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
|
/*-
* Copyright (c) 2013-2015 Gleb Smirnoff <glebius@FreeBSD.org>
* Copyright (c) 1998, David Greenman. 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, 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. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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_kern_tls.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/capsicum.h>
#include <sys/kernel.h>
#include <netinet/in.h>
#include <sys/lock.h>
#include <sys/ktls.h>
#include <sys/mutex.h>
#include <sys/sysproto.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/mman.h>
#include <sys/mount.h>
#include <sys/mbuf.h>
#include <sys/protosw.h>
#include <sys/rwlock.h>
#include <sys/sf_buf.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
#include <sys/vnode.h>
#include <net/vnet.h>
#include <netinet/tcp.h>
#include <security/audit/audit.h>
#include <security/mac/mac_framework.h>
#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_pager.h>
#define EXT_FLAG_SYNC EXT_FLAG_VENDOR1
#define EXT_FLAG_NOCACHE EXT_FLAG_VENDOR2
#define EXT_FLAG_CACHE_LAST EXT_FLAG_VENDOR3
/*
* Structure describing a single sendfile(2) I/O, which may consist of
* several underlying pager I/Os.
*
* The syscall context allocates the structure and initializes 'nios'
* to 1. As sendfile_swapin() runs through pages and starts asynchronous
* paging operations, it increments 'nios'.
*
* Every I/O completion calls sendfile_iodone(), which decrements the 'nios',
* and the syscall also calls sendfile_iodone() after allocating all mbufs,
* linking them and sending to socket. Whoever reaches zero 'nios' is
* responsible to * call pru_ready on the socket, to notify it of readyness
* of the data.
*/
struct sf_io {
volatile u_int nios;
u_int error;
int npages;
struct socket *so;
struct mbuf *m;
#ifdef KERN_TLS
struct ktls_session *tls;
#endif
vm_page_t pa[];
};
/*
* Structure used to track requests with SF_SYNC flag.
*/
struct sendfile_sync {
struct mtx mtx;
struct cv cv;
unsigned count;
};
counter_u64_t sfstat[sizeof(struct sfstat) / sizeof(uint64_t)];
static void
sfstat_init(const void *unused)
{
COUNTER_ARRAY_ALLOC(sfstat, sizeof(struct sfstat) / sizeof(uint64_t),
M_WAITOK);
}
SYSINIT(sfstat, SI_SUB_MBUF, SI_ORDER_FIRST, sfstat_init, NULL);
static int
sfstat_sysctl(SYSCTL_HANDLER_ARGS)
{
struct sfstat s;
COUNTER_ARRAY_COPY(sfstat, &s, sizeof(s) / sizeof(uint64_t));
if (req->newptr)
COUNTER_ARRAY_ZERO(sfstat, sizeof(s) / sizeof(uint64_t));
return (SYSCTL_OUT(req, &s, sizeof(s)));
}
SYSCTL_PROC(_kern_ipc, OID_AUTO, sfstat, CTLTYPE_OPAQUE | CTLFLAG_RW,
NULL, 0, sfstat_sysctl, "I", "sendfile statistics");
static void
sendfile_free_mext(struct mbuf *m)
{
struct sf_buf *sf;
vm_page_t pg;
int flags;
KASSERT(m->m_flags & M_EXT && m->m_ext.ext_type == EXT_SFBUF,
("%s: m %p !M_EXT or !EXT_SFBUF", __func__, m));
sf = m->m_ext.ext_arg1;
pg = sf_buf_page(sf);
flags = (m->m_ext.ext_flags & EXT_FLAG_NOCACHE) != 0 ? VPR_TRYFREE : 0;
sf_buf_free(sf);
vm_page_release(pg, flags);
if (m->m_ext.ext_flags & EXT_FLAG_SYNC) {
struct sendfile_sync *sfs = m->m_ext.ext_arg2;
mtx_lock(&sfs->mtx);
KASSERT(sfs->count > 0, ("Sendfile sync botchup count == 0"));
if (--sfs->count == 0)
cv_signal(&sfs->cv);
mtx_unlock(&sfs->mtx);
}
}
static void
sendfile_free_mext_pg(struct mbuf *m)
{
struct mbuf_ext_pgs *ext_pgs;
vm_page_t pg;
int flags, i;
bool cache_last;
KASSERT(m->m_flags & M_EXT && m->m_ext.ext_type == EXT_PGS,
("%s: m %p !M_EXT or !EXT_PGS", __func__, m));
cache_last = m->m_ext.ext_flags & EXT_FLAG_CACHE_LAST;
ext_pgs = m->m_ext.ext_pgs;
flags = (m->m_ext.ext_flags & EXT_FLAG_NOCACHE) != 0 ? VPR_TRYFREE : 0;
for (i = 0; i < ext_pgs->npgs; i++) {
if (cache_last && i == ext_pgs->npgs - 1)
flags = 0;
pg = PHYS_TO_VM_PAGE(ext_pgs->pa[i]);
vm_page_release(pg, flags);
}
if (m->m_ext.ext_flags & EXT_FLAG_SYNC) {
struct sendfile_sync *sfs = m->m_ext.ext_arg2;
mtx_lock(&sfs->mtx);
KASSERT(sfs->count > 0, ("Sendfile sync botchup count == 0"));
if (--sfs->count == 0)
cv_signal(&sfs->cv);
mtx_unlock(&sfs->mtx);
}
}
/*
* Helper function to calculate how much data to put into page i of n.
