) << 8;
698 v |= qemu_get_byte(f);
699 return v;
702 unsigned int qemu_get_be32(QEMUFile *f)
704 unsigned int v;
705 v = (unsigned int)qemu_get_byte(f) << 24;
706 v |= qemu_get_byte(f) << 16;
707 v |= qemu_get_byte(f) << 8;
708 v |= qemu_get_byte(f);
709 return v;
712 uint64_t qemu_get_be64(QEMUFile *f)
714 uint64_t v;
715 v = (uint64_t)qemu_get_be32(f) << 32;
716 v |= qemu_get_be32(f);
717 return v;
720 /* return the size after compression, or negative value on error */
721 static int qemu_compress_data(z_stream *stream, uint8_t *dest, size_t dest_len,
722 const uint8_t *source, size_t source_len)
724 int err;
726 err = deflateReset(stream);
727 if (err != Z_OK) {
728 return -1;
731 stream->avail_in = source_len;
732 stream->next_in = (uint8_t *)source;
733 stream->avail_out = dest_len;
734 stream->next_out = dest;
736 err = deflate(stream, Z_FINISH);
737 if (err != Z_STREAM_END) {
738 return -1;
741 return stream->next_out - dest;
744 /* Compress size bytes of data start at p and store the compressed
745 * data to the buffer of f.
747 * When f is not writable, return -1 if f has no space to save the
748 * compressed data.
749 * When f is wirtable and it has no space to save the compressed data,
750 * do fflush first, if f still has no space to save the compressed
751 * data, return -1.
753 ssize_t qemu_put_compression_data(QEMUFile *f, z_stream *stream,
754 const uint8_t *p, size_t size)
756 ssize_t blen = IO_BUF_SIZE - f->buf_index - sizeof(int32_t);
758 if (blen < compressBound(size)) {
759 if (!qemu_file_is_writable(f)) {
760 return -1;
762 qemu_fflush(f);
763 blen = IO_BUF_SIZE - sizeof(int32_t);
764 if (blen < compressBound(size)) {
765 return -1;
769 blen = qemu_compress_data(stream, f->buf + f->buf_index + sizeof(int32_t),
770 blen, p, size);
771 if (blen < 0) {
772 return -1;
775 qemu_put_be32(f, blen);
776 add_buf_to_iovec(f, blen);
777 return blen + sizeof(int32_t);
780 /* Put the data in the buffer of f_src to the buffer of f_des, and
781 * then reset the buf_index of f_src to 0.
784 int qemu_put_qemu_file(QEMUFile *f_des, QEMUFile *f_src)
786 int len = 0;
788 if (f_src->buf_index > 0) {
789 len = f_src->buf_index;
790 qemu_put_buffer(f_des, f_src->buf, f_src->buf_index);
791 f_src->buf_index = 0;
792 f_src->iovcnt = 0;
794 return len;
798 * Get a string whose length is determined by a single preceding byte
799 * A preallocated 256 byte buffer must be passed in.
800 * Returns: len on success and a 0 terminated string in the buffer
801 * else 0
802 * (Note a 0 length string will return 0 either way)
804 size_t qemu_get_counted_string(QEMUFile *f, char buf[256])
806 size_t len = qemu_get_byte(f);
807 size_t res = qemu_get_buffer(f, (uint8_t *)buf, len);
809 buf[res] = 0;
811 return res == len ? res : 0;
815 * Put a string with one preceding byte containing its length. The length of
816 * the string should be less than 256.
818 void qemu_put_counted_string(QEMUFile *f, const char *str)
820 size_t len = strlen(str);
822 assert(len < 256);
823 qemu_put_byte(f, len);
824 qemu_put_buffer(f, (const uint8_t *)str, len);
828 * Set the blocking state of the QEMUFile.
829 * Note: On some transports the OS only keeps a single blocking state for
830 * both directions, and thus changing the blocking on the main
831 * QEMUFile can also affect the return path.
833 void qemu_file_set_blocking(QEMUFile *f, bool block)
835 if (f->ops->set_blocking) {
836 f->ops->set_blocking(f->opaque, block, NULL);