aboutsummaryrefslogtreecommitdiff
path: root/sys/kern/kern_mbuf.c
blob: f814d3cab6e8b1f1b73aa4ecdc453703427eef12 (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
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
/*-
 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
 *
 * Copyright (c) 2004, 2005,
 *	Bosko Milekic <bmilekic@FreeBSD.org>.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice unmodified, this list of conditions and the following
 *    disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

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

#include "opt_param.h"
#include "opt_kern_tls.h"

#include <sys/param.h>
#include <sys/conf.h>
#include <sys/domainset.h>
#include <sys/malloc.h>
#include <sys/systm.h>
#include <sys/mbuf.h>
#include <sys/domain.h>
#include <sys/eventhandler.h>
#include <sys/kernel.h>
#include <sys/ktls.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/protosw.h>
#include <sys/refcount.h>
#include <sys/sf_buf.h>
#include <sys/smp.h>
#include <sys/socket.h>
#include <sys/sysctl.h>

#include <net/if.h>
#include <net/if_var.h>

#include <vm/vm.h>
#include <vm/vm_extern.h>
#include <vm/vm_kern.h>
#include <vm/vm_page.h>
#include <vm/vm_pageout.h>
#include <vm/vm_map.h>
#include <vm/uma.h>
#include <vm/uma_dbg.h>

/*
 * In FreeBSD, Mbufs and Mbuf Clusters are allocated from UMA
 * Zones.
 *
 * Mbuf Clusters (2K, contiguous) are allocated from the Cluster
 * Zone.  The Zone can be capped at kern.ipc.nmbclusters, if the
 * administrator so desires.
 *
 * Mbufs are allocated from a UMA Primary Zone called the Mbuf
 * Zone.
 *
 * Additionally, FreeBSD provides a Packet Zone, which it
 * configures as a Secondary Zone to the Mbuf Primary Zone,
 * thus sharing backend Slab kegs with the Mbuf Primary Zone.
 *
 * Thus common-case allocations and locking are simplified:
 *
 *  m_clget()                m_getcl()
 *    |                         |
 *    |   .------------>[(Packet Cache)]    m_get(), m_gethdr()
 *    |   |             [     Packet   ]            |
 *  [(Cluster Cache)]   [    Secondary ]   [ (Mbuf Cache)     ]
 *  [ Cluster Zone  ]   [     Zone     ]   [ Mbuf Primary Zone ]
 *        |                       \________         |
 *  [ Cluster Keg   ]                      \       /
 *        |	                         [ Mbuf Keg   ]
 *  [ Cluster Slabs ]                         |
 *        |                              [ Mbuf Slabs ]
 *         \____________(VM)_________________/
 *
 *
 * Whenever an object is allocated with uma_zalloc() out of
 * one of the Zones its _ctor_ function is executed.  The same
 * for any deallocation through uma_zfree() the _dtor_ function
 * is executed.
 *
 * Caches are per-CPU and are filled from the Primary Zone.
 *
 * Whenever an object is allocated from the underlying global
 * memory pool it gets pre-initialized with the _zinit_ functions.
 * When the Keg's are overfull objects get decommissioned with
 * _zfini_ functions and free'd back to the global memory pool.
 *
 */

int nmbufs;			/* limits number of mbufs */
int nmbclusters;		/* limits number of mbuf clusters */
int nmbjumbop;			/* limits number of page size jumbo clusters */
int nmbjumbo9;			/* limits number of 9k jumbo clusters */
int nmbjumbo16;			/* limits number of 16k jumbo clusters */

bool mb_use_ext_pgs = true;	/* use M_EXTPG mbufs for sendfile & TLS */
SYSCTL_BOOL(_kern_ipc, OID_AUTO, mb_use_ext_pgs, CTLFLAG_RWTUN,
    &mb_use_ext_pgs, 0,
    "Use unmapped mbufs for sendfile(2) and TLS offload");

static quad_t maxmbufmem;	/* overall real memory limit for all mbufs */

SYSCTL_QUAD(_kern_ipc, OID_AUTO, maxmbufmem, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &maxmbufmem, 0,
    "Maximum real memory allocatable to various mbuf types");

static counter_u64_t snd_tag_count;
SYSCTL_COUNTER_U64(_kern_ipc, OID_AUTO, num_snd_tags, CTLFLAG_RW,
    &snd_tag_count, "# of active mbuf send tags");

/*
 * tunable_mbinit() has to be run before any mbuf allocations are done.
 */
static void
tunable_mbinit(void *dummy)
{
	quad_t realmem;

	/*
	 * The default limit for all mbuf related memory is 1/2 of all
	 * available kernel memory (physical or kmem).
	 * At most it can be 3/4 of available kernel memory.
	 */
	realmem = qmin((quad_t)physmem * PAGE_SIZE, vm_kmem_size);
	maxmbufmem = realmem / 2;
	TUNABLE_QUAD_FETCH("kern.ipc.maxmbufmem", &maxmbufmem);
	if (maxmbufmem > realmem / 4 * 3)
		maxmbufmem = realmem / 4 * 3;

	TUNABLE_INT_FETCH("kern.ipc.nmbclusters", &nmbclusters);
	if (nmbclusters == 0)
		nmbclusters = maxmbufmem / MCLBYTES / 4;

	TUNABLE_INT_FETCH("kern.ipc.nmbjumbop", &nmbjumbop);
	if (nmbjumbop == 0)
		nmbjumbop = maxmbufmem / MJUMPAGESIZE / 4;

	TUNABLE_INT_FETCH("kern.ipc.nmbjumbo9", &nmbjumbo9);
	if (nmbjumbo9 == 0)
		nmbjumbo9 = maxmbufmem / MJUM9BYTES / 6;

	TUNABLE_INT_FETCH("kern.ipc.nmbjumbo16", &nmbjumbo16);
	if (nmbjumbo16 == 0)
		nmbjumbo16 = maxmbufmem / MJUM16BYTES / 6;

	/*
	 * We need at least as many mbufs as we have clusters of
	 * the various types added together.
	 */
	TUNABLE_INT_FETCH("kern.ipc.nmbufs", &nmbufs);
	if (nmbufs < nmbclusters + nmbjumbop + nmbjumbo9 + nmbjumbo16)
		nmbufs = lmax(maxmbufmem / MSIZE / 5,
		    nmbclusters + nmbjumbop + nmbjumbo9 + nmbjumbo16);
}
SYSINIT(tunable_mbinit, SI_SUB_KMEM, SI_ORDER_MIDDLE, tunable_mbinit, NULL);

static int
sysctl_nmbclusters(SYSCTL_HANDLER_ARGS)
{
	int error, newnmbclusters;

	newnmbclusters = nmbclusters;
	error = sysctl_handle_int(oidp, &newnmbclusters, 0, req);
	if (error == 0 && req->newptr && newnmbclusters != nmbclusters) {
		if (newnmbclusters > nmbclusters &&
		    nmbufs >= nmbclusters + nmbjumbop + nmbjumbo9 + nmbjumbo16) {
			nmbclusters = newnmbclusters;
			nmbclusters = uma_zone_set_max(zone_clust, nmbclusters);
			EVENTHANDLER_INVOKE(nmbclusters_change);
		} else
			error = EINVAL;
	}
	return (error);
}
SYSCTL_PROC(_kern_ipc, OID_AUTO, nmbclusters,
    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &nmbclusters, 0,
    sysctl_nmbclusters, "IU",
    "Maximum number of mbuf clusters allowed");

