about summary refs log tree commit diff
path: root/nptl/allocatestack.c
blob: 3016a2ed5c9f2722c5706b676a3e563385ea4658 (plain) (blame)
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
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
/* Copyright (C) 2002-2016 Free Software Foundation, Inc.
   This file is part of the GNU C Library.
   Contributed by Ulrich Drepper <drepper@redhat.com>, 2002.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library 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
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library; if not, see
   <http://www.gnu.org/licenses/>.  */

#include <assert.h>
#include <errno.h>
#include <signal.h>
#include <stdint.h>
#include <string.h>
#include <unistd.h>
#include <sys/mman.h>
#include <sys/param.h>
#include <dl-sysdep.h>
#include <dl-tls.h>
#include <tls.h>
#include <list.h>
#include <lowlevellock.h>
#include <futex-internal.h>
#include <kernel-features.h>
#include <stack-aliasing.h>


#ifndef NEED_SEPARATE_REGISTER_STACK

/* Most architectures have exactly one stack pointer.  Some have more.  */
# define STACK_VARIABLES void *stackaddr = NULL

/* How to pass the values to the 'create_thread' function.  */
# define STACK_VARIABLES_ARGS stackaddr

/* How to declare function which gets there parameters.  */
# define STACK_VARIABLES_PARMS void *stackaddr

/* How to declare allocate_stack.  */
# define ALLOCATE_STACK_PARMS void **stack

/* This is how the function is called.  We do it this way to allow
   other variants of the function to have more parameters.  */
# define ALLOCATE_STACK(attr, pd) allocate_stack (attr, pd, &stackaddr)

#else

/* We need two stacks.  The kernel will place them but we have to tell
   the kernel about the size of the reserved address space.  */
# define STACK_VARIABLES void *stackaddr = NULL; size_t stacksize = 0

/* How to pass the values to the 'create_thread' function.  */
# define STACK_VARIABLES_ARGS stackaddr, stacksize

/* How to declare function which gets there parameters.  */
# define STACK_VARIABLES_PARMS void *stackaddr, size_t stacksize

/* How to declare allocate_stack.  */
# define ALLOCATE_STACK_PARMS void **stack, size_t *stacksize

/* This is how the function is called.  We do it this way to allow
   other variants of the function to have more parameters.  */
# define ALLOCATE_STACK(attr, pd) \
  allocate_stack (attr, pd, &stackaddr, &stacksize)

#endif


/* Default alignment of stack.  */
#ifndef STACK_ALIGN
# define STACK_ALIGN __alignof__ (long double)
#endif

/* Default value for minimal stack size after allocating thread
   descriptor and guard.  */
#ifndef MINIMAL_REST_STACK
# define MINIMAL_REST_STACK	4096
#endif


/* Newer kernels have the MAP_STACK flag to indicate a mapping is used for
   a stack.  Use it when possible.  */
#ifndef MAP_STACK
# define MAP_STACK 0
#endif

/* This yields the pointer that TLS support code calls the thread pointer.  */
#if TLS_TCB_AT_TP
# define TLS_TPADJ(pd) (pd)
#elif TLS_DTV_AT_TP
# define TLS_TPADJ(pd) ((struct pthread *)((char *) (pd) + TLS_PRE_TCB_SIZE))
#endif

/* Cache handling for not-yet free stacks.  */

/* Maximum size in kB of cache.  */
static size_t stack_cache_maxsize = 40 * 1024 * 1024; /* 40MiBi by default.  */
static size_t stack_cache_actsize;

/* Mutex protecting this variable.  */
static int stack_cache_lock = LLL_LOCK_INITIALIZER;

/* List of queued stack frames.  */
static LIST_HEAD (stack_cache);

/* List of the stacks in use.  */
static LIST_HEAD (stack_used);

/* We need to record what list operations we are going to do so that,
   in case of an asynchronous interruption due to a fork() call, we
   can correct for the work.  */
static uintptr_t in_flight_stack;

/* List of the threads with user provided stacks in use.  No need to
   initialize this, since it's done in __pthread_initialize_minimal.  */
list_t __stack_user __attribute__ ((nocommon));
hidden_data_def (__stack_user)

#if COLORING_INCREMENT != 0
/* Number of threads created.  */
static unsigned int nptl_ncreated;
#endif


/* Check whether the stack is still used or not.  */
#define FREE_P(descr) ((descr)->tid <= 0)


static void
stack_list_del (list_t *elem)
{
  in_flight_stack = (uintptr_t) elem;

  atomic_write_barrier ();

  list_del (elem);

  atomic_write_barrier ();

  in_flight_stack = 0;
}


static void
stack_list_add (list_t *elem, list_t *list)
{
  in_flight_stack = (uintptr_t) elem | 1;

  atomic_write_barrier ();

  list_add (elem, list);

  atomic_write_barrier ();

  in_flight_stack = 0;
}


/* We create a double linked list of all cache entries.  Double linked
   because this allows removing entries from the end.  */


/* Get a stack frame from the cache.  We have to match by size since
   some blocks might be too small or far too large.  */
static struct pthread *
get_cached_stack (size_t *sizep, void **memp)
{
  size_t size = *sizep;
  struct pthread *result = NULL;
  list_t *entry;

  lll_lock (stack_cache_lock, LLL_PRIVATE);

