aboutsummaryrefslogtreecommitdiffstats
path: root/src/svm/svm.c
blob: 2834524b934b21831ce48861acdf7a559eaa2a45 (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
ef='#n795'>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 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330
/*
 *------------------------------------------------------------------
 * svm.c - shared VM allocation, mmap(...MAP_FIXED...)
 * library
 *
 * Copyright (c) 2009 Cisco and/or its affiliates.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *------------------------------------------------------------------
 */

#include <stdio.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <netinet/in.h>
#include <signal.h>
#include <pthread.h>
#include <unistd.h>
#include <time.h>
#include <fcntl.h>
#include <string.h>
#include <vppinfra/clib.h>
#include <vppinfra/vec.h>
#include <vppinfra/hash.h>
#include <vppinfra/bitmap.h>
#include <vppinfra/fifo.h>
#include <vppinfra/time.h>
#include <vppinfra/heap.h>
#include <vppinfra/pool.h>
#include <vppinfra/format.h>

#include "svm.h"

static svm_region_t *root_rp;
static int root_rp_refcount;

#define MAXLOCK 2
static pthread_mutex_t *mutexes_held[MAXLOCK];
static int nheld;

svm_region_t *
svm_get_root_rp (void)
{
  return root_rp;
}

#define MUTEX_DEBUG

u64
svm_get_global_region_base_va ()
{
#ifdef CLIB_SANITIZE_ADDR
  return 0x200000000000;
#endif

#if __aarch64__
  /* On AArch64 VA space can have different size, from 36 to 48 bits.
     Here we are trying to detect VA bits by parsing /proc/self/maps
     address ranges */
  int fd;
  unformat_input_t input;
  u64 start, end = 0;
  u8 bits = 0;

  if ((fd = open ("/proc/self/maps", 0)) < 0)
    clib_unix_error ("open '/proc/self/maps'");

  unformat_init_clib_file (&input, fd);
  while (unformat_check_input (&input) != UNFORMAT_END_OF_INPUT)
    {
      if (unformat (&input, "%llx-%llx", &start, &end))
	end--;
      unformat_skip_line (&input);
    }
  unformat_free (&input);
  close (fd);

  bits = count_leading_zeros (end);
  bits = 64 - bits;
  if (bits >= 36 && bits <= 48)
    return ((1ul << bits) / 4) - (2 * SVM_GLOBAL_REGION_SIZE);
  else
    clib_unix_error ("unexpected va bits '%u'", bits);
#endif

  /* default value */
  return 0x130000000ULL;
}

static void
region_lock (svm_region_t * rp, int tag)
{
  pthread_mutex_lock (&rp->mutex);
#ifdef MUTEX_DEBUG
  rp->mutex_owner_pid = getpid ();
  rp->mutex_owner_tag = tag;
#endif
  ASSERT (nheld < MAXLOCK);	//NOSONAR
  /*
   * Keep score of held mutexes so we can try to exit
   * cleanly if the world comes to an end at the worst possible
   * moment
   */
  mutexes_held[nheld++] = &rp->mutex;
}

static void
region_unlock (svm_region_t * rp)
{
  int i, j;
#ifdef MUTEX_DEBUG
  rp->mutex_owner_pid = 0;
  rp->mutex_owner_tag = 0;
#endif

  for (i = nheld - 1; i >= 0; i--)
    {
      if (mutexes_held[i] == &rp->mutex)
	{
	  for (j = i; j < MAXLOCK - 1; j++)
	    mutexes_held[j] = mutexes_held[j + 1];
	  nheld--;
	  goto found;
	}
    }
  ASSERT (0);

found:
  CLIB_MEMORY_BARRIER ();
  pthread_mutex_unlock (&rp->mutex);
}


static u8 *
format_svm_flags (u8 * s, va_list * args)
{
  uword f = va_arg (*args, uword);

  if (f & SVM_FLAGS_MHEAP)
    s = format (s, "MHEAP ");
  if (f & SVM_FLAGS_FILE)
    s = format (s, "FILE ");
  if (f & SVM_FLAGS_NODATA)
    s = format (s, "NODATA ");
  if (f & SVM_FLAGS_NEED_DATA_INIT)
    s = format (s, "INIT ");

  return (s);
}

static u8 *
format_svm_size (u8 * s, va_list * args)
{
  uword size = va_arg (*args, uword);

  if (size >= (1 << 20))
    {
      s = format (s, "(%d mb)", size >> 20);
    }
  else if (size >= (1 << 10))
    {
      s = format (s, "(%d kb)", size >> 10);
    }
  else
    {
      s = format (s, "(%d bytes)", size);
    }
  return (s);
}