* Only first and last pages are special.
*/
static inline off_t
xfsize(int i, int n, off_t off, off_t len)
{
if (i == 0)
return (omin(PAGE_SIZE - (off & PAGE_MASK), len));
if (i == n - 1 && ((off + len) & PAGE_MASK) > 0)
return ((off + len) & PAGE_MASK);
return (PAGE_SIZE);
}
/*
* Helper function to get offset within object for i page.
*/
static inline vm_ooffset_t
vmoff(int i, off_t off)
{
if (i == 0)
return ((vm_ooffset_t)off);
return (trunc_page(off + i * PAGE_SIZE));
}
/*
* Helper function used when allocation of a page or sf_buf failed.
* Pretend as if we don't have enough space, subtract xfsize() of
* all pages that failed.
*/
static inline void
fixspace(int old, int new, off_t off, int *space)
{
KASSERT(old > new, ("%s: old %d new %d", __func__, old, new));
/* Subtract last one. */
*space -= xfsize(old - 1, old, off, *space);
old--;
if (new == old)
/* There was only one page. */
return;
/* Subtract first one. */
if (new == 0) {
*space -= xfsize(0, old, off, *space);
new++;
}
/* Rest of pages are full sized. */
*space -= (old - new) * PAGE_SIZE;
KASSERT(*space >= 0, ("%s: space went backwards", __func__));
}
/*
* I/O completion callback.
*/
static void
sendfile_iodone(void *arg, vm_page_t *pg, int count, int error)
{
struct sf_io *sfio = arg;
struct socket *so = sfio->so;
for (int i = 0; i < count; i++)
if (pg[i] != bogus_page)
vm_page_xunbusy_unchecked(pg[i]);
if (error)
sfio->error = error;
if (!refcount_release(&sfio->nios))
return;
if (__predict_false(sfio->error && sfio->m == NULL)) {
/*
* I/O operation failed, but pru_send hadn't been executed -
* nothing had been sent to the socket. The syscall has
* returned error to the user.
*/
free(sfio, M_TEMP);
return;
}
#if defined(KERN_TLS) && defined(INVARIANTS)
if ((sfio->m->m_flags & M_EXT) != 0 &&
sfio->m->m_ext.ext_type == EXT_PGS)
KASSERT(sfio->tls == sfio->m->m_ext.ext_pgs->tls,
("TLS session mismatch"));
else
KASSERT(sfio->tls == NULL,
("non-ext_pgs mbuf with TLS session"));
#endif
CURVNET_SET(so->so_vnet);
if (__predict_false(sfio->error)) {
/*
* I/O operation failed. The state of data in the socket
* is now inconsistent, and all what we can do is to tear
* it down. Protocol abort method would tear down protocol
* state, free all ready mbufs and detach not ready ones.
* We will free the mbufs corresponding to this I/O manually.
*
* The socket would be marked with EIO and made available
* for read, so that application receives EIO on next
* syscall and eventually closes the socket.
*/
so->so_proto->pr_usrreqs->pru_abort(so);
so->so_error = EIO;
mb_free_notready(sfio->m, sfio->npages);
#ifdef KERN_TLS
} else if (sfio->tls != NULL && sfio->tls->mode == TCP_TLS_MODE_SW) {
/*
* I/O operation is complete, but we still need to
* encrypt. We cannot do this in the interrupt thread
* of the disk controller, so forward the mbufs to a
* different thread.
*
* Donate the socket reference from sfio to rather
* than explicitly invoking soref().
*/
ktls_enqueue(sfio->m, so, sfio->npages);
goto out_with_ref;
#endif
} else
(void)(so->so_proto->pr_usrreqs->pru_ready)(so, sfio->m,
sfio->npages);
SOCK_LOCK(so);
sorele(so);
#ifdef KERN_TLS
out_with_ref:
#endif
CURVNET_RESTORE();
free(sfio, M_TEMP);
}
/*
* Iterate through pages vector and request paging for non-valid pages.
