]> git.saurik.com Git - apple/xnu.git/blame_incremental - bsd/nfs/nfs_subs.c
xnu-4903.270.47.tar.gz
[apple/xnu.git] / bsd / nfs / nfs_subs.c
... / ...
CommitLineData
1/*
2 * Copyright (c) 2000-2016 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28/* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
29/*
30 * Copyright (c) 1989, 1993
31 * The Regents of the University of California. All rights reserved.
32 *
33 * This code is derived from software contributed to Berkeley by
34 * Rick Macklem at The University of Guelph.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 3. All advertising materials mentioning features or use of this software
45 * must display the following acknowledgement:
46 * This product includes software developed by the University of
47 * California, Berkeley and its contributors.
48 * 4. Neither the name of the University nor the names of its contributors
49 * may be used to endorse or promote products derived from this software
50 * without specific prior written permission.
51 *
52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * SUCH DAMAGE.
63 *
64 * @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95
65 * FreeBSD-Id: nfs_subs.c,v 1.47 1997/11/07 08:53:24 phk Exp $
66 */
67
68/*
69 * These functions support the macros and help fiddle mbuf chains for
70 * the nfs op functions. They do things like create the rpc header and
71 * copy data between mbuf chains and uio lists.
72 */
73#include <sys/param.h>
74#include <sys/proc.h>
75#include <sys/kauth.h>
76#include <sys/systm.h>
77#include <sys/kernel.h>
78#include <sys/mount_internal.h>
79#include <sys/vnode_internal.h>
80#include <sys/kpi_mbuf.h>
81#include <sys/socket.h>
82#include <sys/stat.h>
83#include <sys/malloc.h>
84#include <sys/syscall.h>
85#include <sys/ubc_internal.h>
86#include <sys/fcntl.h>
87#include <sys/uio.h>
88#include <sys/domain.h>
89#include <libkern/OSAtomic.h>
90#include <kern/thread_call.h>
91
92#include <sys/vm.h>
93#include <sys/vmparam.h>
94
95#include <sys/time.h>
96#include <kern/clock.h>
97
98#include <nfs/rpcv2.h>
99#include <nfs/nfsproto.h>
100#include <nfs/nfs.h>
101#include <nfs/nfsnode.h>
102#if NFSCLIENT
103#define _NFS_XDR_SUBS_FUNCS_ /* define this to get xdrbuf function definitions */
104#endif
105#include <nfs/xdr_subs.h>
106#include <nfs/nfsm_subs.h>
107#include <nfs/nfs_gss.h>
108#include <nfs/nfsmount.h>
109#include <nfs/nfs_lock.h>
110
111#include <miscfs/specfs/specdev.h>
112
113#include <netinet/in.h>
114#include <net/kpi_interface.h>
115
116#include <sys/utfconv.h>
117
118/*
119 * NFS globals
120 */
121struct nfsstats __attribute__((aligned(8))) nfsstats;
122size_t nfs_mbuf_mhlen = 0, nfs_mbuf_minclsize = 0;
123
124/*
125 * functions to convert between NFS and VFS types
126 */
127nfstype
128vtonfs_type(enum vtype vtype, int nfsvers)
129{
130 switch (vtype) {
131 case VNON:
132 return NFNON;
133 case VREG:
134 return NFREG;
135 case VDIR:
136 return NFDIR;
137 case VBLK:
138 return NFBLK;
139 case VCHR:
140 return NFCHR;
141 case VLNK:
142 return NFLNK;
143 case VSOCK:
144 if (nfsvers > NFS_VER2) {
145 return NFSOCK;
146 }
147 case VFIFO:
148 if (nfsvers > NFS_VER2) {
149 return NFFIFO;
150 }
151 case VBAD:
152 case VSTR:
153 case VCPLX:
154 default:
155 return NFNON;
156 }
157}
158
159enum vtype
160nfstov_type(nfstype nvtype, int nfsvers)
161{
162 switch (nvtype) {
163 case NFNON:
164 return VNON;
165 case NFREG:
166 return VREG;
167 case NFDIR:
168 return VDIR;
169 case NFBLK:
170 return VBLK;
171 case NFCHR:
172 return VCHR;
173 case NFLNK:
174 return VLNK;
175 case NFSOCK:
176 if (nfsvers > NFS_VER2) {
177 return VSOCK;
178 }
179 case NFFIFO:
180 if (nfsvers > NFS_VER2) {
181 return VFIFO;
182 }
183 case NFATTRDIR:
184 if (nfsvers > NFS_VER3) {
185 return VDIR;
186 }
187 case NFNAMEDATTR:
188 if (nfsvers > NFS_VER3) {
189 return VREG;
190 }
191 default:
192 return VNON;
193 }
194}
195
196int
197vtonfsv2_mode(enum vtype vtype, mode_t m)
198{
199 switch (vtype) {
200 case VNON:
201 case VREG:
202 case VDIR:
203 case VBLK:
204 case VCHR:
205 case VLNK:
206 case VSOCK:
207 return vnode_makeimode(vtype, m);
208 case VFIFO:
209 return vnode_makeimode(VCHR, m);
210 case VBAD:
211 case VSTR:
212 case VCPLX:
213 default:
214 return vnode_makeimode(VNON, m);
215 }
216}
217
218#if NFSSERVER
219
220/*
221 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
222 */
223int nfsv3_procid[NFS_NPROCS] = {
224 NFSPROC_NULL,
225 NFSPROC_GETATTR,
226 NFSPROC_SETATTR,
227 NFSPROC_NOOP,
228 NFSPROC_LOOKUP,
229 NFSPROC_READLINK,
230 NFSPROC_READ,
231 NFSPROC_NOOP,
232 NFSPROC_WRITE,
233 NFSPROC_CREATE,
234 NFSPROC_REMOVE,
235 NFSPROC_RENAME,
236 NFSPROC_LINK,
237 NFSPROC_SYMLINK,
238 NFSPROC_MKDIR,
239 NFSPROC_RMDIR,
240 NFSPROC_READDIR,
241 NFSPROC_FSSTAT,
242 NFSPROC_NOOP,
243 NFSPROC_NOOP,
244 NFSPROC_NOOP,
245 NFSPROC_NOOP,
246 NFSPROC_NOOP
247};
248
249#endif /* NFSSERVER */
250
251/*
252 * and the reverse mapping from generic to Version 2 procedure numbers
253 */
254int nfsv2_procid[NFS_NPROCS] = {
255 NFSV2PROC_NULL,
256 NFSV2PROC_GETATTR,
257 NFSV2PROC_SETATTR,
258 NFSV2PROC_LOOKUP,
259 NFSV2PROC_NOOP,
260 NFSV2PROC_READLINK,
261 NFSV2PROC_READ,
262 NFSV2PROC_WRITE,
263 NFSV2PROC_CREATE,
264 NFSV2PROC_MKDIR,
265 NFSV2PROC_SYMLINK,
266 NFSV2PROC_CREATE,
267 NFSV2PROC_REMOVE,
268 NFSV2PROC_RMDIR,
269 NFSV2PROC_RENAME,
270 NFSV2PROC_LINK,
271 NFSV2PROC_READDIR,
272 NFSV2PROC_NOOP,
273 NFSV2PROC_STATFS,
274 NFSV2PROC_NOOP,
275 NFSV2PROC_NOOP,
276 NFSV2PROC_NOOP,
277 NFSV2PROC_NOOP
278};
279
280
281/*
282 * initialize NFS's cache of mbuf constants
283 */
284void
285nfs_mbuf_init(void)
286{
287 struct mbuf_stat ms;
288
289 mbuf_stats(&ms);
290 nfs_mbuf_mhlen = ms.mhlen;
291 nfs_mbuf_minclsize = ms.minclsize;
292}
293
294#if NFSSERVER
295
296/*
297 * allocate a list of mbufs to hold the given amount of data
298 */
299int
300nfsm_mbuf_get_list(size_t size, mbuf_t *mp, int *mbcnt)
301{
302 int error, cnt;
303 mbuf_t mhead, mlast, m;
304 size_t len, mlen;
305
306 error = cnt = 0;
307 mhead = mlast = NULL;
308 len = 0;
309
310 while (len < size) {
311 nfsm_mbuf_get(error, &m, (size - len));
312 if (error) {
313 break;
314 }
315 if (!mhead) {
316 mhead = m;
317 }
318 if (mlast && ((error = mbuf_setnext(mlast, m)))) {
319 mbuf_free(m);
320 break;
321 }
322 mlen = mbuf_maxlen(m);
323 if ((len + mlen) > size) {
324 mlen = size - len;
325 }
326 mbuf_setlen(m, mlen);
327 len += mlen;
328 cnt++;
329 mlast = m;
330 }
331
332 if (!error) {
333 *mp = mhead;
334 *mbcnt = cnt;
335 }
336 return error;
337}
338
339#endif /* NFSSERVER */
340
341/*
342 * nfsm_chain_new_mbuf()
343 *
344 * Add a new mbuf to the given chain.
345 */
346int
347nfsm_chain_new_mbuf(struct nfsm_chain *nmc, size_t sizehint)
348{
349 mbuf_t mb;
350 int error = 0;
351
352 if (nmc->nmc_flags & NFSM_CHAIN_FLAG_ADD_CLUSTERS) {
353 sizehint = nfs_mbuf_minclsize;
354 }
355
356 /* allocate a new mbuf */
357 nfsm_mbuf_get(error, &mb, sizehint);
358 if (error) {
359 return error;
360 }
361 if (mb == NULL) {
362 panic("got NULL mbuf?");
363 }
364
365 /* do we have a current mbuf? */
366 if (nmc->nmc_mcur) {
367 /* first cap off current mbuf */
368 mbuf_setlen(nmc->nmc_mcur, nmc->nmc_ptr - (caddr_t)mbuf_data(nmc->nmc_mcur));
369 /* then append the new mbuf */
370 error = mbuf_setnext(nmc->nmc_mcur, mb);
371 if (error) {
372 mbuf_free(mb);
373 return error;
374 }
375 }
376
377 /* set up for using the new mbuf */
378 nmc->nmc_mcur = mb;
379 nmc->nmc_ptr = mbuf_data(mb);
380 nmc->nmc_left = mbuf_trailingspace(mb);
381
382 return 0;
383}
384
385/*
386 * nfsm_chain_add_opaque_f()
387 *
388 * Add "len" bytes of opaque data pointed to by "buf" to the given chain.
389 */
390int
391nfsm_chain_add_opaque_f(struct nfsm_chain *nmc, const u_char *buf, uint32_t len)
392{
393 uint32_t paddedlen, tlen;
394 int error;
395
396 paddedlen = nfsm_rndup(len);
397
398 while (paddedlen) {
399 if (!nmc->nmc_left) {
400 error = nfsm_chain_new_mbuf(nmc, paddedlen);
401 if (error) {
402 return error;
403 }
404 }
405 tlen = MIN(nmc->nmc_left, paddedlen);
406 if (tlen) {
407 if (len) {
408 if (tlen > len) {
409 tlen = len;
410 }
411 bcopy(buf, nmc->nmc_ptr, tlen);
412 } else {
413 bzero(nmc->nmc_ptr, tlen);
414 }
415 nmc->nmc_ptr += tlen;
416 nmc->nmc_left -= tlen;
417 paddedlen -= tlen;
418 if (len) {
419 buf += tlen;
420 len -= tlen;
421 }
422 }
423 }
424 return 0;
425}
426
427/*
428 * nfsm_chain_add_opaque_nopad_f()
429 *
430 * Add "len" bytes of opaque data pointed to by "buf" to the given chain.
431 * Do not XDR pad.
432 */
433int
434nfsm_chain_add_opaque_nopad_f(struct nfsm_chain *nmc, const u_char *buf, uint32_t len)
435{
436 uint32_t tlen;
437 int error;
438
439 while (len > 0) {
440 if (nmc->nmc_left <= 0) {
441 error = nfsm_chain_new_mbuf(nmc, len);
442 if (error) {
443 return error;
444 }
445 }
446 tlen = MIN(nmc->nmc_left, len);
447 bcopy(buf, nmc->nmc_ptr, tlen);
448 nmc->nmc_ptr += tlen;
449 nmc->nmc_left -= tlen;
450 len -= tlen;
451 buf += tlen;
452 }
453 return 0;
454}
455
456/*
457 * nfsm_chain_add_uio()
458 *
459 * Add "len" bytes of data from "uio" to the given chain.
460 */
461int
462nfsm_chain_add_uio(struct nfsm_chain *nmc, uio_t uio, uint32_t len)
463{
464 uint32_t paddedlen, tlen;
465 int error;
466
467 paddedlen = nfsm_rndup(len);
468
469 while (paddedlen) {
470 if (!nmc->nmc_left) {
471 error = nfsm_chain_new_mbuf(nmc, paddedlen);
472 if (error) {
473 return error;
474 }
475 }
476 tlen = MIN(nmc->nmc_left, paddedlen);
477 if (tlen) {
478 if (len) {
479 if (tlen > len) {
480 tlen = len;
481 }
482 uiomove(nmc->nmc_ptr, tlen, uio);
483 } else {
484 bzero(nmc->nmc_ptr, tlen);
485 }
486 nmc->nmc_ptr += tlen;
487 nmc->nmc_left -= tlen;
488 paddedlen -= tlen;
489 if (len) {
490 len -= tlen;
491 }
492 }
493 }
494 return 0;
495}
496
497/*
498 * Find the length of the NFS mbuf chain
499 * up to the current encoding/decoding offset.
500 */
501int
502nfsm_chain_offset(struct nfsm_chain *nmc)
503{
504 mbuf_t mb;
505 int len = 0;
506
507 for (mb = nmc->nmc_mhead; mb; mb = mbuf_next(mb)) {
508 if (mb == nmc->nmc_mcur) {
509 return len + (nmc->nmc_ptr - (caddr_t) mbuf_data(mb));
510 }
511 len += mbuf_len(mb);
512 }
513
514 return len;
515}
516
517/*
518 * nfsm_chain_advance()
519 *
520 * Advance an nfsm_chain by "len" bytes.
521 */
522int
523nfsm_chain_advance(struct nfsm_chain *nmc, uint32_t len)
524{
525 mbuf_t mb;
526
527 while (len) {
528 if (nmc->nmc_left >= len) {
529 nmc->nmc_left -= len;
530 nmc->nmc_ptr += len;
531 return 0;
532 }
533 len -= nmc->nmc_left;
534 nmc->nmc_mcur = mb = mbuf_next(nmc->nmc_mcur);
535 if (!mb) {
536 return EBADRPC;
537 }
538 nmc->nmc_ptr = mbuf_data(mb);
539 nmc->nmc_left = mbuf_len(mb);
540 }
541
542 return 0;
543}
544
545/*
546 * nfsm_chain_reverse()
547 *
548 * Reverse decode offset in an nfsm_chain by "len" bytes.
549 */
550int
551nfsm_chain_reverse(struct nfsm_chain *nmc, uint32_t len)
552{
553 uint32_t mlen, new_offset;
554 int error = 0;
555
556 mlen = nmc->nmc_ptr - (caddr_t) mbuf_data(nmc->nmc_mcur);
557 if (len <= mlen) {
558 nmc->nmc_ptr -= len;
559 nmc->nmc_left += len;
560 return 0;
561 }
562
563 new_offset = nfsm_chain_offset(nmc) - len;
564 nfsm_chain_dissect_init(error, nmc, nmc->nmc_mhead);
565 if (error) {
566 return error;
567 }
568
569 return nfsm_chain_advance(nmc, new_offset);
570}
571
572/*
573 * nfsm_chain_get_opaque_pointer_f()
574 *
575 * Return a pointer to the next "len" bytes of contiguous data in
576 * the mbuf chain. If the next "len" bytes are not contiguous, we
577 * try to manipulate the mbuf chain so that it is.
578 *
579 * The nfsm_chain is advanced by nfsm_rndup("len") bytes.
580 */
581int
582nfsm_chain_get_opaque_pointer_f(struct nfsm_chain *nmc, uint32_t len, u_char **pptr)
583{
584 mbuf_t mbcur, mb;
585 uint32_t left, need, mblen, cplen, padlen;
586 u_char *ptr;
587 int error = 0;
588
589 /* move to next mbuf with data */
590 while (nmc->nmc_mcur && (nmc->nmc_left == 0)) {
591 mb = mbuf_next(nmc->nmc_mcur);
592 nmc->nmc_mcur = mb;
593 if (!mb) {
594 break;
595 }
596 nmc->nmc_ptr = mbuf_data(mb);
597 nmc->nmc_left = mbuf_len(mb);
598 }
599 /* check if we've run out of data */
600 if (!nmc->nmc_mcur) {
601 return EBADRPC;
602 }
603
604 /* do we already have a contiguous buffer? */
605 if (nmc->nmc_left >= len) {
606 /* the returned pointer will be the current pointer */
607 *pptr = (u_char*)nmc->nmc_ptr;
608 error = nfsm_chain_advance(nmc, nfsm_rndup(len));
609 return error;
610 }
611
612 padlen = nfsm_rndup(len) - len;
613
614 /* we need (len - left) more bytes */
615 mbcur = nmc->nmc_mcur;
616 left = nmc->nmc_left;
617 need = len - left;
618
619 if (need > mbuf_trailingspace(mbcur)) {
620 /*
621 * The needed bytes won't fit in the current mbuf so we'll
622 * allocate a new mbuf to hold the contiguous range of data.
