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
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
|
/*-
* Copyright (c) 2008, Jeffrey Roberson <jeff@freebsd.org>
* All rights reserved.
*
* Copyright (c) 2008 Nokia Corporation
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice unmodified, this list of conditions, and the following
* disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_ddb.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/sysproto.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/priv.h>
#include <sys/proc.h>
#include <sys/refcount.h>
#include <sys/sched.h>
#include <sys/smp.h>
#include <sys/syscallsubr.h>
#include <sys/cpuset.h>
#include <sys/sx.h>
#include <sys/refcount.h>
#include <sys/queue.h>
#include <sys/limits.h>
#include <sys/bus.h>
#include <sys/interrupt.h>
#include <sys/jail.h> /* Must come after sys/proc.h */
#include <vm/uma.h>
#ifdef DDB
#include <ddb/ddb.h>
#endif /* DDB */
/*
* cpusets provide a mechanism for creating and manipulating sets of
* processors for the purpose of constraining the scheduling of threads to
* specific processors.
*
* Each process belongs to an identified set, by default this is set 1. Each
* thread may further restrict the cpus it may run on to a subset of this
* named set. This creates an anonymous set which other threads and processes
* may not join by number.
*
* The named set is referred to herein as the 'base' set to avoid ambiguity.
* This set is usually a child of a 'root' set while the anonymous set may
* simply be referred to as a mask. In the syscall api these are referred to
* as the ROOT, CPUSET, and MASK levels where CPUSET is called 'base' here.
*
* Threads inherit their set from their creator whether it be anonymous or
* not. This means that anonymous sets are immutable because they may be
* shared. To modify an anonymous set a new set is created with the desired
* mask and the same parent as the existing anonymous set. This gives the
* illusion of each thread having a private mask.A
*
* Via the syscall apis a user may ask to retrieve or modify the root, base,
* or mask that is discovered via a pid, tid, or setid. Modifying a set
* modifies all numbered and anonymous child sets to comply with the new mask.
* Modifying a pid or tid's mask applies only to that tid but must still
* exist within the assigned parent set.
*
* A thread may not be assigned to a a group seperate from other threads in
* the process. This is to remove ambiguity when the setid is queried with
* a pid argument. There is no other technical limitation.
*
* This somewhat complex arrangement is intended to make it easy for
* applications to query available processors and bind their threads to
* specific processors while also allowing administrators to dynamically
* reprovision by changing sets which apply to groups of processes.
*
* A simple application should not concern itself with sets at all and
* rather apply masks to its own threads via CPU_WHICH_TID and a -1 id
* meaning 'curthread'. It may query availble cpus for that tid with a
* getaffinity call using (CPU_LEVEL_CPUSET, CPU_WHICH_PID, -1, ...).
*/
static uma_zone_t cpuset_zone;
static struct mtx cpuset_lock;
static struct setlist cpuset_ids;
static struct unrhdr *cpuset_unr;
static struct cpuset *cpuset_zero;
cpuset_t *cpuset_root;
/*
* Acquire a reference to a cpuset, all pointers must be tracked with refs.
*/
struct cpuset *
cpuset_ref(struct cpuset *set)
{
refcount_acquire(&set->cs_ref);
return (set);
}
/*
* Walks up the tree from 'set' to find the root. Returns the root
* referenced.
*/
static struct cpuset *
cpuset_refroot(struct cpuset *set)
{
for (; set->cs_parent != NULL; set = set->cs_parent)
if (set->cs_flags & CPU_SET_ROOT)
break;
cpuset_ref(set);
return (set);
}
/*
* Find the first non-anonymous set starting from 'set'. Returns this set
* referenced. May return the passed in set with an extra ref if it is
* not anonymous.
*/
static struct cpuset *
cpuset_refbase(struct cpuset *set)
{
if (set->cs_id == CPUSET_INVALID)
set = set->cs_parent;
cpuset_ref(set);
return (set);
}
/*
* Release a reference in a context where it is safe to allocte.
