xref: /openbmc/qemu/blockdev.c (revision 56411125)
1 /*
2  * QEMU host block devices
3  *
4  * Copyright (c) 2003-2008 Fabrice Bellard
5  *
6  * This work is licensed under the terms of the GNU GPL, version 2 or
7  * later.  See the COPYING file in the top-level directory.
8  *
9  * This file incorporates work covered by the following copyright and
10  * permission notice:
11  *
12  * Copyright (c) 2003-2008 Fabrice Bellard
13  *
14  * Permission is hereby granted, free of charge, to any person obtaining a copy
15  * of this software and associated documentation files (the "Software"), to deal
16  * in the Software without restriction, including without limitation the rights
17  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
18  * copies of the Software, and to permit persons to whom the Software is
19  * furnished to do so, subject to the following conditions:
20  *
21  * The above copyright notice and this permission notice shall be included in
22  * all copies or substantial portions of the Software.
23  *
24  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
27  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
29  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
30  * THE SOFTWARE.
31  */
32 
33 #include "sysemu/block-backend.h"
34 #include "sysemu/blockdev.h"
35 #include "hw/block/block.h"
36 #include "block/blockjob.h"
37 #include "block/throttle-groups.h"
38 #include "monitor/monitor.h"
39 #include "qemu/error-report.h"
40 #include "qemu/option.h"
41 #include "qemu/config-file.h"
42 #include "qapi/qmp/types.h"
43 #include "qapi-visit.h"
44 #include "qapi/qmp/qerror.h"
45 #include "qapi/qmp-output-visitor.h"
46 #include "qapi/util.h"
47 #include "sysemu/sysemu.h"
48 #include "block/block_int.h"
49 #include "qmp-commands.h"
50 #include "trace.h"
51 #include "sysemu/arch_init.h"
52 
53 static const char *const if_name[IF_COUNT] = {
54     [IF_NONE] = "none",
55     [IF_IDE] = "ide",
56     [IF_SCSI] = "scsi",
57     [IF_FLOPPY] = "floppy",
58     [IF_PFLASH] = "pflash",
59     [IF_MTD] = "mtd",
60     [IF_SD] = "sd",
61     [IF_VIRTIO] = "virtio",
62     [IF_XEN] = "xen",
63 };
64 
65 static int if_max_devs[IF_COUNT] = {
66     /*
67      * Do not change these numbers!  They govern how drive option
68      * index maps to unit and bus.  That mapping is ABI.
69      *
70      * All controllers used to imlement if=T drives need to support
71      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
72      * Otherwise, some index values map to "impossible" bus, unit
73      * values.
74      *
75      * For instance, if you change [IF_SCSI] to 255, -drive
76      * if=scsi,index=12 no longer means bus=1,unit=5, but
77      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
78      * the drive can't be set up.  Regression.
79      */
80     [IF_IDE] = 2,
81     [IF_SCSI] = 7,
82 };
83 
84 /**
85  * Boards may call this to offer board-by-board overrides
86  * of the default, global values.
87  */
88 void override_max_devs(BlockInterfaceType type, int max_devs)
89 {
90     BlockBackend *blk;
91     DriveInfo *dinfo;
92 
93     if (max_devs <= 0) {
94         return;
95     }
96 
97     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
98         dinfo = blk_legacy_dinfo(blk);
99         if (dinfo->type == type) {
100             fprintf(stderr, "Cannot override units-per-bus property of"
101                     " the %s interface, because a drive of that type has"
102                     " already been added.\n", if_name[type]);
103             g_assert_not_reached();
104         }
105     }
106 
107     if_max_devs[type] = max_devs;
108 }
109 
110 /*
111  * We automatically delete the drive when a device using it gets
112  * unplugged.  Questionable feature, but we can't just drop it.
113  * Device models call blockdev_mark_auto_del() to schedule the
114  * automatic deletion, and generic qdev code calls blockdev_auto_del()
115  * when deletion is actually safe.
116  */
117 void blockdev_mark_auto_del(BlockBackend *blk)
118 {
119     DriveInfo *dinfo = blk_legacy_dinfo(blk);
120     BlockDriverState *bs = blk_bs(blk);
121     AioContext *aio_context;
122 
123     if (!dinfo) {
124         return;
125     }
126 
127     if (bs) {
128         aio_context = bdrv_get_aio_context(bs);
129         aio_context_acquire(aio_context);
130 
131         if (bs->job) {
132             block_job_cancel(bs->job);
133         }
134 
135         aio_context_release(aio_context);
136     }
137 
138     dinfo->auto_del = 1;
139 }
140 
141 void blockdev_auto_del(BlockBackend *blk)
142 {
143     DriveInfo *dinfo = blk_legacy_dinfo(blk);
144 
145     if (dinfo && dinfo->auto_del) {
146         blk_unref(blk);
147     }
148 }
149 
150 /**
151  * Returns the current mapping of how many units per bus
152  * a particular interface can support.
153  *
154  *  A positive integer indicates n units per bus.
155  *  0 implies the mapping has not been established.
156  * -1 indicates an invalid BlockInterfaceType was given.
157  */
158 int drive_get_max_devs(BlockInterfaceType type)
159 {
160     if (type >= IF_IDE && type < IF_COUNT) {
161         return if_max_devs[type];
162     }
163 
164     return -1;
165 }
166 
167 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
168 {
169     int max_devs = if_max_devs[type];
170     return max_devs ? index / max_devs : 0;
171 }
172 
173 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
174 {
175     int max_devs = if_max_devs[type];
176     return max_devs ? index % max_devs : index;
177 }
178 
179 QemuOpts *drive_def(const char *optstr)
180 {
181     return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
182 }
183 
184 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
185                     const char *optstr)
186 {
187     QemuOpts *opts;
188 
189     opts = drive_def(optstr);
190     if (!opts) {
191         return NULL;
192     }
193     if (type != IF_DEFAULT) {
194         qemu_opt_set(opts, "if", if_name[type], &error_abort);
195     }
196     if (index >= 0) {
197         qemu_opt_set_number(opts, "index", index, &error_abort);
198     }
199     if (file)
200         qemu_opt_set(opts, "file", file, &error_abort);
201     return opts;
202 }
203 
204 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
205 {
206     BlockBackend *blk;
207     DriveInfo *dinfo;
208 
209     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
210         dinfo = blk_legacy_dinfo(blk);
211         if (dinfo && dinfo->type == type
212             && dinfo->bus == bus && dinfo->unit == unit) {
213             return dinfo;
214         }
215     }
216 
217     return NULL;
218 }
219 
220 bool drive_check_orphaned(void)
221 {
222     BlockBackend *blk;
223     DriveInfo *dinfo;
224     bool rs = false;
225 
226     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
227         dinfo = blk_legacy_dinfo(blk);
228         /* If dinfo->bdrv->dev is NULL, it has no device attached. */
229         /* Unless this is a default drive, this may be an oversight. */
230         if (!blk_get_attached_dev(blk) && !dinfo->is_default &&
231             dinfo->type != IF_NONE) {
232             fprintf(stderr, "Warning: Orphaned drive without device: "
233                     "id=%s,file=%s,if=%s,bus=%d,unit=%d\n",
234                     blk_name(blk), blk_bs(blk) ? blk_bs(blk)->filename : "",
235                     if_name[dinfo->type], dinfo->bus, dinfo->unit);
236             rs = true;
237         }
238     }
239 
240     return rs;
241 }
242 
243 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
244 {
245     return drive_get(type,
246                      drive_index_to_bus_id(type, index),
247                      drive_index_to_unit_id(type, index));
248 }
249 
250 int drive_get_max_bus(BlockInterfaceType type)
251 {
252     int max_bus;
253     BlockBackend *blk;
254     DriveInfo *dinfo;
255 
256     max_bus = -1;
257     for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
258         dinfo = blk_legacy_dinfo(blk);
259         if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
260             max_bus = dinfo->bus;
261         }
262     }
263     return max_bus;
264 }
265 
266 /* Get a block device.  This should only be used for single-drive devices
267    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
268    appropriate bus.  */
269 DriveInfo *drive_get_next(BlockInterfaceType type)
270 {
271     static int next_block_unit[IF_COUNT];
272 
273     return drive_get(type, 0, next_block_unit[type]++);
274 }
275 
276 static void bdrv_format_print(void *opaque, const char *name)
277 {
278     error_printf(" %s", name);
279 }
280 
281 typedef struct {
282     QEMUBH *bh;
283     BlockDriverState *bs;
284 } BDRVPutRefBH;
285 
286 static void bdrv_put_ref_bh(void *opaque)
287 {
288     BDRVPutRefBH *s = opaque;
289 
290     bdrv_unref(s->bs);
291     qemu_bh_delete(s->bh);
292     g_free(s);
293 }
294 
295 /*
296  * Release a BDS reference in a BH
297  *
298  * It is not safe to use bdrv_unref() from a callback function when the callers
299  * still need the BlockDriverState.  In such cases we schedule a BH to release
300  * the reference.
301  */
302 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
303 {
304     BDRVPutRefBH *s;
305 
306     s = g_new(BDRVPutRefBH, 1);
307     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
308     s->bs = bs;
309     qemu_bh_schedule(s->bh);
310 }
311 
312 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
313 {
314     if (!strcmp(buf, "ignore")) {
315         return BLOCKDEV_ON_ERROR_IGNORE;
316     } else if (!is_read && !strcmp(buf, "enospc")) {
317         return BLOCKDEV_ON_ERROR_ENOSPC;
318     } else if (!strcmp(buf, "stop")) {
319         return BLOCKDEV_ON_ERROR_STOP;
320     } else if (!strcmp(buf, "report")) {
321         return BLOCKDEV_ON_ERROR_REPORT;
322     } else {
323         error_setg(errp, "'%s' invalid %s error action",
324                    buf, is_read ? "read" : "write");
325         return -1;
326     }
327 }
328 
329 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
330 {
331     if (throttle_conflicting(cfg)) {
332         error_setg(errp, "bps/iops/max total values and read/write values"
333                          " cannot be used at the same time");
334         return false;
335     }
336 
337     if (!throttle_is_valid(cfg)) {
338         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
339         return false;
340     }
341 
342     if (throttle_max_is_missing_limit(cfg)) {
343         error_setg(errp, "bps_max/iops_max require corresponding"
344                          " bps/iops values");
345         return false;
346     }
347 
348     return true;
349 }
350 
351 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
352 
353 /* All parameters but @opts are optional and may be set to NULL. */
354 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
355     const char **throttling_group, ThrottleConfig *throttle_cfg,
356     BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
357 {
358     const char *discard;
359     Error *local_error = NULL;
360     const char *aio;
361 
362     if (bdrv_flags) {
363         if (!qemu_opt_get_bool(opts, "read-only", false)) {
364             *bdrv_flags |= BDRV_O_RDWR;
365         }
366         if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
367             *bdrv_flags |= BDRV_O_COPY_ON_READ;
368         }
369 
370         if ((discard = qemu_opt_get(opts, "discard")) != NULL) {
371             if (bdrv_parse_discard_flags(discard, bdrv_flags) != 0) {
372                 error_setg(errp, "Invalid discard option");
373                 return;
374             }
375         }
376 
377         if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true)) {
378             *bdrv_flags |= BDRV_O_CACHE_WB;
379         }
380         if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_DIRECT, false)) {
381             *bdrv_flags |= BDRV_O_NOCACHE;
382         }
383         if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_NO_FLUSH, false)) {
384             *bdrv_flags |= BDRV_O_NO_FLUSH;
385         }
386 
387         if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
388             if (!strcmp(aio, "native")) {
389                 *bdrv_flags |= BDRV_O_NATIVE_AIO;
390             } else if (!strcmp(aio, "threads")) {
391                 /* this is the default */
392             } else {
393                error_setg(errp, "invalid aio option");
394                return;
395             }
396         }
397     }
398 
399     /* disk I/O throttling */
400     if (throttling_group) {
401         *throttling_group = qemu_opt_get(opts, "throttling.group");
402     }
403 
404     if (throttle_cfg) {
405         memset(throttle_cfg, 0, sizeof(*throttle_cfg));
406         throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
407             qemu_opt_get_number(opts, "throttling.bps-total", 0);
408         throttle_cfg->buckets[THROTTLE_BPS_READ].avg  =
409             qemu_opt_get_number(opts, "throttling.bps-read", 0);
410         throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
411             qemu_opt_get_number(opts, "throttling.bps-write", 0);
412         throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
413             qemu_opt_get_number(opts, "throttling.iops-total", 0);
414         throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
415             qemu_opt_get_number(opts, "throttling.iops-read", 0);
416         throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
417             qemu_opt_get_number(opts, "throttling.iops-write", 0);
418 
419         throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
420             qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
421         throttle_cfg->buckets[THROTTLE_BPS_READ].max  =
422             qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
423         throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
424             qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
425         throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
426             qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
427         throttle_cfg->buckets[THROTTLE_OPS_READ].max =
428             qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
429         throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
430             qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
431 
432         throttle_cfg->op_size =
433             qemu_opt_get_number(opts, "throttling.iops-size", 0);
434 
435         if (!check_throttle_config(throttle_cfg, errp)) {
436             return;
437         }
438     }
439 
440     if (detect_zeroes) {
441         *detect_zeroes =
442             qapi_enum_parse(BlockdevDetectZeroesOptions_lookup,
443                             qemu_opt_get(opts, "detect-zeroes"),
444                             BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
445                             BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
446                             &local_error);
447         if (local_error) {
448             error_propagate(errp, local_error);
449             return;
450         }
451 
452         if (bdrv_flags &&
453             *detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
454             !(*bdrv_flags & BDRV_O_UNMAP))
455         {
456             error_setg(errp, "setting detect-zeroes to unmap is not allowed "
457                              "without setting discard operation to unmap");
458             return;
459         }
460     }
461 }
462 
463 /* Takes the ownership of bs_opts */
464 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
465                                    Error **errp)
466 {
467     const char *buf;
468     int bdrv_flags = 0;
469     int on_read_error, on_write_error;
470     BlockBackend *blk;
471     BlockDriverState *bs;
472     ThrottleConfig cfg;
473     int snapshot = 0;
474     Error *error = NULL;
475     QemuOpts *opts;
476     const char *id;
477     bool has_driver_specific_opts;
478     BlockdevDetectZeroesOptions detect_zeroes =
479         BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
480     const char *throttling_group = NULL;
