2 * QEMU System Emulator block driver
4 * Copyright (c) 2011 IBM Corp.
5 * Copyright (c) 2012 Red Hat, Inc.
7 * Permission is hereby granted, free of charge, to any person obtaining a copy
8 * of this software and associated documentation files (the "Software"), to deal
9 * in the Software without restriction, including without limitation the rights
10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11 * copies of the Software, and to permit persons to whom the Software is
12 * furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
26 #include "qemu/osdep.h"
27 #include "qemu-common.h"
28 #include "block/block.h"
29 #include "block/blockjob_int.h"
30 #include "block/block_int.h"
31 #include "block/trace.h"
32 #include "sysemu/block-backend.h"
33 #include "qapi/error.h"
34 #include "qapi/qapi-events-block-core.h"
35 #include "qapi/qmp/qerror.h"
36 #include "qemu/coroutine.h"
37 #include "qemu/timer.h"
40 * The block job API is composed of two categories of functions.
42 * The first includes functions used by the monitor. The monitor is
43 * peculiar in that it accesses the block job list with block_job_get, and
44 * therefore needs consistency across block_job_get and the actual operation
45 * (e.g. block_job_set_speed). The consistency is achieved with
46 * aio_context_acquire/release. These functions are declared in blockjob.h.
48 * The second includes functions used by the block job drivers and sometimes
49 * by the core block layer. These do not care about locking, because the
50 * whole coroutine runs under the AioContext lock, and are declared in
54 static bool is_block_job(Job
*job
)
56 return job_type(job
) == JOB_TYPE_BACKUP
||
57 job_type(job
) == JOB_TYPE_COMMIT
||
58 job_type(job
) == JOB_TYPE_MIRROR
||
59 job_type(job
) == JOB_TYPE_STREAM
;
62 BlockJob
*block_job_next(BlockJob
*bjob
)
64 Job
*job
= bjob
? &bjob
->job
: NULL
;
68 } while (job
&& !is_block_job(job
));
70 return job
? container_of(job
, BlockJob
, job
) : NULL
;
73 BlockJob
*block_job_get(const char *id
)
75 Job
*job
= job_get(id
);
77 if (job
&& is_block_job(job
)) {
78 return container_of(job
, BlockJob
, job
);
84 static void block_job_attached_aio_context(AioContext
*new_context
,
86 static void block_job_detach_aio_context(void *opaque
);
88 void block_job_free(Job
*job
)
90 BlockJob
*bjob
= container_of(job
, BlockJob
, job
);
91 BlockDriverState
*bs
= blk_bs(bjob
->blk
);
94 block_job_remove_all_bdrv(bjob
);
95 blk_remove_aio_context_notifier(bjob
->blk
,
96 block_job_attached_aio_context
,
97 block_job_detach_aio_context
, bjob
);
99 error_free(bjob
->blocker
);
102 static void block_job_attached_aio_context(AioContext
*new_context
,
105 BlockJob
*job
= opaque
;
107 job
->job
.aio_context
= new_context
;
108 if (job
->driver
->attached_aio_context
) {
109 job
->driver
->attached_aio_context(job
, new_context
);
112 job_resume(&job
->job
);
115 void block_job_drain(Job
*job
)
117 BlockJob
*bjob
= container_of(job
, BlockJob
, job
);
119 blk_drain(bjob
->blk
);
120 if (bjob
->driver
->drain
) {
121 bjob
->driver
->drain(bjob
);
125 static void block_job_detach_aio_context(void *opaque
)
127 BlockJob
*job
= opaque
;
129 /* In case the job terminates during aio_poll()... */
132 job_pause(&job
->job
);
134 while (!job
->job
.paused
&& !job_is_completed(&job
->job
)) {
135 job_drain(&job
->job
);
138 job
->job
.aio_context
= NULL
;
139 job_unref(&job
->job
);
142 static char *child_job_get_parent_desc(BdrvChild
*c
)
144 BlockJob
*job
= c
->opaque
;
145 return g_strdup_printf("%s job '%s'", job_type_str(&job
->job
), job
->job
.id
);
148 static void child_job_drained_begin(BdrvChild
*c
)
150 BlockJob
*job
= c
->opaque
;
151 job_pause(&job
->job
);
154 static void child_job_drained_end(BdrvChild
*c
)