*/
static int
sendfile_swapin(vm_object_t obj, struct sf_io *sfio, int *nios, off_t off,
off_t len, int npages, int rhpages, int flags)
{
vm_page_t *pa = sfio->pa;
int grabbed;
*nios = 0;
flags = (flags & SF_NODISKIO) ? VM_ALLOC_NOWAIT : 0;
/*
* First grab all the pages and wire them. Note that we grab
* only required pages. Readahead pages are dealt with later.
*/
VM_OBJECT_WLOCK(obj);
grabbed = vm_page_grab_pages(obj, OFF_TO_IDX(off),
VM_ALLOC_NORMAL | VM_ALLOC_WIRED | flags, pa, npages);
if (grabbed < npages) {
for (int i = grabbed; i < npages; i++)
pa[i] = NULL;
npages = grabbed;
rhpages = 0;
}
for (int i = 0; i < npages;) {
int j, a, count, rv;
/* Skip valid pages. */
if (vm_page_is_valid(pa[i], vmoff(i, off) & PAGE_MASK,
xfsize(i, npages, off, len))) {
vm_page_xunbusy(pa[i]);
SFSTAT_INC(sf_pages_valid);
i++;
continue;
}
/*
* Next page is invalid. Check if it belongs to pager. It
* may not be there, which is a regular situation for shmem
* pager. For vnode pager this happens only in case of
* a sparse file.
*
* Important feature of vm_pager_has_page() is the hint
* stored in 'a', about how many pages we can pagein after
* this page in a single I/O.
*/
if (!vm_pager_has_page(obj, OFF_TO_IDX(vmoff(i, off)), NULL,
&a)) {
pmap_zero_page(pa[i]);
pa[i]->valid = VM_PAGE_BITS_ALL;
MPASS(pa[i]->dirty == 0);
vm_page_xunbusy(pa[i]);
i++;
continue;
}
/*
* We want to pagein as many pages as possible, limited only
* by the 'a' hint and actual request.
*/
count = min(a + 1, npages - i);
/*
* We should not pagein into a valid page, thus we first trim
* any valid pages off the end of request, and substitute
* to bogus_page those, that are in the middle.
*/
for (j = i + count - 1; j > i; j--) {
if (vm_page_is_valid(pa[j], vmoff(j, off) & PAGE_MASK,
xfsize(j, npages, off, len))) {
count--;
rhpages = 0;
} else
break;
}
for (j = i + 1; j < i + count - 1; j++)
if (vm_page_is_valid(pa[j], vmoff(j, off) & PAGE_MASK,
xfsize(j, npages, off, len))) {
vm_page_xunbusy(pa[j]);
SFSTAT_INC(sf_pages_valid);
SFSTAT_INC(sf_pages_bogus);
pa[j] = bogus_page;
}
refcount_acquire(&sfio->nios);
rv = vm_pager_get_pages_async(obj, pa + i, count, NULL,
i + count == npages ? &rhpages : NULL,
&sendfile_iodone, sfio);
if (__predict_false(rv != VM_PAGER_OK)) {
/*
* Perform full pages recovery before returning EIO.
* Pages from 0 to npages are wired.
* Pages from i to npages are also busied.
* Pages from (i + 1) to (i + count - 1) may be
* substituted to bogus page, and not busied.
*/
for (j = 0; j < npages; j++) {
if (j > i && j < i + count - 1 &&
pa[j] == bogus_page)
pa[j] = vm_page_lookup(obj,
OFF_TO_IDX(vmoff(j, off)));
else if (j >= i)
vm_page_xunbusy(pa[j]);
KASSERT(pa[j] != NULL && pa[j] != bogus_page,
("%s: page %p[%d] I/O recovery failure",
__func__, pa, j));
vm_page_unwire(pa[j], PQ_INACTIVE);
}
VM_OBJECT_WUNLOCK(obj);
return (EIO);
}
SFSTAT_INC(sf_iocnt);
SFSTAT_ADD(sf_pages_read, count);
if (i + count == npages)
SFSTAT_ADD(sf_rhpages_read, rhpages);
/*
* Restore the valid page pointers. They are already
* unbusied, but still wired.
