]> git.proxmox.com Git - mirror_qemu.git/blob - qemu-file.c
Merge remote-tracking branch 'remotes/afaerber/tags/qom-devices-for-2.0' into staging
[mirror_qemu.git] / qemu-file.c
1 #include "qemu-common.h"
2 #include "qemu/iov.h"
3 #include "qemu/sockets.h"
4 #include "block/coroutine.h"
5 #include "migration/migration.h"
6 #include "migration/qemu-file.h"
7 #include "trace.h"
8
9 #define IO_BUF_SIZE 32768
10 #define MAX_IOV_SIZE MIN(IOV_MAX, 64)
11
12 struct QEMUFile {
13 const QEMUFileOps *ops;
14 void *opaque;
15
16 int64_t bytes_xfer;
17 int64_t xfer_limit;
18
19 int64_t pos; /* start of buffer when writing, end of buffer
20 when reading */
21 int buf_index;
22 int buf_size; /* 0 when writing */
23 uint8_t buf[IO_BUF_SIZE];
24
25 struct iovec iov[MAX_IOV_SIZE];
26 unsigned int iovcnt;
27
28 int last_error;
29 };
30
31 typedef struct QEMUFileStdio {
32 FILE *stdio_file;
33 QEMUFile *file;
34 } QEMUFileStdio;
35
36 typedef struct QEMUFileSocket {
37 int fd;
38 QEMUFile *file;
39 } QEMUFileSocket;
40
41 static ssize_t socket_writev_buffer(void *opaque, struct iovec *iov, int iovcnt,
42 int64_t pos)
43 {
44 QEMUFileSocket *s = opaque;
45 ssize_t len;
46 ssize_t size = iov_size(iov, iovcnt);
47
48 len = iov_send(s->fd, iov, iovcnt, 0, size);
49 if (len < size) {
50 len = -socket_error();
51 }
52 return len;
53 }
54
55 static int socket_get_fd(void *opaque)
56 {
57 QEMUFileSocket *s = opaque;
58
59 return s->fd;
60 }
61
62 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
63 {
64 QEMUFileSocket *s = opaque;
65 ssize_t len;
66
67 for (;;) {
68 len = qemu_recv(s->fd, buf, size, 0);
69 if (len != -1) {
70 break;
71 }
72 if (socket_error() == EAGAIN) {
73 yield_until_fd_readable(s->fd);
74 } else if (socket_error() != EINTR) {
75 break;
76 }
77 }
78
79 if (len == -1) {
80 len = -socket_error();
81 }
82 return len;
83 }
84
85 static int socket_close(void *opaque)
86 {
87 QEMUFileSocket *s = opaque;
88 closesocket(s->fd);
89 g_free(s);
90 return 0;
91 }
92
93 static int stdio_get_fd(void *opaque)
94 {
95 QEMUFileStdio *s = opaque;
96
97 return fileno(s->stdio_file);
98 }
99
100 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos,
101 int size)
102 {
103 QEMUFileStdio *s = opaque;
104 int res;
105
106 res = fwrite(buf, 1, size, s->stdio_file);
107
108 if (res != size) {
109 return -errno;
110 }
111 return res;
112 }
113
114 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
115 {
116 QEMUFileStdio *s = opaque;
117 FILE *fp = s->stdio_file;
118 int bytes;
119
120 for (;;) {
121 clearerr(fp);
122 bytes = fread(buf, 1, size, fp);
123 if (bytes != 0 || !ferror(fp)) {
124 break;
125 }
126 if (errno == EAGAIN) {
127 yield_until_fd_readable(fileno(fp));
128 } else if (errno != EINTR) {
129 break;
130 }
131 }
132 return bytes;
133 }
134
135 static int stdio_pclose(void *opaque)
136 {
137 QEMUFileStdio *s = opaque;
138 int ret;
139 ret = pclose(s->stdio_file);
140 if (ret == -1) {
141 ret = -errno;
142 } else if (!WIFEXITED(ret) || WEXITSTATUS(ret) != 0) {
143 /* close succeeded, but non-zero exit code: */
144 ret = -EIO; /* fake errno value */
145 }
146 g_free(s);
147 return ret;
148 }
149
150 static int stdio_fclose(void *opaque)
151 {
152 QEMUFileStdio *s = opaque;
153 int ret = 0;
154
155 if (s->file->ops->put_buffer || s->file->ops->writev_buffer) {
156 int fd = fileno(s->stdio_file);
157 struct stat st;
158
159 ret = fstat(fd, &st);
160 if (ret == 0 && S_ISREG(st.st_mode)) {
161 /*
162 * If the file handle is a regular file make sure the
163 * data is flushed to disk before signaling success.
