xref: /openbmc/qemu/blockdev.c (revision 56983463)
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_delete(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     DriveInfo *dinfo;
238 } DrivePutRefBH;
239 
240 static void drive_put_ref_bh(void *opaque)
241 {
242     DrivePutRefBH *s = opaque;
243 
244     drive_put_ref(s->dinfo);
245     qemu_bh_delete(s->bh);
246     g_free(s);
247 }
248 
249 /*
250  * Release a drive reference in a BH
251  *
252  * It is not possible to use drive_put_ref() from a callback function when the
253  * callers still need the drive.  In such cases we schedule a BH to release the
254  * reference.
255  */
256 static void drive_put_ref_bh_schedule(DriveInfo *dinfo)
257 {
258     DrivePutRefBH *s;
259 
260     s = g_new(DrivePutRefBH, 1);
261     s->bh = qemu_bh_new(drive_put_ref_bh, s);
262     s->dinfo = dinfo;
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 do_check_io_limits(BlockIOLimit *io_limits, Error **errp)
284 {
285     bool bps_flag;
286     bool iops_flag;
287 
288     assert(io_limits);
289 
290     bps_flag  = (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] != 0)
291                  && ((io_limits->bps[BLOCK_IO_LIMIT_READ] != 0)
292                  || (io_limits->bps[BLOCK_IO_LIMIT_WRITE] != 0));
293     iops_flag = (io_limits->iops[BLOCK_IO_LIMIT_TOTAL] != 0)
294                  && ((io_limits->iops[BLOCK_IO_LIMIT_READ] != 0)
295                  || (io_limits->iops[BLOCK_IO_LIMIT_WRITE] != 0));
296     if (bps_flag || iops_flag) {
297         error_setg(errp, "bps(iops) and bps_rd/bps_wr(iops_rd/iops_wr) "
298                          "cannot be used at the same time");
299         return false;
300     }
301 
302     if (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] < 0 ||
303         io_limits->bps[BLOCK_IO_LIMIT_WRITE] < 0 ||
304         io_limits->bps[BLOCK_IO_LIMIT_READ] < 0 ||
305         io_limits->iops[BLOCK_IO_LIMIT_TOTAL] < 0 ||
306         io_limits->iops[BLOCK_IO_LIMIT_WRITE] < 0 ||
307         io_limits->iops[BLOCK_IO_LIMIT_READ] < 0) {
308         error_setg(errp, "bps and iops values must be 0 or greater");
309         return false;
310     }
311 
312     return true;
313 }
314 
315 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
316 {
317     const char *buf;
318     const char *file = NULL;
319     const char *serial;
320     const char *mediastr = "";
321     BlockInterfaceType type;
322     enum { MEDIA_DISK, MEDIA_CDROM } media;
323     int bus_id, unit_id;
324     int cyls, heads, secs, translation;
325     BlockDriver *drv = NULL;
326     int max_devs;
327     int index;
328     int ro = 0;
329     int bdrv_flags = 0;
330     int on_read_error, on_write_error;
331     const char *devaddr;
332     DriveInfo *dinfo;
333     BlockIOLimit io_limits;
334     int snapshot = 0;
335     bool copy_on_read;
336     int ret;
337     Error *error = NULL;
338     QemuOpts *opts;
339     QDict *bs_opts;
340     const char *id;
341 
342     translation = BIOS_ATA_TRANSLATION_AUTO;
343     media = MEDIA_DISK;
344 
345     /* Check common options by copying from all_opts to opts, all other options
346      * are stored in bs_opts. */
347     id = qemu_opts_id(all_opts);
348     opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
349     if (error_is_set(&error)) {
350         qerror_report_err(error);
351         error_free(error);
352         return NULL;
353     }
354 
355     bs_opts = qdict_new();
356     qemu_opts_to_qdict(all_opts, bs_opts);
357     qemu_opts_absorb_qdict(opts, bs_opts, &error);
358     if (error_is_set(&error)) {
359         qerror_report_err(error);
360         error_free(error);
361         return NULL;
362     }
363 
364     if (id) {
365         qdict_del(bs_opts, "id");
366     }
367 
368     /* extract parameters */
369     bus_id  = qemu_opt_get_number(opts, "bus", 0);
370     unit_id = qemu_opt_get_number(opts, "unit", -1);
371     index   = qemu_opt_get_number(opts, "index", -1);
372 
373     cyls  = qemu_opt_get_number(opts, "cyls", 0);
374     heads = qemu_opt_get_number(opts, "heads", 0);
375     secs  = qemu_opt_get_number(opts, "secs", 0);
376 
377     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
378     ro = qemu_opt_get_bool(opts, "readonly", 0);
379     copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
380 
381     file = qemu_opt_get(opts, "file");
382     serial = qemu_opt_get(opts, "serial");
383 
384     if ((buf = qemu_opt_get(opts, "if")) != NULL) {
385         for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
386             ;
387         if (type == IF_COUNT) {
388             error_report("unsupported bus type '%s'", buf);
389             return NULL;
390 	}
391     } else {
392         type = block_default_type;
393     }
394 
395     max_devs = if_max_devs[type];
396 
397     if (cyls || heads || secs) {
398         if (cyls < 1) {
399             error_report("invalid physical cyls number");
400 	    return NULL;
401 	}
402         if (heads < 1) {
403             error_report("invalid physical heads number");
404 	    return NULL;
405 	}
406         if (secs < 1) {
407             error_report("invalid physical secs number");
408 	    return NULL;
409 	}
410     }
411 
412     if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
413         if (!cyls) {
414             error_report("'%s' trans must be used with cyls, heads and secs",
415                          buf);
416             return NULL;
417         }
418         if (!strcmp(buf, "none"))
419             translation = BIOS_ATA_TRANSLATION_NONE;
420         else if (!strcmp(buf, "lba"))
421             translation = BIOS_ATA_TRANSLATION_LBA;
422         else if (!strcmp(buf, "auto"))
423             translation = BIOS_ATA_TRANSLATION_AUTO;
424 	else {
425             error_report("'%s' invalid translation type", buf);
426 	    return NULL;
427 	}
428     }
429 
430     if ((buf = qemu_opt_get(opts, "media")) != NULL) {
431         if (!strcmp(buf, "disk")) {
432 	    media = MEDIA_DISK;
433 	} else if (!strcmp(buf, "cdrom")) {
434             if (cyls || secs || heads) {
435                 error_report("CHS can't be set with media=%s", buf);
