aboutsummaryrefslogtreecommitdiff
path: root/block/throttle-groups.c
blob: 784fa4a16c39e58e6a1eab7dc6934d1c6573a526 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
/*
 * QEMU block throttling group infrastructure
 *
 * Copyright (C) Nodalink, EURL. 2014
 * Copyright (C) Igalia, S.L. 2015
 *
 * Authors:
 *   Benoît Canet <benoit.canet@nodalink.com>
 *   Alberto Garcia <berto@igalia.com>
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation; either version 2 or
 * (at your option) version 3 of the License.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, see <http://www.gnu.org/licenses/>.
 */

#include "qemu/osdep.h"
#include "sysemu/block-backend.h"
#include "block/throttle-groups.h"
#include "qemu/throttle-options.h"
#include "qemu/main-loop.h"
#include "qemu/queue.h"
#include "qemu/thread.h"
#include "sysemu/qtest.h"
#include "qapi/error.h"
#include "qapi/qapi-visit-block-core.h"
#include "qom/object.h"
#include "qom/object_interfaces.h"

static void throttle_group_obj_init(Object *obj);
static void throttle_group_obj_complete(UserCreatable *obj, Error **errp);
static void timer_cb(ThrottleGroupMember *tgm, bool is_write);

/* The ThrottleGroup structure (with its ThrottleState) is shared
 * among different ThrottleGroupMembers and it's independent from
 * AioContext, so in order to use it from different threads it needs
 * its own locking.
 *
 * This locking is however handled internally in this file, so it's
 * transparent to outside users.
 *
 * The whole ThrottleGroup structure is private and invisible to
 * outside users, that only use it through its ThrottleState.
 *
 * In addition to the ThrottleGroup structure, ThrottleGroupMember has
 * fields that need to be accessed by other members of the group and
 * therefore also need to be protected by this lock. Once a
 * ThrottleGroupMember is registered in a group those fields can be accessed
 * by other threads any time.
 *
 * Again, all this is handled internally and is mostly transparent to
 * the outside. The 'throttle_timers' field however has an additional
 * constraint because it may be temporarily invalid (see for example
 * blk_set_aio_context()). Therefore in this file a thread will
 * access some other ThrottleGroupMember's timers only after verifying that
 * that ThrottleGroupMember has throttled requests in the queue.
 */
typedef struct ThrottleGroup {
    Object parent_obj;

    /* refuse individual property change if initialization is complete */
    bool is_initialized;
    char *name; /* This is constant during the lifetime of the group */

    QemuMutex lock; /* This lock protects the following four fields */
    ThrottleState ts;
    QLIST_HEAD(, ThrottleGroupMember) head;
    ThrottleGroupMember *tokens[2];
    bool any_timer_armed[2];
    QEMUClockType clock_type;

    /* This field is protected by the global QEMU mutex */
    QTAILQ_ENTRY(ThrottleGroup) list;
} ThrottleGroup;

/* This is protected by the global QEMU mutex */
static QTAILQ_HEAD(, ThrottleGroup) throttle_groups =
    QTAILQ_HEAD_INITIALIZER(throttle_groups);


/* This function reads throttle_groups and must be called under the global
 * mutex.
 */
static ThrottleGroup *throttle_group_by_name(const char *name)
{
    ThrottleGroup *iter;

    /* Look for an existing group with that name */
    QTAILQ_FOREACH(iter, &throttle_groups, list) {
        if (!g_strcmp0(name, iter->name)) {
            return iter;
        }
    }

    return NULL;
}

/* This function reads throttle_groups and must be called under the global
 * mutex.
 */
bool throttle_group_exists(const char *name)
{
    return throttle_group_by_name(name) != NULL;
}