static int
sysctl_nmbjumbop(SYSCTL_HANDLER_ARGS)
{
	int error, newnmbjumbop;

	newnmbjumbop = nmbjumbop;
	error = sysctl_handle_int(oidp, &newnmbjumbop, 0, req);
	if (error == 0 && req->newptr && newnmbjumbop != nmbjumbop) {
		if (newnmbjumbop > nmbjumbop &&
		    nmbufs >= nmbclusters + nmbjumbop + nmbjumbo9 + nmbjumbo16) {
			nmbjumbop = newnmbjumbop;
			nmbjumbop = uma_zone_set_max(zone_jumbop, nmbjumbop);
		} else
			error = EINVAL;
	}
	return (error);
}
SYSCTL_PROC(_kern_ipc, OID_AUTO, nmbjumbop,
    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &nmbjumbop, 0,
    sysctl_nmbjumbop, "IU",
    "Maximum number of mbuf page size jumbo clusters allowed");

static int
sysctl_nmbjumbo9(SYSCTL_HANDLER_ARGS)
{
	int error, newnmbjumbo9;

	newnmbjumbo9 = nmbjumbo9;
	error = sysctl_handle_int(oidp, &newnmbjumbo9, 0, req);
	if (error == 0 && req->newptr && newnmbjumbo9 != nmbjumbo9) {
		if (newnmbjumbo9 > nmbjumbo9 &&
		    nmbufs >= nmbclusters + nmbjumbop + nmbjumbo9 + nmbjumbo16) {
			nmbjumbo9 = newnmbjumbo9;
			nmbjumbo9 = uma_zone_set_max(zone_jumbo9, nmbjumbo9);
		} else
			error = EINVAL;
	}
	return (error);
}
SYSCTL_PROC(_kern_ipc, OID_AUTO, nmbjumbo9,
    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &nmbjumbo9, 0,
    sysctl_nmbjumbo9, "IU",
    "Maximum number of mbuf 9k jumbo clusters allowed");

static int
sysctl_nmbjumbo16(SYSCTL_HANDLER_ARGS)
{
	int error, newnmbjumbo16;

	newnmbjumbo16 = nmbjumbo16;
	error = sysctl_handle_int(oidp, &newnmbjumbo16, 0, req);
	if (error == 0 && req->newptr && newnmbjumbo16 != nmbjumbo16) {
		if (newnmbjumbo16 > nmbjumbo16 &&
		    nmbufs >= nmbclusters + nmbjumbop + nmbjumbo9 + nmbjumbo16) {
			nmbjumbo16 = newnmbjumbo16;
			nmbjumbo16 = uma_zone_set_max(zone_jumbo16, nmbjumbo16);
		} else
			error = EINVAL;
	}
	return (error);
}
SYSCTL_PROC(_kern_ipc, OID_AUTO, nmbjumbo16,
    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &nmbjumbo16, 0,
    sysctl_nmbjumbo16, "IU",
    "Maximum number of mbuf 16k jumbo clusters allowed");

static int
sysctl_nmbufs(SYSCTL_HANDLER_ARGS)
{
	int error, newnmbufs;

	newnmbufs = nmbufs;
	error = sysctl_handle_int(oidp, &newnmbufs, 0, req);
	if (error == 0 && req->newptr && newnmbufs != nmbufs) {
		if (newnmbufs > nmbufs) {
			nmbufs = newnmbufs;
			nmbufs = uma_zone_set_max(zone_mbuf, nmbufs);
			EVENTHANDLER_INVOKE(nmbufs_change);
		} else
			error = EINVAL;
	}
	return (error);
}
SYSCTL_PROC(_kern_ipc, OID_AUTO, nmbufs,
    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
    &nmbufs, 0, sysctl_nmbufs, "IU",
    "Maximum number of mbufs allowed");

/*
 * Zones from which we allocate.
 */
uma_zone_t	zone_mbuf;
uma_zone_t	zone_clust;
uma_zone_t	zone_pack;
uma_zone_t	zone_jumbop;
uma_zone_t	zone_jumbo9;
uma_zone_t	zone_jumbo16;

/*
 * Local prototypes.
 */
static int	mb_ctor_mbuf(void *, int, void *, int);
static int	mb_ctor_clust(void *, int, void *, int);
static int	mb_ctor_pack(void *, int, void *, int);
static void	mb_dtor_mbuf(void *, int, void *);
static void	mb_dtor_pack(void *, int, void *);
static int	mb_zinit_pack(void *, int, int);
static void	mb_zfini_pack(void *, int);
static void	mb_reclaim(uma_zone_t, int);

/* Ensure that MSIZE is a power of 2. */
CTASSERT((((MSIZE - 1) ^ MSIZE) + 1) >> 1 == MSIZE);

_Static_assert(sizeof(struct mbuf) <= MSIZE,
    "size of mbuf exceeds MSIZE");
/*
 * Initialize FreeBSD Network buffer allocation.
 */
static void
mbuf_init(void *dummy)
{

	/*
	 * Configure UMA zones for Mbufs, Clusters, and Packets.
	 */
	zone_mbuf = uma_zcreate(MBUF_MEM_NAME, MSIZE,
	    mb_ctor_mbuf, mb_dtor_mbuf, NULL, NULL,
	    MSIZE - 1, UMA_ZONE_CONTIG | UMA_ZONE_MAXBUCKET);
	if (nmbufs > 0)
		nmbufs = uma_zone_set_max(zone_mbuf, nmbufs);
	uma_zone_set_warning(zone_mbuf, "kern.ipc.nmbufs limit reached");
	uma_zone_set_maxaction(zone_mbuf, mb_reclaim);

	zone_clust = uma_zcreate(MBUF_CLUSTER_MEM_NAME, MCLBYTES,
	    mb_ctor_clust, NULL, NULL, NULL,
	    UMA_ALIGN_PTR, UMA_ZONE_CONTIG);
	if (nmbclusters > 0)
		nmbclusters = uma_zone_set_max(zone_clust, nmbclusters);
	uma_zone_set_warning(zone_clust, "kern.ipc.nmbclusters limit reached");
	uma_zone_set_maxaction(zone_clust, mb_reclaim);

	zone_pack = uma_zsecond_create(MBUF_PACKET_MEM_NAME, mb_ctor_pack,
	    mb_dtor_pack, mb_zinit_pack, mb_zfini_pack, zone_mbuf);