*/
void
cpuset_rel(struct cpuset *set)
{
cpusetid_t id;
if (refcount_release(&set->cs_ref) == 0)
return;
mtx_lock_spin(&cpuset_lock);
LIST_REMOVE(set, cs_siblings);
id = set->cs_id;
if (id != CPUSET_INVALID)
LIST_REMOVE(set, cs_link);
mtx_unlock_spin(&cpuset_lock);
cpuset_rel(set->cs_parent);
uma_zfree(cpuset_zone, set);
if (id != CPUSET_INVALID)
free_unr(cpuset_unr, id);
}
/*
* Deferred release must be used when in a context that is not safe to
* allocate/free. This places any unreferenced sets on the list 'head'.
*/
static void
cpuset_rel_defer(struct setlist *head, struct cpuset *set)
{
if (refcount_release(&set->cs_ref) == 0)
return;
mtx_lock_spin(&cpuset_lock);
LIST_REMOVE(set, cs_siblings);
if (set->cs_id != CPUSET_INVALID)
LIST_REMOVE(set, cs_link);
LIST_INSERT_HEAD(head, set, cs_link);
mtx_unlock_spin(&cpuset_lock);
}
/*
* Complete a deferred release. Removes the set from the list provided to
* cpuset_rel_defer.
*/
static void
cpuset_rel_complete(struct cpuset *set)
{
LIST_REMOVE(set, cs_link);
cpuset_rel(set->cs_parent);
uma_zfree(cpuset_zone, set);
}
/*
* Find a set based on an id. Returns it with a ref.
*/
static struct cpuset *
cpuset_lookup(cpusetid_t setid, struct thread *td)
{
struct cpuset *set;
if (setid == CPUSET_INVALID)
return (NULL);
mtx_lock_spin(&cpuset_lock);
LIST_FOREACH(set, &cpuset_ids, cs_link)
if (set->cs_id == setid)
break;
if (set)
cpuset_ref(set);
mtx_unlock_spin(&cpuset_lock);
KASSERT(td != NULL, ("[%s:%d] td is NULL", __func__, __LINE__));
if (set != NULL && jailed(td->td_ucred)) {
struct cpuset *rset, *jset;
struct prison *pr;
rset = cpuset_refroot(set);
pr = td->td_ucred->cr_prison;
mtx_lock(&pr->pr_mtx);
cpuset_ref(pr->pr_cpuset);
jset = pr->pr_cpuset;
mtx_unlock(&pr->pr_mtx);
if (jset->cs_id != rset->cs_id) {
cpuset_rel(set);
set = NULL;
}
cpuset_rel(jset);
cpuset_rel(rset);
}
return (set);
}
/*
* Create a set in the space provided in 'set' with the provided parameters.
* The set is returned with a single ref. May return EDEADLK if the set
* will have no valid cpu based on restrictions from the parent.
*/
static int
_cpuset_create(struct cpuset *set, struct cpuset *parent, cpuset_t *mask,
cpusetid_t id)
{
if (!CPU_OVERLAP(&parent->cs_mask, mask))
return (EDEADLK);
CPU_COPY(mask, &set->cs_mask);
LIST_INIT(&set->cs_children);
refcount_init(&set->cs_ref, 1);
set->cs_flags = 0;
mtx_lock_spin(&cpuset_lock);
CPU_AND(mask, &parent->cs_mask);
set->cs_id = id;
set->cs_parent = cpuset_ref(parent);
LIST_INSERT_HEAD(&parent->cs_children, set, cs_siblings);
if (set->cs_id != CPUSET_INVALID)
LIST_INSERT_HEAD(&cpuset_ids, set, cs_link);
mtx_unlock_spin(&cpuset_lock);
return (0);
}
/*
* Create a new non-anonymous set with the requested parent and mask. May
* return failures if the mask is invalid or a new number can not be
* allocated.