/* Increments the reference count of a ThrottleGroup given its name.
 *
 * If no ThrottleGroup is found with the given name a new one is
 * created.
 *
 * This function edits throttle_groups and must be called under the global
 * mutex.
 *
 * @name: the name of the ThrottleGroup
 * @ret:  the ThrottleState member of the ThrottleGroup
 */
ThrottleState *throttle_group_incref(const char *name)
{
    ThrottleGroup *tg = NULL;

    /* Look for an existing group with that name */
    tg = throttle_group_by_name(name);

    if (tg) {
        object_ref(OBJECT(tg));
    } else {
        /* Create a new one if not found */
        /* new ThrottleGroup obj will have a refcnt = 1 */
        tg = THROTTLE_GROUP(object_new(TYPE_THROTTLE_GROUP));
        tg->name = g_strdup(name);
        throttle_group_obj_complete(USER_CREATABLE(tg), &error_abort);
    }

    return &tg->ts;
}

/* Decrease the reference count of a ThrottleGroup.
 *
 * When the reference count reaches zero the ThrottleGroup is
 * destroyed.
 *
 * This function edits throttle_groups and must be called under the global
 * mutex.
 *
 * @ts:  The ThrottleGroup to unref, given by its ThrottleState member
 */
void throttle_group_unref(ThrottleState *ts)
{
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
    object_unref(OBJECT(tg));
}

/* Get the name from a ThrottleGroupMember's group. The name (and the pointer)
 * is guaranteed to remain constant during the lifetime of the group.
 *
 * @tgm:  a ThrottleGroupMember
 * @ret:  the name of the group.
 */
const char *throttle_group_get_name(ThrottleGroupMember *tgm)
{
    ThrottleGroup *tg = container_of(tgm->throttle_state, ThrottleGroup, ts);
    return tg->name;
}

/* Return the next ThrottleGroupMember in the round-robin sequence, simulating
 * a circular list.
 *
 * This assumes that tg->lock is held.
 *
 * @tgm: the current ThrottleGroupMember
 * @ret: the next ThrottleGroupMember in the sequence
 */
static ThrottleGroupMember *throttle_group_next_tgm(ThrottleGroupMember *tgm)
{
    ThrottleState *ts = tgm->throttle_state;
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
    ThrottleGroupMember *next = QLIST_NEXT(tgm, round_robin);

    if (!next) {
        next = QLIST_FIRST(&tg->head);
    }

    return next;
}

/*
 * Return whether a ThrottleGroupMember has pending requests.
 *
 * This assumes that tg->lock is held.
 *
 * @tgm:        the ThrottleGroupMember
 * @is_write:   the type of operation (read/write)
 * @ret:        whether the ThrottleGroupMember has pending requests.
 */
static inline bool tgm_has_pending_reqs(ThrottleGroupMember *tgm,
                                        bool is_write)
{
    return tgm->pending_reqs[is_write];
}

/* Return the next ThrottleGroupMember in the round-robin sequence with pending
 * I/O requests.
 *
 * This assumes that tg->lock is held.
 *
 * @tgm:       the current ThrottleGroupMember
 * @is_write:  the type of operation (read/write)
 * @ret:       the next ThrottleGroupMember with pending requests, or tgm if
 *             there is none.
 */
static ThrottleGroupMember *next_throttle_token(ThrottleGroupMember *tgm,
                                                bool is_write)
{
    ThrottleState *ts = tgm->throttle_state;
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
    ThrottleGroupMember *token, *start;

    /* If this member has its I/O limits disabled then it means that
     * it's being drained. Skip the round-robin search and return tgm
     * immediately if it has pending requests. Otherwise we could be
     * forcing it to wait for other member's throttled requests. */
    if (tgm_has_pending_reqs(tgm, is_write) &&
        atomic_read(&tgm->io_limits_disabled)) {
        return tgm;
    }

    start = token = tg->tokens[is_write];

    /* get next bs round in round robin style */
    token = throttle_group_next_tgm(token);
    while (token != start && !tgm_has_pending_reqs(token, is_write)) {
        token = throttle_group_next_tgm(token);
    }