	/* Make jumbo frame zone too. Page size, 9k and 16k. */
	zone_jumbop = uma_zcreate(MBUF_JUMBOP_MEM_NAME, MJUMPAGESIZE,
	    mb_ctor_clust, NULL, NULL, NULL,
	    UMA_ALIGN_PTR, UMA_ZONE_CONTIG);
	if (nmbjumbop > 0)
		nmbjumbop = uma_zone_set_max(zone_jumbop, nmbjumbop);
	uma_zone_set_warning(zone_jumbop, "kern.ipc.nmbjumbop limit reached");
	uma_zone_set_maxaction(zone_jumbop, mb_reclaim);

	zone_jumbo9 = uma_zcreate(MBUF_JUMBO9_MEM_NAME, MJUM9BYTES,
	    mb_ctor_clust, NULL, NULL, NULL,
	    UMA_ALIGN_PTR, UMA_ZONE_CONTIG);
	if (nmbjumbo9 > 0)
		nmbjumbo9 = uma_zone_set_max(zone_jumbo9, nmbjumbo9);
	uma_zone_set_warning(zone_jumbo9, "kern.ipc.nmbjumbo9 limit reached");
	uma_zone_set_maxaction(zone_jumbo9, mb_reclaim);

	zone_jumbo16 = uma_zcreate(MBUF_JUMBO16_MEM_NAME, MJUM16BYTES,
	    mb_ctor_clust, NULL, NULL, NULL,
	    UMA_ALIGN_PTR, UMA_ZONE_CONTIG);
	if (nmbjumbo16 > 0)
		nmbjumbo16 = uma_zone_set_max(zone_jumbo16, nmbjumbo16);
	uma_zone_set_warning(zone_jumbo16, "kern.ipc.nmbjumbo16 limit reached");
	uma_zone_set_maxaction(zone_jumbo16, mb_reclaim);

	/*
	 * Hook event handler for low-memory situation, used to
	 * drain protocols and push data back to the caches (UMA
	 * later pushes it back to VM).
	 */
	EVENTHANDLER_REGISTER(vm_lowmem, mb_reclaim, NULL,
	    EVENTHANDLER_PRI_FIRST);

	snd_tag_count = counter_u64_alloc(M_WAITOK);
}
SYSINIT(mbuf, SI_SUB_MBUF, SI_ORDER_FIRST, mbuf_init, NULL);

#ifdef DEBUGNET
/*
 * debugnet makes use of a pre-allocated pool of mbufs and clusters.  When
 * debugnet is configured, we initialize a set of UMA cache zones which return
 * items from this pool.  At panic-time, the regular UMA zone pointers are
 * overwritten with those of the cache zones so that drivers may allocate and
 * free mbufs and clusters without attempting to allocate physical memory.
 *
 * We keep mbufs and clusters in a pair of mbuf queues.  In particular, for
 * the purpose of caching clusters, we treat them as mbufs.
 */
static struct mbufq dn_mbufq =
    { STAILQ_HEAD_INITIALIZER(dn_mbufq.mq_head), 0, INT_MAX };
static struct mbufq dn_clustq =
    { STAILQ_HEAD_INITIALIZER(dn_clustq.mq_head), 0, INT_MAX };

static int dn_clsize;
static uma_zone_t dn_zone_mbuf;
static uma_zone_t dn_zone_clust;
static uma_zone_t dn_zone_pack;

static struct debugnet_saved_zones {
	uma_zone_t dsz_mbuf;
	uma_zone_t dsz_clust;
	uma_zone_t dsz_pack;
	uma_zone_t dsz_jumbop;
	uma_zone_t dsz_jumbo9;
	uma_zone_t dsz_jumbo16;
	bool dsz_debugnet_zones_enabled;
} dn_saved_zones;

static int
dn_buf_import(void *arg, void **store, int count, int domain __unused,
    int flags)
{
	struct mbufq *q;
	struct mbuf *m;
	int i;

	q = arg;

	for (i = 0; i < count; i++) {
		m = mbufq_dequeue(q);
		if (m == NULL)
			break;
		trash_init(m, q == &dn_mbufq ? MSIZE : dn_clsize, flags);
		store[i] = m;
	}
	KASSERT((flags & M_WAITOK) == 0 || i == count,
	    ("%s: ran out of pre-allocated mbufs", __func__));
	return (i);
}

static void
dn_buf_release(void *arg, void **store, int count)
{
	struct mbufq *q;
	struct mbuf *m;
	int i;

	q = arg;

	for (i = 0; i < count; i++) {
		m = store[i];
		(void)mbufq_enqueue(q, m);
	}
}

static int
dn_pack_import(void *arg __unused, void **store, int count, int domain __unused,
    int flags __unused)
{
	struct mbuf *m;
	void *clust;
	int i;

	for (i = 0; i < count; i++) {
		m = m_get(MT_DATA, M_NOWAIT);
		if (m == NULL)
			break;
		clust = uma_zalloc(dn_zone_clust, M_NOWAIT);
		if (clust == NULL) {
			m_free(m);
			break;
		}
		mb_ctor_clust(clust, dn_clsize, m, 0);
		store[i] = m;
	}
	KASSERT((flags & M_WAITOK) == 0 || i == count,
	    ("%s: ran out of pre-allocated mbufs", __func__));
	return (i);
}

static void
dn_pack_release(void *arg __unused, void **store, int count)
{
	struct mbuf *m;
	void *clust;
	int i;

	for (i = 0; i < count; i++) {
		m = store[i];
		clust = m->m_ext.ext_buf;
		uma_zfree(dn_zone_clust, clust);
		uma_zfree(dn_zone_mbuf, m);
	}
}

/*
 * Free the pre-allocated mbufs and clusters reserved for debugnet, and destroy
 * the corresponding UMA cache zones.
 */
void
debugnet_mbuf_drain(void)
{
	struct mbuf *m;
	void *item;

	if (dn_zone_mbuf != NULL) {
		uma_zdestroy(dn_zone_mbuf);
		dn_zone_mbuf = NULL;
	}
	if (dn_zone_clust != NULL) {
		uma_zdestroy(dn_zone_clust);
		dn_zone_clust = NULL;
	}
	if (dn_zone_pack != NULL) {
		uma_zdestroy(dn_zone_pack);
		dn_zone_pack = NULL;
	}

	while ((m = mbufq_dequeue(&dn_mbufq)) != NULL)
		m_free(m);
	while ((item = mbufq_dequeue(&dn_clustq)) != NULL)
		uma_zfree(m_getzone(dn_clsize), item);
}

/*
 * Callback invoked immediately prior to starting a debugnet connection.
 */
void
debugnet_mbuf_start(void)
{

	MPASS(!dn_saved_zones.dsz_debugnet_zones_enabled);

	/* Save the old zone pointers to restore when debugnet is closed. */
	dn_saved_zones = (struct debugnet_saved_zones) {
		.dsz_debugnet_zones_enabled = true,
		.dsz_mbuf = zone_mbuf,
		.dsz_clust = zone_clust,
		.dsz_pack = zone_pack,
		.dsz_jumbop = zone_jumbop,
		.dsz_jumbo9 = zone_jumbo9,
		.dsz_jumbo16 = zone_jumbo16,
	};

	/*
	 * All cluster zones return buffers of the size requested by the
	 * drivers.  It's up to the driver to reinitialize the zones if the
	 * MTU of a debugnet-enabled interface changes.
	 */
	printf("debugnet: overwriting mbuf zone pointers\n");
	zone_mbuf = dn_zone_mbuf;
	zone_clust = dn_zone_clust;
	zone_pack = dn_zone_pack;
	zone_jumbop = dn_zone_clust;
	zone_jumbo9 = dn_zone_clust;
	zone_jumbo16 = dn_zone_clust;
}

/*
 * Callback invoked when a debugnet connection is closed/finished.
 */
void
debugnet_mbuf_finish(void)
{