623 */
624 nfsm_mbuf_get(error, &mb, len);
625 if (error) {
626 return error;
627 }
628 /* double check that this mbuf can hold all the data */
629 if (mbuf_maxlen(mb) < len) {
630 mbuf_free(mb);
631 return EOVERFLOW;
632 }
633
634 /* the returned pointer will be the new mbuf's data pointer */
635 *pptr = ptr = mbuf_data(mb);
636
637 /* copy "left" bytes to the new mbuf */
638 bcopy(nmc->nmc_ptr, ptr, left);
639 ptr += left;
640 mbuf_setlen(mb, left);
641
642 /* insert the new mbuf between the current and next mbufs */
643 error = mbuf_setnext(mb, mbuf_next(mbcur));
644 if (!error) {
645 error = mbuf_setnext(mbcur, mb);
646 }
647 if (error) {
648 mbuf_free(mb);
649 return error;
650 }
651
652 /* reduce current mbuf's length by "left" */
653 mbuf_setlen(mbcur, mbuf_len(mbcur) - left);
654
655 /*
656 * update nmc's state to point at the end of the mbuf
657 * where the needed data will be copied to.
658 */
659 nmc->nmc_mcur = mbcur = mb;
660 nmc->nmc_left = 0;
661 nmc->nmc_ptr = (caddr_t)ptr;
662 } else {
663 /* The rest of the data will fit in this mbuf. */
664
665 /* the returned pointer will be the current pointer */
666 *pptr = (u_char*)nmc->nmc_ptr;
667
668 /*
669 * update nmc's state to point at the end of the mbuf
670 * where the needed data will be copied to.
671 */
672 nmc->nmc_ptr += left;
673 nmc->nmc_left = 0;
674 }
675
676 /*
677 * move the next "need" bytes into the current
678 * mbuf from the mbufs that follow
679 */
680
681 /* extend current mbuf length */
682 mbuf_setlen(mbcur, mbuf_len(mbcur) + need);
683
684 /* mb follows mbufs we're copying/compacting data from */
685 mb = mbuf_next(mbcur);
686
687 while (need && mb) {
688 /* copy as much as we need/can */
689 ptr = mbuf_data(mb);
690 mblen = mbuf_len(mb);
691 cplen = MIN(mblen, need);
692 if (cplen) {
693 bcopy(ptr, nmc->nmc_ptr, cplen);
694 /*
695 * update the mbuf's pointer and length to reflect that
696 * the data was shifted to an earlier mbuf in the chain
697 */
698 error = mbuf_setdata(mb, ptr + cplen, mblen - cplen);
699 if (error) {
700 mbuf_setlen(mbcur, mbuf_len(mbcur) - need);
701 return error;
702 }
703 /* update pointer/need */
704 nmc->nmc_ptr += cplen;
705 need -= cplen;
706 }
707 /* if more needed, go to next mbuf */
708 if (need) {
709 mb = mbuf_next(mb);
710 }
711 }
712
713 /* did we run out of data in the mbuf chain? */
714 if (need) {
715 mbuf_setlen(mbcur, mbuf_len(mbcur) - need);
716 return EBADRPC;
717 }
718
719 /*
720 * update nmc's state to point after this contiguous data
721 *
722 * "mb" points to the last mbuf we copied data from so we
723 * just set nmc to point at whatever remains in that mbuf.
724 */
725 nmc->nmc_mcur = mb;
726 nmc->nmc_ptr = mbuf_data(mb);
727 nmc->nmc_left = mbuf_len(mb);
728
729 /* move past any padding */
730 if (padlen) {
731 error = nfsm_chain_advance(nmc, padlen);
732 }
733
734 return error;
735}
736
737/*
738 * nfsm_chain_get_opaque_f()
739 *
740 * Read the next "len" bytes in the chain into "buf".
741 * The nfsm_chain is advanced by nfsm_rndup("len") bytes.
742 */
743int
744nfsm_chain_get_opaque_f(struct nfsm_chain *nmc, uint32_t len, u_char *buf)
745{
746 uint32_t cplen, padlen;
747 int error = 0;
748
749 padlen = nfsm_rndup(len) - len;
750
751 /* loop through mbufs copying all the data we need */
752 while (len && nmc->nmc_mcur) {
753 /* copy as much as we need/can */
754 cplen = MIN(nmc->nmc_left, len);
755 if (cplen) {
756 bcopy(nmc->nmc_ptr, buf, cplen);
757 nmc->nmc_ptr += cplen;
758 nmc->nmc_left -= cplen;
759 buf += cplen;
760 len -= cplen;
761 }
762 /* if more needed, go to next mbuf */
763 if (len) {
764 mbuf_t mb = mbuf_next(nmc->nmc_mcur);
765 nmc->nmc_mcur = mb;
766 nmc->nmc_ptr = mb ? mbuf_data(mb) : NULL;
767 nmc->nmc_left = mb ? mbuf_len(mb) : 0;
768 }
769 }
770
771 /* did we run out of data in the mbuf chain? */
772 if (len) {
773 return EBADRPC;
774 }
775
776 if (padlen) {
777 nfsm_chain_adv(error, nmc, padlen);
778 }
779
780 return error;
781}
782
783/*
784 * nfsm_chain_get_uio()
785 *
786 * Read the next "len" bytes in the chain into the given uio.
787 * The nfsm_chain is advanced by nfsm_rndup("len") bytes.
788 */
789int
790nfsm_chain_get_uio(struct nfsm_chain *nmc, uint32_t len, uio_t uio)
791{
792 uint32_t cplen, padlen;
793 int error = 0;
794
795 padlen = nfsm_rndup(len) - len;
796
797 /* loop through mbufs copying all the data we need */
798 while (len && nmc->nmc_mcur) {
799 /* copy as much as we need/can */
800 cplen = MIN(nmc->nmc_left, len);
801 if (cplen) {
802 error = uiomove(nmc->nmc_ptr, cplen, uio);
803 if (error) {
804 return error;
805 }
806 nmc->nmc_ptr += cplen;
807 nmc->nmc_left -= cplen;
808 len -= cplen;
809 }
810 /* if more needed, go to next mbuf */
811 if (len) {
812 mbuf_t mb = mbuf_next(nmc->nmc_mcur);
813 nmc->nmc_mcur = mb;
814 nmc->nmc_ptr = mb ? mbuf_data(mb) : NULL;
815 nmc->nmc_left = mb ? mbuf_len(mb) : 0;
816 }
817 }
818
819 /* did we run out of data in the mbuf chain? */
820 if (len) {
821 return EBADRPC;
822 }
823
824 if (padlen) {
825 nfsm_chain_adv(error, nmc, padlen);
826 }
827
828 return error;
829}
830
831#if NFSCLIENT
832
833int
834nfsm_chain_add_string_nfc(struct nfsm_chain *nmc, const uint8_t *s, uint32_t slen)
835{
836 uint8_t smallbuf[64];
837 uint8_t *nfcname = smallbuf;
838 size_t buflen = sizeof(smallbuf), nfclen;
839 int error;
840
841 error = utf8_normalizestr(s, slen, nfcname, &nfclen, buflen, UTF_PRECOMPOSED | UTF_NO_NULL_TERM);
842 if (error == ENAMETOOLONG) {
843 buflen = MAXPATHLEN;
844 MALLOC_ZONE(nfcname, uint8_t *, MAXPATHLEN, M_NAMEI, M_WAITOK);
845 if (nfcname) {
846 error = utf8_normalizestr(s, slen, nfcname, &nfclen, buflen, UTF_PRECOMPOSED | UTF_NO_NULL_TERM);
847 }
848 }
849
850 /* if we got an error, just use the original string */
851 if (error) {
852 nfsm_chain_add_string(error, nmc, s, slen);
853 } else {
854 nfsm_chain_add_string(error, nmc, nfcname, nfclen);
855 }
856
857 if (nfcname && (nfcname != smallbuf)) {
858 FREE_ZONE(nfcname, MAXPATHLEN, M_NAMEI);
859 }
860 return error;
861}
862
863/*
864 * Add an NFSv2 "sattr" structure to an mbuf chain
865 */
866int
867nfsm_chain_add_v2sattr_f(struct nfsm_chain *nmc, struct vnode_attr *vap, uint32_t szrdev)
868{
869 int error = 0;
870
871 nfsm_chain_add_32(error, nmc, vtonfsv2_mode(vap->va_type,
872 (VATTR_IS_ACTIVE(vap, va_mode) ? vap->va_mode : 0600)));
873 nfsm_chain_add_32(error, nmc,
874 VATTR_IS_ACTIVE(vap, va_uid) ? vap->va_uid : (uint32_t)-1);
875 nfsm_chain_add_32(error, nmc,
876 VATTR_IS_ACTIVE(vap, va_gid) ? vap->va_gid : (uint32_t)-1);
877 nfsm_chain_add_32(error, nmc, szrdev);
878 nfsm_chain_add_v2time(error, nmc,
879 VATTR_IS_ACTIVE(vap, va_access_time) ?
880 &vap->va_access_time : NULL);
881 nfsm_chain_add_v2time(error, nmc,
882 VATTR_IS_ACTIVE(vap, va_modify_time) ?
883 &vap->va_modify_time : NULL);
884
885 return error;
886}
887
888/*
889 * Add an NFSv3 "sattr" structure to an mbuf chain
890 */
891int
892nfsm_chain_add_v3sattr_f(struct nfsm_chain *nmc, struct vnode_attr *vap)
893{
894 int error = 0;
895
896 if (VATTR_IS_ACTIVE(vap, va_mode)) {
897 nfsm_chain_add_32(error, nmc, TRUE);
898 nfsm_chain_add_32(error, nmc, vap->va_mode);
899 } else {
900 nfsm_chain_add_32(error, nmc, FALSE);
901 }
902 if (VATTR_IS_ACTIVE(vap, va_uid)) {
903 nfsm_chain_add_32(error, nmc, TRUE);
904 nfsm_chain_add_32(error, nmc, vap->va_uid);
905 } else {
906 nfsm_chain_add_32(error, nmc, FALSE);
907 }
908 if (VATTR_IS_ACTIVE(vap, va_gid)) {
909 nfsm_chain_add_32(error, nmc, TRUE);
910 nfsm_chain_add_32(error, nmc, vap->va_gid);
911 } else {
912 nfsm_chain_add_32(error, nmc, FALSE);
913 }
914 if (VATTR_IS_ACTIVE(vap, va_data_size)) {
915 nfsm_chain_add_32(error, nmc, TRUE);
916 nfsm_chain_add_64(error, nmc, vap->va_data_size);
917 } else {
918 nfsm_chain_add_32(error, nmc, FALSE);
919 }
920 if (vap->va_vaflags & VA_UTIMES_NULL) {
921 nfsm_chain_add_32(error, nmc, NFS_TIME_SET_TO_SERVER);
922 nfsm_chain_add_32(error, nmc, NFS_TIME_SET_TO_SERVER);
923 } else {
924 if (VATTR_IS_ACTIVE(vap, va_access_time)) {
925 nfsm_chain_add_32(error, nmc, NFS_TIME_SET_TO_CLIENT);
926 nfsm_chain_add_32(error, nmc, vap->va_access_time.tv_sec);
927 nfsm_chain_add_32(error, nmc, vap->va_access_time.tv_nsec);
928 } else {
929 nfsm_chain_add_32(error, nmc, NFS_TIME_DONT_CHANGE);
930 }
931 if (VATTR_IS_ACTIVE(vap, va_modify_time)) {
932 nfsm_chain_add_32(error, nmc, NFS_TIME_SET_TO_CLIENT);
933 nfsm_chain_add_32(error, nmc, vap->va_modify_time.tv_sec);
934 nfsm_chain_add_32(error, nmc, vap->va_modify_time.tv_nsec);
935 } else {
936 nfsm_chain_add_32(error, nmc, NFS_TIME_DONT_CHANGE);
937 }
938 }
939
940 return error;
941}
942
943
944/*
945 * nfsm_chain_get_fh_attr()
946 *
947 * Get the file handle and attributes from an mbuf chain. (NFSv2/v3)
948 */
949int
950nfsm_chain_get_fh_attr(
951 struct nfsm_chain *nmc,
952 nfsnode_t dnp,
953 vfs_context_t ctx,
954 int nfsvers,
955 uint64_t *xidp,
956 fhandle_t *fhp,
957 struct nfs_vattr *nvap)
958{
959 int error = 0, gotfh, gotattr;
960
961 gotfh = gotattr = 1;
962
963 if (nfsvers == NFS_VER3) { /* check for file handle */
964 nfsm_chain_get_32(error, nmc, gotfh);
965 }
966 if (!error && gotfh) { /* get file handle */
967 nfsm_chain_get_fh(error, nmc, nfsvers, fhp);
968 } else {
969 fhp->fh_len = 0;
970 }
971 if (nfsvers == NFS_VER3) { /* check for file attributes */
972 nfsm_chain_get_32(error, nmc, gotattr);
973 }
974 nfsmout_if(error);
975 if (gotattr) {
976 if (!gotfh) { /* skip attributes */
977 nfsm_chain_adv(error, nmc, NFSX_V3FATTR);
978 } else { /* get attributes */
979 error = nfs_parsefattr(nmc, nfsvers, nvap);
980 }
981 } else if (gotfh) {
982 /* we need valid attributes in order to call nfs_nget() */
983 if (nfs3_getattr_rpc(NULL, NFSTOMP(dnp), fhp->fh_data, fhp->fh_len, 0, ctx, nvap, xidp)) {
984 gotattr = 0;
985 fhp->fh_len = 0;
986 }
987 }
988nfsmout:
989 return error;
990}
991
992/*
993 * Get and process NFSv3 WCC data from an mbuf chain
994 */
995int
996nfsm_chain_get_wcc_data_f(
997 struct nfsm_chain *nmc,
998 nfsnode_t np,
999 struct timespec *premtime,
1000 int *newpostattr,
1001 u_int64_t *xidp)
1002{
1003 int error = 0;
1004 uint32_t flag = 0;
1005
1006 nfsm_chain_get_32(error, nmc, flag);
1007 if (!error && flag) {
1008 nfsm_chain_adv(error, nmc, 2 * NFSX_UNSIGNED);
1009 nfsm_chain_get_32(error, nmc, premtime->tv_sec);
1010 nfsm_chain_get_32(error, nmc, premtime->tv_nsec);
1011 nfsm_chain_adv(error, nmc, 2 * NFSX_UNSIGNED);
1012 } else {
1013 premtime->tv_sec = 0;
1014 premtime->tv_nsec = 0;
1015 }
1016 nfsm_chain_postop_attr_update_flag(error, nmc, np, *newpostattr, xidp);
1017
1018 return error;
1019}
1020
1021/*
1022 * Get the next RPC transaction ID (XID)
1023 */
1024void
1025nfs_get_xid(uint64_t *xidp)
1026{
1027 struct timeval tv;
1028
1029 lck_mtx_lock(nfs_request_mutex);
1030 if (!nfs_xid) {
1031 /*
1032 * Derive initial xid from system time.
1033 *
1034 * Note: it's OK if this code inits nfs_xid to 0 (for example,
1035 * due to a broken clock) because we immediately increment it
1036 * and we guarantee to never use xid 0. So, nfs_xid should only
1037 * ever be 0 the first time this function is called.
1038 */
1039 microtime(&tv);
1040 nfs_xid = tv.tv_sec << 12;
1041 }
1042 if (++nfs_xid == 0) {
1043 /* Skip zero xid if it should ever happen. */
1044 nfs_xidwrap++;
1045 nfs_xid++;
1046 }
1047 *xidp = nfs_xid + ((uint64_t)nfs_xidwrap << 32);
1048 lck_mtx_unlock(nfs_request_mutex);
1049}
1050
1051/*
1052 * Build the RPC header and fill in the authorization info.
1053 * Returns the head of the mbuf list and the xid.
1054 */
1055
1056int
1057nfsm_rpchead(
1058 struct nfsreq *req,
1059 mbuf_t mrest,
1060 u_int64_t *xidp,
1061 mbuf_t *mreqp)
1062{
1063 struct nfsmount *nmp = req->r_nmp;
1064 int nfsvers = nmp->nm_vers;
1065 int proc = ((nfsvers == NFS_VER2) ? nfsv2_procid[req->r_procnum] : (int)req->r_procnum);
1066
1067 return nfsm_rpchead2(nmp, nmp->nm_sotype, NFS_PROG, nfsvers, proc,
1068 req->r_auth, req->r_cred, req, mrest, xidp, mreqp);
1069}
1070
1071/*
1072 * get_auiliary_groups: Gets the supplementary groups from a credential.
1073 *
1074 * IN: cred: credential to get the associated groups from.
1075 * OUT: groups: An array of gids of NGROUPS size.
1076 * IN: count: The number of groups to get; i.e.; the number of groups the server supports
1077 *
1078 * returns: The number of groups found.
1079 *
1080 * Just a wrapper around kauth_cred_getgroups to handle the case of a server supporting less
1081 * than NGROUPS.
1082 */
1083static int
1084get_auxiliary_groups(kauth_cred_t cred, gid_t groups[NGROUPS], int count)
1085{
1086 gid_t pgid;
1087 int maxcount = count < NGROUPS ? count + 1 : NGROUPS;
1088 int i;
1089
1090 for (i = 0; i < NGROUPS; i++) {
1091 groups[i] = -2; /* Initialize to the nobody group */
1092 }
1093 (void)kauth_cred_getgroups(cred, groups, &maxcount);
1094 if (maxcount < 1) {
1095 return maxcount;
1096 }
1097
1098 /*
1099 * kauth_get_groups returns the primary group followed by the
1100 * users auxiliary groups. If the number of groups the server supports
1101 * is less than NGROUPS, then we will drop the first group so that
1102 * we can send one more group over the wire.
