xref: /openbmc/qemu/block.c (revision 03feae73056ba3223151c31871860e30630645ac)
1 /*
2  * QEMU System Emulator block driver
3  *
4  * Copyright (c) 2003 Fabrice Bellard
5  *
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:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
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.
23  */
24 #include "config-host.h"
25 #include "qemu-common.h"
26 #include "trace.h"
27 #include "monitor.h"
28 #include "block_int.h"
29 #include "module.h"
30 #include "qemu-objects.h"
31 
32 #ifdef CONFIG_BSD
33 #include <sys/types.h>
34 #include <sys/stat.h>
35 #include <sys/ioctl.h>
36 #include <sys/queue.h>
37 #ifndef __DragonFly__
38 #include <sys/disk.h>
39 #endif
40 #endif
41 
42 #ifdef _WIN32
43 #include <windows.h>
44 #endif
45 
46 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
47         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
48         BlockDriverCompletionFunc *cb, void *opaque);
49 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
50         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
51         BlockDriverCompletionFunc *cb, void *opaque);
52 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
53         BlockDriverCompletionFunc *cb, void *opaque);
54 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
55         BlockDriverCompletionFunc *cb, void *opaque);
56 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
57                         uint8_t *buf, int nb_sectors);
58 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
59                          const uint8_t *buf, int nb_sectors);
60 
61 static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
62     QTAILQ_HEAD_INITIALIZER(bdrv_states);
63 
64 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
65     QLIST_HEAD_INITIALIZER(bdrv_drivers);
66 
67 /* The device to use for VM snapshots */
68 static BlockDriverState *bs_snapshots;
69 
70 /* If non-zero, use only whitelisted block drivers */
71 static int use_bdrv_whitelist;
72 
73 #ifdef _WIN32
74 static int is_windows_drive_prefix(const char *filename)
75 {
76     return (((filename[0] >= 'a' && filename[0] <= 'z') ||
77              (filename[0] >= 'A' && filename[0] <= 'Z')) &&
78             filename[1] == ':');
79 }
80 
81 int is_windows_drive(const char *filename)
82 {
83     if (is_windows_drive_prefix(filename) &&
84         filename[2] == '\0')
85         return 1;
86     if (strstart(filename, "\\\\.\\", NULL) ||
87         strstart(filename, "//./", NULL))
88         return 1;
89     return 0;
90 }
91 #endif
92 
93 /* check if the path starts with "<protocol>:" */
94 static int path_has_protocol(const char *path)
95 {
96 #ifdef _WIN32
97     if (is_windows_drive(path) ||
98         is_windows_drive_prefix(path)) {
99         return 0;
100     }
101 #endif
102 
103     return strchr(path, ':') != NULL;
104 }
105 
106 int path_is_absolute(const char *path)
107 {
108     const char *p;
109 #ifdef _WIN32
110     /* specific case for names like: "\\.\d:" */
111     if (*path == '/' || *path == '\\')
112         return 1;
113 #endif
114     p = strchr(path, ':');
115     if (p)
116         p++;
117     else
118         p = path;
119 #ifdef _WIN32
120     return (*p == '/' || *p == '\\');
121 #else
122     return (*p == '/');
123 #endif
124 }
125 
126 /* if filename is absolute, just copy it to dest. Otherwise, build a
127    path to it by considering it is relative to base_path. URL are
128    supported. */
129 void path_combine(char *dest, int dest_size,
130                   const char *base_path,
131                   const char *filename)
132 {
133     const char *p, *p1;
134     int len;
135 
136     if (dest_size <= 0)
137         return;
138     if (path_is_absolute(filename)) {
139         pstrcpy(dest, dest_size, filename);
140     } else {
141         p = strchr(base_path, ':');
142         if (p)
143             p++;
144         else
145             p = base_path;
146         p1 = strrchr(base_path, '/');
147 #ifdef _WIN32
148         {
149             const char *p2;
150             p2 = strrchr(base_path, '\\');
151             if (!p1 || p2 > p1)
152                 p1 = p2;
153         }
154 #endif
155         if (p1)
156             p1++;
157         else
158             p1 = base_path;
159         if (p1 > p)
160             p = p1;
161         len = p - base_path;
162         if (len > dest_size - 1)
163             len = dest_size - 1;
164         memcpy(dest, base_path, len);
165         dest[len] = '\0';
166         pstrcat(dest, dest_size, filename);
167     }
168 }
169 
170 void bdrv_register(BlockDriver *bdrv)
171 {
172     if (!bdrv->bdrv_aio_readv) {
173         /* add AIO emulation layer */
174         bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
175         bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
176     } else if (!bdrv->bdrv_read) {
177         /* add synchronous IO emulation layer */
178         bdrv->bdrv_read = bdrv_read_em;
179         bdrv->bdrv_write = bdrv_write_em;
180     }
181 
182     if (!bdrv->bdrv_aio_flush)
183         bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
184 
185     QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
186 }
187 
188 /* create a new block device (by default it is empty) */
189 BlockDriverState *bdrv_new(const char *device_name)
190 {
191     BlockDriverState *bs;
192 
193     bs = qemu_mallocz(sizeof(BlockDriverState));
194     pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
195     if (device_name[0] != '\0') {
196         QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
197     }
198     return bs;
199 }
200 
201 BlockDriver *bdrv_find_format(const char *format_name)
202 {
203     BlockDriver *drv1;
204     QLIST_FOREACH(drv1, &bdrv_drivers, list) {
205         if (!strcmp(drv1->format_name, format_name)) {
206             return drv1;
207         }
208     }
209     return NULL;
210 }
211 
212 static int bdrv_is_whitelisted(BlockDriver *drv)
213 {
214     static const char *whitelist[] = {
215         CONFIG_BDRV_WHITELIST
216     };
217     const char **p;
218 
219     if (!whitelist[0])
220         return 1;               /* no whitelist, anything goes */
221 
222     for (p = whitelist; *p; p++) {
223         if (!strcmp(drv->format_name, *p)) {
224             return 1;
225         }
226     }
227     return 0;
228 }
229 
230 BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
231 {
232     BlockDriver *drv = bdrv_find_format(format_name);
233     return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
234 }
235 
236 int bdrv_create(BlockDriver *drv, const char* filename,
237     QEMUOptionParameter *options)
238 {
239     if (!drv->bdrv_create)
240         return -ENOTSUP;
241 
242     return drv->bdrv_create(filename, options);
243 }
244 
245 int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
246 {
247     BlockDriver *drv;
248 
249     drv = bdrv_find_protocol(filename);
250     if (drv == NULL) {
251         return -ENOENT;
252     }
253 
254     return bdrv_create(drv, filename, options);
255 }
256 
257 #ifdef _WIN32
258 void get_tmp_filename(char *filename, int size)
259 {
260     char temp_dir[MAX_PATH];
261 
262     GetTempPath(MAX_PATH, temp_dir);
263     GetTempFileName(temp_dir, "qem", 0, filename);
264 }
265 #else
266 void get_tmp_filename(char *filename, int size)
267 {
268     int fd;
269     const char *tmpdir;
270     /* XXX: race condition possible */
271     tmpdir = getenv("TMPDIR");
272     if (!tmpdir)
273         tmpdir = "/tmp";
274     snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
275     fd = mkstemp(filename);
276     close(fd);
277 }
278 #endif
279 
280 /*
281  * Detect host devices. By convention, /dev/cdrom[N] is always
282  * recognized as a host CDROM.
283  */
284 static BlockDriver *find_hdev_driver(const char *filename)
285 {
286     int score_max = 0, score;
287     BlockDriver *drv = NULL, *d;
288 
289     QLIST_FOREACH(d, &bdrv_drivers, list) {
290         if (d->bdrv_probe_device) {
291             score = d->bdrv_probe_device(filename);
292             if (score > score_max) {
293                 score_max = score;
294                 drv = d;
295             }
296         }
297     }
298 
299     return drv;
300 }
301 
302 BlockDriver *bdrv_find_protocol(const char *filename)
303 {
304     BlockDriver *drv1;
305     char protocol[128];
306     int len;
307     const char *p;
308 
309     /* TODO Drivers without bdrv_file_open must be specified explicitly */
310 
311     /*
312      * XXX(hch): we really should not let host device detection
313      * override an explicit protocol specification, but moving this
314      * later breaks access to device names with colons in them.
315      * Thanks to the brain-dead persistent naming schemes on udev-
316      * based Linux systems those actually are quite common.
317      */
318     drv1 = find_hdev_driver(filename);
319     if (drv1) {
320         return drv1;
321     }
322 
323     if (!path_has_protocol(filename)) {
324         return bdrv_find_format("file");
325     }
326     p = strchr(filename, ':');
327     assert(p != NULL);
328     len = p - filename;
329     if (len > sizeof(protocol) - 1)
330         len = sizeof(protocol) - 1;
331     memcpy(protocol, filename, len);
332     protocol[len] = '\0';
333     QLIST_FOREACH(drv1, &bdrv_drivers, list) {
334         if (drv1->protocol_name &&
335             !strcmp(drv1->protocol_name, protocol)) {
336             return drv1;
337         }
338     }
339     return NULL;
340 }
341 
342 static int find_image_format(const char *filename, BlockDriver **pdrv)
343 {
344     int ret, score, score_max;
345     BlockDriver *drv1, *drv;
346     uint8_t buf[2048];
347     BlockDriverState *bs;
348 
349     ret = bdrv_file_open(&bs, filename, 0);
350     if (ret < 0) {
351         *pdrv = NULL;
352         return ret;
353     }
354 
355     /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
356     if (bs->sg || !bdrv_is_inserted(bs)) {
357         bdrv_delete(bs);
358         drv = bdrv_find_format("raw");
359         if (!drv) {
360             ret = -ENOENT;
361         }
362         *pdrv = drv;
363         return ret;
364     }
365 
366     ret = bdrv_pread(bs, 0, buf, sizeof(buf));
367     bdrv_delete(bs);
368     if (ret < 0) {
369         *pdrv = NULL;
370         return ret;
371     }
372 
373     score_max = 0;
374     drv = NULL;
375     QLIST_FOREACH(drv1, &bdrv_drivers, list) {
376         if (drv1->bdrv_probe) {
377             score = drv1->bdrv_probe(buf, ret, filename);
378             if (score > score_max) {
379                 score_max = score;
380                 drv = drv1;
381             }
382         }
383     }
384     if (!drv) {
385         ret = -ENOENT;
386     }
387     *pdrv = drv;
388     return ret;
389 }
390 
391 /**
392  * Set the current 'total_sectors' value
393  */
394 static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
395 {
396     BlockDriver *drv = bs->drv;
397 
398     /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
399     if (bs->sg)
400         return 0;
401 
402     /* query actual device if possible, otherwise just trust the hint */
403     if (drv->bdrv_getlength) {
404         int64_t length = drv->bdrv_getlength(bs);
405         if (length < 0) {
406             return length;
407         }
408         hint = length >> BDRV_SECTOR_BITS;
409     }
410 
411     bs->total_sectors = hint;
412     return 0;
413 }
414 
415 /*
416  * Common part for opening disk images and files
417  */
418 static int bdrv_open_common(BlockDriverState *bs, const char *filename,
419     int flags, BlockDriver *drv)
420 {
421     int ret, open_flags;
422 
423     assert(drv != NULL);
424 
425     bs->file = NULL;
426     bs->total_sectors = 0;
427     bs->encrypted = 0;
428     bs->valid_key = 0;
429     bs->open_flags = flags;
430     /* buffer_alignment defaulted to 512, drivers can change this value */
431     bs->buffer_alignment = 512;
432 
433     pstrcpy(bs->filename, sizeof(bs->filename), filename);
434 
435     if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
436         return -ENOTSUP;
437     }
438 
439     bs->drv = drv;
440     bs->opaque = qemu_mallocz(drv->instance_size);
441 
442     /*
443      * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
444      * write cache to the guest.  We do need the fdatasync to flush
445      * out transactions for block allocations, and we maybe have a
446      * volatile write cache in our backing device to deal with.