436 	        return NULL;
437             }
438 	    media = MEDIA_CDROM;
439 	} else {
440 	    error_report("'%s' invalid media", buf);
441 	    return NULL;
442 	}
443     }
444 
445     if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
446         if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
447             error_report("invalid discard option");
448             return NULL;
449         }
450     }
451 
452     bdrv_flags |= BDRV_O_CACHE_WB;
453     if ((buf = qemu_opt_get(opts, "cache")) != NULL) {
454         if (bdrv_parse_cache_flags(buf, &bdrv_flags) != 0) {
455             error_report("invalid cache option");
456             return NULL;
457         }
458     }
459 
460 #ifdef CONFIG_LINUX_AIO
461     if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
462         if (!strcmp(buf, "native")) {
463             bdrv_flags |= BDRV_O_NATIVE_AIO;
464         } else if (!strcmp(buf, "threads")) {
465             /* this is the default */
466         } else {
467            error_report("invalid aio option");
468            return NULL;
469         }
470     }
471 #endif
472 
473     if ((buf = qemu_opt_get(opts, "format")) != NULL) {
474         if (is_help_option(buf)) {
475             error_printf("Supported formats:");
476             bdrv_iterate_format(bdrv_format_print, NULL);
477             error_printf("\n");
478             return NULL;
479         }
480         drv = bdrv_find_whitelisted_format(buf, ro);
481         if (!drv) {
482             error_report("'%s' invalid format", buf);
483             return NULL;
484         }
485     }
486 
487     /* disk I/O throttling */
488     io_limits.bps[BLOCK_IO_LIMIT_TOTAL]  =
489                            qemu_opt_get_number(opts, "bps", 0);
490     io_limits.bps[BLOCK_IO_LIMIT_READ]   =
491                            qemu_opt_get_number(opts, "bps_rd", 0);
492     io_limits.bps[BLOCK_IO_LIMIT_WRITE]  =
493                            qemu_opt_get_number(opts, "bps_wr", 0);
494     io_limits.iops[BLOCK_IO_LIMIT_TOTAL] =
495                            qemu_opt_get_number(opts, "iops", 0);
496     io_limits.iops[BLOCK_IO_LIMIT_READ]  =
497                            qemu_opt_get_number(opts, "iops_rd", 0);
498     io_limits.iops[BLOCK_IO_LIMIT_WRITE] =
499                            qemu_opt_get_number(opts, "iops_wr", 0);
500 
501     if (!do_check_io_limits(&io_limits, &error)) {
502         error_report("%s", error_get_pretty(error));
503         error_free(error);
504         return NULL;
505     }
506 
507     if (qemu_opt_get(opts, "boot") != NULL) {
508         fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
509                 "ignored. Future versions will reject this parameter. Please "
510                 "update your scripts.\n");
511     }
512 
513     on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
514     if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
515         if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
516             error_report("werror is not supported by this bus type");
517             return NULL;
518         }
519 
520         on_write_error = parse_block_error_action(buf, 0);
521         if (on_write_error < 0) {
522             return NULL;
523         }
524     }
525 
526     on_read_error = BLOCKDEV_ON_ERROR_REPORT;
527     if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
528         if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
529             error_report("rerror is not supported by this bus type");
530             return NULL;
531         }
532 
533         on_read_error = parse_block_error_action(buf, 1);
534         if (on_read_error < 0) {
535             return NULL;
536         }
537     }
538 
539     if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
540         if (type != IF_VIRTIO) {
541             error_report("addr is not supported by this bus type");
542             return NULL;
543         }
544     }
545 
546     /* compute bus and unit according index */
547 
548     if (index != -1) {
549         if (bus_id != 0 || unit_id != -1) {
550             error_report("index cannot be used with bus and unit");
551             return NULL;
552         }
553         bus_id = drive_index_to_bus_id(type, index);
554         unit_id = drive_index_to_unit_id(type, index);
555     }
556 
557     /* if user doesn't specify a unit_id,
558      * try to find the first free
559      */
560 
561     if (unit_id == -1) {
562        unit_id = 0;
563        while (drive_get(type, bus_id, unit_id) != NULL) {
564            unit_id++;
565            if (max_devs && unit_id >= max_devs) {
566                unit_id -= max_devs;
567                bus_id++;
568            }
569        }
570     }
571 
572     /* check unit id */
573 
574     if (max_devs && unit_id >= max_devs) {
575         error_report("unit %d too big (max is %d)",
576                      unit_id, max_devs - 1);
577         return NULL;
578     }
579 
580     /*
581      * catch multiple definitions
582      */
583 
584     if (drive_get(type, bus_id, unit_id) != NULL) {
585         error_report("drive with bus=%d, unit=%d (index=%d) exists",
586                      bus_id, unit_id, index);
587         return NULL;
588     }
589 
590     /* init */
591 
592     dinfo = g_malloc0(sizeof(*dinfo));
593     if ((buf = qemu_opts_id(opts)) != NULL) {
594         dinfo->id = g_strdup(buf);
595     } else {
596         /* no id supplied -> create one */
597         dinfo->id = g_malloc0(32);
598         if (type == IF_IDE || type == IF_SCSI)
599             mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
600         if (max_devs)
601             snprintf(dinfo->id, 32, "%s%i%s%i",
602                      if_name[type], bus_id, mediastr, unit_id);
603         else
604             snprintf(dinfo->id, 32, "%s%s%i",
605                      if_name[type], mediastr, unit_id);
606     }
607     dinfo->bdrv = bdrv_new(dinfo->id);
608     dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
609     dinfo->bdrv->read_only = ro;
610     dinfo->devaddr = devaddr;
611     dinfo->type = type;
612     dinfo->bus = bus_id;
