xref: /openbmc/qemu/blockdev.c (revision 5accc840)
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/blockdev.h"
34 #include "hw/block/block.h"
35 #include "block/blockjob.h"
36 #include "monitor/monitor.h"
37 #include "qapi/qmp/qerror.h"
38 #include "qemu/option.h"
39 #include "qemu/config-file.h"
40 #include "qapi/qmp/types.h"
41 #include "sysemu/sysemu.h"
42 #include "block/block_int.h"
43 #include "qmp-commands.h"
44 #include "trace.h"
45 #include "sysemu/arch_init.h"
46 
47 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
48 extern QemuOptsList qemu_common_drive_opts;
49 
50 static const char *const if_name[IF_COUNT] = {
51     [IF_NONE] = "none",
52     [IF_IDE] = "ide",
53     [IF_SCSI] = "scsi",
54     [IF_FLOPPY] = "floppy",
55     [IF_PFLASH] = "pflash",
56     [IF_MTD] = "mtd",
57     [IF_SD] = "sd",
58     [IF_VIRTIO] = "virtio",
59     [IF_XEN] = "xen",
60 };
61 
62 static const int if_max_devs[IF_COUNT] = {
63     /*
64      * Do not change these numbers!  They govern how drive option
65      * index maps to unit and bus.  That mapping is ABI.
66      *
67      * All controllers used to imlement if=T drives need to support
68      * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
69      * Otherwise, some index values map to "impossible" bus, unit
70      * values.
71      *
72      * For instance, if you change [IF_SCSI] to 255, -drive
73      * if=scsi,index=12 no longer means bus=1,unit=5, but
74      * bus=0,unit=12.  With an lsi53c895a controller (7 units max),
75      * the drive can't be set up.  Regression.
76      */
77     [IF_IDE] = 2,
78     [IF_SCSI] = 7,
79 };
80 
81 /*
82  * We automatically delete the drive when a device using it gets
83  * unplugged.  Questionable feature, but we can't just drop it.
84  * Device models call blockdev_mark_auto_del() to schedule the
85  * automatic deletion, and generic qdev code calls blockdev_auto_del()
86  * when deletion is actually safe.
87  */
88 void blockdev_mark_auto_del(BlockDriverState *bs)
89 {
90     DriveInfo *dinfo = drive_get_by_blockdev(bs);
91 
92     if (bs->job) {
93         block_job_cancel(bs->job);
94     }
95     if (dinfo) {
96         dinfo->auto_del = 1;
97     }
98 }
99 
100 void blockdev_auto_del(BlockDriverState *bs)
101 {
102     DriveInfo *dinfo = drive_get_by_blockdev(bs);
103 
104     if (dinfo && dinfo->auto_del) {
105         drive_put_ref(dinfo);
106     }
107 }
108 
109 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
110 {
111     int max_devs = if_max_devs[type];
112     return max_devs ? index / max_devs : 0;
113 }
114 
115 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
116 {
117     int max_devs = if_max_devs[type];
118     return max_devs ? index % max_devs : index;
119 }
120 
121 QemuOpts *drive_def(const char *optstr)
122 {
123     return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
124 }
125 
126 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
127                     const char *optstr)
128 {
129     QemuOpts *opts;
130     char buf[32];
131 
132     opts = drive_def(optstr);
133     if (!opts) {
134         return NULL;
135     }
136     if (type != IF_DEFAULT) {
137         qemu_opt_set(opts, "if", if_name[type]);
138     }
139     if (index >= 0) {
140         snprintf(buf, sizeof(buf), "%d", index);
141         qemu_opt_set(opts, "index", buf);
142     }
143     if (file)
144         qemu_opt_set(opts, "file", file);
145     return opts;
146 }
147 
148 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
149 {
150     DriveInfo *dinfo;
151 
152     /* seek interface, bus and unit */
153 
154     QTAILQ_FOREACH(dinfo, &drives, next) {
155         if (dinfo->type == type &&
156 	    dinfo->bus == bus &&
157 	    dinfo->unit == unit)
158             return dinfo;
159     }
160 
161     return NULL;
162 }
163 
164 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
165 {
166     return drive_get(type,
167                      drive_index_to_bus_id(type, index),
168                      drive_index_to_unit_id(type, index));
169 }
170 
171 int drive_get_max_bus(BlockInterfaceType type)
172 {
173     int max_bus;
174     DriveInfo *dinfo;
175 
176     max_bus = -1;
177     QTAILQ_FOREACH(dinfo, &drives, next) {
178         if(dinfo->type == type &&
179            dinfo->bus > max_bus)
180             max_bus = dinfo->bus;
181     }
182     return max_bus;
183 }
184 
185 /* Get a block device.  This should only be used for single-drive devices
186    (e.g. SD/Floppy/MTD).  Multi-disk devices (scsi/ide) should use the
187    appropriate bus.  */
188 DriveInfo *drive_get_next(BlockInterfaceType type)
189 {
190     static int next_block_unit[IF_COUNT];
191 
192     return drive_get(type, 0, next_block_unit[type]++);
193 }
194 
195 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
196 {
197     DriveInfo *dinfo;
198 
199     QTAILQ_FOREACH(dinfo, &drives, next) {
200         if (dinfo->bdrv == bs) {
201             return dinfo;
202         }
203     }
204     return NULL;
205 }
206 
207 static void bdrv_format_print(void *opaque, const char *name)
208 {
209     error_printf(" %s", name);
210 }
211 
212 static void drive_uninit(DriveInfo *dinfo)
213 {
214     qemu_opts_del(dinfo->opts);
215     bdrv_unref(dinfo->bdrv);
216     g_free(dinfo->id);
217     QTAILQ_REMOVE(&drives, dinfo, next);
218     g_free(dinfo->serial);
219     g_free(dinfo);
220 }
221 
222 void drive_put_ref(DriveInfo *dinfo)
223 {
224     assert(dinfo->refcount);
225     if (--dinfo->refcount == 0) {
226         drive_uninit(dinfo);
227     }
228 }
229 
230 void drive_get_ref(DriveInfo *dinfo)
231 {
232     dinfo->refcount++;
233 }
234 
235 typedef struct {
236     QEMUBH *bh;
237     BlockDriverState *bs;
238 } BDRVPutRefBH;
239 
240 static void bdrv_put_ref_bh(void *opaque)
241 {
242     BDRVPutRefBH *s = opaque;
243 
244     bdrv_unref(s->bs);
245     qemu_bh_delete(s->bh);
246     g_free(s);
247 }
248 
249 /*
250  * Release a BDS reference in a BH
251  *
252  * It is not safe to use bdrv_unref() from a callback function when the callers
253  * still need the BlockDriverState.  In such cases we schedule a BH to release
254  * the reference.