*/
static int
cpuset_create(struct cpuset **setp, struct cpuset *parent, cpuset_t *mask)
{
struct cpuset *set;
cpusetid_t id;
int error;
id = alloc_unr(cpuset_unr);
if (id == -1)
return (ENFILE);
*setp = set = uma_zalloc(cpuset_zone, M_WAITOK);
error = _cpuset_create(set, parent, mask, id);
if (error == 0)
return (0);
free_unr(cpuset_unr, id);
uma_zfree(cpuset_zone, set);
return (error);
}
/*
* Recursively check for errors that would occur from applying mask to
* the tree of sets starting at 'set'. Checks for sets that would become
* empty as well as RDONLY flags.
*/
static int
cpuset_testupdate(struct cpuset *set, cpuset_t *mask)
{
struct cpuset *nset;
cpuset_t newmask;
int error;
mtx_assert(&cpuset_lock, MA_OWNED);
if (set->cs_flags & CPU_SET_RDONLY)
return (EPERM);
if (!CPU_OVERLAP(&set->cs_mask, mask))
return (EDEADLK);
CPU_COPY(&set->cs_mask, &newmask);
CPU_AND(&newmask, mask);
error = 0;
LIST_FOREACH(nset, &set->cs_children, cs_siblings)
if ((error = cpuset_testupdate(nset, &newmask)) != 0)
break;
return (error);
}
/*
* Applies the mask 'mask' without checking for empty sets or permissions.
*/
static void
cpuset_update(struct cpuset *set, cpuset_t *mask)
{
struct cpuset *nset;
mtx_assert(&cpuset_lock, MA_OWNED);
CPU_AND(&set->cs_mask, mask);
LIST_FOREACH(nset, &set->cs_children, cs_siblings)
cpuset_update(nset, &set->cs_mask);
return;
}
/*
* Modify the set 'set' to use a copy of the mask provided. Apply this new
* mask to restrict all children in the tree. Checks for validity before
* applying the changes.
*/
static int
cpuset_modify(struct cpuset *set, cpuset_t *mask)
{
struct cpuset *root;
int error;
error = priv_check(curthread, PRIV_SCHED_CPUSET);
if (error)
return (error);
/*
* Verify that we have access to this set of
* cpus.
*/
root = set->cs_parent;
if (root && !CPU_SUBSET(&root->cs_mask, mask))
return (EINVAL);
mtx_lock_spin(&cpuset_lock);
error = cpuset_testupdate(set, mask);
if (error)
goto out;
cpuset_update(set, mask);
CPU_COPY(mask, &set->cs_mask);
out:
mtx_unlock_spin(&cpuset_lock);
return (error);
}
/*
* Resolve the 'which' parameter of several cpuset apis.
*
* For WHICH_PID and WHICH_TID return a locked proc and valid proc/tid. Also
* checks for permission via p_cansched().
*
* For WHICH_SET returns a valid set with a new reference.
*
* -1 may be supplied for any argument to mean the current proc/thread or
* the base set of the current thread. May fail with ESRCH/EPERM.