*/
for (j = i; j < i + count; j++)
if (pa[j] == bogus_page) {
pa[j] = vm_page_lookup(obj,
OFF_TO_IDX(vmoff(j, off)));
KASSERT(pa[j], ("%s: page %p[%d] disappeared",
__func__, pa, j));
}
i += count;
(*nios)++;
}
VM_OBJECT_WUNLOCK(obj);
if (*nios == 0 && npages != 0)
SFSTAT_INC(sf_noiocnt);
return (0);
}
static int
sendfile_getobj(struct thread *td, struct file *fp, vm_object_t *obj_res,
struct vnode **vp_res, struct shmfd **shmfd_res, off_t *obj_size,
int *bsize)
{
struct vattr va;
vm_object_t obj;
struct vnode *vp;
struct shmfd *shmfd;
int error;
vp = *vp_res = NULL;
obj = NULL;
shmfd = *shmfd_res = NULL;
*bsize = 0;
/*
* The file descriptor must be a regular file and have a
* backing VM object.
*/
if (fp->f_type == DTYPE_VNODE) {
vp = fp->f_vnode;
vn_lock(vp, LK_SHARED | LK_RETRY);
if (vp->v_type != VREG) {
error = EINVAL;
goto out;
}
*bsize = vp->v_mount->mnt_stat.f_iosize;
error = VOP_GETATTR(vp, &va, td->td_ucred);
if (error != 0)
goto out;
*obj_size = va.va_size;
obj = vp->v_object;
if (obj == NULL) {
error = EINVAL;
goto out;
}
} else if (fp->f_type == DTYPE_SHM) {
error = 0;
shmfd = fp->f_data;
obj = shmfd->shm_object;
*obj_size = shmfd->shm_size;
} else {
error = EINVAL;
goto out;
}
VM_OBJECT_WLOCK(obj);
if ((obj->flags & OBJ_DEAD) != 0) {
VM_OBJECT_WUNLOCK(obj);
error = EBADF;
goto out;
}
/*
* Temporarily increase the backing VM object's reference
* count so that a forced reclamation of its vnode does not
* immediately destroy it.
*/
vm_object_reference_locked(obj);
VM_OBJECT_WUNLOCK(obj);
*obj_res = obj;
*vp_res = vp;
*shmfd_res = shmfd;
out:
if (vp != NULL)
VOP_UNLOCK(vp);
return (error);
}
static int
sendfile_getsock(struct thread *td, int s, struct file **sock_fp,
struct socket **so)
{
int error;
*sock_fp = NULL;
*so = NULL;
/*
* The socket must be a stream socket and connected.
*/
error = getsock_cap(td, s, &cap_send_rights,
sock_fp, NULL, NULL);
if (error != 0)
return (error);
*so = (*sock_fp)->f_data;
if ((*so)->so_type != SOCK_STREAM)
return (EINVAL);
if (SOLISTENING(*so))
return (ENOTCONN);
return (0);
}
int
vn_sendfile(struct file *fp, int sockfd, struct uio *hdr_uio,
struct uio *trl_uio, off_t offset, size_t nbytes, off_t *sent, int flags,
struct thread *td)
{
struct file *sock_fp;
struct vnode *vp;
struct vm_object *obj;
struct socket *so;
#ifdef KERN_TLS
struct ktls_session *tls;
#endif
struct mbuf_ext_pgs *ext_pgs;
struct mbuf *m, *mh, *mhtail;
struct sf_buf *sf;
struct shmfd *shmfd;
struct sendfile_sync *sfs;
struct vattr va;
off_t off, sbytes, rem, obj_size;
int bsize, error, ext_pgs_idx, hdrlen, max_pgs, softerr;
#ifdef KERN_TLS
int tls_enq_cnt;
#endif
bool use_ext_pgs;
obj = NULL;
so = NULL;
m = mh = NULL;
sfs = NULL;
#ifdef KERN_TLS
tls = NULL;
#endif
hdrlen = sbytes = 0;
softerr = 0;
use_ext_pgs = false;
error = sendfile_getobj(td, fp, &obj, &vp, &shmfd, &obj_size, &bsize);
if (error != 0)
return (error);
error = sendfile_getsock(td, sockfd, &sock_fp, &so);
if (error != 0)
goto out;
#ifdef MAC
error = mac_socket_check_send(td->td_ucred, so);
if (error != 0)
goto out;
#endif
SFSTAT_INC(sf_syscalls);
SFSTAT_ADD(sf_rhpages_requested, SF_READAHEAD(flags));
if (flags & SF_SYNC) {
sfs = malloc(sizeof *sfs, M_TEMP, M_WAITOK | M_ZERO);
mtx_init(&sfs->mtx, "sendfile", NULL, MTX_DEF);
cv_init(&sfs->cv, "sendfile");
}
rem = nbytes ? omin(nbytes, obj_size - offset) : obj_size - offset;
/*
* Protect against multiple writers to the socket.