    /* If no IO are queued for scheduling on the next round robin token
     * then decide the token is the current tgm because chances are
     * the current tgm got the current request queued.
     */
    if (token == start && !tgm_has_pending_reqs(token, is_write)) {
        token = tgm;
    }

    /* Either we return the original TGM, or one with pending requests */
    assert(token == tgm || tgm_has_pending_reqs(token, is_write));

    return token;
}

/* Check if the next I/O request for a ThrottleGroupMember needs to be
 * throttled or not. If there's no timer set in this group, set one and update
 * the token accordingly.
 *
 * This assumes that tg->lock is held.
 *
 * @tgm:        the current ThrottleGroupMember
 * @is_write:   the type of operation (read/write)
 * @ret:        whether the I/O request needs to be throttled or not
 */
static bool throttle_group_schedule_timer(ThrottleGroupMember *tgm,
                                          bool is_write)
{
    ThrottleState *ts = tgm->throttle_state;
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
    ThrottleTimers *tt = &tgm->throttle_timers;
    bool must_wait;

    if (atomic_read(&tgm->io_limits_disabled)) {
        return false;
    }

    /* Check if any of the timers in this group is already armed */
    if (tg->any_timer_armed[is_write]) {
        return true;
    }

    must_wait = throttle_schedule_timer(ts, tt, is_write);

    /* If a timer just got armed, set tgm as the current token */
    if (must_wait) {
        tg->tokens[is_write] = tgm;
        tg->any_timer_armed[is_write] = true;
    }

    return must_wait;
}

/* Start the next pending I/O request for a ThrottleGroupMember. Return whether
 * any request was actually pending.
 *
 * @tgm:       the current ThrottleGroupMember
 * @is_write:  the type of operation (read/write)
 */
static bool coroutine_fn throttle_group_co_restart_queue(ThrottleGroupMember *tgm,
                                                         bool is_write)
{
    bool ret;

    qemu_co_mutex_lock(&tgm->throttled_reqs_lock);
    ret = qemu_co_queue_next(&tgm->throttled_reqs[is_write]);
    qemu_co_mutex_unlock(&tgm->throttled_reqs_lock);

    return ret;
}

/* Look for the next pending I/O request and schedule it.
 *
 * This assumes that tg->lock is held.
 *
 * @tgm:       the current ThrottleGroupMember
 * @is_write:  the type of operation (read/write)
 */
static void schedule_next_request(ThrottleGroupMember *tgm, bool is_write)
{
    ThrottleState *ts = tgm->throttle_state;
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
    bool must_wait;
    ThrottleGroupMember *token;

    /* Check if there's any pending request to schedule next */
    token = next_throttle_token(tgm, is_write);
    if (!tgm_has_pending_reqs(token, is_write)) {
        return;
    }

    /* Set a timer for the request if it needs to be throttled */
    must_wait = throttle_group_schedule_timer(token, is_write);

    /* If it doesn't have to wait, queue it for immediate execution */
    if (!must_wait) {
        /* Give preference to requests from the current tgm */
        if (qemu_in_coroutine() &&
            throttle_group_co_restart_queue(tgm, is_write)) {
            token = tgm;
        } else {
            ThrottleTimers *tt = &token->throttle_timers;
            int64_t now = qemu_clock_get_ns(tg->clock_type);
            timer_mod(tt->timers[is_write], now);
            tg->any_timer_armed[is_write] = true;
        }
        tg->tokens[is_write] = token;
    }
}

/* Check if an I/O request needs to be throttled, wait and set a timer
 * if necessary, and schedule the next request using a round robin
 * algorithm.
 *
 * @tgm:       the current ThrottleGroupMember
 * @bytes:     the number of bytes for this I/O
 * @is_write:  the type of operation (read/write)
 */
void coroutine_fn throttle_group_co_io_limits_intercept(ThrottleGroupMember *tgm,
                                                        unsigned int bytes,
                                                        bool is_write)
{
    bool must_wait;
    ThrottleGroupMember *token;
    ThrottleGroup *tg = container_of(tgm->throttle_state, ThrottleGroup, ts);
    qemu_mutex_lock(&tg->lock);