*/
static int
cpuset_which(cpuwhich_t which, id_t id, struct proc **pp, struct thread **tdp,
struct cpuset **setp)
{
struct cpuset *set;
struct thread *td;
struct proc *p;
int error;
*pp = p = NULL;
*tdp = td = NULL;
*setp = set = NULL;
switch (which) {
case CPU_WHICH_PID:
if (id == -1) {
PROC_LOCK(curproc);
p = curproc;
break;
}
if ((p = pfind(id)) == NULL)
return (ESRCH);
break;
case CPU_WHICH_TID:
if (id == -1) {
PROC_LOCK(curproc);
p = curproc;
td = curthread;
break;
}
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
PROC_LOCK(p);
FOREACH_THREAD_IN_PROC(p, td)
if (td->td_tid == id)
break;
if (td != NULL)
break;
PROC_UNLOCK(p);
}
sx_sunlock(&allproc_lock);
if (td == NULL)
return (ESRCH);
break;
case CPU_WHICH_CPUSET:
if (id == -1) {
thread_lock(curthread);
set = cpuset_refbase(curthread->td_cpuset);
thread_unlock(curthread);
} else
set = cpuset_lookup(id, curthread);
if (set) {
*setp = set;
return (0);
}
return (ESRCH);
case CPU_WHICH_JAIL:
{
/* Find `set' for prison with given id. */
struct prison *pr;
sx_slock(&allprison_lock);
pr = prison_find(id);
sx_sunlock(&allprison_lock);
if (pr == NULL)
return (ESRCH);
if (jailed(curthread->td_ucred)) {
if (curthread->td_ucred->cr_prison == pr) {
cpuset_ref(pr->pr_cpuset);
set = pr->pr_cpuset;
}
} else {
cpuset_ref(pr->pr_cpuset);
set = pr->pr_cpuset;
}
mtx_unlock(&pr->pr_mtx);
if (set) {
*setp = set;
return (0);
}
return (ESRCH);
}
case CPU_WHICH_IRQ:
return (0);
default:
return (EINVAL);
}
error = p_cansched(curthread, p);
if (error) {
PROC_UNLOCK(p);
return (error);
}
if (td == NULL)
td = FIRST_THREAD_IN_PROC(p);
*pp = p;
*tdp = td;
return (0);
}
/*
* Create an anonymous set with the provided mask in the space provided by
* 'fset'. If the passed in set is anonymous we use its parent otherwise
* the new set is a child of 'set'.
*/
static int
cpuset_shadow(struct cpuset *set, struct cpuset *fset, cpuset_t *mask)
{
struct cpuset *parent;
if (set->cs_id == CPUSET_INVALID)
parent = set->cs_parent;
else
parent = set;
if (!CPU_SUBSET(&parent->cs_mask, mask))
return (EDEADLK);
return (_cpuset_create(fset, parent, mask, CPUSET_INVALID));
}
/*
* Handle two cases for replacing the base set or mask of an entire process.
*
* 1) Set is non-null and mask is null. This reparents all anonymous sets
* to the provided set and replaces all non-anonymous td_cpusets with the
* provided set.
* 2) Mask is non-null and set is null. This replaces or creates anonymous
* sets for every thread with the existing base as a parent.
*
* This is overly complicated because we can't allocate while holding a
* spinlock and spinlocks must be held while changing and examining thread
* state.
*/
static int
cpuset_setproc(pid_t pid, struct cpuset *set, cpuset_t *mask)
{
struct setlist freelist;
struct setlist droplist;
struct cpuset *tdset;
struct cpuset *nset;
struct thread *td;
struct proc *p;
int threads;
int nfree;
int error;
/*
* The algorithm requires two passes due to locking considerations.
*
* 1) Lookup the process and acquire the locks in the required order.
* 2) If enough cpusets have not been allocated release the locks and
* allocate them. Loop.
*/
LIST_INIT(&freelist);
LIST_INIT(&droplist);
nfree = 0;
for (;;) {
error = cpuset_which(CPU_WHICH_PID, pid, &p, &td, &nset);
if (error)
goto out;
if (nfree >= p->p_numthreads)
break;
threads = p->p_numthreads;
PROC_UNLOCK(p);
for (; nfree < threads; nfree++) {
nset = uma_zalloc(cpuset_zone, M_WAITOK);
LIST_INSERT_HEAD(&freelist, nset, cs_link);
}
}
PROC_LOCK_ASSERT(p, MA_OWNED);
/*
* Now that the appropriate locks are held and we have enough cpusets,
* make sure the operation will succeed before applying changes. The
* proc lock prevents td_cpuset from changing between calls.
*/
error = 0;
FOREACH_THREAD_IN_PROC(p, td) {
thread_lock(td);
tdset = td->td_cpuset;
/*
* Verify that a new mask doesn't specify cpus outside of
* the set the thread is a member of.