*
* XXXRW: Historically this has assumed non-interruptibility, so now
* we implement that, but possibly shouldn't.
*/
(void)sblock(&so->so_snd, SBL_WAIT | SBL_NOINTR);
#ifdef KERN_TLS
tls = ktls_hold(so->so_snd.sb_tls_info);
#endif
/*
* Loop through the pages of the file, starting with the requested
* offset. Get a file page (do I/O if necessary), map the file page
* into an sf_buf, attach an mbuf header to the sf_buf, and queue
* it on the socket.
* This is done in two loops. The inner loop turns as many pages
* as it can, up to available socket buffer space, without blocking
* into mbufs to have it bulk delivered into the socket send buffer.
* The outer loop checks the state and available space of the socket
* and takes care of the overall progress.
*/
for (off = offset; rem > 0; ) {
struct sf_io *sfio;
vm_page_t *pa;
struct mbuf *mtail;
int nios, space, npages, rhpages;
mtail = NULL;
/*
* Check the socket state for ongoing connection,
* no errors and space in socket buffer.
* If space is low allow for the remainder of the
* file to be processed if it fits the socket buffer.
* Otherwise block in waiting for sufficient space
* to proceed, or if the socket is nonblocking, return
* to userland with EAGAIN while reporting how far
* we've come.
* We wait until the socket buffer has significant free
* space to do bulk sends. This makes good use of file
* system read ahead and allows packet segmentation
* offloading hardware to take over lots of work. If
* we were not careful here we would send off only one
* sfbuf at a time.
*/
SOCKBUF_LOCK(&so->so_snd);
if (so->so_snd.sb_lowat < so->so_snd.sb_hiwat / 2)
so->so_snd.sb_lowat = so->so_snd.sb_hiwat / 2;
retry_space:
if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
error = EPIPE;
SOCKBUF_UNLOCK(&so->so_snd);
goto done;
} else if (so->so_error) {
error = so->so_error;
so->so_error = 0;
SOCKBUF_UNLOCK(&so->so_snd);
goto done;
}
if ((so->so_state & SS_ISCONNECTED) == 0) {
SOCKBUF_UNLOCK(&so->so_snd);
error = ENOTCONN;
goto done;
}
space = sbspace(&so->so_snd);
if (space < rem &&
(space <= 0 ||
space < so->so_snd.sb_lowat)) {
if (so->so_state & SS_NBIO) {
SOCKBUF_UNLOCK(&so->so_snd);
error = EAGAIN;
goto done;
}
/*
* sbwait drops the lock while sleeping.
* When we loop back to retry_space the
* state may have changed and we retest
* for it.
*/
error = sbwait(&so->so_snd);
/*
* An error from sbwait usually indicates that we've
* been interrupted by a signal. If we've sent anything
* then return bytes sent, otherwise return the error.
*/
if (error != 0) {
SOCKBUF_UNLOCK(&so->so_snd);
goto done;
}
goto retry_space;
}
SOCKBUF_UNLOCK(&so->so_snd);
/*
* At the beginning of the first loop check if any headers
* are specified and copy them into mbufs. Reduce space in
* the socket buffer by the size of the header mbuf chain.
* Clear hdr_uio here and hdrlen at the end of the first loop.
*/
if (hdr_uio != NULL && hdr_uio->uio_resid > 0) {
hdr_uio->uio_td = td;
hdr_uio->uio_rw = UIO_WRITE;
#ifdef KERN_TLS
if (tls != NULL)
mh = m_uiotombuf(hdr_uio, M_WAITOK, space,
tls->params.max_frame_len, M_NOMAP);
else
#endif
mh = m_uiotombuf(hdr_uio, M_WAITOK,
space, 0, 0);
hdrlen = m_length(mh, &mhtail);
space -= hdrlen;
/*
* If header consumed all the socket buffer space,
* don't waste CPU cycles and jump to the end.
*/
if (space == 0) {
sfio = NULL;
nios = 0;
goto prepend_header;
}
hdr_uio = NULL;
}
if (vp != NULL) {
error = vn_lock(vp, LK_SHARED);
if (error != 0)
goto done;
error = VOP_GETATTR(vp, &va, td->td_ucred);
if (error != 0 || off >= va.va_size) {
VOP_UNLOCK(vp);
goto done;
}
if (va.va_size != obj_size) {
obj_size = va.va_size;
rem = nbytes ?
omin(nbytes + offset, obj_size) : obj_size;
rem -= off;
}
}
if (space > rem)
space = rem;
else if (space > PAGE_SIZE) {
/*
* Use page boundaries when possible for large
* requests.