u8 *
format_svm_region (u8 * s, va_list * args)
{
  svm_region_t *rp = va_arg (*args, svm_region_t *);
  int verbose = va_arg (*args, int);
  int i;
  uword lo, hi;

  s = format (s, "%s: base va 0x%x size 0x%x %U\n",
	      rp->region_name, rp->virtual_base,
	      rp->virtual_size, format_svm_size, rp->virtual_size);
  s = format (s, "  user_ctx 0x%x, bitmap_size %d\n",
	      rp->user_ctx, rp->bitmap_size);

  if (verbose)
    {
      s = format (s, "  flags: 0x%x %U\n", rp->flags,
		  format_svm_flags, rp->flags);
      s = format (s,
		  "  region_heap 0x%x data_base 0x%x data_heap 0x%x\n",
		  rp->region_heap, rp->data_base, rp->data_heap);
    }

  s = format (s, "  %d clients, pids: ", vec_len (rp->client_pids));

  for (i = 0; i < vec_len (rp->client_pids); i++)
    s = format (s, "%d ", rp->client_pids[i]);

  s = format (s, "\n");

  if (verbose)
    {
      lo = hi = ~0;

      s = format (s, "  VM in use: ");

      for (i = 0; i < rp->bitmap_size; i++)
	{
	  if (clib_bitmap_get_no_check (rp->bitmap, i) != 0)
	    {
	      if (lo == ~0)
		{
		  hi = lo = rp->virtual_base + i * MMAP_PAGESIZE;
		}
	      else
		{
		  hi = rp->virtual_base + i * MMAP_PAGESIZE;
		}
	    }
	  else
	    {
	      if (lo != ~0)
		{
		  hi = rp->virtual_base + i * MMAP_PAGESIZE - 1;
		  s = format (s, "   0x%x - 0x%x (%dk)\n", lo, hi,
			      (hi - lo) >> 10);
		  lo = hi = ~0;
		}
	    }
	}
    }

  return (s);
}

/*
 * rnd_pagesize
 * Round to a pagesize multiple, presumably 4k works
 */
static u64
rnd_pagesize (u64 size)
{
  u64 rv;

  rv = (size + (MMAP_PAGESIZE - 1)) & ~(MMAP_PAGESIZE - 1);
  return (rv);
}

/*
 * svm_data_region_setup
 */
static int
svm_data_region_create (svm_map_region_args_t * a, svm_region_t * rp)
{
  int fd;
  u8 junk = 0;
  uword map_size;

  map_size = rp->virtual_size - (MMAP_PAGESIZE +
				 (a->pvt_heap_size ? a->pvt_heap_size :
				  SVM_PVT_MHEAP_SIZE));

  if (a->flags & SVM_FLAGS_FILE)
    {
      struct stat statb;

      fd = open (a->backing_file, O_RDWR | O_CREAT, 0777);

      if (fd < 0)
	{
	  clib_unix_warning ("open");
	  return -1;
	}

      if (fstat (fd, &statb) < 0)
	{
	  clib_unix_warning ("fstat");
	  close (fd);
	  return -2;
	}

      if (statb.st_mode & S_IFREG)
	{
	  if (statb.st_size == 0)
	    {
	      if (lseek (fd, map_size, SEEK_SET) == (off_t) - 1)
		{
		  clib_unix_warning ("seek region size");
		  close (fd);
		  return -3;
		}
	      if (write (fd, &junk, 1) != 1)
		{
		  clib_unix_warning ("set region size");
		  close (fd);
		  return -3;
		}
	    }
	  else
	    {
	      map_size = rnd_pagesize (statb.st_size);
	    }
	}
      else
	{
	  map_size = a->backing_mmap_size;
	}

      ASSERT (map_size <= rp->virtual_size -
	      (MMAP_PAGESIZE + SVM_PVT_MHEAP_SIZE));

      if (mmap (rp->data_base, map_size, PROT_READ | PROT_WRITE,
		MAP_SHARED | MAP_FIXED, fd, 0) == MAP_FAILED)
	{
	  clib_unix_warning ("mmap");
	  close (fd);
	  return -3;
	}
      close (fd);
      CLIB_MEM_UNPOISON (rp->data_base, map_size);
      rp->backing_file = (char *) format (0, "%s%c", a->backing_file, 0);
      rp->flags |= SVM_FLAGS_FILE;
    }

  if (a->flags & SVM_FLAGS_MHEAP)
    {
      rp->data_heap = clib_mem_create_heap (rp->data_base, map_size,
					    1 /* locked */ , "svm data");