613     dinfo->unit = unit_id;
614     dinfo->cyls = cyls;
615     dinfo->heads = heads;
616     dinfo->secs = secs;
617     dinfo->trans = translation;
618     dinfo->opts = all_opts;
619     dinfo->refcount = 1;
620     if (serial != NULL) {
621         dinfo->serial = g_strdup(serial);
622     }
623     QTAILQ_INSERT_TAIL(&drives, dinfo, next);
624 
625     bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
626 
627     /* disk I/O throttling */
628     bdrv_set_io_limits(dinfo->bdrv, &io_limits);
629 
630     switch(type) {
631     case IF_IDE:
632     case IF_SCSI:
633     case IF_XEN:
634     case IF_NONE:
635         dinfo->media_cd = media == MEDIA_CDROM;
636         break;
637     case IF_SD:
638     case IF_FLOPPY:
639     case IF_PFLASH:
640     case IF_MTD:
641         break;
642     case IF_VIRTIO:
643     {
644         /* add virtio block device */
645         QemuOpts *devopts;
646         devopts = qemu_opts_create_nofail(qemu_find_opts("device"));
647         if (arch_type == QEMU_ARCH_S390X) {
648             qemu_opt_set(devopts, "driver", "virtio-blk-s390");
649         } else {
650             qemu_opt_set(devopts, "driver", "virtio-blk-pci");
651         }
652         qemu_opt_set(devopts, "drive", dinfo->id);
653         if (devaddr)
654             qemu_opt_set(devopts, "addr", devaddr);
655         break;
656     }
657     default:
658         abort();
659     }
660     if (!file || !*file) {
661         if (qdict_size(bs_opts)) {
662             file = NULL;
663         } else {
664             return dinfo;
665         }
666     }
667     if (snapshot) {
668         /* always use cache=unsafe with snapshot */
669         bdrv_flags &= ~BDRV_O_CACHE_MASK;
670         bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
671     }
672 
673     if (copy_on_read) {
674         bdrv_flags |= BDRV_O_COPY_ON_READ;
675     }
676 
677     if (runstate_check(RUN_STATE_INMIGRATE)) {
678         bdrv_flags |= BDRV_O_INCOMING;
679     }
680 
681     if (media == MEDIA_CDROM) {
682         /* CDROM is fine for any interface, don't check.  */
683         ro = 1;
684     } else if (ro == 1) {
685         if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
686             type != IF_NONE && type != IF_PFLASH) {
687             error_report("readonly not supported by this bus type");
688             goto err;
689         }
690     }
691 
692     bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
693 
694     if (ro && copy_on_read) {
695         error_report("warning: disabling copy_on_read on readonly drive");
696     }
697 
698     ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, drv);
699     bs_opts = NULL;
700 
701     if (ret < 0) {
702         if (ret == -EMEDIUMTYPE) {
703             error_report("could not open disk image %s: not in %s format",
704                          file ?: dinfo->id, drv->format_name);
705         } else {
706             error_report("could not open disk image %s: %s",
707                          file ?: dinfo->id, strerror(-ret));
708         }
709         goto err;
710     }
711 
712     if (bdrv_key_required(dinfo->bdrv))
713         autostart = 0;
714 
715     qemu_opts_del(opts);
716 
717     return dinfo;
718 
719 err:
720     qemu_opts_del(opts);
721     QDECREF(bs_opts);
722     bdrv_delete(dinfo->bdrv);
723     g_free(dinfo->id);
724     QTAILQ_REMOVE(&drives, dinfo, next);
725     g_free(dinfo);
726     return NULL;
727 }
728 
729 void do_commit(Monitor *mon, const QDict *qdict)
730 {
731     const char *device = qdict_get_str(qdict, "device");
732     BlockDriverState *bs;
733     int ret;
734 
735     if (!strcmp(device, "all")) {
736         ret = bdrv_commit_all();
737     } else {
738         bs = bdrv_find(device);
739         if (!bs) {
740             monitor_printf(mon, "Device '%s' not found\n", device);
741             return;
742         }
743         ret = bdrv_commit(bs);
744     }
745     if (ret < 0) {
746         monitor_printf(mon, "'commit' error for '%s': %s\n", device,
747                        strerror(-ret));
748     }
749 }
750 
751 static void blockdev_do_action(int kind, void *data, Error **errp)
752 {
753     TransactionAction action;
754     TransactionActionList list;
755 
756     action.kind = kind;
757     action.data = data;
758     list.value = &action;
759     list.next = NULL;
760     qmp_transaction(&list, errp);
761 }
762 
763 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
764                                 bool has_format, const char *format,
765                                 bool has_mode, enum NewImageMode mode,
766                                 Error **errp)
767 {
768     BlockdevSnapshot snapshot = {
769         .device = (char *) device,
770         .snapshot_file = (char *) snapshot_file,
771         .has_format = has_format,
772         .format = (char *) format,
773         .has_mode = has_mode,
774         .mode = mode,
775     };
776     blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
777                        &snapshot, errp);
778 }
779 
780 
781 /* New and old BlockDriverState structs for group snapshots */
782 
783 typedef struct BlkTransactionState BlkTransactionState;
784 
785 /* Only prepare() may fail. In a single transaction, only one of commit() or
786    abort() will be called, clean() will always be called if it present. */
787 typedef struct BdrvActionOps {
788     /* Size of state struct, in bytes. */
789     size_t instance_size;
790     /* Prepare the work, must NOT be NULL. */
791     void (*prepare)(BlkTransactionState *common, Error **errp);
792     /* Commit the changes, can be NULL. */
793     void (*commit)(BlkTransactionState *common);
794     /* Abort the changes on fail, can be NULL. */
795     void (*abort)(BlkTransactionState *common);
796     /* Clean up resource in the end, can be NULL. */
797     void (*clean)(BlkTransactionState *common);
798 } BdrvActionOps;
799 
800 /*
801  * This structure must be arranged as first member in child type, assuming
802  * that compiler will also arrange it to the same address with parent instance.
803  * Later it will be used in free().
