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
24 #include "qemu/osdep.h"
26 #include "qemu-common.h"
27 #include "qemu/error-report.h"
29 #include "migration.h"
30 #include "qemu-file.h"
33 #define IO_BUF_SIZE 32768
34 #define MAX_IOV_SIZE MIN(IOV_MAX, 64)
37 const QEMUFileOps
*ops
;
38 const QEMUFileHooks
*hooks
;
44 int64_t pos
; /* start of buffer when writing, end of buffer
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
];
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
) {
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
) {
78 return f
->ops
->get_return_path(f
->opaque
);
81 bool qemu_file_mode_is_not_valid(const char *mode
)
84 (mode
[0] != 'r' && mode
[0] != 'w') ||
85 mode
[1] != 'b' || mode
[2] != 0) {
86 fprintf(stderr
, "qemu_fopen: Argument validity check failed\n");
93 QEMUFile
*qemu_fopen_ops(void *opaque
, const QEMUFileOps
*ops
)
97 f
= g_new0(QEMUFile
, 1);
105 void qemu_file_set_hooks(QEMUFile
*f
, const QEMUFileHooks
*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) {
129 bool qemu_file_is_writable(QEMUFile
*f
)
131 return f
->ops
->writev_buffer
;
134 static void qemu_iovec_release_ram(QEMUFile
*f
)
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
) {
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
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
;
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
));
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
)
181 if (!qemu_file_is_writable(f
)) {
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
);
195 /* We expect the QEMUFile write impl to send the full
196 * data set we requested, so sanity check that.
199 qemu_file_set_error(f
, ret
< 0 ? ret
: -EIO
);
205 void ram_control_before_iterate(QEMUFile
*f
, uint64_t flags
)
209 if (f
->hooks
&& f
->hooks
->before_ram_iterate
) {
210 ret
= f
->hooks
->before_ram_iterate(f
, f
->opaque
, flags
, NULL
);
212 qemu_file_set_error(f
, ret
);
217 void ram_control_after_iterate(QEMUFile
*f
, uint64_t flags
)
221 if (f
->hooks
&& f
->hooks
->after_ram_iterate
) {
222 ret
= f
->hooks
->after_ram_iterate(f
, f
->opaque
, flags
, NULL
);
224 qemu_file_set_error(f
, ret
);
229 void ram_control_load_hook(QEMUFile
*f
, uint64_t flags
, void *data
)
233 if (f
->hooks
&& f
->hooks
->hook_ram_load
) {
234 ret
= f
->hooks
->hook_ram_load(f
, f
->opaque
, flags
, data
);
236 qemu_file_set_error(f
, ret
);
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
);
257 if (ret
!= RAM_SAVE_CONTROL_DELAYED
) {
258 if (bytes_sent
&& *bytes_sent
> 0) {
259 qemu_update_position(f
, *bytes_sent
);
260 } else if (ret
< 0) {
261 qemu_file_set_error(f
, ret
);
268 return RAM_SAVE_CONTROL_NOT_SUPP
;
272 * Attempt to fill the buffer from the underlying file
273 * Returns the number of bytes read, or negative value for an error.
275 * Note that it can return a partially full buffer even in a not error/not EOF
276 * case if the underlying file descriptor gives a short read, and that can
277 * happen even on a blocking fd.
279 static ssize_t
qemu_fill_buffer(QEMUFile
*f
)
284 assert(!qemu_file_is_writable(f
));
286 pending
= f
->buf_size
- f
->buf_index
;
288 memmove(f
->buf
, f
->buf
+ f
->buf_index
, pending
);
291 f
->buf_size
= pending
;
293 len
= f
->ops
->get_buffer(f
->opaque
, f
->buf
+ pending
, f
->pos
,
294 IO_BUF_SIZE
- pending
);
298 } else if (len
== 0) {
299 qemu_file_set_error(f
, -EIO
);
300 } else if (len
!= -EAGAIN
) {
301 qemu_file_set_error(f
, len
);
307 void qemu_update_position(QEMUFile
*f
, size_t size
)
314 * Returns negative error value if any error happened on previous operations or
315 * while closing the file. Returns 0 or positive number on success.
317 * The meaning of return value on success depends on the specific backend
320 int qemu_fclose(QEMUFile
*f
)
324 ret
= qemu_file_get_error(f
);
327 int ret2
= f
->ops
->close(f
->opaque
);
332 /* If any error was spotted before closing, we should report it
333 * instead of the close() return value.