      rp->flags |= SVM_FLAGS_MHEAP;
    }
  return 0;
}

static int
svm_data_region_map (svm_map_region_args_t * a, svm_region_t * rp)
{
  int fd;
  u8 junk = 0;
  uword map_size;
  struct stat statb;

  map_size = rp->virtual_size -
    (MMAP_PAGESIZE
     + (a->pvt_heap_size ? a->pvt_heap_size : SVM_PVT_MHEAP_SIZE));

  if (a->flags & SVM_FLAGS_FILE)
    {

      fd = open (a->backing_file, O_RDWR, 0777);

      if (fd < 0)
	{
	  clib_unix_warning ("open");
	  return -1;
	}

      if (fstat (fd, &statb) < 0)
	{
	  clib_unix_warning ("fstat");
	  close (fd);
	  return -2;
	}

      if (statb.st_mode & S_IFREG)
	{
	  if (statb.st_size == 0)
	    {
	      if (lseek (fd, map_size, SEEK_SET) == (off_t) - 1)
		{
		  clib_unix_warning ("seek region size");
		  close (fd);
		  return -3;
		}
	      if (write (fd, &junk, 1) != 1)
		{
		  clib_unix_warning ("set region size");
		  close (fd);
		  return -3;
		}
	    }
	  else
	    {
	      map_size = rnd_pagesize (statb.st_size);
	    }
	}
      else
	{
	  map_size = a->backing_mmap_size;
	}

      ASSERT (map_size <= rp->virtual_size
	      - (MMAP_PAGESIZE
		 +
		 (a->pvt_heap_size ? a->pvt_heap_size : SVM_PVT_MHEAP_SIZE)));

      if (mmap (rp->data_base, map_size, PROT_READ | PROT_WRITE,
		MAP_SHARED | MAP_FIXED, fd, 0) == MAP_FAILED)
	{
	  clib_unix_warning ("mmap");
	  close (fd);
	  return -3;
	}
      close (fd);
      CLIB_MEM_UNPOISON (rp->data_base, map_size);
    }
  return 0;
}

u8 *
shm_name_from_svm_map_region_args (svm_map_region_args_t * a)
{
  u8 *shm_name;
  int root_path_offset = 0;
  int name_offset = 0;

  if (a->root_path)
    {
      /* Tolerate present or absent slashes */
      if (a->root_path[0] == '/')
	root_path_offset++;

      if (a->name[0] == '/')
	name_offset = 1;

      shm_name = format (0, "/%s-%s%c", &a->root_path[root_path_offset],
			 &a->name[name_offset], 0);
    }
  else
    shm_name = format (0, "%s%c", a->name, 0);
  return (shm_name);
}

void
svm_region_init_mapped_region (svm_map_region_args_t * a, svm_region_t * rp)
{
  pthread_mutexattr_t attr;
  pthread_condattr_t cattr;
  int nbits, words, bit;
  int overhead_space;
  void *oldheap;
  uword data_base;
  ASSERT (rp);
  int rv;

  clib_memset (rp, 0, sizeof (*rp));

  if (pthread_mutexattr_init (&attr))
    clib_unix_warning ("mutexattr_init");

  if (pthread_mutexattr_setpshared (&attr, PTHREAD_PROCESS_SHARED))
    clib_unix_warning ("mutexattr_setpshared");

  if (pthread_mutex_init (&rp->mutex, &attr))
    clib_unix_warning ("mutex_init");

  if (pthread_mutexattr_destroy (&attr))
    clib_unix_warning ("mutexattr_destroy");

  if (pthread_condattr_init (&cattr))
    clib_unix_warning ("condattr_init");

  if (pthread_condattr_setpshared (&cattr, PTHREAD_PROCESS_SHARED))
    clib_unix_warning ("condattr_setpshared");

  if (pthread_cond_init (&rp->condvar, &cattr))
    clib_unix_warning ("cond_init");

  if (pthread_condattr_destroy (&cattr))
    clib_unix_warning ("condattr_destroy");

  region_lock (rp, 1);

  rp->virtual_base = a->baseva;
  rp->virtual_size = a->size;

  rp->region_heap = clib_mem_create_heap
    (uword_to_pointer (a->baseva + MMAP_PAGESIZE, void *),
     (a->pvt_heap_size !=
      0) ? a->pvt_heap_size : SVM_PVT_MHEAP_SIZE, 1 /* locked */ ,
     "svm region");

  oldheap = svm_push_pvt_heap (rp);

  rp->region_name = (char *) format (0, "%s%c", a->name, 0);
  vec_add1 (rp->client_pids, getpid ());

  nbits = rp->virtual_size / MMAP_PAGESIZE;