  /* Search the cache for a matching entry.  We search for the
     smallest stack which has at least the required size.  Note that
     in normal situations the size of all allocated stacks is the
     same.  As the very least there are only a few different sizes.
     Therefore this loop will exit early most of the time with an
     exact match.  */
  list_for_each (entry, &stack_cache)
    {
      struct pthread *curr;

      curr = list_entry (entry, struct pthread, list);
      if (FREE_P (curr) && curr->stackblock_size >= size)
	{
	  if (curr->stackblock_size == size)
	    {
	      result = curr;
	      break;
	    }

	  if (result == NULL
	      || result->stackblock_size > curr->stackblock_size)
	    result = curr;
	}
    }

  if (__builtin_expect (result == NULL, 0)
      /* Make sure the size difference is not too excessive.  In that
	 case we do not use the block.  */
      || __builtin_expect (result->stackblock_size > 4 * size, 0))
    {
      /* Release the lock.  */
      lll_unlock (stack_cache_lock, LLL_PRIVATE);

      return NULL;
    }

  /* Don't allow setxid until cloned.  */
  result->setxid_futex = -1;

  /* Dequeue the entry.  */
  stack_list_del (&result->list);

  /* And add to the list of stacks in use.  */
  stack_list_add (&result->list, &stack_used);

  /* And decrease the cache size.  */
  stack_cache_actsize -= result->stackblock_size;

  /* Release the lock early.  */
  lll_unlock (stack_cache_lock, LLL_PRIVATE);

  /* Report size and location of the stack to the caller.  */
  *sizep = result->stackblock_size;
  *memp = result->stackblock;

  /* Cancellation handling is back to the default.  */
  result->cancelhandling = 0;
  result->cleanup = NULL;

  /* No pending event.  */
  result->nextevent = NULL;

  /* Clear the DTV.  */
  dtv_t *dtv = GET_DTV (TLS_TPADJ (result));
  for (size_t cnt = 0; cnt < dtv[-1].counter; ++cnt)
    free (dtv[1 + cnt].pointer.to_free);
  memset (dtv, '\0', (dtv[-1].counter + 1) * sizeof (dtv_t));

  /* Re-initialize the TLS.  */
  _dl_allocate_tls_init (TLS_TPADJ (result));

  return result;
}


/* Free stacks until cache size is lower than LIMIT.  */
void
__free_stacks (size_t limit)
{
  /* We reduce the size of the cache.  Remove the last entries until
     the size is below the limit.  */
  list_t *entry;
  list_t *prev;

  /* Search from the end of the list.  */
  list_for_each_prev_safe (entry, prev, &stack_cache)
    {
      struct pthread *curr;

      curr = list_entry (entry, struct pthread, list);
      if (FREE_P (curr))
	{
	  /* Unlink the block.  */
	  stack_list_del (entry);

	  /* Account for the freed memory.  */
	  stack_cache_actsize -= curr->stackblock_size;

	  /* Free the memory associated with the ELF TLS.  */
	  _dl_deallocate_tls (TLS_TPADJ (curr), false);

	  /* Remove this block.  This should never fail.  If it does
	     something is really wrong.  */
	  if (munmap (curr->stackblock, curr->stackblock_size) != 0)
	    abort ();

	  /* Maybe we have freed enough.  */
	  if (stack_cache_actsize <= limit)
	    break;
	}
    }
}


/* Add a stack frame which is not used anymore to the stack.  Must be
   called with the cache lock held.  */
static inline void
__attribute ((always_inline))
queue_stack (struct pthread *stack)
{
  /* We unconditionally add the stack to the list.  The memory may
     still be in use but it will not be reused until the kernel marks
     the stack as not used anymore.  */
  stack_list_add (&stack->list, &stack_cache);

  stack_cache_actsize += stack->stackblock_size;
  if (__glibc_unlikely (stack_cache_actsize > stack_cache_maxsize))
    __free_stacks (stack_cache_maxsize);
}


static int
internal_function
change_stack_perm (struct pthread *pd
#ifdef NEED_SEPARATE_REGISTER_STACK
		   , size_t pagemask
#endif
		   )
{
#ifdef NEED_SEPARATE_REGISTER_STACK
  void *stack = (pd->stackblock
		 + (((((pd->stackblock_size - pd->guardsize) / 2)
		      & pagemask) + pd->guardsize) & pagemask));
  size_t len = pd->stackblock + pd->stackblock_size - stack;
#elif _STACK_GROWS_DOWN
  void *stack = pd->stackblock + pd->guardsize;
  size_t len = pd->stackblock_size - pd->guardsize;
#elif _STACK_GROWS_UP
  void *stack = pd->stackblock;
  size_t len = (uintptr_t) pd - pd->guardsize - (uintptr_t) pd->stackblock;
#else
# error "Define either _STACK_GROWS_DOWN or _STACK_GROWS_UP"
#endif
  if (mprotect (stack, len, PROT_READ | PROT_WRITE | PROT_EXEC) != 0)
    return errno;

  return 0;
}


/* Returns a usable stack for a new thread either by allocating a
   new stack or reusing a cached stack of sufficient size.
   ATTR must be non-NULL and point to a valid pthread_attr.
   PDP must be non-NULL.  */
static int
allocate_stack (const struct pthread_attr *attr, struct pthread **pdp,
		ALLOCATE_STACK_PARMS)
{
  struct pthread *pd;
  size_t size;
  size_t pagesize_m1 = __getpagesize () - 1;

  assert (powerof2 (pagesize_m1 + 1));
  assert (TCB_ALIGNMENT >= STACK_ALIGN);