164 */
165 ret = fsync(fd);
166 if (ret != 0) {
167 ret = -errno;
168 return ret;
169 }
170 }
171 }
172 if (fclose(s->stdio_file) == EOF) {
173 ret = -errno;
174 }
175 g_free(s);
176 return ret;
177 }
178
179 static const QEMUFileOps stdio_pipe_read_ops = {
180 .get_fd = stdio_get_fd,
181 .get_buffer = stdio_get_buffer,
182 .close = stdio_pclose
183 };
184
185 static const QEMUFileOps stdio_pipe_write_ops = {
186 .get_fd = stdio_get_fd,
187 .put_buffer = stdio_put_buffer,
188 .close = stdio_pclose
189 };
190
191 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
192 {
193 FILE *stdio_file;
194 QEMUFileStdio *s;
195
196 if (mode == NULL || (mode[0] != 'r' && mode[0] != 'w') || mode[1] != 0) {
197 fprintf(stderr, "qemu_popen: Argument validity check failed\n");
198 return NULL;
199 }
200
201 stdio_file = popen(command, mode);
202 if (stdio_file == NULL) {
203 return NULL;
204 }
205
206 s = g_malloc0(sizeof(QEMUFileStdio));
207
208 s->stdio_file = stdio_file;
209
210 if (mode[0] == 'r') {
211 s->file = qemu_fopen_ops(s, &stdio_pipe_read_ops);
212 } else {
213 s->file = qemu_fopen_ops(s, &stdio_pipe_write_ops);
214 }
215 return s->file;
216 }
217
218 static const QEMUFileOps stdio_file_read_ops = {
219 .get_fd = stdio_get_fd,
220 .get_buffer = stdio_get_buffer,
221 .close = stdio_fclose
222 };
223
224 static const QEMUFileOps stdio_file_write_ops = {
225 .get_fd = stdio_get_fd,
226 .put_buffer = stdio_put_buffer,
227 .close = stdio_fclose
228 };
229
230 static ssize_t unix_writev_buffer(void *opaque, struct iovec *iov, int iovcnt,
231 int64_t pos)
232 {
233 QEMUFileSocket *s = opaque;
234 ssize_t len, offset;
235 ssize_t size = iov_size(iov, iovcnt);
236 ssize_t total = 0;
237
238 assert(iovcnt > 0);
239 offset = 0;
240 while (size > 0) {
241 /* Find the next start position; skip all full-sized vector elements */
242 while (offset >= iov[0].iov_len) {
243 offset -= iov[0].iov_len;
244 iov++, iovcnt--;
245 }
246
247 /* skip `offset' bytes from the (now) first element, undo it on exit */
248 assert(iovcnt > 0);
249 iov[0].iov_base += offset;
250 iov[0].iov_len -= offset;
251
252 do {
253 len = writev(s->fd, iov, iovcnt);
254 } while (len == -1 && errno == EINTR);
255 if (len == -1) {
256 return -errno;
257 }
258
259 /* Undo the changes above */
260 iov[0].iov_base -= offset;
261 iov[0].iov_len += offset;
262
263 /* Prepare for the next iteration */
264 offset += len;
265 total += len;
266 size -= len;
267 }
268
269 return total;
270 }
271
272 static int unix_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
273 {
274 QEMUFileSocket *s = opaque;
275 ssize_t len;
276
277 for (;;) {
278 len = read(s->fd, buf, size);
279 if (len != -1) {
280 break;
281 }
282 if (errno == EAGAIN) {
283 yield_until_fd_readable(s->fd);
284 } else if (errno != EINTR) {
285 break;
286 }
287 }
288
289 if (len == -1) {
290 len = -errno;
291 }
292 return len;
293 }