255  */
256 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
257 {
258     BDRVPutRefBH *s;
259 
260     s = g_new(BDRVPutRefBH, 1);
261     s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
262     s->bs = bs;
263     qemu_bh_schedule(s->bh);
264 }
265 
266 static int parse_block_error_action(const char *buf, bool is_read)
267 {
268     if (!strcmp(buf, "ignore")) {
269         return BLOCKDEV_ON_ERROR_IGNORE;
270     } else if (!is_read && !strcmp(buf, "enospc")) {
271         return BLOCKDEV_ON_ERROR_ENOSPC;
272     } else if (!strcmp(buf, "stop")) {
273         return BLOCKDEV_ON_ERROR_STOP;
274     } else if (!strcmp(buf, "report")) {
275         return BLOCKDEV_ON_ERROR_REPORT;
276     } else {
277         error_report("'%s' invalid %s error action",
278                      buf, is_read ? "read" : "write");
279         return -1;
280     }
281 }
282 
283 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
284 {
285     if (throttle_conflicting(cfg)) {
286         error_setg(errp, "bps/iops/max total values and read/write values"
287                          " cannot be used at the same time");
288         return false;
289     }
290 
291     if (!throttle_is_valid(cfg)) {
292         error_setg(errp, "bps/iops/maxs values must be 0 or greater");
293         return false;
294     }
295 
296     return true;
297 }
298 
299 static DriveInfo *blockdev_init(QemuOpts *all_opts,
300                                 BlockInterfaceType block_default_type)
301 {
302     const char *buf;
303     const char *file = NULL;
304     const char *serial;
305     const char *mediastr = "";
306     BlockInterfaceType type;
307     enum { MEDIA_DISK, MEDIA_CDROM } media;
308     int bus_id, unit_id;
309     int cyls, heads, secs, translation;
310     int max_devs;
311     int index;
312     int ro = 0;
313     int bdrv_flags = 0;
314     int on_read_error, on_write_error;
315     const char *devaddr;
316     DriveInfo *dinfo;
317     ThrottleConfig cfg;
318     int snapshot = 0;
319     bool copy_on_read;
320     int ret;
321     Error *error = NULL;
322     QemuOpts *opts;
323     QDict *bs_opts;
324     const char *id;
325     bool has_driver_specific_opts;
326     BlockDriver *drv = NULL;
327 
328     translation = BIOS_ATA_TRANSLATION_AUTO;
329     media = MEDIA_DISK;
330 
331     /* Check common options by copying from all_opts to opts, all other options
332      * are stored in bs_opts. */
333     id = qemu_opts_id(all_opts);
334     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
335     if (error_is_set(&error)) {
336         qerror_report_err(error);
337         error_free(error);
338         return NULL;
339     }
340 
341     bs_opts = qdict_new();
342     qemu_opts_to_qdict(all_opts, bs_opts);
343     qemu_opts_absorb_qdict(opts, bs_opts, &error);
344     if (error_is_set(&error)) {
345         qerror_report_err(error);
346         error_free(error);
347         return NULL;
348     }
349 
350     if (id) {
351         qdict_del(bs_opts, "id");
352     }
353 
354     has_driver_specific_opts = !!qdict_size(bs_opts);
355 
356     /* extract parameters */
357     bus_id  = qemu_opt_get_number(opts, "bus", 0);
358     unit_id = qemu_opt_get_number(opts, "unit", -1);
359     index   = qemu_opt_get_number(opts, "index", -1);
360 
361     cyls  = qemu_opt_get_number(opts, "cyls", 0);
362     heads = qemu_opt_get_number(opts, "heads", 0);
363     secs  = qemu_opt_get_number(opts, "secs", 0);
364 
365     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
366     ro = qemu_opt_get_bool(opts, "read-only", 0);
367     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
368 
369     file = qemu_opt_get(opts, "file");
370     serial = qemu_opt_get(opts, "serial");
371 
372     if ((buf = qemu_opt_get(opts, "if")) != NULL) {
373         for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
374             ;
375         if (type == IF_COUNT) {
376             error_report("unsupported bus type '%s'", buf);
377             return NULL;
378 	}
379     } else {
380         type = block_default_type;
381     }
382 
383     max_devs = if_max_devs[type];
384 
385     if (cyls || heads || secs) {
386         if (cyls < 1) {
387             error_report("invalid physical cyls number");
388 	    return NULL;
389 	}
390         if (heads < 1) {
391             error_report("invalid physical heads number");
392 	    return NULL;
393 	}
394         if (secs < 1) {
395             error_report("invalid physical secs number");
396 	    return NULL;
397 	}
398     }
399 
400     if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
401         if (!cyls) {
402             error_report("'%s' trans must be used with cyls, heads and secs",
403                          buf);
404             return NULL;
405         }
406         if (!strcmp(buf, "none"))
407             translation = BIOS_ATA_TRANSLATION_NONE;
408         else if (!strcmp(buf, "lba"))
409             translation = BIOS_ATA_TRANSLATION_LBA;
410         else if (!strcmp(buf, "auto"))
411             translation = BIOS_ATA_TRANSLATION_AUTO;
412 	else {
413             error_report("'%s' invalid translation type", buf);
414 	    return NULL;
415 	}
416     }
417 
418     if ((buf = qemu_opt_get(opts, "media")) != NULL) {
419         if (!strcmp(buf, "disk")) {
420 	    media = MEDIA_DISK;
421 	} else if (!strcmp(buf, "cdrom")) {
422             if (cyls || secs || heads) {
423                 error_report("CHS can't be set with media=%s", buf);
424 	        return NULL;
425             }
426 	    media = MEDIA_CDROM;
427 	} else {
428 	    error_report("'%s' invalid media", buf);
429 	    return NULL;
430 	}
431     }
432 
433     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
434         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
435             error_report("invalid discard option");
436             return NULL;
437         }
438     }
439 
440     if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
441         bdrv_flags |= BDRV_O_CACHE_WB;
442     }
443     if (qemu_opt_get_bool(opts, "cache.direct", false)) {
444         bdrv_flags |= BDRV_O_NOCACHE;
445     }
446     if (qemu_opt_get_bool(opts, "cache.no-flush", true)) {
447         bdrv_flags |= BDRV_O_NO_FLUSH;
448     }
449 
450 #ifdef CONFIG_LINUX_AIO
451     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
452         if (!strcmp(buf, "native")) {
453             bdrv_flags |= BDRV_O_NATIVE_AIO;
454         } else if (!strcmp(buf, "threads")) {
455             /* this is the default */
456         } else {
457            error_report("invalid aio option");
458            return NULL;
459         }
460     }
461 #endif
462 
463     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
464         if (is_help_option(buf)) {
465             error_printf("Supported formats:");
466             bdrv_iterate_format(bdrv_format_print, NULL);
467             error_printf("\n");
468             return NULL;
469         }
470 
471         drv = bdrv_find_whitelisted_format(buf, ro);
472         if (!drv) {
473             if (!ro && bdrv_find_whitelisted_format(buf, !ro)) {
474                 error_report("'%s' can be only used as read-only device.", buf);
475             } else {
476                 error_report("'%s' invalid format", buf);
477             }
478             return NULL;
479         }
480     }
481 
482     /* disk I/O throttling */
483     memset(&cfg, 0, sizeof(cfg));
484     cfg.buckets[THROTTLE_BPS_TOTAL].avg =
485         qemu_opt_get_number(opts, "throttling.bps-total", 0);
486     cfg.buckets[THROTTLE_BPS_READ].avg  =
487         qemu_opt_get_number(opts, "throttling.bps-read", 0);
488     cfg.buckets[THROTTLE_BPS_WRITE].avg =
489         qemu_opt_get_number(opts, "throttling.bps-write", 0);
490     cfg.buckets[THROTTLE_OPS_TOTAL].avg =
491         qemu_opt_get_number(opts, "throttling.iops-total", 0);
492     cfg.buckets[THROTTLE_OPS_READ].avg =
493         qemu_opt_get_number(opts, "throttling.iops-read", 0);
494     cfg.buckets[THROTTLE_OPS_WRITE].avg =
495         qemu_opt_get_number(opts, "throttling.iops-write", 0);
496 
497     cfg.buckets[THROTTLE_BPS_TOTAL].max =
498         qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
499     cfg.buckets[THROTTLE_BPS_READ].max  =
500         qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
501     cfg.buckets[THROTTLE_BPS_WRITE].max =