  /* Get the stack size from the attribute if it is set.  Otherwise we
     use the default we determined at start time.  */
  if (attr->stacksize != 0)
    size = attr->stacksize;
  else
    {
      lll_lock (__default_pthread_attr_lock, LLL_PRIVATE);
      size = __default_pthread_attr.stacksize;
      lll_unlock (__default_pthread_attr_lock, LLL_PRIVATE);
    }

  /* Get memory for the stack.  */
  if (__glibc_unlikely (attr->flags & ATTR_FLAG_STACKADDR))
    {
      uintptr_t adj;
      char *stackaddr = (char *) attr->stackaddr;

      /* Assume the same layout as the _STACK_GROWS_DOWN case, with struct
	 pthread at the top of the stack block.  Later we adjust the guard
	 location and stack address to match the _STACK_GROWS_UP case.  */
      if (_STACK_GROWS_UP)
	stackaddr += attr->stacksize;

      /* If the user also specified the size of the stack make sure it
	 is large enough.  */
      if (attr->stacksize != 0
	  && attr->stacksize < (__static_tls_size + MINIMAL_REST_STACK))
	return EINVAL;

      /* Adjust stack size for alignment of the TLS block.  */
#if TLS_TCB_AT_TP
      adj = ((uintptr_t) stackaddr - TLS_TCB_SIZE)
	    & __static_tls_align_m1;
      assert (size > adj + TLS_TCB_SIZE);
#elif TLS_DTV_AT_TP
      adj = ((uintptr_t) stackaddr - __static_tls_size)
	    & __static_tls_align_m1;
      assert (size > adj);
#endif

      /* The user provided some memory.  Let's hope it matches the
	 size...  We do not allocate guard pages if the user provided
	 the stack.  It is the user's responsibility to do this if it
	 is wanted.  */
#if TLS_TCB_AT_TP
      pd = (struct pthread *) ((uintptr_t) stackaddr
			       - TLS_TCB_SIZE - adj);
#elif TLS_DTV_AT_TP
      pd = (struct pthread *) (((uintptr_t) stackaddr
				- __static_tls_size - adj)
			       - TLS_PRE_TCB_SIZE);
#endif

      /* The user provided stack memory needs to be cleared.  */
      memset (pd, '\0', sizeof (struct pthread));

      /* The first TSD block is included in the TCB.  */
      pd->specific[0] = pd->specific_1stblock;

      /* Remember the stack-related values.  */
      pd->stackblock = (char *) stackaddr - size;
      pd->stackblock_size = size;

      /* This is a user-provided stack.  It will not be queued in the
	 stack cache nor will the memory (except the TLS memory) be freed.  */
      pd->user_stack = true;

      /* This is at least the second thread.  */
      pd->header.multiple_threads = 1;
#ifndef TLS_MULTIPLE_THREADS_IN_TCB
      __pthread_multiple_threads = *__libc_multiple_threads_ptr = 1;
#endif

#ifndef __ASSUME_PRIVATE_FUTEX
      /* The thread must know when private futexes are supported.  */
      pd->header.private_futex = THREAD_GETMEM (THREAD_SELF,
						header.private_futex);
#endif

#ifdef NEED_DL_SYSINFO
      SETUP_THREAD_SYSINFO (pd);
#endif

      /* The process ID is also the same as that of the caller.  */
      pd->pid = THREAD_GETMEM (THREAD_SELF, pid);

      /* Don't allow setxid until cloned.  */
      pd->setxid_futex = -1;

      /* Allocate the DTV for this thread.  */
      if (_dl_allocate_tls (TLS_TPADJ (pd)) == NULL)
	{
	  /* Something went wrong.  */
	  assert (errno == ENOMEM);
	  return errno;
	}


      /* Prepare to modify global data.  */
      lll_lock (stack_cache_lock, LLL_PRIVATE);

      /* And add to the list of stacks in use.  */
      list_add (&pd->list, &__stack_user);

      lll_unlock (stack_cache_lock, LLL_PRIVATE);
    }
  else
    {
      /* Allocate some anonymous memory.  If possible use the cache.  */
      size_t guardsize;
      size_t reqsize;
      void *mem;
      const int prot = (PROT_READ | PROT_WRITE
			| ((GL(dl_stack_flags) & PF_X) ? PROT_EXEC : 0));

#if COLORING_INCREMENT != 0
      /* Add one more page for stack coloring.  Don't do it for stacks
	 with 16 times pagesize or larger.  This might just cause
	 unnecessary misalignment.  */
      if (size <= 16 * pagesize_m1)
	size += pagesize_m1 + 1;
#endif