1103 */
1104
1105
1106 if (count < NGROUPS) {
1107 pgid = kauth_cred_getgid(cred);
1108 if (pgid == groups[0]) {
1109 maxcount -= 1;
1110 for (i = 0; i < maxcount; i++) {
1111 groups[i] = groups[i + 1];
1112 }
1113 }
1114 }
1115
1116 return maxcount;
1117}
1118
1119int
1120nfsm_rpchead2(struct nfsmount *nmp, int sotype, int prog, int vers, int proc, int auth_type,
1121 kauth_cred_t cred, struct nfsreq *req, mbuf_t mrest, u_int64_t *xidp, mbuf_t *mreqp)
1122{
1123 mbuf_t mreq, mb;
1124 int error, i, auth_len = 0, authsiz, reqlen;
1125 size_t headlen;
1126 struct nfsm_chain nmreq;
1127 gid_t grouplist[NGROUPS];
1128 int groupcount;
1129
1130 /* calculate expected auth length */
1131 switch (auth_type) {
1132 case RPCAUTH_NONE:
1133 auth_len = 0;
1134 break;
1135 case RPCAUTH_SYS:
1136 {
1137 int count = nmp->nm_numgrps < NGROUPS ? nmp->nm_numgrps : NGROUPS;
1138
1139 if (!cred) {
1140 return EINVAL;
1141 }
1142 groupcount = get_auxiliary_groups(cred, grouplist, count);
1143 if (groupcount < 0) {
1144 return EINVAL;
1145 }
1146 auth_len = ((uint32_t)groupcount + 5) * NFSX_UNSIGNED;
1147 break;
1148 }
1149 case RPCAUTH_KRB5:
1150 case RPCAUTH_KRB5I:
1151 case RPCAUTH_KRB5P:
1152 if (!req || !cred) {
1153 return EINVAL;
1154 }
1155 auth_len = 5 * NFSX_UNSIGNED + 0; // zero context handle for now
1156 break;
1157 default:
1158 return EINVAL;
1159 }
1160 authsiz = nfsm_rndup(auth_len);
1161
1162 /* allocate the packet */
1163 headlen = authsiz + 10 * NFSX_UNSIGNED;
1164 if (sotype == SOCK_STREAM) { /* also include room for any RPC Record Mark */
1165 headlen += NFSX_UNSIGNED;
1166 }
1167 if (headlen >= nfs_mbuf_minclsize) {
1168 error = mbuf_getpacket(MBUF_WAITOK, &mreq);
1169 } else {
1170 error = mbuf_gethdr(MBUF_WAITOK, MBUF_TYPE_DATA, &mreq);
1171 if (!error) {
1172 if (headlen < nfs_mbuf_mhlen) {
1173 mbuf_align_32(mreq, headlen);
1174 } else {
1175 mbuf_align_32(mreq, 8 * NFSX_UNSIGNED);
1176 }
1177 }
1178 }
1179 if (error) {
1180 /* unable to allocate packet */
1181 /* XXX should we keep statistics for these errors? */
1182 return error;
1183 }
1184
1185 /*
1186 * If the caller gave us a non-zero XID then use it because
1187 * it may be a higher-level resend with a GSSAPI credential.
1188 * Otherwise, allocate a new one.
1189 */
1190 if (*xidp == 0) {
1191 nfs_get_xid(xidp);
1192 }
1193
1194 /* build the header(s) */
1195 nfsm_chain_init(&nmreq, mreq);
1196
1197 /* First, if it's a TCP stream insert space for an RPC record mark */
1198 if (sotype == SOCK_STREAM) {
1199 nfsm_chain_add_32(error, &nmreq, 0);
1200 }
1201
1202 /* Then the RPC header. */
1203 nfsm_chain_add_32(error, &nmreq, (*xidp & 0xffffffff));
1204 nfsm_chain_add_32(error, &nmreq, RPC_CALL);
1205 nfsm_chain_add_32(error, &nmreq, RPC_VER2);
1206 nfsm_chain_add_32(error, &nmreq, prog);
1207 nfsm_chain_add_32(error, &nmreq, vers);
1208 nfsm_chain_add_32(error, &nmreq, proc);
1209
1210add_cred:
1211 switch (auth_type) {
1212 case RPCAUTH_NONE:
1213 nfsm_chain_add_32(error, &nmreq, RPCAUTH_NONE); /* auth */
1214 nfsm_chain_add_32(error, &nmreq, 0); /* length */
1215 nfsm_chain_add_32(error, &nmreq, RPCAUTH_NONE); /* verf */
1216 nfsm_chain_add_32(error, &nmreq, 0); /* length */
1217 nfsm_chain_build_done(error, &nmreq);
1218 /* Append the args mbufs */
1219 if (!error) {
1220 error = mbuf_setnext(nmreq.nmc_mcur, mrest);
1221 }
1222 break;
1223 case RPCAUTH_SYS: {
1224 nfsm_chain_add_32(error, &nmreq, RPCAUTH_SYS);
1225 nfsm_chain_add_32(error, &nmreq, authsiz);
1226 nfsm_chain_add_32(error, &nmreq, 0); /* stamp */
1227 nfsm_chain_add_32(error, &nmreq, 0); /* zero-length hostname */
1228 nfsm_chain_add_32(error, &nmreq, kauth_cred_getuid(cred)); /* UID */
1229 nfsm_chain_add_32(error, &nmreq, kauth_cred_getgid(cred)); /* GID */
1230 nfsm_chain_add_32(error, &nmreq, groupcount);/* additional GIDs */
1231 for (i = 0; i < groupcount; i++) {
1232 nfsm_chain_add_32(error, &nmreq, grouplist[i]);
1233 }
1234
1235 /* And the verifier... */
1236 nfsm_chain_add_32(error, &nmreq, RPCAUTH_NONE); /* flavor */
1237 nfsm_chain_add_32(error, &nmreq, 0); /* length */
1238 nfsm_chain_build_done(error, &nmreq);
1239
1240 /* Append the args mbufs */
1241 if (!error) {
1242 error = mbuf_setnext(nmreq.nmc_mcur, mrest);
1243 }
1244 break;
1245 }
1246 case RPCAUTH_KRB5:
1247 case RPCAUTH_KRB5I:
1248 case RPCAUTH_KRB5P:
1249 error = nfs_gss_clnt_cred_put(req, &nmreq, mrest);
1250 if (error == ENEEDAUTH) {
1251 int count = nmp->nm_numgrps < NGROUPS ? nmp->nm_numgrps : NGROUPS;
1252
1253 /*
1254 * Use sec=sys for this user
1255 */
1256 error = 0;
1257 req->r_auth = auth_type = RPCAUTH_SYS;
1258 groupcount = get_auxiliary_groups(cred, grouplist, count);
1259 if (groupcount < 0) {
1260 return EINVAL;
1261 }
1262 auth_len = ((uint32_t)groupcount + 5) * NFSX_UNSIGNED;
1263 authsiz = nfsm_rndup(auth_len);
1264 goto add_cred;
1265 }
1266 break;
1267 }
1268 ;
1269
1270 /* finish setting up the packet */
1271 if (!error) {
1272 error = mbuf_pkthdr_setrcvif(mreq, 0);
1273 }
1274
1275 if (error) {
1276 mbuf_freem(mreq);
1277 return error;
1278 }
1279
1280 /* Calculate the size of the request */
1281 reqlen = 0;
1282 for (mb = nmreq.nmc_mhead; mb; mb = mbuf_next(mb)) {
1283 reqlen += mbuf_len(mb);
1284 }
1285
1286 mbuf_pkthdr_setlen(mreq, reqlen);
1287
1288 /*
1289 * If the request goes on a TCP stream,
1290 * set its size in the RPC record mark.
1291 * The record mark count doesn't include itself
1292 * and the last fragment bit is set.
1293 */
1294 if (sotype == SOCK_STREAM) {
1295 nfsm_chain_set_recmark(error, &nmreq,
1296 (reqlen - NFSX_UNSIGNED) | 0x80000000);
1297 }
1298
1299 *mreqp = mreq;
1300 return 0;
1301}
1302
1303/*
1304 * Parse an NFS file attribute structure out of an mbuf chain.
1305 */
1306int
1307nfs_parsefattr(struct nfsm_chain *nmc, int nfsvers, struct nfs_vattr *nvap)
1308{
1309 int error = 0;
1310 enum vtype vtype;
1311 nfstype nvtype;
1312 u_short vmode;
1313 uint32_t val, val2;
1314 dev_t rdev;
1315
1316 val = val2 = 0;
1317 NVATTR_INIT(nvap);
1318
1319 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_TYPE);
1320 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_MODE);
1321 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_NUMLINKS);
1322 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_OWNER);
1323 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_OWNER_GROUP);
1324 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_SIZE);
1325 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_SPACE_USED);
1326 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_RAWDEV);
1327 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_FSID);
1328 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_FILEID);
1329 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_TIME_ACCESS);
1330 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_TIME_MODIFY);
1331 NFS_BITMAP_SET(nvap->nva_bitmap, NFS_FATTR_TIME_METADATA);
1332
1333 nfsm_chain_get_32(error, nmc, nvtype);
1334 nfsm_chain_get_32(error, nmc, vmode);
1335 nfsmout_if(error);
1336
1337 if (nfsvers == NFS_VER3) {
1338 nvap->nva_type = vtype = nfstov_type(nvtype, nfsvers);
1339 } else {
1340 /*
1341 * The duplicate information returned in fa_type and fa_mode
1342 * is an ambiguity in the NFS version 2 protocol.
1343 *
1344 * VREG should be taken literally as a regular file. If a
1345 * server intends to return some type information differently
1346 * in the upper bits of the mode field (e.g. for sockets, or
1347 * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we
1348 * leave the examination of the mode bits even in the VREG
1349 * case to avoid breakage for bogus servers, but we make sure
1350 * that there are actually type bits set in the upper part of
1351 * fa_mode (and failing that, trust the va_type field).
1352 *
1353 * NFSv3 cleared the issue, and requires fa_mode to not
1354 * contain any type information (while also introducing
1355 * sockets and FIFOs for fa_type).
1356 */
1357 vtype = nfstov_type(nvtype, nfsvers);
1358 if ((vtype == VNON) || ((vtype == VREG) && ((vmode & S_IFMT) != 0))) {
1359 vtype = IFTOVT(vmode);
1360 }
1361 nvap->nva_type = vtype;
1362 }
1363
1364 nvap->nva_mode = (vmode & 07777);
1365
1366 nfsm_chain_get_32(error, nmc, nvap->nva_nlink);
1367 nfsm_chain_get_32(error, nmc, nvap->nva_uid);
1368 nfsm_chain_get_32(error, nmc, nvap->nva_gid);
1369
1370 if (nfsvers == NFS_VER3) {
1371 nfsm_chain_get_64(error, nmc, nvap->nva_size);
1372 nfsm_chain_get_64(error, nmc, nvap->nva_bytes);
1373 nfsm_chain_get_32(error, nmc, nvap->nva_rawdev.specdata1);
1374 nfsm_chain_get_32(error, nmc, nvap->nva_rawdev.specdata2);
1375 nfsmout_if(error);
1376 nfsm_chain_get_64(error, nmc, nvap->nva_fsid.major);
1377 nvap->nva_fsid.minor = 0;
1378 nfsm_chain_get_64(error, nmc, nvap->nva_fileid);
1379 } else {
1380 nfsm_chain_get_32(error, nmc, nvap->nva_size);
1381 nfsm_chain_adv(error, nmc, NFSX_UNSIGNED);
1382 nfsm_chain_get_32(error, nmc, rdev);
1383 nfsmout_if(error);
1384 nvap->nva_rawdev.specdata1 = major(rdev);
1385 nvap->nva_rawdev.specdata2 = minor(rdev);
1386 nfsm_chain_get_32(error, nmc, val); /* blocks */
1387 nfsmout_if(error);
1388 nvap->nva_bytes = val * NFS_FABLKSIZE;
1389 nfsm_chain_get_32(error, nmc, val);
1390 nfsmout_if(error);
1391 nvap->nva_fsid.major = (uint64_t)val;
1392 nvap->nva_fsid.minor = 0;
1393 nfsm_chain_get_32(error, nmc, val);
1394 nfsmout_if(error);
1395 nvap->nva_fileid = (uint64_t)val;
1396 /* Really ugly NFSv2 kludge. */
1397 if ((vtype == VCHR) && (rdev == (dev_t)0xffffffff)) {
1398 nvap->nva_type = VFIFO;
1399 }
1400 }
1401 nfsm_chain_get_time(error, nmc, nfsvers,
1402 nvap->nva_timesec[NFSTIME_ACCESS],
1403 nvap->nva_timensec[NFSTIME_ACCESS]);
1404 nfsm_chain_get_time(error, nmc, nfsvers,
1405 nvap->nva_timesec[NFSTIME_MODIFY],
1406 nvap->nva_timensec[NFSTIME_MODIFY]);
1407 nfsm_chain_get_time(error, nmc, nfsvers,
1408 nvap->nva_timesec[NFSTIME_CHANGE],
1409 nvap->nva_timensec[NFSTIME_CHANGE]);
1410nfsmout:
1411 return error;
1412}
1413
1414/*
1415 * Load the attribute cache (that lives in the nfsnode entry) with
1416 * the value pointed to by nvap, unless the file type in the attribute
1417 * cache doesn't match the file type in the nvap, in which case log a
1418 * warning and return ESTALE.
1419 *
1420 * If the dontshrink flag is set, then it's not safe to call ubc_setsize()
1421 * to shrink the size of the file.
1422 */
1423int
1424nfs_loadattrcache(
1425 nfsnode_t np,
1426 struct nfs_vattr *nvap,
1427 u_int64_t *xidp,
1428 int dontshrink)
1429{
1430 mount_t mp;
1431 vnode_t vp;
1432 struct timeval now;
1433 struct nfs_vattr *npnvap;
1434 int xattr = np->n_vattr.nva_flags & NFS_FFLAG_IS_ATTR;
1435 int referral = np->n_vattr.nva_flags & NFS_FFLAG_TRIGGER_REFERRAL;
1436 int aclbit, monitored, error = 0;
1437 kauth_acl_t acl;
1438 struct nfsmount *nmp;
1439 uint32_t events = np->n_events;
1440
1441 if (np->n_hflag & NHINIT) {
1442 vp = NULL;
1443 mp = np->n_mount;
1444 } else {
1445 vp = NFSTOV(np);
1446 mp = vnode_mount(vp);
1447 }
1448 monitored = vp ? vnode_ismonitored(vp) : 0;
1449
1450 FSDBG_TOP(527, np, vp, *xidp >> 32, *xidp);
1451
1452 if (!((nmp = VFSTONFS(mp)))) {
1453 FSDBG_BOT(527, ENXIO, 1, 0, *xidp);
1454 return ENXIO;
1455 }
1456
1457 if (*xidp < np->n_xid) {
1458 /*
1459 * We have already updated attributes with a response from
1460 * a later request. The attributes we have here are probably
1461 * stale so we drop them (just return). However, our
1462 * out-of-order receipt could be correct - if the requests were
1463 * processed out of order at the server. Given the uncertainty
1464 * we invalidate our cached attributes. *xidp is zeroed here
1465 * to indicate the attributes were dropped - only getattr
1466 * cares - it needs to retry the rpc.
1467 */
1468 NATTRINVALIDATE(np);
1469 FSDBG_BOT(527, 0, np, np->n_xid, *xidp);
1470 *xidp = 0;
1471 return 0;
1472 }
1473
1474 if (vp && (nvap->nva_type != vnode_vtype(vp))) {
1475 /*
1476 * The filehandle has changed type on us. This can be
1477 * caused by either the server not having unique filehandles
1478 * or because another client has removed the previous
1479 * filehandle and a new object (of a different type)
1480 * has been created with the same filehandle.
1481 *
1482 * We can't simply switch the type on the vnode because
1483 * there may be type-specific fields that need to be
1484 * cleaned up or set up.
1485 *
1486 * So, what should we do with this vnode?
1487 *
1488 * About the best we can do is log a warning and return
1489 * an error. ESTALE is about the closest error, but it
1490 * is a little strange that we come up with this error
1491 * internally instead of simply passing it through from
1492 * the server. Hopefully, the vnode will be reclaimed
1493 * soon so the filehandle can be reincarnated as the new
1494 * object type.
1495 */
1496 printf("nfs loadattrcache vnode changed type, was %d now %d\n",
1497 vnode_vtype(vp), nvap->nva_type);
1498 error = ESTALE;
1499 if (monitored) {
1500 events |= VNODE_EVENT_DELETE;
1501 }
1502 goto out;
1503 }
1504
1505 npnvap = &np->n_vattr;
1506
1507 /*
1508 * The ACL cache needs special handling because it is not
1509 * always updated. Save current ACL cache state so it can
1510 * be restored after copying the new attributes into place.
1511 */
1512 aclbit = NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_ACL);
1513 acl = npnvap->nva_acl;
1514
1515 if (monitored) {
1516 /*
1517 * For monitored nodes, check for attribute changes that should generate events.