502         qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
503     cfg.buckets[THROTTLE_OPS_TOTAL].max =
504         qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
505     cfg.buckets[THROTTLE_OPS_READ].max =
506         qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
507     cfg.buckets[THROTTLE_OPS_WRITE].max =
508         qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
509 
510     cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
511 
512     if (!check_throttle_config(&cfg, &error)) {
513         error_report("%s", error_get_pretty(error));
514         error_free(error);
515         return NULL;
516     }
517 
518     if (qemu_opt_get(opts, "boot") != NULL) {
519         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
520                 "ignored. Future versions will reject this parameter. Please "
521                 "update your scripts.\n");
522     }
523 
524     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
525     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
526         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
527             error_report("werror is not supported by this bus type");
528             return NULL;
529         }
530 
531         on_write_error = parse_block_error_action(buf, 0);
532         if (on_write_error < 0) {
533             return NULL;
534         }
535     }
536 
537     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
538     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
539         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
540             error_report("rerror is not supported by this bus type");
541             return NULL;
542         }
543 
544         on_read_error = parse_block_error_action(buf, 1);
545         if (on_read_error < 0) {
546             return NULL;
547         }
548     }
549 
550     if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
551         if (type != IF_VIRTIO) {
552             error_report("addr is not supported by this bus type");
553             return NULL;
554         }
555     }
556 
557     /* compute bus and unit according index */
558 
559     if (index != -1) {
560         if (bus_id != 0 || unit_id != -1) {
561             error_report("index cannot be used with bus and unit");
562             return NULL;
563         }
564         bus_id = drive_index_to_bus_id(type, index);
565         unit_id = drive_index_to_unit_id(type, index);
566     }
567 
568     /* if user doesn't specify a unit_id,
569      * try to find the first free
570      */
571 
572     if (unit_id == -1) {
573        unit_id = 0;
574        while (drive_get(type, bus_id, unit_id) != NULL) {
575            unit_id++;
576            if (max_devs && unit_id >= max_devs) {
577                unit_id -= max_devs;
578                bus_id++;
579            }
580        }
581     }
582 
583     /* check unit id */
584 
585     if (max_devs && unit_id >= max_devs) {
586         error_report("unit %d too big (max is %d)",
587                      unit_id, max_devs - 1);
588         return NULL;
589     }
590 
591     /*
592      * catch multiple definitions
593      */
594 
595     if (drive_get(type, bus_id, unit_id) != NULL) {
596         error_report("drive with bus=%d, unit=%d (index=%d) exists",
597                      bus_id, unit_id, index);
598         return NULL;
599     }
600 
601     /* init */
602 
603     dinfo = g_malloc0(sizeof(*dinfo));
604     if ((buf = qemu_opts_id(opts)) != NULL) {
605         dinfo->id = g_strdup(buf);
606     } else {
607         /* no id supplied -> create one */
608         dinfo->id = g_malloc0(32);
609         if (type == IF_IDE || type == IF_SCSI)
610             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
611         if (max_devs)
612             snprintf(dinfo->id, 32, "%s%i%s%i",
613                      if_name[type], bus_id, mediastr, unit_id);
614         else
615             snprintf(dinfo->id, 32, "%s%s%i",
616                      if_name[type], mediastr, unit_id);
617     }
618     dinfo->bdrv = bdrv_new(dinfo->id);
619     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
620     dinfo->bdrv->read_only = ro;
621     dinfo->devaddr = devaddr;
622     dinfo->type = type;
623     dinfo->bus = bus_id;
624     dinfo->unit = unit_id;
625     dinfo->cyls = cyls;
626     dinfo->heads = heads;
627     dinfo->secs = secs;
628     dinfo->trans = translation;
629     dinfo->opts = all_opts;
630     dinfo->refcount = 1;
631     if (serial != NULL) {
632         dinfo->serial = g_strdup(serial);
633     }
634     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
635 
636     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
637 
638     /* disk I/O throttling */
639     if (throttle_enabled(&cfg)) {
640         bdrv_io_limits_enable(dinfo->bdrv);
641         bdrv_set_io_limits(dinfo->bdrv, &cfg);
642     }
643 
644     switch(type) {
645     case IF_IDE:
646     case IF_SCSI:
647     case IF_XEN:
648     case IF_NONE:
649         dinfo->media_cd = media == MEDIA_CDROM;
650         break;
651     case IF_SD:
652     case IF_FLOPPY:
653     case IF_PFLASH:
654     case IF_MTD:
655         break;
656     case IF_VIRTIO:
657     {
658         /* add virtio block device */
659         QemuOpts *devopts;
660         devopts = qemu_opts_create_nofail(qemu_find_opts("device"));
661         if (arch_type == QEMU_ARCH_S390X) {
662             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
663         } else {
664             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
665         }
666         qemu_opt_set(devopts, "drive", dinfo->id);
667         if (devaddr)
668             qemu_opt_set(devopts, "addr", devaddr);
669         break;
670     }
671     default:
672         abort();
673     }
674     if (!file || !*file) {
675         if (has_driver_specific_opts) {
676             file = NULL;
677         } else {
678             return dinfo;
679         }
680     }
681     if (snapshot) {
682         /* always use cache=unsafe with snapshot */
683         bdrv_flags &= ~BDRV_O_CACHE_MASK;
684         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
685     }
686 
687     if (copy_on_read) {
688         bdrv_flags |= BDRV_O_COPY_ON_READ;
689     }
690 
691     if (runstate_check(RUN_STATE_INMIGRATE)) {
692         bdrv_flags |= BDRV_O_INCOMING;
693     }
694 
695     if (media == MEDIA_CDROM) {
696         /* CDROM is fine for any interface, don't check.  */
697         ro = 1;
698     } else if (ro == 1) {
699         if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
700             type != IF_NONE && type != IF_PFLASH) {
701             error_report("read-only not supported by this bus type");
702             goto err;
703         }
704     }
705 
706     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
707 
708     if (ro && copy_on_read) {
709         error_report("warning: disabling copy_on_read on read-only drive");
710     }
711 
712     QINCREF(bs_opts);
713     ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, drv);
714 
715     if (ret < 0) {
716         if (ret == -EMEDIUMTYPE) {
717             error_report("could not open disk image %s: not in %s format",
718                          file ?: dinfo->id, drv ? drv->format_name :
719                          qdict_get_str(bs_opts, "driver"));
720         } else {
721             error_report("could not open disk image %s: %s",
722                          file ?: dinfo->id, strerror(-ret));
723         }
724         goto err;
725     }
726 
727     if (bdrv_key_required(dinfo->bdrv))
728         autostart = 0;
729 
730     QDECREF(bs_opts);
731     qemu_opts_del(opts);
732 
733     return dinfo;
734 
735 err:
736     qemu_opts_del(opts);
737     QDECREF(bs_opts);
738     bdrv_unref(dinfo->bdrv);
739     g_free(dinfo->id);
740     QTAILQ_REMOVE(&drives, dinfo, next);
741     g_free(dinfo);
742     return NULL;
743 }
744 
745 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
746 {
747     const char *value;
748 
749     value = qemu_opt_get(opts, from);
750     if (value) {
751         qemu_opt_set(opts, to, value);
752         qemu_opt_unset(opts, from);
753     }
754 }
755 
756 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
757 {
758     const char *value;
759 
760     /* Change legacy command line options into QMP ones */
761     qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