447      */
448     if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
449         bs->enable_write_cache = 1;
450 
451     /*
452      * Clear flags that are internal to the block layer before opening the
453      * image.
454      */
455     open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
456 
457     /*
458      * Snapshots should be writeable.
459      */
460     if (bs->is_temporary) {
461         open_flags |= BDRV_O_RDWR;
462     }
463 
464     /* Open the image, either directly or using a protocol */
465     if (drv->bdrv_file_open) {
466         ret = drv->bdrv_file_open(bs, filename, open_flags);
467     } else {
468         ret = bdrv_file_open(&bs->file, filename, open_flags);
469         if (ret >= 0) {
470             ret = drv->bdrv_open(bs, open_flags);
471         }
472     }
473 
474     if (ret < 0) {
475         goto free_and_fail;
476     }
477 
478     bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
479 
480     ret = refresh_total_sectors(bs, bs->total_sectors);
481     if (ret < 0) {
482         goto free_and_fail;
483     }
484 
485 #ifndef _WIN32
486     if (bs->is_temporary) {
487         unlink(filename);
488     }
489 #endif
490     return 0;
491 
492 free_and_fail:
493     if (bs->file) {
494         bdrv_delete(bs->file);
495         bs->file = NULL;
496     }
497     qemu_free(bs->opaque);
498     bs->opaque = NULL;
499     bs->drv = NULL;
500     return ret;
501 }
502 
503 /*
504  * Opens a file using a protocol (file, host_device, nbd, ...)
505  */
506 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
507 {
508     BlockDriverState *bs;
509     BlockDriver *drv;
510     int ret;
511 
512     drv = bdrv_find_protocol(filename);
513     if (!drv) {
514         return -ENOENT;
515     }
516 
517     bs = bdrv_new("");
518     ret = bdrv_open_common(bs, filename, flags, drv);
519     if (ret < 0) {
520         bdrv_delete(bs);
521         return ret;
522     }
523     bs->growable = 1;
524     *pbs = bs;
525     return 0;
526 }
527 
528 /*
529  * Opens a disk image (raw, qcow2, vmdk, ...)
530  */
531 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
532               BlockDriver *drv)
533 {
534     int ret;
535 
536     if (flags & BDRV_O_SNAPSHOT) {
537         BlockDriverState *bs1;
538         int64_t total_size;
539         int is_protocol = 0;
540         BlockDriver *bdrv_qcow2;
541         QEMUOptionParameter *options;
542         char tmp_filename[PATH_MAX];
543         char backing_filename[PATH_MAX];
544 
545         /* if snapshot, we create a temporary backing file and open it
546            instead of opening 'filename' directly */
547 
548         /* if there is a backing file, use it */
549         bs1 = bdrv_new("");
550         ret = bdrv_open(bs1, filename, 0, drv);
551         if (ret < 0) {
552             bdrv_delete(bs1);
553             return ret;
554         }
555         total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
556 
557         if (bs1->drv && bs1->drv->protocol_name)
558             is_protocol = 1;
559 
560         bdrv_delete(bs1);
561 
562         get_tmp_filename(tmp_filename, sizeof(tmp_filename));
563 
564         /* Real path is meaningless for protocols */
565         if (is_protocol)
566             snprintf(backing_filename, sizeof(backing_filename),
567                      "%s", filename);
568         else if (!realpath(filename, backing_filename))
569             return -errno;
570 
571         bdrv_qcow2 = bdrv_find_format("qcow2");
572         options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
573 
574         set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size);
575         set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
576         if (drv) {
577             set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
578                 drv->format_name);
579         }
580 
581         ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
582         free_option_parameters(options);
583         if (ret < 0) {
584             return ret;
585         }
586 
587         filename = tmp_filename;
588         drv = bdrv_qcow2;
589         bs->is_temporary = 1;
590     }
591 
592     /* Find the right image format driver */
593     if (!drv) {
594         ret = find_image_format(filename, &drv);
595     }
596 
597     if (!drv) {
598         goto unlink_and_fail;
599     }
600 
601     /* Open the image */
602     ret = bdrv_open_common(bs, filename, flags, drv);
603     if (ret < 0) {
604         goto unlink_and_fail;
605     }
606 
607     /* If there is a backing file, use it */
608     if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
609         char backing_filename[PATH_MAX];
610         int back_flags;
611         BlockDriver *back_drv = NULL;
612 
613         bs->backing_hd = bdrv_new("");
614 
615         if (path_has_protocol(bs->backing_file)) {
616             pstrcpy(backing_filename, sizeof(backing_filename),
617                     bs->backing_file);
618         } else {
619             path_combine(backing_filename, sizeof(backing_filename),
620                          filename, bs->backing_file);
621         }
622 
623         if (bs->backing_format[0] != '\0') {
624             back_drv = bdrv_find_format(bs->backing_format);
625         }
626 
627         /* backing files always opened read-only */
628         back_flags =
629             flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
630 
631         ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
632         if (ret < 0) {
633             bdrv_close(bs);
634             return ret;
635         }
636         if (bs->is_temporary) {
637             bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
638         } else {
639             /* base image inherits from "parent" */
640             bs->backing_hd->keep_read_only = bs->keep_read_only;
641         }
642     }
643 
644     if (!bdrv_key_required(bs)) {
645         /* call the change callback */
646         bs->media_changed = 1;
647         if (bs->change_cb)
648             bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
649     }
650 
651     return 0;
652 
653 unlink_and_fail:
654     if (bs->is_temporary) {
655         unlink(filename);
656     }
657     return ret;
658 }
659 
660 void bdrv_close(BlockDriverState *bs)
661 {
662     if (bs->drv) {
663         if (bs == bs_snapshots) {
664             bs_snapshots = NULL;
665         }
666         if (bs->backing_hd) {
667             bdrv_delete(bs->backing_hd);
668             bs->backing_hd = NULL;
669         }
670         bs->drv->bdrv_close(bs);
671         qemu_free(bs->opaque);
672 #ifdef _WIN32
673         if (bs->is_temporary) {
674             unlink(bs->filename);
675         }
676 #endif
677         bs->opaque = NULL;
678         bs->drv = NULL;
679 
680         if (bs->file != NULL) {
681             bdrv_close(bs->file);
682         }
683 
684         /* call the change callback */
685         bs->media_changed = 1;
686         if (bs->change_cb)
687             bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
688     }
689 }
690 
691 void bdrv_close_all(void)
692 {
693     BlockDriverState *bs;
694 
695     QTAILQ_FOREACH(bs, &bdrv_states, list) {
696         bdrv_close(bs);
697     }
698 }
699 
700 void bdrv_delete(BlockDriverState *bs)
701 {
702     assert(!bs->peer);
703 
704     /* remove from list, if necessary */
705     if (bs->device_name[0] != '\0') {
706         QTAILQ_REMOVE(&bdrv_states, bs, list);
707     }
708 
709     bdrv_close(bs);
710     if (bs->file != NULL) {
711         bdrv_delete(bs->file);
712     }
713 
714     assert(bs != bs_snapshots);
715     qemu_free(bs);
716 }
717 
718 int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
719 {
720     if (bs->peer) {
721         return -EBUSY;
722     }
723     bs->peer = qdev;
724     return 0;
725 }
726 
727 void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
728 {
729     assert(bs->peer == qdev);
730     bs->peer = NULL;
731 }
732 
733 DeviceState *bdrv_get_attached(BlockDriverState *bs)
734 {
735     return bs->peer;
736 }
737 
738 /*
739  * Run consistency checks on an image
740  *
741  * Returns 0 if the check could be completed (it doesn't mean that the image is
742  * free of errors) or -errno when an internal error occured. The results of the
743  * check are stored in res.
744  */
745 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res)
746 {
747     if (bs->drv->bdrv_check == NULL) {
748         return -ENOTSUP;
749     }
750 
751     memset(res, 0, sizeof(*res));
752     return bs->drv->bdrv_check(bs, res);
753 }
754 
755 #define COMMIT_BUF_SECTORS 2048
756 
757 /* commit COW file into the raw image */
758 int bdrv_commit(BlockDriverState *bs)
759 {
760     BlockDriver *drv = bs->drv;
761     BlockDriver *backing_drv;
762     int64_t sector, total_sectors;
763     int n, ro, open_flags;
764     int ret = 0, rw_ret = 0;
765     uint8_t *buf;
766     char filename[1024];
767     BlockDriverState *bs_rw, *bs_ro;
768 
769     if (!drv)
770         return -ENOMEDIUM;
771 
772     if (!bs->backing_hd) {
773         return -ENOTSUP;
774     }
775 
776     if (bs->backing_hd->keep_read_only) {
777         return -EACCES;
778     }
779 
780     backing_drv = bs->backing_hd->drv;
781     ro = bs->backing_hd->read_only;
782     strncpy(filename, bs->backing_hd->filename, sizeof(filename));
783     open_flags =  bs->backing_hd->open_flags;
784 
785     if (ro) {
786         /* re-open as RW */
787         bdrv_delete(bs->backing_hd);
788         bs->backing_hd = NULL;
789         bs_rw = bdrv_new("");
790         rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR,
791             backing_drv);
792         if (rw_ret < 0) {
793             bdrv_delete(bs_rw);
794             /* try to re-open read-only */
795             bs_ro = bdrv_new("");
796             ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
797                 backing_drv);
798             if (ret < 0) {
799                 bdrv_delete(bs_ro);
800                 /* drive not functional anymore */
801                 bs->drv = NULL;
802                 return ret;
803             }
804             bs->backing_hd = bs_ro;
805             return rw_ret;
806         }
807         bs->backing_hd = bs_rw;
808     }
809 
810     total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
811     buf = qemu_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE);
812 
813     for (sector = 0; sector < total_sectors; sector += n) {
814         if (drv->bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) {
815 
816             if (bdrv_read(bs, sector, buf, n) != 0) {
817                 ret = -EIO;
818                 goto ro_cleanup;
819             }
820 
821             if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) {
822                 ret = -EIO;
823                 goto ro_cleanup;
824             }
825         }
826     }
827 
828     if (drv->bdrv_make_empty) {
829         ret = drv->bdrv_make_empty(bs);
830         bdrv_flush(bs);
831     }
832 
833     /*
834      * Make sure all data we wrote to the backing device is actually
835      * stable on disk.