      /* Adjust the stack size for alignment.  */
      size &= ~__static_tls_align_m1;
      assert (size != 0);

      /* Make sure the size of the stack is enough for the guard and
	 eventually the thread descriptor.  */
      guardsize = (attr->guardsize + pagesize_m1) & ~pagesize_m1;
      if (__builtin_expect (size < ((guardsize + __static_tls_size
				     + MINIMAL_REST_STACK + pagesize_m1)
				    & ~pagesize_m1),
			    0))
	/* The stack is too small (or the guard too large).  */
	return EINVAL;

      /* Try to get a stack from the cache.  */
      reqsize = size;
      pd = get_cached_stack (&size, &mem);
      if (pd == NULL)
	{
	  /* To avoid aliasing effects on a larger scale than pages we
	     adjust the allocated stack size if necessary.  This way
	     allocations directly following each other will not have
	     aliasing problems.  */
#if MULTI_PAGE_ALIASING != 0
	  if ((size % MULTI_PAGE_ALIASING) == 0)
	    size += pagesize_m1 + 1;
#endif

	  mem = mmap (NULL, size, prot,
		      MAP_PRIVATE | MAP_ANONYMOUS | MAP_STACK, -1, 0);

	  if (__glibc_unlikely (mem == MAP_FAILED))
	    return errno;

	  /* SIZE is guaranteed to be greater than zero.
	     So we can never get a null pointer back from mmap.  */
	  assert (mem != NULL);

#if COLORING_INCREMENT != 0
	  /* Atomically increment NCREATED.  */
	  unsigned int ncreated = atomic_increment_val (&nptl_ncreated);

	  /* We chose the offset for coloring by incrementing it for
	     every new thread by a fixed amount.  The offset used
	     module the page size.  Even if coloring would be better
	     relative to higher alignment values it makes no sense to
	     do it since the mmap() interface does not allow us to
	     specify any alignment for the returned memory block.  */
	  size_t coloring = (ncreated * COLORING_INCREMENT) & pagesize_m1;

	  /* Make sure the coloring offsets does not disturb the alignment
	     of the TCB and static TLS block.  */
	  if (__glibc_unlikely ((coloring & __static_tls_align_m1) != 0))
	    coloring = (((coloring + __static_tls_align_m1)
			 & ~(__static_tls_align_m1))
			& ~pagesize_m1);
#else
	  /* Unless specified we do not make any adjustments.  */
# define coloring 0
#endif

	  /* Place the thread descriptor at the end of the stack.  */
#if TLS_TCB_AT_TP
	  pd = (struct pthread *) ((char *) mem + size - coloring) - 1;
#elif TLS_DTV_AT_TP
	  pd = (struct pthread *) ((((uintptr_t) mem + size - coloring
				    - __static_tls_size)
				    & ~__static_tls_align_m1)
				   - TLS_PRE_TCB_SIZE);
#endif

	  /* Remember the stack-related values.  */
	  pd->stackblock = mem;
	  pd->stackblock_size = size;

	  /* We allocated the first block thread-specific data array.
	     This address will not change for the lifetime of this
	     descriptor.  */
	  pd->specific[0] = pd->specific_1stblock;

	  /* This is at least the second thread.  */
	  pd->header.multiple_threads = 1;
#ifndef TLS_MULTIPLE_THREADS_IN_TCB
	  __pthread_multiple_threads = *__libc_multiple_threads_ptr = 1;
#endif

#ifndef __ASSUME_PRIVATE_FUTEX
	  /* The thread must know when private futexes are supported.  */
	  pd->header.private_futex = THREAD_GETMEM (THREAD_SELF,
						    header.private_futex);
#endif

#ifdef NEED_DL_SYSINFO
	  SETUP_THREAD_SYSINFO (pd);
#endif

	  /* Don't allow setxid until cloned.  */
	  pd->setxid_futex = -1;

	  /* The process ID is also the same as that of the caller.  */
	  pd->pid = THREAD_GETMEM (THREAD_SELF, pid);

	  /* Allocate the DTV for this thread.  */
	  if (_dl_allocate_tls (TLS_TPADJ (pd)) == NULL)
	    {
	      /* Something went wrong.  */
	      assert (errno == ENOMEM);

	      /* Free the stack memory we just allocated.  */
	      (void) munmap (mem, size);

	      return errno;
	    }


	  /* Prepare to modify global data.  */
	  lll_lock (stack_cache_lock, LLL_PRIVATE);

	  /* And add to the list of stacks in use.  */
	  stack_list_add (&pd->list, &stack_used);

	  lll_unlock (stack_cache_lock, LLL_PRIVATE);


	  /* There might have been a race.  Another thread might have
	     caused the stacks to get exec permission while this new
	     stack was prepared.  Detect if this was possible and
	     change the permission if necessary.  */
	  if (__builtin_expect ((GL(dl_stack_flags) & PF_X) != 0
				&& (prot & PROT_EXEC) == 0, 0))
	    {
	      int err = change_stack_perm (pd
#ifdef NEED_SEPARATE_REGISTER_STACK
					   , ~pagesize_m1
#endif
					   );
	      if (err != 0)
		{
		  /* Free the stack memory we just allocated.  */
		  (void) munmap (mem, size);

		  return err;
		}
	    }


	  /* Note that all of the stack and the thread descriptor is
	     zeroed.  This means we do not have to initialize fields
	     with initial value zero.  This is specifically true for
	     the 'tid' field which is always set back to zero once the
	     stack is not used anymore and for the 'guardsize' field
	     which will be read next.  */
	}