762     qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
763     qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
764 
765     qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
766     qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
767     qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
768 
769     qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
770     qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
771     qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
772 
773     qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
774     qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
775     qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
776 
777     qemu_opt_rename(all_opts,
778                     "iops_size", "throttling.iops-size");
779 
780     qemu_opt_rename(all_opts, "readonly", "read-only");
781 
782     value = qemu_opt_get(all_opts, "cache");
783     if (value) {
784         int flags = 0;
785 
786         if (bdrv_parse_cache_flags(value, &flags) != 0) {
787             error_report("invalid cache option");
788             return NULL;
789         }
790 
791         /* Specific options take precedence */
792         if (!qemu_opt_get(all_opts, "cache.writeback")) {
793             qemu_opt_set_bool(all_opts, "cache.writeback",
794                               !!(flags & BDRV_O_CACHE_WB));
795         }
796         if (!qemu_opt_get(all_opts, "cache.direct")) {
797             qemu_opt_set_bool(all_opts, "cache.direct",
798                               !!(flags & BDRV_O_NOCACHE));
799         }
800         if (!qemu_opt_get(all_opts, "cache.no-flush")) {
801             qemu_opt_set_bool(all_opts, "cache.no-flush",
802                               !!(flags & BDRV_O_NO_FLUSH));
803         }
804         qemu_opt_unset(all_opts, "cache");
805     }
806 
807     return blockdev_init(all_opts, block_default_type);
808 }
809 
810 void do_commit(Monitor *mon, const QDict *qdict)
811 {
812     const char *device = qdict_get_str(qdict, "device");
813     BlockDriverState *bs;
814     int ret;
815 
816     if (!strcmp(device, "all")) {
817         ret = bdrv_commit_all();
818     } else {
819         bs = bdrv_find(device);
820         if (!bs) {
821             monitor_printf(mon, "Device '%s' not found\n", device);
822             return;
823         }
824         ret = bdrv_commit(bs);
825     }
826     if (ret < 0) {
827         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
828                        strerror(-ret));
829     }
830 }
831 
832 static void blockdev_do_action(int kind, void *data, Error **errp)
833 {
834     TransactionAction action;
835     TransactionActionList list;
836 
837     action.kind = kind;
838     action.data = data;
839     list.value = &action;
840     list.next = NULL;
841     qmp_transaction(&list, errp);
842 }
843 
844 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
845                                 bool has_format, const char *format,
846                                 bool has_mode, enum NewImageMode mode,
847                                 Error **errp)
848 {
849     BlockdevSnapshot snapshot = {
850         .device = (char *) device,
851         .snapshot_file = (char *) snapshot_file,
852         .has_format = has_format,
853         .format = (char *) format,
854         .has_mode = has_mode,
855         .mode = mode,
856     };
857     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
858                        &snapshot, errp);
859 }
860 
861 
862 /* New and old BlockDriverState structs for group snapshots */
863 
864 typedef struct BlkTransactionState BlkTransactionState;
865 
866 /* Only prepare() may fail. In a single transaction, only one of commit() or
867    abort() will be called, clean() will always be called if it present. */
868 typedef struct BdrvActionOps {
869     /* Size of state struct, in bytes. */
870     size_t instance_size;
871     /* Prepare the work, must NOT be NULL. */
872     void (*prepare)(BlkTransactionState *common, Error **errp);
873     /* Commit the changes, can be NULL. */
874     void (*commit)(BlkTransactionState *common);
875     /* Abort the changes on fail, can be NULL. */
876     void (*abort)(BlkTransactionState *common);
877     /* Clean up resource in the end, can be NULL. */
878     void (*clean)(BlkTransactionState *common);
879 } BdrvActionOps;
880 
881 /*
882  * This structure must be arranged as first member in child type, assuming
883  * that compiler will also arrange it to the same address with parent instance.
884  * Later it will be used in free().
885  */
886 struct BlkTransactionState {
887     TransactionAction *action;
888     const BdrvActionOps *ops;
889     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
890 };
891 
892 /* external snapshot private data */
893 typedef struct ExternalSnapshotState {
894     BlkTransactionState common;
895     BlockDriverState *old_bs;
896     BlockDriverState *new_bs;
897 } ExternalSnapshotState;
898 
899 static void external_snapshot_prepare(BlkTransactionState *common,
900                                       Error **errp)
901 {
902     BlockDriver *drv;
903     int flags, ret;
904     Error *local_err = NULL;
905     const char *device;
906     const char *new_image_file;
907     const char *format = "qcow2";
908     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
909     ExternalSnapshotState *state =
910                              DO_UPCAST(ExternalSnapshotState, common, common);
911     TransactionAction *action = common->action;
912 
913     /* get parameters */
914     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
915 
916     device = action->blockdev_snapshot_sync->device;
917     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
918     if (action->blockdev_snapshot_sync->has_format) {
919         format = action->blockdev_snapshot_sync->format;
920     }
921     if (action->blockdev_snapshot_sync->has_mode) {
922         mode = action->blockdev_snapshot_sync->mode;
923     }
924 
925     /* start processing */
926     drv = bdrv_find_format(format);
927     if (!drv) {
928         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
929         return;
930     }
931 
932     state->old_bs = bdrv_find(device);
933     if (!state->old_bs) {
934         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
935         return;
936     }
937 
938     if (!bdrv_is_inserted(state->old_bs)) {
939         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
940         return;
941     }
942 
943     if (bdrv_in_use(state->old_bs)) {
944         error_set(errp, QERR_DEVICE_IN_USE, device);
945         return;
946     }
947 
948     if (!bdrv_is_read_only(state->old_bs)) {
949         if (bdrv_flush(state->old_bs)) {
950             error_set(errp, QERR_IO_ERROR);
951             return;
952         }
953     }
954 
955     flags = state->old_bs->open_flags;
956 
957     /* create new image w/backing file */
958     if (mode != NEW_IMAGE_MODE_EXISTING) {
959         bdrv_img_create(new_image_file, format,
960                         state->old_bs->filename,
961                         state->old_bs->drv->format_name,
962                         NULL, -1, flags, &local_err, false);
963         if (error_is_set(&local_err)) {
964             error_propagate(errp, local_err);
965             return;
966         }
967     }
968 
969     /* We will manually add the backing_hd field to the bs later */
970     state->new_bs = bdrv_new("");
971     /* TODO Inherit bs->options or only take explicit options with an
972      * extended QMP command? */
973     ret = bdrv_open(state->new_bs, new_image_file, NULL,
974                     flags | BDRV_O_NO_BACKING, drv);
975     if (ret != 0) {
976         error_setg_file_open(errp, -ret, new_image_file);
977     }
978 }
979 
980 static void external_snapshot_commit(BlkTransactionState *common)
981 {
982     ExternalSnapshotState *state =
983                              DO_UPCAST(ExternalSnapshotState, common, common);
984 
985     /* This removes our old bs and adds the new bs */
986     bdrv_append(state->new_bs, state->old_bs);
987     /* We don't need (or want) to use the transactional
988      * bdrv_reopen_multiple() across all the entries at once, because we
989      * don't want to abort all of them if one of them fails the reopen */
990     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