156 BlockJob
*job
= c
->opaque
;
157 job_resume(&job
->job
);
160 static const BdrvChildRole child_job
= {
161 .get_parent_desc
= child_job_get_parent_desc
,
162 .drained_begin
= child_job_drained_begin
,
163 .drained_end
= child_job_drained_end
,
164 .stay_at_node
= true,
167 void block_job_remove_all_bdrv(BlockJob
*job
)
170 for (l
= job
->nodes
; l
; l
= l
->next
) {
171 BdrvChild
*c
= l
->data
;
172 bdrv_op_unblock_all(c
->bs
, job
->blocker
);
173 bdrv_root_unref_child(c
);
175 g_slist_free(job
->nodes
);
179 int block_job_add_bdrv(BlockJob
*job
, const char *name
, BlockDriverState
*bs
,
180 uint64_t perm
, uint64_t shared_perm
, Error
**errp
)
184 c
= bdrv_root_attach_child(bs
, name
, &child_job
, perm
, shared_perm
,
190 job
->nodes
= g_slist_prepend(job
->nodes
, c
);
192 bdrv_op_block_all(bs
, job
->blocker
);
197 bool block_job_is_internal(BlockJob
*job
)
199 return (job
->job
.id
== NULL
);
202 const BlockJobDriver
*block_job_driver(BlockJob
*job
)
207 /* Assumes the job_mutex is held */
208 static bool job_timer_pending(Job
*job
)
210 return timer_pending(&job
->sleep_timer
);
213 void block_job_set_speed(BlockJob
*job
, int64_t speed
, Error
**errp
)
215 int64_t old_speed
= job
->speed
;
217 if (job_apply_verb(&job
->job
, JOB_VERB_SET_SPEED
, errp
)) {
221 error_setg(errp
, QERR_INVALID_PARAMETER
, "speed");
225 ratelimit_set_speed(&job
->limit
, speed
, BLOCK_JOB_SLICE_TIME
);
228 if (speed
&& speed
<= old_speed
) {
232 /* kick only if a timer is pending */
233 job_enter_cond(&job
->job
, job_timer_pending
);
236 int64_t block_job_ratelimit_get_delay(BlockJob
*job
, uint64_t n
)
242 return ratelimit_calculate_delay(&job
->limit
, n
);
245 void block_job_progress_update(BlockJob
*job
, uint64_t done
)
250 void block_job_progress_set_remaining(BlockJob
*job
, uint64_t remaining
)
252 job
->len
= job
->offset
+ remaining
;
255 BlockJobInfo
*block_job_query(BlockJob
*job
, Error
**errp
)
259 if (block_job_is_internal(job
)) {
260 error_setg(errp
, "Cannot query QEMU internal jobs");
263 info
= g_new0(BlockJobInfo
, 1);
264 info
->type
= g_strdup(job_type_str(&job
->job
));
265 info
->device
= g_strdup(job
->job
.id
);
266 info
->len
= job
->len
;
267 info
->busy
= atomic_read(&job
->job
.busy
);
268 info
->paused
= job
->job
.pause_count
> 0;
269 info
->offset
= job
->offset
;
270 info
->speed
= job
->speed
;
271 info
->io_status
= job
->iostatus
;
272 info
->ready
= job
->ready
;
273 info
->status
= job
->job
.status
;
274 info
->auto_finalize
= job
->job
.auto_finalize
;
275 info
->auto_dismiss
= job
->job
.auto_dismiss
;
276 info
->has_error
= job
->job
.ret
!= 0;
277 info
->error
= job
->job
.ret
? g_strdup(strerror(-job
->job
.ret
)) : NULL
;
281 static void block_job_iostatus_set_err(BlockJob
*job
, int error
)
283 if (job
->iostatus
== BLOCK_DEVICE_IO_STATUS_OK
) {
284 job
->iostatus
= error
== ENOSPC
? BLOCK_DEVICE_IO_STATUS_NOSPACE
:
285 BLOCK_DEVICE_IO_STATUS_FAILED
;
289 static void block_job_event_cancelled(Notifier
*n
, void *opaque
)
291 BlockJob
*job
= opaque
;
293 if (block_job_is_internal(job
)) {
297 qapi_event_send_block_job_cancelled(job_type(&job
->job
),
305 static void block_job_event_completed(Notifier
*n
, void *opaque
)
307 BlockJob
*job
= opaque
;
308 const char *msg
= NULL
;
310 if (block_job_is_internal(job
)) {
314 if (job
->job
.ret
< 0) {
315 msg
= strerror(-job
->job
.ret
);
318 qapi_event_send_block_job_completed(job_type(&job
->job
),
328 static void block_job_event_pending(Notifier
*n
, void *opaque
)
330 BlockJob
*job
= opaque
;
332 if (block_job_is_internal(job
)) {
336 qapi_event_send_block_job_pending(job_type(&job
->job
),
342 * API for block job drivers and the block layer. These functions are
343 * declared in blockjob_int.h.
