hw/s390x: Move virtio-ccw-net code to a separate file
[qemu.git] / migration / qemu-file.c
blob977b9ae07c12a56b6d68ea3dcbe9fec61c94bad6
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-common.h"
27 #include "qemu/error-report.h"
28 #include "qemu/iov.h"
29 #include "migration.h"
30 #include "qemu-file.h"
31 #include "trace.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;
58 * Stop a file from being read/written - not all backing files can do this
59 * typically only sockets can.
61 int qemu_file_shutdown(QEMUFile *f)
63 if (!f->ops->shut_down) {
64 return -ENOSYS;
66 return f->ops->shut_down(f->opaque, true, true);
70 * Result: QEMUFile* for a 'return path' for comms in the opposite direction
71 * NULL if not available
73 QEMUFile *qemu_file_get_return_path(QEMUFile *f)
75 if (!f->ops->get_return_path) {
76 return NULL;
78 return f->ops->get_return_path(f->opaque);
81 bool qemu_file_mode_is_not_valid(const char *mode)
83 if (mode == NULL ||
84 (mode[0] != 'r' && mode[0] != 'w') ||
85 mode[1] != 'b' || mode[2] != 0) {
86 fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
87 return true;
90 return false;
93 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
95 QEMUFile *f;
97 f = g_new0(QEMUFile, 1);
99 f->opaque = opaque;
100 f->ops = ops;
101 return f;
105 void qemu_file_set_hooks(QEMUFile *f, const QEMUFileHooks *hooks)
107 f->hooks = hooks;
111 * Get last error for stream f
113 * Return negative error value if there has been an error on previous
114 * operations, return 0 if no error happened.
117 int qemu_file_get_error(QEMUFile *f)
119 return f->last_error;
122 void qemu_file_set_error(QEMUFile *f, int ret)
124 if (f->last_error == 0) {
125 f->last_error = ret;
129 bool qemu_file_is_writable(QEMUFile *f)
131 return f->ops->writev_buffer;
134 static void qemu_iovec_release_ram(QEMUFile *f)
136 struct iovec iov;
137 unsigned long idx;
139 /* Find and release all the contiguous memory ranges marked as may_free. */
140 idx = find_next_bit(f->may_free, f->iovcnt, 0);
141 if (idx >= f->iovcnt) {
142 return;
144 iov = f->iov[idx];
146 /* The madvise() in the loop is called for iov within a continuous range and
147 * then reinitialize the iov. And in the end, madvise() is called for the
148 * last iov.
150 while ((idx = find_next_bit(f->may_free, f->iovcnt, idx + 1)) < f->iovcnt) {
151 /* check for adjacent buffer and coalesce them */
152 if (iov.iov_base + iov.iov_len == f->iov[idx].iov_base) {
153 iov.iov_len += f->iov[idx].iov_len;
154 continue;
156 if (qemu_madvise(iov.iov_base, iov.iov_len, QEMU_MADV_DONTNEED) < 0) {
157 error_report("migrate: madvise DONTNEED failed %p %zd: %s",
158 iov.iov_base, iov.iov_len, strerror(errno));
160 iov = f->iov[idx];
162 if (qemu_madvise(iov.iov_base, iov.iov_len, QEMU_MADV_DONTNEED) < 0) {
163 error_report("migrate: madvise DONTNEED failed %p %zd: %s",
164 iov.iov_base, iov.iov_len, strerror(errno));
166 memset(f->may_free, 0, sizeof(f->may_free));
170 * Flushes QEMUFile buffer
172 * If there is writev_buffer QEMUFileOps it uses it otherwise uses
173 * put_buffer ops. This will flush all pending data. If data was
174 * only partially flushed, it will set an error state.
176 void qemu_fflush(QEMUFile *f)
178 ssize_t ret = 0;
179 ssize_t expect = 0;
181 if (!qemu_file_is_writable(f)) {
182 return;
185 if (f->iovcnt > 0) {
186 expect = iov_size(f->iov, f->iovcnt);
187 ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt, f->pos);
189 qemu_iovec_release_ram(f);
192 if (ret >= 0) {
193 f->pos += ret;
195 /* We expect the QEMUFile write impl to send the full
196 * data set we requested, so sanity check that.