991                 NULL);
992 }
993 
994 static void external_snapshot_abort(BlkTransactionState *common)
995 {
996     ExternalSnapshotState *state =
997                              DO_UPCAST(ExternalSnapshotState, common, common);
998     if (state->new_bs) {
999         bdrv_unref(state->new_bs);
1000     }
1001 }
1002 
1003 typedef struct DriveBackupState {
1004     BlkTransactionState common;
1005     BlockDriverState *bs;
1006     BlockJob *job;
1007 } DriveBackupState;
1008 
1009 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1010 {
1011     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1012     DriveBackup *backup;
1013     Error *local_err = NULL;
1014 
1015     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1016     backup = common->action->drive_backup;
1017 
1018     qmp_drive_backup(backup->device, backup->target,
1019                      backup->has_format, backup->format,
1020                      backup->sync,
1021                      backup->has_mode, backup->mode,
1022                      backup->has_speed, backup->speed,
1023                      backup->has_on_source_error, backup->on_source_error,
1024                      backup->has_on_target_error, backup->on_target_error,
1025                      &local_err);
1026     if (error_is_set(&local_err)) {
1027         error_propagate(errp, local_err);
1028         state->bs = NULL;
1029         state->job = NULL;
1030         return;
1031     }
1032 
1033     state->bs = bdrv_find(backup->device);
1034     state->job = state->bs->job;
1035 }
1036 
1037 static void drive_backup_abort(BlkTransactionState *common)
1038 {
1039     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1040     BlockDriverState *bs = state->bs;
1041 
1042     /* Only cancel if it's the job we started */
1043     if (bs && bs->job && bs->job == state->job) {
1044         block_job_cancel_sync(bs->job);
1045     }
1046 }
1047 
1048 static void abort_prepare(BlkTransactionState *common, Error **errp)
1049 {
1050     error_setg(errp, "Transaction aborted using Abort action");
1051 }
1052 
1053 static void abort_commit(BlkTransactionState *common)
1054 {
1055     g_assert_not_reached(); /* this action never succeeds */
1056 }
1057 
1058 static const BdrvActionOps actions[] = {
1059     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1060         .instance_size = sizeof(ExternalSnapshotState),
1061         .prepare  = external_snapshot_prepare,
1062         .commit   = external_snapshot_commit,
1063         .abort = external_snapshot_abort,
1064     },
1065     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1066         .instance_size = sizeof(DriveBackupState),
1067         .prepare = drive_backup_prepare,
1068         .abort = drive_backup_abort,
1069     },
1070     [TRANSACTION_ACTION_KIND_ABORT] = {
1071         .instance_size = sizeof(BlkTransactionState),
1072         .prepare = abort_prepare,
1073         .commit = abort_commit,
1074     },
1075 };
1076 
1077 /*
1078  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
1079  *  then we do not pivot any of the devices in the group, and abandon the
1080  *  snapshots
1081  */
1082 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1083 {
1084     TransactionActionList *dev_entry = dev_list;
1085     BlkTransactionState *state, *next;
1086     Error *local_err = NULL;
1087 
1088     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1089     QSIMPLEQ_INIT(&snap_bdrv_states);
1090 
1091     /* drain all i/o before any snapshots */
1092     bdrv_drain_all();
1093 
1094     /* We don't do anything in this loop that commits us to the snapshot */
1095     while (NULL != dev_entry) {
1096         TransactionAction *dev_info = NULL;
1097         const BdrvActionOps *ops;
1098 
1099         dev_info = dev_entry->value;
1100         dev_entry = dev_entry->next;
1101 
1102         assert(dev_info->kind < ARRAY_SIZE(actions));
1103 
1104         ops = &actions[dev_info->kind];
1105         state = g_malloc0(ops->instance_size);
1106         state->ops = ops;
1107         state->action = dev_info;
1108         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1109 
1110         state->ops->prepare(state, &local_err);
1111         if (error_is_set(&local_err)) {
1112             error_propagate(errp, local_err);
1113             goto delete_and_fail;
1114         }
1115     }
1116 
1117     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1118         if (state->ops->commit) {
1119             state->ops->commit(state);
1120         }
1121     }
1122 
1123     /* success */
1124     goto exit;
1125 
1126 delete_and_fail:
1127     /*
1128     * failure, and it is all-or-none; abandon each new bs, and keep using
1129     * the original bs for all images
1130     */
1131     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1132         if (state->ops->abort) {
1133             state->ops->abort(state);
1134         }
1135     }
1136 exit:
1137     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1138         if (state->ops->clean) {
1139             state->ops->clean(state);
1140         }
1141         g_free(state);
1142     }
1143 }
1144 
1145 
1146 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1147 {
1148     if (bdrv_in_use(bs)) {
1149         error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1150         return;
1151     }
1152     if (!bdrv_dev_has_removable_media(bs)) {
1153         error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1154         return;
1155     }
1156 
1157     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1158         bdrv_dev_eject_request(bs, force);
1159         if (!force) {
1160             error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1161             return;
1162         }
1163     }
1164 
1165     bdrv_close(bs);
1166 }
1167 
1168 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1169 {
1170     BlockDriverState *bs;
1171 
1172     bs = bdrv_find(device);
1173     if (!bs) {
1174         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1175         return;
1176     }
1177 
1178     eject_device(bs, force, errp);
1179 }
1180 
1181 void qmp_block_passwd(const char *device, const char *password, Error **errp)
1182 {
1183     BlockDriverState *bs;
1184     int err;
1185 
1186     bs = bdrv_find(device);
1187     if (!bs) {
1188         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1189         return;
1190     }
1191 
1192     err = bdrv_set_key(bs, password);
1193     if (err == -EINVAL) {
1194         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1195         return;
1196     } else if (err < 0) {
1197         error_set(errp, QERR_INVALID_PASSWORD);
1198         return;
1199     }
1200 }
1201 
1202 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1203                                     int bdrv_flags, BlockDriver *drv,
1204                                     const char *password, Error **errp)
1205 {
1206     int ret;
1207 
1208     ret = bdrv_open(bs, filename, NULL, bdrv_flags, drv);
1209     if (ret < 0) {
1210         error_setg_file_open(errp, -ret, filename);
1211         return;
1212     }
1213 
1214     if (bdrv_key_required(bs)) {
1215         if (password) {
1216             if (bdrv_set_key(bs, password) < 0) {
1217                 error_set(errp, QERR_INVALID_PASSWORD);
1218             }
1219         } else {
1220             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1221                       bdrv_get_encrypted_filename(bs));
1222         }
1223     } else if (password) {
1224         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1225     }
1226 }
1227 
1228 void qmp_change_blockdev(const char *device, const char *filename,
1229                          bool has_format, const char *format, Error **errp)
1230 {
1231     BlockDriverState *bs;
1232     BlockDriver *drv = NULL;
1233     int bdrv_flags;
1234     Error *err = NULL;
1235 
1236     bs = bdrv_find(device);
1237     if (!bs) {
1238         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1239         return;
1240     }
1241 
1242     if (format) {
1243         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1244         if (!drv) {
1245             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1246             return;
1247         }
1248     }
1249 
1250     eject_device(bs, 0, &err);
1251     if (error_is_set(&err)) {