*/
if (off & PAGE_MASK)
space -= (PAGE_SIZE - (off & PAGE_MASK));
space = trunc_page(space);
if (off & PAGE_MASK)
space += (PAGE_SIZE - (off & PAGE_MASK));
}
npages = howmany(space + (off & PAGE_MASK), PAGE_SIZE);
/*
* Calculate maximum allowed number of pages for readahead
* at this iteration. If SF_USER_READAHEAD was set, we don't
* do any heuristics and use exactly the value supplied by
* application. Otherwise, we allow readahead up to "rem".
* If application wants more, let it be, but there is no
* reason to go above MAXPHYS. Also check against "obj_size",
* since vm_pager_has_page() can hint beyond EOF.
*/
if (flags & SF_USER_READAHEAD) {
rhpages = SF_READAHEAD(flags);
} else {
rhpages = howmany(rem + (off & PAGE_MASK), PAGE_SIZE) -
npages;
rhpages += SF_READAHEAD(flags);
}
rhpages = min(howmany(MAXPHYS, PAGE_SIZE), rhpages);
rhpages = min(howmany(obj_size - trunc_page(off), PAGE_SIZE) -
npages, rhpages);
sfio = malloc(sizeof(struct sf_io) +
npages * sizeof(vm_page_t), M_TEMP, M_WAITOK);
refcount_init(&sfio->nios, 1);
sfio->so = so;
sfio->error = 0;
#ifdef KERN_TLS
/*
* This doesn't use ktls_hold() because sfio->m will
* also have a reference on 'tls' that will be valid
* for all of sfio's lifetime.
*/
sfio->tls = tls;
#endif
error = sendfile_swapin(obj, sfio, &nios, off, space, npages,
rhpages, flags);
if (error != 0) {
if (vp != NULL)
VOP_UNLOCK(vp);
sfio->m = NULL;
sendfile_iodone(sfio, NULL, 0, error);
goto done;
}
/*
* Loop and construct maximum sized mbuf chain to be bulk
* dumped into socket buffer.
*/
pa = sfio->pa;
/*
* Use unmapped mbufs if enabled for TCP. Unmapped
* bufs are restricted to TCP as that is what has been
* tested. In particular, unmapped mbufs have not
* been tested with UNIX-domain sockets.
*
* TLS frames always require unmapped mbufs.
*/
if ((mb_use_ext_pgs &&
so->so_proto->pr_protocol == IPPROTO_TCP)
#ifdef KERN_TLS
|| tls != NULL
#endif
) {
use_ext_pgs = true;
#ifdef KERN_TLS
if (tls != NULL)
max_pgs = num_pages(tls->params.max_frame_len);
else
#endif
max_pgs = MBUF_PEXT_MAX_PGS;
/* Start at last index, to wrap on first use. */
ext_pgs_idx = max_pgs - 1;
}
for (int i = 0; i < npages; i++) {
struct mbuf *m0;
/*
* If a page wasn't grabbed successfully, then
* trim the array. Can happen only with SF_NODISKIO.
*/
if (pa[i] == NULL) {
SFSTAT_INC(sf_busy);
fixspace(npages, i, off, &space);
npages = i;
softerr = EBUSY;
break;
}
if (use_ext_pgs) {
off_t xfs;
ext_pgs_idx++;
if (ext_pgs_idx == max_pgs) {
m0 = mb_alloc_ext_pgs(M_WAITOK, false,
sendfile_free_mext_pg);
if (flags & SF_NOCACHE) {
m0->m_ext.ext_flags |=
EXT_FLAG_NOCACHE;
/*
* See comment below regarding
* ignoring SF_NOCACHE for the
* last page.
*/
if ((npages - i <= max_pgs) &&
((off + space) & PAGE_MASK) &&
(rem > space || rhpages > 0))
m0->m_ext.ext_flags |=
EXT_FLAG_CACHE_LAST;
}
if (sfs != NULL) {
m0->m_ext.ext_flags |=
EXT_FLAG_SYNC;
m0->m_ext.ext_arg2 = sfs;
mtx_lock(&sfs->mtx);
sfs->count++;
mtx_unlock(&sfs->mtx);
}
ext_pgs = m0->m_ext.ext_pgs;
if (i == 0)
sfio->m = m0;
ext_pgs_idx = 0;
/* Append to mbuf chain. */
if (mtail != NULL)
mtail->m_next = m0;
else
m = m0;
mtail = m0;
ext_pgs->first_pg_off =
vmoff(i, off) & PAGE_MASK;
}
if (nios) {
mtail->m_flags |= M_NOTREADY;
ext_pgs->nrdy++;
}
ext_pgs->pa[ext_pgs_idx] = VM_PAGE_TO_PHYS(pa[i]);
ext_pgs->npgs++;
xfs = xfsize(i, npages, off, space);
ext_pgs->last_pg_len = xfs;
MBUF_EXT_PGS_ASSERT_SANITY(ext_pgs);
mtail->m_len += xfs;
mtail->m_ext.ext_size += PAGE_SIZE;
continue;
}
/*
* Get a sendfile buf. When allocating the
* first buffer for mbuf chain, we usually
* wait as long as necessary, but this wait
* can be interrupted. For consequent
* buffers, do not sleep, since several
* threads might exhaust the buffers and then
* deadlock.
