spapr: Add kvm-type property
[qemu/kevin.git] / qemu-file.c
bloba8e39127f287ec55324699cb1d98fd65f191af68
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"
9 #define IO_BUF_SIZE 32768
10 #define MAX_IOV_SIZE MIN(IOV_MAX, 64)
12 struct QEMUFile {
13 const QEMUFileOps *ops;
14 void *opaque;
16 int64_t bytes_xfer;
17 int64_t xfer_limit;
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];
25 struct iovec iov[MAX_IOV_SIZE];
26 unsigned int iovcnt;
28 int last_error;
31 typedef struct QEMUFileStdio {
32 FILE *stdio_file;
33 QEMUFile *file;
34 } QEMUFileStdio;
36 typedef struct QEMUFileSocket {
37 int fd;
38 QEMUFile *file;
39 } QEMUFileSocket;
41 static ssize_t socket_writev_buffer(void *opaque, struct iovec *iov, int iovcnt,
42 int64_t pos)
44 QEMUFileSocket *s = opaque;
45 ssize_t len;
46 ssize_t size = iov_size(iov, iovcnt);
48 len = iov_send(s->fd, iov, iovcnt, 0, size);
49 if (len < size) {
50 len = -socket_error();
52 return len;
55 static int socket_get_fd(void *opaque)
57 QEMUFileSocket *s = opaque;
59 return s->fd;
62 static int socket_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
64 QEMUFileSocket *s = opaque;
65 ssize_t len;
67 for (;;) {
68 len = qemu_recv(s->fd, buf, size, 0);
69 if (len != -1) {
70 break;
72 if (socket_error() == EAGAIN) {
73 yield_until_fd_readable(s->fd);
74 } else if (socket_error() != EINTR) {
75 break;
79 if (len == -1) {
80 len = -socket_error();
82 return len;
85 static int socket_close(void *opaque)
87 QEMUFileSocket *s = opaque;
88 closesocket(s->fd);
89 g_free(s);
90 return 0;
93 static int stdio_get_fd(void *opaque)
95 QEMUFileStdio *s = opaque;
97 return fileno(s->stdio_file);
100 static int stdio_put_buffer(void *opaque, const uint8_t *buf, int64_t pos,
101 int size)
103 QEMUFileStdio *s = opaque;
104 int res;
106 res = fwrite(buf, 1, size, s->stdio_file);
108 if (res != size) {
109 return -errno;
111 return res;
114 static int stdio_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
116 QEMUFileStdio *s = opaque;
117 FILE *fp = s->stdio_file;
118 int bytes;
120 for (;;) {
121 clearerr(fp);
122 bytes = fread(buf, 1, size, fp);
123 if (bytes != 0 || !ferror(fp)) {
124 break;
126 if (errno == EAGAIN) {
127 yield_until_fd_readable(fileno(fp));
128 } else if (errno != EINTR) {
129 break;
132 return bytes;
135 static int stdio_pclose(void *opaque)
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 */
146 g_free(s);
147 return ret;
150 static int stdio_fclose(void *opaque)
152 QEMUFileStdio *s = opaque;
153 int ret = 0;
155 if (s->file->ops->put_buffer || s->file->ops->writev_buffer) {
156 int fd = fileno(s->stdio_file);
157 struct stat st;
159 ret = fstat(fd, &st);
160 if (ret == 0 && S_ISREG(st.st_mode)) {
162 * If the file handle is a regular file make sure the
163 * data is flushed to disk before signaling success.