481 
482     /* Check common options by copying from bs_opts to opts, all other options
483      * stay in bs_opts for processing by bdrv_open(). */
484     id = qdict_get_try_str(bs_opts, "id");
485     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
486     if (error) {
487         error_propagate(errp, error);
488         goto err_no_opts;
489     }
490 
491     qemu_opts_absorb_qdict(opts, bs_opts, &error);
492     if (error) {
493         error_propagate(errp, error);
494         goto early_err;
495     }
496 
497     if (id) {
498         qdict_del(bs_opts, "id");
499     }
500 
501     has_driver_specific_opts = !!qdict_size(bs_opts);
502 
503     /* extract parameters */
504     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
505 
506     extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
507                                     &detect_zeroes, &error);
508     if (error) {
509         error_propagate(errp, error);
510         goto early_err;
511     }
512 
513     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
514         if (is_help_option(buf)) {
515             error_printf("Supported formats:");
516             bdrv_iterate_format(bdrv_format_print, NULL);
517             error_printf("\n");
518             goto early_err;
519         }
520 
521         if (qdict_haskey(bs_opts, "driver")) {
522             error_setg(errp, "Cannot specify both 'driver' and 'format'");
523             goto early_err;
524         }
525         qdict_put(bs_opts, "driver", qstring_from_str(buf));
526     }
527 
528     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
529     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
530         on_write_error = parse_block_error_action(buf, 0, &error);
531         if (error) {
532             error_propagate(errp, error);
533             goto early_err;
534         }
535     }
536 
537     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
538     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
539         on_read_error = parse_block_error_action(buf, 1, &error);
540         if (error) {
541             error_propagate(errp, error);
542             goto early_err;
543         }
544     }
545 
546     if (snapshot) {
547         /* always use cache=unsafe with snapshot */
548         bdrv_flags &= ~BDRV_O_CACHE_MASK;
549         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
550     }
551 
552     /* init */
553     if ((!file || !*file) && !has_driver_specific_opts) {
554         BlockBackendRootState *blk_rs;
555 
556         blk = blk_new(qemu_opts_id(opts), errp);
557         if (!blk) {
558             goto early_err;
559         }
560 
561         blk_rs = blk_get_root_state(blk);
562         blk_rs->open_flags    = bdrv_flags;
563         blk_rs->read_only     = !(bdrv_flags & BDRV_O_RDWR);
564         blk_rs->detect_zeroes = detect_zeroes;
565 
566         if (throttle_enabled(&cfg)) {
567             if (!throttling_group) {
568                 throttling_group = blk_name(blk);
569             }
570             blk_rs->throttle_group = g_strdup(throttling_group);
571             blk_rs->throttle_state = throttle_group_incref(throttling_group);
572             blk_rs->throttle_state->cfg = cfg;
573         }
574 
575         QDECREF(bs_opts);
576     } else {
577         if (file && !*file) {
578             file = NULL;
579         }
580 
581         blk = blk_new_open(qemu_opts_id(opts), file, NULL, bs_opts, bdrv_flags,
582                            errp);
583         if (!blk) {
584             goto err_no_bs_opts;
585         }
586         bs = blk_bs(blk);
587 
588         bs->detect_zeroes = detect_zeroes;
589 
590         /* disk I/O throttling */
591         if (throttle_enabled(&cfg)) {
592             if (!throttling_group) {
593                 throttling_group = blk_name(blk);
594             }
595             bdrv_io_limits_enable(bs, throttling_group);
596             bdrv_set_io_limits(bs, &cfg);
597         }
598 
599         if (bdrv_key_required(bs)) {
600             autostart = 0;
601         }
602     }
603 
604     blk_set_on_error(blk, on_read_error, on_write_error);
605 
606 err_no_bs_opts:
607     qemu_opts_del(opts);
608     return blk;
609 
610 early_err:
611     qemu_opts_del(opts);
612 err_no_opts:
613     QDECREF(bs_opts);
614     return NULL;
615 }
616 
617 static QemuOptsList qemu_root_bds_opts;
618 
619 /* Takes the ownership of bs_opts */
620 static BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
621 {
622     BlockDriverState *bs;
623     QemuOpts *opts;
624     Error *local_error = NULL;
625     BlockdevDetectZeroesOptions detect_zeroes;
626     int ret;
627     int bdrv_flags = 0;
628 
629     opts = qemu_opts_create(&qemu_root_bds_opts, NULL, 1, errp);
630     if (!opts) {
631         goto fail;
632     }
633 
634     qemu_opts_absorb_qdict(opts, bs_opts, &local_error);
635     if (local_error) {
636         error_propagate(errp, local_error);
637         goto fail;
638     }
639 
640     extract_common_blockdev_options(opts, &bdrv_flags, NULL, NULL,
641                                     &detect_zeroes, &local_error);
642     if (local_error) {
643         error_propagate(errp, local_error);
644         goto fail;
645     }
646 
647     bs = NULL;
648     ret = bdrv_open(&bs, NULL, NULL, bs_opts, bdrv_flags, errp);
649     if (ret < 0) {
650         goto fail_no_bs_opts;
651     }
652 
653     bs->detect_zeroes = detect_zeroes;
654 
655 fail_no_bs_opts:
656     qemu_opts_del(opts);
657     return bs;
658 
659 fail:
660     qemu_opts_del(opts);
661     QDECREF(bs_opts);
662     return NULL;
663 }
664 
665 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
666                             Error **errp)
667 {
668     const char *value;
669 
670     value = qemu_opt_get(opts, from);
671     if (value) {
672         if (qemu_opt_find(opts, to)) {
673             error_setg(errp, "'%s' and its alias '%s' can't be used at the "
674                        "same time", to, from);
675             return;
676         }
677     }
678 
679     /* rename all items in opts */
680     while ((value = qemu_opt_get(opts, from))) {
681         qemu_opt_set(opts, to, value, &error_abort);
682         qemu_opt_unset(opts, from);
683     }
684 }
685 
686 QemuOptsList qemu_legacy_drive_opts = {
687     .name = "drive",
688     .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
689     .desc = {
690         {
691             .name = "bus",
692             .type = QEMU_OPT_NUMBER,
693             .help = "bus number",
694         },{
695             .name = "unit",
696             .type = QEMU_OPT_NUMBER,
697             .help = "unit number (i.e. lun for scsi)",
698         },{
699             .name = "index",
700             .type = QEMU_OPT_NUMBER,
701             .help = "index number",
702         },{
703             .name = "media",
704             .type = QEMU_OPT_STRING,
705             .help = "media type (disk, cdrom)",
706         },{
707             .name = "if",
708             .type = QEMU_OPT_STRING,
709             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
710         },{
711             .name = "cyls",
712             .type = QEMU_OPT_NUMBER,
713             .help = "number of cylinders (ide disk geometry)",
714         },{
715             .name = "heads",
716             .type = QEMU_OPT_NUMBER,
717             .help = "number of heads (ide disk geometry)",
718         },{
719             .name = "secs",
720             .type = QEMU_OPT_NUMBER,
721             .help = "number of sectors (ide disk geometry)",
722         },{
723             .name = "trans",
724             .type = QEMU_OPT_STRING,
725             .help = "chs translation (auto, lba, none)",
726         },{
727             .name = "boot",
728             .type = QEMU_OPT_BOOL,
729             .help = "(deprecated, ignored)",
730         },{
731             .name = "addr",
732             .type = QEMU_OPT_STRING,
733             .help = "pci address (virtio only)",
734         },{
735             .name = "serial",
736             .type = QEMU_OPT_STRING,
737             .help = "disk serial number",
738         },{
739             .name = "file",
740             .type = QEMU_OPT_STRING,
741             .help = "file name",
742         },
743 
744         /* Options that are passed on, but have special semantics with -drive */
745         {
746             .name = "read-only",
747             .type = QEMU_OPT_BOOL,
748             .help = "open drive file as read-only",
749         },{
750             .name = "rerror",
751             .type = QEMU_OPT_STRING,
752             .help = "read error action",
753         },{
754             .name = "werror",
755             .type = QEMU_OPT_STRING,
756             .help = "write error action",
757         },{
758             .name = "copy-on-read",
759             .type = QEMU_OPT_BOOL,
760             .help = "copy read data from backing file into image file",
761         },
762 
763         { /* end of list */ }
764     },
765 };
766 
767 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
768 {
769     const char *value;
770     BlockBackend *blk;
771     DriveInfo *dinfo = NULL;
772     QDict *bs_opts;
773     QemuOpts *legacy_opts;
774     DriveMediaType media = MEDIA_DISK;
775     BlockInterfaceType type;
776     int cyls, heads, secs, translation;
777     int max_devs, bus_id, unit_id, index;
778     const char *devaddr;
779     const char *werror, *rerror;
780     bool read_only = false;
781     bool copy_on_read;
782     const char *serial;
783     const char *filename;
784     Error *local_err = NULL;
785     int i;
786 
787     /* Change legacy command line options into QMP ones */
788     static const struct {
789         const char *from;
790         const char *to;
791     } opt_renames[] = {
792         { "iops",           "throttling.iops-total" },
793         { "iops_rd",        "throttling.iops-read" },
794         { "iops_wr",        "throttling.iops-write" },
795 
796         { "bps",            "throttling.bps-total" },
797         { "bps_rd",         "throttling.bps-read" },
798         { "bps_wr",         "throttling.bps-write" },
799 
800         { "iops_max",       "throttling.iops-total-max" },
801         { "iops_rd_max",    "throttling.iops-read-max" },
802         { "iops_wr_max",    "throttling.iops-write-max" },
803 
804         { "bps_max",        "throttling.bps-total-max" },
805         { "bps_rd_max",     "throttling.bps-read-max" },
806         { "bps_wr_max",     "throttling.bps-write-max" },
807 
808         { "iops_size",      "throttling.iops-size" },
809 
810         { "group",          "throttling.group" },
811 
812         { "readonly",       "read-only" },
813     };
814 
815     for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
816         qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
817                         &local_err);
818         if (local_err) {
819             error_report_err(local_err);
820             return NULL;
821         }
822     }
823 
824     value = qemu_opt_get(all_opts, "cache");
825     if (value) {
826         int flags = 0;
827 
828         if (bdrv_parse_cache_flags(value, &flags) != 0) {
829             error_report("invalid cache option");
830             return NULL;
831         }
832 
833         /* Specific options take precedence */
834         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
835             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
836                               !!(flags & BDRV_O_CACHE_WB), &error_abort);
837         }
838         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
839             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
840                               !!(flags & BDRV_O_NOCACHE), &error_abort);
841         }
842         if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
843             qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
844                               !!(flags & BDRV_O_NO_FLUSH), &error_abort);
845         }
846         qemu_opt_unset(all_opts, "cache");
847     }
848 
849     /* Get a QDict for processing the options */
850     bs_opts = qdict_new();
851     qemu_opts_to_qdict(all_opts, bs_opts);
852 
853     legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
854                                    &error_abort);
855     qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
856     if (local_err) {
857         error_report_err(local_err);
858         goto fail;
859     }
860 
861     /* Deprecated option boot=[on|off] */
862     if (qemu_opt_get(legacy_opts, "boot") != NULL) {
863         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
864                 "ignored. Future versions will reject this parameter. Please "
865                 "update your scripts.\n");
866     }
867 
868     /* Media type */
869     value = qemu_opt_get(legacy_opts, "media");
870     if (value) {
871         if (!strcmp(value, "disk")) {
872             media = MEDIA_DISK;
873         } else if (!strcmp(value, "cdrom")) {
874             media = MEDIA_CDROM;
875             read_only = true;
876         } else {
877             error_report("'%s' invalid media", value);
878             goto fail;
879         }
880     }
881 
882     /* copy-on-read is disabled with a warning for read-only devices */
883     read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
884     copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
885 
886     if (read_only && copy_on_read) {
887         error_report("warning: disabling copy-on-read on read-only drive");
888         copy_on_read = false;
889     }
890 
891     qdict_put(bs_opts, "read-only",
892               qstring_from_str(read_only ? "on" : "off"));
893     qdict_put(bs_opts, "copy-on-read",
894               qstring_from_str(copy_on_read ? "on" :"off"));
895 
896     /* Controller type */
897     value = qemu_opt_get(legacy_opts, "if");
898     if (value) {
899         for (type = 0;
900              type < IF_COUNT && strcmp(value, if_name[type]);
901              type++) {
902         }
903         if (type == IF_COUNT) {
904             error_report("unsupported bus type '%s'", value);
905             goto fail;
906         }
907     } else {
908         type = block_default_type;
909     }
910 
911     /* Geometry */
912     cyls  = qemu_opt_get_number(legacy_opts, "cyls", 0);
913     heads = qemu_opt_get_number(legacy_opts, "heads", 0);
914     secs  = qemu_opt_get_number(legacy_opts, "secs", 0);
915 
916     if (cyls || heads || secs) {
917         if (cyls < 1) {
918             error_report("invalid physical cyls number");
919             goto fail;
920         }
921         if (heads < 1) {
922             error_report("invalid physical heads number");
923             goto fail;
924         }
925         if (secs < 1) {
926             error_report("invalid physical secs number");
927             goto fail;
928         }
929     }
930 
931     translation = BIOS_ATA_TRANSLATION_AUTO;
932     value = qemu_opt_get(legacy_opts, "trans");
933     if (value != NULL) {
934         if (!cyls) {
935             error_report("'%s' trans must be used with cyls, heads and secs",
936                          value);
937             goto fail;
938         }
939         if (!strcmp(value, "none")) {
940             translation = BIOS_ATA_TRANSLATION_NONE;
941         } else if (!strcmp(value, "lba")) {
942             translation = BIOS_ATA_TRANSLATION_LBA;
943         } else if (!strcmp(value, "large")) {
944             translation = BIOS_ATA_TRANSLATION_LARGE;
945         } else if (!strcmp(value, "rechs")) {
946             translation = BIOS_ATA_TRANSLATION_RECHS;
947         } else if (!strcmp(value, "auto")) {
948             translation = BIOS_ATA_TRANSLATION_AUTO;
949         } else {
950             error_report("'%s' invalid translation type", value);
951             goto fail;
952         }
953     }
954 
955     if (media == MEDIA_CDROM) {
956         if (cyls || secs || heads) {
957             error_report("CHS can't be set with media=cdrom");
958             goto fail;
959         }
960     }
961 
962     /* Device address specified by bus/unit or index.