1252         error_propagate(errp, err);
1253         return;
1254     }
1255 
1256     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1257     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1258 
1259     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1260 }
1261 
1262 /* throttling disk I/O limits */
1263 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1264                                int64_t bps_wr,
1265                                int64_t iops,
1266                                int64_t iops_rd,
1267                                int64_t iops_wr,
1268                                bool has_bps_max,
1269                                int64_t bps_max,
1270                                bool has_bps_rd_max,
1271                                int64_t bps_rd_max,
1272                                bool has_bps_wr_max,
1273                                int64_t bps_wr_max,
1274                                bool has_iops_max,
1275                                int64_t iops_max,
1276                                bool has_iops_rd_max,
1277                                int64_t iops_rd_max,
1278                                bool has_iops_wr_max,
1279                                int64_t iops_wr_max,
1280                                bool has_iops_size,
1281                                int64_t iops_size, Error **errp)
1282 {
1283     ThrottleConfig cfg;
1284     BlockDriverState *bs;
1285 
1286     bs = bdrv_find(device);
1287     if (!bs) {
1288         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1289         return;
1290     }
1291 
1292     memset(&cfg, 0, sizeof(cfg));
1293     cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1294     cfg.buckets[THROTTLE_BPS_READ].avg  = bps_rd;
1295     cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1296 
1297     cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1298     cfg.buckets[THROTTLE_OPS_READ].avg  = iops_rd;
1299     cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1300 
1301     if (has_bps_max) {
1302         cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1303     }
1304     if (has_bps_rd_max) {
1305         cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1306     }
1307     if (has_bps_wr_max) {
1308         cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1309     }
1310     if (has_iops_max) {
1311         cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1312     }
1313     if (has_iops_rd_max) {
1314         cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1315     }
1316     if (has_iops_wr_max) {
1317         cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1318     }
1319 
1320     if (has_iops_size) {
1321         cfg.op_size = iops_size;
1322     }
1323 
1324     if (!check_throttle_config(&cfg, errp)) {
1325         return;
1326     }
1327 
1328     if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1329         bdrv_io_limits_enable(bs);
1330     } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1331         bdrv_io_limits_disable(bs);
1332     }
1333 
1334     if (bs->io_limits_enabled) {
1335         bdrv_set_io_limits(bs, &cfg);
1336     }
1337 }
1338 
1339 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1340 {
1341     const char *id = qdict_get_str(qdict, "id");
1342     BlockDriverState *bs;
1343 
1344     bs = bdrv_find(id);
1345     if (!bs) {
1346         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1347         return -1;
1348     }
1349     if (bdrv_in_use(bs)) {
1350         qerror_report(QERR_DEVICE_IN_USE, id);
1351         return -1;
1352     }
1353 
1354     /* quiesce block driver; prevent further io */
1355     bdrv_drain_all();
1356     bdrv_flush(bs);
1357     bdrv_close(bs);
1358 
1359     /* if we have a device attached to this BlockDriverState
1360      * then we need to make the drive anonymous until the device
1361      * can be removed.  If this is a drive with no device backing
1362      * then we can just get rid of the block driver state right here.
1363      */
1364     if (bdrv_get_attached_dev(bs)) {
1365         bdrv_make_anon(bs);
1366 
1367         /* Further I/O must not pause the guest */
1368         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1369                           BLOCKDEV_ON_ERROR_REPORT);
1370     } else {
1371         drive_uninit(drive_get_by_blockdev(bs));
1372     }
1373 
1374     return 0;
1375 }
1376 
1377 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1378 {
1379     BlockDriverState *bs;
1380     int ret;
1381 
1382     bs = bdrv_find(device);
1383     if (!bs) {
1384         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1385         return;
1386     }
1387 
1388     if (size < 0) {
1389         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1390         return;
1391     }
1392 
1393     /* complete all in-flight operations before resizing the device */
1394     bdrv_drain_all();
1395 
1396     ret = bdrv_truncate(bs, size);
1397     switch (ret) {
1398     case 0:
1399         break;
1400     case -ENOMEDIUM:
1401         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1402         break;
1403     case -ENOTSUP:
1404         error_set(errp, QERR_UNSUPPORTED);
1405         break;
1406     case -EACCES:
1407         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1408         break;
1409     case -EBUSY:
1410         error_set(errp, QERR_DEVICE_IN_USE, device);
1411         break;
1412     default:
1413         error_setg_errno(errp, -ret, "Could not resize");
1414         break;
1415     }
1416 }
1417 
1418 static void block_job_cb(void *opaque, int ret)
1419 {
1420     BlockDriverState *bs = opaque;
1421     QObject *obj;
1422 
1423     trace_block_job_cb(bs, bs->job, ret);
1424 
1425     assert(bs->job);
1426     obj = qobject_from_block_job(bs->job);
1427     if (ret < 0) {
1428         QDict *dict = qobject_to_qdict(obj);
1429         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1430     }
1431 
1432     if (block_job_is_cancelled(bs->job)) {
1433         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1434     } else {
1435         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1436     }
1437     qobject_decref(obj);
1438 
1439     bdrv_put_ref_bh_schedule(bs);
1440 }
1441 
1442 void qmp_block_stream(const char *device, bool has_base,
1443                       const char *base, bool has_speed, int64_t speed,
1444                       bool has_on_error, BlockdevOnError on_error,
1445                       Error **errp)
1446 {
1447     BlockDriverState *bs;
1448     BlockDriverState *base_bs = NULL;
1449     Error *local_err = NULL;
1450 
1451     if (!has_on_error) {
1452         on_error = BLOCKDEV_ON_ERROR_REPORT;
1453     }
1454 
1455     bs = bdrv_find(device);
1456     if (!bs) {
1457         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1458         return;
1459     }
1460 
1461     if (base) {
1462         base_bs = bdrv_find_backing_image(bs, base);
1463         if (base_bs == NULL) {
1464             error_set(errp, QERR_BASE_NOT_FOUND, base);
1465             return;
1466         }
1467     }
1468 
1469     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1470                  on_error, block_job_cb, bs, &local_err);
1471     if (error_is_set(&local_err)) {
1472         error_propagate(errp, local_err);
1473         return;
1474     }
1475 
1476     trace_qmp_block_stream(bs, bs->job);
1477 }
1478 
1479 void qmp_block_commit(const char *device,
1480                       bool has_base, const char *base, const char *top,
1481                       bool has_speed, int64_t speed,
1482                       Error **errp)
1483 {
1484     BlockDriverState *bs;
1485     BlockDriverState *base_bs, *top_bs;
1486     Error *local_err = NULL;
1487     /* This will be part of the QMP command, if/when the
1488      * BlockdevOnError change for blkmirror makes it in
1489      */
1490     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1491 
1492     /* drain all i/o before commits */
1493     bdrv_drain_all();
1494 
1495     bs = bdrv_find(device);
1496     if (!bs) {
1497         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1498         return;
1499     }
1500 
1501     /* default top_bs is the active layer */
1502     top_bs = bs;
1503 
1504     if (top) {
1505         if (strcmp(bs->filename, top) != 0) {
1506             top_bs = bdrv_find_backing_image(bs, top);