804  */
805 struct BlkTransactionState {
806     TransactionAction *action;
807     const BdrvActionOps *ops;
808     QSIMPLEQ_ENTRY(BlkTransactionState) entry;
809 };
810 
811 /* external snapshot private data */
812 typedef struct ExternalSnapshotState {
813     BlkTransactionState common;
814     BlockDriverState *old_bs;
815     BlockDriverState *new_bs;
816 } ExternalSnapshotState;
817 
818 static void external_snapshot_prepare(BlkTransactionState *common,
819                                       Error **errp)
820 {
821     BlockDriver *drv;
822     int flags, ret;
823     Error *local_err = NULL;
824     const char *device;
825     const char *new_image_file;
826     const char *format = "qcow2";
827     enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
828     ExternalSnapshotState *state =
829                              DO_UPCAST(ExternalSnapshotState, common, common);
830     TransactionAction *action = common->action;
831 
832     /* get parameters */
833     g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
834 
835     device = action->blockdev_snapshot_sync->device;
836     new_image_file = action->blockdev_snapshot_sync->snapshot_file;
837     if (action->blockdev_snapshot_sync->has_format) {
838         format = action->blockdev_snapshot_sync->format;
839     }
840     if (action->blockdev_snapshot_sync->has_mode) {
841         mode = action->blockdev_snapshot_sync->mode;
842     }
843 
844     /* start processing */
845     drv = bdrv_find_format(format);
846     if (!drv) {
847         error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
848         return;
849     }
850 
851     state->old_bs = bdrv_find(device);
852     if (!state->old_bs) {
853         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
854         return;
855     }
856 
857     if (!bdrv_is_inserted(state->old_bs)) {
858         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
859         return;
860     }
861 
862     if (bdrv_in_use(state->old_bs)) {
863         error_set(errp, QERR_DEVICE_IN_USE, device);
864         return;
865     }
866 
867     if (!bdrv_is_read_only(state->old_bs)) {
868         if (bdrv_flush(state->old_bs)) {
869             error_set(errp, QERR_IO_ERROR);
870             return;
871         }
872     }
873 
874     flags = state->old_bs->open_flags;
875 
876     /* create new image w/backing file */
877     if (mode != NEW_IMAGE_MODE_EXISTING) {
878         bdrv_img_create(new_image_file, format,
879                         state->old_bs->filename,
880                         state->old_bs->drv->format_name,
881                         NULL, -1, flags, &local_err, false);
882         if (error_is_set(&local_err)) {
883             error_propagate(errp, local_err);
884             return;
885         }
886     }
887 
888     /* We will manually add the backing_hd field to the bs later */
889     state->new_bs = bdrv_new("");
890     /* TODO Inherit bs->options or only take explicit options with an
891      * extended QMP command? */
892     ret = bdrv_open(state->new_bs, new_image_file, NULL,
893                     flags | BDRV_O_NO_BACKING, drv);
894     if (ret != 0) {
895         error_setg_file_open(errp, -ret, new_image_file);
896     }
897 }
898 
899 static void external_snapshot_commit(BlkTransactionState *common)
900 {
901     ExternalSnapshotState *state =
902                              DO_UPCAST(ExternalSnapshotState, common, common);
903 
904     /* This removes our old bs and adds the new bs */
905     bdrv_append(state->new_bs, state->old_bs);
906     /* We don't need (or want) to use the transactional
907      * bdrv_reopen_multiple() across all the entries at once, because we
908      * don't want to abort all of them if one of them fails the reopen */
909     bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
910                 NULL);
911 }
912 
913 static void external_snapshot_abort(BlkTransactionState *common)
914 {
915     ExternalSnapshotState *state =
916                              DO_UPCAST(ExternalSnapshotState, common, common);
917     if (state->new_bs) {
918         bdrv_delete(state->new_bs);
919     }
920 }
921 
922 typedef struct DriveBackupState {
923     BlkTransactionState common;
924     BlockDriverState *bs;
925     BlockJob *job;
926 } DriveBackupState;
927 
928 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
929 {
930     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
931     DriveBackup *backup;
932     Error *local_err = NULL;
933 
934     assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
935     backup = common->action->drive_backup;
936 
937     qmp_drive_backup(backup->device, backup->target,
938                      backup->has_format, backup->format,
939                      backup->sync,
940                      backup->has_mode, backup->mode,
941                      backup->has_speed, backup->speed,
942                      backup->has_on_source_error, backup->on_source_error,
943                      backup->has_on_target_error, backup->on_target_error,
944                      &local_err);
945     if (error_is_set(&local_err)) {
946         error_propagate(errp, local_err);
947         state->bs = NULL;
948         state->job = NULL;
949         return;
950     }
951 
952     state->bs = bdrv_find(backup->device);
953     state->job = state->bs->job;
954 }
955 
956 static void drive_backup_abort(BlkTransactionState *common)
957 {
958     DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
959     BlockDriverState *bs = state->bs;
960 
961     /* Only cancel if it's the job we started */
962     if (bs && bs->job && bs->job == state->job) {
963         block_job_cancel_sync(bs->job);
964     }
965 }
966 
967 static void abort_prepare(BlkTransactionState *common, Error **errp)
968 {
969     error_setg(errp, "Transaction aborted using Abort action");
970 }
971 
972 static void abort_commit(BlkTransactionState *common)
973 {
974     assert(false); /* this action never succeeds */
975 }
976 
977 static const BdrvActionOps actions[] = {
978     [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
979         .instance_size = sizeof(ExternalSnapshotState),
980         .prepare  = external_snapshot_prepare,
981         .commit   = external_snapshot_commit,
982         .abort = external_snapshot_abort,
983     },
984     [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
985         .instance_size = sizeof(DriveBackupState),
986         .prepare = drive_backup_prepare,
987         .abort = drive_backup_abort,
988     },
989     [TRANSACTION_ACTION_KIND_ABORT] = {
990         .instance_size = sizeof(BlkTransactionState),
991         .prepare = abort_prepare,
992         .commit = abort_commit,
993     },
994 };
995 
996 /*
997  * 'Atomic' group snapshots.  The snapshots are taken as a set, and if any fail
998  *  then we do not pivot any of the devices in the group, and abandon the
999  *  snapshots
1000  */
1001 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1002 {
1003     TransactionActionList *dev_entry = dev_list;
1004     BlkTransactionState *state, *next;
1005     Error *local_err = NULL;
1006 
1007     QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1008     QSIMPLEQ_INIT(&snap_bdrv_states);
1009 
1010     /* drain all i/o before any snapshots */
1011     bdrv_drain_all();
1012 