  ASSERT (nbits > 0);
  rp->bitmap_size = nbits;
  words = (nbits + BITS (uword) - 1) / BITS (uword);
  vec_validate (rp->bitmap, words - 1);

  overhead_space = MMAP_PAGESIZE /* header */  +
    ((a->pvt_heap_size != 0) ? a->pvt_heap_size : SVM_PVT_MHEAP_SIZE);

  bit = 0;
  data_base = (uword) rp->virtual_base;

  if (a->flags & SVM_FLAGS_NODATA)
    rp->flags |= SVM_FLAGS_NEED_DATA_INIT;

  do
    {
      clib_bitmap_set_no_check (rp->bitmap, bit, 1);
      bit++;
      overhead_space -= MMAP_PAGESIZE;
      data_base += MMAP_PAGESIZE;
    }
  while (overhead_space > 0);

  rp->data_base = (void *) data_base;

  /*
   * Note: although the POSIX spec guarantees that only one
   * process enters this block, we have to play games
   * to hold off clients until e.g. the mutex is ready
   */
  rp->version = SVM_VERSION;

  /* setup the data portion of the region */

  rv = svm_data_region_create (a, rp);
  if (rv)
    {
      clib_warning ("data_region_create: %d", rv);
    }

  region_unlock (rp);

  svm_pop_heap (oldheap);
}

/*
 * svm_map_region
 */
void *
svm_map_region (svm_map_region_args_t * a)
{
  int svm_fd;
  svm_region_t *rp;
  int deadman = 0;
  u8 junk = 0;
  void *oldheap;
  int rv;
  int pid_holding_region_lock;
  u8 *shm_name;
  int dead_region_recovery = 0;
  int time_left;
  struct stat stat;
  struct timespec ts, tsrem;

  ASSERT ((a->size & ~(MMAP_PAGESIZE - 1)) == a->size);
  ASSERT (a->name);

  shm_name = shm_name_from_svm_map_region_args (a);

  if (CLIB_DEBUG > 1)
    clib_warning ("[%d] map region %s: shm_open (%s)",
		  getpid (), a->name, shm_name);

  svm_fd = shm_open ((char *) shm_name, O_RDWR | O_CREAT | O_EXCL, 0777);

  if (svm_fd >= 0)
    {
      if (fchmod (svm_fd, S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP) < 0)
	clib_unix_warning ("segment chmod");
      /* This turns out to fail harmlessly if the client starts first */
      if (fchown (svm_fd, a->uid, a->gid) < 0)
	clib_unix_warning ("segment chown [ok if client starts first]");

      vec_free (shm_name);

      if (lseek (svm_fd, a->size, SEEK_SET) == (off_t) - 1)
	{
	  clib_warning ("seek region size");
	  close (svm_fd);
	  return (0);
	}
      if (write (svm_fd, &junk, 1) != 1)
	{
	  clib_warning ("set region size");
	  close (svm_fd);
	  return (0);
	}

      rp = mmap (uword_to_pointer (a->baseva, void *), a->size,
		 PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, svm_fd, 0);

      if (rp == (svm_region_t *) MAP_FAILED)
	{
	  clib_unix_warning ("mmap create");
	  close (svm_fd);
	  return (0);
	}
      close (svm_fd);
      CLIB_MEM_UNPOISON (rp, a->size);

      svm_region_init_mapped_region (a, rp);

      return ((void *) rp);
    }
  else
    {
      svm_fd = shm_open ((char *) shm_name, O_RDWR, 0777);

      vec_free (shm_name);

      if (svm_fd < 0)
	{
	  perror ("svm_region_map(mmap open)");
	  return (0);
	}

      /* Reset ownership in case the client started first */
      if (fchown (svm_fd, a->uid, a->gid) < 0)
	clib_unix_warning ("segment chown [ok if client starts first]");

      time_left = 20;
      while (1)
	{
	  if (0 != fstat (svm_fd, &stat))
	    {
	      clib_warning ("fstat failed: %d", errno);
	      close (svm_fd);
	      return (0);
	    }
	  if (stat.st_size > 0)
	    {
	      break;
	    }
	  if (0 == time_left)
	    {
	      clib_warning ("waiting for resize of shm file timed out");
	      close (svm_fd);
	      return (0);
	    }
	  ts.tv_sec = 0;
	  ts.tv_nsec = 100000000;
	  while (nanosleep (&ts, &tsrem) < 0)
	    ts = tsrem;
	  time_left--;
	}

      rp = mmap (0, MMAP_PAGESIZE,
		 PROT_READ | PROT_WRITE, MAP_SHARED, svm_fd, 0);

      if (rp == (svm_region_t *) MAP_FAILED)
	{
	  close (svm_fd);
	  clib_warning ("mmap");
	  return (0);
	}