294
295 static int unix_close(void *opaque)
296 {
297 QEMUFileSocket *s = opaque;
298 close(s->fd);
299 g_free(s);
300 return 0;
301 }
302
303 static const QEMUFileOps unix_read_ops = {
304 .get_fd = socket_get_fd,
305 .get_buffer = unix_get_buffer,
306 .close = unix_close
307 };
308
309 static const QEMUFileOps unix_write_ops = {
310 .get_fd = socket_get_fd,
311 .writev_buffer = unix_writev_buffer,
312 .close = unix_close
313 };
314
315 QEMUFile *qemu_fdopen(int fd, const char *mode)
316 {
317 QEMUFileSocket *s;
318
319 if (mode == NULL ||
320 (mode[0] != 'r' && mode[0] != 'w') ||
321 mode[1] != 'b' || mode[2] != 0) {
322 fprintf(stderr, "qemu_fdopen: Argument validity check failed\n");
323 return NULL;
324 }
325
326 s = g_malloc0(sizeof(QEMUFileSocket));
327 s->fd = fd;
328
329 if (mode[0] == 'r') {
330 s->file = qemu_fopen_ops(s, &unix_read_ops);
331 } else {
332 s->file = qemu_fopen_ops(s, &unix_write_ops);
333 }
334 return s->file;
335 }
336
337 static const QEMUFileOps socket_read_ops = {
338 .get_fd = socket_get_fd,
339 .get_buffer = socket_get_buffer,
340 .close = socket_close
341 };
342
343 static const QEMUFileOps socket_write_ops = {
344 .get_fd = socket_get_fd,
345 .writev_buffer = socket_writev_buffer,
346 .close = socket_close
347 };
348
349 bool qemu_file_mode_is_not_valid(const char *mode)
350 {
351 if (mode == NULL ||
352 (mode[0] != 'r' && mode[0] != 'w') ||
353 mode[1] != 'b' || mode[2] != 0) {
354 fprintf(stderr, "qemu_fopen: Argument validity check failed\n");
355 return true;
356 }
357
358 return false;
359 }
360
361 QEMUFile *qemu_fopen_socket(int fd, const char *mode)
362 {
363 QEMUFileSocket *s;
364
365 if (qemu_file_mode_is_not_valid(mode)) {
366 return NULL;
367 }
368
369 s = g_malloc0(sizeof(QEMUFileSocket));
370 s->fd = fd;
371 if (mode[0] == 'w') {
372 qemu_set_block(s->fd);
373 s->file = qemu_fopen_ops(s, &socket_write_ops);
374 } else {
375 s->file = qemu_fopen_ops(s, &socket_read_ops);
376 }
377 return s->file;
378 }
379
380 QEMUFile *qemu_fopen(const char *filename, const char *mode)
381 {
382 QEMUFileStdio *s;
383
384 if (qemu_file_mode_is_not_valid(mode)) {
385 return NULL;
386 }
387
388 s = g_malloc0(sizeof(QEMUFileStdio));
389
390 s->stdio_file = fopen(filename, mode);
391 if (!s->stdio_file) {
392 goto fail;
393 }
394
395 if (mode[0] == 'w') {
396 s->file = qemu_fopen_ops(s, &stdio_file_write_ops);
397 } else {
398 s->file = qemu_fopen_ops(s, &stdio_file_read_ops);
399 }
400 return s->file;
401 fail:
402 g_free(s);
403 return NULL;
404 }
405
406 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
407 {
408 QEMUFile *f;
409
410 f = g_malloc0(sizeof(QEMUFile));
411
412 f->opaque = opaque;
413 f->ops = ops;
414 return f;
415 }
416
417 /*
418 * Get last error for stream f
419 *
420 * Return negative error value if there has been an error on previous
421 * operations, return 0 if no error happened.