339 trace_qemu_file_fclose();
343 static void add_to_iovec(QEMUFile
*f
, const uint8_t *buf
, size_t size
,
346 /* check for adjacent buffer and coalesce them */
347 if (f
->iovcnt
> 0 && buf
== f
->iov
[f
->iovcnt
- 1].iov_base
+
348 f
->iov
[f
->iovcnt
- 1].iov_len
&&
349 may_free
== test_bit(f
->iovcnt
- 1, f
->may_free
))
351 f
->iov
[f
->iovcnt
- 1].iov_len
+= size
;
354 set_bit(f
->iovcnt
, f
->may_free
);
356 f
->iov
[f
->iovcnt
].iov_base
= (uint8_t *)buf
;
357 f
->iov
[f
->iovcnt
++].iov_len
= size
;
360 if (f
->iovcnt
>= MAX_IOV_SIZE
) {
365 void qemu_put_buffer_async(QEMUFile
*f
, const uint8_t *buf
, size_t size
,
372 f
->bytes_xfer
+= size
;
373 add_to_iovec(f
, buf
, size
, may_free
);
376 void qemu_put_buffer(QEMUFile
*f
, const uint8_t *buf
, size_t size
)
385 l
= IO_BUF_SIZE
- f
->buf_index
;
389 memcpy(f
->buf
+ f
->buf_index
, buf
, l
);
391 add_to_iovec(f
, f
->buf
+ f
->buf_index
, l
, false);
393 if (f
->buf_index
== IO_BUF_SIZE
) {
396 if (qemu_file_get_error(f
)) {
404 void qemu_put_byte(QEMUFile
*f
, int v
)
410 f
->buf
[f
->buf_index
] = v
;
412 add_to_iovec(f
, f
->buf
+ f
->buf_index
, 1, false);
414 if (f
->buf_index
== IO_BUF_SIZE
) {
419 void qemu_file_skip(QEMUFile
*f
, int size
)
421 if (f
->buf_index
+ size
<= f
->buf_size
) {
422 f
->buf_index
+= size
;
427 * Read 'size' bytes from file (at 'offset') without moving the
428 * pointer and set 'buf' to point to that data.
430 * It will return size bytes unless there was an error, in which case it will
431 * return as many as it managed to read (assuming blocking fd's which
432 * all current QEMUFile are)
434 size_t qemu_peek_buffer(QEMUFile
*f
, uint8_t **buf
, size_t size
, size_t offset
)
439 assert(!qemu_file_is_writable(f
));
440 assert(offset
< IO_BUF_SIZE
);
441 assert(size
<= IO_BUF_SIZE
- offset
);
443 /* The 1st byte to read from */
444 index
= f
->buf_index
+ offset
;
445 /* The number of available bytes starting at index */
446 pending
= f
->buf_size
- index
;
449 * qemu_fill_buffer might return just a few bytes, even when there isn't
450 * an error, so loop collecting them until we get enough.
452 while (pending
< size
) {
453 int received
= qemu_fill_buffer(f
);
459 index
= f
->buf_index
+ offset
;
460 pending
= f
->buf_size
- index
;
466 if (size
> pending
) {
470 *buf
= f
->buf
+ index
;
475 * Read 'size' bytes of data from the file into buf.
476 * 'size' can be larger than the internal buffer.
478 * It will return size bytes unless there was an error, in which case it will
479 * return as many as it managed to read (assuming blocking fd's which
480 * all current QEMUFile are)
482 size_t qemu_get_buffer(QEMUFile
*f
, uint8_t *buf
, size_t size
)
484 size_t pending
= size
;
487 while (pending
> 0) {
491 res
= qemu_peek_buffer(f
, &src
, MIN(pending
, IO_BUF_SIZE
), 0);
495 memcpy(buf
, src
, res
);
496 qemu_file_skip(f
, res
);
505 * Read 'size' bytes of data from the file.
506 * 'size' can be larger than the internal buffer.
509 * may be held on an internal buffer (in which case *buf is updated
510 * to point to it) that is valid until the next qemu_file operation.
512 * will be copied to the *buf that was passed in.
514 * The code tries to avoid the copy if possible.
516 * It will return size bytes unless there was an error, in which case it will
517 * return as many as it managed to read (assuming blocking fd's which
518 * all current QEMUFile are)
520 * Note: Since **buf may get changed, the caller should take care to
521 * keep a pointer to the original buffer if it needs to deallocate it.
523 size_t qemu_get_buffer_in_place(QEMUFile
*f
, uint8_t **buf
, size_t size
)
525 if (size
< IO_BUF_SIZE
) {
529 res
= qemu_peek_buffer(f
, &src
, size
, 0);
532 qemu_file_skip(f
, res
);
538 return qemu_get_buffer(f
, *buf
, size
);
542 * Peeks a single byte from the buffer; this isn't guaranteed to work if
543 * offset leaves a gap after the previous read/peeked data.