*/
sf = sf_buf_alloc(pa[i],
m != NULL ? SFB_NOWAIT : SFB_CATCH);
if (sf == NULL) {
SFSTAT_INC(sf_allocfail);
for (int j = i; j < npages; j++)
vm_page_unwire(pa[j], PQ_INACTIVE);
if (m == NULL)
softerr = ENOBUFS;
fixspace(npages, i, off, &space);
npages = i;
break;
}
m0 = m_get(M_WAITOK, MT_DATA);
m0->m_ext.ext_buf = (char *)sf_buf_kva(sf);
m0->m_ext.ext_size = PAGE_SIZE;
m0->m_ext.ext_arg1 = sf;
m0->m_ext.ext_type = EXT_SFBUF;
m0->m_ext.ext_flags = EXT_FLAG_EMBREF;
m0->m_ext.ext_free = sendfile_free_mext;
/*
* SF_NOCACHE sets the page as being freed upon send.
* However, we ignore it for the last page in 'space',
* if the page is truncated, and we got more data to
* send (rem > space), or if we have readahead
* configured (rhpages > 0).
*/
if ((flags & SF_NOCACHE) &&
(i != npages - 1 ||
!((off + space) & PAGE_MASK) ||
!(rem > space || rhpages > 0)))
m0->m_ext.ext_flags |= EXT_FLAG_NOCACHE;
if (sfs != NULL) {
m0->m_ext.ext_flags |= EXT_FLAG_SYNC;
m0->m_ext.ext_arg2 = sfs;
mtx_lock(&sfs->mtx);
sfs->count++;
mtx_unlock(&sfs->mtx);
}
m0->m_ext.ext_count = 1;
m0->m_flags |= (M_EXT | M_RDONLY);
if (nios)
m0->m_flags |= M_NOTREADY;
m0->m_data = (char *)sf_buf_kva(sf) +
(vmoff(i, off) & PAGE_MASK);
m0->m_len = xfsize(i, npages, off, space);
if (i == 0)
sfio->m = m0;
/* Append to mbuf chain. */
if (mtail != NULL)
mtail->m_next = m0;
else
m = m0;
mtail = m0;
}
if (vp != NULL)
VOP_UNLOCK(vp);
/* Keep track of bytes processed. */
off += space;
rem -= space;
/* Prepend header, if any. */
if (hdrlen) {
prepend_header:
mhtail->m_next = m;
m = mh;
mh = NULL;
}
if (m == NULL) {
KASSERT(softerr, ("%s: m NULL, no error", __func__));
error = softerr;
free(sfio, M_TEMP);
goto done;
}
/* Add the buffer chain to the socket buffer. */
KASSERT(m_length(m, NULL) == space + hdrlen,
("%s: mlen %u space %d hdrlen %d",
__func__, m_length(m, NULL), space, hdrlen));
CURVNET_SET(so->so_vnet);
#ifdef KERN_TLS
if (tls != NULL) {
error = ktls_frame(m, tls, &tls_enq_cnt,
TLS_RLTYPE_APP);
if (error != 0)
goto done;
}
#endif
if (nios == 0) {
/*
* If sendfile_swapin() didn't initiate any I/Os,
* which happens if all data is cached in VM, then
* we can send data right now without the
* PRUS_NOTREADY flag.
*/
free(sfio, M_TEMP);
#ifdef KERN_TLS
if (tls != NULL && tls->mode == TCP_TLS_MODE_SW) {
error = (*so->so_proto->pr_usrreqs->pru_send)
(so, PRUS_NOTREADY, m, NULL, NULL, td);
soref(so);
ktls_enqueue(m, so, tls_enq_cnt);
} else
#endif
error = (*so->so_proto->pr_usrreqs->pru_send)
(so, 0, m, NULL, NULL, td);
} else {
sfio->npages = npages;
soref(so);
error = (*so->so_proto->pr_usrreqs->pru_send)
(so, PRUS_NOTREADY, m, NULL, NULL, td);
sendfile_iodone(sfio, NULL, 0, 0);
}
CURVNET_RESTORE();
m = NULL; /* pru_send always consumes */
if (error)
goto done;
sbytes += space + hdrlen;
if (hdrlen)
hdrlen = 0;
if (softerr) {
error = softerr;
goto done;
}
}
/*
* Send trailers. Wimp out and use writev(2).