	MPASS(dn_saved_zones.dsz_debugnet_zones_enabled);

	printf("debugnet: restoring mbuf zone pointers\n");
	zone_mbuf = dn_saved_zones.dsz_mbuf;
	zone_clust = dn_saved_zones.dsz_clust;
	zone_pack = dn_saved_zones.dsz_pack;
	zone_jumbop = dn_saved_zones.dsz_jumbop;
	zone_jumbo9 = dn_saved_zones.dsz_jumbo9;
	zone_jumbo16 = dn_saved_zones.dsz_jumbo16;

	memset(&dn_saved_zones, 0, sizeof(dn_saved_zones));
}

/*
 * Reinitialize the debugnet mbuf+cluster pool and cache zones.
 */
void
debugnet_mbuf_reinit(int nmbuf, int nclust, int clsize)
{
	struct mbuf *m;
	void *item;

	debugnet_mbuf_drain();

	dn_clsize = clsize;

	dn_zone_mbuf = uma_zcache_create("debugnet_" MBUF_MEM_NAME,
	    MSIZE, mb_ctor_mbuf, mb_dtor_mbuf, NULL, NULL,
	    dn_buf_import, dn_buf_release,
	    &dn_mbufq, UMA_ZONE_NOBUCKET);

	dn_zone_clust = uma_zcache_create("debugnet_" MBUF_CLUSTER_MEM_NAME,
	    clsize, mb_ctor_clust, NULL, NULL, NULL,
	    dn_buf_import, dn_buf_release,
	    &dn_clustq, UMA_ZONE_NOBUCKET);

	dn_zone_pack = uma_zcache_create("debugnet_" MBUF_PACKET_MEM_NAME,
	    MCLBYTES, mb_ctor_pack, mb_dtor_pack, NULL, NULL,
	    dn_pack_import, dn_pack_release,
	    NULL, UMA_ZONE_NOBUCKET);

	while (nmbuf-- > 0) {
		m = m_get(MT_DATA, M_WAITOK);
		uma_zfree(dn_zone_mbuf, m);
	}
	while (nclust-- > 0) {
		item = uma_zalloc(m_getzone(dn_clsize), M_WAITOK);
		uma_zfree(dn_zone_clust, item);
	}
}
#endif /* DEBUGNET */

/*
 * Constructor for Mbuf primary zone.
 *
 * The 'arg' pointer points to a mb_args structure which
 * contains call-specific information required to support the
 * mbuf allocation API.  See mbuf.h.
 */
static int
mb_ctor_mbuf(void *mem, int size, void *arg, int how)
{
	struct mbuf *m;
	struct mb_args *args;
	int error;
	int flags;
	short type;

	args = (struct mb_args *)arg;
	type = args->type;

	/*
	 * The mbuf is initialized later.  The caller has the
	 * responsibility to set up any MAC labels too.
	 */
	if (type == MT_NOINIT)
		return (0);

	m = (struct mbuf *)mem;
	flags = args->flags;
	MPASS((flags & M_NOFREE) == 0);

	error = m_init(m, how, type, flags);

	return (error);
}

/*
 * The Mbuf primary zone destructor.
 */
static void
mb_dtor_mbuf(void *mem, int size, void *arg)
{
	struct mbuf *m;
	unsigned long flags;

	m = (struct mbuf *)mem;
	flags = (unsigned long)arg;

	KASSERT((m->m_flags & M_NOFREE) == 0, ("%s: M_NOFREE set", __func__));
	KASSERT((flags & 0x1) == 0, ("%s: obsolete MB_DTOR_SKIP passed", __func__));
	if ((m->m_flags & M_PKTHDR) && !SLIST_EMPTY(&m->m_pkthdr.tags))
		m_tag_delete_chain(m, NULL);
}

/*
 * The Mbuf Packet zone destructor.
 */
static void
mb_dtor_pack(void *mem, int size, void *arg)
{
	struct mbuf *m;

	m = (struct mbuf *)mem;
	if ((m->m_flags & M_PKTHDR) != 0)
		m_tag_delete_chain(m, NULL);

	/* Make sure we've got a clean cluster back. */
	KASSERT((m->m_flags & M_EXT) == M_EXT, ("%s: M_EXT not set", __func__));
	KASSERT(m->m_ext.ext_buf != NULL, ("%s: ext_buf == NULL", __func__));
	KASSERT(m->m_ext.ext_free == NULL, ("%s: ext_free != NULL", __func__));
	KASSERT(m->m_ext.ext_arg1 == NULL, ("%s: ext_arg1 != NULL", __func__));
	KASSERT(m->m_ext.ext_arg2 == NULL, ("%s: ext_arg2 != NULL", __func__));
	KASSERT(m->m_ext.ext_size == MCLBYTES, ("%s: ext_size != MCLBYTES", __func__));
	KASSERT(m->m_ext.ext_type == EXT_PACKET, ("%s: ext_type != EXT_PACKET", __func__));
#ifdef INVARIANTS
	trash_dtor(m->m_ext.ext_buf, MCLBYTES, arg);
#endif
	/*
	 * If there are processes blocked on zone_clust, waiting for pages
	 * to be freed up, cause them to be woken up by draining the
	 * packet zone.  We are exposed to a race here (in the check for
	 * the UMA_ZFLAG_FULL) where we might miss the flag set, but that
	 * is deliberate. We don't want to acquire the zone lock for every
	 * mbuf free.
	 */
	if (uma_zone_exhausted(zone_clust))
		uma_zone_reclaim(zone_pack, UMA_RECLAIM_DRAIN);
}

/*
 * The Cluster and Jumbo[PAGESIZE|9|16] zone constructor.
 *
 * Here the 'arg' pointer points to the Mbuf which we
 * are configuring cluster storage for.  If 'arg' is
 * empty we allocate just the cluster without setting
 * the mbuf to it.  See mbuf.h.
 */
static int
mb_ctor_clust(void *mem, int size, void *arg, int how)
{
	struct mbuf *m;

	m = (struct mbuf *)arg;
	if (m != NULL) {
		m->m_ext.ext_buf = (char *)mem;
		m->m_data = m->m_ext.ext_buf;
		m->m_flags |= M_EXT;
		m->m_ext.ext_free = NULL;
		m->m_ext.ext_arg1 = NULL;
		m->m_ext.ext_arg2 = NULL;
		m->m_ext.ext_size = size;
		m->m_ext.ext_type = m_gettype(size);
		m->m_ext.ext_flags = EXT_FLAG_EMBREF;
		m->m_ext.ext_count = 1;
	}

	return (0);
}

/*
 * The Packet secondary zone's init routine, executed on the
 * object's transition from mbuf keg slab to zone cache.
 */
static int
mb_zinit_pack(void *mem, int size, int how)
{
	struct mbuf *m;

	m = (struct mbuf *)mem;		/* m is virgin. */
	if (uma_zalloc_arg(zone_clust, m, how) == NULL ||
	    m->m_ext.ext_buf == NULL)
		return (ENOMEM);
	m->m_ext.ext_type = EXT_PACKET;	/* Override. */
#ifdef INVARIANTS
	trash_init(m->m_ext.ext_buf, MCLBYTES, how);
#endif
	return (0);
}

/*
 * The Packet secondary zone's fini routine, executed on the
 * object's transition from zone cache to keg slab.
 */
static void
mb_zfini_pack(void *mem, int size)
{
	struct mbuf *m;

	m = (struct mbuf *)mem;
#ifdef INVARIANTS
	trash_fini(m->m_ext.ext_buf, MCLBYTES);
#endif
	uma_zfree_arg(zone_clust, m->m_ext.ext_buf, NULL);
#ifdef INVARIANTS
	trash_dtor(mem, size, NULL);
#endif
}