422 *
423 */
424 int qemu_file_get_error(QEMUFile *f)
425 {
426 return f->last_error;
427 }
428
429 void qemu_file_set_error(QEMUFile *f, int ret)
430 {
431 if (f->last_error == 0) {
432 f->last_error = ret;
433 }
434 }
435
436 static inline bool qemu_file_is_writable(QEMUFile *f)
437 {
438 return f->ops->writev_buffer || f->ops->put_buffer;
439 }
440
441 /**
442 * Flushes QEMUFile buffer
443 *
444 * If there is writev_buffer QEMUFileOps it uses it otherwise uses
445 * put_buffer ops.
446 */
447 void qemu_fflush(QEMUFile *f)
448 {
449 ssize_t ret = 0;
450
451 if (!qemu_file_is_writable(f)) {
452 return;
453 }
454
455 if (f->ops->writev_buffer) {
456 if (f->iovcnt > 0) {
457 ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt, f->pos);
458 }
459 } else {
460 if (f->buf_index > 0) {
461 ret = f->ops->put_buffer(f->opaque, f->buf, f->pos, f->buf_index);
462 }
463 }
464 if (ret >= 0) {
465 f->pos += ret;
466 }
467 f->buf_index = 0;
468 f->iovcnt = 0;
469 if (ret < 0) {
470 qemu_file_set_error(f, ret);
471 }
472 }
473
474 void ram_control_before_iterate(QEMUFile *f, uint64_t flags)
475 {
476 int ret = 0;
477
478 if (f->ops->before_ram_iterate) {
479 ret = f->ops->before_ram_iterate(f, f->opaque, flags);
480 if (ret < 0) {
481 qemu_file_set_error(f, ret);
482 }
483 }
484 }
485
486 void ram_control_after_iterate(QEMUFile *f, uint64_t flags)
487 {
488 int ret = 0;
489
490 if (f->ops->after_ram_iterate) {
491 ret = f->ops->after_ram_iterate(f, f->opaque, flags);
492 if (ret < 0) {
493 qemu_file_set_error(f, ret);
494 }
495 }
496 }
497
498 void ram_control_load_hook(QEMUFile *f, uint64_t flags)
499 {
500 int ret = -EINVAL;
501
502 if (f->ops->hook_ram_load) {
503 ret = f->ops->hook_ram_load(f, f->opaque, flags);
504 if (ret < 0) {
505 qemu_file_set_error(f, ret);
506 }
507 } else {
508 qemu_file_set_error(f, ret);
509 }
510 }
511
512 size_t ram_control_save_page(QEMUFile *f, ram_addr_t block_offset,
513 ram_addr_t offset, size_t size, int *bytes_sent)
514 {
515 if (f->ops->save_page) {
516 int ret = f->ops->save_page(f, f->opaque, block_offset,
517 offset, size, bytes_sent);
518
519 if (ret != RAM_SAVE_CONTROL_DELAYED) {
520 if (bytes_sent && *bytes_sent > 0) {
521 qemu_update_position(f, *bytes_sent);
522 } else if (ret < 0) {
523 qemu_file_set_error(f, ret);
524 }
525 }
526
527 return ret;
528 }
529
530 return RAM_SAVE_CONTROL_NOT_SUPP;
531 }
532
533 static void qemu_fill_buffer(QEMUFile *f)
534 {
535 int len;
536 int pending;
537
538 assert(!qemu_file_is_writable(f));
539
540 pending = f->buf_size - f->buf_index;
541 if (pending > 0) {
542 memmove(f->buf, f->buf + f->buf_index, pending);
543 }
544 f->buf_index = 0;
545 f->buf_size = pending;
546
547 len = f->ops->get_buffer(f->opaque, f->buf + pending, f->pos,
548 IO_BUF_SIZE - pending);
549 if (len > 0) {
550 f->buf_size += len;
551 f->pos += len;
552 } else if (len == 0) {
553 qemu_file_set_error(f, -EIO);
554 } else if (len != -EAGAIN) {
555 qemu_file_set_error(f, len);
556 }
557 }
558
559 int qemu_get_fd(QEMUFile *f)
560 {
561 if (f->ops->get_fd) {
562 return f->ops->get_fd(f->opaque);
563 }
564 return -1;
565 }
566
567 void qemu_update_position(QEMUFile *f, size_t size)
568 {
569 f->pos += size;
570 }
571
572 /** Closes the file
573 *
574 * Returns negative error value if any error happened on previous operations or
575 * while closing the file. Returns 0 or positive number on success.