963      * If none was specified, try to find the first free one. */
964     bus_id  = qemu_opt_get_number(legacy_opts, "bus", 0);
965     unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
966     index   = qemu_opt_get_number(legacy_opts, "index", -1);
967 
968     max_devs = if_max_devs[type];
969 
970     if (index != -1) {
971         if (bus_id != 0 || unit_id != -1) {
972             error_report("index cannot be used with bus and unit");
973             goto fail;
974         }
975         bus_id = drive_index_to_bus_id(type, index);
976         unit_id = drive_index_to_unit_id(type, index);
977     }
978 
979     if (unit_id == -1) {
980        unit_id = 0;
981        while (drive_get(type, bus_id, unit_id) != NULL) {
982            unit_id++;
983            if (max_devs && unit_id >= max_devs) {
984                unit_id -= max_devs;
985                bus_id++;
986            }
987        }
988     }
989 
990     if (max_devs && unit_id >= max_devs) {
991         error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
992         goto fail;
993     }
994 
995     if (drive_get(type, bus_id, unit_id) != NULL) {
996         error_report("drive with bus=%d, unit=%d (index=%d) exists",
997                      bus_id, unit_id, index);
998         goto fail;
999     }
1000 
1001     /* Serial number */
1002     serial = qemu_opt_get(legacy_opts, "serial");
1003 
1004     /* no id supplied -> create one */
1005     if (qemu_opts_id(all_opts) == NULL) {
1006         char *new_id;
1007         const char *mediastr = "";
1008         if (type == IF_IDE || type == IF_SCSI) {
1009             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
1010         }
1011         if (max_devs) {
1012             new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
1013                                      mediastr, unit_id);
1014         } else {
1015             new_id = g_strdup_printf("%s%s%i", if_name[type],
1016                                      mediastr, unit_id);
1017         }
1018         qdict_put(bs_opts, "id", qstring_from_str(new_id));
1019         g_free(new_id);
1020     }
1021 
1022     /* Add virtio block device */
1023     devaddr = qemu_opt_get(legacy_opts, "addr");
1024     if (devaddr && type != IF_VIRTIO) {
1025         error_report("addr is not supported by this bus type");
1026         goto fail;
1027     }
1028 
1029     if (type == IF_VIRTIO) {
1030         QemuOpts *devopts;
1031         devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
1032                                    &error_abort);
1033         if (arch_type == QEMU_ARCH_S390X) {
1034             qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
1035         } else {
1036             qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
1037         }
1038         qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
1039                      &error_abort);
1040         if (devaddr) {
1041             qemu_opt_set(devopts, "addr", devaddr, &error_abort);
1042         }
1043     }
1044 
1045     filename = qemu_opt_get(legacy_opts, "file");
1046 
1047     /* Check werror/rerror compatibility with if=... */
1048     werror = qemu_opt_get(legacy_opts, "werror");
1049     if (werror != NULL) {
1050         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
1051             type != IF_NONE) {
1052             error_report("werror is not supported by this bus type");
1053             goto fail;
1054         }
1055         qdict_put(bs_opts, "werror", qstring_from_str(werror));
1056     }
1057 
1058     rerror = qemu_opt_get(legacy_opts, "rerror");
1059     if (rerror != NULL) {
1060         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
1061             type != IF_NONE) {
1062             error_report("rerror is not supported by this bus type");
1063             goto fail;
1064         }
1065         qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
1066     }
1067 
1068     /* Actual block device init: Functionality shared with blockdev-add */
1069     blk = blockdev_init(filename, bs_opts, &local_err);
1070     bs_opts = NULL;
1071     if (!blk) {
1072         if (local_err) {
1073             error_report_err(local_err);
1074         }
1075         goto fail;
1076     } else {
1077         assert(!local_err);
1078     }
1079 
1080     /* Create legacy DriveInfo */
1081     dinfo = g_malloc0(sizeof(*dinfo));
1082     dinfo->opts = all_opts;
1083 
1084     dinfo->cyls = cyls;
1085     dinfo->heads = heads;
1086     dinfo->secs = secs;
1087     dinfo->trans = translation;
1088 
1089     dinfo->type = type;
1090     dinfo->bus = bus_id;
1091     dinfo->unit = unit_id;
1092     dinfo->devaddr = devaddr;
1093     dinfo->serial = g_strdup(serial);
1094 
1095     blk_set_legacy_dinfo(blk, dinfo);
1096 
1097     switch(type) {
1098     case IF_IDE:
1099     case IF_SCSI:
1100     case IF_XEN:
1101     case IF_NONE:
1102         dinfo->media_cd = media == MEDIA_CDROM;
1103         break;
1104     default:
1105         break;
1106     }
1107 
1108 fail:
1109     qemu_opts_del(legacy_opts);
1110     QDECREF(bs_opts);
1111     return dinfo;
1112 }
1113 
1114 void hmp_commit(Monitor *mon, const QDict *qdict)
1115 {
1116     const char *device = qdict_get_str(qdict, "device");
1117     BlockBackend *blk;
1118     int ret;
1119 
1120     if (!strcmp(device, "all")) {
1121         ret = bdrv_commit_all();
1122     } else {
1123         blk = blk_by_name(device);
1124         if (!blk) {
1125             monitor_printf(mon, "Device '%s' not found\n", device);
1126             return;
1127         }
1128         if (!blk_is_available(blk)) {
1129             monitor_printf(mon, "Device '%s' has no medium\n", device);
1130             return;
1131         }
1132         ret = bdrv_commit(blk_bs(blk));
1133     }
1134     if (ret < 0) {
1135         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1136                        strerror(-ret));
1137     }
1138 }
1139 
1140 static void blockdev_do_action(int kind, void *data, Error **errp)
1141 {
1142     TransactionAction action;
1143     TransactionActionList list;
1144 
1145     action.kind = kind;
1146     action.data = data;
1147     list.value = &action;
1148     list.next = NULL;
1149     qmp_transaction(&list, errp);
1150 }
1151 
1152 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1153                                 bool has_node_name, const char *node_name,
1154                                 const char *snapshot_file,
1155                                 bool has_snapshot_node_name,
1156                                 const char *snapshot_node_name,
1157                                 bool has_format, const char *format,
1158                                 bool has_mode, NewImageMode mode, Error **errp)
1159 {
1160     BlockdevSnapshot snapshot = {
1161         .has_device = has_device,
1162         .device = (char *) device,
1163         .has_node_name = has_node_name,
1164         .node_name = (char *) node_name,
1165         .snapshot_file = (char *) snapshot_file,
1166         .has_snapshot_node_name = has_snapshot_node_name,
1167         .snapshot_node_name = (char *) snapshot_node_name,
1168         .has_format = has_format,
1169         .format = (char *) format,
1170         .has_mode = has_mode,
1171         .mode = mode,
1172     };
1173     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1174                        &snapshot, errp);
1175 }
1176 
1177 void qmp_blockdev_snapshot_internal_sync(const char *device,
1178                                          const char *name,
1179                                          Error **errp)
1180 {
1181     BlockdevSnapshotInternal snapshot = {
1182         .device = (char *) device,
1183         .name = (char *) name
1184     };
1185 
1186     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1187                        &snapshot, errp);
1188 }
1189 
1190 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1191                                                          bool has_id,
1192                                                          const char *id,
1193                                                          bool has_name,
1194                                                          const char *name,
1195                                                          Error **errp)
1196 {
1197     BlockDriverState *bs;
1198     BlockBackend *blk;
1199     AioContext *aio_context;
1200     QEMUSnapshotInfo sn;
1201     Error *local_err = NULL;
1202     SnapshotInfo *info = NULL;
1203     int ret;
1204 
1205     blk = blk_by_name(device);
1206     if (!blk) {
1207         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1208                   "Device '%s' not found", device);
1209         return NULL;
1210     }
1211 
1212     aio_context = blk_get_aio_context(blk);
1213     aio_context_acquire(aio_context);
1214 
1215     if (!has_id) {
1216         id = NULL;
1217     }
1218 
1219     if (!has_name) {
1220         name = NULL;
1221     }
1222 
1223     if (!id && !name) {
1224         error_setg(errp, "Name or id must be provided");
1225         goto out_aio_context;
1226     }
1227 
1228     if (!blk_is_available(blk)) {
1229         error_setg(errp, "Device '%s' has no medium", device);
1230         goto out_aio_context;
1231     }
1232     bs = blk_bs(blk);
1233 
1234     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1235         goto out_aio_context;
1236     }
1237 
1238     ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1239     if (local_err) {
1240         error_propagate(errp, local_err);
1241         goto out_aio_context;
1242     }
1243     if (!ret) {
1244         error_setg(errp,
1245                    "Snapshot with id '%s' and name '%s' does not exist on "
1246                    "device '%s'",
1247                    STR_OR_NULL(id), STR_OR_NULL(name), device);
1248         goto out_aio_context;
1249     }
1250 
1251     bdrv_snapshot_delete(bs, id, name, &local_err);
1252     if (local_err) {
1253         error_propagate(errp, local_err);
1254         goto out_aio_context;
1255     }
1256 
1257     aio_context_release(aio_context);
1258 
1259     info = g_new0(SnapshotInfo, 1);
1260     info->id = g_strdup(sn.id_str);
1261     info->name = g_strdup(sn.name);
1262     info->date_nsec = sn.date_nsec;
1263     info->date_sec = sn.date_sec;
1264     info->vm_state_size = sn.vm_state_size;
1265     info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1266     info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1267 
1268     return info;
1269 
1270 out_aio_context:
1271     aio_context_release(aio_context);
1272     return NULL;
1273 }
1274 
1275 /**
1276  * block_dirty_bitmap_lookup:
1277  * Return a dirty bitmap (if present), after validating
1278  * the node reference and bitmap names.
1279  *
1280  * @node: The name of the BDS node to search for bitmaps
1281  * @name: The name of the bitmap to search for
1282  * @pbs: Output pointer for BDS lookup, if desired. Can be NULL.
1283  * @paio: Output pointer for aio_context acquisition, if desired. Can be NULL.
1284  * @errp: Output pointer for error information. Can be NULL.
1285  *
1286  * @return: A bitmap object on success, or NULL on failure.