/*
 * The "packet" keg constructor.
 */
static int
mb_ctor_pack(void *mem, int size, void *arg, int how)
{
	struct mbuf *m;
	struct mb_args *args;
	int error, flags;
	short type;

	m = (struct mbuf *)mem;
	args = (struct mb_args *)arg;
	flags = args->flags;
	type = args->type;
	MPASS((flags & M_NOFREE) == 0);

#ifdef INVARIANTS
	trash_ctor(m->m_ext.ext_buf, MCLBYTES, arg, how);
#endif

	error = m_init(m, how, type, flags);

	/* m_ext is already initialized. */
	m->m_data = m->m_ext.ext_buf;
 	m->m_flags = (flags | M_EXT);

	return (error);
}

/*
 * This is the protocol drain routine.  Called by UMA whenever any of the
 * mbuf zones is closed to its limit.
 *
 * No locks should be held when this is called.  The drain routines have to
 * presently acquire some locks which raises the possibility of lock order
 * reversal.
 */
static void
mb_reclaim(uma_zone_t zone __unused, int pending __unused)
{
	struct epoch_tracker et;
	struct domain *dp;
	struct protosw *pr;

	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK | WARN_PANIC, NULL, __func__);

	NET_EPOCH_ENTER(et);
	for (dp = domains; dp != NULL; dp = dp->dom_next)
		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++)
			if (pr->pr_drain != NULL)
				(*pr->pr_drain)();
	NET_EPOCH_EXIT(et);
}

/*
 * Free "count" units of I/O from an mbuf chain.  They could be held
 * in M_EXTPG or just as a normal mbuf.  This code is intended to be
 * called in an error path (I/O error, closed connection, etc).
 */
void
mb_free_notready(struct mbuf *m, int count)
{
	int i;

	for (i = 0; i < count && m != NULL; i++) {
		if ((m->m_flags & M_EXTPG) != 0) {
			m->m_epg_nrdy--;
			if (m->m_epg_nrdy != 0)
				continue;
		}
		m = m_free(m);
	}
	KASSERT(i == count, ("Removed only %d items from %p", i, m));
}

/*
 * Compress an unmapped mbuf into a simple mbuf when it holds a small
 * amount of data.  This is used as a DOS defense to avoid having
 * small packets tie up wired pages, an ext_pgs structure, and an
 * mbuf.  Since this converts the existing mbuf in place, it can only
 * be used if there are no other references to 'm'.
 */
int
mb_unmapped_compress(struct mbuf *m)
{
	volatile u_int *refcnt;
	char buf[MLEN];

	/*
	 * Assert that 'm' does not have a packet header.  If 'm' had
	 * a packet header, it would only be able to hold MHLEN bytes
	 * and m_data would have to be initialized differently.
	 */
	KASSERT((m->m_flags & M_PKTHDR) == 0 && (m->m_flags & M_EXTPG),
            ("%s: m %p !M_EXTPG or M_PKTHDR", __func__, m));
	KASSERT(m->m_len <= MLEN, ("m_len too large %p", m));

	if (m->m_ext.ext_flags & EXT_FLAG_EMBREF) {
		refcnt = &m->m_ext.ext_count;
	} else {
		KASSERT(m->m_ext.ext_cnt != NULL,
		    ("%s: no refcounting pointer on %p", __func__, m));
		refcnt = m->m_ext.ext_cnt;
	}

	if (*refcnt != 1)
		return (EBUSY);

	m_copydata(m, 0, m->m_len, buf);

	/* Free the backing pages. */
	m->m_ext.ext_free(m);

	/* Turn 'm' into a "normal" mbuf. */
	m->m_flags &= ~(M_EXT | M_RDONLY | M_EXTPG);
	m->m_data = m->m_dat;

	/* Copy data back into m. */
	bcopy(buf, mtod(m, char *), m->m_len);

	return (0);
}

/*
 * These next few routines are used to permit downgrading an unmapped
 * mbuf to a chain of mapped mbufs.  This is used when an interface
 * doesn't supported unmapped mbufs or if checksums need to be
 * computed in software.
 *
 * Each unmapped mbuf is converted to a chain of mbufs.  First, any
 * TLS header data is stored in a regular mbuf.  Second, each page of
 * unmapped data is stored in an mbuf with an EXT_SFBUF external
 * cluster.  These mbufs use an sf_buf to provide a valid KVA for the
 * associated physical page.  They also hold a reference on the
 * original M_EXTPG mbuf to ensure the physical page doesn't go away.
 * Finally, any TLS trailer data is stored in a regular mbuf.
 *
 * mb_unmapped_free_mext() is the ext_free handler for the EXT_SFBUF
 * mbufs.  It frees the associated sf_buf and releases its reference
 * on the original M_EXTPG mbuf.
 *
 * _mb_unmapped_to_ext() is a helper function that converts a single
 * unmapped mbuf into a chain of mbufs.
 *
 * mb_unmapped_to_ext() is the public function that walks an mbuf
 * chain converting any unmapped mbufs to mapped mbufs.  It returns
 * the new chain of unmapped mbufs on success.  On failure it frees
 * the original mbuf chain and returns NULL.
 */
static void
mb_unmapped_free_mext(struct mbuf *m)
{
	struct sf_buf *sf;
	struct mbuf *old_m;

	sf = m->m_ext.ext_arg1;
	sf_buf_free(sf);

	/* Drop the reference on the backing M_EXTPG mbuf. */
	old_m = m->m_ext.ext_arg2;
	mb_free_extpg(old_m);
}

static struct mbuf *
_mb_unmapped_to_ext(struct mbuf *m)
{
	struct mbuf *m_new, *top, *prev, *mref;
	struct sf_buf *sf;
	vm_page_t pg;
	int i, len, off, pglen, pgoff, seglen, segoff;
	volatile u_int *refcnt;
	u_int ref_inc = 0;

	M_ASSERTEXTPG(m);
	len = m->m_len;
	KASSERT(m->m_epg_tls == NULL, ("%s: can't convert TLS mbuf %p",
	    __func__, m));

	/* See if this is the mbuf that holds the embedded refcount. */
	if (m->m_ext.ext_flags & EXT_FLAG_EMBREF) {
		refcnt = &m->m_ext.ext_count;
		mref = m;
	} else {
		KASSERT(m->m_ext.ext_cnt != NULL,
		    ("%s: no refcounting pointer on %p", __func__, m));
		refcnt = m->m_ext.ext_cnt;
		mref = __containerof(refcnt, struct mbuf, m_ext.ext_count);
	}