1507         }
1508     }
1509 
1510     if (top_bs == NULL) {
1511         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1512         return;
1513     }
1514 
1515     if (has_base && base) {
1516         base_bs = bdrv_find_backing_image(top_bs, base);
1517     } else {
1518         base_bs = bdrv_find_base(top_bs);
1519     }
1520 
1521     if (base_bs == NULL) {
1522         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1523         return;
1524     }
1525 
1526     commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1527                 &local_err);
1528     if (local_err != NULL) {
1529         error_propagate(errp, local_err);
1530         return;
1531     }
1532 }
1533 
1534 void qmp_drive_backup(const char *device, const char *target,
1535                       bool has_format, const char *format,
1536                       enum MirrorSyncMode sync,
1537                       bool has_mode, enum NewImageMode mode,
1538                       bool has_speed, int64_t speed,
1539                       bool has_on_source_error, BlockdevOnError on_source_error,
1540                       bool has_on_target_error, BlockdevOnError on_target_error,
1541                       Error **errp)
1542 {
1543     BlockDriverState *bs;
1544     BlockDriverState *target_bs;
1545     BlockDriverState *source = NULL;
1546     BlockDriver *drv = NULL;
1547     Error *local_err = NULL;
1548     int flags;
1549     int64_t size;
1550     int ret;
1551 
1552     if (!has_speed) {
1553         speed = 0;
1554     }
1555     if (!has_on_source_error) {
1556         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1557     }
1558     if (!has_on_target_error) {
1559         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1560     }
1561     if (!has_mode) {
1562         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1563     }
1564 
1565     bs = bdrv_find(device);
1566     if (!bs) {
1567         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1568         return;
1569     }
1570 
1571     if (!bdrv_is_inserted(bs)) {
1572         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1573         return;
1574     }
1575 
1576     if (!has_format) {
1577         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1578     }
1579     if (format) {
1580         drv = bdrv_find_format(format);
1581         if (!drv) {
1582             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1583             return;
1584         }
1585     }
1586 
1587     if (bdrv_in_use(bs)) {
1588         error_set(errp, QERR_DEVICE_IN_USE, device);
1589         return;
1590     }
1591 
1592     flags = bs->open_flags | BDRV_O_RDWR;
1593 
1594     /* See if we have a backing HD we can use to create our new image
1595      * on top of. */
1596     if (sync == MIRROR_SYNC_MODE_TOP) {
1597         source = bs->backing_hd;
1598         if (!source) {
1599             sync = MIRROR_SYNC_MODE_FULL;
1600         }
1601     }
1602     if (sync == MIRROR_SYNC_MODE_NONE) {
1603         source = bs;
1604     }
1605 
1606     size = bdrv_getlength(bs);
1607     if (size < 0) {
1608         error_setg_errno(errp, -size, "bdrv_getlength failed");
1609         return;
1610     }
1611 
1612     if (mode != NEW_IMAGE_MODE_EXISTING) {
1613         assert(format && drv);
1614         if (source) {
1615             bdrv_img_create(target, format, source->filename,
1616                             source->drv->format_name, NULL,
1617                             size, flags, &local_err, false);
1618         } else {
1619             bdrv_img_create(target, format, NULL, NULL, NULL,
1620                             size, flags, &local_err, false);
1621         }
1622     }
1623 
1624     if (error_is_set(&local_err)) {
1625         error_propagate(errp, local_err);
1626         return;
1627     }
1628 
1629     target_bs = bdrv_new("");
1630     ret = bdrv_open(target_bs, target, NULL, flags, drv);
1631     if (ret < 0) {
1632         bdrv_unref(target_bs);
1633         error_setg_file_open(errp, -ret, target);
1634         return;
1635     }
1636 
1637     backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
1638                  block_job_cb, bs, &local_err);
1639     if (local_err != NULL) {
1640         bdrv_unref(target_bs);
1641         error_propagate(errp, local_err);
1642         return;
1643     }
1644 }
1645 
1646 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
1647 
1648 void qmp_drive_mirror(const char *device, const char *target,
1649                       bool has_format, const char *format,
1650                       enum MirrorSyncMode sync,
1651                       bool has_mode, enum NewImageMode mode,
1652                       bool has_speed, int64_t speed,
1653                       bool has_granularity, uint32_t granularity,
1654                       bool has_buf_size, int64_t buf_size,
1655                       bool has_on_source_error, BlockdevOnError on_source_error,
1656                       bool has_on_target_error, BlockdevOnError on_target_error,
1657                       Error **errp)
1658 {
1659     BlockDriverState *bs;
1660     BlockDriverState *source, *target_bs;
1661     BlockDriver *drv = NULL;
1662     Error *local_err = NULL;
1663     int flags;
1664     int64_t size;
1665     int ret;
1666 
1667     if (!has_speed) {
1668         speed = 0;
1669     }
1670     if (!has_on_source_error) {
1671         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1672     }
1673     if (!has_on_target_error) {
1674         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1675     }
1676     if (!has_mode) {
1677         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1678     }
1679     if (!has_granularity) {
1680         granularity = 0;
1681     }
1682     if (!has_buf_size) {
1683         buf_size = DEFAULT_MIRROR_BUF_SIZE;
1684     }
1685 
1686     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
1687         error_set(errp, QERR_INVALID_PARAMETER, device);
1688         return;
1689     }
1690     if (granularity & (granularity - 1)) {
1691         error_set(errp, QERR_INVALID_PARAMETER, device);
1692         return;
1693     }
1694 
1695     bs = bdrv_find(device);
1696     if (!bs) {
1697         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1698         return;
1699     }
1700 
1701     if (!bdrv_is_inserted(bs)) {
1702         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1703         return;
1704     }
1705 
1706     if (!has_format) {
1707         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1708     }
1709     if (format) {
1710         drv = bdrv_find_format(format);
1711         if (!drv) {
1712             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1713             return;
1714         }
1715     }
1716 
1717     if (bdrv_in_use(bs)) {
1718         error_set(errp, QERR_DEVICE_IN_USE, device);
1719         return;
1720     }
1721 
1722     flags = bs->open_flags | BDRV_O_RDWR;
1723     source = bs->backing_hd;
1724     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1725         sync = MIRROR_SYNC_MODE_FULL;
1726     }
1727 
1728     size = bdrv_getlength(bs);
1729     if (size < 0) {
1730         error_setg_errno(errp, -size, "bdrv_getlength failed");
1731         return;
1732     }
1733 
1734     if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1735         /* create new image w/o backing file */
1736         assert(format && drv);
1737         bdrv_img_create(target, format,
1738                         NULL, NULL, NULL, size, flags, &local_err, false);
1739     } else {
1740         switch (mode) {
1741         case NEW_IMAGE_MODE_EXISTING:
1742             ret = 0;
1743             break;
1744         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1745             /* create new image with backing file */
1746             bdrv_img_create(target, format,
1747                             source->filename,
1748                             source->drv->format_name,
1749                             NULL, size, flags, &local_err, false);
1750             break;
1751         default:
1752             abort();
1753         }
1754     }
1755 
1756     if (error_is_set(&local_err)) {
1757         error_propagate(errp, local_err);
1758         return;
1759     }
1760 
1761     /* Mirroring takes care of copy-on-write using the source's backing
1762      * file.