      /* Create or resize the guard area if necessary.  */
      if (__glibc_unlikely (guardsize > pd->guardsize))
	{
#ifdef NEED_SEPARATE_REGISTER_STACK
	  char *guard = mem + (((size - guardsize) / 2) & ~pagesize_m1);
#elif _STACK_GROWS_DOWN
	  char *guard = mem;
#elif _STACK_GROWS_UP
	  char *guard = (char *) (((uintptr_t) pd - guardsize) & ~pagesize_m1);
#endif
	  if (mprotect (guard, guardsize, PROT_NONE) != 0)
	    {
	    mprot_error:
	      lll_lock (stack_cache_lock, LLL_PRIVATE);

	      /* Remove the thread from the list.  */
	      stack_list_del (&pd->list);

	      lll_unlock (stack_cache_lock, LLL_PRIVATE);

	      /* Get rid of the TLS block we allocated.  */
	      _dl_deallocate_tls (TLS_TPADJ (pd), false);

	      /* Free the stack memory regardless of whether the size
		 of the cache is over the limit or not.  If this piece
		 of memory caused problems we better do not use it
		 anymore.  Uh, and we ignore possible errors.  There
		 is nothing we could do.  */
	      (void) munmap (mem, size);

	      return errno;
	    }

	  pd->guardsize = guardsize;
	}
      else if (__builtin_expect (pd->guardsize - guardsize > size - reqsize,
				 0))
	{
	  /* The old guard area is too large.  */

#ifdef NEED_SEPARATE_REGISTER_STACK
	  char *guard = mem + (((size - guardsize) / 2) & ~pagesize_m1);
	  char *oldguard = mem + (((size - pd->guardsize) / 2) & ~pagesize_m1);

	  if (oldguard < guard
	      && mprotect (oldguard, guard - oldguard, prot) != 0)
	    goto mprot_error;

	  if (mprotect (guard + guardsize,
			oldguard + pd->guardsize - guard - guardsize,
			prot) != 0)
	    goto mprot_error;
#elif _STACK_GROWS_DOWN
	  if (mprotect ((char *) mem + guardsize, pd->guardsize - guardsize,
			prot) != 0)
	    goto mprot_error;
#elif _STACK_GROWS_UP
	  if (mprotect ((char *) pd - pd->guardsize,
			pd->guardsize - guardsize, prot) != 0)
	    goto mprot_error;
#endif

	  pd->guardsize = guardsize;
	}
      /* The pthread_getattr_np() calls need to get passed the size
	 requested in the attribute, regardless of how large the
	 actually used guardsize is.  */
      pd->reported_guardsize = guardsize;
    }

  /* Initialize the lock.  We have to do this unconditionally since the
     stillborn thread could be canceled while the lock is taken.  */
  pd->lock = LLL_LOCK_INITIALIZER;

  /* The robust mutex lists also need to be initialized
     unconditionally because the cleanup for the previous stack owner
     might have happened in the kernel.  */
  pd->robust_head.futex_offset = (offsetof (pthread_mutex_t, __data.__lock)
				  - offsetof (pthread_mutex_t,
					      __data.__list.__next));
  pd->robust_head.list_op_pending = NULL;
#ifdef __PTHREAD_MUTEX_HAVE_PREV
  pd->robust_prev = &pd->robust_head;
#endif
  pd->robust_head.list = &pd->robust_head;

  /* We place the thread descriptor at the end of the stack.  */
  *pdp = pd;

#if _STACK_GROWS_DOWN
  void *stacktop;

# if TLS_TCB_AT_TP
  /* The stack begins before the TCB and the static TLS block.  */
  stacktop = ((char *) (pd + 1) - __static_tls_size);
# elif TLS_DTV_AT_TP
  stacktop = (char *) (pd - 1);
# endif

# ifdef NEED_SEPARATE_REGISTER_STACK
  *stack = pd->stackblock;
  *stacksize = stacktop - *stack;
# else
  *stack = stacktop;
# endif
#else
  *stack = pd->stackblock;
#endif

  return 0;
}


void
internal_function
__deallocate_stack (struct pthread *pd)
{
  lll_lock (stack_cache_lock, LLL_PRIVATE);

  /* Remove the thread from the list of threads with user defined
     stacks.  */
  stack_list_del (&pd->list);

  /* Not much to do.  Just free the mmap()ed memory.  Note that we do
     not reset the 'used' flag in the 'tid' field.  This is done by
     the kernel.  If no thread has been created yet this field is
     still zero.  */
  if (__glibc_likely (! pd->user_stack))
    (void) queue_stack (pd);
  else
    /* Free the memory associated with the ELF TLS.  */
    _dl_deallocate_tls (TLS_TPADJ (pd), false);

  lll_unlock (stack_cache_lock, LLL_PRIVATE);
}


int
internal_function
__make_stacks_executable (void **stack_endp)
{
  /* First the main thread's stack.  */
  int err = _dl_make_stack_executable (stack_endp);
  if (err != 0)
    return err;