	/* Skip over any data removed from the front. */
	off = mtod(m, vm_offset_t);

	top = NULL;
	if (m->m_epg_hdrlen != 0) {
		if (off >= m->m_epg_hdrlen) {
			off -= m->m_epg_hdrlen;
		} else {
			seglen = m->m_epg_hdrlen - off;
			segoff = off;
			seglen = min(seglen, len);
			off = 0;
			len -= seglen;
			m_new = m_get(M_NOWAIT, MT_DATA);
			if (m_new == NULL)
				goto fail;
			m_new->m_len = seglen;
			prev = top = m_new;
			memcpy(mtod(m_new, void *), &m->m_epg_hdr[segoff],
			    seglen);
		}
	}
	pgoff = m->m_epg_1st_off;
	for (i = 0; i < m->m_epg_npgs && len > 0; i++) {
		pglen = m_epg_pagelen(m, i, pgoff);
		if (off >= pglen) {
			off -= pglen;
			pgoff = 0;
			continue;
		}
		seglen = pglen - off;
		segoff = pgoff + off;
		off = 0;
		seglen = min(seglen, len);
		len -= seglen;

		pg = PHYS_TO_VM_PAGE(m->m_epg_pa[i]);
		m_new = m_get(M_NOWAIT, MT_DATA);
		if (m_new == NULL)
			goto fail;
		if (top == NULL) {
			top = prev = m_new;
		} else {
			prev->m_next = m_new;
			prev = m_new;
		}
		sf = sf_buf_alloc(pg, SFB_NOWAIT);
		if (sf == NULL)
			goto fail;

		ref_inc++;
		m_extadd(m_new, (char *)sf_buf_kva(sf), PAGE_SIZE,
		    mb_unmapped_free_mext, sf, mref, M_RDONLY, EXT_SFBUF);
		m_new->m_data += segoff;
		m_new->m_len = seglen;

		pgoff = 0;
	};
	if (len != 0) {
		KASSERT((off + len) <= m->m_epg_trllen,
		    ("off + len > trail (%d + %d > %d)", off, len,
		    m->m_epg_trllen));
		m_new = m_get(M_NOWAIT, MT_DATA);
		if (m_new == NULL)
			goto fail;
		if (top == NULL)
			top = m_new;
		else
			prev->m_next = m_new;
		m_new->m_len = len;
		memcpy(mtod(m_new, void *), &m->m_epg_trail[off], len);
	}

	if (ref_inc != 0) {
		/*
		 * Obtain an additional reference on the old mbuf for
		 * each created EXT_SFBUF mbuf.  They will be dropped
		 * in mb_unmapped_free_mext().
		 */
		if (*refcnt == 1)
			*refcnt += ref_inc;
		else
			atomic_add_int(refcnt, ref_inc);
	}
	m_free(m);
	return (top);

fail:
	if (ref_inc != 0) {
		/*
		 * Obtain an additional reference on the old mbuf for
		 * each created EXT_SFBUF mbuf.  They will be
		 * immediately dropped when these mbufs are freed
		 * below.
		 */
		if (*refcnt == 1)
			*refcnt += ref_inc;
		else
			atomic_add_int(refcnt, ref_inc);
	}
	m_free(m);
	m_freem(top);
	return (NULL);
}

struct mbuf *
mb_unmapped_to_ext(struct mbuf *top)
{
	struct mbuf *m, *next, *prev = NULL;

	prev = NULL;
	for (m = top; m != NULL; m = next) {
		/* m might be freed, so cache the next pointer. */
		next = m->m_next;
		if (m->m_flags & M_EXTPG) {
			if (prev != NULL) {
				/*
				 * Remove 'm' from the new chain so
				 * that the 'top' chain terminates
				 * before 'm' in case 'top' is freed
				 * due to an error.
				 */
				prev->m_next = NULL;
			}
			m = _mb_unmapped_to_ext(m);
			if (m == NULL) {
				m_freem(top);
				m_freem(next);
				return (NULL);
			}
			if (prev == NULL) {
				top = m;
			} else {
				prev->m_next = m;
			}

			/*
			 * Replaced one mbuf with a chain, so we must
			 * find the end of chain.
			 */
			prev = m_last(m);
		} else {
			if (prev != NULL) {
				prev->m_next = m;
			}
			prev = m;
		}
	}
	return (top);
}

/*
 * Allocate an empty M_EXTPG mbuf.  The ext_free routine is
 * responsible for freeing any pages backing this mbuf when it is
 * freed.
 */
struct mbuf *
mb_alloc_ext_pgs(int how, m_ext_free_t ext_free)
{
	struct mbuf *m;

	m = m_get(how, MT_DATA);
	if (m == NULL)
		return (NULL);

	m->m_epg_npgs = 0;
	m->m_epg_nrdy = 0;
	m->m_epg_1st_off = 0;
	m->m_epg_last_len = 0;
	m->m_epg_flags = 0;
	m->m_epg_hdrlen = 0;
	m->m_epg_trllen = 0;
	m->m_epg_tls = NULL;
	m->m_epg_so = NULL;
	m->m_data = NULL;
	m->m_flags |= (M_EXT | M_RDONLY | M_EXTPG);
	m->m_ext.ext_flags = EXT_FLAG_EMBREF;
	m->m_ext.ext_count = 1;
	m->m_ext.ext_size = 0;
	m->m_ext.ext_free = ext_free;
	return (m);
}

/*
 * Clean up after mbufs with M_EXT storage attached to them if the
 * reference count hits 1.
 */
void
mb_free_ext(struct mbuf *m)
{
	volatile u_int *refcnt;
	struct mbuf *mref;
	int freembuf;

	KASSERT(m->m_flags & M_EXT, ("%s: M_EXT not set on %p", __func__, m));

	/* See if this is the mbuf that holds the embedded refcount. */
	if (m->m_ext.ext_flags & EXT_FLAG_EMBREF) {
		refcnt = &m->m_ext.ext_count;
		mref = m;
	} else {
		KASSERT(m->m_ext.ext_cnt != NULL,
		    ("%s: no refcounting pointer on %p", __func__, m));
		refcnt = m->m_ext.ext_cnt;
		mref = __containerof(refcnt, struct mbuf, m_ext.ext_count);
	}

	/*
	 * Check if the header is embedded in the cluster.  It is
	 * important that we can't touch any of the mbuf fields
	 * after we have freed the external storage, since mbuf
	 * could have been embedded in it.  For now, the mbufs
	 * embedded into the cluster are always of type EXT_EXTREF,
	 * and for this type we won't free the mref.
	 */
	if (m->m_flags & M_NOFREE) {
		freembuf = 0;
		KASSERT(m->m_ext.ext_type == EXT_EXTREF ||
		    m->m_ext.ext_type == EXT_RXRING,
		    ("%s: no-free mbuf %p has wrong type", __func__, m));
	} else
		freembuf = 1;

	/* Free attached storage if this mbuf is the only reference to it. */
	if (*refcnt == 1 || atomic_fetchadd_int(refcnt, -1) == 1) {
		switch (m->m_ext.ext_type) {
		case EXT_PACKET:
			/* The packet zone is special. */
			if (*refcnt == 0)
				*refcnt = 1;
			uma_zfree(zone_pack, mref);
			break;
		case EXT_CLUSTER:
			uma_zfree(zone_clust, m->m_ext.ext_buf);
			uma_zfree(zone_mbuf, mref);
			break;
		case EXT_JUMBOP:
			uma_zfree(zone_jumbop, m->m_ext.ext_buf);
			uma_zfree(zone_mbuf, mref);
			break;
		case EXT_JUMBO9:
			uma_zfree(zone_jumbo9, m->m_ext.ext_buf);
			uma_zfree(zone_mbuf, mref);
			break;
		case EXT_JUMBO16:
			uma_zfree(zone_jumbo16, m->m_ext.ext_buf);
			uma_zfree(zone_mbuf, mref);
			break;
		case EXT_SFBUF:
		case EXT_NET_DRV:
		case EXT_MOD_TYPE:
		case EXT_DISPOSABLE:
			KASSERT(mref->m_ext.ext_free != NULL,
			    ("%s: ext_free not set", __func__));
			mref->m_ext.ext_free(mref);
			uma_zfree(zone_mbuf, mref);
			break;
		case EXT_EXTREF:
			KASSERT(m->m_ext.ext_free != NULL,
			    ("%s: ext_free not set", __func__));
			m->m_ext.ext_free(m);
			break;
		case EXT_RXRING:
			KASSERT(m->m_ext.ext_free == NULL,
			    ("%s: ext_free is set", __func__));
			break;
		default:
			KASSERT(m->m_ext.ext_type == 0,
			    ("%s: unknown ext_type", __func__));
		}
	}

	if (freembuf && m != mref)
		uma_zfree(zone_mbuf, m);
}

/*
 * Clean up after mbufs with M_EXTPG storage attached to them if the
 * reference count hits 1.
 */
void
mb_free_extpg(struct mbuf *m)
{
	volatile u_int *refcnt;
	struct mbuf *mref;

	M_ASSERTEXTPG(m);

	/* See if this is the mbuf that holds the embedded refcount. */
	if (m->m_ext.ext_flags & EXT_FLAG_EMBREF) {
		refcnt = &m->m_ext.ext_count;
		mref = m;
	} else {
		KASSERT(m->m_ext.ext_cnt != NULL,
		    ("%s: no refcounting pointer on %p", __func__, m));
		refcnt = m->m_ext.ext_cnt;
		mref = __containerof(refcnt, struct mbuf, m_ext.ext_count);
	}

	/* Free attached storage if this mbuf is the only reference to it. */
	if (*refcnt == 1 || atomic_fetchadd_int(refcnt, -1) == 1) {
		KASSERT(mref->m_ext.ext_free != NULL,
		    ("%s: ext_free not set", __func__));

		mref->m_ext.ext_free(mref);
#ifdef KERN_TLS
		if (mref->m_epg_tls != NULL &&
		    !refcount_release_if_not_last(&mref->m_epg_tls->refcount))
			ktls_enqueue_to_free(mref);
		else
#endif
			uma_zfree(zone_mbuf, mref);
	}

	if (m != mref)
		uma_zfree(zone_mbuf, m);
}

/*
 * Official mbuf(9) allocation KPI for stack and drivers:
 *
 * m_get()	- a single mbuf without any attachments, sys/mbuf.h.
 * m_gethdr()	- a single mbuf initialized as M_PKTHDR, sys/mbuf.h.
 * m_getcl()	- an mbuf + 2k cluster, sys/mbuf.h.
 * m_clget()	- attach cluster to already allocated mbuf.
 * m_cljget()	- attach jumbo cluster to already allocated mbuf.
 * m_get2()	- allocate minimum mbuf that would fit size argument.
 * m_getm2()	- allocate a chain of mbufs/clusters.
 * m_extadd()	- attach external cluster to mbuf.
 *
 * m_free()	- free single mbuf with its tags and ext, sys/mbuf.h.
 * m_freem()	- free chain of mbufs.
 */

int
m_clget(struct mbuf *m, int how)
{

	KASSERT((m->m_flags & M_EXT) == 0, ("%s: mbuf %p has M_EXT",
	    __func__, m));
	m->m_ext.ext_buf = (char *)NULL;
	uma_zalloc_arg(zone_clust, m, how);
	/*
	 * On a cluster allocation failure, drain the packet zone and retry,
	 * we might be able to loosen a few clusters up on the drain.
	 */
	if ((how & M_NOWAIT) && (m->m_ext.ext_buf == NULL)) {
		uma_zone_reclaim(zone_pack, UMA_RECLAIM_DRAIN);
		uma_zalloc_arg(zone_clust, m, how);
	}
	MBUF_PROBE2(m__clget, m, how);
	return (m->m_flags & M_EXT);
}

/*
 * m_cljget() is different from m_clget() as it can allocate clusters without
 * attaching them to an mbuf.  In that case the return value is the pointer
 * to the cluster of the requested size.  If an mbuf was specified, it gets
 * the cluster attached to it and the return value can be safely ignored.
 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
 */
void *
m_cljget(struct mbuf *m, int how, int size)
{
	uma_zone_t zone;
	void *retval;

	if (m != NULL) {
		KASSERT((m->m_flags & M_EXT) == 0, ("%s: mbuf %p has M_EXT",
		    __func__, m));
		m->m_ext.ext_buf = NULL;
	}

	zone = m_getzone(size);
	retval = uma_zalloc_arg(zone, m, how);

	MBUF_PROBE4(m__cljget, m, how, size, retval);

	return (retval);
}

/*
 * m_get2() allocates minimum mbuf that would fit "size" argument.
 */
struct mbuf *
m_get2(int size, int how, short type, int flags)
{
	struct mb_args args;
	struct mbuf *m, *n;

	args.flags = flags;
	args.type = type;

	if (size <= MHLEN || (size <= MLEN && (flags & M_PKTHDR) == 0))
		return (uma_zalloc_arg(zone_mbuf, &args, how));
	if (size <= MCLBYTES)
		return (uma_zalloc_arg(zone_pack, &args, how));

	if (size > MJUMPAGESIZE)
		return (NULL);

	m = uma_zalloc_arg(zone_mbuf, &args, how);
	if (m == NULL)
		return (NULL);

	n = uma_zalloc_arg(zone_jumbop, m, how);
	if (n == NULL) {
		uma_zfree(zone_mbuf, m);
		return (NULL);
	}

	return (m);
}

/*
 * m_getjcl() returns an mbuf with a cluster of the specified size attached.
 * For size it takes MCLBYTES, MJUMPAGESIZE, MJUM9BYTES, MJUM16BYTES.
 */
struct mbuf *
m_getjcl(int how, short type, int flags, int size)
{
	struct mb_args args;
	struct mbuf *m, *n;
	uma_zone_t zone;

	if (size == MCLBYTES)
		return m_getcl(how, type, flags);

	args.flags = flags;
	args.type = type;

	m = uma_zalloc_arg(zone_mbuf, &args, how);
	if (m == NULL)
		return (NULL);

	zone = m_getzone(size);
	n = uma_zalloc_arg(zone, m, how);
	if (n == NULL) {
		uma_zfree(zone_mbuf, m);
		return (NULL);
	}
	MBUF_PROBE5(m__getjcl, how, type, flags, size, m);
	return (m);
}

/*
 * Allocate a given length worth of mbufs and/or clusters (whatever fits
 * best) and return a pointer to the top of the allocated chain.  If an
 * existing mbuf chain is provided, then we will append the new chain
 * to the existing one and return a pointer to the provided mbuf.
 */
struct mbuf *
m_getm2(struct mbuf *m, int len, int how, short type, int flags)
{
	struct mbuf *mb, *nm = NULL, *mtail = NULL;

	KASSERT(len >= 0, ("%s: len is < 0", __func__));

	/* Validate flags. */
	flags &= (M_PKTHDR | M_EOR);

	/* Packet header mbuf must be first in chain. */
	if ((flags & M_PKTHDR) && m != NULL)
		flags &= ~M_PKTHDR;

	/* Loop and append maximum sized mbufs to the chain tail. */
	while (len > 0) {
		mb = NULL;
		if (len > MCLBYTES) {
			mb = m_getjcl(M_NOWAIT, type, (flags & M_PKTHDR),
			    MJUMPAGESIZE);
		}
		if (mb == NULL) {
			if (len >= MINCLSIZE)
				mb = m_getcl(how, type, (flags & M_PKTHDR));
			else if (flags & M_PKTHDR)
				mb = m_gethdr(how, type);
			else
				mb = m_get(how, type);

			/*
			 * Fail the whole operation if one mbuf can't be
			 * allocated.
			 */
			if (mb == NULL) {
				m_freem(nm);
				return (NULL);
			}
		}

		/* Book keeping. */
		len -= M_SIZE(mb);
		if (mtail != NULL)
			mtail->m_next = mb;
		else
			nm = mb;
		mtail = mb;
		flags &= ~M_PKTHDR;	/* Only valid on the first mbuf. */
	}
	if (flags & M_EOR)
		mtail->m_flags |= M_EOR;  /* Only valid on the last mbuf. */

	/* If mbuf was supplied, append new chain to the end of it. */
	if (m != NULL) {
		for (mtail = m; mtail->m_next != NULL; mtail = mtail->m_next)
			;
		mtail->m_next = nm;
		mtail->m_flags &= ~M_EOR;
	} else
		m = nm;

	return (m);
}

/*-
 * Configure a provided mbuf to refer to the provided external storage
 * buffer and setup a reference count for said buffer.
 *
 * Arguments:
 *    mb     The existing mbuf to which to attach the provided buffer.
 *    buf    The address of the provided external storage buffer.
 *    size   The size of the provided buffer.
 *    freef  A pointer to a routine that is responsible for freeing the
 *           provided external storage buffer.
 *    args   A pointer to an argument structure (of any type) to be passed
 *           to the provided freef routine (may be NULL).
 *    flags  Any other flags to be passed to the provided mbuf.
 *    type   The type that the external storage buffer should be
 *           labeled with.
 *
 * Returns:
 *    Nothing.
 */
void
m_extadd(struct mbuf *mb, char *buf, u_int size, m_ext_free_t freef,
    void *arg1, void *arg2, int flags, int type)
{

	KASSERT(type != EXT_CLUSTER, ("%s: EXT_CLUSTER not allowed", __func__));

	mb->m_flags |= (M_EXT | flags);
	mb->m_ext.ext_buf = buf;
	mb->m_data = mb->m_ext.ext_buf;
	mb->m_ext.ext_size = size;
	mb->m_ext.ext_free = freef;
	mb->m_ext.ext_arg1 = arg1;
	mb->m_ext.ext_arg2 = arg2;
	mb->m_ext.ext_type = type;

	if (type != EXT_EXTREF) {
		mb->m_ext.ext_count = 1;
		mb->m_ext.ext_flags = EXT_FLAG_EMBREF;
	} else
		mb->m_ext.ext_flags = 0;
}

/*
 * Free an entire chain of mbufs and associated external buffers, if
 * applicable.
 */
void
m_freem(struct mbuf *mb)
{

	MBUF_PROBE1(m__freem, mb);
	while (mb != NULL)
		mb = m_free(mb);
}

int
m_snd_tag_alloc(struct ifnet *ifp, union if_snd_tag_alloc_params *params,
    struct m_snd_tag **mstp)
{

	if (ifp->if_snd_tag_alloc == NULL)
		return (EOPNOTSUPP);
	return (ifp->if_snd_tag_alloc(ifp, params, mstp));
}

void
m_snd_tag_init(struct m_snd_tag *mst, struct ifnet *ifp, u_int type)
{

	if_ref(ifp);
	mst->ifp = ifp;
	refcount_init(&mst->refcount, 1);
	mst->type = type;
	counter_u64_add(snd_tag_count, 1);
}

void
m_snd_tag_destroy(struct m_snd_tag *mst)
{
	struct ifnet *ifp;

	ifp = mst->ifp;
	ifp->if_snd_tag_free(mst);
	if_rele(ifp);
	counter_u64_add(snd_tag_count, -1);
}

/*
 * Allocate an mbuf with anonymous external pages.
 */
struct mbuf *
mb_alloc_ext_plus_pages(int len, int how)
{
	struct mbuf *m;
	vm_page_t pg;
	int i, npgs;

	m = mb_alloc_ext_pgs(how, mb_free_mext_pgs);
	if (m == NULL)
		return (NULL);
	m->m_epg_flags |= EPG_FLAG_ANON;
	npgs = howmany(len, PAGE_SIZE);
	for (i = 0; i < npgs; i++) {
		do {
			pg = vm_page_alloc(NULL, 0, VM_ALLOC_NORMAL |
			    VM_ALLOC_NOOBJ | VM_ALLOC_NODUMP | VM_ALLOC_WIRED);
			if (pg == NULL) {
				if (how == M_NOWAIT) {
					m->m_epg_npgs = i;
					m_free(m);
					return (NULL);
				}
				vm_wait(NULL);
			}
		} while (pg == NULL);
		m->m_epg_pa[i] = VM_PAGE_TO_PHYS(pg);
	}
	m->m_epg_npgs = npgs;
	return (m);
}

/*
 * Copy the data in the mbuf chain to a chain of mbufs with anonymous external
 * unmapped pages.
 * len is the length of data in the input mbuf chain.
 * mlen is the maximum number of bytes put into each ext_page mbuf.
 */
struct mbuf *
mb_mapped_to_unmapped(struct mbuf *mp, int len, int mlen, int how,
    struct mbuf **mlast)
{
	struct mbuf *m, *mout;
	char *pgpos, *mbpos;
	int i, mblen, mbufsiz, pglen, xfer;

	if (len == 0)
		return (NULL);
	mbufsiz = min(mlen, len);
	m = mout = mb_alloc_ext_plus_pages(mbufsiz, how);
	if (m == NULL)
		return (m);
	pgpos = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[0]);
	pglen = PAGE_SIZE;
	mblen = 0;
	i = 0;
	do {
		if (pglen == 0) {
			if (++i == m->m_epg_npgs) {
				m->m_epg_last_len = PAGE_SIZE;
				mbufsiz = min(mlen, len);
				m->m_next = mb_alloc_ext_plus_pages(mbufsiz,
				    how);
				m = m->m_next;
				if (m == NULL) {
					m_freem(mout);
					return (m);
				}
				i = 0;
			}
			pgpos = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[i]);
			pglen = PAGE_SIZE;
		}
		while (mblen == 0) {
			if (mp == NULL) {
				m_freem(mout);
				return (NULL);
			}
			KASSERT((mp->m_flags & M_EXTPG) == 0,
			    ("mb_copym_ext_pgs: ext_pgs input mbuf"));
			mbpos = mtod(mp, char *);
			mblen = mp->m_len;
			mp = mp->m_next;
		}
		xfer = min(mblen, pglen);
		memcpy(pgpos, mbpos, xfer);
		pgpos += xfer;
		mbpos += xfer;
		pglen -= xfer;
		mblen -= xfer;
		len -= xfer;
		m->m_len += xfer;
	} while (len > 0);
	m->m_epg_last_len = PAGE_SIZE - pglen;
	if (mlast != NULL)
		*mlast = m;
	return (mout);
}