576 *
577 * The meaning of return value on success depends on the specific backend
578 * being used.
579 */
580 int qemu_fclose(QEMUFile *f)
581 {
582 int ret;
583 qemu_fflush(f);
584 ret = qemu_file_get_error(f);
585
586 if (f->ops->close) {
587 int ret2 = f->ops->close(f->opaque);
588 if (ret >= 0) {
589 ret = ret2;
590 }
591 }
592 /* If any error was spotted before closing, we should report it
593 * instead of the close() return value.
594 */
595 if (f->last_error) {
596 ret = f->last_error;
597 }
598 g_free(f);
599 trace_qemu_file_fclose();
600 return ret;
601 }
602
603 static void add_to_iovec(QEMUFile *f, const uint8_t *buf, int size)
604 {
605 /* check for adjacent buffer and coalesce them */
606 if (f->iovcnt > 0 && buf == f->iov[f->iovcnt - 1].iov_base +
607 f->iov[f->iovcnt - 1].iov_len) {
608 f->iov[f->iovcnt - 1].iov_len += size;
609 } else {
610 f->iov[f->iovcnt].iov_base = (uint8_t *)buf;
611 f->iov[f->iovcnt++].iov_len = size;
612 }
613
614 if (f->iovcnt >= MAX_IOV_SIZE) {
615 qemu_fflush(f);
616 }
617 }
618
619 void qemu_put_buffer_async(QEMUFile *f, const uint8_t *buf, int size)
620 {
621 if (!f->ops->writev_buffer) {
622 qemu_put_buffer(f, buf, size);
623 return;
624 }
625
626 if (f->last_error) {
627 return;
628 }
629
630 f->bytes_xfer += size;
631 add_to_iovec(f, buf, size);
632 }
633
634 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
635 {
636 int l;
637
638 if (f->last_error) {
639 return;
640 }
641
642 while (size > 0) {
643 l = IO_BUF_SIZE - f->buf_index;
644 if (l > size) {
645 l = size;
646 }
647 memcpy(f->buf + f->buf_index, buf, l);
648 f->bytes_xfer += l;
649 if (f->ops->writev_buffer) {
650 add_to_iovec(f, f->buf + f->buf_index, l);
651 }
652 f->buf_index += l;
653 if (f->buf_index == IO_BUF_SIZE) {
654 qemu_fflush(f);
655 }
656 if (qemu_file_get_error(f)) {
657 break;
658 }
659 buf += l;
660 size -= l;
661 }
662 }
663
664 void qemu_put_byte(QEMUFile *f, int v)
665 {
666 if (f->last_error) {
667 return;
668 }
669
670 f->buf[f->buf_index] = v;
671 f->bytes_xfer++;
672 if (f->ops->writev_buffer) {
673 add_to_iovec(f, f->buf + f->buf_index, 1);
674 }
675 f->buf_index++;
676 if (f->buf_index == IO_BUF_SIZE) {
677 qemu_fflush(f);
678 }
679 }
680
681 void qemu_file_skip(QEMUFile *f, int size)
682 {
683 if (f->buf_index + size <= f->buf_size) {
684 f->buf_index += size;
685 }
686 }
687
688 int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset)
689 {
690 int pending;
691 int index;
692
693 assert(!qemu_file_is_writable(f));
694
695 index = f->buf_index + offset;
696 pending = f->buf_size - index;
697 if (pending < size) {
698 qemu_fill_buffer(f);