165 ret = fsync(fd);
166 if (ret != 0) {
167 ret = -errno;
168 return ret;
172 if (fclose(s->stdio_file) == EOF) {
173 ret = -errno;
175 g_free(s);
176 return ret;
179 static const QEMUFileOps stdio_pipe_read_ops = {
180 .get_fd = stdio_get_fd,
181 .get_buffer = stdio_get_buffer,
182 .close = stdio_pclose
185 static const QEMUFileOps stdio_pipe_write_ops = {
186 .get_fd = stdio_get_fd,
187 .put_buffer = stdio_put_buffer,
188 .close = stdio_pclose
191 QEMUFile *qemu_popen_cmd(const char *command, const char *mode)
193 FILE *stdio_file;
194 QEMUFileStdio *s;
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;
201 stdio_file = popen(command, mode);
202 if (stdio_file == NULL) {
203 return NULL;
206 s = g_malloc0(sizeof(QEMUFileStdio));
208 s->stdio_file = stdio_file;
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);
215 return s->file;
218 static const QEMUFileOps stdio_file_read_ops = {
219 .get_fd = stdio_get_fd,
220 .get_buffer = stdio_get_buffer,
221 .close = stdio_fclose
224 static const QEMUFileOps stdio_file_write_ops = {
225 .get_fd = stdio_get_fd,
226 .put_buffer = stdio_put_buffer,
227 .close = stdio_fclose
230 static ssize_t unix_writev_buffer(void *opaque, struct iovec *iov, int iovcnt,
231 int64_t pos)
233 QEMUFileSocket *s = opaque;
234 ssize_t len, offset;
235 ssize_t size = iov_size(iov, iovcnt);
236 ssize_t total = 0;
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--;
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;
252 do {
253 len = writev(s->fd, iov, iovcnt);
254 } while (len == -1 && errno == EINTR);
255 if (len == -1) {
256 return -errno;
259 /* Undo the changes above */
260 iov[0].iov_base -= offset;
261 iov[0].iov_len += offset;
263 /* Prepare for the next iteration */
264 offset += len;
265 total += len;
266 size -= len;
269 return total;
272 static int unix_get_buffer(void *opaque, uint8_t *buf, int64_t pos, int size)
274 QEMUFileSocket *s = opaque;
275 ssize_t len;
277 for (;;) {
278 len = read(s->fd, buf, size);
279 if (len != -1) {
280 break;
282 if (errno == EAGAIN) {
283 yield_until_fd_readable(s->fd);
284 } else if (errno != EINTR) {
285 break;
289 if (len == -1) {
290 len = -errno;
292 return len;
295 static int unix_close(void *opaque)
297 QEMUFileSocket *s = opaque;
298 close(s->fd);
299 g_free(s);
300 return 0;
303 static const QEMUFileOps unix_read_ops = {
304 .get_fd = socket_get_fd,
305 .get_buffer = unix_get_buffer,
306 .close = unix_close
309 static const QEMUFileOps unix_write_ops = {
310 .get_fd = socket_get_fd,
311 .writev_buffer = unix_writev_buffer,
312 .close = unix_close
315 QEMUFile *qemu_fdopen(int fd, const char *mode)
317 QEMUFileSocket *s;
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;
326 s = g_malloc0(sizeof(QEMUFileSocket));
327 s->fd = fd;
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);
334 return s->file;
337 static const QEMUFileOps socket_read_ops = {
338 .get_fd = socket_get_fd,
339 .get_buffer = socket_get_buffer,
340 .close = socket_close
343 static const QEMUFileOps socket_write_ops = {
344 .get_fd = socket_get_fd,
345 .writev_buffer = socket_writev_buffer,
346 .close = socket_close
349 bool qemu_file_mode_is_not_valid(const char *mode)
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;
358 return false;
361 QEMUFile *qemu_fopen_socket(int fd, const char *mode)
363 QEMUFileSocket *s;
365 if (qemu_file_mode_is_not_valid(mode)) {
366 return NULL;
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);
377 return s->file;
380 QEMUFile *qemu_fopen(const char *filename, const char *mode)
382 QEMUFileStdio *s;
384 if (qemu_file_mode_is_not_valid(mode)) {
385 return NULL;
388 s = g_malloc0(sizeof(QEMUFileStdio));
390 s->stdio_file = fopen(filename, mode);
391 if (!s->stdio_file) {
392 goto fail;
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);
400 return s->file;
401 fail:
402 g_free(s);
403 return NULL;
406 QEMUFile *qemu_fopen_ops(void *opaque, const QEMUFileOps *ops)
408 QEMUFile *f;
410 f = g_malloc0(sizeof(QEMUFile));
412 f->opaque = opaque;
413 f->ops = ops;
414 return f;
418 * Get last error for stream f
420 * Return negative error value if there has been an error on previous
421 * operations, return 0 if no error happened.
424 int qemu_file_get_error(QEMUFile *f)
426 return f->last_error;
429 void qemu_file_set_error(QEMUFile *f, int ret)
431 if (f->last_error == 0) {
432 f->last_error = ret;
436 static inline bool qemu_file_is_writable(QEMUFile *f)
438 return f->ops->writev_buffer || f->ops->put_buffer;
442 * Flushes QEMUFile buffer
444 * If there is writev_buffer QEMUFileOps it uses it otherwise uses
445 * put_buffer ops.