#ifdef NEED_SEPARATE_REGISTER_STACK
  const size_t pagemask = ~(__getpagesize () - 1);
#endif

  lll_lock (stack_cache_lock, LLL_PRIVATE);

  list_t *runp;
  list_for_each (runp, &stack_used)
    {
      err = change_stack_perm (list_entry (runp, struct pthread, list)
#ifdef NEED_SEPARATE_REGISTER_STACK
			       , pagemask
#endif
			       );
      if (err != 0)
	break;
    }

  /* Also change the permission for the currently unused stacks.  This
     might be wasted time but better spend it here than adding a check
     in the fast path.  */
  if (err == 0)
    list_for_each (runp, &stack_cache)
      {
	err = change_stack_perm (list_entry (runp, struct pthread, list)
#ifdef NEED_SEPARATE_REGISTER_STACK
				 , pagemask
#endif
				 );
	if (err != 0)
	  break;
      }

  lll_unlock (stack_cache_lock, LLL_PRIVATE);

  return err;
}


/* In case of a fork() call the memory allocation in the child will be
   the same but only one thread is running.  All stacks except that of
   the one running thread are not used anymore.  We have to recycle
   them.  */
void
__reclaim_stacks (void)
{
  struct pthread *self = (struct pthread *) THREAD_SELF;

  /* No locking necessary.  The caller is the only stack in use.  But
     we have to be aware that we might have interrupted a list
     operation.  */

  if (in_flight_stack != 0)
    {
      bool add_p = in_flight_stack & 1;
      list_t *elem = (list_t *) (in_flight_stack & ~(uintptr_t) 1);

      if (add_p)
	{
	  /* We always add at the beginning of the list.  So in this case we
	     only need to check the beginning of these lists to see if the
	     pointers at the head of the list are inconsistent.  */
	  list_t *l = NULL;

	  if (stack_used.next->prev != &stack_used)
	    l = &stack_used;
	  else if (stack_cache.next->prev != &stack_cache)
	    l = &stack_cache;

	  if (l != NULL)
	    {
	      assert (l->next->prev == elem);
	      elem->next = l->next;
	      elem->prev = l;
	      l->next = elem;
	    }
	}
      else
	{
	  /* We can simply always replay the delete operation.  */
	  elem->next->prev = elem->prev;
	  elem->prev->next = elem->next;
	}
    }

  /* Mark all stacks except the still running one as free.  */
  list_t *runp;
  list_for_each (runp, &stack_used)
    {
      struct pthread *curp = list_entry (runp, struct pthread, list);
      if (curp != self)
	{
	  /* This marks the stack as free.  */
	  curp->tid = 0;

	  /* The PID field must be initialized for the new process.  */
	  curp->pid = self->pid;

	  /* Account for the size of the stack.  */
	  stack_cache_actsize += curp->stackblock_size;

	  if (curp->specific_used)
	    {
	      /* Clear the thread-specific data.  */
	      memset (curp->specific_1stblock, '\0',
		      sizeof (curp->specific_1stblock));

	      curp->specific_used = false;

	      for (size_t cnt = 1; cnt < PTHREAD_KEY_1STLEVEL_SIZE; ++cnt)
		if (curp->specific[cnt] != NULL)
		  {
		    memset (curp->specific[cnt], '\0',
			    sizeof (curp->specific_1stblock));

		    /* We have allocated the block which we do not
		       free here so re-set the bit.  */
		    curp->specific_used = true;
		  }
	    }
	}
    }

  /* Reset the PIDs in any cached stacks.  */
  list_for_each (runp, &stack_cache)
    {
      struct pthread *curp = list_entry (runp, struct pthread, list);
      curp->pid = self->pid;
    }

  /* Add the stack of all running threads to the cache.  */
  list_splice (&stack_used, &stack_cache);

  /* Remove the entry for the current thread to from the cache list
     and add it to the list of running threads.  Which of the two
     lists is decided by the user_stack flag.  */
  stack_list_del (&self->list);

  /* Re-initialize the lists for all the threads.  */
  INIT_LIST_HEAD (&stack_used);
  INIT_LIST_HEAD (&__stack_user);

  if (__glibc_unlikely (THREAD_GETMEM (self, user_stack)))
    list_add (&self->list, &__stack_user);
  else
    list_add (&self->list, &stack_used);

  /* There is one thread running.  */
  __nptl_nthreads = 1;

  in_flight_stack = 0;

  /* Initialize locks.  */
  stack_cache_lock = LLL_LOCK_INITIALIZER;
  __default_pthread_attr_lock = LLL_LOCK_INITIALIZER;
}


#if HP_TIMING_AVAIL
# undef __find_thread_by_id
/* Find a thread given the thread ID.  */
attribute_hidden
struct pthread *
__find_thread_by_id (pid_t tid)
{
  struct pthread *result = NULL;

  lll_lock (stack_cache_lock, LLL_PRIVATE);