      CLIB_MEM_UNPOISON (rp, MMAP_PAGESIZE);

      /*
       * We lost the footrace to create this region; make sure
       * the winner has crossed the finish line.
       */
      while (rp->version == 0 && deadman++ < 5)
	{
	  sleep (1);
	}

      /*
       * <bleep>-ed?
       */
      if (rp->version == 0)
	{
	  clib_warning ("rp->version %d not %d", rp->version, SVM_VERSION);
	  close (svm_fd);
	  munmap (rp, a->size);
	  return (0);
	}
      /* Remap now that the region has been placed */
      a->baseva = rp->virtual_base;
      a->size = rp->virtual_size;
      munmap (rp, MMAP_PAGESIZE);

      rp = (void *) mmap (uword_to_pointer (a->baseva, void *), a->size,
			  PROT_READ | PROT_WRITE,
			  MAP_SHARED | MAP_FIXED, svm_fd, 0);
      if ((uword) rp == (uword) MAP_FAILED)
	{
	  clib_unix_warning ("mmap");
	  close (svm_fd);
	  return (0);
	}

      close (svm_fd);

      CLIB_MEM_UNPOISON (rp, a->size);

      if ((uword) rp != rp->virtual_base)
	{
	  clib_warning ("mmap botch");
	}

      /*
       * Try to fix the region mutex if it is held by
       * a dead process
       */
      pid_holding_region_lock = rp->mutex_owner_pid;
      if (pid_holding_region_lock && kill (pid_holding_region_lock, 0) < 0)
	{
	  clib_warning
	    ("region %s mutex held by dead pid %d, tag %d, force unlock",
	     rp->region_name, pid_holding_region_lock, rp->mutex_owner_tag);
	  /* owner pid is nonexistent */
	  rp->mutex.__data.__owner = 0;
	  rp->mutex.__data.__lock = 0;
	  dead_region_recovery = 1;
	}

      if (dead_region_recovery)
	clib_warning ("recovery: attempt to re-lock region");

      region_lock (rp, 2);
      oldheap = svm_push_pvt_heap (rp);
      vec_add1 (rp->client_pids, getpid ());

      if (dead_region_recovery)
	clib_warning ("recovery: attempt svm_data_region_map");

      rv = svm_data_region_map (a, rp);
      if (rv)
	{
	  clib_warning ("data_region_map: %d", rv);
	}

      if (dead_region_recovery)
	clib_warning ("unlock and continue");

      region_unlock (rp);

      svm_pop_heap (oldheap);

      return ((void *) rp);

    }
  return 0;			/* NOTREACHED *///NOSONAR
}

static void
svm_mutex_cleanup (void)
{
  int i;
  for (i = 0; i < nheld; i++)
    {
      pthread_mutex_unlock (mutexes_held[i]);	//NOSONAR
    }
}

static int
svm_region_init_internal (svm_map_region_args_t * a)
{
  svm_region_t *rp;
  u64 ticks = clib_cpu_time_now ();
  uword randomize_baseva;

  /* guard against klutz calls */
  if (root_rp)
    return -1;

  root_rp_refcount++;

  atexit (svm_mutex_cleanup);

  /* Randomize the shared-VM base at init time */
  if (MMAP_PAGESIZE <= (4 << 10))
    randomize_baseva = (ticks & 15) * MMAP_PAGESIZE;
  else
    randomize_baseva = (ticks & 3) * MMAP_PAGESIZE;

  a->baseva += randomize_baseva;

  rp = svm_map_region (a);
  if (!rp)
    return -1;

  region_lock (rp, 3);

  /* Set up the main region data structures */
  if (rp->flags & SVM_FLAGS_NEED_DATA_INIT)
    {
      svm_main_region_t *mp = 0;
      void *oldheap;

      rp->flags &= ~(SVM_FLAGS_NEED_DATA_INIT);

      oldheap = svm_push_pvt_heap (rp);
      vec_validate (mp, 0);
      mp->name_hash = hash_create_string (0, sizeof (uword));
      mp->root_path = a->root_path ? format (0, "%s%c", a->root_path, 0) : 0;
      mp->uid = a->uid;
      mp->gid = a->gid;
      rp->data_base = mp;
      svm_pop_heap (oldheap);
    }
  region_unlock (rp);
  root_rp = rp;

  return 0;
}

void
svm_region_init (void)
{
  svm_map_region_args_t _a, *a = &_a;