447 void qemu_fflush(QEMUFile *f)
449 ssize_t ret = 0;
451 if (!qemu_file_is_writable(f)) {
452 return;
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);
459 } else {
460 if (f->buf_index > 0) {
461 ret = f->ops->put_buffer(f->opaque, f->buf, f->pos, f->buf_index);
464 if (ret >= 0) {
465 f->pos += ret;
467 f->buf_index = 0;
468 f->iovcnt = 0;
469 if (ret < 0) {
470 qemu_file_set_error(f, ret);
474 void ram_control_before_iterate(QEMUFile *f, uint64_t flags)
476 int ret = 0;
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);
486 void ram_control_after_iterate(QEMUFile *f, uint64_t flags)
488 int ret = 0;
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);
498 void ram_control_load_hook(QEMUFile *f, uint64_t flags)
500 int ret = -EINVAL;
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);
507 } else {
508 qemu_file_set_error(f, ret);
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)
515 if (f->ops->save_page) {
516 int ret = f->ops->save_page(f, f->opaque, block_offset,
517 offset, size, bytes_sent);
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);
527 return ret;
530 return RAM_SAVE_CONTROL_NOT_SUPP;
534 * Attempt to fill the buffer from the underlying file
535 * Returns the number of bytes read, or negative value for an error.
537 * Note that it can return a partially full buffer even in a not error/not EOF
538 * case if the underlying file descriptor gives a short read, and that can
539 * happen even on a blocking fd.
541 static ssize_t qemu_fill_buffer(QEMUFile *f)
543 int len;
544 int pending;
546 assert(!qemu_file_is_writable(f));
548 pending = f->buf_size - f->buf_index;
549 if (pending > 0) {
550 memmove(f->buf, f->buf + f->buf_index, pending);
552 f->buf_index = 0;
553 f->buf_size = pending;
555 len = f->ops->get_buffer(f->opaque, f->buf + pending, f->pos,
556 IO_BUF_SIZE - pending);
557 if (len > 0) {
558 f->buf_size += len;
559 f->pos += len;
560 } else if (len == 0) {
561 qemu_file_set_error(f, -EIO);
562 } else if (len != -EAGAIN) {
563 qemu_file_set_error(f, len);
566 return len;
569 int qemu_get_fd(QEMUFile *f)
571 if (f->ops->get_fd) {
572 return f->ops->get_fd(f->opaque);
574 return -1;
577 void qemu_update_position(QEMUFile *f, size_t size)
579 f->pos += size;
582 /** Closes the file
584 * Returns negative error value if any error happened on previous operations or
585 * while closing the file. Returns 0 or positive number on success.
587 * The meaning of return value on success depends on the specific backend
588 * being used.
590 int qemu_fclose(QEMUFile *f)
592 int ret;
593 qemu_fflush(f);
594 ret = qemu_file_get_error(f);
596 if (f->ops->close) {
597 int ret2 = f->ops->close(f->opaque);
598 if (ret >= 0) {
599 ret = ret2;
602 /* If any error was spotted before closing, we should report it
603 * instead of the close() return value.
605 if (f->last_error) {
606 ret = f->last_error;
608 g_free(f);
609 trace_qemu_file_fclose();
610 return ret;
613 static void add_to_iovec(QEMUFile *f, const uint8_t *buf, int size)
615 /* check for adjacent buffer and coalesce them */
616 if (f->iovcnt > 0 && buf == f->iov[f->iovcnt - 1].iov_base +
617 f->iov[f->iovcnt - 1].iov_len) {
618 f->iov[f->iovcnt - 1].iov_len += size;
619 } else {
620 f->iov[f->iovcnt].iov_base = (uint8_t *)buf;
621 f->iov[f->iovcnt++].iov_len = size;
624 if (f->iovcnt >= MAX_IOV_SIZE) {
625 qemu_fflush(f);
629 void qemu_put_buffer_async(QEMUFile *f, const uint8_t *buf, int size)
631 if (!f->ops->writev_buffer) {
632 qemu_put_buffer(f, buf, size);
633 return;
636 if (f->last_error) {
637 return;
640 f->bytes_xfer += size;
641 add_to_iovec(f, buf, size);
644 void qemu_put_buffer(QEMUFile *f, const uint8_t *buf, int size)
646 int l;
648 if (f->last_error) {
649 return;
652 while (size > 0) {
653 l = IO_BUF_SIZE - f->buf_index;
654 if (l > size) {
655 l = size;
657 memcpy(f->buf + f->buf_index, buf, l);
658 f->bytes_xfer += l;
659 if (f->ops->writev_buffer) {
660 add_to_iovec(f, f->buf + f->buf_index, l);
662 f->buf_index += l;
663 if (f->buf_index == IO_BUF_SIZE) {
664 qemu_fflush(f);
666 if (qemu_file_get_error(f)) {
667 break;
669 buf += l;
670 size -= l;
674 void qemu_put_byte(QEMUFile *f, int v)
676 if (f->last_error) {
677 return;
680 f->buf[f->buf_index] = v;
681 f->bytes_xfer++;
682 if (f->ops->writev_buffer) {
683 add_to_iovec(f, f->buf + f->buf_index, 1);
685 f->buf_index++;
686 if (f->buf_index == IO_BUF_SIZE) {
687 qemu_fflush(f);
691 void qemu_file_skip(QEMUFile *f, int size)
693 if (f->buf_index + size <= f->buf_size) {
694 f->buf_index += size;
699 * Read 'size' bytes from file (at 'offset') into buf without moving the
700 * pointer.