    /* First we check if this I/O has to be throttled. */
    token = next_throttle_token(tgm, is_write);
    must_wait = throttle_group_schedule_timer(token, is_write);

    /* Wait if there's a timer set or queued requests of this type */
    if (must_wait || tgm->pending_reqs[is_write]) {
        tgm->pending_reqs[is_write]++;
        qemu_mutex_unlock(&tg->lock);
        qemu_co_mutex_lock(&tgm->throttled_reqs_lock);
        qemu_co_queue_wait(&tgm->throttled_reqs[is_write],
                           &tgm->throttled_reqs_lock);
        qemu_co_mutex_unlock(&tgm->throttled_reqs_lock);
        qemu_mutex_lock(&tg->lock);
        tgm->pending_reqs[is_write]--;
    }

    /* The I/O will be executed, so do the accounting */
    throttle_account(tgm->throttle_state, is_write, bytes);

    /* Schedule the next request */
    schedule_next_request(tgm, is_write);

    qemu_mutex_unlock(&tg->lock);
}

typedef struct {
    ThrottleGroupMember *tgm;
    bool is_write;
} RestartData;

static void coroutine_fn throttle_group_restart_queue_entry(void *opaque)
{
    RestartData *data = opaque;
    ThrottleGroupMember *tgm = data->tgm;
    ThrottleState *ts = tgm->throttle_state;
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
    bool is_write = data->is_write;
    bool empty_queue;

    empty_queue = !throttle_group_co_restart_queue(tgm, is_write);

    /* If the request queue was empty then we have to take care of
     * scheduling the next one */
    if (empty_queue) {
        qemu_mutex_lock(&tg->lock);
        schedule_next_request(tgm, is_write);
        qemu_mutex_unlock(&tg->lock);
    }

    g_free(data);

    atomic_dec(&tgm->restart_pending);
    aio_wait_kick();
}

static void throttle_group_restart_queue(ThrottleGroupMember *tgm, bool is_write)
{
    Coroutine *co;
    RestartData *rd = g_new0(RestartData, 1);

    rd->tgm = tgm;
    rd->is_write = is_write;

    /* This function is called when a timer is fired or when
     * throttle_group_restart_tgm() is called. Either way, there can
     * be no timer pending on this tgm at this point */
    assert(!timer_pending(tgm->throttle_timers.timers[is_write]));

    atomic_inc(&tgm->restart_pending);

    co = qemu_coroutine_create(throttle_group_restart_queue_entry, rd);
    aio_co_enter(tgm->aio_context, co);
}

void throttle_group_restart_tgm(ThrottleGroupMember *tgm)
{
    int i;

    if (tgm->throttle_state) {
        for (i = 0; i < 2; i++) {
            QEMUTimer *t = tgm->throttle_timers.timers[i];
            if (timer_pending(t)) {
                /* If there's a pending timer on this tgm, fire it now */
                timer_del(t);
                timer_cb(tgm, i);
            } else {
                /* Else run the next request from the queue manually */
                throttle_group_restart_queue(tgm, i);
            }
        }
    }
}