*/
if (mask) {
if (tdset->cs_id == CPUSET_INVALID)
tdset = tdset->cs_parent;
if (!CPU_SUBSET(&tdset->cs_mask, mask))
error = EDEADLK;
/*
* Verify that a new set won't leave an existing thread
* mask without a cpu to run on. It can, however, restrict
* the set.
*/
} else if (tdset->cs_id == CPUSET_INVALID) {
if (!CPU_OVERLAP(&set->cs_mask, &tdset->cs_mask))
error = EDEADLK;
}
thread_unlock(td);
if (error)
goto unlock_out;
}
/*
* Replace each thread's cpuset while using deferred release. We
* must do this because the thread lock must be held while operating
* on the thread and this limits the type of operations allowed.
*/
FOREACH_THREAD_IN_PROC(p, td) {
thread_lock(td);
/*
* If we presently have an anonymous set or are applying a
* mask we must create an anonymous shadow set. That is
* either parented to our existing base or the supplied set.
*
* If we have a base set with no anonymous shadow we simply
* replace it outright.
*/
tdset = td->td_cpuset;
if (tdset->cs_id == CPUSET_INVALID || mask) {
nset = LIST_FIRST(&freelist);
LIST_REMOVE(nset, cs_link);
if (mask)
error = cpuset_shadow(tdset, nset, mask);
else
error = _cpuset_create(nset, set,
&tdset->cs_mask, CPUSET_INVALID);
if (error) {
LIST_INSERT_HEAD(&freelist, nset, cs_link);
thread_unlock(td);
break;
}
} else
nset = cpuset_ref(set);
cpuset_rel_defer(&droplist, tdset);
td->td_cpuset = nset;
sched_affinity(td);
thread_unlock(td);
}
unlock_out:
PROC_UNLOCK(p);
out:
while ((nset = LIST_FIRST(&droplist)) != NULL)
cpuset_rel_complete(nset);
while ((nset = LIST_FIRST(&freelist)) != NULL) {
LIST_REMOVE(nset, cs_link);
uma_zfree(cpuset_zone, nset);
}
return (error);
}
/*
* Apply an anonymous mask to a single thread.
*/
int
cpuset_setthread(lwpid_t id, cpuset_t *mask)
{
struct cpuset *nset;
struct cpuset *set;
struct thread *td;
struct proc *p;
int error;
nset = uma_zalloc(cpuset_zone, M_WAITOK);
error = cpuset_which(CPU_WHICH_TID, id, &p, &td, &set);
if (error)
goto out;
set = NULL;
thread_lock(td);
error = cpuset_shadow(td->td_cpuset, nset, mask);
if (error == 0) {
set = td->td_cpuset;
td->td_cpuset = nset;
sched_affinity(td);
nset = NULL;
}
thread_unlock(td);
PROC_UNLOCK(p);
if (set)
cpuset_rel(set);
out:
if (nset)
uma_zfree(cpuset_zone, nset);
return (error);
}
/*
* Creates the cpuset for thread0. We make two sets:
*
* 0 - The root set which should represent all valid processors in the
* system. It is initially created with a mask of all processors
* because we don't know what processors are valid until cpuset_init()
* runs. This set is immutable.
* 1 - The default set which all processes are a member of until changed.
* This allows an administrator to move all threads off of given cpus to
* dedicate them to high priority tasks or save power etc.
*/
struct cpuset *
cpuset_thread0(void)
{
struct cpuset *set;
int error;
cpuset_zone = uma_zcreate("cpuset", sizeof(struct cpuset), NULL, NULL,
NULL, NULL, UMA_ALIGN_PTR, 0);
mtx_init(&cpuset_lock, "cpuset", NULL, MTX_SPIN | MTX_RECURSE);
/*
* Create the root system set for the whole machine. Doesn't use
* cpuset_create() due to NULL parent.