1518 */
1519 if (NFS_BITMAP_ISSET(nvap->nva_bitmap, NFS_FATTR_NUMLINKS) &&
1520 (nvap->nva_nlink != npnvap->nva_nlink)) {
1521 events |= VNODE_EVENT_ATTRIB | VNODE_EVENT_LINK;
1522 }
1523 if (events & VNODE_EVENT_PERMS) {
1524 /* no need to do all the checking if it's already set */;
1525 } else if (NFS_BITMAP_ISSET(nvap->nva_bitmap, NFS_FATTR_MODE) &&
1526 (nvap->nva_mode != npnvap->nva_mode)) {
1527 events |= VNODE_EVENT_ATTRIB | VNODE_EVENT_PERMS;
1528 } else if (NFS_BITMAP_ISSET(nvap->nva_bitmap, NFS_FATTR_OWNER) &&
1529 (nvap->nva_uid != npnvap->nva_uid)) {
1530 events |= VNODE_EVENT_ATTRIB | VNODE_EVENT_PERMS;
1531 } else if (NFS_BITMAP_ISSET(nvap->nva_bitmap, NFS_FATTR_OWNER_GROUP) &&
1532 (nvap->nva_gid != npnvap->nva_gid)) {
1533 events |= VNODE_EVENT_ATTRIB | VNODE_EVENT_PERMS;
1534 } else if (nmp->nm_vers >= NFS_VER4) {
1535 if (NFS_BITMAP_ISSET(nvap->nva_bitmap, NFS_FATTR_OWNER) &&
1536 !kauth_guid_equal(&nvap->nva_uuuid, &npnvap->nva_uuuid)) {
1537 events |= VNODE_EVENT_ATTRIB | VNODE_EVENT_PERMS;
1538 } else if (NFS_BITMAP_ISSET(nvap->nva_bitmap, NFS_FATTR_OWNER_GROUP) &&
1539 !kauth_guid_equal(&nvap->nva_guuid, &npnvap->nva_guuid)) {
1540 events |= VNODE_EVENT_ATTRIB | VNODE_EVENT_PERMS;
1541 } else if ((NFS_BITMAP_ISSET(nvap->nva_bitmap, NFS_FATTR_ACL) &&
1542 nvap->nva_acl && npnvap->nva_acl &&
1543 ((nvap->nva_acl->acl_entrycount != npnvap->nva_acl->acl_entrycount) ||
1544 bcmp(nvap->nva_acl, npnvap->nva_acl, KAUTH_ACL_COPYSIZE(nvap->nva_acl))))) {
1545 events |= VNODE_EVENT_ATTRIB | VNODE_EVENT_PERMS;
1546 }
1547 }
1548 if (((nmp->nm_vers >= NFS_VER4) && (nvap->nva_change != npnvap->nva_change)) ||
1549 (NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_TIME_MODIFY) &&
1550 ((nvap->nva_timesec[NFSTIME_MODIFY] != npnvap->nva_timesec[NFSTIME_MODIFY]) ||
1551 (nvap->nva_timensec[NFSTIME_MODIFY] != npnvap->nva_timensec[NFSTIME_MODIFY])))) {
1552 events |= VNODE_EVENT_ATTRIB | VNODE_EVENT_WRITE;
1553 }
1554 if (!events && NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_RAWDEV) &&
1555 ((nvap->nva_rawdev.specdata1 != npnvap->nva_rawdev.specdata1) ||
1556 (nvap->nva_rawdev.specdata2 != npnvap->nva_rawdev.specdata2))) {
1557 events |= VNODE_EVENT_ATTRIB;
1558 }
1559 if (!events && NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_FILEID) &&
1560 (nvap->nva_fileid != npnvap->nva_fileid)) {
1561 events |= VNODE_EVENT_ATTRIB;
1562 }
1563 if (!events && NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_ARCHIVE) &&
1564 ((nvap->nva_flags & NFS_FFLAG_ARCHIVED) != (npnvap->nva_flags & NFS_FFLAG_ARCHIVED))) {
1565 events |= VNODE_EVENT_ATTRIB;
1566 }
1567 if (!events && NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_HIDDEN) &&
1568 ((nvap->nva_flags & NFS_FFLAG_HIDDEN) != (npnvap->nva_flags & NFS_FFLAG_HIDDEN))) {
1569 events |= VNODE_EVENT_ATTRIB;
1570 }
1571 if (!events && NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_TIME_CREATE) &&
1572 ((nvap->nva_timesec[NFSTIME_CREATE] != npnvap->nva_timesec[NFSTIME_CREATE]) ||
1573 (nvap->nva_timensec[NFSTIME_CREATE] != npnvap->nva_timensec[NFSTIME_CREATE]))) {
1574 events |= VNODE_EVENT_ATTRIB;
1575 }
1576 if (!events && NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_TIME_BACKUP) &&
1577 ((nvap->nva_timesec[NFSTIME_BACKUP] != npnvap->nva_timesec[NFSTIME_BACKUP]) ||
1578 (nvap->nva_timensec[NFSTIME_BACKUP] != npnvap->nva_timensec[NFSTIME_BACKUP]))) {
1579 events |= VNODE_EVENT_ATTRIB;
1580 }
1581 }
1582
1583 /* Copy the attributes to the attribute cache */
1584 bcopy((caddr_t)nvap, (caddr_t)npnvap, sizeof(*nvap));
1585
1586 microuptime(&now);
1587 np->n_attrstamp = now.tv_sec;
1588 np->n_xid = *xidp;
1589 /* NFS_FFLAG_IS_ATTR and NFS_FFLAG_TRIGGER_REFERRAL need to be sticky... */
1590 if (vp && xattr) {
1591 nvap->nva_flags |= xattr;
1592 }
1593 if (vp && referral) {
1594 nvap->nva_flags |= referral;
1595 }
1596
1597 if (NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_ACL)) {
1598 /* we're updating the ACL */
1599 if (nvap->nva_acl) {
1600 /* make a copy of the acl for the cache */
1601 npnvap->nva_acl = kauth_acl_alloc(nvap->nva_acl->acl_entrycount);
1602 if (npnvap->nva_acl) {
1603 bcopy(nvap->nva_acl, npnvap->nva_acl, KAUTH_ACL_COPYSIZE(nvap->nva_acl));
1604 } else {
1605 /* can't make a copy to cache, invalidate ACL cache */
1606 NFS_BITMAP_CLR(npnvap->nva_bitmap, NFS_FATTR_ACL);
1607 NACLINVALIDATE(np);
1608 aclbit = 0;
1609 }
1610 }
1611 if (acl) {
1612 kauth_acl_free(acl);
1613 acl = NULL;
1614 }
1615 }
1616 if (NFS_BITMAP_ISSET(npnvap->nva_bitmap, NFS_FATTR_ACL)) {
1617 /* update the ACL timestamp */
1618 np->n_aclstamp = now.tv_sec;
1619 } else {
1620 /* we aren't updating the ACL, so restore original values */
1621 if (aclbit) {
1622 NFS_BITMAP_SET(npnvap->nva_bitmap, NFS_FATTR_ACL);
1623 }
1624 npnvap->nva_acl = acl;
1625 }
1626
1627#if CONFIG_TRIGGERS
1628 /*
1629 * For NFSv4, if the fsid doesn't match the fsid for the mount, then
1630 * this node is for a different file system on the server. So we mark
1631 * this node as a trigger node that will trigger the mirror mount.
1632 */
1633 if ((nmp->nm_vers >= NFS_VER4) && (nvap->nva_type == VDIR) &&
1634 ((np->n_vattr.nva_fsid.major != nmp->nm_fsid.major) ||
1635 (np->n_vattr.nva_fsid.minor != nmp->nm_fsid.minor))) {
1636 np->n_vattr.nva_flags |= NFS_FFLAG_TRIGGER;
1637 }
1638#endif
1639
1640 if (!vp || (nvap->nva_type != VREG)) {
1641 np->n_size = nvap->nva_size;
1642 } else if (nvap->nva_size != np->n_size) {
1643 FSDBG(527, np, nvap->nva_size, np->n_size, (nvap->nva_type == VREG) | (np->n_flag & NMODIFIED ? 6 : 4));
1644 if (!UBCINFOEXISTS(vp) || (dontshrink && (nvap->nva_size < np->n_size))) {
1645 /* asked not to shrink, so stick with current size */
1646 FSDBG(527, np, np->n_size, np->n_vattr.nva_size, 0xf00d0001);
1647 nvap->nva_size = np->n_size;
1648 NATTRINVALIDATE(np);
1649 } else if ((np->n_flag & NMODIFIED) && (nvap->nva_size < np->n_size)) {
1650 /* if we've modified, stick with larger size */
1651 FSDBG(527, np, np->n_size, np->n_vattr.nva_size, 0xf00d0002);
1652 nvap->nva_size = np->n_size;
1653 npnvap->nva_size = np->n_size;
1654 } else {
1655 /*
1656 * n_size is protected by the data lock, so we need to
1657 * defer updating it until it's safe. We save the new size
1658 * and set a flag and it'll get updated the next time we get/drop
1659 * the data lock or the next time we do a getattr.
1660 */
1661 np->n_newsize = nvap->nva_size;
1662 SET(np->n_flag, NUPDATESIZE);
1663 if (monitored) {
1664 events |= VNODE_EVENT_ATTRIB | VNODE_EVENT_EXTEND;
1665 }
1666 }
1667 }
1668
1669 if (np->n_flag & NCHG) {
1670 if (np->n_flag & NACC) {
1671 nvap->nva_timesec[NFSTIME_ACCESS] = np->n_atim.tv_sec;
1672 nvap->nva_timensec[NFSTIME_ACCESS] = np->n_atim.tv_nsec;
1673 }
1674 if (np->n_flag & NUPD) {
1675 nvap->nva_timesec[NFSTIME_MODIFY] = np->n_mtim.tv_sec;
1676 nvap->nva_timensec[NFSTIME_MODIFY] = np->n_mtim.tv_nsec;
1677 }
1678 }
1679
1680out:
1681 if (monitored && events) {
1682 nfs_vnode_notify(np, events);
1683 }
1684 FSDBG_BOT(527, error, np, np->n_size, *xidp);
1685 return error;
1686}
1687
1688/*
1689 * Calculate the attribute timeout based on
1690 * how recently the file has been modified.
1691 */
1692int
1693nfs_attrcachetimeout(nfsnode_t np)
1694{
1695 struct nfsmount *nmp;
1696 struct timeval now;
1697 int isdir;
1698 uint32_t timeo;
1699
1700 nmp = NFSTONMP(np);
1701 if (nfs_mount_gone(nmp)) {
1702 return 0;
1703 }
1704
1705 isdir = vnode_isdir(NFSTOV(np));
1706
1707 if ((nmp->nm_vers >= NFS_VER4) && (np->n_openflags & N_DELEG_MASK)) {
1708 /* If we have a delegation, we always use the max timeout. */
1709 timeo = isdir ? nmp->nm_acdirmax : nmp->nm_acregmax;
1710 } else if ((np)->n_flag & NMODIFIED) {
1711 /* If we have modifications, we always use the min timeout. */
1712 timeo = isdir ? nmp->nm_acdirmin : nmp->nm_acregmin;
1713 } else {
1714 /* Otherwise, we base the timeout on how old the file seems. */
1715 /* Note that if the client and server clocks are way out of sync, */
1716 /* timeout will probably get clamped to a min or max value */
1717 microtime(&now);
1718 timeo = (now.tv_sec - (np)->n_vattr.nva_timesec[NFSTIME_MODIFY]) / 10;
1719 if (isdir) {
1720 if (timeo < nmp->nm_acdirmin) {
1721 timeo = nmp->nm_acdirmin;
1722 } else if (timeo > nmp->nm_acdirmax) {
1723 timeo = nmp->nm_acdirmax;
1724 }
1725 } else {
1726 if (timeo < nmp->nm_acregmin) {
1727 timeo = nmp->nm_acregmin;
1728 } else if (timeo > nmp->nm_acregmax) {
1729 timeo = nmp->nm_acregmax;
1730 }
1731 }
1732 }
1733
1734 return timeo;
1735}
1736
1737/*
1738 * Check the attribute cache time stamp.
1739 * If the cache is valid, copy contents to *nvaper and return 0
1740 * otherwise return an error.
1741 * Must be called with the node locked.
1742 */
1743int
1744nfs_getattrcache(nfsnode_t np, struct nfs_vattr *nvaper, int flags)
1745{
1746 struct nfs_vattr *nvap;
1747 struct timeval nowup;
1748 int32_t timeo;
1749 struct nfsmount *nmp;
1750
1751 /* Check if the attributes are valid. */
1752 if (!NATTRVALID(np) || ((flags & NGA_ACL) && !NACLVALID(np))) {
1753 FSDBG(528, np, 0, 0xffffff01, ENOENT);
1754 OSAddAtomic64(1, &nfsstats.attrcache_misses);
1755 return ENOENT;
1756 }
1757
1758 nmp = NFSTONMP(np);
1759 if (nfs_mount_gone(nmp)) {
1760 return ENXIO;
1761 }
1762 /*
1763 * Verify the cached attributes haven't timed out.
1764 * If the server isn't responding, skip the check
1765 * and return cached attributes.
1766 */
1767 if (!nfs_use_cache(nmp)) {
1768 microuptime(&nowup);
1769 if (np->n_attrstamp > nowup.tv_sec) {
1770 printf("NFS: Attribute time stamp is in the future by %ld seconds. Invalidating cache\n",
1771 np->n_attrstamp - nowup.tv_sec);
1772 NATTRINVALIDATE(np);
1773 NACCESSINVALIDATE(np);
1774 return ENOENT;
1775 }
1776 timeo = nfs_attrcachetimeout(np);
1777 if ((nowup.tv_sec - np->n_attrstamp) >= timeo) {
1778 FSDBG(528, np, 0, 0xffffff02, ENOENT);
1779 OSAddAtomic64(1, &nfsstats.attrcache_misses);
1780 return ENOENT;
1781 }
1782 if ((flags & NGA_ACL) && ((nowup.tv_sec - np->n_aclstamp) >= timeo)) {
1783 FSDBG(528, np, 0, 0xffffff02, ENOENT);
1784 OSAddAtomic64(1, &nfsstats.attrcache_misses);
1785 return ENOENT;
1786 }
1787 }
1788
1789 nvap = &np->n_vattr;
1790 FSDBG(528, np, nvap->nva_size, np->n_size, 0xcace);
1791 OSAddAtomic64(1, &nfsstats.attrcache_hits);
1792
1793 if (nvap->nva_type != VREG) {
1794 np->n_size = nvap->nva_size;
1795 } else if (nvap->nva_size != np->n_size) {
1796 FSDBG(528, np, nvap->nva_size, np->n_size, (nvap->nva_type == VREG) | (np->n_flag & NMODIFIED ? 6 : 4));
1797 if ((np->n_flag & NMODIFIED) && (nvap->nva_size < np->n_size)) {
1798 /* if we've modified, stick with larger size */
1799 nvap->nva_size = np->n_size;
1800 } else {
1801 /*
1802 * n_size is protected by the data lock, so we need to
1803 * defer updating it until it's safe. We save the new size
1804 * and set a flag and it'll get updated the next time we get/drop
1805 * the data lock or the next time we do a getattr.
1806 */
1807 np->n_newsize = nvap->nva_size;
1808 SET(np->n_flag, NUPDATESIZE);
1809 }
1810 }
1811
1812 bcopy((caddr_t)nvap, (caddr_t)nvaper, sizeof(struct nfs_vattr));
1813 if (np->n_flag & NCHG) {
1814 if (np->n_flag & NACC) {
1815 nvaper->nva_timesec[NFSTIME_ACCESS] = np->n_atim.tv_sec;
1816 nvaper->nva_timensec[NFSTIME_ACCESS] = np->n_atim.tv_nsec;
1817 }
1818 if (np->n_flag & NUPD) {
1819 nvaper->nva_timesec[NFSTIME_MODIFY] = np->n_mtim.tv_sec;
1820 nvaper->nva_timensec[NFSTIME_MODIFY] = np->n_mtim.tv_nsec;
1821 }
1822 }
1823 if (nvap->nva_acl) {
1824 if (flags & NGA_ACL) {
1825 nvaper->nva_acl = kauth_acl_alloc(nvap->nva_acl->acl_entrycount);
1826 if (!nvaper->nva_acl) {
1827 return ENOMEM;
1828 }
1829 bcopy(nvap->nva_acl, nvaper->nva_acl, KAUTH_ACL_COPYSIZE(nvap->nva_acl));
1830 } else {
1831 nvaper->nva_acl = NULL;
1832 }
1833 }
1834 return 0;
1835}
1836
1837/*
1838 * When creating file system objects:
1839 * Don't bother setting UID if it's the same as the credential performing the create.
1840 * Don't bother setting GID if it's the same as the directory or credential.
1841 */
1842void
1843nfs_avoid_needless_id_setting_on_create(nfsnode_t dnp, struct vnode_attr *vap, vfs_context_t ctx)
1844{
1845 if (VATTR_IS_ACTIVE(vap, va_uid)) {
1846 if (kauth_cred_getuid(vfs_context_ucred(ctx)) == vap->va_uid) {
1847 VATTR_CLEAR_ACTIVE(vap, va_uid);
1848 VATTR_CLEAR_ACTIVE(vap, va_uuuid);
1849 }
1850 }
1851 if (VATTR_IS_ACTIVE(vap, va_gid)) {
1852 if ((vap->va_gid == dnp->n_vattr.nva_gid) ||
1853 (kauth_cred_getgid(vfs_context_ucred(ctx)) == vap->va_gid)) {
1854 VATTR_CLEAR_ACTIVE(vap, va_gid);
1855 VATTR_CLEAR_ACTIVE(vap, va_guuid);
1856 }
1857 }
1858}
1859
1860/*
1861 * Convert a universal address string to a sockaddr structure.
1862 *
1863 * Universal addresses can be in the following formats:
1864 *
1865 * d = decimal (IPv4)
1866 * x = hexadecimal (IPv6)
1867 * p = port (decimal)
1868 *
1869 * d.d.d.d
1870 * d.d.d.d.p.p
1871 * x:x:x:x:x:x:x:x
1872 * x:x:x:x:x:x:x:x.p.p
1873 * x:x:x:x:x:x:d.d.d.d
1874 * x:x:x:x:x:x:d.d.d.d.p.p
1875 *
1876 * IPv6 strings can also have a series of zeroes elided
1877 * IPv6 strings can also have a %scope suffix at the end (after any port)
1878 *
1879 * rules & exceptions:
1880 * - value before : is hex
1881 * - value before . is dec
1882 * - once . hit, all values are dec
1883 * - hex+port case means value before first dot is actually hex
1884 * - . is always preceded by digits except if last hex was double-colon
1885 *
1886 * scan, converting #s to bytes
1887 * first time a . is encountered, scan the rest to count them.