836      */
837     if (bs->backing_hd)
838         bdrv_flush(bs->backing_hd);
839 
840 ro_cleanup:
841     qemu_free(buf);
842 
843     if (ro) {
844         /* re-open as RO */
845         bdrv_delete(bs->backing_hd);
846         bs->backing_hd = NULL;
847         bs_ro = bdrv_new("");
848         ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
849             backing_drv);
850         if (ret < 0) {
851             bdrv_delete(bs_ro);
852             /* drive not functional anymore */
853             bs->drv = NULL;
854             return ret;
855         }
856         bs->backing_hd = bs_ro;
857         bs->backing_hd->keep_read_only = 0;
858     }
859 
860     return ret;
861 }
862 
863 void bdrv_commit_all(void)
864 {
865     BlockDriverState *bs;
866 
867     QTAILQ_FOREACH(bs, &bdrv_states, list) {
868         bdrv_commit(bs);
869     }
870 }
871 
872 /*
873  * Return values:
874  * 0        - success
875  * -EINVAL  - backing format specified, but no file
876  * -ENOSPC  - can't update the backing file because no space is left in the
877  *            image file header
878  * -ENOTSUP - format driver doesn't support changing the backing file
879  */
880 int bdrv_change_backing_file(BlockDriverState *bs,
881     const char *backing_file, const char *backing_fmt)
882 {
883     BlockDriver *drv = bs->drv;
884 
885     if (drv->bdrv_change_backing_file != NULL) {
886         return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
887     } else {
888         return -ENOTSUP;
889     }
890 }
891 
892 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
893                                    size_t size)
894 {
895     int64_t len;
896 
897     if (!bdrv_is_inserted(bs))
898         return -ENOMEDIUM;
899 
900     if (bs->growable)
901         return 0;
902 
903     len = bdrv_getlength(bs);
904 
905     if (offset < 0)
906         return -EIO;
907 
908     if ((offset > len) || (len - offset < size))
909         return -EIO;
910 
911     return 0;
912 }
913 
914 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
915                               int nb_sectors)
916 {
917     return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
918                                    nb_sectors * BDRV_SECTOR_SIZE);
919 }
920 
921 /* return < 0 if error. See bdrv_write() for the return codes */
922 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
923               uint8_t *buf, int nb_sectors)
924 {
925     BlockDriver *drv = bs->drv;
926 
927     if (!drv)
928         return -ENOMEDIUM;
929     if (bdrv_check_request(bs, sector_num, nb_sectors))
930         return -EIO;
931 
932     return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
933 }
934 
935 static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
936                              int nb_sectors, int dirty)
937 {
938     int64_t start, end;
939     unsigned long val, idx, bit;
940 
941     start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
942     end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
943 
944     for (; start <= end; start++) {
945         idx = start / (sizeof(unsigned long) * 8);
946         bit = start % (sizeof(unsigned long) * 8);
947         val = bs->dirty_bitmap[idx];
948         if (dirty) {
949             if (!(val & (1UL << bit))) {
950                 bs->dirty_count++;
951                 val |= 1UL << bit;
952             }
953         } else {
954             if (val & (1UL << bit)) {
955                 bs->dirty_count--;
956                 val &= ~(1UL << bit);
957             }
958         }
959         bs->dirty_bitmap[idx] = val;
960     }
961 }
962 
963 /* Return < 0 if error. Important errors are:
964   -EIO         generic I/O error (may happen for all errors)
965   -ENOMEDIUM   No media inserted.
966   -EINVAL      Invalid sector number or nb_sectors
967   -EACCES      Trying to write a read-only device
968 */
969 int bdrv_write(BlockDriverState *bs, int64_t sector_num,
970                const uint8_t *buf, int nb_sectors)
971 {
972     BlockDriver *drv = bs->drv;
973     if (!bs->drv)
974         return -ENOMEDIUM;
975     if (bs->read_only)
976         return -EACCES;
977     if (bdrv_check_request(bs, sector_num, nb_sectors))
978         return -EIO;
979 
980     if (bs->dirty_bitmap) {
981         set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
982     }
983 
984     if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
985         bs->wr_highest_sector = sector_num + nb_sectors - 1;
986     }
987 
988     return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
989 }
990 
991 int bdrv_pread(BlockDriverState *bs, int64_t offset,
992                void *buf, int count1)
993 {
994     uint8_t tmp_buf[BDRV_SECTOR_SIZE];
995     int len, nb_sectors, count;
996     int64_t sector_num;
997     int ret;
998 
999     count = count1;
1000     /* first read to align to sector start */
1001     len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1002     if (len > count)
1003         len = count;
1004     sector_num = offset >> BDRV_SECTOR_BITS;
1005     if (len > 0) {
1006         if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1007             return ret;
1008         memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
1009         count -= len;
1010         if (count == 0)
1011             return count1;
1012         sector_num++;
1013         buf += len;
1014     }
1015 
1016     /* read the sectors "in place" */
1017     nb_sectors = count >> BDRV_SECTOR_BITS;
1018     if (nb_sectors > 0) {
1019         if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
1020             return ret;
1021         sector_num += nb_sectors;
1022         len = nb_sectors << BDRV_SECTOR_BITS;
1023         buf += len;
1024         count -= len;
1025     }
1026 
1027     /* add data from the last sector */
1028     if (count > 0) {
1029         if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1030             return ret;
1031         memcpy(buf, tmp_buf, count);
1032     }
1033     return count1;
1034 }
1035 
1036 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
1037                 const void *buf, int count1)
1038 {
1039     uint8_t tmp_buf[BDRV_SECTOR_SIZE];
1040     int len, nb_sectors, count;
1041     int64_t sector_num;
1042     int ret;
1043 
1044     count = count1;
1045     /* first write to align to sector start */
1046     len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1047     if (len > count)
1048         len = count;
1049     sector_num = offset >> BDRV_SECTOR_BITS;
1050     if (len > 0) {
1051         if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1052             return ret;
1053         memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
1054         if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1055             return ret;
1056         count -= len;
1057         if (count == 0)
1058             return count1;
1059         sector_num++;
1060         buf += len;
1061     }
1062 
1063     /* write the sectors "in place" */
1064     nb_sectors = count >> BDRV_SECTOR_BITS;
1065     if (nb_sectors > 0) {
1066         if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
1067             return ret;
1068         sector_num += nb_sectors;
1069         len = nb_sectors << BDRV_SECTOR_BITS;
1070         buf += len;
1071         count -= len;
1072     }
1073 
1074     /* add data from the last sector */
1075     if (count > 0) {
1076         if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1077             return ret;
1078         memcpy(tmp_buf, buf, count);
1079         if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1080             return ret;
1081     }
1082     return count1;
1083 }
1084 
1085 /*
1086  * Writes to the file and ensures that no writes are reordered across this
1087  * request (acts as a barrier)
1088  *
1089  * Returns 0 on success, -errno in error cases.
1090  */
1091 int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset,
1092     const void *buf, int count)
1093 {
1094     int ret;
1095 
1096     ret = bdrv_pwrite(bs, offset, buf, count);
1097     if (ret < 0) {
1098         return ret;
1099     }
1100 
1101     /* No flush needed for cache=writethrough, it uses O_DSYNC */
1102     if ((bs->open_flags & BDRV_O_CACHE_MASK) != 0) {
1103         bdrv_flush(bs);
1104     }
1105 
1106     return 0;
1107 }
1108 
1109 /*
1110  * Writes to the file and ensures that no writes are reordered across this
1111  * request (acts as a barrier)
1112  *
1113  * Returns 0 on success, -errno in error cases.
1114  */
1115 int bdrv_write_sync(BlockDriverState *bs, int64_t sector_num,
1116     const uint8_t *buf, int nb_sectors)
1117 {
1118     return bdrv_pwrite_sync(bs, BDRV_SECTOR_SIZE * sector_num,
1119         buf, BDRV_SECTOR_SIZE * nb_sectors);
1120 }
1121 
1122 /**
1123  * Truncate file to 'offset' bytes (needed only for file protocols)
1124  */
1125 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
1126 {
1127     BlockDriver *drv = bs->drv;
1128     int ret;
1129     if (!drv)
1130         return -ENOMEDIUM;
1131     if (!drv->bdrv_truncate)
1132         return -ENOTSUP;
1133     if (bs->read_only)
1134         return -EACCES;
1135     if (bdrv_in_use(bs))
1136         return -EBUSY;
1137     ret = drv->bdrv_truncate(bs, offset);
1138     if (ret == 0) {
1139         ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
1140         if (bs->change_cb) {
1141             bs->change_cb(bs->change_opaque, CHANGE_SIZE);
1142         }
1143     }
1144     return ret;
1145 }
1146 
1147 /**
1148  * Length of a file in bytes. Return < 0 if error or unknown.