702 * It will return size bytes unless there was an error, in which case it will
703 * return as many as it managed to read (assuming blocking fd's which
704 * all current QEMUFile are)
706 int qemu_peek_buffer(QEMUFile *f, uint8_t *buf, int size, size_t offset)
708 int pending;
709 int index;
711 assert(!qemu_file_is_writable(f));
712 assert(offset < IO_BUF_SIZE);
713 assert(size <= IO_BUF_SIZE - offset);
715 /* The 1st byte to read from */
716 index = f->buf_index + offset;
717 /* The number of available bytes starting at index */
718 pending = f->buf_size - index;
721 * qemu_fill_buffer might return just a few bytes, even when there isn't
722 * an error, so loop collecting them until we get enough.
724 while (pending < size) {
725 int received = qemu_fill_buffer(f);
727 if (received <= 0) {
728 break;
731 index = f->buf_index + offset;
732 pending = f->buf_size - index;
735 if (pending <= 0) {
736 return 0;
738 if (size > pending) {
739 size = pending;
742 memcpy(buf, f->buf + index, size);
743 return size;
747 * Read 'size' bytes of data from the file into buf.
748 * 'size' can be larger than the internal buffer.
750 * It will return size bytes unless there was an error, in which case it will
751 * return as many as it managed to read (assuming blocking fd's which
752 * all current QEMUFile are)
754 int qemu_get_buffer(QEMUFile *f, uint8_t *buf, int size)
756 int pending = size;
757 int done = 0;
759 while (pending > 0) {
760 int res;
762 res = qemu_peek_buffer(f, buf, MIN(pending, IO_BUF_SIZE), 0);
763 if (res == 0) {
764 return done;
766 qemu_file_skip(f, res);
767 buf += res;
768 pending -= res;
769 done += res;
771 return done;
775 * Peeks a single byte from the buffer; this isn't guaranteed to work if
776 * offset leaves a gap after the previous read/peeked data.
778 int qemu_peek_byte(QEMUFile *f, int offset)
780 int index = f->buf_index + offset;
782 assert(!qemu_file_is_writable(f));
783 assert(offset < IO_BUF_SIZE);
785 if (index >= f->buf_size) {
786 qemu_fill_buffer(f);
787 index = f->buf_index + offset;
788 if (index >= f->buf_size) {
789 return 0;
792 return f->buf[index];
795 int qemu_get_byte(QEMUFile *f)
797 int result;
799 result = qemu_peek_byte(f, 0);
800 qemu_file_skip(f, 1);
801 return result;
804 int64_t qemu_ftell(QEMUFile *f)
806 qemu_fflush(f);
807 return f->pos;
810 int qemu_file_rate_limit(QEMUFile *f)
812 if (qemu_file_get_error(f)) {
813 return 1;
815 if (f->xfer_limit > 0 && f->bytes_xfer > f->xfer_limit) {
816 return 1;
818 return 0;
821 int64_t qemu_file_get_rate_limit(QEMUFile *f)
823 return f->xfer_limit;
826 void qemu_file_set_rate_limit(QEMUFile *f, int64_t limit)
828 f->xfer_limit = limit;
831 void qemu_file_reset_rate_limit(QEMUFile *f)
833 f->bytes_xfer = 0;
836 void qemu_put_be16(QEMUFile *f, unsigned int v)
838 qemu_put_byte(f, v >> 8);
839 qemu_put_byte(f, v);
842 void qemu_put_be32(QEMUFile *f, unsigned int v)
844 qemu_put_byte(f, v >> 24);
845 qemu_put_byte(f, v >> 16);
846 qemu_put_byte(f, v >> 8);
847 qemu_put_byte(f, v);
850 void qemu_put_be64(QEMUFile *f, uint64_t v)
852 qemu_put_be32(f, v >> 32);
853 qemu_put_be32(f, v);
856 unsigned int qemu_get_be16(QEMUFile *f)
858 unsigned int v;
859 v = qemu_get_byte(f) << 8;
860 v |= qemu_get_byte(f);
861 return v;
864 unsigned int qemu_get_be32(QEMUFile *f)
866 unsigned int v;
867 v = qemu_get_byte(f) << 24;
868 v |= qemu_get_byte(f) << 16;
869 v |= qemu_get_byte(f) << 8;
870 v |= qemu_get_byte(f);
871 return v;
874 uint64_t qemu_get_be64(QEMUFile *f)
876 uint64_t v;
877 v = (uint64_t)qemu_get_be32(f) << 32;
878 v |= qemu_get_be32(f);
879 return v;