198 if (ret != expect) {
199 qemu_file_set_error(f, ret < 0 ? ret : -EIO);
201 f->buf_index = 0;
202 f->iovcnt = 0;
205 void ram_control_before_iterate(QEMUFile *f, uint64_t flags)
207 int ret = 0;
209 if (f->hooks && f->hooks->before_ram_iterate) {
210 ret = f->hooks->before_ram_iterate(f, f->opaque, flags, NULL);
211 if (ret < 0) {
212 qemu_file_set_error(f, ret);
217 void ram_control_after_iterate(QEMUFile *f, uint64_t flags)
219 int ret = 0;
221 if (f->hooks && f->hooks->after_ram_iterate) {
222 ret = f->hooks->after_ram_iterate(f, f->opaque, flags, NULL);
223 if (ret < 0) {
224 qemu_file_set_error(f, ret);
229 void ram_control_load_hook(QEMUFile *f, uint64_t flags, void *data)
231 int ret = -EINVAL;
233 if (f->hooks && f->hooks->hook_ram_load) {
234 ret = f->hooks->hook_ram_load(f, f->opaque, flags, data);
235 if (ret < 0) {
236 qemu_file_set_error(f, ret);
238 } else {
240 * Hook is a hook specifically requested by the source sending a flag
241 * that expects there to be a hook on the destination.
243 if (flags == RAM_CONTROL_HOOK) {
244 qemu_file_set_error(f, ret);
249 size_t ram_control_save_page(QEMUFile *f, ram_addr_t block_offset,
250 ram_addr_t offset, size_t size,
251 uint64_t *bytes_sent)
253 if (f->hooks && f->hooks->save_page) {
254 int ret = f->hooks->save_page(f, f->opaque, block_offset,
255 offset, size, bytes_sent);
256 if (ret != RAM_SAVE_CONTROL_NOT_SUPP) {
257 f->bytes_xfer += size;
260 if (ret != RAM_SAVE_CONTROL_DELAYED &&
261 ret != RAM_SAVE_CONTROL_NOT_SUPP) {
262 if (bytes_sent && *bytes_sent > 0) {
263 qemu_update_position(f, *bytes_sent);
264 } else if (ret < 0) {
265 qemu_file_set_error(f, ret);
269 return ret;
272 return RAM_SAVE_CONTROL_NOT_SUPP;
276 * Attempt to fill the buffer from the underlying file
277 * Returns the number of bytes read, or negative value for an error.
279 * Note that it can return a partially full buffer even in a not error/not EOF
280 * case if the underlying file descriptor gives a short read, and that can
281 * happen even on a blocking fd.
283 static ssize_t qemu_fill_buffer(QEMUFile *f)
285 int len;
286 int pending;
288 assert(!qemu_file_is_writable(f));
290 pending = f->buf_size - f->buf_index;
291 if (pending > 0) {
292 memmove(f->buf, f->buf + f->buf_index, pending);
294 f->buf_index = 0;
295 f->buf_size = pending;
297 len = f->ops->get_buffer(f->opaque, f->buf + pending, f->pos,
298 IO_BUF_SIZE - pending);
299 if (len > 0) {
300 f->buf_size += len;
301 f->pos += len;
302 } else if (len == 0) {
303 qemu_file_set_error(f, -EIO);
304 } else if (len != -EAGAIN) {
305 qemu_file_set_error(f, len);
308 return len;
311 void qemu_update_position(QEMUFile *f, size_t size)
313 f->pos += size;
316 /** Closes the file
318 * Returns negative error value if any error happened on previous operations or
319 * while closing the file. Returns 0 or positive number on success.
321 * The meaning of return value on success depends on the specific backend
322 * being used.
324 int qemu_fclose(QEMUFile *f)
326 int ret;
327 qemu_fflush(f);
328 ret = qemu_file_get_error(f);
330 if (f->ops->close) {
331 int ret2 = f->ops->close(f->opaque);
332 if (ret >= 0) {
333 ret = ret2;
336 /* If any error was spotted before closing, we should report it
337 * instead of the close() return value.