*/
set = uma_zalloc(cpuset_zone, M_WAITOK | M_ZERO);
set->cs_mask.__bits[0] = -1;
LIST_INIT(&set->cs_children);
LIST_INSERT_HEAD(&cpuset_ids, set, cs_link);
set->cs_ref = 1;
set->cs_flags = CPU_SET_ROOT;
cpuset_zero = set;
cpuset_root = &set->cs_mask;
/*
* Now derive a default, modifiable set from that to give out.
*/
set = uma_zalloc(cpuset_zone, M_WAITOK);
error = _cpuset_create(set, cpuset_zero, &cpuset_zero->cs_mask, 1);
KASSERT(error == 0, ("Error creating default set: %d\n", error));
/*
* Initialize the unit allocator. 0 and 1 are allocated above.
*/
cpuset_unr = new_unrhdr(2, INT_MAX, NULL);
return (set);
}
/*
* Create a cpuset, which would be cpuset_create() but
* mark the new 'set' as root.
*
* We are not going to reparent the td to it. Use cpuset_reparentproc() for that.
*
* In case of no error, returns the set in *setp locked with a reference.
*/
int
cpuset_create_root(struct thread *td, struct cpuset **setp)
{
struct cpuset *root;
struct cpuset *set;
int error;
KASSERT(td != NULL, ("[%s:%d] invalid td", __func__, __LINE__));
KASSERT(setp != NULL, ("[%s:%d] invalid setp", __func__, __LINE__));
thread_lock(td);
root = cpuset_refroot(td->td_cpuset);
thread_unlock(td);
error = cpuset_create(setp, td->td_cpuset, &root->cs_mask);
cpuset_rel(root);
if (error)
return (error);
KASSERT(*setp != NULL, ("[%s:%d] cpuset_create returned invalid data",
__func__, __LINE__));
/* Mark the set as root. */
set = *setp;
set->cs_flags |= CPU_SET_ROOT;
return (0);
}
int
cpuset_setproc_update_set(struct proc *p, struct cpuset *set)
{
int error;
KASSERT(p != NULL, ("[%s:%d] invalid proc", __func__, __LINE__));
KASSERT(set != NULL, ("[%s:%d] invalid set", __func__, __LINE__));
cpuset_ref(set);
error = cpuset_setproc(p->p_pid, set, NULL);
if (error)
return (error);
cpuset_rel(set);
return (0);
}
/*
* This is called once the final set of system cpus is known. Modifies
* the root set and all children and mark the root readonly.
*/
static void
cpuset_init(void *arg)
{
cpuset_t mask;
CPU_ZERO(&mask);
#ifdef SMP
mask.__bits[0] = all_cpus;
#else
mask.__bits[0] = 1;
#endif
if (cpuset_modify(cpuset_zero, &mask))
panic("Can't set initial cpuset mask.\n");
cpuset_zero->cs_flags |= CPU_SET_RDONLY;
}
SYSINIT(cpuset, SI_SUB_SMP, SI_ORDER_ANY, cpuset_init, NULL);
#ifndef _SYS_SYSPROTO_H_
struct cpuset_args {
cpusetid_t *setid;
};
#endif
int
cpuset(struct thread *td, struct cpuset_args *uap)
{
struct cpuset *root;
struct cpuset *set;
int error;
thread_lock(td);
root = cpuset_refroot(td->td_cpuset);
thread_unlock(td);
error = cpuset_create(&set, root, &root->cs_mask);
cpuset_rel(root);
if (error)
return (error);
error = copyout(&set->cs_id, uap->setid, sizeof(set->cs_id));
if (error == 0)
error = cpuset_setproc(-1, set, NULL);
cpuset_rel(set);
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct cpuset_setid_args {
cpuwhich_t which;
id_t id;
cpusetid_t setid;
};
#endif
int
cpuset_setid(struct thread *td, struct cpuset_setid_args *uap)
{
struct cpuset *set;
int error;
/*
* Presently we only support per-process sets.