  /* Iterate over the list with system-allocated threads first.  */
  list_t *runp;
  list_for_each (runp, &stack_used)
    {
      struct pthread *curp;

      curp = list_entry (runp, struct pthread, list);

      if (curp->tid == tid)
	{
	  result = curp;
	  goto out;
	}
    }

  /* Now the list with threads using user-allocated stacks.  */
  list_for_each (runp, &__stack_user)
    {
      struct pthread *curp;

      curp = list_entry (runp, struct pthread, list);

      if (curp->tid == tid)
	{
	  result = curp;
	  goto out;
	}
    }

 out:
  lll_unlock (stack_cache_lock, LLL_PRIVATE);

  return result;
}
#endif


#ifdef SIGSETXID
static void
internal_function
setxid_mark_thread (struct xid_command *cmdp, struct pthread *t)
{
  int ch;

  /* Wait until this thread is cloned.  */
  if (t->setxid_futex == -1
      && ! atomic_compare_and_exchange_bool_acq (&t->setxid_futex, -2, -1))
    do
      futex_wait_simple (&t->setxid_futex, -2, FUTEX_PRIVATE);
    while (t->setxid_futex == -2);

  /* Don't let the thread exit before the setxid handler runs.  */
  t->setxid_futex = 0;

  do
    {
      ch = t->cancelhandling;

      /* If the thread is exiting right now, ignore it.  */
      if ((ch & EXITING_BITMASK) != 0)
	{
	  /* Release the futex if there is no other setxid in
	     progress.  */
	  if ((ch & SETXID_BITMASK) == 0)
	    {
	      t->setxid_futex = 1;
	      futex_wake (&t->setxid_futex, 1, FUTEX_PRIVATE);
	    }
	  return;
	}
    }
  while (atomic_compare_and_exchange_bool_acq (&t->cancelhandling,
					       ch | SETXID_BITMASK, ch));
}


static void
internal_function
setxid_unmark_thread (struct xid_command *cmdp, struct pthread *t)
{
  int ch;

  do
    {
      ch = t->cancelhandling;
      if ((ch & SETXID_BITMASK) == 0)
	return;
    }
  while (atomic_compare_and_exchange_bool_acq (&t->cancelhandling,
					       ch & ~SETXID_BITMASK, ch));

  /* Release the futex just in case.  */
  t->setxid_futex = 1;
  futex_wake (&t->setxid_futex, 1, FUTEX_PRIVATE);
}


static int
internal_function
setxid_signal_thread (struct xid_command *cmdp, struct pthread *t)
{
  if ((t->cancelhandling & SETXID_BITMASK) == 0)
    return 0;

  int val;
  INTERNAL_SYSCALL_DECL (err);
  val = INTERNAL_SYSCALL (tgkill, err, 3, THREAD_GETMEM (THREAD_SELF, pid),
			  t->tid, SIGSETXID);

  /* If this failed, it must have had not started yet or else exited.  */
  if (!INTERNAL_SYSCALL_ERROR_P (val, err))
    {
      atomic_increment (&cmdp->cntr);
      return 1;
    }
  else
    return 0;
}

/* Check for consistency across set*id system call results.  The abort
   should not happen as long as all privileges changes happen through
   the glibc wrappers.  ERROR must be 0 (no error) or an errno
   code.  */
void
attribute_hidden
__nptl_setxid_error (struct xid_command *cmdp, int error)
{
  do
    {
      int olderror = cmdp->error;
      if (olderror == error)
	break;
      if (olderror != -1)
	/* Mismatch between current and previous results.  */
	abort ();
    }
  while (atomic_compare_and_exchange_bool_acq (&cmdp->error, error, -1));
}

int
attribute_hidden
__nptl_setxid (struct xid_command *cmdp)
{
  int signalled;
  int result;
  lll_lock (stack_cache_lock, LLL_PRIVATE);

  __xidcmd = cmdp;
  cmdp->cntr = 0;
  cmdp->error = -1;

  struct pthread *self = THREAD_SELF;

  /* Iterate over the list with system-allocated threads first.  */
  list_t *runp;
  list_for_each (runp, &stack_used)
    {
      struct pthread *t = list_entry (runp, struct pthread, list);
      if (t == self)
	continue;

      setxid_mark_thread (cmdp, t);
    }

  /* Now the list with threads using user-allocated stacks.  */
  list_for_each (runp, &__stack_user)
    {
      struct pthread *t = list_entry (runp, struct pthread, list);
      if (t == self)
	continue;

      setxid_mark_thread (cmdp, t);
    }

  /* Iterate until we don't succeed in signalling anyone.  That means
     we have gotten all running threads, and their children will be
     automatically correct once started.  */
  do
    {
      signalled = 0;

      list_for_each (runp, &stack_used)
	{
	  struct pthread *t = list_entry (runp, struct pthread, list);
	  if (t == self)
	    continue;

	  signalled += setxid_signal_thread (cmdp, t);
	}

      list_for_each (runp, &__stack_user)
	{
	  struct pthread *t = list_entry (runp, struct pthread, list);
	  if (t == self)
	    continue;

	  signalled += setxid_signal_thread (cmdp, t);
	}

      int cur = cmdp->cntr;
      while (cur != 0)
	{
	  futex_wait_simple ((unsigned int *) &cmdp->cntr, cur,
			     FUTEX_PRIVATE);
	  cur = cmdp->cntr;
	}
    }
  while (signalled != 0);