339 if (f->last_error) {
340 ret = f->last_error;
342 g_free(f);
343 trace_qemu_file_fclose();
344 return ret;
347 static void add_to_iovec(QEMUFile *f, const uint8_t *buf, size_t size,
348 bool may_free)
350 /* check for adjacent buffer and coalesce them */
351 if (f->iovcnt > 0 && buf == f->iov[f->iovcnt - 1].iov_base +
352 f->iov[f->iovcnt - 1].iov_len &&
353 may_free == test_bit(f->iovcnt - 1, f->may_free))
355 f->iov[f->iovcnt - 1].iov_len += size;
356 } else {
357 if (may_free) {
358 set_bit(f->iovcnt, f->may_free);
360 f->iov[f->iovcnt].iov_base = (uint8_t *)buf;
361 f->iov[f->iovcnt++].iov_len = size;
364 if (f->iovcnt >= MAX_IOV_SIZE) {
365 qemu_fflush(f);
369 void qemu_put_buffer_async(QEMUFile *f, const uint8_t *buf, size_t size,
370 bool may_free)
372 if (f->last_error) {
373 return;
376 f->bytes_xfer += size;
377 add_to_iovec(f, buf, size, may_free);
380 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, size_t size)
382 size_t l;
384 if (f->last_error) {
385 return;
388 while (size > 0) {
389 l = IO_BUF_SIZE - f->buf_index;
390 if (l > size) {
391 l = size;
393 memcpy(f->buf + f->buf_index, buf, l);
394 f->bytes_xfer += l;
395 add_to_iovec(f, f->buf + f->buf_index, l, false);
396 f->buf_index += l;
397 if (f->buf_index == IO_BUF_SIZE) {
398 qemu_fflush(f);
400 if (qemu_file_get_error(f)) {
401 break;
403 buf += l;
404 size -= l;
408 void qemu_put_byte(QEMUFile *f, int v)
410 if (f->last_error) {
411 return;
414 f->buf[f->buf_index] = v;
415 f->bytes_xfer++;
416 add_to_iovec(f, f->buf + f->buf_index, 1, false);
417 f->buf_index++;
418 if (f->buf_index == IO_BUF_SIZE) {
419 qemu_fflush(f);
423 void qemu_file_skip(QEMUFile *f, int size)
425 if (f->buf_index + size <= f->buf_size) {
426 f->buf_index += size;
431 * Read 'size' bytes from file (at 'offset') without moving the
432 * pointer and set 'buf' to point to that data.
434 * It will return size bytes unless there was an error, in which case it will
435 * return as many as it managed to read (assuming blocking fd's which
436 * all current QEMUFile are)
438 size_t qemu_peek_buffer(QEMUFile *f, uint8_t **buf, size_t size, size_t offset)
440 ssize_t pending;
441 size_t index;
443 assert(!qemu_file_is_writable(f));
444 assert(offset < IO_BUF_SIZE);
445 assert(size <= IO_BUF_SIZE - offset);
447 /* The 1st byte to read from */
448 index = f->buf_index + offset;
449 /* The number of available bytes starting at index */
450 pending = f->buf_size - index;
453 * qemu_fill_buffer might return just a few bytes, even when there isn't
454 * an error, so loop collecting them until we get enough.
456 while (pending < size) {
457 int received = qemu_fill_buffer(f);
459 if (received <= 0) {
460 break;
463 index = f->buf_index + offset;
464 pending = f->buf_size - index;
467 if (pending <= 0) {
468 return 0;
470 if (size > pending) {
471 size = pending;
474 *buf = f->buf + index;
475 return size;
479 * Read 'size' bytes of data from the file into buf.
480 * 'size' can be larger than the internal buffer.