/* Update the throttle configuration for a particular group. Similar
 * to throttle_config(), but guarantees atomicity within the
 * throttling group.
 *
 * @tgm:    a ThrottleGroupMember that is a member of the group
 * @cfg: the configuration to set
 */
void throttle_group_config(ThrottleGroupMember *tgm, ThrottleConfig *cfg)
{
    ThrottleState *ts = tgm->throttle_state;
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
    qemu_mutex_lock(&tg->lock);
    throttle_config(ts, tg->clock_type, cfg);
    qemu_mutex_unlock(&tg->lock);

    throttle_group_restart_tgm(tgm);
}

/* Get the throttle configuration from a particular group. Similar to
 * throttle_get_config(), but guarantees atomicity within the
 * throttling group.
 *
 * @tgm:    a ThrottleGroupMember that is a member of the group
 * @cfg: the configuration will be written here
 */
void throttle_group_get_config(ThrottleGroupMember *tgm, ThrottleConfig *cfg)
{
    ThrottleState *ts = tgm->throttle_state;
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
    qemu_mutex_lock(&tg->lock);
    throttle_get_config(ts, cfg);
    qemu_mutex_unlock(&tg->lock);
}

/* ThrottleTimers callback. This wakes up a request that was waiting
 * because it had been throttled.
 *
 * @tgm:       the ThrottleGroupMember whose request had been throttled
 * @is_write:  the type of operation (read/write)
 */
static void timer_cb(ThrottleGroupMember *tgm, bool is_write)
{
    ThrottleState *ts = tgm->throttle_state;
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);

    /* The timer has just been fired, so we can update the flag */
    qemu_mutex_lock(&tg->lock);
    tg->any_timer_armed[is_write] = false;
    qemu_mutex_unlock(&tg->lock);

    /* Run the request that was waiting for this timer */
    throttle_group_restart_queue(tgm, is_write);
}

static void read_timer_cb(void *opaque)
{
    timer_cb(opaque, false);
}

static void write_timer_cb(void *opaque)
{
    timer_cb(opaque, true);
}

/* Register a ThrottleGroupMember from the throttling group, also initializing
 * its timers and updating its throttle_state pointer to point to it. If a
 * throttling group with that name does not exist yet, it will be created.
 *
 * This function edits throttle_groups and must be called under the global
 * mutex.
 *
 * @tgm:       the ThrottleGroupMember to insert
 * @groupname: the name of the group
 * @ctx:       the AioContext to use
 */
void throttle_group_register_tgm(ThrottleGroupMember *tgm,
                                 const char *groupname,
                                 AioContext *ctx)
{
    int i;
    ThrottleState *ts = throttle_group_incref(groupname);
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);

    tgm->throttle_state = ts;
    tgm->aio_context = ctx;
    atomic_set(&tgm->restart_pending, 0);

    qemu_mutex_lock(&tg->lock);
    /* If the ThrottleGroup is new set this ThrottleGroupMember as the token */
    for (i = 0; i < 2; i++) {
        if (!tg->tokens[i]) {
            tg->tokens[i] = tgm;
        }
    }

    QLIST_INSERT_HEAD(&tg->head, tgm, round_robin);

    throttle_timers_init(&tgm->throttle_timers,
                         tgm->aio_context,
                         tg->clock_type,
                         read_timer_cb,
                         write_timer_cb,
                         tgm);
    qemu_co_mutex_init(&tgm->throttled_reqs_lock);
    qemu_co_queue_init(&tgm->throttled_reqs[0]);
    qemu_co_queue_init(&tgm->throttled_reqs[1]);

    qemu_mutex_unlock(&tg->lock);
}

/* Unregister a ThrottleGroupMember from its group, removing it from the list,
 * destroying the timers and setting the throttle_state pointer to NULL.
 *
 * The ThrottleGroupMember must not have pending throttled requests, so the
 * caller has to drain them first.
 *
 * The group will be destroyed if it's empty after this operation.
 *
 * @tgm the ThrottleGroupMember to remove
 */
void throttle_group_unregister_tgm(ThrottleGroupMember *tgm)
{
    ThrottleState *ts = tgm->throttle_state;
    ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
    ThrottleGroupMember *token;
    int i;

    if (!ts) {
        /* Discard already unregistered tgm */
        return;
    }

    /* Wait for throttle_group_restart_queue_entry() coroutines to finish */
    AIO_WAIT_WHILE(tgm->aio_context, atomic_read(&tgm->restart_pending) > 0);