1287  */
1288 static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
1289                                                   const char *name,
1290                                                   BlockDriverState **pbs,
1291                                                   AioContext **paio,
1292                                                   Error **errp)
1293 {
1294     BlockDriverState *bs;
1295     BdrvDirtyBitmap *bitmap;
1296     AioContext *aio_context;
1297 
1298     if (!node) {
1299         error_setg(errp, "Node cannot be NULL");
1300         return NULL;
1301     }
1302     if (!name) {
1303         error_setg(errp, "Bitmap name cannot be NULL");
1304         return NULL;
1305     }
1306     bs = bdrv_lookup_bs(node, node, NULL);
1307     if (!bs) {
1308         error_setg(errp, "Node '%s' not found", node);
1309         return NULL;
1310     }
1311 
1312     aio_context = bdrv_get_aio_context(bs);
1313     aio_context_acquire(aio_context);
1314 
1315     bitmap = bdrv_find_dirty_bitmap(bs, name);
1316     if (!bitmap) {
1317         error_setg(errp, "Dirty bitmap '%s' not found", name);
1318         goto fail;
1319     }
1320 
1321     if (pbs) {
1322         *pbs = bs;
1323     }
1324     if (paio) {
1325         *paio = aio_context;
1326     } else {
1327         aio_context_release(aio_context);
1328     }
1329 
1330     return bitmap;
1331 
1332  fail:
1333     aio_context_release(aio_context);
1334     return NULL;
1335 }
1336 
1337 /* New and old BlockDriverState structs for atomic group operations */
1338 
1339 typedef struct BlkTransactionState BlkTransactionState;
1340 
1341 /* Only prepare() may fail. In a single transaction, only one of commit() or
1342    abort() will be called, clean() will always be called if it present. */
1343 typedef struct BdrvActionOps {
1344     /* Size of state struct, in bytes. */
1345     size_t instance_size;
1346     /* Prepare the work, must NOT be NULL. */
1347     void (*prepare)(BlkTransactionState *common, Error **errp);
1348     /* Commit the changes, can be NULL. */
1349     void (*commit)(BlkTransactionState *common);
1350     /* Abort the changes on fail, can be NULL. */
1351     void (*abort)(BlkTransactionState *common);
1352     /* Clean up resource in the end, can be NULL. */
1353     void (*clean)(BlkTransactionState *common);
1354 } BdrvActionOps;
1355 
1356 /*
1357  * This structure must be arranged as first member in child type, assuming
1358  * that compiler will also arrange it to the same address with parent instance.
1359  * Later it will be used in free().
1360  */
1361 struct BlkTransactionState {
1362     TransactionAction *action;
1363     const BdrvActionOps *ops;
1364     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1365 };
1366 
1367 /* internal snapshot private data */
1368 typedef struct InternalSnapshotState {
1369     BlkTransactionState common;
1370     BlockDriverState *bs;
1371     AioContext *aio_context;
1372     QEMUSnapshotInfo sn;
1373     bool created;
1374 } InternalSnapshotState;
1375 
1376 static void internal_snapshot_prepare(BlkTransactionState *common,
1377                                       Error **errp)
1378 {
1379     Error *local_err = NULL;
1380     const char *device;
1381     const char *name;
1382     BlockBackend *blk;
1383     BlockDriverState *bs;
1384     QEMUSnapshotInfo old_sn, *sn;
1385     bool ret;
1386     qemu_timeval tv;
1387     BlockdevSnapshotInternal *internal;
1388     InternalSnapshotState *state;
1389     int ret1;
1390 
1391     g_assert(common->action->kind ==
1392              TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1393     internal = common->action->blockdev_snapshot_internal_sync;
1394     state = DO_UPCAST(InternalSnapshotState, common, common);
1395 
1396     /* 1. parse input */
1397     device = internal->device;
1398     name = internal->name;
1399 
1400     /* 2. check for validation */
1401     blk = blk_by_name(device);
1402     if (!blk) {
1403         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1404                   "Device '%s' not found", device);
1405         return;
1406     }
1407 
1408     /* AioContext is released in .clean() */
1409     state->aio_context = blk_get_aio_context(blk);
1410     aio_context_acquire(state->aio_context);
1411 
1412     if (!blk_is_available(blk)) {
1413         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1414         return;
1415     }
1416     bs = blk_bs(blk);
1417 
1418     state->bs = bs;
1419     bdrv_drained_begin(bs);
1420 
1421     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1422         return;
1423     }
1424 
1425     if (bdrv_is_read_only(bs)) {
1426         error_setg(errp, "Device '%s' is read only", device);
1427         return;
1428     }
1429 
1430     if (!bdrv_can_snapshot(bs)) {
1431         error_setg(errp, "Block format '%s' used by device '%s' "
1432                    "does not support internal snapshots",
1433                    bs->drv->format_name, device);
1434         return;
1435     }
1436 
1437     if (!strlen(name)) {
1438         error_setg(errp, "Name is empty");
1439         return;
1440     }
1441 
1442     /* check whether a snapshot with name exist */
1443     ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1444                                             &local_err);
1445     if (local_err) {
1446         error_propagate(errp, local_err);
1447         return;
1448     } else if (ret) {
1449         error_setg(errp,
1450                    "Snapshot with name '%s' already exists on device '%s'",
1451                    name, device);
1452         return;
1453     }
1454 
1455     /* 3. take the snapshot */
1456     sn = &state->sn;
1457     pstrcpy(sn->name, sizeof(sn->name), name);
1458     qemu_gettimeofday(&tv);
1459     sn->date_sec = tv.tv_sec;
1460     sn->date_nsec = tv.tv_usec * 1000;
1461     sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1462 
1463     ret1 = bdrv_snapshot_create(bs, sn);
1464     if (ret1 < 0) {
1465         error_setg_errno(errp, -ret1,
1466                          "Failed to create snapshot '%s' on device '%s'",
1467                          name, device);
1468         return;
1469     }
1470 
1471     /* 4. succeed, mark a snapshot is created */
1472     state->created = true;
1473 }
1474 
1475 static void internal_snapshot_abort(BlkTransactionState *common)
1476 {
1477     InternalSnapshotState *state =
1478                              DO_UPCAST(InternalSnapshotState, common, common);
1479     BlockDriverState *bs = state->bs;
1480     QEMUSnapshotInfo *sn = &state->sn;
1481     Error *local_error = NULL;
1482 
1483     if (!state->created) {
1484         return;
1485     }
1486 
1487     if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1488         error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1489                      "device '%s' in abort: %s",
1490                      sn->id_str,
1491                      sn->name,
1492                      bdrv_get_device_name(bs),
1493                      error_get_pretty(local_error));
1494         error_free(local_error);
1495     }
1496 }
1497 
1498 static void internal_snapshot_clean(BlkTransactionState *common)
1499 {
1500     InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1501                                              common, common);
1502 
1503     if (state->aio_context) {
1504         if (state->bs) {
1505             bdrv_drained_end(state->bs);
1506         }
1507         aio_context_release(state->aio_context);
1508     }
1509 }
1510 
1511 /* external snapshot private data */
1512 typedef struct ExternalSnapshotState {
1513     BlkTransactionState common;
1514     BlockDriverState *old_bs;
1515     BlockDriverState *new_bs;
1516     AioContext *aio_context;
1517 } ExternalSnapshotState;
1518 
1519 static void external_snapshot_prepare(BlkTransactionState *common,
1520                                       Error **errp)
1521 {
1522     int flags, ret;
1523     QDict *options;
1524     Error *local_err = NULL;
1525     bool has_device = false;
1526     const char *device;
1527     bool has_node_name = false;
1528     const char *node_name;
1529     bool has_snapshot_node_name = false;
1530     const char *snapshot_node_name;
1531     const char *new_image_file;
1532     const char *format = "qcow2";
1533     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1534     ExternalSnapshotState *state =
1535                              DO_UPCAST(ExternalSnapshotState, common, common);
1536     TransactionAction *action = common->action;
1537 
1538     /* get parameters */
1539     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1540 
1541     has_device = action->blockdev_snapshot_sync->has_device;
1542     device = action->blockdev_snapshot_sync->device;
1543     has_node_name = action->blockdev_snapshot_sync->has_node_name;
1544     node_name = action->blockdev_snapshot_sync->node_name;
1545     has_snapshot_node_name =
1546         action->blockdev_snapshot_sync->has_snapshot_node_name;
1547     snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1548 
1549     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1550     if (action->blockdev_snapshot_sync->has_format) {
1551         format = action->blockdev_snapshot_sync->format;
1552     }
1553     if (action->blockdev_snapshot_sync->has_mode) {
1554         mode = action->blockdev_snapshot_sync->mode;
1555     }
1556 
1557     /* start processing */
1558     state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1559                                    has_node_name ? node_name : NULL,
1560                                    &local_err);
1561     if (local_err) {
1562         error_propagate(errp, local_err);
1563         return;
1564     }
1565 
1566     if (has_node_name && !has_snapshot_node_name) {
1567         error_setg(errp, "New snapshot node name missing");
1568         return;
1569     }
1570 
1571     if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1572         error_setg(errp, "New snapshot node name already existing");
1573         return;
1574     }
1575 
1576     /* Acquire AioContext now so any threads operating on old_bs stop */
1577     state->aio_context = bdrv_get_aio_context(state->old_bs);
1578     aio_context_acquire(state->aio_context);
1579     bdrv_drained_begin(state->old_bs);
1580 
1581     if (!bdrv_is_inserted(state->old_bs)) {
1582         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1583         return;
1584     }
1585 
1586     if (bdrv_op_is_blocked(state->old_bs,
1587                            BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1588         return;
1589     }
1590 
1591     if (!bdrv_is_read_only(state->old_bs)) {
1592         if (bdrv_flush(state->old_bs)) {
1593             error_setg(errp, QERR_IO_ERROR);
1594             return;
1595         }
1596     }
1597 
1598     if (!bdrv_is_first_non_filter(state->old_bs)) {
1599         error_setg(errp, QERR_FEATURE_DISABLED, "snapshot");
1600         return;
1601     }
1602 
1603     flags = state->old_bs->open_flags;
1604 
1605     /* create new image w/backing file */
1606     if (mode != NEW_IMAGE_MODE_EXISTING) {
1607         bdrv_img_create(new_image_file, format,
1608                         state->old_bs->filename,
1609                         state->old_bs->drv->format_name,
1610                         NULL, -1, flags, &local_err, false);
1611         if (local_err) {
1612             error_propagate(errp, local_err);
1613             return;
1614         }
1615     }
1616 
1617     options = qdict_new();
1618     if (has_snapshot_node_name) {
1619         qdict_put(options, "node-name",
1620                   qstring_from_str(snapshot_node_name));
1621     }
1622     qdict_put(options, "driver", qstring_from_str(format));
1623 
1624     /* TODO Inherit bs->options or only take explicit options with an
1625      * extended QMP command? */
1626     assert(state->new_bs == NULL);
1627     ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1628                     flags | BDRV_O_NO_BACKING, &local_err);
1629     /* We will manually add the backing_hd field to the bs later */
1630     if (ret != 0) {
1631         error_propagate(errp, local_err);
1632     }
1633 }
1634 
1635 static void external_snapshot_commit(BlkTransactionState *common)
1636 {
1637     ExternalSnapshotState *state =
1638                              DO_UPCAST(ExternalSnapshotState, common, common);
1639 
1640     bdrv_set_aio_context(state->new_bs, state->aio_context);
1641 
1642     /* This removes our old bs and adds the new bs */
1643     bdrv_append(state->new_bs, state->old_bs);
1644     /* We don't need (or want) to use the transactional
1645      * bdrv_reopen_multiple() across all the entries at once, because we
1646      * don't want to abort all of them if one of them fails the reopen */
1647     bdrv_reopen(state->old_bs, state->old_bs->open_flags & ~BDRV_O_RDWR,
1648                 NULL);
1649 }
1650 
1651 static void external_snapshot_abort(BlkTransactionState *common)
1652 {
1653     ExternalSnapshotState *state =
1654                              DO_UPCAST(ExternalSnapshotState, common, common);
1655     if (state->new_bs) {
1656         bdrv_unref(state->new_bs);
1657     }
1658 }
1659 
1660 static void external_snapshot_clean(BlkTransactionState *common)
1661 {
1662     ExternalSnapshotState *state =
1663                              DO_UPCAST(ExternalSnapshotState, common, common);
1664     if (state->aio_context) {
1665         bdrv_drained_end(state->old_bs);
1666         aio_context_release(state->aio_context);
1667     }
1668 }
1669 
1670 typedef struct DriveBackupState {
1671     BlkTransactionState common;
1672     BlockDriverState *bs;
1673     AioContext *aio_context;
1674     BlockJob *job;
1675 } DriveBackupState;
1676 
1677 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1678 {
1679     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1680     BlockBackend *blk;
1681     DriveBackup *backup;
1682     Error *local_err = NULL;
1683 
1684     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1685     backup = common->action->drive_backup;
1686 
1687     blk = blk_by_name(backup->device);
1688     if (!blk) {
1689         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1690                   "Device '%s' not found", backup->device);
1691         return;
1692     }
1693 
1694     if (!blk_is_available(blk)) {
1695         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, backup->device);
1696         return;
1697     }
1698 
1699     /* AioContext is released in .clean() */
1700     state->aio_context = blk_get_aio_context(blk);
1701     aio_context_acquire(state->aio_context);
1702     bdrv_drained_begin(blk_bs(blk));
1703     state->bs = blk_bs(blk);
1704 
1705     qmp_drive_backup(backup->device, backup->target,
1706                      backup->has_format, backup->format,
1707                      backup->sync,
1708                      backup->has_mode, backup->mode,
1709                      backup->has_speed, backup->speed,
1710                      backup->has_bitmap, backup->bitmap,
1711                      backup->has_on_source_error, backup->on_source_error,
1712                      backup->has_on_target_error, backup->on_target_error,
1713                      &local_err);
1714     if (local_err) {
1715         error_propagate(errp, local_err);
1716         return;
1717     }
1718 
1719     state->job = state->bs->job;
1720 }
1721 
1722 static void drive_backup_abort(BlkTransactionState *common)
1723 {
1724     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1725     BlockDriverState *bs = state->bs;
1726 
1727     /* Only cancel if it's the job we started */
1728     if (bs && bs->job && bs->job == state->job) {
1729         block_job_cancel_sync(bs->job);
1730     }
1731 }
1732 
1733 static void drive_backup_clean(BlkTransactionState *common)
1734 {
1735     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1736 
1737     if (state->aio_context) {
1738         bdrv_drained_end(state->bs);
1739         aio_context_release(state->aio_context);
1740     }
1741 }
1742 
1743 typedef struct BlockdevBackupState {
1744     BlkTransactionState common;
1745     BlockDriverState *bs;
1746     BlockJob *job;
1747     AioContext *aio_context;
1748 } BlockdevBackupState;
1749 
1750 static void blockdev_backup_prepare(BlkTransactionState *common, Error **errp)
1751 {
1752     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1753     BlockdevBackup *backup;
1754     BlockBackend *blk, *target;