  clib_memset (a, 0, sizeof (*a));
  a->root_path = 0;
  a->name = SVM_GLOBAL_REGION_NAME;
  a->baseva = svm_get_global_region_base_va ();
  a->size = SVM_GLOBAL_REGION_SIZE;
  a->flags = SVM_FLAGS_NODATA;
  a->uid = 0;
  a->gid = 0;

  svm_region_init_internal (a);
}

int
svm_region_init_chroot (const char *root_path)
{
  svm_map_region_args_t _a, *a = &_a;

  clib_memset (a, 0, sizeof (*a));
  a->root_path = root_path;
  a->name = SVM_GLOBAL_REGION_NAME;
  a->baseva = svm_get_global_region_base_va ();
  a->size = SVM_GLOBAL_REGION_SIZE;
  a->flags = SVM_FLAGS_NODATA;
  a->uid = 0;
  a->gid = 0;

  return svm_region_init_internal (a);
}

void
svm_region_init_chroot_uid_gid (const char *root_path, int uid, int gid)
{
  svm_map_region_args_t _a, *a = &_a;

  clib_memset (a, 0, sizeof (*a));
  a->root_path = root_path;
  a->name = SVM_GLOBAL_REGION_NAME;
  a->baseva = svm_get_global_region_base_va ();
  a->size = SVM_GLOBAL_REGION_SIZE;
  a->flags = SVM_FLAGS_NODATA;
  a->uid = uid;
  a->gid = gid;

  svm_region_init_internal (a);
}

void
svm_region_init_args (svm_map_region_args_t * a)
{
  svm_region_init_internal (a);
}

void *
svm_region_find_or_create (svm_map_region_args_t * a)
{
  svm_main_region_t *mp;
  svm_region_t *rp;
  uword need_nbits;
  int index, i;
  void *oldheap;
  uword *p;
  u8 *name;
  svm_subregion_t *subp;

  ASSERT (root_rp);

  a->size += MMAP_PAGESIZE +
    ((a->pvt_heap_size != 0) ? a->pvt_heap_size : SVM_PVT_MHEAP_SIZE);
  a->size = rnd_pagesize (a->size);

  region_lock (root_rp, 4);
  oldheap = svm_push_pvt_heap (root_rp);
  mp = root_rp->data_base;

  ASSERT (mp);

  /* Map the named region from the correct chroot environment */
  if (a->root_path == NULL)
    a->root_path = (char *) mp->root_path;

  /*
   * See if this region is already known. If it is, we're
   * almost done...
   */
  p = hash_get_mem (mp->name_hash, a->name);

  if (p)
    {
      rp = svm_map_region (a);
      region_unlock (root_rp);
      svm_pop_heap (oldheap);
      return rp;
    }

  /* Create the region. */
  ASSERT ((a->size & ~(MMAP_PAGESIZE - 1)) == a->size);

  need_nbits = a->size / MMAP_PAGESIZE;

  index = 1;			/* $$$ fixme, figure out how many bit to really skip */

  /*
   * Scan the virtual space allocation bitmap, looking for a large
   * enough chunk
   */
  do
    {
      if (clib_bitmap_get_no_check (root_rp->bitmap, index) == 0)
	{
	  for (i = 0; i < (need_nbits - 1); i++)
	    {
	      if (clib_bitmap_get_no_check (root_rp->bitmap, index + i) == 1)
		{
		  index = index + i;
		  goto next;
		}
	    }
	  break;
	}
      index++;
    next:;
    }
  while (index < root_rp->bitmap_size);

  /* Completely out of VM? */
  if (index >= root_rp->bitmap_size)
    {
      clib_warning ("region %s: not enough VM to allocate 0x%llx (%lld)",
		    root_rp->region_name, a->size, a->size);
      svm_pop_heap (oldheap);
      region_unlock (root_rp);
      return 0;
    }

  /*
   * Mark virtual space allocated
   */
#if CLIB_DEBUG > 1
  clib_warning ("set %d bits at index %d", need_nbits, index);
#endif

  for (i = 0; i < need_nbits; i++)
    {
      clib_bitmap_set_no_check (root_rp->bitmap, index + i, 1);
    }

  /* Place this region where it goes... */
  a->baseva = root_rp->virtual_base + index * MMAP_PAGESIZE;

  rp = svm_map_region (a);

  pool_get (mp->subregions, subp);
  name = format (0, "%s%c", a->name, 0);
  subp->subregion_name = name;

  hash_set_mem (mp->name_hash, name, subp - mp->subregions);

  svm_pop_heap (oldheap);

  region_unlock (root_rp);

  return (rp);
}

void
svm_region_unlink (svm_region_t * rp)
{
  svm_map_region_args_t _a, *a = &_a;
  svm_main_region_t *mp;
  u8 *shm_name;