*/
if (uap->which != CPU_WHICH_PID)
return (EINVAL);
set = cpuset_lookup(uap->setid, td);
if (set == NULL)
return (ESRCH);
error = cpuset_setproc(uap->id, set, NULL);
cpuset_rel(set);
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct cpuset_getid_args {
cpulevel_t level;
cpuwhich_t which;
id_t id;
cpusetid_t *setid;
#endif
int
cpuset_getid(struct thread *td, struct cpuset_getid_args *uap)
{
struct cpuset *nset;
struct cpuset *set;
struct thread *ttd;
struct proc *p;
cpusetid_t id;
int error;
if (uap->level == CPU_LEVEL_WHICH && uap->which != CPU_WHICH_CPUSET)
return (EINVAL);
error = cpuset_which(uap->which, uap->id, &p, &ttd, &set);
if (error)
return (error);
switch (uap->which) {
case CPU_WHICH_TID:
case CPU_WHICH_PID:
thread_lock(ttd);
set = cpuset_refbase(ttd->td_cpuset);
thread_unlock(ttd);
PROC_UNLOCK(p);
break;
case CPU_WHICH_CPUSET:
case CPU_WHICH_JAIL:
break;
case CPU_WHICH_IRQ:
return (EINVAL);
}
switch (uap->level) {
case CPU_LEVEL_ROOT:
nset = cpuset_refroot(set);
cpuset_rel(set);
set = nset;
break;
case CPU_LEVEL_CPUSET:
break;
case CPU_LEVEL_WHICH:
break;
}
id = set->cs_id;
cpuset_rel(set);
if (error == 0)
error = copyout(&id, uap->setid, sizeof(id));
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct cpuset_getaffinity_args {
cpulevel_t level;
cpuwhich_t which;
id_t id;
size_t cpusetsize;
cpuset_t *mask;
};
#endif
int
cpuset_getaffinity(struct thread *td, struct cpuset_getaffinity_args *uap)
{
struct thread *ttd;
struct cpuset *nset;
struct cpuset *set;
struct proc *p;
cpuset_t *mask;
int error;
size_t size;
if (uap->cpusetsize < sizeof(cpuset_t) ||
uap->cpusetsize > CPU_MAXSIZE / NBBY)
return (ERANGE);
size = uap->cpusetsize;
mask = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
error = cpuset_which(uap->which, uap->id, &p, &ttd, &set);
if (error)
goto out;
switch (uap->level) {
case CPU_LEVEL_ROOT:
case CPU_LEVEL_CPUSET:
switch (uap->which) {
case CPU_WHICH_TID:
case CPU_WHICH_PID:
thread_lock(ttd);
set = cpuset_ref(ttd->td_cpuset);
thread_unlock(ttd);
break;
case CPU_WHICH_CPUSET:
case CPU_WHICH_JAIL:
break;
case CPU_WHICH_IRQ:
error = EINVAL;
goto out;
}
if (uap->level == CPU_LEVEL_ROOT)
nset = cpuset_refroot(set);
else
nset = cpuset_refbase(set);
CPU_COPY(&nset->cs_mask, mask);
cpuset_rel(nset);
break;
case CPU_LEVEL_WHICH:
switch (uap->which) {
case CPU_WHICH_TID:
thread_lock(ttd);
CPU_COPY(&ttd->td_cpuset->cs_mask, mask);
thread_unlock(ttd);
break;
case CPU_WHICH_PID:
FOREACH_THREAD_IN_PROC(p, ttd) {
thread_lock(ttd);
CPU_OR(mask, &ttd->td_cpuset->cs_mask);
thread_unlock(ttd);
}
break;
case CPU_WHICH_CPUSET:
case CPU_WHICH_JAIL:
CPU_COPY(&set->cs_mask, mask);
break;
case CPU_WHICH_IRQ:
error = intr_getaffinity(uap->id, mask);
break;
}
break;
default:
error = EINVAL;
break;
}
if (set)
cpuset_rel(set);
if (p)
PROC_UNLOCK(p);
if (error == 0)
error = copyout(mask, uap->mask, size);
out:
free(mask, M_TEMP);
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct cpuset_setaffinity_args {
cpulevel_t level;
cpuwhich_t which;
id_t id;
size_t cpusetsize;
const cpuset_t *mask;
};
#endif
int
cpuset_setaffinity(struct thread *td, struct cpuset_setaffinity_args *uap)
{
struct cpuset *nset;
struct cpuset *set;
struct thread *ttd;
struct proc *p;
cpuset_t *mask;
int error;
if (uap->cpusetsize < sizeof(cpuset_t) ||
uap->cpusetsize > CPU_MAXSIZE / NBBY)
return (ERANGE);
mask = malloc(uap->cpusetsize, M_TEMP, M_WAITOK | M_ZERO);
error = copyin(uap->mask, mask, uap->cpusetsize);
if (error)
goto out;
/*
* Verify that no high bits are set.