1755     Error *local_err = NULL;
1756 
1757     assert(common->action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1758     backup = common->action->blockdev_backup;
1759 
1760     blk = blk_by_name(backup->device);
1761     if (!blk) {
1762         error_setg(errp, "Device '%s' not found", backup->device);
1763         return;
1764     }
1765 
1766     if (!blk_is_available(blk)) {
1767         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, backup->device);
1768         return;
1769     }
1770 
1771     target = blk_by_name(backup->target);
1772     if (!target) {
1773         error_setg(errp, "Device '%s' not found", backup->target);
1774         return;
1775     }
1776 
1777     /* AioContext is released in .clean() */
1778     state->aio_context = blk_get_aio_context(blk);
1779     if (state->aio_context != blk_get_aio_context(target)) {
1780         state->aio_context = NULL;
1781         error_setg(errp, "Backup between two IO threads is not implemented");
1782         return;
1783     }
1784     aio_context_acquire(state->aio_context);
1785     state->bs = blk_bs(blk);
1786     bdrv_drained_begin(state->bs);
1787 
1788     qmp_blockdev_backup(backup->device, backup->target,
1789                         backup->sync,
1790                         backup->has_speed, backup->speed,
1791                         backup->has_on_source_error, backup->on_source_error,
1792                         backup->has_on_target_error, backup->on_target_error,
1793                         &local_err);
1794     if (local_err) {
1795         error_propagate(errp, local_err);
1796         return;
1797     }
1798 
1799     state->job = state->bs->job;
1800 }
1801 
1802 static void blockdev_backup_abort(BlkTransactionState *common)
1803 {
1804     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1805     BlockDriverState *bs = state->bs;
1806 
1807     /* Only cancel if it's the job we started */
1808     if (bs && bs->job && bs->job == state->job) {
1809         block_job_cancel_sync(bs->job);
1810     }
1811 }
1812 
1813 static void blockdev_backup_clean(BlkTransactionState *common)
1814 {
1815     BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1816 
1817     if (state->aio_context) {
1818         bdrv_drained_end(state->bs);
1819         aio_context_release(state->aio_context);
1820     }
1821 }
1822 
1823 static void abort_prepare(BlkTransactionState *common, Error **errp)
1824 {
1825     error_setg(errp, "Transaction aborted using Abort action");
1826 }
1827 
1828 static void abort_commit(BlkTransactionState *common)
1829 {
1830     g_assert_not_reached(); /* this action never succeeds */
1831 }
1832 
1833 static const BdrvActionOps actions[] = {
1834     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1835         .instance_size = sizeof(ExternalSnapshotState),
1836         .prepare  = external_snapshot_prepare,
1837         .commit   = external_snapshot_commit,
1838         .abort = external_snapshot_abort,
1839         .clean = external_snapshot_clean,
1840     },
1841     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1842         .instance_size = sizeof(DriveBackupState),
1843         .prepare = drive_backup_prepare,
1844         .abort = drive_backup_abort,
1845         .clean = drive_backup_clean,
1846     },
1847     [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
1848         .instance_size = sizeof(BlockdevBackupState),
1849         .prepare = blockdev_backup_prepare,
1850         .abort = blockdev_backup_abort,
1851         .clean = blockdev_backup_clean,
1852     },
1853     [TRANSACTION_ACTION_KIND_ABORT] = {
1854         .instance_size = sizeof(BlkTransactionState),
1855         .prepare = abort_prepare,
1856         .commit = abort_commit,
1857     },
1858     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1859         .instance_size = sizeof(InternalSnapshotState),
1860         .prepare  = internal_snapshot_prepare,
1861         .abort = internal_snapshot_abort,
1862         .clean = internal_snapshot_clean,
1863     },
1864 };
1865 
1866 /*
1867  * 'Atomic' group operations.  The operations are performed as a set, and if
1868  * any fail then we roll back all operations in the group.
1869  */
1870 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1871 {
1872     TransactionActionList *dev_entry = dev_list;
1873     BlkTransactionState *state, *next;
1874     Error *local_err = NULL;
1875 
1876     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1877     QSIMPLEQ_INIT(&snap_bdrv_states);
1878 
1879     /* drain all i/o before any operations */
1880     bdrv_drain_all();
1881 
1882     /* We don't do anything in this loop that commits us to the operations */
1883     while (NULL != dev_entry) {
1884         TransactionAction *dev_info = NULL;
1885         const BdrvActionOps *ops;
1886 
1887         dev_info = dev_entry->value;
1888         dev_entry = dev_entry->next;
1889 
1890         assert(dev_info->kind < ARRAY_SIZE(actions));
1891 
1892         ops = &actions[dev_info->kind];
1893         assert(ops->instance_size > 0);
1894 
1895         state = g_malloc0(ops->instance_size);
1896         state->ops = ops;
1897         state->action = dev_info;
1898         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1899 
1900         state->ops->prepare(state, &local_err);
1901         if (local_err) {
1902             error_propagate(errp, local_err);
1903             goto delete_and_fail;
1904         }
1905     }
1906 
1907     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1908         if (state->ops->commit) {
1909             state->ops->commit(state);
1910         }
1911     }
1912 
1913     /* success */
1914     goto exit;
1915 
1916 delete_and_fail:
1917     /* failure, and it is all-or-none; roll back all operations */
1918     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1919         if (state->ops->abort) {
1920             state->ops->abort(state);
1921         }
1922     }
1923 exit:
1924     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1925         if (state->ops->clean) {
1926             state->ops->clean(state);
1927         }
1928         g_free(state);
1929     }
1930 }
1931 
1932 
1933 static void eject_device(BlockBackend *blk, int force, Error **errp)
1934 {
1935     BlockDriverState *bs = blk_bs(blk);
1936     AioContext *aio_context;
1937 
1938     if (!bs) {
1939         /* No medium inserted, so there is nothing to do */
1940         return;
1941     }
1942 
1943     aio_context = bdrv_get_aio_context(bs);
1944     aio_context_acquire(aio_context);
1945 
1946     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1947         goto out;
1948     }
1949     if (!blk_dev_has_removable_media(blk)) {
1950         error_setg(errp, "Device '%s' is not removable",
1951                    bdrv_get_device_name(bs));
1952         goto out;
1953     }
1954 
1955     if (blk_dev_is_medium_locked(blk) && !blk_dev_is_tray_open(blk)) {
1956         blk_dev_eject_request(blk, force);
1957         if (!force) {
1958             error_setg(errp, "Device '%s' is locked",
1959                        bdrv_get_device_name(bs));
1960             goto out;
1961         }
1962     }
1963 
1964     bdrv_close(bs);
1965 
1966 out:
1967     aio_context_release(aio_context);
1968 }
1969 
1970 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1971 {
1972     BlockBackend *blk;
1973 
1974     blk = blk_by_name(device);
1975     if (!blk) {
1976         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
1977                   "Device '%s' not found", device);
1978         return;
1979     }
1980 
1981     eject_device(blk, force, errp);
1982 }
1983 
1984 void qmp_block_passwd(bool has_device, const char *device,
1985                       bool has_node_name, const char *node_name,
1986                       const char *password, Error **errp)
1987 {
1988     Error *local_err = NULL;
1989     BlockDriverState *bs;
1990     AioContext *aio_context;
1991 
1992     bs = bdrv_lookup_bs(has_device ? device : NULL,
1993                         has_node_name ? node_name : NULL,
1994                         &local_err);
1995     if (local_err) {
1996         error_propagate(errp, local_err);
1997         return;
1998     }
1999 
2000     aio_context = bdrv_get_aio_context(bs);
2001     aio_context_acquire(aio_context);
2002 
2003     bdrv_add_key(bs, password, errp);
2004 
2005     aio_context_release(aio_context);
2006 }
2007 
2008 /* Assumes AioContext is held */
2009 static void qmp_bdrv_open_encrypted(BlockDriverState **pbs,
2010                                     const char *filename,
2011                                     int bdrv_flags, const char *format,
2012                                     const char *password, Error **errp)
2013 {
2014     Error *local_err = NULL;
2015     QDict *options = NULL;
2016     int ret;
2017 
2018     if (format) {
2019         options = qdict_new();
2020         qdict_put(options, "driver", qstring_from_str(format));
2021     }
2022 
2023     ret = bdrv_open(pbs, filename, NULL, options, bdrv_flags, &local_err);
2024     if (ret < 0) {
2025         error_propagate(errp, local_err);
2026         return;
2027     }
2028 
2029     bdrv_add_key(*pbs, password, errp);
2030 }
2031 
2032 void qmp_change_blockdev(const char *device, const char *filename,
2033                          const char *format, Error **errp)
2034 {
2035     BlockBackend *blk;
2036     BlockDriverState *bs;
2037     AioContext *aio_context;
2038     int bdrv_flags;
2039     bool new_bs;
2040     Error *err = NULL;
2041 
2042     blk = blk_by_name(device);
2043     if (!blk) {
2044         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2045                   "Device '%s' not found", device);
2046         return;
2047     }
2048     bs = blk_bs(blk);
2049     new_bs = !bs;
2050 
2051     aio_context = blk_get_aio_context(blk);
2052     aio_context_acquire(aio_context);
2053 
2054     eject_device(blk, 0, &err);
2055     if (err) {
2056         error_propagate(errp, err);
2057         goto out;
2058     }
2059 
2060     bdrv_flags = blk_is_read_only(blk) ? 0 : BDRV_O_RDWR;
2061     bdrv_flags |= blk_get_root_state(blk)->open_flags & ~BDRV_O_RDWR;
2062 
2063     qmp_bdrv_open_encrypted(&bs, filename, bdrv_flags, format, NULL, &err);
2064     if (err) {
2065         error_propagate(errp, err);
2066         goto out;
2067     }
2068 
2069     if (new_bs) {
2070         blk_insert_bs(blk, bs);
2071         /* Has been sent automatically by bdrv_open() if blk_bs(blk) was not
2072          * NULL */
2073         blk_dev_change_media_cb(blk, true);
2074     }
2075 
2076 out:
2077     aio_context_release(aio_context);
2078 }
2079 
2080 /* throttling disk I/O limits */
2081 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
2082                                int64_t bps_wr,
2083                                int64_t iops,
2084                                int64_t iops_rd,
2085                                int64_t iops_wr,
2086                                bool has_bps_max,
2087                                int64_t bps_max,
2088                                bool has_bps_rd_max,
2089                                int64_t bps_rd_max,
2090                                bool has_bps_wr_max,
2091                                int64_t bps_wr_max,
2092                                bool has_iops_max,
2093                                int64_t iops_max,
2094                                bool has_iops_rd_max,
2095                                int64_t iops_rd_max,
2096                                bool has_iops_wr_max,
2097                                int64_t iops_wr_max,
2098                                bool has_iops_size,
2099                                int64_t iops_size,
2100                                bool has_group,
2101                                const char *group, Error **errp)
2102 {
2103     ThrottleConfig cfg;
2104     BlockDriverState *bs;
2105     BlockBackend *blk;
2106     AioContext *aio_context;
2107 
2108     blk = blk_by_name(device);
2109     if (!blk) {
2110         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2111                   "Device '%s' not found", device);
2112         return;
2113     }
2114 
2115     aio_context = blk_get_aio_context(blk);
2116     aio_context_acquire(aio_context);
2117 
2118     bs = blk_bs(blk);
2119     if (!bs) {
2120         error_setg(errp, "Device '%s' has no medium", device);
2121         goto out;
2122     }
2123 
2124     memset(&cfg, 0, sizeof(cfg));
2125     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
2126     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
2127     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
2128 
2129     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
2130     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
2131     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
2132 
2133     if (has_bps_max) {
2134         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
2135     }
2136     if (has_bps_rd_max) {
2137         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
2138     }
2139     if (has_bps_wr_max) {
2140         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
2141     }
2142     if (has_iops_max) {
2143         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
2144     }
2145     if (has_iops_rd_max) {
2146         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
2147     }
2148     if (has_iops_wr_max) {
2149         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
2150     }
2151 
2152     if (has_iops_size) {
2153         cfg.op_size = iops_size;
2154     }
2155 
2156     if (!check_throttle_config(&cfg, errp)) {
2157         goto out;
2158     }
2159 
2160     if (throttle_enabled(&cfg)) {
2161         /* Enable I/O limits if they're not enabled yet, otherwise
2162          * just update the throttling group. */
2163         if (!bs->io_limits_enabled) {
2164             bdrv_io_limits_enable(bs, has_group ? group : device);
2165         } else if (has_group) {
2166             bdrv_io_limits_update_group(bs, group);
2167         }
2168         /* Set the new throttling configuration */
2169         bdrv_set_io_limits(bs, &cfg);
2170     } else if (bs->io_limits_enabled) {
2171         /* If all throttling settings are set to 0, disable I/O limits */
2172         bdrv_io_limits_disable(bs);
2173     }
2174 
2175 out:
2176     aio_context_release(aio_context);
2177 }
2178 
2179 void qmp_block_dirty_bitmap_add(const char *node, const char *name,
2180                                 bool has_granularity, uint32_t granularity,
2181                                 Error **errp)
2182 {
2183     AioContext *aio_context;
2184     BlockDriverState *bs;
2185 
2186     if (!name || name[0] == '\0') {
2187         error_setg(errp, "Bitmap name cannot be empty");
2188         return;
2189     }
2190 
2191     bs = bdrv_lookup_bs(node, node, errp);
2192     if (!bs) {
2193         return;
2194     }
2195 
2196     aio_context = bdrv_get_aio_context(bs);
2197     aio_context_acquire(aio_context);
2198 
2199     if (has_granularity) {
2200         if (granularity < 512 || !is_power_of_2(granularity)) {
2201             error_setg(errp, "Granularity must be power of 2 "
2202                              "and at least 512");
2203             goto out;
2204         }
2205     } else {
2206         /* Default to cluster size, if available: */
2207         granularity = bdrv_get_default_bitmap_granularity(bs);
2208     }
2209 
2210     bdrv_create_dirty_bitmap(bs, granularity, name, errp);
2211 
2212  out:
2213     aio_context_release(aio_context);
2214 }
2215 
2216 void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
2217                                    Error **errp)
2218 {
2219     AioContext *aio_context;
2220     BlockDriverState *bs;
2221     BdrvDirtyBitmap *bitmap;
2222 
2223     bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2224     if (!bitmap || !bs) {
2225         return;
2226     }
2227 
2228     if (bdrv_dirty_bitmap_frozen(bitmap)) {
2229         error_setg(errp,
2230                    "Bitmap '%s' is currently frozen and cannot be removed",
2231                    name);
2232         goto out;
2233     }
2234     bdrv_dirty_bitmap_make_anon(bitmap);
2235     bdrv_release_dirty_bitmap(bs, bitmap);
2236 
2237  out:
2238     aio_context_release(aio_context);
2239 }
2240 
2241 /**
2242  * Completely clear a bitmap, for the purposes of synchronizing a bitmap
2243  * immediately after a full backup operation.