  ASSERT (root_rp);
  ASSERT (rp);
  ASSERT (vec_c_string_is_terminated (rp->region_name));

  mp = root_rp->data_base;
  ASSERT (mp);

  a->root_path = (char *) mp->root_path;
  a->name = rp->region_name;
  shm_name = shm_name_from_svm_map_region_args (a);
  if (CLIB_DEBUG > 1)
    clib_warning ("[%d] shm_unlink (%s)", getpid (), shm_name);
  shm_unlink ((const char *) shm_name);
  vec_free (shm_name);
}

/*
 * svm_region_unmap
 *
 * Let go of the indicated region. If the calling process
 * is the last customer, throw it away completely.
 * The root region mutex guarantees atomicity with respect to
 * a new region client showing up at the wrong moment.
 */
void
svm_region_unmap_internal (void *rp_arg, u8 is_client)
{
  int i, mypid = getpid ();
  int nclients_left;
  void *oldheap;
  uword virtual_base, virtual_size;
  svm_region_t *rp = rp_arg;
  char *name;

  /*
   * If we take a signal while holding one or more shared-memory
   * mutexes, we may end up back here from an otherwise
   * benign exit handler. Bail out to avoid a recursive
   * mutex screw-up.
   */
  if (nheld)
    return;

  ASSERT (rp);
  ASSERT (root_rp);

  if (CLIB_DEBUG > 1)
    clib_warning ("[%d] unmap region %s", getpid (), rp->region_name);

  region_lock (root_rp, 5);
  region_lock (rp, 6);

  oldheap = svm_push_pvt_heap (rp);	/* nb vec_delete() in the loop */

  /* Remove the caller from the list of mappers */
  CLIB_MEM_UNPOISON (rp->client_pids, vec_bytes (rp->client_pids));
  for (i = 0; i < vec_len (rp->client_pids); i++)
    {
      if (rp->client_pids[i] == mypid)
	{
	  vec_delete (rp->client_pids, 1, i);
	  goto found;
	}
    }
  clib_warning ("pid %d AWOL", mypid);

found:

  svm_pop_heap (oldheap);

  nclients_left = vec_len (rp->client_pids);
  virtual_base = rp->virtual_base;
  virtual_size = rp->virtual_size;

  if (nclients_left == 0)
    {
      int index, nbits, i;
      svm_main_region_t *mp;
      uword *p;
      svm_subregion_t *subp;

      /* Kill the region, last guy on his way out */

      oldheap = svm_push_pvt_heap (root_rp);
      name = vec_dup (rp->region_name);

      virtual_base = rp->virtual_base;
      virtual_size = rp->virtual_size;

      /* Figure out which bits to clear in the root region bitmap */
      index = (virtual_base - root_rp->virtual_base) / MMAP_PAGESIZE;

      nbits = (virtual_size + MMAP_PAGESIZE - 1) / MMAP_PAGESIZE;

#if CLIB_DEBUG > 1
      clib_warning ("clear %d bits at index %d", nbits, index);
#endif
      /* Give back the allocated VM */
      for (i = 0; i < nbits; i++)
	{
	  clib_bitmap_set_no_check (root_rp->bitmap, index + i, 0);
	}

      mp = root_rp->data_base;

      p = hash_get_mem (mp->name_hash, name);

      /* Better never happen ... */
      if (p == NULL)
	{
	  region_unlock (rp);
	  region_unlock (root_rp);
	  svm_pop_heap (oldheap);
	  clib_warning ("Region name '%s' not found?", name);
	  return;
	}

      /* Remove from the root region subregion pool */
      subp = mp->subregions + p[0];
      pool_put (mp->subregions, subp);

      hash_unset_mem (mp->name_hash, name);

      vec_free (name);

      region_unlock (rp);

      /* If a client asks for the cleanup, don't unlink the backing
       * file since we can't tell if it has been recreated. */
      if (!is_client)
	svm_region_unlink (rp);

      munmap ((void *) virtual_base, virtual_size);
      region_unlock (root_rp);
      svm_pop_heap (oldheap);
      return;
    }

  region_unlock (rp);
  region_unlock (root_rp);

  munmap ((void *) virtual_base, virtual_size);
}

void
svm_region_unmap (void *rp_arg)
{
  svm_region_unmap_internal (rp_arg, 0 /* is_client */ );
}

void
svm_region_unmap_client (void *rp_arg)
{
  svm_region_unmap_internal (rp_arg, 1 /* is_client */ );
}