*/
if (uap->cpusetsize > sizeof(cpuset_t)) {
char *end;
char *cp;
end = cp = (char *)&mask->__bits;
end += uap->cpusetsize;
cp += sizeof(cpuset_t);
while (cp != end)
if (*cp++ != 0) {
error = EINVAL;
goto out;
}
}
switch (uap->level) {
case CPU_LEVEL_ROOT:
case CPU_LEVEL_CPUSET:
error = cpuset_which(uap->which, uap->id, &p, &ttd, &set);
if (error)
break;
switch (uap->which) {
case CPU_WHICH_TID:
case CPU_WHICH_PID:
thread_lock(ttd);
set = cpuset_ref(ttd->td_cpuset);
thread_unlock(ttd);
PROC_UNLOCK(p);
break;
case CPU_WHICH_CPUSET:
case CPU_WHICH_JAIL:
break;
case CPU_WHICH_IRQ:
error = EINVAL;
goto out;
}
if (uap->level == CPU_LEVEL_ROOT)
nset = cpuset_refroot(set);
else
nset = cpuset_refbase(set);
error = cpuset_modify(nset, mask);
cpuset_rel(nset);
cpuset_rel(set);
break;
case CPU_LEVEL_WHICH:
switch (uap->which) {
case CPU_WHICH_TID:
error = cpuset_setthread(uap->id, mask);
break;
case CPU_WHICH_PID:
error = cpuset_setproc(uap->id, NULL, mask);
break;
case CPU_WHICH_CPUSET:
case CPU_WHICH_JAIL:
error = cpuset_which(uap->which, uap->id, &p,
&ttd, &set);
if (error == 0) {
error = cpuset_modify(set, mask);
cpuset_rel(set);
}
break;
case CPU_WHICH_IRQ:
error = intr_setaffinity(uap->id, mask);
break;
default:
error = EINVAL;
break;
}
break;
default:
error = EINVAL;
break;
}
out:
free(mask, M_TEMP);
return (error);
}
#ifdef DDB
DB_SHOW_COMMAND(cpusets, db_show_cpusets)
{
struct cpuset *set;
int cpu, once;
LIST_FOREACH(set, &cpuset_ids, cs_link) {
db_printf("set=%p id=%-6u ref=%-6d flags=0x%04x parent id=%d\n",
set, set->cs_id, set->cs_ref, set->cs_flags,
(set->cs_parent != NULL) ? set->cs_parent->cs_id : 0);
db_printf(" mask=");
for (once = 0, cpu = 0; cpu < CPU_SETSIZE; cpu++) {
if (CPU_ISSET(cpu, &set->cs_mask)) {
if (once == 0) {
db_printf("%d", cpu);
once = 1;
} else
db_printf(",%d", cpu);
}
}
db_printf("\n");
if (db_pager_quit)
break;
}
}
#endif /* DDB */
|