1013     /* We don't do anything in this loop that commits us to the snapshot */
1014     while (NULL != dev_entry) {
1015         TransactionAction *dev_info = NULL;
1016         const BdrvActionOps *ops;
1017 
1018         dev_info = dev_entry->value;
1019         dev_entry = dev_entry->next;
1020 
1021         assert(dev_info->kind < ARRAY_SIZE(actions));
1022 
1023         ops = &actions[dev_info->kind];
1024         state = g_malloc0(ops->instance_size);
1025         state->ops = ops;
1026         state->action = dev_info;
1027         QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1028 
1029         state->ops->prepare(state, &local_err);
1030         if (error_is_set(&local_err)) {
1031             error_propagate(errp, local_err);
1032             goto delete_and_fail;
1033         }
1034     }
1035 
1036     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1037         if (state->ops->commit) {
1038             state->ops->commit(state);
1039         }
1040     }
1041 
1042     /* success */
1043     goto exit;
1044 
1045 delete_and_fail:
1046     /*
1047     * failure, and it is all-or-none; abandon each new bs, and keep using
1048     * the original bs for all images
1049     */
1050     QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1051         if (state->ops->abort) {
1052             state->ops->abort(state);
1053         }
1054     }
1055 exit:
1056     QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1057         if (state->ops->clean) {
1058             state->ops->clean(state);
1059         }
1060         g_free(state);
1061     }
1062 }
1063 
1064 
1065 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1066 {
1067     if (bdrv_in_use(bs)) {
1068         error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1069         return;
1070     }
1071     if (!bdrv_dev_has_removable_media(bs)) {
1072         error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1073         return;
1074     }
1075 
1076     if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1077         bdrv_dev_eject_request(bs, force);
1078         if (!force) {
1079             error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1080             return;
1081         }
1082     }
1083 
1084     bdrv_close(bs);
1085 }
1086 
1087 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1088 {
1089     BlockDriverState *bs;
1090 
1091     bs = bdrv_find(device);
1092     if (!bs) {
1093         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1094         return;
1095     }
1096 
1097     eject_device(bs, force, errp);
1098 }
1099 
1100 void qmp_block_passwd(const char *device, const char *password, Error **errp)
1101 {
1102     BlockDriverState *bs;
1103     int err;
1104 
1105     bs = bdrv_find(device);
1106     if (!bs) {
1107         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1108         return;
1109     }
1110 
1111     err = bdrv_set_key(bs, password);
1112     if (err == -EINVAL) {
1113         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1114         return;
1115     } else if (err < 0) {
1116         error_set(errp, QERR_INVALID_PASSWORD);
1117         return;
1118     }
1119 }
1120 
1121 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1122                                     int bdrv_flags, BlockDriver *drv,
1123                                     const char *password, Error **errp)
1124 {
1125     int ret;
1126 
1127     ret = bdrv_open(bs, filename, NULL, bdrv_flags, drv);
1128     if (ret < 0) {
1129         error_setg_file_open(errp, -ret, filename);
1130         return;
1131     }
1132 
1133     if (bdrv_key_required(bs)) {
1134         if (password) {
1135             if (bdrv_set_key(bs, password) < 0) {
1136                 error_set(errp, QERR_INVALID_PASSWORD);
1137             }
1138         } else {
1139             error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1140                       bdrv_get_encrypted_filename(bs));
1141         }
1142     } else if (password) {
1143         error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1144     }
1145 }
1146 
1147 void qmp_change_blockdev(const char *device, const char *filename,
1148                          bool has_format, const char *format, Error **errp)
1149 {
1150     BlockDriverState *bs;
1151     BlockDriver *drv = NULL;
1152     int bdrv_flags;
1153     Error *err = NULL;
1154 
1155     bs = bdrv_find(device);
1156     if (!bs) {
1157         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1158         return;
1159     }
1160 
1161     if (format) {
1162         drv = bdrv_find_whitelisted_format(format, bs->read_only);
1163         if (!drv) {
1164             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1165             return;
1166         }
1167     }
1168 
1169     eject_device(bs, 0, &err);
1170     if (error_is_set(&err)) {
1171         error_propagate(errp, err);
1172         return;
1173     }
1174 
1175     bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1176     bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1177 
1178     qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1179 }
1180 
1181 /* throttling disk I/O limits */
1182 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1183                                int64_t bps_wr, int64_t iops, int64_t iops_rd,
1184                                int64_t iops_wr, Error **errp)
1185 {
1186     BlockIOLimit io_limits;
1187     BlockDriverState *bs;
1188 
1189     bs = bdrv_find(device);
1190     if (!bs) {
1191         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1192         return;
1193     }
1194 
1195     io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
1196     io_limits.bps[BLOCK_IO_LIMIT_READ]  = bps_rd;
1197     io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
1198     io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
1199     io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
1200     io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
1201 
1202     if (!do_check_io_limits(&io_limits, errp)) {
1203         return;
1204     }
1205 
1206     bs->io_limits = io_limits;
1207 
1208     if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
1209         bdrv_io_limits_enable(bs);
1210     } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
1211         bdrv_io_limits_disable(bs);
1212     } else {
1213         if (bs->block_timer) {
1214             qemu_mod_timer(bs->block_timer, qemu_get_clock_ns(vm_clock));
1215         }
1216     }
1217 }
1218 
1219 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1220 {
1221     const char *id = qdict_get_str(qdict, "id");
1222     BlockDriverState *bs;
1223 
1224     bs = bdrv_find(id);
1225     if (!bs) {
1226         qerror_report(QERR_DEVICE_NOT_FOUND, id);
1227         return -1;
1228     }
1229     if (bdrv_in_use(bs)) {
1230         qerror_report(QERR_DEVICE_IN_USE, id);
1231         return -1;
1232     }
1233 
1234     /* quiesce block driver; prevent further io */
1235     bdrv_drain_all();
1236     bdrv_flush(bs);
1237     bdrv_close(bs);
1238 
1239     /* if we have a device attached to this BlockDriverState
1240      * then we need to make the drive anonymous until the device
1241      * can be removed.  If this is a drive with no device backing
1242      * then we can just get rid of the block driver state right here.