2244  */
2245 void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
2246                                   Error **errp)
2247 {
2248     AioContext *aio_context;
2249     BdrvDirtyBitmap *bitmap;
2250     BlockDriverState *bs;
2251 
2252     bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
2253     if (!bitmap || !bs) {
2254         return;
2255     }
2256 
2257     if (bdrv_dirty_bitmap_frozen(bitmap)) {
2258         error_setg(errp,
2259                    "Bitmap '%s' is currently frozen and cannot be modified",
2260                    name);
2261         goto out;
2262     } else if (!bdrv_dirty_bitmap_enabled(bitmap)) {
2263         error_setg(errp,
2264                    "Bitmap '%s' is currently disabled and cannot be cleared",
2265                    name);
2266         goto out;
2267     }
2268 
2269     bdrv_clear_dirty_bitmap(bitmap);
2270 
2271  out:
2272     aio_context_release(aio_context);
2273 }
2274 
2275 void hmp_drive_del(Monitor *mon, const QDict *qdict)
2276 {
2277     const char *id = qdict_get_str(qdict, "id");
2278     BlockBackend *blk;
2279     BlockDriverState *bs;
2280     AioContext *aio_context;
2281     Error *local_err = NULL;
2282 
2283     blk = blk_by_name(id);
2284     if (!blk) {
2285         error_report("Device '%s' not found", id);
2286         return;
2287     }
2288 
2289     if (!blk_legacy_dinfo(blk)) {
2290         error_report("Deleting device added with blockdev-add"
2291                      " is not supported");
2292         return;
2293     }
2294 
2295     aio_context = blk_get_aio_context(blk);
2296     aio_context_acquire(aio_context);
2297 
2298     bs = blk_bs(blk);
2299     if (bs) {
2300         if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
2301             error_report_err(local_err);
2302             aio_context_release(aio_context);
2303             return;
2304         }
2305 
2306         bdrv_close(bs);
2307     }
2308 
2309     /* if we have a device attached to this BlockDriverState
2310      * then we need to make the drive anonymous until the device
2311      * can be removed.  If this is a drive with no device backing
2312      * then we can just get rid of the block driver state right here.
2313      */
2314     if (blk_get_attached_dev(blk)) {
2315         blk_hide_on_behalf_of_hmp_drive_del(blk);
2316         /* Further I/O must not pause the guest */
2317         blk_set_on_error(blk, BLOCKDEV_ON_ERROR_REPORT,
2318                          BLOCKDEV_ON_ERROR_REPORT);
2319     } else {
2320         blk_unref(blk);
2321     }
2322 
2323     aio_context_release(aio_context);
2324 }
2325 
2326 void qmp_block_resize(bool has_device, const char *device,
2327                       bool has_node_name, const char *node_name,
2328                       int64_t size, Error **errp)
2329 {
2330     Error *local_err = NULL;
2331     BlockDriverState *bs;
2332     AioContext *aio_context;
2333     int ret;
2334 
2335     bs = bdrv_lookup_bs(has_device ? device : NULL,
2336                         has_node_name ? node_name : NULL,
2337                         &local_err);
2338     if (local_err) {
2339         error_propagate(errp, local_err);
2340         return;
2341     }
2342 
2343     aio_context = bdrv_get_aio_context(bs);
2344     aio_context_acquire(aio_context);
2345 
2346     if (!bdrv_is_first_non_filter(bs)) {
2347         error_setg(errp, QERR_FEATURE_DISABLED, "resize");
2348         goto out;
2349     }
2350 
2351     if (size < 0) {
2352         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2353         goto out;
2354     }
2355 
2356     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2357         error_setg(errp, QERR_DEVICE_IN_USE, device);
2358         goto out;
2359     }
2360 
2361     /* complete all in-flight operations before resizing the device */
2362     bdrv_drain_all();
2363 
2364     ret = bdrv_truncate(bs, size);
2365     switch (ret) {
2366     case 0:
2367         break;
2368     case -ENOMEDIUM:
2369         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2370         break;
2371     case -ENOTSUP:
2372         error_setg(errp, QERR_UNSUPPORTED);
2373         break;
2374     case -EACCES:
2375         error_setg(errp, "Device '%s' is read only", device);
2376         break;
2377     case -EBUSY:
2378         error_setg(errp, QERR_DEVICE_IN_USE, device);
2379         break;
2380     default:
2381         error_setg_errno(errp, -ret, "Could not resize");
2382         break;
2383     }
2384 
2385 out:
2386     aio_context_release(aio_context);
2387 }
2388 
2389 static void block_job_cb(void *opaque, int ret)
2390 {
2391     /* Note that this function may be executed from another AioContext besides
2392      * the QEMU main loop.  If you need to access anything that assumes the
2393      * QEMU global mutex, use a BH or introduce a mutex.
2394      */
2395 
2396     BlockDriverState *bs = opaque;
2397     const char *msg = NULL;
2398 
2399     trace_block_job_cb(bs, bs->job, ret);
2400 
2401     assert(bs->job);
2402 
2403     if (ret < 0) {
2404         msg = strerror(-ret);
2405     }
2406 
2407     if (block_job_is_cancelled(bs->job)) {
2408         block_job_event_cancelled(bs->job);
2409     } else {
2410         block_job_event_completed(bs->job, msg);
2411     }
2412 
2413     bdrv_put_ref_bh_schedule(bs);
2414 }
2415 
2416 void qmp_block_stream(const char *device,
2417                       bool has_base, const char *base,
2418                       bool has_backing_file, const char *backing_file,
2419                       bool has_speed, int64_t speed,
2420                       bool has_on_error, BlockdevOnError on_error,
2421                       Error **errp)
2422 {
2423     BlockBackend *blk;
2424     BlockDriverState *bs;
2425     BlockDriverState *base_bs = NULL;
2426     AioContext *aio_context;
2427     Error *local_err = NULL;
2428     const char *base_name = NULL;
2429 
2430     if (!has_on_error) {
2431         on_error = BLOCKDEV_ON_ERROR_REPORT;
2432     }
2433 
2434     blk = blk_by_name(device);
2435     if (!blk) {
2436         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2437                   "Device '%s' not found", device);
2438         return;
2439     }
2440 
2441     aio_context = blk_get_aio_context(blk);
2442     aio_context_acquire(aio_context);
2443 
2444     if (!blk_is_available(blk)) {
2445         error_setg(errp, "Device '%s' has no medium", device);
2446         goto out;
2447     }
2448     bs = blk_bs(blk);
2449 
2450     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
2451         goto out;
2452     }
2453 
2454     if (has_base) {
2455         base_bs = bdrv_find_backing_image(bs, base);
2456         if (base_bs == NULL) {
2457             error_setg(errp, QERR_BASE_NOT_FOUND, base);
2458             goto out;
2459         }
2460         assert(bdrv_get_aio_context(base_bs) == aio_context);
2461         base_name = base;
2462     }
2463 
2464     /* if we are streaming the entire chain, the result will have no backing
2465      * file, and specifying one is therefore an error */
2466     if (base_bs == NULL && has_backing_file) {
2467         error_setg(errp, "backing file specified, but streaming the "
2468                          "entire chain");
2469         goto out;
2470     }
2471 
2472     /* backing_file string overrides base bs filename */
2473     base_name = has_backing_file ? backing_file : base_name;
2474 
2475     stream_start(bs, base_bs, base_name, has_speed ? speed : 0,
2476                  on_error, block_job_cb, bs, &local_err);
2477     if (local_err) {
2478         error_propagate(errp, local_err);
2479         goto out;
2480     }
2481 
2482     trace_qmp_block_stream(bs, bs->job);
2483 
2484 out:
2485     aio_context_release(aio_context);
2486 }
2487 
2488 void qmp_block_commit(const char *device,
2489                       bool has_base, const char *base,
2490                       bool has_top, const char *top,
2491                       bool has_backing_file, const char *backing_file,
2492                       bool has_speed, int64_t speed,
2493                       Error **errp)
2494 {
2495     BlockBackend *blk;
2496     BlockDriverState *bs;
2497     BlockDriverState *base_bs, *top_bs;
2498     AioContext *aio_context;
2499     Error *local_err = NULL;
2500     /* This will be part of the QMP command, if/when the
2501      * BlockdevOnError change for blkmirror makes it in
2502      */
2503     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
2504 
2505     if (!has_speed) {
2506         speed = 0;
2507     }
2508 
2509     /* Important Note:
2510      *  libvirt relies on the DeviceNotFound error class in order to probe for
2511      *  live commit feature versions; for this to work, we must make sure to
2512      *  perform the device lookup before any generic errors that may occur in a
2513      *  scenario in which all optional arguments are omitted. */
2514     blk = blk_by_name(device);
2515     if (!blk) {
2516         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2517                   "Device '%s' not found", device);
2518         return;
2519     }
2520 
2521     aio_context = blk_get_aio_context(blk);
2522     aio_context_acquire(aio_context);
2523 
2524     if (!blk_is_available(blk)) {
2525         error_setg(errp, "Device '%s' has no medium", device);
2526         goto out;
2527     }
2528     bs = blk_bs(blk);
2529 
2530     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2531         goto out;
2532     }
2533 
2534     /* default top_bs is the active layer */
2535     top_bs = bs;
2536 
2537     if (has_top && top) {
2538         if (strcmp(bs->filename, top) != 0) {
2539             top_bs = bdrv_find_backing_image(bs, top);
2540         }
2541     }
2542 
2543     if (top_bs == NULL) {
2544         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2545         goto out;
2546     }
2547 
2548     assert(bdrv_get_aio_context(top_bs) == aio_context);
2549 
2550     if (has_base && base) {
2551         base_bs = bdrv_find_backing_image(top_bs, base);
2552     } else {
2553         base_bs = bdrv_find_base(top_bs);
2554     }
2555 
2556     if (base_bs == NULL) {
2557         error_setg(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2558         goto out;
2559     }
2560 
2561     assert(bdrv_get_aio_context(base_bs) == aio_context);
2562 
2563     if (bdrv_op_is_blocked(base_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2564         goto out;
2565     }
2566 
2567     /* Do not allow attempts to commit an image into itself */
2568     if (top_bs == base_bs) {
2569         error_setg(errp, "cannot commit an image into itself");
2570         goto out;
2571     }
2572 
2573     if (top_bs == bs) {
2574         if (has_backing_file) {
2575             error_setg(errp, "'backing-file' specified,"
2576                              " but 'top' is the active layer");
2577             goto out;
2578         }
2579         commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
2580                             bs, &local_err);
2581     } else {
2582         commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
2583                      has_backing_file ? backing_file : NULL, &local_err);
2584     }
2585     if (local_err != NULL) {
2586         error_propagate(errp, local_err);
2587         goto out;
2588     }
2589 
2590 out:
2591     aio_context_release(aio_context);
2592 }
2593 
2594 void qmp_drive_backup(const char *device, const char *target,
2595                       bool has_format, const char *format,
2596                       enum MirrorSyncMode sync,
2597                       bool has_mode, enum NewImageMode mode,
2598                       bool has_speed, int64_t speed,
2599                       bool has_bitmap, const char *bitmap,
2600                       bool has_on_source_error, BlockdevOnError on_source_error,
2601                       bool has_on_target_error, BlockdevOnError on_target_error,