*/
if (trl_uio != NULL) {
sbunlock(&so->so_snd);
error = kern_writev(td, sockfd, trl_uio);
if (error == 0)
sbytes += td->td_retval[0];
goto out;
}
done:
sbunlock(&so->so_snd);
out:
/*
* If there was no error we have to clear td->td_retval[0]
* because it may have been set by writev.
*/
if (error == 0) {
td->td_retval[0] = 0;
}
if (sent != NULL) {
(*sent) = sbytes;
}
if (obj != NULL)
vm_object_deallocate(obj);
if (so)
fdrop(sock_fp, td);
if (m)
m_freem(m);
if (mh)
m_freem(mh);
if (sfs != NULL) {
mtx_lock(&sfs->mtx);
if (sfs->count != 0)
cv_wait(&sfs->cv, &sfs->mtx);
KASSERT(sfs->count == 0, ("sendfile sync still busy"));
cv_destroy(&sfs->cv);
mtx_destroy(&sfs->mtx);
free(sfs, M_TEMP);
}
#ifdef KERN_TLS
if (tls != NULL)
ktls_free(tls);
#endif
if (error == ERESTART)
error = EINTR;
return (error);
}
static int
sendfile(struct thread *td, struct sendfile_args *uap, int compat)
{
struct sf_hdtr hdtr;
struct uio *hdr_uio, *trl_uio;
struct file *fp;
off_t sbytes;
int error;
/*
* File offset must be positive. If it goes beyond EOF
* we send only the header/trailer and no payload data.
*/
if (uap->offset < 0)
return (EINVAL);
sbytes = 0;
hdr_uio = trl_uio = NULL;
if (uap->hdtr != NULL) {
error = copyin(uap->hdtr, &hdtr, sizeof(hdtr));
if (error != 0)
goto out;
if (hdtr.headers != NULL) {
error = copyinuio(hdtr.headers, hdtr.hdr_cnt,
&hdr_uio);
if (error != 0)
goto out;
#ifdef COMPAT_FREEBSD4
/*
* In FreeBSD < 5.0 the nbytes to send also included
* the header. If compat is specified subtract the
* header size from nbytes.
*/
if (compat) {
if (uap->nbytes > hdr_uio->uio_resid)
uap->nbytes -= hdr_uio->uio_resid;
else
uap->nbytes = 0;
}
#endif
}
if (hdtr.trailers != NULL) {
error = copyinuio(hdtr.trailers, hdtr.trl_cnt,
&trl_uio);
if (error != 0)
goto out;
}
}
AUDIT_ARG_FD(uap->fd);
/*
* sendfile(2) can start at any offset within a file so we require
* CAP_READ+CAP_SEEK = CAP_PREAD.
*/
if ((error = fget_read(td, uap->fd, &cap_pread_rights, &fp)) != 0)
goto out;
error = fo_sendfile(fp, uap->s, hdr_uio, trl_uio, uap->offset,
uap->nbytes, &sbytes, uap->flags, td);
fdrop(fp, td);
if (uap->sbytes != NULL)
copyout(&sbytes, uap->sbytes, sizeof(off_t));
out:
free(hdr_uio, M_IOV);
free(trl_uio, M_IOV);
return (error);
}
/*
* sendfile(2)
*
* int sendfile(int fd, int s, off_t offset, size_t nbytes,
* struct sf_hdtr *hdtr, off_t *sbytes, int flags)
*
* Send a file specified by 'fd' and starting at 'offset' to a socket
* specified by 's'. Send only 'nbytes' of the file or until EOF if nbytes ==
* 0. Optionally add a header and/or trailer to the socket output. If
* specified, write the total number of bytes sent into *sbytes.
*/
int
sys_sendfile(struct thread *td, struct sendfile_args *uap)
{
return (sendfile(td, uap, 0));
}
#ifdef COMPAT_FREEBSD4
int
freebsd4_sendfile(struct thread *td, struct freebsd4_sendfile_args *uap)
{
struct sendfile_args args;
args.fd = uap->fd;
args.s = uap->s;
args.offset = uap->offset;
args.nbytes = uap->nbytes;
args.hdtr = uap->hdtr;
args.sbytes = uap->sbytes;
args.flags = uap->flags;
return (sendfile(td, &args, 1));
}
#endif /* COMPAT_FREEBSD4 */
|