346 void *block_job_create(const char *job_id
, const BlockJobDriver
*driver
,
347 JobTxn
*txn
, BlockDriverState
*bs
, uint64_t perm
,
348 uint64_t shared_perm
, int64_t speed
, int flags
,
349 BlockCompletionFunc
*cb
, void *opaque
, Error
**errp
)
356 error_setg(errp
, QERR_DEVICE_IN_USE
, bdrv_get_device_name(bs
));
360 if (job_id
== NULL
&& !(flags
& JOB_INTERNAL
)) {
361 job_id
= bdrv_get_device_name(bs
);
364 blk
= blk_new(perm
, shared_perm
);
365 ret
= blk_insert_bs(blk
, bs
, errp
);
371 job
= job_create(job_id
, &driver
->job_driver
, txn
, blk_get_aio_context(blk
),
372 flags
, cb
, opaque
, errp
);
378 assert(is_block_job(&job
->job
));
379 assert(job
->job
.driver
->free
== &block_job_free
);
380 assert(job
->job
.driver
->user_resume
== &block_job_user_resume
);
381 assert(job
->job
.driver
->drain
== &block_job_drain
);
383 job
->driver
= driver
;
386 job
->finalize_cancelled_notifier
.notify
= block_job_event_cancelled
;
387 job
->finalize_completed_notifier
.notify
= block_job_event_completed
;
388 job
->pending_notifier
.notify
= block_job_event_pending
;
390 notifier_list_add(&job
->job
.on_finalize_cancelled
,
391 &job
->finalize_cancelled_notifier
);
392 notifier_list_add(&job
->job
.on_finalize_completed
,
393 &job
->finalize_completed_notifier
);
394 notifier_list_add(&job
->job
.on_pending
, &job
->pending_notifier
);
396 error_setg(&job
->blocker
, "block device is in use by block job: %s",
397 job_type_str(&job
->job
));
398 block_job_add_bdrv(job
, "main node", bs
, 0, BLK_PERM_ALL
, &error_abort
);
401 bdrv_op_unblock(bs
, BLOCK_OP_TYPE_DATAPLANE
, job
->blocker
);
403 blk_add_aio_context_notifier(blk
, block_job_attached_aio_context
,
404 block_job_detach_aio_context
, job
);
406 /* Only set speed when necessary to avoid NotSupported error */
408 Error
*local_err
= NULL
;
410 block_job_set_speed(job
, speed
, &local_err
);
412 job_early_fail(&job
->job
);
413 error_propagate(errp
, local_err
);
421 void block_job_iostatus_reset(BlockJob
*job
)
423 if (job
->iostatus
== BLOCK_DEVICE_IO_STATUS_OK
) {
426 assert(job
->job
.user_paused
&& job
->job
.pause_count
> 0);
427 job
->iostatus
= BLOCK_DEVICE_IO_STATUS_OK
;
430 void block_job_user_resume(Job
*job
)
432 BlockJob
*bjob
= container_of(job
, BlockJob
, job
);
433 block_job_iostatus_reset(bjob
);
436 void block_job_event_ready(BlockJob
*job
)
438 job_state_transition(&job
->job
, JOB_STATUS_READY
);
441 if (block_job_is_internal(job
)) {
445 qapi_event_send_block_job_ready(job_type(&job
->job
),
449 job
->speed
, &error_abort
);
452 BlockErrorAction
block_job_error_action(BlockJob
*job
, BlockdevOnError on_err
,
453 int is_read
, int error
)
455 BlockErrorAction action
;
458 case BLOCKDEV_ON_ERROR_ENOSPC
:
459 case BLOCKDEV_ON_ERROR_AUTO
:
460 action
= (error
== ENOSPC
) ?
461 BLOCK_ERROR_ACTION_STOP
: BLOCK_ERROR_ACTION_REPORT
;
463 case BLOCKDEV_ON_ERROR_STOP
:
464 action
= BLOCK_ERROR_ACTION_STOP
;
466 case BLOCKDEV_ON_ERROR_REPORT
:
467 action
= BLOCK_ERROR_ACTION_REPORT
;
469 case BLOCKDEV_ON_ERROR_IGNORE
:
470 action
= BLOCK_ERROR_ACTION_IGNORE
;
475 if (!block_job_is_internal(job
)) {
476 qapi_event_send_block_job_error(job
->job
.id
,
477 is_read
? IO_OPERATION_TYPE_READ
:
478 IO_OPERATION_TYPE_WRITE
,
479 action
, &error_abort
);
481 if (action
== BLOCK_ERROR_ACTION_STOP
) {
482 job_pause(&job
->job
);
483 /* make the pause user visible, which will be resumed from QMP. */
484 job
->job
.user_paused
= true;
485 block_job_iostatus_set_err(job
, error
);