2602                       Error **errp)
2603 {
2604     BlockBackend *blk;
2605     BlockDriverState *bs;
2606     BlockDriverState *target_bs;
2607     BlockDriverState *source = NULL;
2608     BdrvDirtyBitmap *bmap = NULL;
2609     AioContext *aio_context;
2610     QDict *options = NULL;
2611     Error *local_err = NULL;
2612     int flags;
2613     int64_t size;
2614     int ret;
2615 
2616     if (!has_speed) {
2617         speed = 0;
2618     }
2619     if (!has_on_source_error) {
2620         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2621     }
2622     if (!has_on_target_error) {
2623         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2624     }
2625     if (!has_mode) {
2626         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2627     }
2628 
2629     blk = blk_by_name(device);
2630     if (!blk) {
2631         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2632                   "Device '%s' not found", device);
2633         return;
2634     }
2635 
2636     aio_context = blk_get_aio_context(blk);
2637     aio_context_acquire(aio_context);
2638 
2639     /* Although backup_run has this check too, we need to use bs->drv below, so
2640      * do an early check redundantly. */
2641     if (!blk_is_available(blk)) {
2642         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2643         goto out;
2644     }
2645     bs = blk_bs(blk);
2646 
2647     if (!has_format) {
2648         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2649     }
2650 
2651     /* Early check to avoid creating target */
2652     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2653         goto out;
2654     }
2655 
2656     flags = bs->open_flags | BDRV_O_RDWR;
2657 
2658     /* See if we have a backing HD we can use to create our new image
2659      * on top of. */
2660     if (sync == MIRROR_SYNC_MODE_TOP) {
2661         source = backing_bs(bs);
2662         if (!source) {
2663             sync = MIRROR_SYNC_MODE_FULL;
2664         }
2665     }
2666     if (sync == MIRROR_SYNC_MODE_NONE) {
2667         source = bs;
2668     }
2669 
2670     size = bdrv_getlength(bs);
2671     if (size < 0) {
2672         error_setg_errno(errp, -size, "bdrv_getlength failed");
2673         goto out;
2674     }
2675 
2676     if (mode != NEW_IMAGE_MODE_EXISTING) {
2677         assert(format);
2678         if (source) {
2679             bdrv_img_create(target, format, source->filename,
2680                             source->drv->format_name, NULL,
2681                             size, flags, &local_err, false);
2682         } else {
2683             bdrv_img_create(target, format, NULL, NULL, NULL,
2684                             size, flags, &local_err, false);
2685         }
2686     }
2687 
2688     if (local_err) {
2689         error_propagate(errp, local_err);
2690         goto out;
2691     }
2692 
2693     if (format) {
2694         options = qdict_new();
2695         qdict_put(options, "driver", qstring_from_str(format));
2696     }
2697 
2698     target_bs = NULL;
2699     ret = bdrv_open(&target_bs, target, NULL, options, flags, &local_err);
2700     if (ret < 0) {
2701         error_propagate(errp, local_err);
2702         goto out;
2703     }
2704 
2705     bdrv_set_aio_context(target_bs, aio_context);
2706 
2707     if (has_bitmap) {
2708         bmap = bdrv_find_dirty_bitmap(bs, bitmap);
2709         if (!bmap) {
2710             error_setg(errp, "Bitmap '%s' could not be found", bitmap);
2711             goto out;
2712         }
2713     }
2714 
2715     backup_start(bs, target_bs, speed, sync, bmap,
2716                  on_source_error, on_target_error,
2717                  block_job_cb, bs, &local_err);
2718     if (local_err != NULL) {
2719         bdrv_unref(target_bs);
2720         error_propagate(errp, local_err);
2721         goto out;
2722     }
2723 
2724 out:
2725     aio_context_release(aio_context);
2726 }
2727 
2728 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2729 {
2730     return bdrv_named_nodes_list(errp);
2731 }
2732 
2733 void qmp_blockdev_backup(const char *device, const char *target,
2734                          enum MirrorSyncMode sync,
2735                          bool has_speed, int64_t speed,
2736                          bool has_on_source_error,
2737                          BlockdevOnError on_source_error,
2738                          bool has_on_target_error,
2739                          BlockdevOnError on_target_error,
2740                          Error **errp)
2741 {
2742     BlockBackend *blk, *target_blk;
2743     BlockDriverState *bs;
2744     BlockDriverState *target_bs;
2745     Error *local_err = NULL;
2746     AioContext *aio_context;
2747 
2748     if (!has_speed) {
2749         speed = 0;
2750     }
2751     if (!has_on_source_error) {
2752         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2753     }
2754     if (!has_on_target_error) {
2755         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2756     }
2757 
2758     blk = blk_by_name(device);
2759     if (!blk) {
2760         error_setg(errp, "Device '%s' not found", device);
2761         return;
2762     }
2763 
2764     aio_context = blk_get_aio_context(blk);
2765     aio_context_acquire(aio_context);
2766 
2767     if (!blk_is_available(blk)) {
2768         error_setg(errp, "Device '%s' has no medium", device);
2769         goto out;
2770     }
2771     bs = blk_bs(blk);
2772 
2773     target_blk = blk_by_name(target);
2774     if (!target_blk) {
2775         error_setg(errp, "Device '%s' not found", target);
2776         goto out;
2777     }
2778 
2779     if (!blk_is_available(target_blk)) {
2780         error_setg(errp, "Device '%s' has no medium", target);
2781         goto out;
2782     }
2783     target_bs = blk_bs(target_blk);
2784 
2785     bdrv_ref(target_bs);
2786     bdrv_set_aio_context(target_bs, aio_context);
2787     backup_start(bs, target_bs, speed, sync, NULL, on_source_error,
2788                  on_target_error, block_job_cb, bs, &local_err);
2789     if (local_err != NULL) {
2790         bdrv_unref(target_bs);
2791         error_propagate(errp, local_err);
2792     }
2793 out:
2794     aio_context_release(aio_context);
2795 }
2796 
2797 void qmp_drive_mirror(const char *device, const char *target,
2798                       bool has_format, const char *format,
2799                       bool has_node_name, const char *node_name,
2800                       bool has_replaces, const char *replaces,
2801                       enum MirrorSyncMode sync,
2802                       bool has_mode, enum NewImageMode mode,
2803                       bool has_speed, int64_t speed,
2804                       bool has_granularity, uint32_t granularity,
2805                       bool has_buf_size, int64_t buf_size,
2806                       bool has_on_source_error, BlockdevOnError on_source_error,
2807                       bool has_on_target_error, BlockdevOnError on_target_error,
2808                       bool has_unmap, bool unmap,
2809                       Error **errp)
2810 {
2811     BlockBackend *blk;
2812     BlockDriverState *bs;
2813     BlockDriverState *source, *target_bs;
2814     AioContext *aio_context;
2815     Error *local_err = NULL;
2816     QDict *options;
2817     int flags;
2818     int64_t size;
2819     int ret;
2820 
2821     if (!has_speed) {
2822         speed = 0;
2823     }
2824     if (!has_on_source_error) {
2825         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2826     }
2827     if (!has_on_target_error) {
2828         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2829     }
2830     if (!has_mode) {
2831         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2832     }
2833     if (!has_granularity) {
2834         granularity = 0;
2835     }
2836     if (!has_buf_size) {
2837         buf_size = 0;
2838     }
2839     if (!has_unmap) {
2840         unmap = true;
2841     }
2842 
2843     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2844         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2845                    "a value in range [512B, 64MB]");
2846         return;
2847     }
2848     if (granularity & (granularity - 1)) {
2849         error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2850                    "power of 2");
2851         return;
2852     }
2853 
2854     blk = blk_by_name(device);
2855     if (!blk) {
2856         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2857                   "Device '%s' not found", device);
2858         return;
2859     }
2860 
2861     aio_context = blk_get_aio_context(blk);
2862     aio_context_acquire(aio_context);
2863 
2864     if (!blk_is_available(blk)) {
2865         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2866         goto out;
2867     }
2868     bs = blk_bs(blk);
2869 
2870     if (!has_format) {
2871         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2872     }
2873 
2874     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2875         goto out;
2876     }
2877 
2878     flags = bs->open_flags | BDRV_O_RDWR;
2879     source = backing_bs(bs);
2880     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2881         sync = MIRROR_SYNC_MODE_FULL;
2882     }
2883     if (sync == MIRROR_SYNC_MODE_NONE) {
2884         source = bs;
2885     }
2886 
2887     size = bdrv_getlength(bs);
2888     if (size < 0) {
2889         error_setg_errno(errp, -size, "bdrv_getlength failed");
2890         goto out;
2891     }
2892 
2893     if (has_replaces) {
2894         BlockDriverState *to_replace_bs;
2895         AioContext *replace_aio_context;
2896         int64_t replace_size;
2897 
2898         if (!has_node_name) {
2899             error_setg(errp, "a node-name must be provided when replacing a"
2900                              " named node of the graph");
2901             goto out;
2902         }
2903 
2904         to_replace_bs = check_to_replace_node(bs, replaces, &local_err);
2905 
2906         if (!to_replace_bs) {
2907             error_propagate(errp, local_err);
2908             goto out;
2909         }
2910 
2911         replace_aio_context = bdrv_get_aio_context(to_replace_bs);
2912         aio_context_acquire(replace_aio_context);
2913         replace_size = bdrv_getlength(to_replace_bs);
2914         aio_context_release(replace_aio_context);
2915 
2916         if (size != replace_size) {
2917             error_setg(errp, "cannot replace image with a mirror image of "
2918                              "different size");
2919             goto out;
2920         }
2921     }
2922 
2923     if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2924         && mode != NEW_IMAGE_MODE_EXISTING)
2925     {
2926         /* create new image w/o backing file */
2927         assert(format);
2928         bdrv_img_create(target, format,
2929                         NULL, NULL, NULL, size, flags, &local_err, false);
2930     } else {
2931         switch (mode) {
2932         case NEW_IMAGE_MODE_EXISTING:
2933             break;
2934         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2935             /* create new image with backing file */
2936             bdrv_img_create(target, format,
2937                             source->filename,
2938                             source->drv->format_name,
2939                             NULL, size, flags, &local_err, false);
2940             break;
2941         default:
2942             abort();
2943         }
2944     }
2945 
2946     if (local_err) {
2947         error_propagate(errp, local_err);
2948         goto out;
2949     }
2950 
2951     options = qdict_new();
2952     if (has_node_name) {
2953         qdict_put(options, "node-name", qstring_from_str(node_name));
2954     }
2955     if (format) {
2956         qdict_put(options, "driver", qstring_from_str(format));
2957     }
2958 
2959     /* Mirroring takes care of copy-on-write using the source's backing
2960      * file.