482 * It will return size bytes unless there was an error, in which case it will
483 * return as many as it managed to read (assuming blocking fd's which
484 * all current QEMUFile are)
486 size_t qemu_get_buffer(QEMUFile *f, uint8_t *buf, size_t size)
488 size_t pending = size;
489 size_t done = 0;
491 while (pending > 0) {
492 size_t res;
493 uint8_t *src;
495 res = qemu_peek_buffer(f, &src, MIN(pending, IO_BUF_SIZE), 0);
496 if (res == 0) {
497 return done;
499 memcpy(buf, src, res);
500 qemu_file_skip(f, res);
501 buf += res;
502 pending -= res;
503 done += res;
505 return done;
509 * Read 'size' bytes of data from the file.
510 * 'size' can be larger than the internal buffer.
512 * The data:
513 * may be held on an internal buffer (in which case *buf is updated
514 * to point to it) that is valid until the next qemu_file operation.
515 * OR
516 * will be copied to the *buf that was passed in.
518 * The code tries to avoid the copy if possible.
520 * It will return size bytes unless there was an error, in which case it will
521 * return as many as it managed to read (assuming blocking fd's which
522 * all current QEMUFile are)
524 * Note: Since **buf may get changed, the caller should take care to
525 * keep a pointer to the original buffer if it needs to deallocate it.
527 size_t qemu_get_buffer_in_place(QEMUFile *f, uint8_t **buf, size_t size)
529 if (size < IO_BUF_SIZE) {
530 size_t res;
531 uint8_t *src;
533 res = qemu_peek_buffer(f, &src, size, 0);
535 if (res == size) {
536 qemu_file_skip(f, res);
537 *buf = src;
538 return res;
542 return qemu_get_buffer(f, *buf, size);
546 * Peeks a single byte from the buffer; this isn't guaranteed to work if
547 * offset leaves a gap after the previous read/peeked data.
549 int qemu_peek_byte(QEMUFile *f, int offset)
551 int index = f->buf_index + offset;
553 assert(!qemu_file_is_writable(f));
554 assert(offset < IO_BUF_SIZE);
556 if (index >= f->buf_size) {
557 qemu_fill_buffer(f);
558 index = f->buf_index + offset;
559 if (index >= f->buf_size) {
560 return 0;
563 return f->buf[index];
566 int qemu_get_byte(QEMUFile *f)
568 int result;
570 result = qemu_peek_byte(f, 0);
571 qemu_file_skip(f, 1);
572 return result;
575 int64_t qemu_ftell_fast(QEMUFile *f)
577 int64_t ret = f->pos;
578 int i;
580 for (i = 0; i < f->iovcnt; i++) {
581 ret += f->iov[i].iov_len;
584 return ret;
587 int64_t qemu_ftell(QEMUFile *f)
589 qemu_fflush(f);
590 return f->pos;
593 int qemu_file_rate_limit(QEMUFile *f)
595 if (qemu_file_get_error(f)) {
596 return 1;
598 if (f->xfer_limit > 0 && f->bytes_xfer > f->xfer_limit) {
599 return 1;
601 return 0;
604 int64_t qemu_file_get_rate_limit(QEMUFile *f)
606 return f->xfer_limit;
609 void qemu_file_set_rate_limit(QEMUFile *f, int64_t limit)
611 f->xfer_limit = limit;
614 void qemu_file_reset_rate_limit(QEMUFile *f)
616 f->bytes_xfer = 0;
619 void qemu_put_be16(QEMUFile *f, unsigned int v)
621 qemu_put_byte(f, v >> 8);
622 qemu_put_byte(f, v);
625 void qemu_put_be32(QEMUFile *f, unsigned int v)
627 qemu_put_byte(f, v >> 24);
628 qemu_put_byte(f, v >> 16);
629 qemu_put_byte(f, v >> 8);
630 qemu_put_byte(f, v);
633 void qemu_put_be64(QEMUFile *f, uint64_t v)
635 qemu_put_be32(f, v >> 32);
636 qemu_put_be32(f, v);
639 unsigned int qemu_get_be16(QEMUFile *f)
641 unsigned int v;
642 v = qemu_get_byte(f) << 8;
643 v |= qemu_get_byte(f);
644 return v;
647 unsigned int qemu_get_be32(QEMUFile *f)
649 unsigned int v;
650 v = (unsigned int)qemu_get_byte(f) << 24;
651 v |= qemu_get_byte(f) << 16;
652 v |= qemu_get_byte(f) << 8;
653 v |= qemu_get_byte(f);
654 return v;
657 uint64_t qemu_get_be64(QEMUFile *f)
659 uint64_t v;
660 v = (uint64_t)qemu_get_be32(f) << 32;
661 v |= qemu_get_be32(f);
662 return v;
665 /* return the size after compression, or negative value on error */
666 static int qemu_compress_data(z_stream *stream, uint8_t *dest, size_t dest_len,
667 const uint8_t *source, size_t source_len)
669 int err;
671 err = deflateReset(stream);
672 if (err != Z_OK) {
673 return -1;
676 stream->avail_in = source_len;
677 stream->next_in = (uint8_t *)source;
678 stream->avail_out = dest_len;
679 stream->next_out = dest;
681 err = deflate(stream, Z_FINISH);
682 if (err != Z_STREAM_END) {
683 return -1;
686 return stream->next_out - dest;
689 /* Compress size bytes of data start at p and store the compressed
690 * data to the buffer of f.