/*
 * svm_region_exit
 */
static void
svm_region_exit_internal (u8 is_client)
{
  void *oldheap;
  int i, mypid = getpid ();
  uword virtual_base, virtual_size;

  /* It felt so nice we did it twice... */
  if (root_rp == 0)
    return;

  if (--root_rp_refcount > 0)
    return;

  /*
   * If we take a signal while holding one or more shared-memory
   * mutexes, we may end up back here from an otherwise
   * benign exit handler. Bail out to avoid a recursive
   * mutex screw-up.
   */
  if (nheld)
    return;

  region_lock (root_rp, 7);
  oldheap = svm_push_pvt_heap (root_rp);

  virtual_base = root_rp->virtual_base;
  virtual_size = root_rp->virtual_size;

  CLIB_MEM_UNPOISON (root_rp->client_pids, vec_bytes (root_rp->client_pids));
  for (i = 0; i < vec_len (root_rp->client_pids); i++)
    {
      if (root_rp->client_pids[i] == mypid)
	{
	  vec_delete (root_rp->client_pids, 1, i);
	  goto found;
	}
    }
  clib_warning ("pid %d AWOL", mypid);

found:

  if (!is_client && vec_len (root_rp->client_pids) == 0)
    svm_region_unlink (root_rp);

  region_unlock (root_rp);
  svm_pop_heap (oldheap);

  root_rp = 0;
  munmap ((void *) virtual_base, virtual_size);
}

void
svm_region_exit (void)
{
  svm_region_exit_internal (0 /* is_client */ );
}

void
svm_region_exit_client (void)
{
  svm_region_exit_internal (1 /* is_client */ );
}

void
svm_client_scan_this_region_nolock (svm_region_t * rp)
{
  int j;
  int mypid = getpid ();
  void *oldheap;

  for (j = 0; j < vec_len (rp->client_pids); j++)
    {
      if (mypid == rp->client_pids[j])
	continue;
      if (rp->client_pids[j] && (kill (rp->client_pids[j], 0) < 0))
	{
	  clib_warning ("%s: cleanup ghost pid %d",
			rp->region_name, rp->client_pids[j]);
	  /* nb: client vec in rp->region_heap */
	  oldheap = svm_push_pvt_heap (rp);
	  vec_delete (rp->client_pids, 1, j);
	  j--;
	  svm_pop_heap (oldheap);
	}
    }
}


/*
 * Scan svm regions for dead clients
 */
void
svm_client_scan (const char *root_path)
{
  int i, j;
  svm_main_region_t *mp;
  svm_map_region_args_t *a = 0;
  svm_region_t *root_rp;
  svm_region_t *rp;
  svm_subregion_t *subp;
  u8 *name = 0;
  u8 **svm_names = 0;
  void *oldheap;
  int mypid = getpid ();

  vec_validate (a, 0);

  svm_region_init_chroot (root_path);

  root_rp = svm_get_root_rp ();

  pthread_mutex_lock (&root_rp->mutex);

  mp = root_rp->data_base;

  for (j = 0; j < vec_len (root_rp->client_pids); j++)
    {
      if (mypid == root_rp->client_pids[j])
	continue;
      if (root_rp->client_pids[j] && (kill (root_rp->client_pids[j], 0) < 0))
	{
	  clib_warning ("%s: cleanup ghost pid %d",
			root_rp->region_name, root_rp->client_pids[j]);
	  /* nb: client vec in root_rp->region_heap */
	  oldheap = svm_push_pvt_heap (root_rp);
	  vec_delete (root_rp->client_pids, 1, j);
	  j--;
	  svm_pop_heap (oldheap);
	}
    }

  /*
   * Snapshoot names, can't hold root rp mutex across
   * find_or_create.
   */
  /* *INDENT-OFF* */
  pool_foreach (subp, mp->subregions)  {
        name = vec_dup (subp->subregion_name);
        vec_add1(svm_names, name);
      }
  /* *INDENT-ON* */

  pthread_mutex_unlock (&root_rp->mutex);

  for (i = 0; i < vec_len (svm_names); i++)
    {
      vec_validate (a, 0);
      a->root_path = root_path;
      a->name = (char *) svm_names[i];
      rp = svm_region_find_or_create (a);
      if (rp)
	{
	  pthread_mutex_lock (&rp->mutex);

	  svm_client_scan_this_region_nolock (rp);

	  pthread_mutex_unlock (&rp->mutex);
	  svm_region_unmap (rp);
	  vec_free (svm_names[i]);
	}
      vec_free (a);
    }
  vec_free (svm_names);

  svm_region_exit ();

  vec_free (a);
}

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */