aboutsummaryrefslogtreecommitdiffstats
path: root/sys/kern/sys_pipe.c
blob: 22468e22622bc50a0b453fc20966fb6b1a92bd94 (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
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
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
/*-
 * Copyright (c) 1996 John S. Dyson
 * 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 immediately at the beginning of the file, without modification,
 *    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.
 * 3. Absolutely no warranty of function or purpose is made by the author
 *    John S. Dyson.
 * 4. Modifications may be freely made to this file if the above conditions
 *    are met.
 */

/*
 * This file contains a high-performance replacement for the socket-based
 * pipes scheme originally used in FreeBSD/4.4Lite.  It does not support
 * all features of sockets, but does do everything that pipes normally
 * do.
 */

/*
 * This code has two modes of operation, a small write mode and a large
 * write mode.  The small write mode acts like conventional pipes with
 * a kernel buffer.  If the buffer is less than PIPE_MINDIRECT, then the
 * "normal" pipe buffering is done.  If the buffer is between PIPE_MINDIRECT
 * and PIPE_SIZE in size, it is fully mapped and wired into the kernel, and
 * the receiving process can copy it directly from the pages in the sending
 * process.
 *
 * If the sending process receives a signal, it is possible that it will
 * go away, and certainly its address space can change, because control
 * is returned back to the user-mode side.  In that case, the pipe code
 * arranges to copy the buffer supplied by the user process, to a pageable
 * kernel buffer, and the receiving process will grab the data from the
 * pageable kernel buffer.  Since signals don't happen all that often,
 * the copy operation is normally eliminated.
 *
 * The constant PIPE_MINDIRECT is chosen to make sure that buffering will
 * happen for small transfers so that the system will not spend all of
 * its time context switching.
 *
 * In order to limit the resource use of pipes, two sysctls exist:
 *
 * kern.ipc.maxpipekva - This is a hard limit on the amount of pageable
 * address space available to us in pipe_map. This value is normally
 * autotuned, but may also be loader tuned.
 *
 * kern.ipc.pipekva - This read-only sysctl tracks the current amount of
 * memory in use by pipes.
 *
 * Based on how large pipekva is relative to maxpipekva, the following
 * will happen:
 *
 * 0% - 50%:
 *     New pipes are given 16K of memory backing, pipes may dynamically
 *     grow to as large as 64K where needed.
 * 50% - 75%:
 *     New pipes are given 4K (or PAGE_SIZE) of memory backing,
 *     existing pipes may NOT grow.
 * 75% - 100%:
 *     New pipes are given 4K (or PAGE_SIZE) of memory backing,
 *     existing pipes will be shrunk down to 4K whenever possible.
 *
 * Resizing may be disabled by setting kern.ipc.piperesizeallowed=0.  If
 * that is set,  the only resize that will occur is the 0 -> SMALL_PIPE_SIZE
 * resize which MUST occur for reverse-direction pipes when they are
 * first used.
 *
 * Additional information about the current state of pipes may be obtained
 * from kern.ipc.pipes, kern.ipc.pipefragretry, kern.ipc.pipeallocfail,
 * and kern.ipc.piperesizefail.
 *
 * Locking rules:  There are two locks present here:  A mutex, used via
 * PIPE_LOCK, and a flag, used via pipelock().  All locking is done via
 * the flag, as mutexes can not persist over uiomove.  The mutex
 * exists only to guard access to the flag, and is not in itself a
 * locking mechanism.  Also note that there is only a single mutex for
 * both directions of a pipe.
 *
 * As pipelock() may have to sleep before it can acquire the flag, it
 * is important to reread all data after a call to pipelock(); everything
 * in the structure may have changed.
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include "opt_mac.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/fcntl.h>
#include <sys/file.h>
#include <sys/filedesc.h>
#include <sys/filio.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/ttycom.h>
#include <sys/stat.h>
#include <sys/malloc.h>
#include <sys/poll.h>
#include <sys/selinfo.h>
#include <sys/signalvar.h>
#include <sys/sysctl.h>
#include <sys/sysproto.h>
#include <sys/pipe.h>
#include <sys/proc.h>
#include <sys/vnode.h>
#include <sys/uio.h>
#include <sys/event.h>

#include <security/mac/mac_framework.h>

#include <vm/vm.h>
#include <vm/vm_param.h>
#include <vm/vm_object.h>
#include <vm/vm_kern.h>
#include <vm/vm_extern.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
#include <vm/vm_page.h>
#include <vm/uma.h>

/*
 * Use this define if you want to disable *fancy* VM things.  Expect an
 * approx 30% decrease in transfer rate.  This could be useful for
 * NetBSD or OpenBSD.
 */
/* #define PIPE_NODIRECT */

/*
 * interfaces to the outside world
 */
static fo_rdwr_t	pipe_read;
static fo_rdwr_t	pipe_write;
static fo_ioctl_t	pipe_ioctl;
static fo_poll_t	pipe_poll;
static fo_kqfilter_t	pipe_kqfilter;
static fo_stat_t	pipe_stat;
static fo_close_t	pipe_close;

static struct fileops pipeops = {
	.fo_read = pipe_read,
	.fo_write = pipe_write,
	.fo_ioctl = pipe_ioctl,
	.fo_poll = pipe_poll,
	.fo_kqfilter = pipe_kqfilter,
	.fo_stat = pipe_stat,
	.fo_close = pipe_close,
	.fo_flags = DFLAG_PASSABLE
};

static void	filt_pipedetach(struct knote *kn);
static int	filt_piperead(struct knote *kn, long hint);
static int	filt_pipewrite(struct knote *kn, long hint);

static struct filterops pipe_rfiltops =
	{ 1, NULL, filt_pipedetach, filt_piperead };
static struct filterops pipe_wfiltops =
	{ 1, NULL, filt_pipedetach, filt_pipewrite };

/*
 * Default pipe buffer size(s), this can be kind-of large now because pipe
 * space is pageable.  The pipe code will try to maintain locality of
 * reference for performance reasons, so small amounts of outstanding I/O
 * will not wipe the cache.
 */
#define MINPIPESIZE (PIPE_SIZE/3)
#define MAXPIPESIZE (2*PIPE_SIZE/3)

static int amountpipekva;
static int pipefragretry;
static int pipeallocfail;
static int piperesizefail;
static int piperesizeallowed = 1;

SYSCTL_INT(_kern_ipc, OID_AUTO, maxpipekva, CTLFLAG_RDTUN,
	   &maxpipekva, 0, "Pipe KVA limit");
SYSCTL_INT(_kern_ipc, OID_AUTO, pipekva, CTLFLAG_RD,
	   &amountpipekva, 0, "Pipe KVA usage");
SYSCTL_INT(_kern_ipc, OID_AUTO, pipefragretry, CTLFLAG_RD,
	  &pipefragretry, 0, "Pipe allocation retries due to fragmentation");
SYSCTL_INT(_kern_ipc, OID_AUTO, pipeallocfail, CTLFLAG_RD,
	  &pipeallocfail, 0, "Pipe allocation failures");
SYSCTL_INT(_kern_ipc, OID_AUTO, piperesizefail, CTLFLAG_RD,
	  &piperesizefail, 0, "Pipe resize failures");
SYSCTL_INT(_kern_ipc, OID_AUTO, piperesizeallowed, CTLFLAG_RW,
	  &piperesizeallowed, 0, "Pipe resizing allowed");

static void pipeinit(void *dummy __unused);
static void pipeclose(struct pipe *cpipe);
static void pipe_free_kmem(struct pipe *cpipe);
static int pipe_create(struct pipe *pipe, int backing);
static __inline int pipelock(struct pipe *cpipe, int catch);
static __inline void pipeunlock(struct pipe *cpipe);
static __inline void pipeselwakeup(struct pipe *cpipe);
#ifndef PIPE_NODIRECT
static int pipe_build_write_buffer(struct pipe *wpipe, struct uio *uio);
static void pipe_destroy_write_buffer(struct pipe *wpipe);
static int pipe_direct_write(struct pipe *wpipe, struct uio *uio);
static void pipe_clone_write_buffer(struct pipe *wpipe);
#endif
static int pipespace(struct pipe *cpipe, int size);
static int pipespace_new(struct pipe *cpipe, int size);

static int	pipe_zone_ctor(void *mem, int size, void *arg, int flags);
static int	pipe_zone_init(void *mem, int size, int flags);
static void	pipe_zone_fini(void *mem, int size);

static uma_zone_t pipe_zone;

SYSINIT(vfs, SI_SUB_VFS, SI_ORDER_ANY, pipeinit, NULL);

static void
pipeinit(void *dummy __unused)
{

	pipe_zone = uma_zcreate("pipe", sizeof(struct pipepair),
	    pipe_zone_ctor, NULL, pipe_zone_init, pipe_zone_fini,
	    UMA_ALIGN_PTR, 0);
	KASSERT(pipe_zone != NULL, ("pipe_zone not initialized"));
}

static int
pipe_zone_ctor(void *mem, int size, void *arg, int flags)
{
	struct pipepair *pp;
	struct pipe *rpipe, *wpipe;

	KASSERT(size == sizeof(*pp), ("pipe_zone_ctor: wrong size"));

	pp = (struct pipepair *)mem;

	/*
	 * We zero both pipe endpoints to make sure all the kmem pointers
	 * are NULL, flag fields are zero'd, etc.  We timestamp both
	 * endpoints with the same time.
	 */
	rpipe = &pp->pp_rpipe;
	bzero(rpipe, sizeof(*rpipe));
	vfs_timestamp(&rpipe->pipe_ctime);
	rpipe->pipe_atime = rpipe->pipe_mtime = rpipe->pipe_ctime;

	wpipe = &pp->pp_wpipe;
	bzero(wpipe, sizeof(*wpipe));
	wpipe->pipe_ctime = rpipe->pipe_ctime;
	wpipe->pipe_atime = wpipe->pipe_mtime = rpipe->pipe_ctime;

	rpipe->pipe_peer = wpipe;
	rpipe->pipe_pair = pp;
	wpipe->pipe_peer = rpipe;
	wpipe->pipe_pair = pp;

	/*
	 * Mark both endpoints as present; they will later get free'd
	 * one at a time.  When both are free'd, then the whole pair
	 * is released.
	 */
	rpipe->pipe_present = PIPE_ACTIVE;
	wpipe->pipe_present = PIPE_ACTIVE;

	/*
	 * Eventually, the MAC Framework may initialize the label
	 * in ctor or init, but for now we do it elswhere to avoid
	 * blocking in ctor or init.
	 */
	pp->pp_label = NULL;

	return (0);
}

static int
pipe_zone_init(void *mem, int size, int flags)
{
	struct pipepair *pp;

	KASSERT(size == sizeof(*pp), ("pipe_zone_init: wrong size"));

	pp = (struct pipepair *)mem;

	mtx_init(&pp->pp_mtx, "pipe mutex", NULL, MTX_DEF | MTX_RECURSE);
	return (0);
}

static void
pipe_zone_fini(void *mem, int size)
{
	struct pipepair *pp;

	KASSERT(size == sizeof(*pp), ("pipe_zone_fini: wrong size"));

	pp = (struct pipepair *)mem;

	mtx_destroy(&pp->pp_mtx);
}

/*
 * The pipe system call for the DTYPE_PIPE type of pipes.  If we fail, let
 * the zone pick up the pieces via pipeclose().
 */
/* ARGSUSED */
int
pipe(td, uap)
	struct thread *td;
	struct pipe_args /* {
		int	dummy;
	} */ *uap;
{
	struct filedesc *fdp = td->td_proc->p_fd;
	struct file *rf, *wf;
	struct pipepair *pp;
	struct pipe *rpipe, *wpipe;
	int fd, error;

	pp = uma_zalloc(pipe_zone, M_WAITOK);
#ifdef MAC
	/*
	 * The MAC label is shared between the connected endpoints.  As a
	 * result mac_init_pipe() and mac_create_pipe() are called once
	 * for the pair, and not on the endpoints.
	 */
	mac_init_pipe(pp);
	mac_create_pipe(td->td_ucred, pp);
#endif
	rpipe = &pp->pp_rpipe;
	wpipe = &pp->pp_wpipe;

	knlist_init(&rpipe->pipe_sel.si_note, PIPE_MTX(rpipe), NULL, NULL,
	    NULL);
	knlist_init(&wpipe->pipe_sel.si_note, PIPE_MTX(wpipe), NULL, NULL,
	    NULL);

	/* Only the forward direction pipe is backed by default */
	if ((error = pipe_create(rpipe, 1)) != 0 ||
	    (error = pipe_create(wpipe, 0)) != 0) {
		pipeclose(rpipe);
		pipeclose(wpipe);
		return (error);
	}

	rpipe->pipe_state |= PIPE_DIRECTOK;
	wpipe->pipe_state |= PIPE_DIRECTOK;

	error = falloc(td, &rf, &fd);
	if (error) {
		pipeclose(rpipe);
		pipeclose(wpipe);
		return (error);
	}
	/* An extra reference on `rf' has been held for us by falloc(). */
	td->td_retval[0] = fd;

	/*
	 * Warning: once we've gotten past allocation of the fd for the
	 * read-side, we can only drop the read side via fdrop() in order
	 * to avoid races against processes which manage to dup() the read
	 * side while we are blocked trying to allocate the write side.
	 */
	FILE_LOCK(rf);
	rf->f_flag = FREAD | FWRITE;
	rf->f_type = DTYPE_PIPE;
	rf->f_data = rpipe;
	rf->f_ops = &pipeops;
	FILE_UNLOCK(rf);
	error = falloc(td, &wf, &fd);
	if (error) {
		fdclose(fdp, rf, td->td_retval[0], td);
		fdrop(rf, td);
		/* rpipe has been closed by fdrop(). */
		pipeclose(wpipe);
		return (error);
	}
	/* An extra reference on `wf' has been held for us by falloc(). */
	FILE_LOCK(wf);
	wf->f_flag = FREAD | FWRITE;
	wf->f_type = DTYPE_PIPE;
	wf->f_data = wpipe;
	wf->f_ops = &pipeops;
	FILE_UNLOCK(wf);
	fdrop(wf, td);
	td->td_retval[1] = fd;
	fdrop(rf, td);

	return (0);
}

/*
 * Allocate kva for pipe circular buffer, the space is pageable
 * This routine will 'realloc' the size of a pipe safely, if it fails
 * it will retain the old buffer.
 * If it fails it will return ENOMEM.
 */
static int
pipespace_new(cpipe, size)
	struct pipe *cpipe;
	int size;
{
	caddr_t buffer;
	int error, cnt, firstseg;
	static int curfail = 0;
	static struct timeval lastfail;

	KASSERT(!mtx_owned(PIPE_MTX(cpipe)), ("pipespace: pipe mutex locked"));
	KASSERT(!(cpipe->pipe_state & PIPE_DIRECTW),
		("pipespace: resize of direct writes not allowed"));
retry:
	cnt = cpipe->pipe_buffer.cnt;
	if (cnt > size)
		size = cnt;

	size = round_page(size);
	buffer = (caddr_t) vm_map_min(pipe_map);

	error = vm_map_find(pipe_map, NULL, 0,
		(vm_offset_t *) &buffer, size, 1,
		VM_PROT_ALL, VM_PROT_ALL, 0);
	if (error != KERN_SUCCESS) {
		if ((cpipe->pipe_buffer.buffer == NULL) &&
			(size > SMALL_PIPE_SIZE)) {
			size = SMALL_PIPE_SIZE;
			pipefragretry++;
			goto retry;
		}
		if (cpipe->pipe_buffer.buffer == NULL) {
			pipeallocfail++;
			if (ppsratecheck(&lastfail, &curfail, 1))
				printf("kern.ipc.maxpipekva exceeded; see tuning(7)\n");
		} else {
			piperesizefail++;
		}
		return (ENOMEM);
	}

	/* copy data, then free old resources if we're resizing */
	if (cnt > 0) {
		if (cpipe->pipe_buffer.in <= cpipe->pipe_buffer.out) {
			firstseg = cpipe->pipe_buffer.size - cpipe->pipe_buffer.out;
			bcopy(&cpipe->pipe_buffer.buffer[cpipe->pipe_buffer.out],
				buffer, firstseg);
			if ((cnt - firstseg) > 0)
				bcopy(cpipe->pipe_buffer.buffer, &buffer[firstseg],
					cpipe->pipe_buffer.in);
		} else {
			bcopy(&cpipe->pipe_buffer.buffer[cpipe->pipe_buffer.out],
				buffer, cnt);
		}
	}
	pipe_free_kmem(cpipe);
	cpipe->pipe_buffer.buffer = buffer;
	cpipe->pipe_buffer.size = size;
	cpipe->pipe_buffer.in = cnt;
	cpipe->pipe_buffer.out = 0;
	cpipe->pipe_buffer.cnt = cnt;
	atomic_add_int(&amountpipekva, cpipe->pipe_buffer.size);
	return (0);
}

/*
 * Wrapper for pipespace_new() that performs locking assertions.
 */
static int
pipespace(cpipe, size)
	struct pipe *cpipe;
	int size;
{

	KASSERT(cpipe->pipe_state & PIPE_LOCKFL,
		("Unlocked pipe passed to pipespace"));
	return (pipespace_new(cpipe, size));
}

/*
 * lock a pipe for I/O, blocking other access
 */
static __inline int
pipelock(cpipe, catch)
	struct pipe *cpipe;
	int catch;
{
	int error;

	PIPE_LOCK_ASSERT(cpipe, MA_OWNED);
	while (cpipe->pipe_state & PIPE_LOCKFL) {
		cpipe->pipe_state |= PIPE_LWANT;
		error = msleep(cpipe, PIPE_MTX(cpipe),
		    catch ? (PRIBIO | PCATCH) : PRIBIO,
		    "pipelk", 0);
		if (error != 0)
			return (error);
	}
	cpipe->pipe_state |= PIPE_LOCKFL;
	return (0);
}

/*
 * unlock a pipe I/O lock
 */
static __inline void
pipeunlock(cpipe)
	struct pipe *cpipe;
{

	PIPE_LOCK_ASSERT(cpipe, MA_OWNED);
	KASSERT(cpipe->pipe_state & PIPE_LOCKFL,
		("Unlocked pipe passed to pipeunlock"));
	cpipe->pipe_state &= ~PIPE_LOCKFL;
	if (cpipe->pipe_state & PIPE_LWANT) {
		cpipe->pipe_state &= ~PIPE_LWANT;
		wakeup(cpipe);
	}
}

static __inline void
pipeselwakeup(cpipe)
	struct pipe *cpipe;
{

	PIPE_LOCK_ASSERT(cpipe, MA_OWNED);
	if (cpipe->pipe_state & PIPE_SEL) {
		cpipe->pipe_state &= ~PIPE_SEL;
		selwakeuppri(&cpipe->pipe_sel, PSOCK);
	}
	if ((cpipe->pipe_state & PIPE_ASYNC) && cpipe->pipe_sigio)
		pgsigio(&cpipe->pipe_sigio, SIGIO, 0);
	KNOTE_LOCKED(&cpipe->pipe_sel.si_note, 0);
}

/*
 * Initialize and allocate VM and memory for pipe.  The structure
 * will start out zero'd from the ctor, so we just manage the kmem.
 */
static int
pipe_create(pipe, backing)
	struct pipe *pipe;
	int backing;
{
	int error;

	if (backing) {
		if (amountpipekva > maxpipekva / 2)
			error = pipespace_new(pipe, SMALL_PIPE_SIZE);
		else
			error = pipespace_new(pipe, PIPE_SIZE);
	} else {
		/* If we're not backing this pipe, no need to do anything. */
		error = 0;
	}
	return (error);
}

/* ARGSUSED */
static int
pipe_read(fp, uio, active_cred, flags, td)
	struct file *fp;
	struct uio *uio;
	struct ucred *active_cred;
	struct thread *td;
	int flags;
{
	struct pipe *rpipe = fp->f_data;
	int error;
	int nread = 0;
	u_int size;

	PIPE_LOCK(rpipe);
	++rpipe->pipe_busy;
	error = pipelock(rpipe, 1);
	if (error)
		goto unlocked_error;

#ifdef MAC
	error = mac_check_pipe_read(active_cred, rpipe->pipe_pair);
	if (error)
		goto locked_error;
#endif
	if (amountpipekva > (3 * maxpipekva) / 4) {
		if (!(rpipe->pipe_state & PIPE_DIRECTW) &&
			(rpipe->pipe_buffer.size > SMALL_PIPE_SIZE) &&
			(rpipe->pipe_buffer.cnt <= SMALL_PIPE_SIZE) &&
			(piperesizeallowed == 1)) {
			PIPE_UNLOCK(rpipe);
			pipespace(rpipe, SMALL_PIPE_SIZE);
			PIPE_LOCK(rpipe);
		}
	}

	while (uio->uio_resid) {
		/*
		 * normal pipe buffer receive
		 */
		if (rpipe->pipe_buffer.cnt > 0) {
			size = rpipe->pipe_buffer.size - rpipe->pipe_buffer.out;
			if (size > rpipe->pipe_buffer.cnt)
				size = rpipe->pipe_buffer.cnt;
			if (size > (u_int) uio->uio_resid)
				size = (u_int) uio->uio_resid;

			PIPE_UNLOCK(rpipe);
			error = uiomove(
			    &rpipe->pipe_buffer.buffer[rpipe->pipe_buffer.out],
			    size, uio);
			PIPE_LOCK(rpipe);
			if (error)
				break;

			rpipe->pipe_buffer.out += size;
			if (rpipe->pipe_buffer.out >= rpipe->pipe_buffer.size)
				rpipe->pipe_buffer.out = 0;

			rpipe->pipe_buffer.cnt -= size;

			/*
			 * If there is no more to read in the pipe, reset
			 * its pointers to the beginning.  This improves
			 * cache hit stats.
			 */
			if (rpipe->pipe_buffer.cnt == 0) {
				rpipe->pipe_buffer.in = 0;
				rpipe->pipe_buffer.out = 0;
			}
			nread += size;
#ifndef PIPE_NODIRECT
		/*
		 * Direct copy, bypassing a kernel buffer.
		 */
		} else if ((size = rpipe->pipe_map.cnt) &&
			   (rpipe->pipe_state & PIPE_DIRECTW)) {
			if (size > (u_int) uio->uio_resid)
				size = (u_int) uio->uio_resid;

			PIPE_UNLOCK(rpipe);
			error = uiomove_fromphys(rpipe->pipe_map.ms,
			    rpipe->pipe_map.pos, size, uio);
			PIPE_LOCK(rpipe);
			if (error)
				break;
			nread += size;
			rpipe->pipe_map.pos += size;
			rpipe->pipe_map.cnt -= size;
			if (rpipe->pipe_map.cnt == 0) {
				rpipe->pipe_state &= ~PIPE_DIRECTW;
				wakeup(rpipe);
			}
#endif
		} else {
			/*
			 * detect EOF condition
			 * read returns 0 on EOF, no need to set error
			 */
			if (rpipe->pipe_state & PIPE_EOF)
				break;

			/*
			 * If the "write-side" has been blocked, wake it up now.
			 */
			if (rpipe->pipe_state & PIPE_WANTW) {
				rpipe->pipe_state &= ~PIPE_WANTW;
				wakeup(rpipe);
			}

			/*
			 * Break if some data was read.
			 */
			if (nread > 0)
				break;

			/*
			 * Unlock the pipe buffer for our remaining processing.
			 * We will either break out with an error or we will
			 * sleep and relock to loop.
			 */
			pipeunlock(rpipe);

			/*
			 * Handle non-blocking mode operation or
			 * wait for more data.
			 */
			if (fp->f_flag & FNONBLOCK) {
				error = EAGAIN;
			} else {
				rpipe->pipe_state |= PIPE_WANTR;
				if ((error = msleep(rpipe, PIPE_MTX(rpipe),
				    PRIBIO | PCATCH,
				    "piperd", 0)) == 0)
					error = pipelock(rpipe, 1);
			}
			if (error)
				goto unlocked_error;
		}
	}
#ifdef MAC
locked_error:
#endif
	pipeunlock(rpipe);

	/* XXX: should probably do this before getting any locks. */
	if (error == 0)
		vfs_timestamp(&rpipe->pipe_atime);
unlocked_error:
	--rpipe->pipe_busy;

	/*
	 * PIPE_WANT processing only makes sense if pipe_busy is 0.
	 */
	if ((rpipe->pipe_busy == 0) && (rpipe->pipe_state & PIPE_WANT)) {
		rpipe->pipe_state &= ~(PIPE_WANT|PIPE_WANTW);
		wakeup(rpipe);
	} else if (rpipe->pipe_buffer.cnt < MINPIPESIZE) {
		/*
		 * Handle write blocking hysteresis.
		 */
		if (rpipe->pipe_state & PIPE_WANTW) {
			rpipe->pipe_state &= ~PIPE_WANTW;
			wakeup(rpipe);
		}
	}

	if ((rpipe->pipe_buffer.size - rpipe->pipe_buffer.cnt) >= PIPE_BUF)
		pipeselwakeup(rpipe);

	PIPE_UNLOCK(rpipe);
	return (error);
}

#ifndef PIPE_NODIRECT
/*
 * Map the sending processes' buffer into kernel space and wire it.
 * This is similar to a physical write operation.
 */
static int
pipe_build_write_buffer(wpipe, uio)
	struct pipe *wpipe;
	struct uio *uio;
{
	pmap_t pmap;
	u_int size;
	int i, j;
	vm_offset_t addr, endaddr;

	PIPE_LOCK_ASSERT(wpipe, MA_NOTOWNED);
	KASSERT(wpipe->pipe_state & PIPE_DIRECTW,
		("Clone attempt on non-direct write pipe!"));

	size = (u_int) uio->uio_iov->iov_len;
	if (size > wpipe->pipe_buffer.size)
		size = wpipe->pipe_buffer.size;

	pmap = vmspace_pmap(curproc->p_vmspace);
	endaddr = round_page((vm_offset_t)uio->uio_iov->iov_base + size);
	addr = trunc_page((vm_offset_t)uio->uio_iov->iov_base);
	for (i = 0; addr < endaddr; addr += PAGE_SIZE, i++) {
		/*
		 * vm_fault_quick() can sleep.  Consequently,
		 * vm_page_lock_queue() and vm_page_unlock_queue()
		 * should not be performed outside of this loop.
		 */
	race:
		if (vm_fault_quick((caddr_t)addr, VM_PROT_READ) < 0) {
			vm_page_lock_queues();
			for (j = 0; j < i; j++)
				vm_page_unhold(wpipe->pipe_map.ms[j]);
			vm_page_unlock_queues();
			return (EFAULT);
		}
		wpipe->pipe_map.ms[i] = pmap_extract_and_hold(pmap, addr,
		    VM_PROT_READ);
		if (wpipe->pipe_map.ms[i] == NULL)
			goto race;
	}

/*
 * set up the control block
 */
	wpipe->pipe_map.npages = i;
	wpipe->pipe_map.pos =
	    ((vm_offset_t) uio->uio_iov->iov_base) & PAGE_MASK;
	wpipe->pipe_map.cnt = size;

/*
 * and update the uio data
 */

	uio->uio_iov->iov_len -= size;
	uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + size;
	if (uio->uio_iov->iov_len == 0)
		uio->uio_iov++;
	uio->uio_resid -= size;
	uio->uio_offset += size;
	return (0);
}

/*
 * unmap and unwire the process buffer
 */
static void
pipe_destroy_write_buffer(wpipe)
	struct pipe *wpipe;
{
	int i;

	PIPE_LOCK_ASSERT(wpipe, MA_OWNED);
	vm_page_lock_queues();
	for (i = 0; i < wpipe->pipe_map.npages; i++) {
		vm_page_unhold(wpipe->pipe_map.ms[i]);
	}
	vm_page_unlock_queues();
	wpipe->pipe_map.npages = 0;
}

/*
 * In the case of a signal, the writing process might go away.  This
 * code copies the data into the circular buffer so that the source
 * pages can be freed without loss of data.
 */
static void
pipe_clone_write_buffer(wpipe)
	struct pipe *wpipe;
{
	struct uio uio;
	struct iovec iov;
	int size;
	int pos;

	PIPE_LOCK_ASSERT(wpipe, MA_OWNED);
	size = wpipe->pipe_map.cnt;
	pos = wpipe->pipe_map.pos;

	wpipe->pipe_buffer.in = size;
	wpipe->pipe_buffer.out = 0;
	wpipe->pipe_buffer.cnt = size;
	wpipe->pipe_state &= ~PIPE_DIRECTW;

	PIPE_UNLOCK(wpipe);
	iov.iov_base = wpipe->pipe_buffer.buffer;
	iov.iov_len = size;
	uio.uio_iov = &iov;
	uio.uio_iovcnt = 1;
	uio.uio_offset = 0;
	uio.uio_resid = size;
	uio.uio_segflg = UIO_SYSSPACE;
	uio.uio_rw = UIO_READ;
	uio.uio_td = curthread;
	uiomove_fromphys(wpipe->pipe_map.ms, pos, size, &uio);
	PIPE_LOCK(wpipe);
	pipe_destroy_write_buffer(wpipe);
}

/*
 * This implements the pipe buffer write mechanism.  Note that only
 * a direct write OR a normal pipe write can be pending at any given time.
 * If there are any characters in the pipe buffer, the direct write will
 * be deferred until the receiving process grabs all of the bytes from
 * the pipe buffer.  Then the direct mapping write is set-up.
 */
static int
pipe_direct_write(wpipe, uio)
	struct pipe *wpipe;
	struct uio *uio;
{
	int error;

retry:
	PIPE_LOCK_ASSERT(wpipe, MA_OWNED);
	error = pipelock(wpipe, 1);
	if (wpipe->pipe_state & PIPE_EOF)
		error = EPIPE;
	if (error) {
		pipeunlock(wpipe);
		goto error1;
	}
	while (wpipe->pipe_state & PIPE_DIRECTW) {
		if (wpipe->pipe_state & PIPE_WANTR) {
			wpipe->pipe_state &= ~PIPE_WANTR;
			wakeup(wpipe);
		}
		pipeselwakeup(wpipe);
		wpipe->pipe_state |= PIPE_WANTW;
		pipeunlock(wpipe);
		error = msleep(wpipe, PIPE_MTX(wpipe),
		    PRIBIO | PCATCH, "pipdww", 0);
		if (error)
			goto error1;
		else
			goto retry;
	}
	wpipe->pipe_map.cnt = 0;	/* transfer not ready yet */
	if (wpipe->pipe_buffer.cnt > 0) {
		if (wpipe->pipe_state & PIPE_WANTR) {
			wpipe->pipe_state &= ~PIPE_WANTR;
			wakeup(wpipe);
		}
		pipeselwakeup(wpipe);
		wpipe->pipe_state |= PIPE_WANTW;
		pipeunlock(wpipe);
		error = msleep(wpipe, PIPE_MTX(wpipe),
		    PRIBIO | PCATCH, "pipdwc", 0);
		if (error)
			goto error1;
		else
			goto retry;
	}

	wpipe->pipe_state |= PIPE_DIRECTW;

	PIPE_UNLOCK(wpipe);
	error = pipe_build_write_buffer(wpipe, uio);
	PIPE_LOCK(wpipe);
	if (error) {
		wpipe->pipe_state &= ~PIPE_DIRECTW;
		pipeunlock(wpipe);
		goto error1;
	}

	error = 0;
	while (!error && (wpipe->pipe_state & PIPE_DIRECTW)) {
		if (wpipe->pipe_state & PIPE_EOF) {
			pipe_destroy_write_buffer(wpipe);
			pipeselwakeup(wpipe);
			pipeunlock(wpipe);
			error = EPIPE;
			goto error1;
		}
		if (wpipe->pipe_state & PIPE_WANTR) {
			wpipe->pipe_state &= ~PIPE_WANTR;
			wakeup(wpipe);
		}
		pipeselwakeup(wpipe);
		pipeunlock(wpipe);
		error = msleep(wpipe, PIPE_MTX(wpipe), PRIBIO | PCATCH,
		    "pipdwt", 0);
		pipelock(wpipe, 0);
	}

	if (wpipe->pipe_state & PIPE_EOF)
		error = EPIPE;
	if (wpipe->pipe_state & PIPE_DIRECTW) {
		/*
		 * this bit of trickery substitutes a kernel buffer for
		 * the process that might be going away.
		 */
		pipe_clone_write_buffer(wpipe);
	} else {
		pipe_destroy_write_buffer(wpipe);
	}
	pipeunlock(wpipe);
	return (error);

error1:
	wakeup(wpipe);
	return (error);
}
#endif

static int
pipe_write(fp, uio, active_cred, flags, td)
	struct file *fp;
	struct uio *uio;
	struct ucred *active_cred;
	struct thread *td;
	int flags;
{
	int error = 0;
	int desiredsize, orig_resid;
	struct pipe *wpipe, *rpipe;

	rpipe = fp->f_data;
	wpipe = rpipe->pipe_peer;

	PIPE_LOCK(rpipe);
	error = pipelock(wpipe, 1);
	if (error) {
		PIPE_UNLOCK(rpipe);
		return (error);
	}
	/*
	 * detect loss of pipe read side, issue SIGPIPE if lost.
	 */
	if (wpipe->pipe_present != PIPE_ACTIVE ||
	    (wpipe->pipe_state & PIPE_EOF)) {
		pipeunlock(wpipe);
		PIPE_UNLOCK(rpipe);
		return (EPIPE);
	}
#ifdef MAC
	error = mac_check_pipe_write(active_cred, wpipe->pipe_pair);
	if (error) {
		pipeunlock(wpipe);
		PIPE_UNLOCK(rpipe);
		return (error);
	}
#endif
	++wpipe->pipe_busy;

	/* Choose a larger size if it's advantageous */
	desiredsize = max(SMALL_PIPE_SIZE, wpipe->pipe_buffer.size);
	while (desiredsize < wpipe->pipe_buffer.cnt + uio->uio_resid) {
		if (piperesizeallowed != 1)
			break;
		if (amountpipekva > maxpipekva / 2)
			break;
		if (desiredsize == BIG_PIPE_SIZE)
			break;
		desiredsize = desiredsize * 2;
	}

	/* Choose a smaller size if we're in a OOM situation */
	if ((amountpipekva > (3 * maxpipekva) / 4) &&
		(wpipe->pipe_buffer.size > SMALL_PIPE_SIZE) &&
		(wpipe->pipe_buffer.cnt <= SMALL_PIPE_SIZE) &&
		(piperesizeallowed == 1))
		desiredsize = SMALL_PIPE_SIZE;

	/* Resize if the above determined that a new size was necessary */
	if ((desiredsize != wpipe->pipe_buffer.size) &&
		((wpipe->pipe_state & PIPE_DIRECTW) == 0)) {
		PIPE_UNLOCK(wpipe);
		pipespace(wpipe, desiredsize);
		PIPE_LOCK(wpipe);
	}
	if (wpipe->pipe_buffer.size == 0) {
		/*
		 * This can only happen for reverse direction use of pipes
		 * in a complete OOM situation.
		 */
		error = ENOMEM;
		--wpipe->pipe_busy;
		pipeunlock(wpipe);
		PIPE_UNLOCK(wpipe);
		return (error);
	}

	pipeunlock(wpipe);

	orig_resid = uio->uio_resid;

	while (uio->uio_resid) {
		int space;

		pipelock(wpipe, 0);
		if (wpipe->pipe_state & PIPE_EOF) {
			pipeunlock(wpipe);
			error = EPIPE;
			break;
		}
#ifndef PIPE_NODIRECT
		/*
		 * If the transfer is large, we can gain performance if
		 * we do process-to-process copies directly.
		 * If the write is non-blocking, we don't use the
		 * direct write mechanism.
		 *
		 * The direct write mechanism will detect the reader going
		 * away on us.
		 */
		if (uio->uio_segflg == UIO_USERSPACE &&
		    uio->uio_iov->iov_len >= PIPE_MINDIRECT &&
		    wpipe->pipe_buffer.size >= PIPE_MINDIRECT &&
		    (fp->f_flag & FNONBLOCK) == 0) {
			pipeunlock(wpipe);
			error = pipe_direct_write(wpipe, uio);
			if (error)
				break;
			continue;
		}
#endif

		/*
		 * Pipe buffered writes cannot be coincidental with
		 * direct writes.  We wait until the currently executing
		 * direct write is completed before we start filling the
		 * pipe buffer.  We break out if a signal occurs or the
		 * reader goes away.
		 */
		if (wpipe->pipe_state & PIPE_DIRECTW) {
			if (wpipe->pipe_state & PIPE_WANTR) {
				wpipe->pipe_state &= ~PIPE_WANTR;
				wakeup(wpipe);
			}
			pipeselwakeup(wpipe);
			wpipe->pipe_state |= PIPE_WANTW;
			pipeunlock(wpipe);
			error = msleep(wpipe, PIPE_MTX(rpipe), PRIBIO | PCATCH,
			    "pipbww", 0);
			if (error)
				break;
			else
				continue;
		}

		space = wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt;

		/* Writes of size <= PIPE_BUF must be atomic. */
		if ((space < uio->uio_resid) && (orig_resid <= PIPE_BUF))
			space = 0;

		if (space > 0) {
			int size;	/* Transfer size */
			int segsize;	/* first segment to transfer */

			/*
			 * Transfer size is minimum of uio transfer
			 * and free space in pipe buffer.
			 */
			if (space > uio->uio_resid)
				size = uio->uio_resid;
			else
				size = space;
			/*
			 * First segment to transfer is minimum of
			 * transfer size and contiguous space in
			 * pipe buffer.  If first segment to transfer
			 * is less than the transfer size, we've got
			 * a wraparound in the buffer.
			 */
			segsize = wpipe->pipe_buffer.size -
				wpipe->pipe_buffer.in;
			if (segsize > size)
				segsize = size;

			/* Transfer first segment */

			PIPE_UNLOCK(rpipe);
			error = uiomove(&wpipe->pipe_buffer.buffer[wpipe->pipe_buffer.in],
					segsize, uio);
			PIPE_LOCK(rpipe);

			if (error == 0 && segsize < size) {
				KASSERT(wpipe->pipe_buffer.in + segsize ==
					wpipe->pipe_buffer.size,
					("Pipe buffer wraparound disappeared"));
				/*
				 * Transfer remaining part now, to
				 * support atomic writes.  Wraparound
				 * happened.
				 */

				PIPE_UNLOCK(rpipe);
				error = uiomove(
				    &wpipe->pipe_buffer.buffer[0],
				    size - segsize, uio);
				PIPE_LOCK(rpipe);
			}
			if (error == 0) {
				wpipe->pipe_buffer.in += size;
				if (wpipe->pipe_buffer.in >=
				    wpipe->pipe_buffer.size) {
					KASSERT(wpipe->pipe_buffer.in ==
						size - segsize +
						wpipe->pipe_buffer.size,
						("Expected wraparound bad"));
					wpipe->pipe_buffer.in = size - segsize;
				}

				wpipe->pipe_buffer.cnt += size;
				KASSERT(wpipe->pipe_buffer.cnt <=
					wpipe->pipe_buffer.size,
					("Pipe buffer overflow"));
			}
			pipeunlock(wpipe);
			if (error != 0)
				break;
		} else {
			/*
			 * If the "read-side" has been blocked, wake it up now.
			 */
			if (wpipe->pipe_state & PIPE_WANTR) {
				wpipe->pipe_state &= ~PIPE_WANTR;
				wakeup(wpipe);
			}

			/*
			 * don't block on non-blocking I/O
			 */
			if (fp->f_flag & FNONBLOCK) {
				error = EAGAIN;
				pipeunlock(wpipe);
				break;
			}

			/*
			 * We have no more space and have something to offer,
			 * wake up select/poll.
			 */
			pipeselwakeup(wpipe);

			wpipe->pipe_state |= PIPE_WANTW;
			pipeunlock(wpipe);
			error = msleep(wpipe, PIPE_MTX(rpipe),
			    PRIBIO | PCATCH, "pipewr", 0);
			if (error != 0)
				break;
		}
	}

	pipelock(wpipe, 0);
	--wpipe->pipe_busy;

	if ((wpipe->pipe_busy == 0) && (wpipe->pipe_state & PIPE_WANT)) {
		wpipe->pipe_state &= ~(PIPE_WANT | PIPE_WANTR);
		wakeup(wpipe);
	} else if (wpipe->pipe_buffer.cnt > 0) {
		/*
		 * If we have put any characters in the buffer, we wake up
		 * the reader.
		 */
		if (wpipe->pipe_state & PIPE_WANTR) {
			wpipe->pipe_state &= ~PIPE_WANTR;
			wakeup(wpipe);
		}
	}

	/*
	 * Don't return EPIPE if I/O was successful
	 */
	if ((wpipe->pipe_buffer.cnt == 0) &&
	    (uio->uio_resid == 0) &&
	    (error == EPIPE)) {
		error = 0;
	}

	if (error == 0)
		vfs_timestamp(&wpipe->pipe_mtime);

	/*
	 * We have something to offer,
	 * wake up select/poll.
	 */
	if (wpipe->pipe_buffer.cnt)
		pipeselwakeup(wpipe);

	pipeunlock(wpipe);
	PIPE_UNLOCK(rpipe);
	return (error);
}

/*
 * we implement a very minimal set of ioctls for compatibility with sockets.
 */
static int
pipe_ioctl(fp, cmd, data, active_cred, td)
	struct file *fp;
	u_long cmd;
	void *data;
	struct ucred *active_cred;
	struct thread *td;
{
	struct pipe *mpipe = fp->f_data;
	int error;

	PIPE_LOCK(mpipe);

#ifdef MAC
	error = mac_check_pipe_ioctl(active_cred, mpipe->pipe_pair, cmd, data);
	if (error) {
		PIPE_UNLOCK(mpipe);
		return (error);
	}
#endif

	error = 0;
	switch (cmd) {

	case FIONBIO:
		break;

	case FIOASYNC:
		if (*(int *)data) {
			mpipe->pipe_state |= PIPE_ASYNC;
		} else {
			mpipe->pipe_state &= ~PIPE_ASYNC;
		}
		break;

	case FIONREAD:
		if (mpipe->pipe_state & PIPE_DIRECTW)
			*(int *)data = mpipe->pipe_map.cnt;
		else
			*(int *)data = mpipe->pipe_buffer.cnt;
		break;

	case FIOSETOWN:
		PIPE_UNLOCK(mpipe);
		error = fsetown(*(int *)data, &mpipe->pipe_sigio);
		goto out_unlocked;

	case FIOGETOWN:
		*(int *)data = fgetown(&mpipe->pipe_sigio);
		break;

	/* This is deprecated, FIOSETOWN should be used instead. */
	case TIOCSPGRP:
		PIPE_UNLOCK(mpipe);
		error = fsetown(-(*(int *)data), &mpipe->pipe_sigio);
		goto out_unlocked;

	/* This is deprecated, FIOGETOWN should be used instead. */
	case TIOCGPGRP:
		*(int *)data = -fgetown(&mpipe->pipe_sigio);
		break;

	default:
		error = ENOTTY;
		break;
	}
	PIPE_UNLOCK(mpipe);
out_unlocked:
	return (error);
}

static int
pipe_poll(fp, events, active_cred, td)
	struct file *fp;
	int events;
	struct ucred *active_cred;
	struct thread *td;
{
	struct pipe *rpipe = fp->f_data;
	struct pipe *wpipe;
	int revents = 0;
#ifdef MAC
	int error;
#endif

	wpipe = rpipe->pipe_peer;
	PIPE_LOCK(rpipe);
#ifdef MAC
	error = mac_check_pipe_poll(active_cred, rpipe->pipe_pair);
	if (error)
		goto locked_error;
#endif
	if (events & (POLLIN | POLLRDNORM))
		if ((rpipe->pipe_state & PIPE_DIRECTW) ||
		    (rpipe->pipe_buffer.cnt > 0) ||
		    (rpipe->pipe_state & PIPE_EOF))
			revents |= events & (POLLIN | POLLRDNORM);

	if (events & (POLLOUT | POLLWRNORM))
		if (wpipe->pipe_present != PIPE_ACTIVE ||
		    (wpipe->pipe_state & PIPE_EOF) ||
		    (((wpipe->pipe_state & PIPE_DIRECTW) == 0) &&
		     (wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt) >= PIPE_BUF))
			revents |= events & (POLLOUT | POLLWRNORM);

	if ((rpipe->pipe_state & PIPE_EOF) ||
	    wpipe->pipe_present != PIPE_ACTIVE ||
	    (wpipe->pipe_state & PIPE_EOF))
		revents |= POLLHUP;

	if (revents == 0) {
		if (events & (POLLIN | POLLRDNORM)) {
			selrecord(td, &rpipe->pipe_sel);
			rpipe->pipe_state |= PIPE_SEL;
		}

		if (events & (POLLOUT | POLLWRNORM)) {
			selrecord(td, &wpipe->pipe_sel);
			wpipe->pipe_state |= PIPE_SEL;
		}
	}
#ifdef MAC
locked_error:
#endif
	PIPE_UNLOCK(rpipe);

	return (revents);
}

/*
 * We shouldn't need locks here as we're doing a read and this should
 * be a natural race.
 */
static int
pipe_stat(fp, ub, active_cred, td)
	struct file *fp;
	struct stat *ub;
	struct ucred *active_cred;
	struct thread *td;
{
	struct pipe *pipe = fp->f_data;
#ifdef MAC
	int error;

	PIPE_LOCK(pipe);
	error = mac_check_pipe_stat(active_cred, pipe->pipe_pair);
	PIPE_UNLOCK(pipe);
	if (error)
		return (error);
#endif
	bzero(ub, sizeof(*ub));
	ub->st_mode = S_IFIFO;
	ub->st_blksize = PAGE_SIZE;
	if (pipe->pipe_state & PIPE_DIRECTW)
		ub->st_size = pipe->pipe_map.cnt;
	else
		ub->st_size = pipe->pipe_buffer.cnt;
	ub->st_blocks = (ub->st_size + ub->st_blksize - 1) / ub->st_blksize;
	ub->st_atimespec = pipe->pipe_atime;
	ub->st_mtimespec = pipe->pipe_mtime;
	ub->st_ctimespec = pipe->pipe_ctime;
	ub->st_uid = fp->f_cred->cr_uid;
	ub->st_gid = fp->f_cred->cr_gid;
	/*
	 * Left as 0: st_dev, st_ino, st_nlink, st_rdev, st_flags, st_gen.
	 * XXX (st_dev, st_ino) should be unique.
	 */
	return (0);
}