1149  */
1150 int64_t bdrv_getlength(BlockDriverState *bs)
1151 {
1152     BlockDriver *drv = bs->drv;
1153     if (!drv)
1154         return -ENOMEDIUM;
1155 
1156     /* Fixed size devices use the total_sectors value for speed instead of
1157        issuing a length query (like lseek) on each call.  Also, legacy block
1158        drivers don't provide a bdrv_getlength function and must use
1159        total_sectors. */
1160     if (!bs->growable || !drv->bdrv_getlength) {
1161         return bs->total_sectors * BDRV_SECTOR_SIZE;
1162     }
1163     return drv->bdrv_getlength(bs);
1164 }
1165 
1166 /* return 0 as number of sectors if no device present or error */
1167 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
1168 {
1169     int64_t length;
1170     length = bdrv_getlength(bs);
1171     if (length < 0)
1172         length = 0;
1173     else
1174         length = length >> BDRV_SECTOR_BITS;
1175     *nb_sectors_ptr = length;
1176 }
1177 
1178 struct partition {
1179         uint8_t boot_ind;           /* 0x80 - active */
1180         uint8_t head;               /* starting head */
1181         uint8_t sector;             /* starting sector */
1182         uint8_t cyl;                /* starting cylinder */
1183         uint8_t sys_ind;            /* What partition type */
1184         uint8_t end_head;           /* end head */
1185         uint8_t end_sector;         /* end sector */
1186         uint8_t end_cyl;            /* end cylinder */
1187         uint32_t start_sect;        /* starting sector counting from 0 */
1188         uint32_t nr_sects;          /* nr of sectors in partition */
1189 } __attribute__((packed));
1190 
1191 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1192 static int guess_disk_lchs(BlockDriverState *bs,
1193                            int *pcylinders, int *pheads, int *psectors)
1194 {
1195     uint8_t buf[BDRV_SECTOR_SIZE];
1196     int ret, i, heads, sectors, cylinders;
1197     struct partition *p;
1198     uint32_t nr_sects;
1199     uint64_t nb_sectors;
1200 
1201     bdrv_get_geometry(bs, &nb_sectors);
1202 
1203     ret = bdrv_read(bs, 0, buf, 1);
1204     if (ret < 0)
1205         return -1;
1206     /* test msdos magic */
1207     if (buf[510] != 0x55 || buf[511] != 0xaa)
1208         return -1;
1209     for(i = 0; i < 4; i++) {
1210         p = ((struct partition *)(buf + 0x1be)) + i;
1211         nr_sects = le32_to_cpu(p->nr_sects);
1212         if (nr_sects && p->end_head) {
1213             /* We make the assumption that the partition terminates on
1214                a cylinder boundary */
1215             heads = p->end_head + 1;
1216             sectors = p->end_sector & 63;
1217             if (sectors == 0)
1218                 continue;
1219             cylinders = nb_sectors / (heads * sectors);
1220             if (cylinders < 1 || cylinders > 16383)
1221                 continue;
1222             *pheads = heads;
1223             *psectors = sectors;
1224             *pcylinders = cylinders;
1225 #if 0
1226             printf("guessed geometry: LCHS=%d %d %d\n",
1227                    cylinders, heads, sectors);
1228 #endif
1229             return 0;
1230         }
1231     }
1232     return -1;
1233 }
1234 
1235 void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1236 {
1237     int translation, lba_detected = 0;
1238     int cylinders, heads, secs;
1239     uint64_t nb_sectors;
1240 
1241     /* if a geometry hint is available, use it */
1242     bdrv_get_geometry(bs, &nb_sectors);
1243     bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1244     translation = bdrv_get_translation_hint(bs);
1245     if (cylinders != 0) {
1246         *pcyls = cylinders;
1247         *pheads = heads;
1248         *psecs = secs;
1249     } else {
1250         if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1251             if (heads > 16) {
1252                 /* if heads > 16, it means that a BIOS LBA
1253                    translation was active, so the default
1254                    hardware geometry is OK */
1255                 lba_detected = 1;
1256                 goto default_geometry;
1257             } else {
1258                 *pcyls = cylinders;
1259                 *pheads = heads;
1260                 *psecs = secs;
1261                 /* disable any translation to be in sync with
1262                    the logical geometry */
1263                 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1264                     bdrv_set_translation_hint(bs,
1265                                               BIOS_ATA_TRANSLATION_NONE);
1266                 }
1267             }
1268         } else {
1269         default_geometry:
1270             /* if no geometry, use a standard physical disk geometry */
1271             cylinders = nb_sectors / (16 * 63);
1272 
1273             if (cylinders > 16383)
1274                 cylinders = 16383;
1275             else if (cylinders < 2)
1276                 cylinders = 2;
1277             *pcyls = cylinders;
1278             *pheads = 16;
1279             *psecs = 63;
1280             if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1281                 if ((*pcyls * *pheads) <= 131072) {
1282                     bdrv_set_translation_hint(bs,
1283                                               BIOS_ATA_TRANSLATION_LARGE);
1284                 } else {
1285                     bdrv_set_translation_hint(bs,
1286                                               BIOS_ATA_TRANSLATION_LBA);
1287                 }
1288             }
1289         }
1290         bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1291     }
1292 }
1293 
1294 void bdrv_set_geometry_hint(BlockDriverState *bs,
1295                             int cyls, int heads, int secs)
1296 {
1297     bs->cyls = cyls;
1298     bs->heads = heads;
1299     bs->secs = secs;
1300 }
1301 
1302 void bdrv_set_type_hint(BlockDriverState *bs, int type)
1303 {
1304     bs->type = type;
1305     bs->removable = ((type == BDRV_TYPE_CDROM ||
1306                       type == BDRV_TYPE_FLOPPY));
1307 }
1308 
1309 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1310 {
1311     bs->translation = translation;
1312 }
1313 
1314 void bdrv_get_geometry_hint(BlockDriverState *bs,
1315                             int *pcyls, int *pheads, int *psecs)
1316 {
1317     *pcyls = bs->cyls;
1318     *pheads = bs->heads;
1319     *psecs = bs->secs;
1320 }
1321 
1322 /* Recognize floppy formats */
1323 typedef struct FDFormat {
1324     FDriveType drive;
1325     uint8_t last_sect;
1326     uint8_t max_track;
1327     uint8_t max_head;
1328 } FDFormat;
1329 
1330 static const FDFormat fd_formats[] = {
1331     /* First entry is default format */
1332     /* 1.44 MB 3"1/2 floppy disks */
1333     { FDRIVE_DRV_144, 18, 80, 1, },
1334     { FDRIVE_DRV_144, 20, 80, 1, },
1335     { FDRIVE_DRV_144, 21, 80, 1, },
1336     { FDRIVE_DRV_144, 21, 82, 1, },
1337     { FDRIVE_DRV_144, 21, 83, 1, },
1338     { FDRIVE_DRV_144, 22, 80, 1, },
1339     { FDRIVE_DRV_144, 23, 80, 1, },
1340     { FDRIVE_DRV_144, 24, 80, 1, },
1341     /* 2.88 MB 3"1/2 floppy disks */
1342     { FDRIVE_DRV_288, 36, 80, 1, },
1343     { FDRIVE_DRV_288, 39, 80, 1, },
1344     { FDRIVE_DRV_288, 40, 80, 1, },
1345     { FDRIVE_DRV_288, 44, 80, 1, },
1346     { FDRIVE_DRV_288, 48, 80, 1, },
1347     /* 720 kB 3"1/2 floppy disks */
1348     { FDRIVE_DRV_144,  9, 80, 1, },
1349     { FDRIVE_DRV_144, 10, 80, 1, },
1350     { FDRIVE_DRV_144, 10, 82, 1, },
1351     { FDRIVE_DRV_144, 10, 83, 1, },
1352     { FDRIVE_DRV_144, 13, 80, 1, },
1353     { FDRIVE_DRV_144, 14, 80, 1, },
1354     /* 1.2 MB 5"1/4 floppy disks */
1355     { FDRIVE_DRV_120, 15, 80, 1, },
1356     { FDRIVE_DRV_120, 18, 80, 1, },
1357     { FDRIVE_DRV_120, 18, 82, 1, },
1358     { FDRIVE_DRV_120, 18, 83, 1, },
1359     { FDRIVE_DRV_120, 20, 80, 1, },
1360     /* 720 kB 5"1/4 floppy disks */
1361     { FDRIVE_DRV_120,  9, 80, 1, },
1362     { FDRIVE_DRV_120, 11, 80, 1, },
1363     /* 360 kB 5"1/4 floppy disks */
1364     { FDRIVE_DRV_120,  9, 40, 1, },
1365     { FDRIVE_DRV_120,  9, 40, 0, },
1366     { FDRIVE_DRV_120, 10, 41, 1, },
1367     { FDRIVE_DRV_120, 10, 42, 1, },
1368     /* 320 kB 5"1/4 floppy disks */
1369     { FDRIVE_DRV_120,  8, 40, 1, },
1370     { FDRIVE_DRV_120,  8, 40, 0, },
1371     /* 360 kB must match 5"1/4 better than 3"1/2... */
1372     { FDRIVE_DRV_144,  9, 80, 0, },
1373     /* end */
1374     { FDRIVE_DRV_NONE, -1, -1, 0, },
1375 };
1376 
1377 void bdrv_get_floppy_geometry_hint(BlockDriverState *bs, int *nb_heads,
1378                                    int *max_track, int *last_sect,
1379                                    FDriveType drive_in, FDriveType *drive)
1380 {
1381     const FDFormat *parse;
1382     uint64_t nb_sectors, size;
1383     int i, first_match, match;
1384 
1385     bdrv_get_geometry_hint(bs, nb_heads, max_track, last_sect);
1386     if (*nb_heads != 0 && *max_track != 0 && *last_sect != 0) {
1387         /* User defined disk */
1388     } else {
1389         bdrv_get_geometry(bs, &nb_sectors);
1390         match = -1;
1391         first_match = -1;
1392         for (i = 0; ; i++) {
1393             parse = &fd_formats[i];
1394             if (parse->drive == FDRIVE_DRV_NONE) {
1395                 break;
1396             }
1397             if (drive_in == parse->drive ||
1398                 drive_in == FDRIVE_DRV_NONE) {
1399                 size = (parse->max_head + 1) * parse->max_track *
1400                     parse->last_sect;
1401                 if (nb_sectors == size) {
1402                     match = i;
1403                     break;
1404                 }
1405                 if (first_match == -1) {
1406                     first_match = i;
1407                 }
1408             }
1409         }
1410         if (match == -1) {
1411             if (first_match == -1) {
1412                 match = 1;
1413             } else {
1414                 match = first_match;
1415             }
1416             parse = &fd_formats[match];
1417         }
1418         *nb_heads = parse->max_head + 1;
1419         *max_track = parse->max_track;
1420         *last_sect = parse->last_sect;
1421         *drive = parse->drive;
1422     }
1423 }
1424 
1425 int bdrv_get_type_hint(BlockDriverState *bs)
1426 {
1427     return bs->type;
1428 }
1429 
1430 int bdrv_get_translation_hint(BlockDriverState *bs)
1431 {
1432     return bs->translation;
1433 }
1434 
1435 void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
1436                        BlockErrorAction on_write_error)
1437 {
1438     bs->on_read_error = on_read_error;
1439     bs->on_write_error = on_write_error;
1440 }
1441 
1442 BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
1443 {
1444     return is_read ? bs->on_read_error : bs->on_write_error;
1445 }
1446 
1447 void bdrv_set_removable(BlockDriverState *bs, int removable)
1448 {
1449     bs->removable = removable;
1450     if (removable && bs == bs_snapshots) {
1451         bs_snapshots = NULL;
1452     }
1453 }
1454 
1455 int bdrv_is_removable(BlockDriverState *bs)
1456 {
1457     return bs->removable;
1458 }
1459 
1460 int bdrv_is_read_only(BlockDriverState *bs)
1461 {
1462     return bs->read_only;
1463 }
1464 
1465 int bdrv_is_sg(BlockDriverState *bs)
1466 {
1467     return bs->sg;
1468 }
1469 
1470 int bdrv_enable_write_cache(BlockDriverState *bs)
1471 {
1472     return bs->enable_write_cache;
1473 }
1474 
1475 /* XXX: no longer used */
1476 void bdrv_set_change_cb(BlockDriverState *bs,
1477                         void (*change_cb)(void *opaque, int reason),
1478                         void *opaque)
1479 {
1480     bs->change_cb = change_cb;
1481     bs->change_opaque = opaque;
1482 }
1483 
1484 int bdrv_is_encrypted(BlockDriverState *bs)
1485 {
1486     if (bs->backing_hd && bs->backing_hd->encrypted)
1487         return 1;
1488     return bs->encrypted;
1489 }
1490 
1491 int bdrv_key_required(BlockDriverState *bs)
1492 {
1493     BlockDriverState *backing_hd = bs->backing_hd;
1494 
1495     if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1496         return 1;
1497     return (bs->encrypted && !bs->valid_key);
1498 }
1499 
1500 int bdrv_set_key(BlockDriverState *bs, const char *key)
1501 {
1502     int ret;
1503     if (bs->backing_hd && bs->backing_hd->encrypted) {
1504         ret = bdrv_set_key(bs->backing_hd, key);
1505         if (ret < 0)
1506             return ret;
1507         if (!bs->encrypted)
1508             return 0;
1509     }
1510     if (!bs->encrypted) {
1511         return -EINVAL;
1512     } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1513         return -ENOMEDIUM;
1514     }
1515     ret = bs->drv->bdrv_set_key(bs, key);
1516     if (ret < 0) {
1517         bs->valid_key = 0;
1518     } else if (!bs->valid_key) {
1519         bs->valid_key = 1;
1520         /* call the change callback now, we skipped it on open */
1521         bs->media_changed = 1;
1522         if (bs->change_cb)
1523             bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
1524     }
1525     return ret;
1526 }
1527 
1528 void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1529 {
1530     if (!bs->drv) {
1531         buf[0] = '\0';
1532     } else {
1533         pstrcpy(buf, buf_size, bs->drv->format_name);
1534     }
1535 }
1536 
1537 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1538                          void *opaque)
1539 {
1540     BlockDriver *drv;
1541 
1542     QLIST_FOREACH(drv, &bdrv_drivers, list) {
1543         it(opaque, drv->format_name);
1544     }
1545 }
1546 
1547 BlockDriverState *bdrv_find(const char *name)
1548 {
1549     BlockDriverState *bs;
1550 
1551     QTAILQ_FOREACH(bs, &bdrv_states, list) {
1552         if (!strcmp(name, bs->device_name)) {
1553             return bs;
1554         }
1555     }
1556     return NULL;
1557 }
1558 
1559 BlockDriverState *bdrv_next(BlockDriverState *bs)
1560 {
1561     if (!bs) {
1562         return QTAILQ_FIRST(&bdrv_states);
1563     }
1564     return QTAILQ_NEXT(bs, list);
1565 }
1566 
1567 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1568 {
1569     BlockDriverState *bs;
1570 
1571     QTAILQ_FOREACH(bs, &bdrv_states, list) {
1572         it(opaque, bs);
1573     }
1574 }
1575 
1576 const char *bdrv_get_device_name(BlockDriverState *bs)
1577 {
1578     return bs->device_name;
1579 }
1580 
1581 int bdrv_flush(BlockDriverState *bs)
1582 {
1583     if (bs->open_flags & BDRV_O_NO_FLUSH) {
1584         return 0;
1585     }
1586 
1587     if (bs->drv && bs->drv->bdrv_flush) {
1588         return bs->drv->bdrv_flush(bs);
1589     }
1590 
1591     /*
1592      * Some block drivers always operate in either writethrough or unsafe mode
1593      * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1594      * the server works (because the behaviour is hardcoded or depends on
1595      * server-side configuration), so we can't ensure that everything is safe
1596      * on disk. Returning an error doesn't work because that would break guests
1597      * even if the server operates in writethrough mode.