1243      */
1244     if (bdrv_get_attached_dev(bs)) {
1245         bdrv_make_anon(bs);
1246 
1247         /* Further I/O must not pause the guest */
1248         bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1249                           BLOCKDEV_ON_ERROR_REPORT);
1250     } else {
1251         drive_uninit(drive_get_by_blockdev(bs));
1252     }
1253 
1254     return 0;
1255 }
1256 
1257 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1258 {
1259     BlockDriverState *bs;
1260     int ret;
1261 
1262     bs = bdrv_find(device);
1263     if (!bs) {
1264         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1265         return;
1266     }
1267 
1268     if (size < 0) {
1269         error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1270         return;
1271     }
1272 
1273     /* complete all in-flight operations before resizing the device */
1274     bdrv_drain_all();
1275 
1276     ret = bdrv_truncate(bs, size);
1277     switch (ret) {
1278     case 0:
1279         break;
1280     case -ENOMEDIUM:
1281         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1282         break;
1283     case -ENOTSUP:
1284         error_set(errp, QERR_UNSUPPORTED);
1285         break;
1286     case -EACCES:
1287         error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1288         break;
1289     case -EBUSY:
1290         error_set(errp, QERR_DEVICE_IN_USE, device);
1291         break;
1292     default:
1293         error_setg_errno(errp, -ret, "Could not resize");
1294         break;
1295     }
1296 }
1297 
1298 static void block_job_cb(void *opaque, int ret)
1299 {
1300     BlockDriverState *bs = opaque;
1301     QObject *obj;
1302 
1303     trace_block_job_cb(bs, bs->job, ret);
1304 
1305     assert(bs->job);
1306     obj = qobject_from_block_job(bs->job);
1307     if (ret < 0) {
1308         QDict *dict = qobject_to_qdict(obj);
1309         qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1310     }
1311 
1312     if (block_job_is_cancelled(bs->job)) {
1313         monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1314     } else {
1315         monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1316     }
1317     qobject_decref(obj);
1318 
1319     drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1320 }
1321 
1322 void qmp_block_stream(const char *device, bool has_base,
1323                       const char *base, bool has_speed, int64_t speed,
1324                       bool has_on_error, BlockdevOnError on_error,
1325                       Error **errp)
1326 {
1327     BlockDriverState *bs;
1328     BlockDriverState *base_bs = NULL;
1329     Error *local_err = NULL;
1330 
1331     if (!has_on_error) {
1332         on_error = BLOCKDEV_ON_ERROR_REPORT;
1333     }
1334 
1335     bs = bdrv_find(device);
1336     if (!bs) {
1337         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1338         return;
1339     }
1340 
1341     if (base) {
1342         base_bs = bdrv_find_backing_image(bs, base);
1343         if (base_bs == NULL) {
1344             error_set(errp, QERR_BASE_NOT_FOUND, base);
1345             return;
1346         }
1347     }
1348 
1349     stream_start(bs, base_bs, base, has_speed ? speed : 0,
1350                  on_error, block_job_cb, bs, &local_err);
1351     if (error_is_set(&local_err)) {
1352         error_propagate(errp, local_err);
1353         return;
1354     }
1355 
1356     /* Grab a reference so hotplug does not delete the BlockDriverState from
1357      * underneath us.
1358      */
1359     drive_get_ref(drive_get_by_blockdev(bs));
1360 
1361     trace_qmp_block_stream(bs, bs->job);
1362 }
1363 
1364 void qmp_block_commit(const char *device,
1365                       bool has_base, const char *base, const char *top,
1366                       bool has_speed, int64_t speed,
1367                       Error **errp)
1368 {
1369     BlockDriverState *bs;
1370     BlockDriverState *base_bs, *top_bs;
1371     Error *local_err = NULL;
1372     /* This will be part of the QMP command, if/when the
1373      * BlockdevOnError change for blkmirror makes it in
1374      */
1375     BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1376 
1377     /* drain all i/o before commits */
1378     bdrv_drain_all();
1379 
1380     bs = bdrv_find(device);
1381     if (!bs) {
1382         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1383         return;
1384     }
1385 
1386     /* default top_bs is the active layer */
1387     top_bs = bs;
1388 
1389     if (top) {
1390         if (strcmp(bs->filename, top) != 0) {
1391             top_bs = bdrv_find_backing_image(bs, top);
1392         }
1393     }
1394 
1395     if (top_bs == NULL) {
1396         error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1397         return;
1398     }
1399 
1400     if (has_base && base) {
1401         base_bs = bdrv_find_backing_image(top_bs, base);
1402     } else {
1403         base_bs = bdrv_find_base(top_bs);
1404     }
1405 
1406     if (base_bs == NULL) {
1407         error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1408         return;
1409     }
1410 
1411     commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1412                 &local_err);
1413     if (local_err != NULL) {
1414         error_propagate(errp, local_err);
1415         return;
1416     }
1417     /* Grab a reference so hotplug does not delete the BlockDriverState from
1418      * underneath us.
1419      */
1420     drive_get_ref(drive_get_by_blockdev(bs));
1421 }
1422 
1423 void qmp_drive_backup(const char *device, const char *target,
1424                       bool has_format, const char *format,
1425                       enum MirrorSyncMode sync,
1426                       bool has_mode, enum NewImageMode mode,
1427                       bool has_speed, int64_t speed,
1428                       bool has_on_source_error, BlockdevOnError on_source_error,
1429                       bool has_on_target_error, BlockdevOnError on_target_error,
1430                       Error **errp)
1431 {
1432     BlockDriverState *bs;
1433     BlockDriverState *target_bs;
1434     BlockDriver *drv = NULL;
1435     Error *local_err = NULL;
1436     int flags;
1437     int64_t size;
1438     int ret;
1439 
1440     if (sync != MIRROR_SYNC_MODE_FULL) {
1441         error_setg(errp, "only sync mode 'full' is currently supported");
1442         return;
1443     }
1444     if (!has_speed) {
1445         speed = 0;
1446     }
1447     if (!has_on_source_error) {
1448         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1449     }
1450     if (!has_on_target_error) {
1451         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1452     }
1453     if (!has_mode) {
1454         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1455     }
1456 
1457     bs = bdrv_find(device);
1458     if (!bs) {
1459         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1460         return;
1461     }
1462 
1463     if (!bdrv_is_inserted(bs)) {
1464         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1465         return;
1466     }
1467 
1468     if (!has_format) {
1469         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1470     }
1471     if (format) {
1472         drv = bdrv_find_format(format);
1473         if (!drv) {
1474             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1475             return;
1476         }
1477     }
1478 
1479     if (bdrv_in_use(bs)) {
1480         error_set(errp, QERR_DEVICE_IN_USE, device);
1481         return;
1482     }
1483 
1484     flags = bs->open_flags | BDRV_O_RDWR;
1485 
1486     size = bdrv_getlength(bs);
1487     if (size < 0) {
1488         error_setg_errno(errp, -size, "bdrv_getlength failed");
1489         return;
1490     }
1491 
1492     if (mode != NEW_IMAGE_MODE_EXISTING) {
1493         assert(format && drv);
1494         bdrv_img_create(target, format,
1495                         NULL, NULL, NULL, size, flags, &local_err, false);
1496     }
1497 
1498     if (error_is_set(&local_err)) {
1499         error_propagate(errp, local_err);
1500         return;
1501     }
1502 
1503     target_bs = bdrv_new("");
1504     ret = bdrv_open(target_bs, target, NULL, flags, drv);
1505     if (ret < 0) {
1506         bdrv_delete(target_bs);
1507         error_setg_file_open(errp, -ret, target);
1508         return;
1509     }
1510 
1511     backup_start(bs, target_bs, speed, on_source_error, on_target_error,
1512                  block_job_cb, bs, &local_err);
1513     if (local_err != NULL) {
1514         bdrv_delete(target_bs);
1515         error_propagate(errp, local_err);
1516         return;
1517     }
1518 
1519     /* Grab a reference so hotplug does not delete the BlockDriverState from
1520      * underneath us.
1521      */
1522     drive_get_ref(drive_get_by_blockdev(bs));
1523 }
1524 
1525 #define DEFAULT_MIRROR_BUF_SIZE   (10 << 20)
1526 
1527 void qmp_drive_mirror(const char *device, const char *target,
1528                       bool has_format, const char *format,
1529                       enum MirrorSyncMode sync,
1530                       bool has_mode, enum NewImageMode mode,
1531                       bool has_speed, int64_t speed,
1532                       bool has_granularity, uint32_t granularity,
1533                       bool has_buf_size, int64_t buf_size,
1534                       bool has_on_source_error, BlockdevOnError on_source_error,
1535                       bool has_on_target_error, BlockdevOnError on_target_error,
1536                       Error **errp)
1537 {
1538     BlockDriverState *bs;
1539     BlockDriverState *source, *target_bs;
1540     BlockDriver *drv = NULL;
1541     Error *local_err = NULL;
1542     int flags;
1543     int64_t size;
1544     int ret;
1545 
1546     if (!has_speed) {
1547         speed = 0;
1548     }
1549     if (!has_on_source_error) {
1550         on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1551     }
1552     if (!has_on_target_error) {
1553         on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1554     }
1555     if (!has_mode) {
1556         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1557     }
1558     if (!has_granularity) {
1559         granularity = 0;
1560     }
1561     if (!has_buf_size) {
1562         buf_size = DEFAULT_MIRROR_BUF_SIZE;
1563     }
1564 
1565     if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
1566         error_set(errp, QERR_INVALID_PARAMETER, device);
1567         return;
1568     }
1569     if (granularity & (granularity - 1)) {
1570         error_set(errp, QERR_INVALID_PARAMETER, device);
1571         return;
1572     }
1573 
1574     bs = bdrv_find(device);
1575     if (!bs) {
1576         error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1577         return;
1578     }
1579 
1580     if (!bdrv_is_inserted(bs)) {
1581         error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1582         return;
1583     }
1584 
1585     if (!has_format) {
1586         format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1587     }
1588     if (format) {
1589         drv = bdrv_find_format(format);
1590         if (!drv) {
1591             error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1592             return;
1593         }
1594     }
1595 
1596     if (bdrv_in_use(bs)) {
1597         error_set(errp, QERR_DEVICE_IN_USE, device);
1598         return;
1599     }
1600 
1601     flags = bs->open_flags | BDRV_O_RDWR;
1602     source = bs->backing_hd;
1603     if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1604         sync = MIRROR_SYNC_MODE_FULL;
1605     }
1606 
1607     size = bdrv_getlength(bs);
1608     if (size < 0) {
1609         error_setg_errno(errp, -size, "bdrv_getlength failed");
1610         return;
1611     }
1612 
1613     if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1614         /* create new image w/o backing file */
1615         assert(format && drv);
1616         bdrv_img_create(target, format,
1617                         NULL, NULL, NULL, size, flags, &local_err, false);
1618     } else {
1619         switch (mode) {
1620         case NEW_IMAGE_MODE_EXISTING:
1621             ret = 0;
1622             break;
1623         case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1624             /* create new image with backing file */
1625             bdrv_img_create(target, format,
1626                             source->filename,
1627                             source->drv->format_name,
1628                             NULL, size, flags, &local_err, false);
1629             break;
1630         default:
1631             abort();
1632         }
1633     }
1634 
1635     if (error_is_set(&local_err)) {
1636         error_propagate(errp, local_err);
1637         return;
1638     }
1639 
1640     /* Mirroring takes care of copy-on-write using the source's backing
1641      * file.
1642      */
1643     target_bs = bdrv_new("");
1644     ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv);
1645     if (ret < 0) {
1646         bdrv_delete(target_bs);
1647         error_setg_file_open(errp, -ret, target);
1648         return;
1649     }
1650 
1651     mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
1652                  on_source_error, on_target_error,
1653                  block_job_cb, bs, &local_err);
1654     if (local_err != NULL) {
1655         bdrv_delete(target_bs);
1656         error_propagate(errp, local_err);
1657         return;
1658     }
1659 
1660     /* Grab a reference so hotplug does not delete the BlockDriverState from
1661      * underneath us.