  /* Clean up flags, so that no thread blocks during exit waiting
     for a signal which will never come.  */
  list_for_each (runp, &stack_used)
    {
      struct pthread *t = list_entry (runp, struct pthread, list);
      if (t == self)
	continue;

      setxid_unmark_thread (cmdp, t);
    }

  list_for_each (runp, &__stack_user)
    {
      struct pthread *t = list_entry (runp, struct pthread, list);
      if (t == self)
	continue;

      setxid_unmark_thread (cmdp, t);
    }

  /* This must be last, otherwise the current thread might not have
     permissions to send SIGSETXID syscall to the other threads.  */
  INTERNAL_SYSCALL_DECL (err);
  result = INTERNAL_SYSCALL_NCS (cmdp->syscall_no, err, 3,
				 cmdp->id[0], cmdp->id[1], cmdp->id[2]);
  int error = 0;
  if (__glibc_unlikely (INTERNAL_SYSCALL_ERROR_P (result, err)))
    {
      error = INTERNAL_SYSCALL_ERRNO (result, err);
      __set_errno (error);
      result = -1;
    }
  __nptl_setxid_error (cmdp, error);

  lll_unlock (stack_cache_lock, LLL_PRIVATE);
  return result;
}
#endif  /* SIGSETXID.  */


static inline void __attribute__((always_inline))
init_one_static_tls (struct pthread *curp, struct link_map *map)
{
# if TLS_TCB_AT_TP
  void *dest = (char *) curp - map->l_tls_offset;
# elif TLS_DTV_AT_TP
  void *dest = (char *) curp + map->l_tls_offset + TLS_PRE_TCB_SIZE;
# else
#  error "Either TLS_TCB_AT_TP or TLS_DTV_AT_TP must be defined"
# endif

  /* Fill in the DTV slot so that a later LD/GD access will find it.  */
  dtv_t *dtv = GET_DTV (TLS_TPADJ (curp));
  dtv[map->l_tls_modid].pointer.to_free = NULL;
  dtv[map->l_tls_modid].pointer.val = dest;

  /* Initialize the memory.  */
  memset (__mempcpy (dest, map->l_tls_initimage, map->l_tls_initimage_size),
	  '\0', map->l_tls_blocksize - map->l_tls_initimage_size);
}

void
attribute_hidden
__pthread_init_static_tls (struct link_map *map)
{
  lll_lock (stack_cache_lock, LLL_PRIVATE);

  /* Iterate over the list with system-allocated threads first.  */
  list_t *runp;
  list_for_each (runp, &stack_used)
    init_one_static_tls (list_entry (runp, struct pthread, list), map);

  /* Now the list with threads using user-allocated stacks.  */
  list_for_each (runp, &__stack_user)
    init_one_static_tls (list_entry (runp, struct pthread, list), map);

  lll_unlock (stack_cache_lock, LLL_PRIVATE);
}


void
attribute_hidden
__wait_lookup_done (void)
{
  lll_lock (stack_cache_lock, LLL_PRIVATE);

  struct pthread *self = THREAD_SELF;

  /* Iterate over the list with system-allocated threads first.  */
  list_t *runp;
  list_for_each (runp, &stack_used)
    {
      struct pthread *t = list_entry (runp, struct pthread, list);
      if (t == self || t->header.gscope_flag == THREAD_GSCOPE_FLAG_UNUSED)
	continue;

      int *const gscope_flagp = &t->header.gscope_flag;

      /* We have to wait until this thread is done with the global
	 scope.  First tell the thread that we are waiting and
	 possibly have to be woken.  */
      if (atomic_compare_and_exchange_bool_acq (gscope_flagp,
						THREAD_GSCOPE_FLAG_WAIT,
						THREAD_GSCOPE_FLAG_USED))
	continue;

      do
	futex_wait_simple ((unsigned int *) gscope_flagp,
			   THREAD_GSCOPE_FLAG_WAIT, FUTEX_PRIVATE);
      while (*gscope_flagp == THREAD_GSCOPE_FLAG_WAIT);
    }

  /* Now the list with threads using user-allocated stacks.  */
  list_for_each (runp, &__stack_user)
    {
      struct pthread *t = list_entry (runp, struct pthread, list);
      if (t == self || t->header.gscope_flag == THREAD_GSCOPE_FLAG_UNUSED)
	continue;

      int *const gscope_flagp = &t->header.gscope_flag;

      /* We have to wait until this thread is done with the global
	 scope.  First tell the thread that we are waiting and
	 possibly have to be woken.  */
      if (atomic_compare_and_exchange_bool_acq (gscope_flagp,
						THREAD_GSCOPE_FLAG_WAIT,
						THREAD_GSCOPE_FLAG_USED))
	continue;

      do
	futex_wait_simple ((unsigned int *) gscope_flagp,
			   THREAD_GSCOPE_FLAG_WAIT, FUTEX_PRIVATE);
      while (*gscope_flagp == THREAD_GSCOPE_FLAG_WAIT);
    }

  lll_unlock (stack_cache_lock, LLL_PRIVATE);
}