    qemu_mutex_lock(&tg->lock);
    for (i = 0; i < 2; i++) {
        assert(tgm->pending_reqs[i] == 0);
        assert(qemu_co_queue_empty(&tgm->throttled_reqs[i]));
        assert(!timer_pending(tgm->throttle_timers.timers[i]));
        if (tg->tokens[i] == tgm) {
            token = throttle_group_next_tgm(tgm);
            /* Take care of the case where this is the last tgm in the group */
            if (token == tgm) {
                token = NULL;
            }
            tg->tokens[i] = token;
        }
    }

    /* remove the current tgm from the list */
    QLIST_REMOVE(tgm, round_robin);
    throttle_timers_destroy(&tgm->throttle_timers);
    qemu_mutex_unlock(&tg->lock);

    throttle_group_unref(&tg->ts);
    tgm->throttle_state = NULL;
}

void throttle_group_attach_aio_context(ThrottleGroupMember *tgm,
                                       AioContext *new_context)
{
    ThrottleTimers *tt = &tgm->throttle_timers;
    throttle_timers_attach_aio_context(tt, new_context);
    tgm->aio_context = new_context;
}

void throttle_group_detach_aio_context(ThrottleGroupMember *tgm)
{
    ThrottleGroup *tg = container_of(tgm->throttle_state, ThrottleGroup, ts);
    ThrottleTimers *tt = &tgm->throttle_timers;
    int i;

    /* Requests must have been drained */
    assert(tgm->pending_reqs[0] == 0 && tgm->pending_reqs[1] == 0);
    assert(qemu_co_queue_empty(&tgm->throttled_reqs[0]));
    assert(qemu_co_queue_empty(&tgm->throttled_reqs[1]));

    /* Kick off next ThrottleGroupMember, if necessary */
    qemu_mutex_lock(&tg->lock);
    for (i = 0; i < 2; i++) {
        if (timer_pending(tt->timers[i])) {
            tg->any_timer_armed[i] = false;
            schedule_next_request(tgm, i);
        }
    }
    qemu_mutex_unlock(&tg->lock);

    throttle_timers_detach_aio_context(tt);
    tgm->aio_context = NULL;
}

#undef THROTTLE_OPT_PREFIX
#define THROTTLE_OPT_PREFIX "x-"

/* Helper struct and array for QOM property setter/getter */
typedef struct {
    const char *name;
    BucketType type;
    enum {
        AVG,
        MAX,
        BURST_LENGTH,
        IOPS_SIZE,
    } category;
} ThrottleParamInfo;

static ThrottleParamInfo properties[] = {
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_TOTAL,
        THROTTLE_OPS_TOTAL, AVG,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_TOTAL_MAX,
        THROTTLE_OPS_TOTAL, MAX,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_TOTAL_MAX_LENGTH,
        THROTTLE_OPS_TOTAL, BURST_LENGTH,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_READ,
        THROTTLE_OPS_READ, AVG,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_READ_MAX,
        THROTTLE_OPS_READ, MAX,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_READ_MAX_LENGTH,
        THROTTLE_OPS_READ, BURST_LENGTH,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_WRITE,
        THROTTLE_OPS_WRITE, AVG,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_WRITE_MAX,
        THROTTLE_OPS_WRITE, MAX,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_WRITE_MAX_LENGTH,
        THROTTLE_OPS_WRITE, BURST_LENGTH,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_BPS_TOTAL,
        THROTTLE_BPS_TOTAL, AVG,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_BPS_TOTAL_MAX,
        THROTTLE_BPS_TOTAL, MAX,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_BPS_TOTAL_MAX_LENGTH,
        THROTTLE_BPS_TOTAL, BURST_LENGTH,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_BPS_READ,
        THROTTLE_BPS_READ, AVG,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_BPS_READ_MAX,
        THROTTLE_BPS_READ, MAX,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_BPS_READ_MAX_LENGTH,
        THROTTLE_BPS_READ, BURST_LENGTH,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_BPS_WRITE,
        THROTTLE_BPS_WRITE, AVG,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_BPS_WRITE_MAX,
        THROTTLE_BPS_WRITE, MAX,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_BPS_WRITE_MAX_LENGTH,
        THROTTLE_BPS_WRITE, BURST_LENGTH,
    },
    {
        THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_SIZE,
        0, IOPS_SIZE,
    }
};