1888 * 2 dots = just port
1889 * 3 dots = just IPv4 no port
1890 * 5 dots = IPv4 and port
1891 */
1892
1893#define IS_DIGIT(C) \
1894 (((C) >= '0') && ((C) <= '9'))
1895
1896#define IS_XDIGIT(C) \
1897 (IS_DIGIT(C) || \
1898 (((C) >= 'A') && ((C) <= 'F')) || \
1899 (((C) >= 'a') && ((C) <= 'f')))
1900
1901int
1902nfs_uaddr2sockaddr(const char *uaddr, struct sockaddr *addr)
1903{
1904 const char *p, *pd; /* pointers to current character in scan */
1905 const char *pnum; /* pointer to current number to decode */
1906 const char *pscope; /* pointer to IPv6 scope ID */
1907 uint8_t a[18]; /* octet array to store address bytes */
1908 int i; /* index of next octet to decode */
1909 int dci; /* index of octet to insert double-colon zeroes */
1910 int dcount, xdcount; /* count of digits in current number */
1911 int needmore; /* set when we know we need more input (e.g. after colon, period) */
1912 int dots; /* # of dots */
1913 int hex; /* contains hex values */
1914 unsigned long val; /* decoded value */
1915 int s; /* index used for sliding array to insert elided zeroes */
1916
1917#define HEXVALUE 0
1918#define DECIMALVALUE 1
1919#define GET(TYPE) \
1920 do { \
1921 if ((dcount <= 0) || (dcount > (((TYPE) == DECIMALVALUE) ? 3 : 4))) \
1922 return (0); \
1923 if (((TYPE) == DECIMALVALUE) && xdcount) \
1924 return (0); \
1925 val = strtoul(pnum, NULL, ((TYPE) == DECIMALVALUE) ? 10 : 16); \
1926 if (((TYPE) == DECIMALVALUE) && (val >= 256)) \
1927 return (0); \
1928 /* check if there is room left in the array */ \
1929 if (i > (int)(sizeof(a) - (((TYPE) == HEXVALUE) ? 2 : 1) - ((dci != -1) ? 2 : 0))) \
1930 return (0); \
1931 if ((TYPE) == HEXVALUE) \
1932 a[i++] = ((val >> 8) & 0xff); \
1933 a[i++] = (val & 0xff); \
1934 } while (0)
1935
1936 hex = 0;
1937 dots = 0;
1938 dci = -1;
1939 i = dcount = xdcount = 0;
1940 pnum = p = uaddr;
1941 pscope = NULL;
1942 needmore = 1;
1943 if ((*p == ':') && (*++p != ':')) { /* if it starts with colon, gotta be a double */
1944 return 0;
1945 }
1946
1947 while (*p) {
1948 if (IS_XDIGIT(*p)) {
1949 dcount++;
1950 if (!IS_DIGIT(*p)) {
1951 xdcount++;
1952 }
1953 needmore = 0;
1954 p++;
1955 } else if (*p == '.') {
1956 /* rest is decimal IPv4 dotted quad and/or port */
1957 if (!dots) {
1958 /* this is the first, so count them */
1959 for (pd = p; *pd; pd++) {
1960 if (*pd == '.') {
1961 if (++dots > 5) {
1962 return 0;
1963 }
1964 } else if (hex && (*pd == '%')) {
1965 break;
1966 } else if ((*pd < '0') || (*pd > '9')) {
1967 return 0;
1968 }
1969 }
1970 if ((dots != 2) && (dots != 3) && (dots != 5)) {
1971 return 0;
1972 }
1973 if (hex && (dots == 2)) { /* hex+port */
1974 if (!dcount && needmore) {
1975 return 0;
1976 }
1977 if (dcount) { /* last hex may be elided zero */
1978 GET(HEXVALUE);
1979 }
1980 } else {
1981 GET(DECIMALVALUE);
1982 }
1983 } else {
1984 GET(DECIMALVALUE);
1985 }
1986 dcount = xdcount = 0;
1987 needmore = 1;
1988 pnum = ++p;
1989 } else if (*p == ':') {
1990 hex = 1;
1991 if (dots) {
1992 return 0;
1993 }
1994 if (!dcount) { /* missing number, probably double colon */
1995 if (dci >= 0) { /* can only have one double colon */
1996 return 0;
1997 }
1998 dci = i;
1999 needmore = 0;
2000 } else {
2001 GET(HEXVALUE);
2002 dcount = xdcount = 0;
2003 needmore = 1;
2004 }
2005 pnum = ++p;
2006 } else if (*p == '%') { /* scope ID delimiter */
2007 if (!hex) {
2008 return 0;
2009 }
2010 p++;
2011 pscope = p;
2012 break;
2013 } else { /* unexpected character */
2014 return 0;
2015 }
2016 }
2017 if (needmore && !dcount) {
2018 return 0;
2019 }
2020 if (dcount) { /* decode trailing number */
2021 GET(dots ? DECIMALVALUE : HEXVALUE);
2022 }
2023 if (dci >= 0) { /* got a double-colon at i, need to insert a range of zeroes */
2024 /* if we got a port, slide to end of array */
2025 /* otherwise, slide to end of address (non-port) values */
2026 int end = ((dots == 2) || (dots == 5)) ? sizeof(a) : (sizeof(a) - 2);
2027 if (i % 2) { /* length of zero range must be multiple of 2 */
2028 return 0;
2029 }
2030 if (i >= end) { /* no room? */
2031 return 0;
2032 }
2033 /* slide (i-dci) numbers up from index dci */
2034 for (s = 0; s < (i - dci); s++) {
2035 a[end - 1 - s] = a[i - 1 - s];
2036 }
2037 /* zero (end-i) numbers at index dci */
2038 for (s = 0; s < (end - i); s++) {
2039 a[dci + s] = 0;
2040 }
2041 i = end;
2042 }
2043
2044 /* copy out resulting socket address */
2045 if (hex) {
2046 struct sockaddr_in6 *sin6 = (struct sockaddr_in6*)addr;
2047 if ((((dots == 0) || (dots == 3)) && (i != (sizeof(a) - 2)))) {
2048 return 0;
2049 }
2050 if ((((dots == 2) || (dots == 5)) && (i != sizeof(a)))) {
2051 return 0;
2052 }
2053 bzero(sin6, sizeof(struct sockaddr_in6));
2054 sin6->sin6_len = sizeof(struct sockaddr_in6);
2055 sin6->sin6_family = AF_INET6;
2056 bcopy(a, &sin6->sin6_addr.s6_addr, sizeof(struct in6_addr));
2057 if ((dots == 5) || (dots == 2)) {
2058 sin6->sin6_port = htons((a[16] << 8) | a[17]);
2059 }
2060 if (pscope) {
2061 for (p = pscope; IS_DIGIT(*p); p++) {
2062 ;
2063 }
2064 if (*p && !IS_DIGIT(*p)) { /* name */
2065 ifnet_t interface = NULL;
2066 if (ifnet_find_by_name(pscope, &interface) == 0) {
2067 sin6->sin6_scope_id = ifnet_index(interface);
2068 }
2069 if (interface) {
2070 ifnet_release(interface);
2071 }
2072 } else { /* decimal number */
2073 sin6->sin6_scope_id = strtoul(pscope, NULL, 10);
2074 }
2075 /* XXX should we also embed scope id for linklocal? */
2076 }
2077 } else {
2078 struct sockaddr_in *sin = (struct sockaddr_in*)addr;
2079 if ((dots != 3) && (dots != 5)) {
2080 return 0;
2081 }
2082 if ((dots == 3) && (i != 4)) {
2083 return 0;
2084 }
2085 if ((dots == 5) && (i != 6)) {
2086 return 0;
2087 }
2088 bzero(sin, sizeof(struct sockaddr_in));
2089 sin->sin_len = sizeof(struct sockaddr_in);
2090 sin->sin_family = AF_INET;
2091 bcopy(a, &sin->sin_addr.s_addr, sizeof(struct in_addr));
2092 if (dots == 5) {
2093 sin->sin_port = htons((a[4] << 8) | a[5]);
2094 }
2095 }
2096 return 1;
2097}
2098
2099
2100/* NFS Client debugging support */
2101uint32_t nfs_debug_ctl;
2102
2103#include <libkern/libkern.h>
2104#include <stdarg.h>
2105
2106void
2107nfs_printf(int facility, int level, const char *fmt, ...)
2108{
2109 va_list ap;
2110
2111 if ((uint32_t)level > NFS_DEBUG_LEVEL) {
2112 return;
2113 }
2114 if (NFS_DEBUG_FACILITY && !((uint32_t)facility & NFS_DEBUG_FACILITY)) {
2115 return;
2116 }
2117
2118 va_start(ap, fmt);
2119 vprintf(fmt, ap);
2120 va_end(ap);
2121}
2122
2123/* Is a mount gone away? */
2124int
2125nfs_mount_gone(struct nfsmount *nmp)
2126{
2127 return !nmp || vfs_isforce(nmp->nm_mountp) || (nmp->nm_state & (NFSSTA_FORCE | NFSSTA_DEAD));
2128}
2129
2130/*
2131 * Return some of the more significant mount options
2132 * as a string, e.g. "'ro,hard,intr,tcp,vers=3,sec=krb5,deadtimeout=0'
2133 */
2134int
2135nfs_mountopts(struct nfsmount *nmp, char *buf, int buflen)
2136{
2137 int c;
2138
2139 c = snprintf(buf, buflen, "%s,%s,%s,%s,vers=%d,sec=%s,%sdeadtimeout=%d",
2140 (vfs_flags(nmp->nm_mountp) & MNT_RDONLY) ? "ro" : "rw",
2141 NMFLAG(nmp, SOFT) ? "soft" : "hard",
2142 NMFLAG(nmp, INTR) ? "intr" : "nointr",
2143 nmp->nm_sotype == SOCK_STREAM ? "tcp" : "udp",
2144 nmp->nm_vers,
2145 nmp->nm_auth == RPCAUTH_KRB5 ? "krb5" :
2146 nmp->nm_auth == RPCAUTH_KRB5I ? "krb5i" :
2147 nmp->nm_auth == RPCAUTH_KRB5P ? "krb5p" :
2148 nmp->nm_auth == RPCAUTH_SYS ? "sys" : "none",
2149 nmp->nm_lockmode == NFS_LOCK_MODE_ENABLED ? "locks," :
2150 nmp->nm_lockmode == NFS_LOCK_MODE_DISABLED ? "nolocks," :
2151 nmp->nm_lockmode == NFS_LOCK_MODE_LOCAL ? "locallocks," : "",
2152 nmp->nm_deadtimeout);
2153
2154 return c > buflen ? ENOMEM : 0;
2155}
2156
2157#endif /* NFSCLIENT */
2158
2159/*
2160 * Schedule a callout thread to run an NFS timer function
2161 * interval milliseconds in the future.
2162 */
2163void
2164nfs_interval_timer_start(thread_call_t call, int interval)
2165{
2166 uint64_t deadline;
2167
2168 clock_interval_to_deadline(interval, 1000 * 1000, &deadline);
2169 thread_call_enter_delayed(call, deadline);
2170}
2171
2172
2173#if NFSSERVER
2174
2175int nfsrv_cmp_secflavs(struct nfs_sec *, struct nfs_sec *);
2176int nfsrv_hang_addrlist(struct nfs_export *, struct user_nfs_export_args *);
2177int nfsrv_free_netopt(struct radix_node *, void *);
2178int nfsrv_free_addrlist(struct nfs_export *, struct user_nfs_export_args *);
2179struct nfs_export_options *nfsrv_export_lookup(struct nfs_export *, mbuf_t);
2180struct nfs_export *nfsrv_fhtoexport(struct nfs_filehandle *);
2181struct nfs_user_stat_node *nfsrv_get_user_stat_node(struct nfs_active_user_list *, struct sockaddr *, uid_t);
2182void nfsrv_init_user_list(struct nfs_active_user_list *);
2183void nfsrv_free_user_list(struct nfs_active_user_list *);
2184
2185/*
2186 * add NFSv3 WCC data to an mbuf chain
2187 */
2188int
2189nfsm_chain_add_wcc_data_f(
2190 struct nfsrv_descript *nd,
2191 struct nfsm_chain *nmc,
2192 int preattrerr,
2193 struct vnode_attr *prevap,
2194 int postattrerr,
2195 struct vnode_attr *postvap)
2196{
2197 int error = 0;
2198
2199 if (preattrerr) {
2200 nfsm_chain_add_32(error, nmc, FALSE);
2201 } else {
2202 nfsm_chain_add_32(error, nmc, TRUE);
2203 nfsm_chain_add_64(error, nmc, prevap->va_data_size);
2204 nfsm_chain_add_time(error, nmc, NFS_VER3, &prevap->va_modify_time);
2205 nfsm_chain_add_time(error, nmc, NFS_VER3, &prevap->va_change_time);
2206 }
2207 nfsm_chain_add_postop_attr(error, nd, nmc, postattrerr, postvap);
2208
2209 return error;
2210}
2211
2212/*
2213 * Extract a lookup path from the given mbufs and store it in
2214 * a newly allocated buffer saved in the given nameidata structure.
2215 */
2216int
2217nfsm_chain_get_path_namei(
2218 struct nfsm_chain *nmc,
2219 uint32_t len,
2220 struct nameidata *nip)
2221{
2222 struct componentname *cnp = &nip->ni_cnd;
2223 int error = 0;
2224 char *cp;
2225
2226 if (len > (MAXPATHLEN - 1)) {
2227 return ENAMETOOLONG;
2228 }
2229
2230 /*
2231 * Get a buffer for the name to be translated, and copy the
2232 * name into the buffer.
2233 */
2234 MALLOC_ZONE(cnp->cn_pnbuf, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
2235 if (!cnp->cn_pnbuf) {
2236 return ENOMEM;
2237 }
2238 cnp->cn_pnlen = MAXPATHLEN;
2239 cnp->cn_flags |= HASBUF;
2240
2241 /* Copy the name from the mbuf list to the string */
2242 cp = cnp->cn_pnbuf;
2243 nfsm_chain_get_opaque(error, nmc, len, cp);
2244 if (error) {
2245 goto out;
2246 }
2247 cnp->cn_pnbuf[len] = '\0';
2248
2249 /* sanity check the string */
2250 if ((strlen(cp) != len) || strchr(cp, '/')) {
2251 error = EACCES;
2252 }
2253out:
2254 if (error) {
2255 if (cnp->cn_pnbuf) {
2256 FREE_ZONE(cnp->cn_pnbuf, MAXPATHLEN, M_NAMEI);
2257 }
2258 cnp->cn_flags &= ~HASBUF;
2259 } else {
2260 nip->ni_pathlen = len;
2261 }
2262 return error;
2263}
2264
2265/*
2266 * Set up nameidata for a lookup() call and do it.
2267 */
2268int
2269nfsrv_namei(
2270 struct nfsrv_descript *nd,
2271 vfs_context_t ctx,
2272 struct nameidata *nip,
2273 struct nfs_filehandle *nfhp,
2274 vnode_t *retdirp,
2275 struct nfs_export **nxp,
2276 struct nfs_export_options **nxop)
2277{
2278 vnode_t dp;
2279 int error;
2280 struct componentname *cnp = &nip->ni_cnd;
2281 uint32_t cnflags;
2282 char *tmppn;
2283
2284 *retdirp = NULL;
2285
2286 /*
2287 * Extract and set starting directory.
2288 */
2289 error = nfsrv_fhtovp(nfhp, nd, &dp, nxp, nxop);
2290 if (error) {
2291 goto out;
2292 }
2293 error = nfsrv_credcheck(nd, ctx, *nxp, *nxop);
2294 if (error || (vnode_vtype(dp) != VDIR)) {
2295 vnode_put(dp);
2296 error = ENOTDIR;
2297 goto out;
2298 }
2299 *retdirp = dp;
2300
2301 nip->ni_cnd.cn_context = ctx;
2302
2303 if (*nxop && ((*nxop)->nxo_flags & NX_READONLY)) {
2304 cnp->cn_flags |= RDONLY;
2305 }
2306
2307 cnp->cn_flags |= NOCROSSMOUNT;
2308 cnp->cn_nameptr = cnp->cn_pnbuf;
2309 nip->ni_usedvp = nip->ni_startdir = dp;
2310 nip->ni_rootdir = rootvnode;
2311
2312 /*
2313 * And call lookup() to do the real work
2314 */
2315 cnflags = nip->ni_cnd.cn_flags; /* store in case we have to restore */
2316 while ((error = lookup(nip)) == ERECYCLE) {
2317 nip->ni_cnd.cn_flags = cnflags;
2318 cnp->cn_nameptr = cnp->cn_pnbuf;
2319 nip->ni_usedvp = nip->ni_dvp = nip->ni_startdir = dp;
2320 }
2321 if (error) {
2322 goto out;
2323 }
2324
2325 /* Check for encountering a symbolic link */
2326 if (cnp->cn_flags & ISSYMLINK) {
2327 if (cnp->cn_flags & (LOCKPARENT | WANTPARENT)) {
2328 vnode_put(nip->ni_dvp);
2329 }
2330 if (nip->ni_vp) {
2331 vnode_put(nip->ni_vp);
2332 nip->ni_vp = NULL;
2333 }
2334 error = EINVAL;
2335 }
2336out:
2337 if (error) {
2338 tmppn = cnp->cn_pnbuf;
2339 cnp->cn_pnbuf = NULL;
2340 cnp->cn_flags &= ~HASBUF;
2341 FREE_ZONE(tmppn, cnp->cn_pnlen, M_NAMEI);
2342 }
2343 return error;
2344}
2345
2346/*
2347 * A fiddled version of m_adj() that ensures null fill to a 4-byte
2348 * boundary and only trims off the back end
2349 */
2350void
2351nfsm_adj(mbuf_t mp, int len, int nul)
2352{
2353 mbuf_t m, mnext;
2354 int count, i, mlen;
2355 char *cp;
2356
2357 /*
2358 * Trim from tail. Scan the mbuf chain,
2359 * calculating its length and finding the last mbuf.
2360 * If the adjustment only affects this mbuf, then just
2361 * adjust and return. Otherwise, rescan and truncate
2362 * after the remaining size.