/* ARGSUSED */
static int
pipe_close(fp, td)
	struct file *fp;
	struct thread *td;
{
	struct pipe *cpipe = fp->f_data;

	fp->f_ops = &badfileops;
	fp->f_data = NULL;
	funsetown(&cpipe->pipe_sigio);
	pipeclose(cpipe);
	return (0);
}

static void
pipe_free_kmem(cpipe)
	struct pipe *cpipe;
{

	KASSERT(!mtx_owned(PIPE_MTX(cpipe)),
	    ("pipe_free_kmem: pipe mutex locked"));

	if (cpipe->pipe_buffer.buffer != NULL) {
		atomic_subtract_int(&amountpipekva, cpipe->pipe_buffer.size);
		vm_map_remove(pipe_map,
		    (vm_offset_t)cpipe->pipe_buffer.buffer,
		    (vm_offset_t)cpipe->pipe_buffer.buffer + cpipe->pipe_buffer.size);
		cpipe->pipe_buffer.buffer = NULL;
	}
#ifndef PIPE_NODIRECT
	{
		cpipe->pipe_map.cnt = 0;
		cpipe->pipe_map.pos = 0;
		cpipe->pipe_map.npages = 0;
	}
#endif
}

/*
 * shutdown the pipe
 */
static void
pipeclose(cpipe)
	struct pipe *cpipe;
{
	struct pipepair *pp;
	struct pipe *ppipe;

	KASSERT(cpipe != NULL, ("pipeclose: cpipe == NULL"));

	PIPE_LOCK(cpipe);
	pipelock(cpipe, 0);
	pp = cpipe->pipe_pair;

	pipeselwakeup(cpipe);

	/*
	 * If the other side is blocked, wake it up saying that
	 * we want to close it down.
	 */
	cpipe->pipe_state |= PIPE_EOF;
	while (cpipe->pipe_busy) {
		wakeup(cpipe);
		cpipe->pipe_state |= PIPE_WANT;
		pipeunlock(cpipe);
		msleep(cpipe, PIPE_MTX(cpipe), PRIBIO, "pipecl", 0);
		pipelock(cpipe, 0);
	}


	/*
	 * Disconnect from peer, if any.
	 */
	ppipe = cpipe->pipe_peer;
	if (ppipe->pipe_present == PIPE_ACTIVE) {
		pipeselwakeup(ppipe);

		ppipe->pipe_state |= PIPE_EOF;
		wakeup(ppipe);
		KNOTE_LOCKED(&ppipe->pipe_sel.si_note, 0);
	}

	/*
	 * Mark this endpoint as free.  Release kmem resources.  We
	 * don't mark this endpoint as unused until we've finished
	 * doing that, or the pipe might disappear out from under
	 * us.
	 */
	PIPE_UNLOCK(cpipe);
	pipe_free_kmem(cpipe);
	PIPE_LOCK(cpipe);
	cpipe->pipe_present = PIPE_CLOSING;
	pipeunlock(cpipe);

	/*
	 * knlist_clear() may sleep dropping the PIPE_MTX. Set the
	 * PIPE_FINALIZED, that allows other end to free the
	 * pipe_pair, only after the knotes are completely dismantled.
	 */
	knlist_clear(&cpipe->pipe_sel.si_note, 1);
	cpipe->pipe_present = PIPE_FINALIZED;
	knlist_destroy(&cpipe->pipe_sel.si_note);

	/*
	 * If both endpoints are now closed, release the memory for the
	 * pipe pair.  If not, unlock.
	 */
	if (ppipe->pipe_present == PIPE_FINALIZED) {
		PIPE_UNLOCK(cpipe);
#ifdef MAC
		mac_destroy_pipe(pp);
#endif
		uma_zfree(pipe_zone, cpipe->pipe_pair);
	} else
		PIPE_UNLOCK(cpipe);
}

/*ARGSUSED*/
static int
pipe_kqfilter(struct file *fp, struct knote *kn)
{
	struct pipe *cpipe;

	cpipe = kn->kn_fp->f_data;
	PIPE_LOCK(cpipe);
	switch (kn->kn_filter) {
	case EVFILT_READ:
		kn->kn_fop = &pipe_rfiltops;
		break;
	case EVFILT_WRITE:
		kn->kn_fop = &pipe_wfiltops;
		if (cpipe->pipe_peer->pipe_present != PIPE_ACTIVE) {
			/* other end of pipe has been closed */
			PIPE_UNLOCK(cpipe);
			return (EPIPE);
		}
		cpipe = cpipe->pipe_peer;
		break;
	default:
		PIPE_UNLOCK(cpipe);
		return (EINVAL);
	}

	knlist_add(&cpipe->pipe_sel.si_note, kn, 1);
	PIPE_UNLOCK(cpipe);
	return (0);
}

static void
filt_pipedetach(struct knote *kn)
{
	struct pipe *cpipe = (struct pipe *)kn->kn_fp->f_data;

	PIPE_LOCK(cpipe);
	if (kn->kn_filter == EVFILT_WRITE)
		cpipe = cpipe->pipe_peer;
	knlist_remove(&cpipe->pipe_sel.si_note, kn, 1);
	PIPE_UNLOCK(cpipe);
}

/*ARGSUSED*/
static int
filt_piperead(struct knote *kn, long hint)
{
	struct pipe *rpipe = kn->kn_fp->f_data;
	struct pipe *wpipe = rpipe->pipe_peer;
	int ret;

	PIPE_LOCK(rpipe);
	kn->kn_data = rpipe->pipe_buffer.cnt;
	if ((kn->kn_data == 0) && (rpipe->pipe_state & PIPE_DIRECTW))
		kn->kn_data = rpipe->pipe_map.cnt;

	if ((rpipe->pipe_state & PIPE_EOF) ||
	    wpipe->pipe_present != PIPE_ACTIVE ||
	    (wpipe->pipe_state & PIPE_EOF)) {
		kn->kn_flags |= EV_EOF;
		PIPE_UNLOCK(rpipe);
		return (1);
	}
	ret = kn->kn_data > 0;
	PIPE_UNLOCK(rpipe);
	return ret;
}

/*ARGSUSED*/
static int
filt_pipewrite(struct knote *kn, long hint)
{
	struct pipe *rpipe = kn->kn_fp->f_data;
	struct pipe *wpipe = rpipe->pipe_peer;

	PIPE_LOCK(rpipe);
	if (wpipe->pipe_present != PIPE_ACTIVE ||
	    (wpipe->pipe_state & PIPE_EOF)) {
		kn->kn_data = 0;
		kn->kn_flags |= EV_EOF;
		PIPE_UNLOCK(rpipe);
		return (1);
	}
	kn->kn_data = wpipe->pipe_buffer.size - wpipe->pipe_buffer.cnt;
	if (wpipe->pipe_state & PIPE_DIRECTW)
		kn->kn_data = 0;

	PIPE_UNLOCK(rpipe);
	return (kn->kn_data >= PIPE_BUF);
}