1598      *
1599      * Let's hope the user knows what he's doing.
1600      */
1601     return 0;
1602 }
1603 
1604 void bdrv_flush_all(void)
1605 {
1606     BlockDriverState *bs;
1607 
1608     QTAILQ_FOREACH(bs, &bdrv_states, list) {
1609         if (bs->drv && !bdrv_is_read_only(bs) &&
1610             (!bdrv_is_removable(bs) || bdrv_is_inserted(bs))) {
1611             bdrv_flush(bs);
1612         }
1613     }
1614 }
1615 
1616 int bdrv_has_zero_init(BlockDriverState *bs)
1617 {
1618     assert(bs->drv);
1619 
1620     if (bs->drv->bdrv_has_zero_init) {
1621         return bs->drv->bdrv_has_zero_init(bs);
1622     }
1623 
1624     return 1;
1625 }
1626 
1627 int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
1628 {
1629     if (!bs->drv) {
1630         return -ENOMEDIUM;
1631     }
1632     if (!bs->drv->bdrv_discard) {
1633         return 0;
1634     }
1635     return bs->drv->bdrv_discard(bs, sector_num, nb_sectors);
1636 }
1637 
1638 /*
1639  * Returns true iff the specified sector is present in the disk image. Drivers
1640  * not implementing the functionality are assumed to not support backing files,
1641  * hence all their sectors are reported as allocated.
1642  *
1643  * 'pnum' is set to the number of sectors (including and immediately following
1644  * the specified sector) that are known to be in the same
1645  * allocated/unallocated state.
1646  *
1647  * 'nb_sectors' is the max value 'pnum' should be set to.
1648  */
1649 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1650 	int *pnum)
1651 {
1652     int64_t n;
1653     if (!bs->drv->bdrv_is_allocated) {
1654         if (sector_num >= bs->total_sectors) {
1655             *pnum = 0;
1656             return 0;
1657         }
1658         n = bs->total_sectors - sector_num;
1659         *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1660         return 1;
1661     }
1662     return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1663 }
1664 
1665 void bdrv_mon_event(const BlockDriverState *bdrv,
1666                     BlockMonEventAction action, int is_read)
1667 {
1668     QObject *data;
1669     const char *action_str;
1670 
1671     switch (action) {
1672     case BDRV_ACTION_REPORT:
1673         action_str = "report";
1674         break;
1675     case BDRV_ACTION_IGNORE:
1676         action_str = "ignore";
1677         break;
1678     case BDRV_ACTION_STOP:
1679         action_str = "stop";
1680         break;
1681     default:
1682         abort();
1683     }
1684 
1685     data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1686                               bdrv->device_name,
1687                               action_str,
1688                               is_read ? "read" : "write");
1689     monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1690 
1691     qobject_decref(data);
1692 }
1693 
1694 static void bdrv_print_dict(QObject *obj, void *opaque)
1695 {
1696     QDict *bs_dict;
1697     Monitor *mon = opaque;
1698 
1699     bs_dict = qobject_to_qdict(obj);
1700 
1701     monitor_printf(mon, "%s: type=%s removable=%d",
1702                         qdict_get_str(bs_dict, "device"),
1703                         qdict_get_str(bs_dict, "type"),
1704                         qdict_get_bool(bs_dict, "removable"));
1705 
1706     if (qdict_get_bool(bs_dict, "removable")) {
1707         monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1708     }
1709 
1710     if (qdict_haskey(bs_dict, "inserted")) {
1711         QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1712 
1713         monitor_printf(mon, " file=");
1714         monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1715         if (qdict_haskey(qdict, "backing_file")) {
1716             monitor_printf(mon, " backing_file=");
1717             monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1718         }
1719         monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1720                             qdict_get_bool(qdict, "ro"),
1721                             qdict_get_str(qdict, "drv"),
1722                             qdict_get_bool(qdict, "encrypted"));
1723     } else {
1724         monitor_printf(mon, " [not inserted]");
1725     }
1726 
1727     monitor_printf(mon, "\n");
1728 }
1729 
1730 void bdrv_info_print(Monitor *mon, const QObject *data)
1731 {
1732     qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1733 }
1734 
1735 void bdrv_info(Monitor *mon, QObject **ret_data)
1736 {
1737     QList *bs_list;
1738     BlockDriverState *bs;
1739 
1740     bs_list = qlist_new();
1741 
1742     QTAILQ_FOREACH(bs, &bdrv_states, list) {
1743         QObject *bs_obj;
1744         const char *type = "unknown";
1745 
1746         switch(bs->type) {
1747         case BDRV_TYPE_HD:
1748             type = "hd";
1749             break;
1750         case BDRV_TYPE_CDROM:
1751             type = "cdrom";
1752             break;
1753         case BDRV_TYPE_FLOPPY:
1754             type = "floppy";
1755             break;
1756         }
1757 
1758         bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1759                                     "'removable': %i, 'locked': %i }",
1760                                     bs->device_name, type, bs->removable,
1761                                     bs->locked);
1762 
1763         if (bs->drv) {
1764             QObject *obj;
1765             QDict *bs_dict = qobject_to_qdict(bs_obj);
1766 
1767             obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1768                                      "'encrypted': %i }",
1769                                      bs->filename, bs->read_only,
1770                                      bs->drv->format_name,
1771                                      bdrv_is_encrypted(bs));
1772             if (bs->backing_file[0] != '\0') {
1773                 QDict *qdict = qobject_to_qdict(obj);
1774                 qdict_put(qdict, "backing_file",
1775                           qstring_from_str(bs->backing_file));
1776             }
1777 
1778             qdict_put_obj(bs_dict, "inserted", obj);
1779         }
1780         qlist_append_obj(bs_list, bs_obj);
1781     }
1782 
1783     *ret_data = QOBJECT(bs_list);
1784 }
1785 
1786 static void bdrv_stats_iter(QObject *data, void *opaque)
1787 {
1788     QDict *qdict;
1789     Monitor *mon = opaque;
1790 
1791     qdict = qobject_to_qdict(data);
1792     monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1793 
1794     qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1795     monitor_printf(mon, " rd_bytes=%" PRId64
1796                         " wr_bytes=%" PRId64
1797                         " rd_operations=%" PRId64
1798                         " wr_operations=%" PRId64
1799                         "\n",
1800                         qdict_get_int(qdict, "rd_bytes"),
1801                         qdict_get_int(qdict, "wr_bytes"),
1802                         qdict_get_int(qdict, "rd_operations"),
1803                         qdict_get_int(qdict, "wr_operations"));
1804 }
1805 
1806 void bdrv_stats_print(Monitor *mon, const QObject *data)
1807 {
1808     qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1809 }
1810 
1811 static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1812 {
1813     QObject *res;
1814     QDict *dict;
1815 
1816     res = qobject_from_jsonf("{ 'stats': {"
1817                              "'rd_bytes': %" PRId64 ","
1818                              "'wr_bytes': %" PRId64 ","
1819                              "'rd_operations': %" PRId64 ","
1820                              "'wr_operations': %" PRId64 ","
1821                              "'wr_highest_offset': %" PRId64
1822                              "} }",
1823                              bs->rd_bytes, bs->wr_bytes,
1824                              bs->rd_ops, bs->wr_ops,
1825                              bs->wr_highest_sector *
1826                              (uint64_t)BDRV_SECTOR_SIZE);
1827     dict  = qobject_to_qdict(res);
1828 
1829     if (*bs->device_name) {
1830         qdict_put(dict, "device", qstring_from_str(bs->device_name));
1831     }
1832 
1833     if (bs->file) {
1834         QObject *parent = bdrv_info_stats_bs(bs->file);
1835         qdict_put_obj(dict, "parent", parent);
1836     }
1837 
1838     return res;
1839 }
1840 
1841 void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1842 {
1843     QObject *obj;
1844     QList *devices;
1845     BlockDriverState *bs;
1846 
1847     devices = qlist_new();
1848 
1849     QTAILQ_FOREACH(bs, &bdrv_states, list) {
1850         obj = bdrv_info_stats_bs(bs);
1851         qlist_append_obj(devices, obj);
1852     }
1853 
1854     *ret_data = QOBJECT(devices);
1855 }
1856 
1857 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1858 {
1859     if (bs->backing_hd && bs->backing_hd->encrypted)
1860         return bs->backing_file;
1861     else if (bs->encrypted)
1862         return bs->filename;
1863     else
1864         return NULL;
1865 }
1866 
1867 void bdrv_get_backing_filename(BlockDriverState *bs,
1868                                char *filename, int filename_size)
1869 {
1870     if (!bs->backing_file) {
1871         pstrcpy(filename, filename_size, "");
1872     } else {
1873         pstrcpy(filename, filename_size, bs->backing_file);
1874     }
1875 }
1876 
1877 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1878                           const uint8_t *buf, int nb_sectors)
1879 {
1880     BlockDriver *drv = bs->drv;
1881     if (!drv)
1882         return -ENOMEDIUM;
1883     if (!drv->bdrv_write_compressed)
1884         return -ENOTSUP;
1885     if (bdrv_check_request(bs, sector_num, nb_sectors))
1886         return -EIO;
1887 
1888     if (bs->dirty_bitmap) {
1889         set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1890     }
1891 
1892     return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1893 }
1894 
1895 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1896 {
1897     BlockDriver *drv = bs->drv;
1898     if (!drv)
1899         return -ENOMEDIUM;
1900     if (!drv->bdrv_get_info)
1901         return -ENOTSUP;
1902     memset(bdi, 0, sizeof(*bdi));
1903     return drv->bdrv_get_info(bs, bdi);
1904 }
1905 
1906 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1907                       int64_t pos, int size)
1908 {
1909     BlockDriver *drv = bs->drv;
1910     if (!drv)
1911         return -ENOMEDIUM;
1912     if (drv->bdrv_save_vmstate)
1913         return drv->bdrv_save_vmstate(bs, buf, pos, size);
1914     if (bs->file)
1915         return bdrv_save_vmstate(bs->file, buf, pos, size);
1916     return -ENOTSUP;
1917 }
1918 
1919 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1920                       int64_t pos, int size)
1921 {
1922     BlockDriver *drv = bs->drv;
1923     if (!drv)
1924         return -ENOMEDIUM;
1925     if (drv->bdrv_load_vmstate)
1926         return drv->bdrv_load_vmstate(bs, buf, pos, size);
1927     if (bs->file)
1928         return bdrv_load_vmstate(bs->file, buf, pos, size);
1929     return -ENOTSUP;
1930 }
1931 
1932 void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
1933 {
1934     BlockDriver *drv = bs->drv;
1935 
1936     if (!drv || !drv->bdrv_debug_event) {
1937         return;
1938     }
1939 
1940     return drv->bdrv_debug_event(bs, event);
1941 
1942 }
1943 
1944 /**************************************************************/
1945 /* handling of snapshots */
1946 
1947 int bdrv_can_snapshot(BlockDriverState *bs)
1948 {
1949     BlockDriver *drv = bs->drv;
1950     if (!drv || bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1951         return 0;
1952     }
1953 
1954     if (!drv->bdrv_snapshot_create) {
1955         if (bs->file != NULL) {
1956             return bdrv_can_snapshot(bs->file);
1957         }
1958         return 0;
1959     }
1960 
1961     return 1;
1962 }
1963 
1964 int bdrv_is_snapshot(BlockDriverState *bs)
1965 {
1966     return !!(bs->open_flags & BDRV_O_SNAPSHOT);
1967 }
1968 
1969 BlockDriverState *bdrv_snapshots(void)
1970 {
1971     BlockDriverState *bs;
1972 
1973     if (bs_snapshots) {
1974         return bs_snapshots;
1975     }
1976 
1977     bs = NULL;
1978     while ((bs = bdrv_next(bs))) {
1979         if (bdrv_can_snapshot(bs)) {
1980             bs_snapshots = bs;
1981             return bs;
1982         }
1983     }
1984     return NULL;
1985 }
1986 
1987 int bdrv_snapshot_create(BlockDriverState *bs,
1988                          QEMUSnapshotInfo *sn_info)
1989 {
1990     BlockDriver *drv = bs->drv;
1991     if (!drv)
1992         return -ENOMEDIUM;
1993     if (drv->bdrv_snapshot_create)
1994         return drv->bdrv_snapshot_create(bs, sn_info);
1995     if (bs->file)
1996         return bdrv_snapshot_create(bs->file, sn_info);
1997     return -ENOTSUP;
1998 }
1999 
2000 int bdrv_snapshot_goto(BlockDriverState *bs,
2001                        const char *snapshot_id)
2002 {
2003     BlockDriver *drv = bs->drv;
2004     int ret, open_ret;
2005 
2006     if (!drv)
2007         return -ENOMEDIUM;
2008     if (drv->bdrv_snapshot_goto)
2009         return drv->bdrv_snapshot_goto(bs, snapshot_id);
2010 
2011     if (bs->file) {
2012         drv->bdrv_close(bs);
2013         ret = bdrv_snapshot_goto(bs->file, snapshot_id);
2014         open_ret = drv->bdrv_open(bs, bs->open_flags);
2015         if (open_ret < 0) {
2016             bdrv_delete(bs->file);
2017             bs->drv = NULL;
2018             return open_ret;
2019         }
2020         return ret;
2021     }
2022 
2023     return -ENOTSUP;
2024 }
2025 
2026 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
2027 {
2028     BlockDriver *drv = bs->drv;
2029     if (!drv)
2030         return -ENOMEDIUM;
2031     if (drv->bdrv_snapshot_delete)
2032         return drv->bdrv_snapshot_delete(bs, snapshot_id);
2033     if (bs->file)
2034         return bdrv_snapshot_delete(bs->file, snapshot_id);
2035     return -ENOTSUP;
2036 }
2037 
2038 int bdrv_snapshot_list(BlockDriverState *bs,
2039                        QEMUSnapshotInfo **psn_info)
2040 {
2041     BlockDriver *drv = bs->drv;
2042     if (!drv)
2043         return -ENOMEDIUM;
2044     if (drv->bdrv_snapshot_list)
2045         return drv->bdrv_snapshot_list(bs, psn_info);
2046     if (bs->file)
2047         return bdrv_snapshot_list(bs->file, psn_info);
2048     return -ENOTSUP;
2049 }
2050 
2051 int bdrv_snapshot_load_tmp(BlockDriverState *bs,
2052         const char *snapshot_name)
2053 {
2054     BlockDriver *drv = bs->drv;
2055     if (!drv) {
2056         return -ENOMEDIUM;
2057     }
2058     if (!bs->read_only) {
2059         return -EINVAL;
2060     }
2061     if (drv->bdrv_snapshot_load_tmp) {
2062         return drv->bdrv_snapshot_load_tmp(bs, snapshot_name);
2063     }
2064     return -ENOTSUP;
2065 }
2066 
2067 #define NB_SUFFIXES 4
2068 
2069 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
2070 {
2071     static const char suffixes[NB_SUFFIXES] = "KMGT";
2072     int64_t base;
2073     int i;
2074 
2075     if (size <= 999) {
2076         snprintf(buf, buf_size, "%" PRId64, size);
2077     } else {
2078         base = 1024;
2079         for(i = 0; i < NB_SUFFIXES; i++) {
2080             if (size < (10 * base)) {
2081                 snprintf(buf, buf_size, "%0.1f%c",
2082                          (double)size / base,
2083                          suffixes[i]);
2084                 break;
2085             } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
2086                 snprintf(buf, buf_size, "%" PRId64 "%c",
2087                          ((size + (base >> 1)) / base),
2088                          suffixes[i]);
2089                 break;
2090             }
2091             base = base * 1024;
2092         }
2093     }
2094     return buf;
2095 }
2096 
2097 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
2098 {
2099     char buf1[128], date_buf[128], clock_buf[128];
2100 #ifdef _WIN32
2101     struct tm *ptm;
2102 #else
2103     struct tm tm;
2104 #endif
2105     time_t ti;
2106     int64_t secs;
2107 
2108     if (!sn) {
2109         snprintf(buf, buf_size,
2110                  "%-10s%-20s%7s%20s%15s",
2111                  "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2112     } else {
2113         ti = sn->date_sec;
2114 #ifdef _WIN32
2115         ptm = localtime(&ti);
2116         strftime(date_buf, sizeof(date_buf),
2117                  "%Y-%m-%d %H:%M:%S", ptm);
2118 #else
2119         localtime_r(&ti, &tm);
2120         strftime(date_buf, sizeof(date_buf),
2121                  "%Y-%m-%d %H:%M:%S", &tm);
2122 #endif
2123         secs = sn->vm_clock_nsec / 1000000000;
2124         snprintf(clock_buf, sizeof(clock_buf),
2125                  "%02d:%02d:%02d.%03d",
2126                  (int)(secs / 3600),
2127                  (int)((secs / 60) % 60),
2128                  (int)(secs % 60),
2129                  (int)((sn->vm_clock_nsec / 1000000) % 1000));
2130         snprintf(buf, buf_size,
2131                  "%-10s%-20s%7s%20s%15s",
2132                  sn->id_str, sn->name,
2133                  get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
2134                  date_buf,
2135                  clock_buf);
2136     }
2137     return buf;
2138 }
2139 
2140 
2141 /**************************************************************/
2142 /* async I/Os */
2143 
2144 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
2145                                  QEMUIOVector *qiov, int nb_sectors,
2146                                  BlockDriverCompletionFunc *cb, void *opaque)
2147 {
2148     BlockDriver *drv = bs->drv;
2149     BlockDriverAIOCB *ret;
2150 
2151     trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
2152 
2153     if (!drv)
2154         return NULL;
2155     if (bdrv_check_request(bs, sector_num, nb_sectors))
2156         return NULL;
2157 
2158     ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
2159                               cb, opaque);
2160 
2161     if (ret) {
2162 	/* Update stats even though technically transfer has not happened. */
2163 	bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2164 	bs->rd_ops ++;
2165     }
2166 
2167     return ret;
2168 }
2169 
2170 typedef struct BlockCompleteData {
2171     BlockDriverCompletionFunc *cb;
2172     void *opaque;
2173     BlockDriverState *bs;
2174     int64_t sector_num;
2175     int nb_sectors;
2176 } BlockCompleteData;
2177 
2178 static void block_complete_cb(void *opaque, int ret)
2179 {
2180     BlockCompleteData *b = opaque;
2181 
2182     if (b->bs->dirty_bitmap) {
2183         set_dirty_bitmap(b->bs, b->sector_num, b->nb_sectors, 1);
2184     }
2185     b->cb(b->opaque, ret);
2186     qemu_free(b);
2187 }
2188 
2189 static BlockCompleteData *blk_dirty_cb_alloc(BlockDriverState *bs,
2190                                              int64_t sector_num,
2191                                              int nb_sectors,
2192                                              BlockDriverCompletionFunc *cb,
2193                                              void *opaque)
2194 {
2195     BlockCompleteData *blkdata = qemu_mallocz(sizeof(BlockCompleteData));
2196 
2197     blkdata->bs = bs;
2198     blkdata->cb = cb;
2199     blkdata->opaque = opaque;
2200     blkdata->sector_num = sector_num;
2201     blkdata->nb_sectors = nb_sectors;
2202 
2203     return blkdata;
2204 }
2205 
2206 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
2207                                   QEMUIOVector *qiov, int nb_sectors,
2208                                   BlockDriverCompletionFunc *cb, void *opaque)
2209 {
2210     BlockDriver *drv = bs->drv;
2211     BlockDriverAIOCB *ret;
2212     BlockCompleteData *blk_cb_data;
2213 
2214     trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
2215 
2216     if (!drv)
2217         return NULL;
2218     if (bs->read_only)
2219         return NULL;
2220     if (bdrv_check_request(bs, sector_num, nb_sectors))
2221         return NULL;
2222 
2223     if (bs->dirty_bitmap) {
2224         blk_cb_data = blk_dirty_cb_alloc(bs, sector_num, nb_sectors, cb,
2225                                          opaque);
2226         cb = &block_complete_cb;
2227         opaque = blk_cb_data;
2228     }
2229 
2230     ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
2231                                cb, opaque);
2232 
2233     if (ret) {
2234         /* Update stats even though technically transfer has not happened. */
2235         bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2236         bs->wr_ops ++;
2237         if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
2238             bs->wr_highest_sector = sector_num + nb_sectors - 1;
2239         }
2240     }
2241 
2242     return ret;
2243 }
2244 
2245 
2246 typedef struct MultiwriteCB {
2247     int error;
2248     int num_requests;
2249     int num_callbacks;
2250     struct {
2251         BlockDriverCompletionFunc *cb;
2252         void *opaque;
2253         QEMUIOVector *free_qiov;
2254         void *free_buf;
2255     } callbacks[];
2256 } MultiwriteCB;
2257 
2258 static void multiwrite_user_cb(MultiwriteCB *mcb)
2259 {
2260     int i;
2261 
2262     for (i = 0; i < mcb->num_callbacks; i++) {
2263         mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
2264         if (mcb->callbacks[i].free_qiov) {
2265             qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
2266         }
2267         qemu_free(mcb->callbacks[i].free_qiov);
2268         qemu_vfree(mcb->callbacks[i].free_buf);
2269     }
2270 }
2271 
2272 static void multiwrite_cb(void *opaque, int ret)
2273 {
2274     MultiwriteCB *mcb = opaque;
2275 
2276     trace_multiwrite_cb(mcb, ret);
2277 
2278     if (ret < 0 && !mcb->error) {
2279         mcb->error = ret;
2280     }
2281 
2282     mcb->num_requests--;
2283     if (mcb->num_requests == 0) {
2284         multiwrite_user_cb(mcb);
2285         qemu_free(mcb);
2286     }
2287 }
2288 
2289 static int multiwrite_req_compare(const void *a, const void *b)
2290 {
2291     const BlockRequest *req1 = a, *req2 = b;
2292 
2293     /*
2294      * Note that we can't simply subtract req2->sector from req1->sector
2295      * here as that could overflow the return value.
2296      */
2297     if (req1->sector > req2->sector) {
2298         return 1;
2299     } else if (req1->sector < req2->sector) {
2300         return -1;
2301     } else {
2302         return 0;
2303     }
2304 }
2305 
2306 /*
2307  * Takes a bunch of requests and tries to merge them. Returns the number of
2308  * requests that remain after merging.
2309  */
2310 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
2311     int num_reqs, MultiwriteCB *mcb)
2312 {
2313     int i, outidx;
2314 
2315     // Sort requests by start sector
2316     qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
2317 
2318     // Check if adjacent requests touch the same clusters. If so, combine them,
2319     // filling up gaps with zero sectors.
2320     outidx = 0;
2321     for (i = 1; i < num_reqs; i++) {
2322         int merge = 0;
2323         int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
2324 
2325         // This handles the cases that are valid for all block drivers, namely
2326         // exactly sequential writes and overlapping writes.
2327         if (reqs[i].sector <= oldreq_last) {
2328             merge = 1;
2329         }
2330 
2331         // The block driver may decide that it makes sense to combine requests
2332         // even if there is a gap of some sectors between them. In this case,
2333         // the gap is filled with zeros (therefore only applicable for yet
2334         // unused space in format like qcow2).
2335         if (!merge && bs->drv->bdrv_merge_requests) {
2336             merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
2337         }
2338 
2339         if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
2340             merge = 0;
2341         }
2342 
2343         if (merge) {
2344             size_t size;
2345             QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
2346             qemu_iovec_init(qiov,
2347                 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
2348 
2349             // Add the first request to the merged one. If the requests are
2350             // overlapping, drop the last sectors of the first request.
2351             size = (reqs[i].sector - reqs[outidx].sector) << 9;
2352             qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
2353 
2354             // We might need to add some zeros between the two requests
2355             if (reqs[i].sector > oldreq_last) {
2356                 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
2357                 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
2358                 memset(buf, 0, zero_bytes);
2359                 qemu_iovec_add(qiov, buf, zero_bytes);
2360                 mcb->callbacks[i].free_buf = buf;
2361             }
2362 
2363             // Add the second request
2364             qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2365 
2366             reqs[outidx].nb_sectors = qiov->size >> 9;
2367             reqs[outidx].qiov = qiov;
2368 
2369             mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2370         } else {
2371             outidx++;
2372             reqs[outidx].sector     = reqs[i].sector;
2373             reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2374             reqs[outidx].qiov       = reqs[i].qiov;
2375         }
2376     }
2377 
2378     return outidx + 1;
2379 }
2380 
2381 /*
2382  * Submit multiple AIO write requests at once.
2383  *
2384  * On success, the function returns 0 and all requests in the reqs array have
2385  * been submitted. In error case this function returns -1, and any of the
2386  * requests may or may not be submitted yet. In particular, this means that the
2387  * callback will be called for some of the requests, for others it won't. The
2388  * caller must check the error field of the BlockRequest to wait for the right
2389  * callbacks (if error != 0, no callback will be called).
2390  *
2391  * The implementation may modify the contents of the reqs array, e.g. to merge
2392  * requests. However, the fields opaque and error are left unmodified as they
2393  * are used to signal failure for a single request to the caller.
2394  */
2395 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2396 {
2397     BlockDriverAIOCB *acb;
2398     MultiwriteCB *mcb;
2399     int i;
2400 
2401     /* don't submit writes if we don't have a medium */
2402     if (bs->drv == NULL) {
2403         for (i = 0; i < num_reqs; i++) {
2404             reqs[i].error = -ENOMEDIUM;
2405         }
2406         return -1;
2407     }
2408 
2409     if (num_reqs == 0) {
2410         return 0;
2411     }
2412 
2413     // Create MultiwriteCB structure
2414     mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2415     mcb->num_requests = 0;
2416     mcb->num_callbacks = num_reqs;
2417 
2418     for (i = 0; i < num_reqs; i++) {
2419         mcb->callbacks[i].cb = reqs[i].cb;
2420         mcb->callbacks[i].opaque = reqs[i].opaque;
2421     }
2422 
2423     // Check for mergable requests
2424     num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2425 
2426     trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs);
2427 
2428     /*
2429      * Run the aio requests. As soon as one request can't be submitted
2430      * successfully, fail all requests that are not yet submitted (we must
2431      * return failure for all requests anyway)
2432      *
2433      * num_requests cannot be set to the right value immediately: If
2434      * bdrv_aio_writev fails for some request, num_requests would be too high
2435      * and therefore multiwrite_cb() would never recognize the multiwrite
2436      * request as completed. We also cannot use the loop variable i to set it
2437      * when the first request fails because the callback may already have been
2438      * called for previously submitted requests. Thus, num_requests must be
2439      * incremented for each request that is submitted.
2440      *
2441      * The problem that callbacks may be called early also means that we need
2442      * to take care that num_requests doesn't become 0 before all requests are
2443      * submitted - multiwrite_cb() would consider the multiwrite request
2444      * completed. A dummy request that is "completed" by a manual call to
2445      * multiwrite_cb() takes care of this.
2446      */
2447     mcb->num_requests = 1;
2448 
2449     // Run the aio requests
2450     for (i = 0; i < num_reqs; i++) {
2451         mcb->num_requests++;
2452         acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2453             reqs[i].nb_sectors, multiwrite_cb, mcb);
2454 
2455         if (acb == NULL) {
2456             // We can only fail the whole thing if no request has been
2457             // submitted yet. Otherwise we'll wait for the submitted AIOs to
2458             // complete and report the error in the callback.
2459             if (i == 0) {
2460                 trace_bdrv_aio_multiwrite_earlyfail(mcb);
2461                 goto fail;
2462             } else {
2463                 trace_bdrv_aio_multiwrite_latefail(mcb, i);
2464                 multiwrite_cb(mcb, -EIO);
2465                 break;
2466             }
2467         }
2468     }
2469 
2470     /* Complete the dummy request */
2471     multiwrite_cb(mcb, 0);
2472 
2473     return 0;
2474 
2475 fail:
2476     for (i = 0; i < mcb->num_callbacks; i++) {
2477         reqs[i].error = -EIO;
2478     }
2479     qemu_free(mcb);
2480     return -1;
2481 }
2482 
2483 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2484         BlockDriverCompletionFunc *cb, void *opaque)
2485 {
2486     BlockDriver *drv = bs->drv;
2487 
2488     trace_bdrv_aio_flush(bs, opaque);
2489 
2490     if (bs->open_flags & BDRV_O_NO_FLUSH) {
2491         return bdrv_aio_noop_em(bs, cb, opaque);
2492     }
2493 
2494     if (!drv)
2495         return NULL;
2496     return drv->bdrv_aio_flush(bs, cb, opaque);
2497 }
2498 
2499 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2500 {
2501     acb->pool->cancel(acb);
2502 }
2503 
2504 
2505 /**************************************************************/
2506 /* async block device emulation */
2507 
2508 typedef struct BlockDriverAIOCBSync {
2509     BlockDriverAIOCB common;
2510     QEMUBH *bh;
2511     int ret;
2512     /* vector translation state */
2513     QEMUIOVector *qiov;
2514     uint8_t *bounce;
2515     int is_write;
2516 } BlockDriverAIOCBSync;
2517 
2518 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2519 {
2520     BlockDriverAIOCBSync *acb =
2521         container_of(blockacb, BlockDriverAIOCBSync, common);
2522     qemu_bh_delete(acb->bh);
2523     acb->bh = NULL;
2524     qemu_aio_release(acb);
2525 }
2526 
2527 static AIOPool bdrv_em_aio_pool = {
2528     .aiocb_size         = sizeof(BlockDriverAIOCBSync),
2529     .cancel             = bdrv_aio_cancel_em,
2530 };
2531 
2532 static void bdrv_aio_bh_cb(void *opaque)
2533 {
2534     BlockDriverAIOCBSync *acb = opaque;
2535 
2536     if (!acb->is_write)
2537         qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2538     qemu_vfree(acb->bounce);
2539     acb->common.cb(acb->common.opaque, acb->ret);
2540     qemu_bh_delete(acb->bh);
2541     acb->bh = NULL;
2542     qemu_aio_release(acb);
2543 }
2544 
2545 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2546                                             int64_t sector_num,
2547                                             QEMUIOVector *qiov,
2548                                             int nb_sectors,
2549                                             BlockDriverCompletionFunc *cb,
2550                                             void *opaque,
2551                                             int is_write)
2552 
2553 {
2554     BlockDriverAIOCBSync *acb;
2555 
2556     acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2557     acb->is_write = is_write;
2558     acb->qiov = qiov;
2559     acb->bounce = qemu_blockalign(bs, qiov->size);
2560 
2561     if (!acb->bh)
2562         acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2563 
2564     if (is_write) {
2565         qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2566         acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2567     } else {
2568         acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2569     }
2570 
2571     qemu_bh_schedule(acb->bh);
2572 
2573     return &acb->common;
2574 }
2575 
2576 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2577         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2578         BlockDriverCompletionFunc *cb, void *opaque)
2579 {
2580     return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2581 }
2582 
2583 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2584         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2585         BlockDriverCompletionFunc *cb, void *opaque)
2586 {
2587     return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2588 }
2589 
2590 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2591         BlockDriverCompletionFunc *cb, void *opaque)
2592 {
2593     BlockDriverAIOCBSync *acb;
2594 
2595     acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2596     acb->is_write = 1; /* don't bounce in the completion hadler */
2597     acb->qiov = NULL;
2598     acb->bounce = NULL;
2599     acb->ret = 0;
2600 
2601     if (!acb->bh)
2602         acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2603 
2604     bdrv_flush(bs);
2605     qemu_bh_schedule(acb->bh);
2606     return &acb->common;
2607 }
2608 
2609 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2610         BlockDriverCompletionFunc *cb, void *opaque)
2611 {
2612     BlockDriverAIOCBSync *acb;
2613 
2614     acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2615     acb->is_write = 1; /* don't bounce in the completion handler */
2616     acb->qiov = NULL;
2617     acb->bounce = NULL;
2618     acb->ret = 0;
2619 
2620     if (!acb->bh) {
2621         acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2622     }
2623 
2624     qemu_bh_schedule(acb->bh);
2625     return &acb->common;
2626 }
2627 
2628 /**************************************************************/
2629 /* sync block device emulation */
2630 
2631 static void bdrv_rw_em_cb(void *opaque, int ret)
2632 {
2633     *(int *)opaque = ret;
2634 }
2635 
2636 #define NOT_DONE 0x7fffffff
2637 
2638 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2639                         uint8_t *buf, int nb_sectors)
2640 {
2641     int async_ret;
2642     BlockDriverAIOCB *acb;
2643     struct iovec iov;
2644     QEMUIOVector qiov;
2645 
2646     async_context_push();
2647 
2648     async_ret = NOT_DONE;
2649     iov.iov_base = (void *)buf;
2650     iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2651     qemu_iovec_init_external(&qiov, &iov, 1);
2652     acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2653         bdrv_rw_em_cb, &async_ret);
2654     if (acb == NULL) {
2655         async_ret = -1;
2656         goto fail;
2657     }
2658 
2659     while (async_ret == NOT_DONE) {
2660         qemu_aio_wait();
2661     }
2662 
2663 
2664 fail:
2665     async_context_pop();
2666     return async_ret;
2667 }
2668 
2669 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2670                          const uint8_t *buf, int nb_sectors)
2671 {
2672     int async_ret;
2673     BlockDriverAIOCB *acb;
2674     struct iovec iov;
2675     QEMUIOVector qiov;
2676 
2677     async_context_push();
2678 
2679     async_ret = NOT_DONE;
2680     iov.iov_base = (void *)buf;
2681     iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2682     qemu_iovec_init_external(&qiov, &iov, 1);
2683     acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2684         bdrv_rw_em_cb, &async_ret);
2685     if (acb == NULL) {
2686         async_ret = -1;
2687         goto fail;
2688     }
2689     while (async_ret == NOT_DONE) {
2690         qemu_aio_wait();
2691     }
2692 
2693 fail:
2694     async_context_pop();
2695     return async_ret;
2696 }
2697 
2698 void bdrv_init(void)
2699 {
2700     module_call_init(MODULE_INIT_BLOCK);
2701 }
2702 
2703 void bdrv_init_with_whitelist(void)
2704 {
2705     use_bdrv_whitelist = 1;
2706     bdrv_init();
2707 }
2708 
2709 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2710                    BlockDriverCompletionFunc *cb, void *opaque)
2711 {
2712     BlockDriverAIOCB *acb;
2713 
2714     if (pool->free_aiocb) {
2715         acb = pool->free_aiocb;
2716         pool->free_aiocb = acb->next;
2717     } else {
2718         acb = qemu_mallocz(pool->aiocb_size);
2719         acb->pool = pool;
2720     }
2721     acb->bs = bs;
2722     acb->cb = cb;
2723     acb->opaque = opaque;
2724     return acb;
2725 }
2726 
2727 void qemu_aio_release(void *p)
2728 {
2729     BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2730     AIOPool *pool = acb->pool;
2731     acb->next = pool->free_aiocb;
2732     pool->free_aiocb = acb;
2733 }
2734 
2735 /**************************************************************/
2736 /* removable device support */
2737 
2738 /**
2739  * Return TRUE if the media is present
2740  */
2741 int bdrv_is_inserted(BlockDriverState *bs)
2742 {
2743     BlockDriver *drv = bs->drv;
2744     int ret;
2745     if (!drv)
2746         return 0;
2747     if (!drv->bdrv_is_inserted)
2748         return !bs->tray_open;
2749     ret = drv->bdrv_is_inserted(bs);
2750     return ret;
2751 }
2752 
2753 /**
2754  * Return TRUE if the media changed since the last call to this
2755  * function. It is currently only used for floppy disks
2756  */
2757 int bdrv_media_changed(BlockDriverState *bs)
2758 {
2759     BlockDriver *drv = bs->drv;
2760     int ret;
2761 
2762     if (!drv || !drv->bdrv_media_changed)
2763         ret = -ENOTSUP;
2764     else
2765         ret = drv->bdrv_media_changed(bs);
2766     if (ret == -ENOTSUP)
2767         ret = bs->media_changed;
2768     bs->media_changed = 0;
2769     return ret;
2770 }
2771 
2772 /**
2773  * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2774  */
2775 int bdrv_eject(BlockDriverState *bs, int eject_flag)
2776 {
2777     BlockDriver *drv = bs->drv;
2778     int ret;
2779 
2780     if (bs->locked) {
2781         return -EBUSY;
2782     }
2783 
2784     if (!drv || !drv->bdrv_eject) {
2785         ret = -ENOTSUP;
2786     } else {
2787         ret = drv->bdrv_eject(bs, eject_flag);
2788     }
2789     if (ret == -ENOTSUP) {
2790         ret = 0;
2791     }
2792     if (ret >= 0) {
2793         bs->tray_open = eject_flag;
2794     }
2795 
2796     return ret;
2797 }
2798 
2799 int bdrv_is_locked(BlockDriverState *bs)
2800 {
2801     return bs->locked;
2802 }
2803 
2804 /**
2805  * Lock or unlock the media (if it is locked, the user won't be able
2806  * to eject it manually).
2807  */
2808 void bdrv_set_locked(BlockDriverState *bs, int locked)
2809 {
2810     BlockDriver *drv = bs->drv;
2811 
2812     bs->locked = locked;
2813     if (drv && drv->bdrv_set_locked) {
2814         drv->bdrv_set_locked(bs, locked);
2815     }
2816 }
2817 
2818 /* needed for generic scsi interface */
2819 
2820 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2821 {
2822     BlockDriver *drv = bs->drv;
2823 
2824     if (drv && drv->bdrv_ioctl)
2825         return drv->bdrv_ioctl(bs, req, buf);
2826     return -ENOTSUP;
2827 }
2828 
2829 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2830         unsigned long int req, void *buf,
2831         BlockDriverCompletionFunc *cb, void *opaque)
2832 {
2833     BlockDriver *drv = bs->drv;
2834 
2835     if (drv && drv->bdrv_aio_ioctl)
2836         return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2837     return NULL;
2838 }
2839 
2840 
2841 
2842 void *qemu_blockalign(BlockDriverState *bs, size_t size)
2843 {
2844     return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2845 }
2846 
2847 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2848 {
2849     int64_t bitmap_size;
2850 
2851     bs->dirty_count = 0;
2852     if (enable) {
2853         if (!bs->dirty_bitmap) {
2854             bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2855                     BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2856             bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2857 
2858             bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2859         }
2860     } else {
2861         if (bs->dirty_bitmap) {
2862             qemu_free(bs->dirty_bitmap);
2863             bs->dirty_bitmap = NULL;
2864         }
2865     }
2866 }
2867 
2868 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2869 {
2870     int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2871 
2872     if (bs->dirty_bitmap &&
2873         (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2874         return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2875             (1UL << (chunk % (sizeof(unsigned long) * 8))));
2876     } else {
2877         return 0;
2878     }
2879 }
2880 
2881 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2882                       int nr_sectors)
2883 {
2884     set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2885 }
2886 
2887 int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2888 {
2889     return bs->dirty_count;
2890 }
2891 
2892 void bdrv_set_in_use(BlockDriverState *bs, int in_use)
2893 {
2894     assert(bs->in_use != in_use);
2895     bs->in_use = in_use;
2896 }
2897 
2898 int bdrv_in_use(BlockDriverState *bs)
2899 {
2900     return bs->in_use;
2901 }
2902 
2903 int bdrv_img_create(const char *filename, const char *fmt,
2904                     const char *base_filename, const char *base_fmt,
2905                     char *options, uint64_t img_size, int flags)
2906 {
2907     QEMUOptionParameter *param = NULL, *create_options = NULL;
2908     QEMUOptionParameter *backing_fmt, *backing_file;
2909     BlockDriverState *bs = NULL;
2910     BlockDriver *drv, *proto_drv;
2911     BlockDriver *backing_drv = NULL;
2912     int ret = 0;
2913 
2914     /* Find driver and parse its options */
2915     drv = bdrv_find_format(fmt);
2916     if (!drv) {
2917         error_report("Unknown file format '%s'", fmt);
2918         ret = -EINVAL;
2919         goto out;
2920     }
2921 
2922     proto_drv = bdrv_find_protocol(filename);
2923     if (!proto_drv) {
2924         error_report("Unknown protocol '%s'", filename);
2925         ret = -EINVAL;
2926         goto out;
2927     }
2928 
2929     create_options = append_option_parameters(create_options,
2930                                               drv->create_options);
2931     create_options = append_option_parameters(create_options,
2932                                               proto_drv->create_options);
2933 
2934     /* Create parameter list with default values */
2935     param = parse_option_parameters("", create_options, param);
2936 
2937     set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size);
2938 
2939     /* Parse -o options */
2940     if (options) {
2941         param = parse_option_parameters(options, create_options, param);
2942         if (param == NULL) {
2943             error_report("Invalid options for file format '%s'.", fmt);
2944             ret = -EINVAL;
2945             goto out;
2946         }
2947     }
2948 
2949     if (base_filename) {
2950         if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE,
2951                                  base_filename)) {
2952             error_report("Backing file not supported for file format '%s'",
2953                          fmt);
2954             ret = -EINVAL;
2955             goto out;
2956         }
2957     }
2958 
2959     if (base_fmt) {
2960         if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) {
2961             error_report("Backing file format not supported for file "
2962                          "format '%s'", fmt);
2963             ret = -EINVAL;
2964             goto out;
2965         }
2966     }
2967 
2968     backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE);
2969     if (backing_file && backing_file->value.s) {
2970         if (!strcmp(filename, backing_file->value.s)) {
2971             error_report("Error: Trying to create an image with the "
2972                          "same filename as the backing file");
2973             ret = -EINVAL;
2974             goto out;
2975         }
2976     }
2977 
2978     backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT);
2979     if (backing_fmt && backing_fmt->value.s) {
2980         backing_drv = bdrv_find_format(backing_fmt->value.s);
2981         if (!backing_drv) {
2982             error_report("Unknown backing file format '%s'",
2983                          backing_fmt->value.s);
2984             ret = -EINVAL;
2985             goto out;
2986         }
2987     }
2988 
2989     // The size for the image must always be specified, with one exception:
2990     // If we are using a backing file, we can obtain the size from there
2991     if (get_option_parameter(param, BLOCK_OPT_SIZE)->value.n == -1) {
2992         if (backing_file && backing_file->value.s) {
2993             uint64_t size;
2994             char buf[32];
2995 
2996             bs = bdrv_new("");
2997 
2998             ret = bdrv_open(bs, backing_file->value.s, flags, backing_drv);
2999             if (ret < 0) {
3000                 error_report("Could not open '%s'", backing_file->value.s);
3001                 goto out;
3002             }
3003             bdrv_get_geometry(bs, &size);
3004             size *= 512;
3005 
3006             snprintf(buf, sizeof(buf), "%" PRId64, size);
3007             set_option_parameter(param, BLOCK_OPT_SIZE, buf);
3008         } else {
3009             error_report("Image creation needs a size parameter");
3010             ret = -EINVAL;
3011             goto out;
3012         }
3013     }
3014 
3015     printf("Formatting '%s', fmt=%s ", filename, fmt);
3016     print_option_parameters(param);
3017     puts("");
3018 
3019     ret = bdrv_create(drv, filename, param);
3020 
3021     if (ret < 0) {
3022         if (ret == -ENOTSUP) {
3023             error_report("Formatting or formatting option not supported for "
3024                          "file format '%s'", fmt);
3025         } else if (ret == -EFBIG) {
3026             error_report("The image size is too large for file format '%s'",
3027                          fmt);
3028         } else {
3029             error_report("%s: error while creating %s: %s", filename, fmt,
3030                          strerror(-ret));
3031         }
3032     }
3033 
3034 out:
3035     free_option_parameters(create_options);
3036     free_option_parameters(param);
3037 
3038     if (bs) {
3039         bdrv_delete(bs);
3040     }
3041 
3042     return ret;
3043 }
3044