545 int qemu_peek_byte(QEMUFile
*f
, int offset
)
547 int index
= f
->buf_index
+ offset
;
549 assert(!qemu_file_is_writable(f
));
550 assert(offset
< IO_BUF_SIZE
);
552 if (index
>= f
->buf_size
) {
554 index
= f
->buf_index
+ offset
;
555 if (index
>= f
->buf_size
) {
559 return f
->buf
[index
];
562 int qemu_get_byte(QEMUFile
*f
)
566 result
= qemu_peek_byte(f
, 0);
567 qemu_file_skip(f
, 1);
571 int64_t qemu_ftell_fast(QEMUFile
*f
)
573 int64_t ret
= f
->pos
;
576 for (i
= 0; i
< f
->iovcnt
; i
++) {
577 ret
+= f
->iov
[i
].iov_len
;
583 int64_t qemu_ftell(QEMUFile
*f
)
589 int qemu_file_rate_limit(QEMUFile
*f
)
591 if (qemu_file_get_error(f
)) {
594 if (f
->xfer_limit
> 0 && f
->bytes_xfer
> f
->xfer_limit
) {
600 int64_t qemu_file_get_rate_limit(QEMUFile
*f
)
602 return f
->xfer_limit
;
605 void qemu_file_set_rate_limit(QEMUFile
*f
, int64_t limit
)
607 f
->xfer_limit
= limit
;
610 void qemu_file_reset_rate_limit(QEMUFile
*f
)
615 void qemu_put_be16(QEMUFile
*f
, unsigned int v
)
617 qemu_put_byte(f
, v
>> 8);
621 void qemu_put_be32(QEMUFile
*f
, unsigned int v
)
623 qemu_put_byte(f
, v
>> 24);
624 qemu_put_byte(f
, v
>> 16);
625 qemu_put_byte(f
, v
>> 8);
629 void qemu_put_be64(QEMUFile
*f
, uint64_t v
)
631 qemu_put_be32(f
, v
>> 32);
635 unsigned int qemu_get_be16(QEMUFile
*f
)
638 v
= qemu_get_byte(f
) << 8;
639 v
|= qemu_get_byte(f
);
643 unsigned int qemu_get_be32(QEMUFile
*f
)
646 v
= (unsigned int)qemu_get_byte(f
) << 24;
647 v
|= qemu_get_byte(f
) << 16;
648 v
|= qemu_get_byte(f
) << 8;
649 v
|= qemu_get_byte(f
);
653 uint64_t qemu_get_be64(QEMUFile
*f
)
656 v
= (uint64_t)qemu_get_be32(f
) << 32;
657 v
|= qemu_get_be32(f
);
661 /* Compress size bytes of data start at p with specific compression
662 * level and store the compressed data to the buffer of f.
664 * When f is not writable, return -1 if f has no space to save the
666 * When f is wirtable and it has no space to save the compressed data,
667 * do fflush first, if f still has no space to save the compressed
671 ssize_t
qemu_put_compression_data(QEMUFile
*f
, const uint8_t *p
, size_t size
,
674 ssize_t blen
= IO_BUF_SIZE
- f
->buf_index
- sizeof(int32_t);
676 if (blen
< compressBound(size
)) {
677 if (!qemu_file_is_writable(f
)) {
681 blen
= IO_BUF_SIZE
- sizeof(int32_t);
682 if (blen
< compressBound(size
)) {
686 if (compress2(f
->buf
+ f
->buf_index
+ sizeof(int32_t), (uLongf
*)&blen
,
687 (Bytef
*)p
, size
, level
) != Z_OK
) {
688 error_report("Compress Failed!");
691 qemu_put_be32(f
, blen
);
692 if (f
->ops
->writev_buffer
) {
693 add_to_iovec(f
, f
->buf
+ f
->buf_index
, blen
, false);
695 f
->buf_index
+= blen
;
696 if (f
->buf_index
== IO_BUF_SIZE
) {
699 return blen
+ sizeof(int32_t);
702 /* Put the data in the buffer of f_src to the buffer of f_des, and
703 * then reset the buf_index of f_src to 0.
706 int qemu_put_qemu_file(QEMUFile
*f_des
, QEMUFile
*f_src
)
710 if (f_src
->buf_index
> 0) {
711 len
= f_src
->buf_index
;
712 qemu_put_buffer(f_des
, f_src
->buf
, f_src
->buf_index
);
713 f_src
->buf_index
= 0;
720 * Get a string whose length is determined by a single preceding byte
721 * A preallocated 256 byte buffer must be passed in.
722 * Returns: len on success and a 0 terminated string in the buffer
724 * (Note a 0 length string will return 0 either way)
726 size_t qemu_get_counted_string(QEMUFile
*f
, char buf
[256])
728 size_t len
= qemu_get_byte(f
);
729 size_t res
= qemu_get_buffer(f
, (uint8_t *)buf
, len
);
733 return res
== len
? res
: 0;
737 * Put a string with one preceding byte containing its length. The length of
738 * the string should be less than 256.
740 void qemu_put_counted_string(QEMUFile
*f
, const char *str
)
742 size_t len
= strlen(str
);
745 qemu_put_byte(f
, len
);
746 qemu_put_buffer(f
, (const uint8_t *)str
, len
);
750 * Set the blocking state of the QEMUFile.
751 * Note: On some transports the OS only keeps a single blocking state for
752 * both directions, and thus changing the blocking on the main
753 * QEMUFile can also affect the return path.
755 void qemu_file_set_blocking(QEMUFile
*f
, bool block
)
757 if (f
->ops
->set_blocking
) {
758 f
->ops
->set_blocking(f
->opaque
, block
);