2363 */
2364 count = 0;
2365 m = mp;
2366 for (;;) {
2367 mlen = mbuf_len(m);
2368 count += mlen;
2369 mnext = mbuf_next(m);
2370 if (mnext == NULL) {
2371 break;
2372 }
2373 m = mnext;
2374 }
2375 if (mlen > len) {
2376 mlen -= len;
2377 mbuf_setlen(m, mlen);
2378 if (nul > 0) {
2379 cp = (caddr_t)mbuf_data(m) + mlen - nul;
2380 for (i = 0; i < nul; i++) {
2381 *cp++ = '\0';
2382 }
2383 }
2384 return;
2385 }
2386 count -= len;
2387 if (count < 0) {
2388 count = 0;
2389 }
2390 /*
2391 * Correct length for chain is "count".
2392 * Find the mbuf with last data, adjust its length,
2393 * and toss data from remaining mbufs on chain.
2394 */
2395 for (m = mp; m; m = mbuf_next(m)) {
2396 mlen = mbuf_len(m);
2397 if (mlen >= count) {
2398 mlen = count;
2399 mbuf_setlen(m, count);
2400 if (nul > 0) {
2401 cp = (caddr_t)mbuf_data(m) + mlen - nul;
2402 for (i = 0; i < nul; i++) {
2403 *cp++ = '\0';
2404 }
2405 }
2406 break;
2407 }
2408 count -= mlen;
2409 }
2410 for (m = mbuf_next(m); m; m = mbuf_next(m)) {
2411 mbuf_setlen(m, 0);
2412 }
2413}
2414
2415/*
2416 * Trim the header out of the mbuf list and trim off any trailing
2417 * junk so that the mbuf list has only the write data.
2418 */
2419int
2420nfsm_chain_trim_data(struct nfsm_chain *nmc, int len, int *mlen)
2421{
2422 int cnt = 0, dlen, adjust;
2423 caddr_t data;
2424 mbuf_t m;
2425
2426 if (mlen) {
2427 *mlen = 0;
2428 }
2429
2430 /* trim header */
2431 for (m = nmc->nmc_mhead; m && (m != nmc->nmc_mcur); m = mbuf_next(m)) {
2432 mbuf_setlen(m, 0);
2433 }
2434 if (!m) {
2435 return EIO;
2436 }
2437
2438 /* trim current mbuf */
2439 data = mbuf_data(m);
2440 dlen = mbuf_len(m);
2441 adjust = nmc->nmc_ptr - data;
2442 dlen -= adjust;
2443 if ((dlen > 0) && (adjust > 0)) {
2444 if (mbuf_setdata(m, nmc->nmc_ptr, dlen)) {
2445 return EIO;
2446 }
2447 } else {
2448 mbuf_setlen(m, dlen);
2449 }
2450
2451 /* skip next len bytes */
2452 for (; m && (cnt < len); m = mbuf_next(m)) {
2453 dlen = mbuf_len(m);
2454 cnt += dlen;
2455 if (cnt > len) {
2456 /* truncate to end of data */
2457 mbuf_setlen(m, dlen - (cnt - len));
2458 if (m == nmc->nmc_mcur) {
2459 nmc->nmc_left -= (cnt - len);
2460 }
2461 cnt = len;
2462 }
2463 }
2464 if (mlen) {
2465 *mlen = cnt;
2466 }
2467
2468 /* trim any trailing data */
2469 if (m == nmc->nmc_mcur) {
2470 nmc->nmc_left = 0;
2471 }
2472 for (; m; m = mbuf_next(m)) {
2473 mbuf_setlen(m, 0);
2474 }
2475
2476 return 0;
2477}
2478
2479int
2480nfsm_chain_add_fattr(
2481 struct nfsrv_descript *nd,
2482 struct nfsm_chain *nmc,
2483 struct vnode_attr *vap)
2484{
2485 int error = 0;
2486
2487 // XXX Should we assert here that all fields are supported?
2488
2489 nfsm_chain_add_32(error, nmc, vtonfs_type(vap->va_type, nd->nd_vers));
2490 if (nd->nd_vers == NFS_VER3) {
2491 nfsm_chain_add_32(error, nmc, vap->va_mode & 07777);
2492 } else {
2493 nfsm_chain_add_32(error, nmc, vtonfsv2_mode(vap->va_type, vap->va_mode));
2494 }
2495 nfsm_chain_add_32(error, nmc, vap->va_nlink);
2496 nfsm_chain_add_32(error, nmc, vap->va_uid);
2497 nfsm_chain_add_32(error, nmc, vap->va_gid);
2498 if (nd->nd_vers == NFS_VER3) {
2499 nfsm_chain_add_64(error, nmc, vap->va_data_size);
2500 nfsm_chain_add_64(error, nmc, vap->va_data_alloc);
2501 nfsm_chain_add_32(error, nmc, major(vap->va_rdev));
2502 nfsm_chain_add_32(error, nmc, minor(vap->va_rdev));
2503 nfsm_chain_add_64(error, nmc, vap->va_fsid);
2504 nfsm_chain_add_64(error, nmc, vap->va_fileid);
2505 } else {
2506 nfsm_chain_add_32(error, nmc, vap->va_data_size);
2507 nfsm_chain_add_32(error, nmc, NFS_FABLKSIZE);
2508 if (vap->va_type == VFIFO) {
2509 nfsm_chain_add_32(error, nmc, 0xffffffff);
2510 } else {
2511 nfsm_chain_add_32(error, nmc, vap->va_rdev);
2512 }
2513 nfsm_chain_add_32(error, nmc, vap->va_data_alloc / NFS_FABLKSIZE);
2514 nfsm_chain_add_32(error, nmc, vap->va_fsid);
2515 nfsm_chain_add_32(error, nmc, vap->va_fileid);
2516 }
2517 nfsm_chain_add_time(error, nmc, nd->nd_vers, &vap->va_access_time);
2518 nfsm_chain_add_time(error, nmc, nd->nd_vers, &vap->va_modify_time);
2519 nfsm_chain_add_time(error, nmc, nd->nd_vers, &vap->va_change_time);
2520
2521 return error;
2522}
2523
2524int
2525nfsm_chain_get_sattr(
2526 struct nfsrv_descript *nd,
2527 struct nfsm_chain *nmc,
2528 struct vnode_attr *vap)
2529{
2530 int error = 0;
2531 uint32_t val = 0;
2532 uint64_t val64 = 0;
2533 struct timespec now;
2534
2535 if (nd->nd_vers == NFS_VER2) {
2536 /*
2537 * There is/was a bug in the Sun client that puts 0xffff in the mode
2538 * field of sattr when it should put in 0xffffffff. The u_short
2539 * doesn't sign extend. So check the low order 2 bytes for 0xffff.
2540 */
2541 nfsm_chain_get_32(error, nmc, val);
2542 if ((val & 0xffff) != 0xffff) {
2543 VATTR_SET(vap, va_mode, val & 07777);
2544 /* save the "type" bits for NFSv2 create */
2545 VATTR_SET(vap, va_type, IFTOVT(val));
2546 VATTR_CLEAR_ACTIVE(vap, va_type);
2547 }
2548 nfsm_chain_get_32(error, nmc, val);
2549 if (val != (uint32_t)-1) {
2550 VATTR_SET(vap, va_uid, val);
2551 }
2552 nfsm_chain_get_32(error, nmc, val);
2553 if (val != (uint32_t)-1) {
2554 VATTR_SET(vap, va_gid, val);
2555 }
2556 /* save the "size" bits for NFSv2 create (even if they appear unset) */
2557 nfsm_chain_get_32(error, nmc, val);
2558 VATTR_SET(vap, va_data_size, val);
2559 if (val == (uint32_t)-1) {
2560 VATTR_CLEAR_ACTIVE(vap, va_data_size);
2561 }
2562 nfsm_chain_get_time(error, nmc, NFS_VER2,
2563 vap->va_access_time.tv_sec,
2564 vap->va_access_time.tv_nsec);
2565 if (vap->va_access_time.tv_sec != -1) {
2566 VATTR_SET_ACTIVE(vap, va_access_time);
2567 }
2568 nfsm_chain_get_time(error, nmc, NFS_VER2,
2569 vap->va_modify_time.tv_sec,
2570 vap->va_modify_time.tv_nsec);
2571 if (vap->va_modify_time.tv_sec != -1) {
2572 VATTR_SET_ACTIVE(vap, va_modify_time);
2573 }
2574 return error;
2575 }
2576
2577 /* NFSv3 */
2578 nfsm_chain_get_32(error, nmc, val);
2579 if (val) {
2580 nfsm_chain_get_32(error, nmc, val);
2581 VATTR_SET(vap, va_mode, val & 07777);
2582 }
2583 nfsm_chain_get_32(error, nmc, val);
2584 if (val) {
2585 nfsm_chain_get_32(error, nmc, val);
2586 VATTR_SET(vap, va_uid, val);
2587 }
2588 nfsm_chain_get_32(error, nmc, val);
2589 if (val) {
2590 nfsm_chain_get_32(error, nmc, val);
2591 VATTR_SET(vap, va_gid, val);
2592 }
2593 nfsm_chain_get_32(error, nmc, val);
2594 if (val) {
2595 nfsm_chain_get_64(error, nmc, val64);
2596 VATTR_SET(vap, va_data_size, val64);
2597 }
2598 nanotime(&now);
2599 nfsm_chain_get_32(error, nmc, val);
2600 switch (val) {
2601 case NFS_TIME_SET_TO_CLIENT:
2602 nfsm_chain_get_time(error, nmc, nd->nd_vers,
2603 vap->va_access_time.tv_sec,
2604 vap->va_access_time.tv_nsec);
2605 VATTR_SET_ACTIVE(vap, va_access_time);
2606 vap->va_vaflags &= ~VA_UTIMES_NULL;
2607 break;
2608 case NFS_TIME_SET_TO_SERVER:
2609 VATTR_SET(vap, va_access_time, now);
2610 vap->va_vaflags |= VA_UTIMES_NULL;
2611 break;
2612 }
2613 nfsm_chain_get_32(error, nmc, val);
2614 switch (val) {
2615 case NFS_TIME_SET_TO_CLIENT:
2616 nfsm_chain_get_time(error, nmc, nd->nd_vers,
2617 vap->va_modify_time.tv_sec,
2618 vap->va_modify_time.tv_nsec);
2619 VATTR_SET_ACTIVE(vap, va_modify_time);
2620 vap->va_vaflags &= ~VA_UTIMES_NULL;
2621 break;
2622 case NFS_TIME_SET_TO_SERVER:
2623 VATTR_SET(vap, va_modify_time, now);
2624 if (!VATTR_IS_ACTIVE(vap, va_access_time)) {
2625 vap->va_vaflags |= VA_UTIMES_NULL;
2626 }
2627 break;
2628 }
2629
2630 return error;
2631}
2632
2633/*
2634 * Compare two security flavor structs
2635 */
2636int
2637nfsrv_cmp_secflavs(struct nfs_sec *sf1, struct nfs_sec *sf2)
2638{
2639 int i;
2640
2641 if (sf1->count != sf2->count) {
2642 return 1;
2643 }
2644 for (i = 0; i < sf1->count; i++) {
2645 if (sf1->flavors[i] != sf2->flavors[i]) {
2646 return 1;
2647 }
2648 }
2649 return 0;
2650}
2651
2652/*
2653 * Build hash lists of net addresses and hang them off the NFS export.
2654 * Called by nfsrv_export() to set up the lists of export addresses.
2655 */
2656int
2657nfsrv_hang_addrlist(struct nfs_export *nx, struct user_nfs_export_args *unxa)
2658{
2659 struct nfs_export_net_args nxna;
2660 struct nfs_netopt *no, *rn_no;
2661 struct radix_node_head *rnh;
2662 struct radix_node *rn;
2663 struct sockaddr *saddr, *smask;
2664 struct domain *dom;
2665 int i, error;
2666 unsigned int net;
2667 user_addr_t uaddr;
2668 kauth_cred_t cred;
2669
2670 uaddr = unxa->nxa_nets;
2671 for (net = 0; net < unxa->nxa_netcount; net++, uaddr += sizeof(nxna)) {
2672 error = copyin(uaddr, &nxna, sizeof(nxna));
2673 if (error) {
2674 return error;
2675 }
2676
2677 if (nxna.nxna_addr.ss_len > sizeof(struct sockaddr_storage) ||
2678 nxna.nxna_mask.ss_len > sizeof(struct sockaddr_storage) ||
2679 nxna.nxna_addr.ss_family > AF_MAX ||
2680 nxna.nxna_mask.ss_family > AF_MAX) {
2681 return EINVAL;
2682 }
2683
2684 if (nxna.nxna_flags & (NX_MAPROOT | NX_MAPALL)) {
2685 struct posix_cred temp_pcred;
2686 bzero(&temp_pcred, sizeof(temp_pcred));
2687 temp_pcred.cr_uid = nxna.nxna_cred.cr_uid;
2688 temp_pcred.cr_ngroups = nxna.nxna_cred.cr_ngroups;
2689 for (i = 0; i < nxna.nxna_cred.cr_ngroups && i < NGROUPS; i++) {
2690 temp_pcred.cr_groups[i] = nxna.nxna_cred.cr_groups[i];
2691 }
2692 cred = posix_cred_create(&temp_pcred);
2693 if (!IS_VALID_CRED(cred)) {
2694 return ENOMEM;
2695 }
2696 } else {
2697 cred = NOCRED;
2698 }
2699
2700 if (nxna.nxna_addr.ss_len == 0) {
2701 /* No address means this is a default/world export */
2702 if (nx->nx_flags & NX_DEFAULTEXPORT) {
2703 if (IS_VALID_CRED(cred)) {
2704 kauth_cred_unref(&cred);
2705 }
2706 return EEXIST;
2707 }
2708 nx->nx_flags |= NX_DEFAULTEXPORT;
2709 nx->nx_defopt.nxo_flags = nxna.nxna_flags;
2710 nx->nx_defopt.nxo_cred = cred;
2711 bcopy(&nxna.nxna_sec, &nx->nx_defopt.nxo_sec, sizeof(struct nfs_sec));
2712 nx->nx_expcnt++;
2713 continue;
2714 }
2715
2716 i = sizeof(struct nfs_netopt);
2717 i += nxna.nxna_addr.ss_len + nxna.nxna_mask.ss_len;
2718 MALLOC(no, struct nfs_netopt *, i, M_NETADDR, M_WAITOK);
2719 if (!no) {
2720 if (IS_VALID_CRED(cred)) {
2721 kauth_cred_unref(&cred);
2722 }
2723 return ENOMEM;
2724 }
2725 bzero(no, sizeof(struct nfs_netopt));
2726 no->no_opt.nxo_flags = nxna.nxna_flags;
2727 no->no_opt.nxo_cred = cred;
2728 bcopy(&nxna.nxna_sec, &no->no_opt.nxo_sec, sizeof(struct nfs_sec));
2729
2730 saddr = (struct sockaddr *)(no + 1);
2731 bcopy(&nxna.nxna_addr, saddr, nxna.nxna_addr.ss_len);
2732 if (nxna.nxna_mask.ss_len) {
2733 smask = (struct sockaddr *)((caddr_t)saddr + nxna.nxna_addr.ss_len);
2734 bcopy(&nxna.nxna_mask, smask, nxna.nxna_mask.ss_len);
2735 } else {
2736 smask = NULL;
2737 }
2738 i = saddr->sa_family;
2739 if ((rnh = nx->nx_rtable[i]) == 0) {
2740 /*
2741 * Seems silly to initialize every AF when most are not
2742 * used, do so on demand here
2743 */
2744 TAILQ_FOREACH(dom, &domains, dom_entry) {
2745 if (dom->dom_family == i && dom->dom_rtattach) {
2746 dom->dom_rtattach((void **)&nx->nx_rtable[i],
2747 dom->dom_rtoffset);
2748 break;
2749 }
2750 }
2751 if ((rnh = nx->nx_rtable[i]) == 0) {
2752 if (IS_VALID_CRED(cred)) {
2753 kauth_cred_unref(&cred);
2754 }
2755 _FREE(no, M_NETADDR);
2756 return ENOBUFS;
2757 }
2758 }
2759 rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh, no->no_rnodes);
2760 if (rn == 0) {
2761 /*
2762 * One of the reasons that rnh_addaddr may fail is that
2763 * the entry already exists. To check for this case, we
2764 * look up the entry to see if it is there. If so, we
2765 * do not need to make a new entry but do continue.
2766 *
2767 * XXX should this be rnh_lookup() instead?
2768 */
2769 int matched = 0;
2770 rn = (*rnh->rnh_matchaddr)((caddr_t)saddr, rnh);
2771 rn_no = (struct nfs_netopt *)rn;
2772 if (rn != 0 && (rn->rn_flags & RNF_ROOT) == 0 &&
2773 (rn_no->no_opt.nxo_flags == nxna.nxna_flags) &&
2774 (!nfsrv_cmp_secflavs(&rn_no->no_opt.nxo_sec, &nxna.nxna_sec))) {
2775 kauth_cred_t cred2 = rn_no->no_opt.nxo_cred;
2776 if (cred == cred2) {
2777 /* creds are same (or both NULL) */
2778 matched = 1;
2779 } else if (cred && cred2 && (kauth_cred_getuid(cred) == kauth_cred_getuid(cred2))) {
2780 /*
2781 * Now compare the effective and
2782 * supplementary groups...
2783 *
2784 * Note: This comparison, as written,
2785 * does not correctly indicate that
2786 * the groups are equivalent, since
2787 * other than the first supplementary
2788 * group, which is also the effective
2789 * group, order on the remaining groups
2790 * doesn't matter, and this is an
2791 * ordered compare.