1662      */
1663     drive_get_ref(drive_get_by_blockdev(bs));
1664 }
1665 
1666 static BlockJob *find_block_job(const char *device)
1667 {
1668     BlockDriverState *bs;
1669 
1670     bs = bdrv_find(device);
1671     if (!bs || !bs->job) {
1672         return NULL;
1673     }
1674     return bs->job;
1675 }
1676 
1677 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1678 {
1679     BlockJob *job = find_block_job(device);
1680 
1681     if (!job) {
1682         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1683         return;
1684     }
1685 
1686     block_job_set_speed(job, speed, errp);
1687 }
1688 
1689 void qmp_block_job_cancel(const char *device,
1690                           bool has_force, bool force, Error **errp)
1691 {
1692     BlockJob *job = find_block_job(device);
1693 
1694     if (!has_force) {
1695         force = false;
1696     }
1697 
1698     if (!job) {
1699         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1700         return;
1701     }
1702     if (job->paused && !force) {
1703         error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1704         return;
1705     }
1706 
1707     trace_qmp_block_job_cancel(job);
1708     block_job_cancel(job);
1709 }
1710 
1711 void qmp_block_job_pause(const char *device, Error **errp)
1712 {
1713     BlockJob *job = find_block_job(device);
1714 
1715     if (!job) {
1716         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1717         return;
1718     }
1719 
1720     trace_qmp_block_job_pause(job);
1721     block_job_pause(job);
1722 }
1723 
1724 void qmp_block_job_resume(const char *device, Error **errp)
1725 {
1726     BlockJob *job = find_block_job(device);
1727 
1728     if (!job) {
1729         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1730         return;
1731     }
1732 
1733     trace_qmp_block_job_resume(job);
1734     block_job_resume(job);
1735 }
1736 
1737 void qmp_block_job_complete(const char *device, Error **errp)
1738 {
1739     BlockJob *job = find_block_job(device);
1740 
1741     if (!job) {
1742         error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1743         return;
1744     }
1745 
1746     trace_qmp_block_job_complete(job);
1747     block_job_complete(job, errp);
1748 }
1749 
1750 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1751 {
1752     BlockJobInfoList **prev = opaque;
1753     BlockJob *job = bs->job;
1754 
1755     if (job) {
1756         BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1757         elem->value = block_job_query(bs->job);
1758         (*prev)->next = elem;
1759         *prev = elem;
1760     }
1761 }
1762 
1763 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1764 {
1765     /* Dummy is a fake list element for holding the head pointer */
1766     BlockJobInfoList dummy = {};
1767     BlockJobInfoList *prev = &dummy;
1768     bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1769     return dummy.next;
1770 }
1771 
1772 QemuOptsList qemu_common_drive_opts = {
1773     .name = "drive",
1774     .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
1775     .desc = {
1776         {
1777             .name = "bus",
1778             .type = QEMU_OPT_NUMBER,
1779             .help = "bus number",
1780         },{
1781             .name = "unit",
1782             .type = QEMU_OPT_NUMBER,
1783             .help = "unit number (i.e. lun for scsi)",
1784         },{
1785             .name = "if",
1786             .type = QEMU_OPT_STRING,
1787             .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1788         },{
1789             .name = "index",
1790             .type = QEMU_OPT_NUMBER,
1791             .help = "index number",
1792         },{
1793             .name = "cyls",
1794             .type = QEMU_OPT_NUMBER,
1795             .help = "number of cylinders (ide disk geometry)",
1796         },{
1797             .name = "heads",
1798             .type = QEMU_OPT_NUMBER,
1799             .help = "number of heads (ide disk geometry)",
1800         },{
1801             .name = "secs",
1802             .type = QEMU_OPT_NUMBER,
1803             .help = "number of sectors (ide disk geometry)",
1804         },{
1805             .name = "trans",
1806             .type = QEMU_OPT_STRING,
1807             .help = "chs translation (auto, lba. none)",
1808         },{
1809             .name = "media",
1810             .type = QEMU_OPT_STRING,
1811             .help = "media type (disk, cdrom)",
1812         },{
1813             .name = "snapshot",
1814             .type = QEMU_OPT_BOOL,
1815             .help = "enable/disable snapshot mode",
1816         },{
1817             .name = "file",
1818             .type = QEMU_OPT_STRING,
1819             .help = "disk image",
1820         },{
1821             .name = "discard",
1822             .type = QEMU_OPT_STRING,
1823             .help = "discard operation (ignore/off, unmap/on)",
1824         },{
1825             .name = "cache",
1826             .type = QEMU_OPT_STRING,
1827             .help = "host cache usage (none, writeback, writethrough, "
1828                     "directsync, unsafe)",
1829         },{
1830             .name = "aio",
1831             .type = QEMU_OPT_STRING,
1832             .help = "host AIO implementation (threads, native)",
1833         },{
1834             .name = "format",
1835             .type = QEMU_OPT_STRING,
1836             .help = "disk format (raw, qcow2, ...)",
1837         },{
1838             .name = "serial",
1839             .type = QEMU_OPT_STRING,
1840             .help = "disk serial number",
1841         },{
1842             .name = "rerror",
1843             .type = QEMU_OPT_STRING,
1844             .help = "read error action",
1845         },{
1846             .name = "werror",
1847             .type = QEMU_OPT_STRING,
1848             .help = "write error action",
1849         },{
1850             .name = "addr",
1851             .type = QEMU_OPT_STRING,
1852             .help = "pci address (virtio only)",
1853         },{
1854             .name = "readonly",
1855             .type = QEMU_OPT_BOOL,
1856             .help = "open drive file as read-only",
1857         },{
1858             .name = "iops",
1859             .type = QEMU_OPT_NUMBER,
1860             .help = "limit total I/O operations per second",
1861         },{
1862             .name = "iops_rd",
1863             .type = QEMU_OPT_NUMBER,
1864             .help = "limit read operations per second",
1865         },{
1866             .name = "iops_wr",
1867             .type = QEMU_OPT_NUMBER,
1868             .help = "limit write operations per second",
1869         },{
1870             .name = "bps",
1871             .type = QEMU_OPT_NUMBER,
1872             .help = "limit total bytes per second",
1873         },{
1874             .name = "bps_rd",
1875             .type = QEMU_OPT_NUMBER,
1876             .help = "limit read bytes per second",
1877         },{
1878             .name = "bps_wr",
1879             .type = QEMU_OPT_NUMBER,
1880             .help = "limit write bytes per second",
1881         },{
1882             .name = "copy-on-read",
1883             .type = QEMU_OPT_BOOL,
1884             .help = "copy read data from backing file into image file",
1885         },{
1886             .name = "boot",
1887             .type = QEMU_OPT_BOOL,
1888             .help = "(deprecated, ignored)",
1889         },
1890         { /* end of list */ }
1891     },
1892 };
1893 
1894 QemuOptsList qemu_drive_opts = {
1895     .name = "drive",
1896     .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
1897     .desc = {
1898         /*
1899          * no elements => accept any params
1900          * validation will happen later
1901          */
1902         { /* end of list */ }
1903     },
1904 };
1905