/* This function edits throttle_groups and must be called under the global
 * mutex */
static void throttle_group_obj_init(Object *obj)
{
    ThrottleGroup *tg = THROTTLE_GROUP(obj);

    tg->clock_type = QEMU_CLOCK_REALTIME;
    if (qtest_enabled()) {
        /* For testing block IO throttling only */
        tg->clock_type = QEMU_CLOCK_VIRTUAL;
    }
    tg->is_initialized = false;
    qemu_mutex_init(&tg->lock);
    throttle_init(&tg->ts);
    QLIST_INIT(&tg->head);
}

/* This function edits throttle_groups and must be called under the global
 * mutex */
static void throttle_group_obj_complete(UserCreatable *obj, Error **errp)
{
    ThrottleGroup *tg = THROTTLE_GROUP(obj);
    ThrottleConfig cfg;

    /* set group name to object id if it exists */
    if (!tg->name && tg->parent_obj.parent) {
        tg->name = object_get_canonical_path_component(OBJECT(obj));
    }
    /* We must have a group name at this point */
    assert(tg->name);

    /* error if name is duplicate */
    if (throttle_group_exists(tg->name)) {
        error_setg(errp, "A group with this name already exists");
        return;
    }

    /* check validity */
    throttle_get_config(&tg->ts, &cfg);
    if (!throttle_is_valid(&cfg, errp)) {
        return;
    }
    throttle_config(&tg->ts, tg->clock_type, &cfg);
    QTAILQ_INSERT_TAIL(&throttle_groups, tg, list);
    tg->is_initialized = true;
}

/* This function edits throttle_groups and must be called under the global
 * mutex */
static void throttle_group_obj_finalize(Object *obj)
{
    ThrottleGroup *tg = THROTTLE_GROUP(obj);
    if (tg->is_initialized) {
        QTAILQ_REMOVE(&throttle_groups, tg, list);
    }
    qemu_mutex_destroy(&tg->lock);
    g_free(tg->name);
}

static void throttle_group_set(Object *obj, Visitor *v, const char * name,
                               void *opaque, Error **errp)

{
    ThrottleGroup *tg = THROTTLE_GROUP(obj);
    ThrottleConfig *cfg;
    ThrottleParamInfo *info = opaque;
    Error *local_err = NULL;
    int64_t value;

    /* If we have finished initialization, don't accept individual property
     * changes through QOM. Throttle configuration limits must be set in one
     * transaction, as certain combinations are invalid.
     */
    if (tg->is_initialized) {
        error_setg(&local_err, "Property cannot be set after initialization");
        goto ret;
    }

    visit_type_int64(v, name, &value, &local_err);
    if (local_err) {
        goto ret;
    }
    if (value < 0) {
        error_setg(&local_err, "Property values cannot be negative");
        goto ret;
    }

    cfg = &tg->ts.cfg;
    switch (info->category) {
    case AVG:
        cfg->buckets[info->type].avg = value;
        break;
    case MAX:
        cfg->buckets[info->type].max = value;
        break;
    case BURST_LENGTH:
        if (value > UINT_MAX) {
            error_setg(&local_err, "%s value must be in the"
                       "range [0, %u]", info->name, UINT_MAX);
            goto ret;
        }
        cfg->buckets[info->type].burst_length = value;
        break;
    case IOPS_SIZE:
        cfg->op_size = value;
        break;
    }

ret:
    error_propagate(errp, local_err);
    return;