2792 */
2793 gid_t groups[NGROUPS];
2794 gid_t groups2[NGROUPS];
2795 int groupcount = NGROUPS;
2796 int group2count = NGROUPS;
2797
2798 if (!kauth_cred_getgroups(cred, groups, &groupcount) &&
2799 !kauth_cred_getgroups(cred2, groups2, &group2count) &&
2800 groupcount == group2count) {
2801 for (i = 0; i < group2count; i++) {
2802 if (groups[i] != groups2[i]) {
2803 break;
2804 }
2805 }
2806 if (i >= group2count || i >= NGROUPS) {
2807 matched = 1;
2808 }
2809 }
2810 }
2811 }
2812 if (IS_VALID_CRED(cred)) {
2813 kauth_cred_unref(&cred);
2814 }
2815 _FREE(no, M_NETADDR);
2816 if (matched) {
2817 continue;
2818 }
2819 return EPERM;
2820 }
2821 nx->nx_expcnt++;
2822 }
2823
2824 return 0;
2825}
2826
2827/*
2828 * In order to properly track an export's netopt count, we need to pass
2829 * an additional argument to nfsrv_free_netopt() so that it can decrement
2830 * the export's netopt count.
2831 */
2832struct nfsrv_free_netopt_arg {
2833 uint32_t *cnt;
2834 struct radix_node_head *rnh;
2835};
2836
2837int
2838nfsrv_free_netopt(struct radix_node *rn, void *w)
2839{
2840 struct nfsrv_free_netopt_arg *fna = (struct nfsrv_free_netopt_arg *)w;
2841 struct radix_node_head *rnh = fna->rnh;
2842 uint32_t *cnt = fna->cnt;
2843 struct nfs_netopt *nno = (struct nfs_netopt *)rn;
2844
2845 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh);
2846 if (IS_VALID_CRED(nno->no_opt.nxo_cred)) {
2847 kauth_cred_unref(&nno->no_opt.nxo_cred);
2848 }
2849 _FREE((caddr_t)rn, M_NETADDR);
2850 *cnt -= 1;
2851 return 0;
2852}
2853
2854/*
2855 * Free the net address hash lists that are hanging off the mount points.
2856 */
2857int
2858nfsrv_free_addrlist(struct nfs_export *nx, struct user_nfs_export_args *unxa)
2859{
2860 struct nfs_export_net_args nxna;
2861 struct radix_node_head *rnh;
2862 struct radix_node *rn;
2863 struct nfsrv_free_netopt_arg fna;
2864 struct nfs_netopt *nno;
2865 user_addr_t uaddr;
2866 unsigned int net;
2867 int i, error;
2868
2869 if (!unxa || !unxa->nxa_netcount) {
2870 /* delete everything */
2871 for (i = 0; i <= AF_MAX; i++) {
2872 if ((rnh = nx->nx_rtable[i])) {
2873 fna.rnh = rnh;
2874 fna.cnt = &nx->nx_expcnt;
2875 (*rnh->rnh_walktree)(rnh, nfsrv_free_netopt, (caddr_t)&fna);
2876 _FREE((caddr_t)rnh, M_RTABLE);
2877 nx->nx_rtable[i] = 0;
2878 }
2879 }
2880 return 0;
2881 }
2882
2883 /* delete only the exports specified */
2884 uaddr = unxa->nxa_nets;
2885 for (net = 0; net < unxa->nxa_netcount; net++, uaddr += sizeof(nxna)) {
2886 error = copyin(uaddr, &nxna, sizeof(nxna));
2887 if (error) {
2888 return error;
2889 }
2890
2891 if (nxna.nxna_addr.ss_len == 0) {
2892 /* No address means this is a default/world export */
2893 if (nx->nx_flags & NX_DEFAULTEXPORT) {
2894 nx->nx_flags &= ~NX_DEFAULTEXPORT;
2895 if (IS_VALID_CRED(nx->nx_defopt.nxo_cred)) {
2896 kauth_cred_unref(&nx->nx_defopt.nxo_cred);
2897 }
2898 nx->nx_expcnt--;
2899 }
2900 continue;
2901 }
2902
2903 if ((rnh = nx->nx_rtable[nxna.nxna_addr.ss_family]) == 0) {
2904 /* AF not initialized? */
2905 if (!(unxa->nxa_flags & NXA_ADD)) {
2906 printf("nfsrv_free_addrlist: address not found (0)\n");
2907 }
2908 continue;
2909 }
2910
2911 rn = (*rnh->rnh_lookup)(&nxna.nxna_addr,
2912 nxna.nxna_mask.ss_len ? &nxna.nxna_mask : NULL, rnh);
2913 if (!rn || (rn->rn_flags & RNF_ROOT)) {
2914 if (!(unxa->nxa_flags & NXA_ADD)) {
2915 printf("nfsrv_free_addrlist: address not found (1)\n");
2916 }
2917 continue;
2918 }
2919
2920 (*rnh->rnh_deladdr)(rn->rn_key, rn->rn_mask, rnh);
2921 nno = (struct nfs_netopt *)rn;
2922 if (IS_VALID_CRED(nno->no_opt.nxo_cred)) {
2923 kauth_cred_unref(&nno->no_opt.nxo_cred);
2924 }
2925 _FREE((caddr_t)rn, M_NETADDR);
2926
2927 nx->nx_expcnt--;
2928 if (nx->nx_expcnt == ((nx->nx_flags & NX_DEFAULTEXPORT) ? 1 : 0)) {
2929 /* no more entries in rnh, so free it up */
2930 _FREE((caddr_t)rnh, M_RTABLE);
2931 nx->nx_rtable[nxna.nxna_addr.ss_family] = 0;
2932 }
2933 }
2934
2935 return 0;
2936}
2937
2938void enablequotas(struct mount *mp, vfs_context_t ctx); // XXX
2939
2940int
2941nfsrv_export(struct user_nfs_export_args *unxa, vfs_context_t ctx)
2942{
2943 int error = 0;
2944 size_t pathlen;
2945 struct nfs_exportfs *nxfs, *nxfs2, *nxfs3;
2946 struct nfs_export *nx, *nx2, *nx3;
2947 struct nfs_filehandle nfh;
2948 struct nameidata mnd, xnd;
2949 vnode_t mvp = NULL, xvp = NULL;
2950 mount_t mp = NULL;
2951 char path[MAXPATHLEN];
2952 int expisroot;
2953
2954 if (unxa->nxa_flags == NXA_CHECK) {
2955 /* just check if the path is an NFS-exportable file system */
2956 error = copyinstr(unxa->nxa_fspath, path, MAXPATHLEN, &pathlen);
2957 if (error) {
2958 return error;
2959 }
2960 NDINIT(&mnd, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
2961 UIO_SYSSPACE, CAST_USER_ADDR_T(path), ctx);
2962 error = namei(&mnd);
2963 if (error) {
2964 return error;
2965 }
2966 mvp = mnd.ni_vp;
2967 mp = vnode_mount(mvp);
2968 /* make sure it's the root of a file system */
2969 if (!vnode_isvroot(mvp)) {
2970 error = EINVAL;
2971 }
2972 /* make sure the file system is NFS-exportable */
2973 if (!error) {
2974 nfh.nfh_len = NFSV3_MAX_FID_SIZE;
2975 error = VFS_VPTOFH(mvp, (int*)&nfh.nfh_len, &nfh.nfh_fid[0], NULL);
2976 }
2977 if (!error && (nfh.nfh_len > (int)NFSV3_MAX_FID_SIZE)) {
2978 error = EIO;
2979 }
2980 if (!error && !(mp->mnt_vtable->vfc_vfsflags & VFC_VFSREADDIR_EXTENDED)) {
2981 error = EISDIR;
2982 }
2983 vnode_put(mvp);
2984 nameidone(&mnd);
2985 return error;
2986 }
2987
2988 /* all other operations: must be super user */
2989 if ((error = vfs_context_suser(ctx))) {
2990 return error;
2991 }
2992
2993 if (unxa->nxa_flags & NXA_DELETE_ALL) {
2994 /* delete all exports on all file systems */
2995 lck_rw_lock_exclusive(&nfsrv_export_rwlock);
2996 while ((nxfs = LIST_FIRST(&nfsrv_exports))) {
2997 mp = vfs_getvfs_by_mntonname(nxfs->nxfs_path);
2998 if (mp) {
2999 vfs_clearflags(mp, MNT_EXPORTED);
3000 mount_iterdrop(mp);
3001 mp = NULL;
3002 }
3003 /* delete all exports on this file system */
3004 while ((nx = LIST_FIRST(&nxfs->nxfs_exports))) {
3005 LIST_REMOVE(nx, nx_next);
3006 LIST_REMOVE(nx, nx_hash);
3007 /* delete all netopts for this export */
3008 nfsrv_free_addrlist(nx, NULL);
3009 nx->nx_flags &= ~NX_DEFAULTEXPORT;
3010 if (IS_VALID_CRED(nx->nx_defopt.nxo_cred)) {
3011 kauth_cred_unref(&nx->nx_defopt.nxo_cred);
3012 }
3013 /* free active user list for this export */
3014 nfsrv_free_user_list(&nx->nx_user_list);
3015 FREE(nx->nx_path, M_TEMP);
3016 FREE(nx, M_TEMP);
3017 }
3018 LIST_REMOVE(nxfs, nxfs_next);
3019 FREE(nxfs->nxfs_path, M_TEMP);
3020 FREE(nxfs, M_TEMP);
3021 }
3022 if (nfsrv_export_hashtbl) {
3023 /* all exports deleted, clean up export hash table */
3024 FREE(nfsrv_export_hashtbl, M_TEMP);
3025 nfsrv_export_hashtbl = NULL;
3026 }
3027 lck_rw_done(&nfsrv_export_rwlock);
3028 return 0;
3029 }
3030
3031 error = copyinstr(unxa->nxa_fspath, path, MAXPATHLEN, &pathlen);
3032 if (error) {
3033 return error;
3034 }
3035
3036 lck_rw_lock_exclusive(&nfsrv_export_rwlock);
3037
3038 /* init export hash table if not already */
3039 if (!nfsrv_export_hashtbl) {
3040 if (nfsrv_export_hash_size <= 0) {
3041 nfsrv_export_hash_size = NFSRVEXPHASHSZ;
3042 }
3043 nfsrv_export_hashtbl = hashinit(nfsrv_export_hash_size, M_TEMP, &nfsrv_export_hash);
3044 }
3045
3046 // first check if we've already got an exportfs with the given ID
3047 LIST_FOREACH(nxfs, &nfsrv_exports, nxfs_next) {
3048 if (nxfs->nxfs_id == unxa->nxa_fsid) {
3049 break;
3050 }
3051 }
3052 if (nxfs) {
3053 /* verify exported FS path matches given path */
3054 if (strncmp(path, nxfs->nxfs_path, MAXPATHLEN)) {
3055 error = EEXIST;
3056 goto unlock_out;
3057 }
3058 if ((unxa->nxa_flags & (NXA_ADD | NXA_OFFLINE)) == NXA_ADD) {
3059 /* if adding, verify that the mount is still what we expect */
3060 mp = vfs_getvfs_by_mntonname(nxfs->nxfs_path);
3061 if (mp) {
3062 mount_ref(mp, 0);
3063 mount_iterdrop(mp);
3064 }
3065 /* find exported FS root vnode */
3066 NDINIT(&mnd, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
3067 UIO_SYSSPACE, CAST_USER_ADDR_T(nxfs->nxfs_path), ctx);
3068 error = namei(&mnd);
3069 if (error) {
3070 goto unlock_out;
3071 }
3072 mvp = mnd.ni_vp;
3073 /* make sure it's (still) the root of a file system */
3074 if (!vnode_isvroot(mvp)) {
3075 error = EINVAL;
3076 goto out;
3077 }
3078 /* sanity check: this should be same mount */
3079 if (mp != vnode_mount(mvp)) {
3080 error = EINVAL;
3081 goto out;
3082 }
3083 }
3084 } else {
3085 /* no current exported file system with that ID */
3086 if (!(unxa->nxa_flags & NXA_ADD)) {
3087 error = ENOENT;
3088 goto unlock_out;
3089 }
3090
3091 /* find exported FS root vnode */
3092 NDINIT(&mnd, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
3093 UIO_SYSSPACE, CAST_USER_ADDR_T(path), ctx);
3094 error = namei(&mnd);
3095 if (error) {
3096 if (!(unxa->nxa_flags & NXA_OFFLINE)) {
3097 goto unlock_out;
3098 }
3099 } else {
3100 mvp = mnd.ni_vp;
3101 /* make sure it's the root of a file system */
3102 if (!vnode_isvroot(mvp)) {
3103 /* bail if not marked offline */
3104 if (!(unxa->nxa_flags & NXA_OFFLINE)) {
3105 error = EINVAL;
3106 goto out;
3107 }
3108 vnode_put(mvp);
3109 nameidone(&mnd);
3110 mvp = NULL;
3111 } else {
3112 mp = vnode_mount(mvp);
3113 mount_ref(mp, 0);
3114
3115 /* make sure the file system is NFS-exportable */
3116 nfh.nfh_len = NFSV3_MAX_FID_SIZE;
3117 error = VFS_VPTOFH(mvp, (int*)&nfh.nfh_len, &nfh.nfh_fid[0], NULL);
3118 if (!error && (nfh.nfh_len > (int)NFSV3_MAX_FID_SIZE)) {
3119 error = EIO;
3120 }
3121 if (!error && !(mp->mnt_vtable->vfc_vfsflags & VFC_VFSREADDIR_EXTENDED)) {
3122 error = EISDIR;
3123 }
3124 if (error) {
3125 goto out;
3126 }
3127 }
3128 }
3129
3130 /* add an exportfs for it */
3131 MALLOC(nxfs, struct nfs_exportfs *, sizeof(struct nfs_exportfs), M_TEMP, M_WAITOK);
3132 if (!nxfs) {
3133 error = ENOMEM;
3134 goto out;
3135 }
3136 bzero(nxfs, sizeof(struct nfs_exportfs));
3137 nxfs->nxfs_id = unxa->nxa_fsid;
3138 MALLOC(nxfs->nxfs_path, char*, pathlen, M_TEMP, M_WAITOK);
3139 if (!nxfs->nxfs_path) {
3140 FREE(nxfs, M_TEMP);
3141 error = ENOMEM;
3142 goto out;
3143 }
3144 bcopy(path, nxfs->nxfs_path, pathlen);
3145 /* insert into list in reverse-sorted order */
3146 nxfs3 = NULL;
3147 LIST_FOREACH(nxfs2, &nfsrv_exports, nxfs_next) {
3148 if (strncmp(nxfs->nxfs_path, nxfs2->nxfs_path, MAXPATHLEN) > 0) {
3149 break;
3150 }
3151 nxfs3 = nxfs2;
3152 }
3153 if (nxfs2) {
3154 LIST_INSERT_BEFORE(nxfs2, nxfs, nxfs_next);
3155 } else if (nxfs3) {
3156 LIST_INSERT_AFTER(nxfs3, nxfs, nxfs_next);
3157 } else {
3158 LIST_INSERT_HEAD(&nfsrv_exports, nxfs, nxfs_next);
3159 }
3160
3161 /* make sure any quotas are enabled before we export the file system */
3162 if (mp) {
3163 enablequotas(mp, ctx);
3164 }
3165 }
3166
3167 if (unxa->nxa_exppath) {
3168 error = copyinstr(unxa->nxa_exppath, path, MAXPATHLEN, &pathlen);
3169 if (error) {
3170 goto out;
3171 }
3172 LIST_FOREACH(nx, &nxfs->nxfs_exports, nx_next) {
3173 if (nx->nx_id == unxa->nxa_expid) {
3174 break;
3175 }
3176 }
3177 if (nx) {
3178 /* verify exported FS path matches given path */
3179 if (strncmp(path, nx->nx_path, MAXPATHLEN)) {
3180 error = EEXIST;
3181 goto out;
3182 }
3183 } else {
3184 /* no current export with that ID */
3185 if (!(unxa->nxa_flags & NXA_ADD)) {
3186 error = ENOENT;
3187 goto out;
3188 }
3189 /* add an export for it */
3190 MALLOC(nx, struct nfs_export *, sizeof(struct nfs_export), M_TEMP, M_WAITOK);
3191 if (!nx) {
3192 error = ENOMEM;
3193 goto out1;
3194 }
3195 bzero(nx, sizeof(struct nfs_export));
3196 nx->nx_id = unxa->nxa_expid;
3197 nx->nx_fs = nxfs;
3198 microtime(&nx->nx_exptime);
3199 MALLOC(nx->nx_path, char*, pathlen, M_TEMP, M_WAITOK);
3200 if (!nx->nx_path) {
3201 error = ENOMEM;
3202 FREE(nx, M_TEMP);
3203 nx = NULL;
3204 goto out1;
3205 }
3206 bcopy(path, nx->nx_path, pathlen);
3207 /* initialize the active user list */
3208 nfsrv_init_user_list(&nx->nx_user_list);
3209 /* insert into list in reverse-sorted order */
3210 nx3 = NULL;
3211 LIST_FOREACH(nx2, &nxfs->nxfs_exports, nx_next) {
3212 if (strncmp(nx->nx_path, nx2->nx_path, MAXPATHLEN) > 0) {
3213 break;
3214 }
3215 nx3 = nx2;
3216 }
3217 if (nx2) {
3218 LIST_INSERT_BEFORE(nx2, nx, nx_next);
3219 } else if (nx3) {
3220 LIST_INSERT_AFTER(nx3, nx, nx_next);
3221 } else {
3222 LIST_INSERT_HEAD(&nxfs->nxfs_exports, nx, nx_next);
3223 }
3224 /* insert into hash */
3225 LIST_INSERT_HEAD(NFSRVEXPHASH(nxfs->nxfs_id, nx->nx_id), nx, nx_hash);
3226
3227 /*
3228 * We don't allow/support nested exports. Check if the new entry
3229 * nests with the entries before and after or if there's an
3230 * entry for the file system root and subdirs.