2961      */
2962     target_bs = NULL;
2963     ret = bdrv_open(&target_bs, target, NULL, options,
2964                     flags | BDRV_O_NO_BACKING, &local_err);
2965     if (ret < 0) {
2966         error_propagate(errp, local_err);
2967         goto out;
2968     }
2969 
2970     bdrv_set_aio_context(target_bs, aio_context);
2971 
2972     /* pass the node name to replace to mirror start since it's loose coupling
2973      * and will allow to check whether the node still exist at mirror completion
2974      */
2975     mirror_start(bs, target_bs,
2976                  has_replaces ? replaces : NULL,
2977                  speed, granularity, buf_size, sync,
2978                  on_source_error, on_target_error,
2979                  unmap,
2980                  block_job_cb, bs, &local_err);
2981     if (local_err != NULL) {
2982         bdrv_unref(target_bs);
2983         error_propagate(errp, local_err);
2984         goto out;
2985     }
2986 
2987 out:
2988     aio_context_release(aio_context);
2989 }
2990 
2991 /* Get the block job for a given device name and acquire its AioContext */
2992 static BlockJob *find_block_job(const char *device, AioContext **aio_context,
2993                                 Error **errp)
2994 {
2995     BlockBackend *blk;
2996     BlockDriverState *bs;
2997 
2998     *aio_context = NULL;
2999 
3000     blk = blk_by_name(device);
3001     if (!blk) {
3002         goto notfound;
3003     }
3004 
3005     *aio_context = blk_get_aio_context(blk);
3006     aio_context_acquire(*aio_context);
3007 
3008     if (!blk_is_available(blk)) {
3009         goto notfound;
3010     }
3011     bs = blk_bs(blk);
3012 
3013     if (!bs->job) {
3014         goto notfound;
3015     }
3016 
3017     return bs->job;
3018 
3019 notfound:
3020     error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3021               "No active block job on device '%s'", device);
3022     if (*aio_context) {
3023         aio_context_release(*aio_context);
3024         *aio_context = NULL;
3025     }
3026     return NULL;
3027 }
3028 
3029 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3030 {
3031     AioContext *aio_context;
3032     BlockJob *job = find_block_job(device, &aio_context, errp);
3033 
3034     if (!job) {
3035         return;
3036     }
3037 
3038     block_job_set_speed(job, speed, errp);
3039     aio_context_release(aio_context);
3040 }
3041 
3042 void qmp_block_job_cancel(const char *device,
3043                           bool has_force, bool force, Error **errp)
3044 {
3045     AioContext *aio_context;
3046     BlockJob *job = find_block_job(device, &aio_context, errp);
3047 
3048     if (!job) {
3049         return;
3050     }
3051 
3052     if (!has_force) {
3053         force = false;
3054     }
3055 
3056     if (job->user_paused && !force) {
3057         error_setg(errp, "The block job for device '%s' is currently paused",
3058                    device);
3059         goto out;
3060     }
3061 
3062     trace_qmp_block_job_cancel(job);
3063     block_job_cancel(job);
3064 out:
3065     aio_context_release(aio_context);
3066 }
3067 
3068 void qmp_block_job_pause(const char *device, Error **errp)
3069 {
3070     AioContext *aio_context;
3071     BlockJob *job = find_block_job(device, &aio_context, errp);
3072 
3073     if (!job || job->user_paused) {
3074         return;
3075     }
3076 
3077     job->user_paused = true;
3078     trace_qmp_block_job_pause(job);
3079     block_job_pause(job);
3080     aio_context_release(aio_context);
3081 }
3082 
3083 void qmp_block_job_resume(const char *device, Error **errp)
3084 {
3085     AioContext *aio_context;
3086     BlockJob *job = find_block_job(device, &aio_context, errp);
3087 
3088     if (!job || !job->user_paused) {
3089         return;
3090     }
3091 
3092     job->user_paused = false;
3093     trace_qmp_block_job_resume(job);
3094     block_job_resume(job);
3095     aio_context_release(aio_context);
3096 }
3097 
3098 void qmp_block_job_complete(const char *device, Error **errp)
3099 {
3100     AioContext *aio_context;
3101     BlockJob *job = find_block_job(device, &aio_context, errp);
3102 
3103     if (!job) {
3104         return;
3105     }
3106 
3107     trace_qmp_block_job_complete(job);
3108     block_job_complete(job, errp);
3109     aio_context_release(aio_context);
3110 }
3111 
3112 void qmp_change_backing_file(const char *device,
3113                              const char *image_node_name,
3114                              const char *backing_file,
3115                              Error **errp)
3116 {
3117     BlockBackend *blk;
3118     BlockDriverState *bs = NULL;
3119     AioContext *aio_context;
3120     BlockDriverState *image_bs = NULL;
3121     Error *local_err = NULL;
3122     bool ro;
3123     int open_flags;
3124     int ret;
3125 
3126     blk = blk_by_name(device);
3127     if (!blk) {
3128         error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
3129                   "Device '%s' not found", device);
3130         return;
3131     }
3132 
3133     aio_context = blk_get_aio_context(blk);
3134     aio_context_acquire(aio_context);
3135 
3136     if (!blk_is_available(blk)) {
3137         error_setg(errp, "Device '%s' has no medium", device);
3138         goto out;
3139     }
3140     bs = blk_bs(blk);
3141 
3142     image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3143     if (local_err) {
3144         error_propagate(errp, local_err);
3145         goto out;
3146     }
3147 
3148     if (!image_bs) {
3149         error_setg(errp, "image file not found");
3150         goto out;
3151     }
3152 
3153     if (bdrv_find_base(image_bs) == image_bs) {
3154         error_setg(errp, "not allowing backing file change on an image "
3155                          "without a backing file");
3156         goto out;
3157     }
3158 
3159     /* even though we are not necessarily operating on bs, we need it to
3160      * determine if block ops are currently prohibited on the chain */
3161     if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3162         goto out;
3163     }
3164 
3165     /* final sanity check */
3166     if (!bdrv_chain_contains(bs, image_bs)) {
3167         error_setg(errp, "'%s' and image file are not in the same chain",
3168                    device);
3169         goto out;
3170     }
3171 
3172     /* if not r/w, reopen to make r/w */
3173     open_flags = image_bs->open_flags;
3174     ro = bdrv_is_read_only(image_bs);
3175 
3176     if (ro) {
3177         bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err);
3178         if (local_err) {
3179             error_propagate(errp, local_err);
3180             goto out;
3181         }
3182     }
3183 
3184     ret = bdrv_change_backing_file(image_bs, backing_file,
3185                                image_bs->drv ? image_bs->drv->format_name : "");
3186 
3187     if (ret < 0) {
3188         error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3189                          backing_file);
3190         /* don't exit here, so we can try to restore open flags if
3191          * appropriate */
3192     }
3193 
3194     if (ro) {
3195         bdrv_reopen(image_bs, open_flags, &local_err);
3196         if (local_err) {
3197             error_propagate(errp, local_err); /* will preserve prior errp */
3198         }
3199     }
3200 
3201 out:
3202     aio_context_release(aio_context);
3203 }
3204 
3205 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3206 {
3207     QmpOutputVisitor *ov = qmp_output_visitor_new();
3208     BlockDriverState *bs;
3209     BlockBackend *blk = NULL;
3210     QObject *obj;
3211     QDict *qdict;
3212     Error *local_err = NULL;
3213 
3214     /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
3215      * cache.direct=false instead of silently switching to aio=threads, except
3216      * when called from drive_new().
3217      *
3218      * For now, simply forbidding the combination for all drivers will do. */
3219     if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
3220         bool direct = options->has_cache &&
3221                       options->cache->has_direct &&
3222                       options->cache->direct;
3223         if (!direct) {
3224             error_setg(errp, "aio=native requires cache.direct=true");
3225             goto fail;
3226         }
3227     }
3228 
3229     visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
3230                                &options, NULL, &local_err);
3231     if (local_err) {
3232         error_propagate(errp, local_err);
3233         goto fail;
3234     }
3235 
3236     obj = qmp_output_get_qobject(ov);
3237     qdict = qobject_to_qdict(obj);
3238 
3239     qdict_flatten(qdict);
3240 
3241     if (options->has_id) {
3242         blk = blockdev_init(NULL, qdict, &local_err);
3243         if (local_err) {
3244             error_propagate(errp, local_err);
3245             goto fail;
3246         }
3247 
3248         bs = blk_bs(blk);
3249     } else {
3250         if (!qdict_get_try_str(qdict, "node-name")) {
3251             error_setg(errp, "'id' and/or 'node-name' need to be specified for "
3252                        "the root node");
3253             goto fail;
3254         }
3255 
3256         bs = bds_tree_init(qdict, errp);
3257         if (!bs) {
3258             goto fail;
3259         }
3260     }
3261 
3262     if (bs && bdrv_key_required(bs)) {
3263         if (blk) {
3264             blk_unref(blk);
3265         } else {
3266             bdrv_unref(bs);
3267         }
3268         error_setg(errp, "blockdev-add doesn't support encrypted devices");
3269         goto fail;
3270     }
3271 
3272 fail:
3273     qmp_output_visitor_cleanup(ov);
3274 }
3275 
3276 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3277 {
3278     BlockJobInfoList *head = NULL, **p_next = &head;
3279     BlockDriverState *bs;
3280 
3281     for (bs = bdrv_next(NULL); bs; bs = bdrv_next(bs)) {
3282         AioContext *aio_context = bdrv_get_aio_context(bs);
3283 
3284         aio_context_acquire(aio_context);
3285 
3286         if (bs->job) {
3287             BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
3288             elem->value = block_job_query(bs->job);
3289             *p_next = elem;
3290             p_next = &elem->next;
3291         }
3292 
3293         aio_context_release(aio_context);
3294     }
3295 
3296     return head;
3297 }
3298 
3299 QemuOptsList qemu_common_drive_opts = {
3300     .name = "drive",
3301     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3302     .desc = {
3303         {
3304             .name = "snapshot",
3305             .type = QEMU_OPT_BOOL,
3306             .help = "enable/disable snapshot mode",
3307         },{
3308             .name = "discard",
3309             .type = QEMU_OPT_STRING,
3310             .help = "discard operation (ignore/off, unmap/on)",
3311         },{
3312             .name = BDRV_OPT_CACHE_WB,
3313             .type = QEMU_OPT_BOOL,
3314             .help = "enables writeback mode for any caches",
3315         },{
3316             .name = BDRV_OPT_CACHE_DIRECT,
3317             .type = QEMU_OPT_BOOL,
3318             .help = "enables use of O_DIRECT (bypass the host page cache)",
3319         },{
3320             .name = BDRV_OPT_CACHE_NO_FLUSH,
3321             .type = QEMU_OPT_BOOL,
3322             .help = "ignore any flush requests for the device",
3323         },{
3324             .name = "aio",
3325             .type = QEMU_OPT_STRING,
3326             .help = "host AIO implementation (threads, native)",
3327         },{
3328             .name = "format",
3329             .type = QEMU_OPT_STRING,
3330             .help = "disk format (raw, qcow2, ...)",
3331         },{
3332             .name = "rerror",
3333             .type = QEMU_OPT_STRING,
3334             .help = "read error action",
3335         },{
3336             .name = "werror",
3337             .type = QEMU_OPT_STRING,
3338             .help = "write error action",
3339         },{
3340             .name = "read-only",
3341             .type = QEMU_OPT_BOOL,
3342             .help = "open drive file as read-only",
3343         },{
3344             .name = "throttling.iops-total",
3345             .type = QEMU_OPT_NUMBER,
3346             .help = "limit total I/O operations per second",
3347         },{
3348             .name = "throttling.iops-read",
3349             .type = QEMU_OPT_NUMBER,
3350             .help = "limit read operations per second",
3351         },{
3352             .name = "throttling.iops-write",
3353             .type = QEMU_OPT_NUMBER,
3354             .help = "limit write operations per second",
3355         },{
3356             .name = "throttling.bps-total",
3357             .type = QEMU_OPT_NUMBER,
3358             .help = "limit total bytes per second",
3359         },{
3360             .name = "throttling.bps-read",
3361             .type = QEMU_OPT_NUMBER,
3362             .help = "limit read bytes per second",
3363         },{
3364             .name = "throttling.bps-write",
3365             .type = QEMU_OPT_NUMBER,
3366             .help = "limit write bytes per second",
3367         },{
3368             .name = "throttling.iops-total-max",
3369             .type = QEMU_OPT_NUMBER,
3370             .help = "I/O operations burst",
3371         },{
3372             .name = "throttling.iops-read-max",
3373             .type = QEMU_OPT_NUMBER,
3374             .help = "I/O operations read burst",
3375         },{
3376             .name = "throttling.iops-write-max",
3377             .type = QEMU_OPT_NUMBER,
3378             .help = "I/O operations write burst",
3379         },{
3380             .name = "throttling.bps-total-max",
3381             .type = QEMU_OPT_NUMBER,
3382             .help = "total bytes burst",
3383         },{
3384             .name = "throttling.bps-read-max",
3385             .type = QEMU_OPT_NUMBER,
3386             .help = "total bytes read burst",
3387         },{
3388             .name = "throttling.bps-write-max",
3389             .type = QEMU_OPT_NUMBER,
3390             .help = "total bytes write burst",
3391         },{
3392             .name = "throttling.iops-size",
3393             .type = QEMU_OPT_NUMBER,
3394             .help = "when limiting by iops max size of an I/O in bytes",
3395         },{
3396             .name = "throttling.group",
3397             .type = QEMU_OPT_STRING,
3398             .help = "name of the block throttling group",
3399         },{
3400             .name = "copy-on-read",
3401             .type = QEMU_OPT_BOOL,
3402             .help = "copy read data from backing file into image file",
3403         },{
3404             .name = "detect-zeroes",
3405             .type = QEMU_OPT_STRING,
3406             .help = "try to optimize zero writes (off, on, unmap)",
3407         },
3408         { /* end of list */ }
3409     },
3410 };
3411 
3412 static QemuOptsList qemu_root_bds_opts = {
3413     .name = "root-bds",
3414     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3415     .desc = {
3416         {
3417             .name = "discard",
3418             .type = QEMU_OPT_STRING,
3419             .help = "discard operation (ignore/off, unmap/on)",
3420         },{
3421             .name = "cache.writeback",
3422             .type = QEMU_OPT_BOOL,
3423             .help = "enables writeback mode for any caches",
3424         },{
3425             .name = "cache.direct",
3426             .type = QEMU_OPT_BOOL,
3427             .help = "enables use of O_DIRECT (bypass the host page cache)",
3428         },{
3429             .name = "cache.no-flush",
3430             .type = QEMU_OPT_BOOL,
3431             .help = "ignore any flush requests for the device",
3432         },{
3433             .name = "aio",
3434             .type = QEMU_OPT_STRING,
3435             .help = "host AIO implementation (threads, native)",
3436         },{
3437             .name = "read-only",
3438             .type = QEMU_OPT_BOOL,
3439             .help = "open drive file as read-only",
3440         },{
3441             .name = "copy-on-read",
3442             .type = QEMU_OPT_BOOL,
3443             .help = "copy read data from backing file into image file",
3444         },{
3445             .name = "detect-zeroes",
3446             .type = QEMU_OPT_STRING,
3447             .help = "try to optimize zero writes (off, on, unmap)",
3448         },
3449         { /* end of list */ }
3450     },
3451 };
3452 
3453 QemuOptsList qemu_drive_opts = {
3454     .name = "drive",
3455     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3456     .desc = {
3457         /*
3458          * no elements => accept any params
3459          * validation will happen later
3460          */
3461         { /* end of list */ }
3462     },
3463 };
3464