}

static void throttle_group_get(Object *obj, Visitor *v, const char *name,
                               void *opaque, Error **errp)
{
    ThrottleGroup *tg = THROTTLE_GROUP(obj);
    ThrottleConfig cfg;
    ThrottleParamInfo *info = opaque;
    int64_t value;

    throttle_get_config(&tg->ts, &cfg);
    switch (info->category) {
    case AVG:
        value = cfg.buckets[info->type].avg;
        break;
    case MAX:
        value = cfg.buckets[info->type].max;
        break;
    case BURST_LENGTH:
        value = cfg.buckets[info->type].burst_length;
        break;
    case IOPS_SIZE:
        value = cfg.op_size;
        break;
    }

    visit_type_int64(v, name, &value, errp);
}

static void throttle_group_set_limits(Object *obj, Visitor *v,
                                      const char *name, void *opaque,
                                      Error **errp)

{
    ThrottleGroup *tg = THROTTLE_GROUP(obj);
    ThrottleConfig cfg;
    ThrottleLimits *argp;
    Error *local_err = NULL;

    visit_type_ThrottleLimits(v, name, &argp, &local_err);
    if (local_err) {
        goto ret;
    }
    qemu_mutex_lock(&tg->lock);
    throttle_get_config(&tg->ts, &cfg);
    throttle_limits_to_config(argp, &cfg, &local_err);
    if (local_err) {
        goto unlock;
    }
    throttle_config(&tg->ts, tg->clock_type, &cfg);

unlock:
    qemu_mutex_unlock(&tg->lock);
ret:
    qapi_free_ThrottleLimits(argp);
    error_propagate(errp, local_err);
    return;
}

static void throttle_group_get_limits(Object *obj, Visitor *v,
                                      const char *name, void *opaque,
                                      Error **errp)
{
    ThrottleGroup *tg = THROTTLE_GROUP(obj);
    ThrottleConfig cfg;
    ThrottleLimits arg = { 0 };
    ThrottleLimits *argp = &arg;

    qemu_mutex_lock(&tg->lock);
    throttle_get_config(&tg->ts, &cfg);
    qemu_mutex_unlock(&tg->lock);

    throttle_config_to_limits(&cfg, argp);

    visit_type_ThrottleLimits(v, name, &argp, errp);
}

static bool throttle_group_can_be_deleted(UserCreatable *uc)
{
    return OBJECT(uc)->ref == 1;
}

static void throttle_group_obj_class_init(ObjectClass *klass, void *class_data)
{
    size_t i = 0;
    UserCreatableClass *ucc = USER_CREATABLE_CLASS(klass);

    ucc->complete = throttle_group_obj_complete;
    ucc->can_be_deleted = throttle_group_can_be_deleted;

    /* individual properties */
    for (i = 0; i < sizeof(properties) / sizeof(ThrottleParamInfo); i++) {
        object_class_property_add(klass,
                                  properties[i].name,
                                  "int",
                                  throttle_group_get,
                                  throttle_group_set,
                                  NULL, &properties[i]);
    }

    /* ThrottleLimits */
    object_class_property_add(klass,
                              "limits", "ThrottleLimits",
                              throttle_group_get_limits,
                              throttle_group_set_limits,
                              NULL, NULL);
}

static const TypeInfo throttle_group_info = {
    .name = TYPE_THROTTLE_GROUP,
    .parent = TYPE_OBJECT,
    .class_init = throttle_group_obj_class_init,
    .instance_size = sizeof(ThrottleGroup),
    .instance_init = throttle_group_obj_init,
    .instance_finalize = throttle_group_obj_finalize,
    .interfaces = (InterfaceInfo[]) {
        { TYPE_USER_CREATABLE },
        { }
    },
};

static void throttle_groups_init(void)
{
    type_register_static(&throttle_group_info);
}

type_init(throttle_groups_init);