699 index = f->buf_index + offset;
700 pending = f->buf_size - index;
701 }
702
703 if (pending <= 0) {
704 return 0;
705 }
706 if (size > pending) {
707 size = pending;
708 }
709
710 memcpy(buf, f->buf + index, size);
711 return size;
712 }
713
714 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size)
715 {
716 int pending = size;
717 int done = 0;
718
719 while (pending > 0) {
720 int res;
721
722 res = qemu_peek_buffer(f, buf, pending, 0);
723 if (res == 0) {
724 return done;
725 }
726 qemu_file_skip(f, res);
727 buf += res;
728 pending -= res;
729 done += res;
730 }
731 return done;
732 }
733
734 int qemu_peek_byte(QEMUFile *f, int offset)
735 {
736 int index = f->buf_index + offset;
737
738 assert(!qemu_file_is_writable(f));
739
740 if (index >= f->buf_size) {
741 qemu_fill_buffer(f);
742 index = f->buf_index + offset;
743 if (index >= f->buf_size) {
744 return 0;
745 }
746 }
747 return f->buf[index];
748 }
749
750 int qemu_get_byte(QEMUFile *f)
751 {
752 int result;
753
754 result = qemu_peek_byte(f, 0);
755 qemu_file_skip(f, 1);
756 return result;
757 }
758
759 int64_t qemu_ftell(QEMUFile *f)
760 {
761 qemu_fflush(f);
762 return f->pos;
763 }
764
765 int qemu_file_rate_limit(QEMUFile *f)
766 {
767 if (qemu_file_get_error(f)) {
768 return 1;
769 }
770 if (f->xfer_limit > 0 && f->bytes_xfer > f->xfer_limit) {
771 return 1;
772 }
773 return 0;
774 }
775
776 int64_t qemu_file_get_rate_limit(QEMUFile *f)
777 {
778 return f->xfer_limit;
779 }
780
781 void qemu_file_set_rate_limit(QEMUFile *f, int64_t limit)
782 {
783 f->xfer_limit = limit;
784 }
785
786 void qemu_file_reset_rate_limit(QEMUFile *f)
787 {
788 f->bytes_xfer = 0;
789 }
790
791 void qemu_put_be16(QEMUFile *f, unsigned int v)
792 {
793 qemu_put_byte(f, v >> 8);
794 qemu_put_byte(f, v);
795 }
796
797 void qemu_put_be32(QEMUFile *f, unsigned int v)
798 {
799 qemu_put_byte(f, v >> 24);
800 qemu_put_byte(f, v >> 16);
801 qemu_put_byte(f, v >> 8);
802 qemu_put_byte(f, v);
803 }
804
805 void qemu_put_be64(QEMUFile *f, uint64_t v)
806 {
807 qemu_put_be32(f, v >> 32);
808 qemu_put_be32(f, v);
809 }
810
811 unsigned int qemu_get_be16(QEMUFile *f)
812 {
813 unsigned int v;
814 v = qemu_get_byte(f) << 8;
815 v |= qemu_get_byte(f);
816 return v;
817 }
818
819 unsigned int qemu_get_be32(QEMUFile *f)
820 {
821 unsigned int v;
822 v = qemu_get_byte(f) << 24;
823 v |= qemu_get_byte(f) << 16;
824 v |= qemu_get_byte(f) << 8;
825 v |= qemu_get_byte(f);
826 return v;
827 }
828
829 uint64_t qemu_get_be64(QEMUFile *f)
830 {
831 uint64_t v;
832 v = (uint64_t)qemu_get_be32(f) << 32;
833 v |= qemu_get_be32(f);
834 return v;
835 }