1763      */
1764     target_bs = bdrv_new("");
1765     ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv);
1766     if (ret < 0) {
1767         bdrv_unref(target_bs);
1768         error_setg_file_open(errp, -ret, target);
1769         return;
1770     }
1771 
1772     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
1773                  on_source_error, on_target_error,
1774                  block_job_cb, bs, &local_err);
1775     if (local_err != NULL) {
1776         bdrv_unref(target_bs);
1777         error_propagate(errp, local_err);
1778         return;
1779     }
1780 }
1781 
1782 static BlockJob *find_block_job(const char *device)
1783 {
1784     BlockDriverState *bs;
1785 
1786     bs = bdrv_find(device);
1787     if (!bs || !bs->job) {
1788         return NULL;
1789     }
1790     return bs->job;
1791 }
1792 
1793 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1794 {
1795     BlockJob *job = find_block_job(device);
1796 
1797     if (!job) {
1798         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1799         return;
1800     }
1801 
1802     block_job_set_speed(job, speed, errp);
1803 }
1804 
1805 void qmp_block_job_cancel(const char *device,
1806                           bool has_force, bool force, Error **errp)
1807 {
1808     BlockJob *job = find_block_job(device);
1809 
1810     if (!has_force) {
1811         force = false;
1812     }
1813 
1814     if (!job) {
1815         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1816         return;
1817     }
1818     if (job->paused && !force) {
1819         error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1820         return;
1821     }
1822 
1823     trace_qmp_block_job_cancel(job);
1824     block_job_cancel(job);
1825 }
1826 
1827 void qmp_block_job_pause(const char *device, Error **errp)
1828 {
1829     BlockJob *job = find_block_job(device);
1830 
1831     if (!job) {
1832         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1833         return;
1834     }
1835 
1836     trace_qmp_block_job_pause(job);
1837     block_job_pause(job);
1838 }
1839 
1840 void qmp_block_job_resume(const char *device, Error **errp)
1841 {
1842     BlockJob *job = find_block_job(device);
1843 
1844     if (!job) {
1845         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1846         return;
1847     }
1848 
1849     trace_qmp_block_job_resume(job);
1850     block_job_resume(job);
1851 }
1852 
1853 void qmp_block_job_complete(const char *device, Error **errp)
1854 {
1855     BlockJob *job = find_block_job(device);
1856 
1857     if (!job) {
1858         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1859         return;
1860     }
1861 
1862     trace_qmp_block_job_complete(job);
1863     block_job_complete(job, errp);
1864 }
1865 
1866 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1867 {
1868     BlockJobInfoList **prev = opaque;
1869     BlockJob *job = bs->job;
1870 
1871     if (job) {
1872         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1873         elem->value = block_job_query(bs->job);
1874         (*prev)->next = elem;
1875         *prev = elem;
1876     }
1877 }
1878 
1879 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1880 {
1881     /* Dummy is a fake list element for holding the head pointer */
1882     BlockJobInfoList dummy = {};
1883     BlockJobInfoList *prev = &dummy;
1884     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1885     return dummy.next;
1886 }
1887 
1888 QemuOptsList qemu_common_drive_opts = {
1889     .name = "drive",
1890     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
1891     .desc = {
1892         {
1893             .name = "bus",
1894             .type = QEMU_OPT_NUMBER,
1895             .help = "bus number",
1896         },{
1897             .name = "unit",
1898             .type = QEMU_OPT_NUMBER,
1899             .help = "unit number (i.e. lun for scsi)",
1900         },{
1901             .name = "if",
1902             .type = QEMU_OPT_STRING,
1903             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1904         },{
1905             .name = "index",
1906             .type = QEMU_OPT_NUMBER,
1907             .help = "index number",
1908         },{
1909             .name = "cyls",
1910             .type = QEMU_OPT_NUMBER,
1911             .help = "number of cylinders (ide disk geometry)",
1912         },{
1913             .name = "heads",
1914             .type = QEMU_OPT_NUMBER,
1915             .help = "number of heads (ide disk geometry)",
1916         },{
1917             .name = "secs",
1918             .type = QEMU_OPT_NUMBER,
1919             .help = "number of sectors (ide disk geometry)",
1920         },{
1921             .name = "trans",
1922             .type = QEMU_OPT_STRING,
1923             .help = "chs translation (auto, lba. none)",
1924         },{
1925             .name = "media",
1926             .type = QEMU_OPT_STRING,
1927             .help = "media type (disk, cdrom)",
1928         },{
1929             .name = "snapshot",
1930             .type = QEMU_OPT_BOOL,
1931             .help = "enable/disable snapshot mode",
1932         },{
1933             .name = "file",
1934             .type = QEMU_OPT_STRING,
1935             .help = "disk image",
1936         },{
1937             .name = "discard",
1938             .type = QEMU_OPT_STRING,
1939             .help = "discard operation (ignore/off, unmap/on)",
1940         },{
1941             .name = "cache.writeback",
1942             .type = QEMU_OPT_BOOL,
1943             .help = "enables writeback mode for any caches",
1944         },{
1945             .name = "cache.direct",
1946             .type = QEMU_OPT_BOOL,
1947             .help = "enables use of O_DIRECT (bypass the host page cache)",
1948         },{
1949             .name = "cache.no-flush",
1950             .type = QEMU_OPT_BOOL,
1951             .help = "ignore any flush requests for the device",
1952         },{
1953             .name = "aio",
1954             .type = QEMU_OPT_STRING,
1955             .help = "host AIO implementation (threads, native)",
1956         },{
1957             .name = "format",
1958             .type = QEMU_OPT_STRING,
1959             .help = "disk format (raw, qcow2, ...)",
1960         },{
1961             .name = "serial",
1962             .type = QEMU_OPT_STRING,
1963             .help = "disk serial number",
1964         },{
1965             .name = "rerror",
1966             .type = QEMU_OPT_STRING,
1967             .help = "read error action",
1968         },{
1969             .name = "werror",
1970             .type = QEMU_OPT_STRING,
1971             .help = "write error action",
1972         },{
1973             .name = "addr",
1974             .type = QEMU_OPT_STRING,
1975             .help = "pci address (virtio only)",
1976         },{
1977             .name = "read-only",
1978             .type = QEMU_OPT_BOOL,
1979             .help = "open drive file as read-only",
1980         },{
1981             .name = "throttling.iops-total",
1982             .type = QEMU_OPT_NUMBER,
1983             .help = "limit total I/O operations per second",
1984         },{
1985             .name = "throttling.iops-read",
1986             .type = QEMU_OPT_NUMBER,
1987             .help = "limit read operations per second",
1988         },{
1989             .name = "throttling.iops-write",
1990             .type = QEMU_OPT_NUMBER,
1991             .help = "limit write operations per second",
1992         },{
1993             .name = "throttling.bps-total",
1994             .type = QEMU_OPT_NUMBER,
1995             .help = "limit total bytes per second",
1996         },{
1997             .name = "throttling.bps-read",
1998             .type = QEMU_OPT_NUMBER,
1999             .help = "limit read bytes per second",
2000         },{
2001             .name = "throttling.bps-write",
2002             .type = QEMU_OPT_NUMBER,
2003             .help = "limit write bytes per second",
2004         },{
2005             .name = "throttling.iops-total-max",
2006             .type = QEMU_OPT_NUMBER,
2007             .help = "I/O operations burst",
2008         },{
2009             .name = "throttling.iops-read-max",
2010             .type = QEMU_OPT_NUMBER,
2011             .help = "I/O operations read burst",
2012         },{
2013             .name = "throttling.iops-write-max",
2014             .type = QEMU_OPT_NUMBER,
2015             .help = "I/O operations write burst",
2016         },{
2017             .name = "throttling.bps-total-max",
2018             .type = QEMU_OPT_NUMBER,
2019             .help = "total bytes burst",
2020         },{
2021             .name = "throttling.bps-read-max",
2022             .type = QEMU_OPT_NUMBER,
2023             .help = "total bytes read burst",
2024         },{
2025             .name = "throttling.bps-write-max",
2026             .type = QEMU_OPT_NUMBER,
2027             .help = "total bytes write burst",
2028         },{
2029             .name = "throttling.iops-size",
2030             .type = QEMU_OPT_NUMBER,
2031             .help = "when limiting by iops max size of an I/O in bytes",
2032         },{
2033             .name = "copy-on-read",
2034             .type = QEMU_OPT_BOOL,
2035             .help = "copy read data from backing file into image file",
2036         },{
2037             .name = "boot",
2038             .type = QEMU_OPT_BOOL,
2039             .help = "(deprecated, ignored)",
2040         },
2041         { /* end of list */ }
2042     },
2043 };
2044 
2045 QemuOptsList qemu_drive_opts = {
2046     .name = "drive",
2047     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2048     .desc = {
2049         /*
2050          * no elements => accept any params
2051          * validation will happen later
2052          */
2053         { /* end of list */ }
2054     },
2055 };
2056