692 * When f is not writable, return -1 if f has no space to save the
693 * compressed data.
694 * When f is wirtable and it has no space to save the compressed data,
695 * do fflush first, if f still has no space to save the compressed
696 * data, return -1.
698 ssize_t qemu_put_compression_data(QEMUFile *f, z_stream *stream,
699 const uint8_t *p, size_t size)
701 ssize_t blen = IO_BUF_SIZE - f->buf_index - sizeof(int32_t);
703 if (blen < compressBound(size)) {
704 if (!qemu_file_is_writable(f)) {
705 return -1;
707 qemu_fflush(f);
708 blen = IO_BUF_SIZE - sizeof(int32_t);
709 if (blen < compressBound(size)) {
710 return -1;
714 blen = qemu_compress_data(stream, f->buf + f->buf_index + sizeof(int32_t),
715 blen, p, size);
716 if (blen < 0) {
717 return -1;
720 qemu_put_be32(f, blen);
721 if (f->ops->writev_buffer) {
722 add_to_iovec(f, f->buf + f->buf_index, blen, false);
724 f->buf_index += blen;
725 if (f->buf_index == IO_BUF_SIZE) {
726 qemu_fflush(f);
728 return blen + sizeof(int32_t);
731 /* Put the data in the buffer of f_src to the buffer of f_des, and
732 * then reset the buf_index of f_src to 0.
735 int qemu_put_qemu_file(QEMUFile *f_des, QEMUFile *f_src)
737 int len = 0;
739 if (f_src->buf_index > 0) {
740 len = f_src->buf_index;
741 qemu_put_buffer(f_des, f_src->buf, f_src->buf_index);
742 f_src->buf_index = 0;
743 f_src->iovcnt = 0;
745 return len;
749 * Get a string whose length is determined by a single preceding byte
750 * A preallocated 256 byte buffer must be passed in.
751 * Returns: len on success and a 0 terminated string in the buffer
752 * else 0
753 * (Note a 0 length string will return 0 either way)
755 size_t qemu_get_counted_string(QEMUFile *f, char buf[256])
757 size_t len = qemu_get_byte(f);
758 size_t res = qemu_get_buffer(f, (uint8_t *)buf, len);
760 buf[res] = 0;
762 return res == len ? res : 0;
766 * Put a string with one preceding byte containing its length. The length of
767 * the string should be less than 256.
769 void qemu_put_counted_string(QEMUFile *f, const char *str)
771 size_t len = strlen(str);
773 assert(len < 256);
774 qemu_put_byte(f, len);
775 qemu_put_buffer(f, (const uint8_t *)str, len);
779 * Set the blocking state of the QEMUFile.
780 * Note: On some transports the OS only keeps a single blocking state for
781 * both directions, and thus changing the blocking on the main
782 * QEMUFile can also affect the return path.
784 void qemu_file_set_blocking(QEMUFile *f, bool block)
786 if (f->ops->set_blocking) {
787 f->ops->set_blocking(f->opaque, block);