3231 */
3232 error = 0;
3233 if ((nx3 && !strncmp(nx3->nx_path, nx->nx_path, pathlen - 1) &&
3234 (nx3->nx_path[pathlen - 1] == '/')) ||
3235 (nx2 && !strncmp(nx2->nx_path, nx->nx_path, strlen(nx2->nx_path)) &&
3236 (nx->nx_path[strlen(nx2->nx_path)] == '/'))) {
3237 error = EINVAL;
3238 }
3239 if (!error) {
3240 /* check export conflict with fs root export and vice versa */
3241 expisroot = !nx->nx_path[0] ||
3242 ((nx->nx_path[0] == '.') && !nx->nx_path[1]);
3243 LIST_FOREACH(nx2, &nxfs->nxfs_exports, nx_next) {
3244 if (expisroot) {
3245 if (nx2 != nx) {
3246 break;
3247 }
3248 } else if (!nx2->nx_path[0]) {
3249 break;
3250 } else if ((nx2->nx_path[0] == '.') && !nx2->nx_path[1]) {
3251 break;
3252 }
3253 }
3254 if (nx2) {
3255 error = EINVAL;
3256 }
3257 }
3258 if (error) {
3259 /*
3260 * Don't actually return an error because mountd is
3261 * probably about to delete the conflicting export.
3262 * This can happen when a new export momentarily conflicts
3263 * with an old export while the transition is being made.
3264 * Theoretically, mountd could be written to avoid this
3265 * transient situation - but it would greatly increase the
3266 * complexity of mountd for very little overall benefit.
3267 */
3268 printf("nfsrv_export: warning: nested exports: %s/%s\n",
3269 nxfs->nxfs_path, nx->nx_path);
3270 error = 0;
3271 }
3272 nx->nx_fh.nfh_xh.nxh_flags = NXHF_INVALIDFH;
3273 }
3274 /* make sure file handle is set up */
3275 if ((nx->nx_fh.nfh_xh.nxh_version != htonl(NFS_FH_VERSION)) ||
3276 (nx->nx_fh.nfh_xh.nxh_flags & NXHF_INVALIDFH)) {
3277 /* try to set up export root file handle */
3278 nx->nx_fh.nfh_xh.nxh_version = htonl(NFS_FH_VERSION);
3279 nx->nx_fh.nfh_xh.nxh_fsid = htonl(nx->nx_fs->nxfs_id);
3280 nx->nx_fh.nfh_xh.nxh_expid = htonl(nx->nx_id);
3281 nx->nx_fh.nfh_xh.nxh_flags = 0;
3282 nx->nx_fh.nfh_xh.nxh_reserved = 0;
3283 nx->nx_fh.nfh_fhp = (u_char*)&nx->nx_fh.nfh_xh;
3284 bzero(&nx->nx_fh.nfh_fid[0], NFSV2_MAX_FID_SIZE);
3285 if (mvp) {
3286 /* find export root vnode */
3287 if (!nx->nx_path[0] || ((nx->nx_path[0] == '.') && !nx->nx_path[1])) {
3288 /* exporting file system's root directory */
3289 xvp = mvp;
3290 vnode_get(xvp);
3291 } else {
3292 xnd.ni_cnd.cn_nameiop = LOOKUP;
3293#if CONFIG_TRIGGERS
3294 xnd.ni_op = OP_LOOKUP;
3295#endif
3296 xnd.ni_cnd.cn_flags = LOCKLEAF;
3297 xnd.ni_pathlen = pathlen - 1;
3298 xnd.ni_cnd.cn_nameptr = xnd.ni_cnd.cn_pnbuf = path;
3299 xnd.ni_startdir = mvp;
3300 xnd.ni_usedvp = mvp;
3301 xnd.ni_rootdir = rootvnode;
3302 xnd.ni_cnd.cn_context = ctx;
3303 while ((error = lookup(&xnd)) == ERECYCLE) {
3304 xnd.ni_cnd.cn_flags = LOCKLEAF;
3305 xnd.ni_cnd.cn_nameptr = xnd.ni_cnd.cn_pnbuf;
3306 xnd.ni_usedvp = xnd.ni_dvp = xnd.ni_startdir = mvp;
3307 }
3308 if (error) {
3309 goto out1;
3310 }
3311 xvp = xnd.ni_vp;
3312 }
3313
3314 if (vnode_vtype(xvp) != VDIR) {
3315 error = EINVAL;
3316 vnode_put(xvp);
3317 goto out1;
3318 }
3319
3320 /* grab file handle */
3321 nx->nx_fh.nfh_len = NFSV3_MAX_FID_SIZE;
3322 error = VFS_VPTOFH(xvp, (int*)&nx->nx_fh.nfh_len, &nx->nx_fh.nfh_fid[0], NULL);
3323 if (!error && (nx->nx_fh.nfh_len > (int)NFSV3_MAX_FID_SIZE)) {
3324 error = EIO;
3325 } else {
3326 nx->nx_fh.nfh_xh.nxh_fidlen = nx->nx_fh.nfh_len;
3327 nx->nx_fh.nfh_len += sizeof(nx->nx_fh.nfh_xh);
3328 }
3329
3330 vnode_put(xvp);
3331 if (error) {
3332 goto out1;
3333 }
3334 } else {
3335 nx->nx_fh.nfh_xh.nxh_flags = NXHF_INVALIDFH;
3336 nx->nx_fh.nfh_xh.nxh_fidlen = 0;
3337 nx->nx_fh.nfh_len = sizeof(nx->nx_fh.nfh_xh);
3338 }
3339 }
3340 } else {
3341 nx = NULL;
3342 }
3343
3344 /* perform the export changes */
3345 if (unxa->nxa_flags & NXA_DELETE) {
3346 if (!nx) {
3347 /* delete all exports on this file system */
3348 while ((nx = LIST_FIRST(&nxfs->nxfs_exports))) {
3349 LIST_REMOVE(nx, nx_next);
3350 LIST_REMOVE(nx, nx_hash);
3351 /* delete all netopts for this export */
3352 nfsrv_free_addrlist(nx, NULL);
3353 nx->nx_flags &= ~NX_DEFAULTEXPORT;
3354 if (IS_VALID_CRED(nx->nx_defopt.nxo_cred)) {
3355 kauth_cred_unref(&nx->nx_defopt.nxo_cred);
3356 }
3357 /* delete active user list for this export */
3358 nfsrv_free_user_list(&nx->nx_user_list);
3359 FREE(nx->nx_path, M_TEMP);
3360 FREE(nx, M_TEMP);
3361 }
3362 goto out1;
3363 } else if (!unxa->nxa_netcount) {
3364 /* delete all netopts for this export */
3365 nfsrv_free_addrlist(nx, NULL);
3366 nx->nx_flags &= ~NX_DEFAULTEXPORT;
3367 if (IS_VALID_CRED(nx->nx_defopt.nxo_cred)) {
3368 kauth_cred_unref(&nx->nx_defopt.nxo_cred);
3369 }
3370 } else {
3371 /* delete only the netopts for the given addresses */
3372 error = nfsrv_free_addrlist(nx, unxa);
3373 if (error) {
3374 goto out1;
3375 }
3376 }
3377 }
3378 if (unxa->nxa_flags & NXA_ADD) {
3379 /*
3380 * If going offline set the export time so that when
3381 * coming back on line we will present a new write verifier
3382 * to the client.
3383 */
3384 if (unxa->nxa_flags & NXA_OFFLINE) {
3385 microtime(&nx->nx_exptime);
3386 }
3387
3388 error = nfsrv_hang_addrlist(nx, unxa);
3389 if (!error && mp) {
3390 vfs_setflags(mp, MNT_EXPORTED);
3391 }
3392 }
3393
3394out1:
3395 if (nx && !nx->nx_expcnt) {
3396 /* export has no export options */
3397 LIST_REMOVE(nx, nx_next);
3398 LIST_REMOVE(nx, nx_hash);
3399 /* delete active user list for this export */
3400 nfsrv_free_user_list(&nx->nx_user_list);
3401 FREE(nx->nx_path, M_TEMP);
3402 FREE(nx, M_TEMP);
3403 }
3404 if (LIST_EMPTY(&nxfs->nxfs_exports)) {
3405 /* exported file system has no more exports */
3406 LIST_REMOVE(nxfs, nxfs_next);
3407 FREE(nxfs->nxfs_path, M_TEMP);
3408 FREE(nxfs, M_TEMP);
3409 if (mp) {
3410 vfs_clearflags(mp, MNT_EXPORTED);
3411 }
3412 }
3413
3414out:
3415 if (mvp) {
3416 vnode_put(mvp);
3417 nameidone(&mnd);
3418 }
3419unlock_out:
3420 if (mp) {
3421 mount_drop(mp, 0);
3422 }
3423 lck_rw_done(&nfsrv_export_rwlock);
3424 return error;
3425}
3426
3427/*
3428 * Check if there is a least one export that will allow this address.
3429 *
3430 * Return 0, if there is an export that will allow this address,
3431 * else return EACCES
3432 */
3433int
3434nfsrv_check_exports_allow_address(mbuf_t nam)
3435{
3436 struct nfs_exportfs *nxfs;
3437 struct nfs_export *nx;
3438 struct nfs_export_options *nxo = NULL;
3439
3440 if (nam == NULL) {
3441 return EACCES;
3442 }
3443
3444 lck_rw_lock_shared(&nfsrv_export_rwlock);
3445 LIST_FOREACH(nxfs, &nfsrv_exports, nxfs_next) {
3446 LIST_FOREACH(nx, &nxfs->nxfs_exports, nx_next) {
3447 /* A little optimizing by checking for the default first */
3448 if (nx->nx_flags & NX_DEFAULTEXPORT) {
3449 nxo = &nx->nx_defopt;
3450 }
3451 if (nxo || (nxo = nfsrv_export_lookup(nx, nam))) {
3452 goto found;
3453 }
3454 }
3455 }
3456found:
3457 lck_rw_done(&nfsrv_export_rwlock);
3458
3459 return nxo ? 0 : EACCES;
3460}
3461
3462struct nfs_export_options *
3463nfsrv_export_lookup(struct nfs_export *nx, mbuf_t nam)
3464{
3465 struct nfs_export_options *nxo = NULL;
3466 struct nfs_netopt *no = NULL;
3467 struct radix_node_head *rnh;
3468 struct sockaddr *saddr;
3469
3470 /* Lookup in the export list first. */
3471 if (nam != NULL) {
3472 saddr = mbuf_data(nam);
3473 if (saddr->sa_family > AF_MAX) {
3474 /* Bogus sockaddr? Don't match anything. */
3475 return NULL;
3476 }
3477 rnh = nx->nx_rtable[saddr->sa_family];
3478 if (rnh != NULL) {
3479 no = (struct nfs_netopt *)
3480 (*rnh->rnh_matchaddr)((caddr_t)saddr, rnh);
3481 if (no && no->no_rnodes->rn_flags & RNF_ROOT) {
3482 no = NULL;
3483 }
3484 if (no) {
3485 nxo = &no->no_opt;
3486 }
3487 }
3488 }
3489 /* If no address match, use the default if it exists. */
3490 if ((nxo == NULL) && (nx->nx_flags & NX_DEFAULTEXPORT)) {
3491 nxo = &nx->nx_defopt;
3492 }
3493 return nxo;
3494}
3495
3496/* find an export for the given handle */
3497struct nfs_export *
3498nfsrv_fhtoexport(struct nfs_filehandle *nfhp)
3499{
3500 struct nfs_exphandle *nxh = (struct nfs_exphandle*)nfhp->nfh_fhp;
3501 struct nfs_export *nx;
3502 uint32_t fsid, expid;
3503
3504 if (!nfsrv_export_hashtbl) {
3505 return NULL;
3506 }
3507 fsid = ntohl(nxh->nxh_fsid);
3508 expid = ntohl(nxh->nxh_expid);
3509 nx = NFSRVEXPHASH(fsid, expid)->lh_first;
3510 for (; nx; nx = LIST_NEXT(nx, nx_hash)) {
3511 if (nx->nx_fs->nxfs_id != fsid) {
3512 continue;
3513 }
3514 if (nx->nx_id != expid) {
3515 continue;
3516 }
3517 break;
3518 }
3519 return nx;
3520}
3521
3522/*
3523 * nfsrv_fhtovp() - convert FH to vnode and export info
3524 */
3525int
3526nfsrv_fhtovp(
3527 struct nfs_filehandle *nfhp,
3528 struct nfsrv_descript *nd,
3529 vnode_t *vpp,
3530 struct nfs_export **nxp,
3531 struct nfs_export_options **nxop)
3532{
3533 struct nfs_exphandle *nxh = (struct nfs_exphandle*)nfhp->nfh_fhp;
3534 struct nfs_export_options *nxo;
3535 u_char *fidp;
3536 int error;
3537 struct mount *mp;
3538 mbuf_t nam = NULL;
3539 uint32_t v;
3540 int i, valid;
3541
3542 *vpp = NULL;
3543 *nxp = NULL;
3544 *nxop = NULL;
3545
3546 if (nd != NULL) {
3547 nam = nd->nd_nam;
3548 }
3549
3550 v = ntohl(nxh->nxh_version);
3551 if (v != NFS_FH_VERSION) {
3552 /* file handle format not supported */
3553 return ESTALE;
3554 }
3555 if (nfhp->nfh_len > NFSV3_MAX_FH_SIZE) {
3556 return EBADRPC;
3557 }
3558 if (nfhp->nfh_len < (int)sizeof(struct nfs_exphandle)) {
3559 return ESTALE;
3560 }
3561 v = ntohs(nxh->nxh_flags);
3562 if (v & NXHF_INVALIDFH) {
3563 return ESTALE;
3564 }
3565
3566 *nxp = nfsrv_fhtoexport(nfhp);
3567 if (!*nxp) {
3568 return ESTALE;
3569 }
3570
3571 /* Get the export option structure for this <export, client> tuple. */
3572 *nxop = nxo = nfsrv_export_lookup(*nxp, nam);
3573 if (nam && (*nxop == NULL)) {
3574 return EACCES;
3575 }
3576
3577 if (nd != NULL) {
3578 /* Validate the security flavor of the request */
3579 for (i = 0, valid = 0; i < nxo->nxo_sec.count; i++) {
3580 if (nd->nd_sec == nxo->nxo_sec.flavors[i]) {
3581 valid = 1;
3582 break;
3583 }
3584 }
3585 if (!valid) {
3586 /*
3587 * RFC 2623 section 2.3.2 recommends no authentication
3588 * requirement for certain NFS procedures used for mounting.
3589 * This allows an unauthenticated superuser on the client
3590 * to do mounts for the benefit of authenticated users.
3591 */
3592 if (nd->nd_vers == NFS_VER2) {
3593 if (nd->nd_procnum == NFSV2PROC_GETATTR ||
3594 nd->nd_procnum == NFSV2PROC_STATFS) {
3595 valid = 1;
3596 }
3597 }
3598 if (nd->nd_vers == NFS_VER3) {
3599 if (nd->nd_procnum == NFSPROC_FSINFO) {
3600 valid = 1;
3601 }
3602 }
3603
3604 if (!valid) {
3605 return NFSERR_AUTHERR | AUTH_REJECTCRED;
3606 }
3607 }
3608 }
3609
3610 if (nxo && (nxo->nxo_flags & NX_OFFLINE)) {
3611 return (nd == NULL || nd->nd_vers == NFS_VER2) ? ESTALE : NFSERR_TRYLATER;
3612 }
3613
3614 /* find mount structure */
3615 mp = vfs_getvfs_by_mntonname((*nxp)->nx_fs->nxfs_path);
3616 if (mp) {
3617 error = vfs_busy(mp, LK_NOWAIT);
3618 mount_iterdrop(mp);
3619 if (error) {
3620 mp = NULL;
3621 }
3622 }
3623 if (!mp) {
3624 /*
3625 * We have an export, but no mount?
3626 * Perhaps the export just hasn't been marked offline yet.
3627 */
3628 return (nd == NULL || nd->nd_vers == NFS_VER2) ? ESTALE : NFSERR_TRYLATER;
3629 }
3630
3631 fidp = nfhp->nfh_fhp + sizeof(*nxh);
3632 error = VFS_FHTOVP(mp, nxh->nxh_fidlen, fidp, vpp, NULL);
3633 vfs_unbusy(mp);
3634 if (error) {
3635 return error;
3636 }
3637 /* vnode pointer should be good at this point or ... */
3638 if (*vpp == NULL) {
3639 return ESTALE;
3640 }
3641 return 0;
3642}
3643
3644/*
3645 * nfsrv_credcheck() - check/map credentials according
3646 * to given export options.
3647 */
3648int
3649nfsrv_credcheck(
3650 struct nfsrv_descript *nd,
3651 vfs_context_t ctx,
3652 __unused struct nfs_export *nx,
3653 struct nfs_export_options *nxo)
3654{
3655 if (nxo && nxo->nxo_cred) {
3656 if ((nxo->nxo_flags & NX_MAPALL) ||
3657 ((nxo->nxo_flags & NX_MAPROOT) && !suser(nd->nd_cr, NULL))) {
3658 kauth_cred_ref(nxo->nxo_cred);
3659 kauth_cred_unref(&nd->nd_cr);
3660 nd->nd_cr = nxo->nxo_cred;
3661 }
3662 }
3663 ctx->vc_ucred = nd->nd_cr;
3664 return 0;
3665}
3666
3667/*
3668 * nfsrv_vptofh() - convert vnode to file handle for given export
3669 *
3670 * If the caller is passing in a vnode for a ".." directory entry,
3671 * they can pass a directory NFS file handle (dnfhp) which will be
3672 * checked against the root export file handle. If it matches, we
3673 * refuse to provide the file handle for the out-of-export directory.
3674 */
3675int
3676nfsrv_vptofh(
3677 struct nfs_export *nx,
3678 int nfsvers,
3679 struct nfs_filehandle *dnfhp,
3680 vnode_t vp,
3681 vfs_context_t ctx,
3682 struct nfs_filehandle *nfhp)
3683{
3684 int error;
3685 uint32_t maxfidsize;
3686
3687 nfhp->nfh_fhp = (u_char*)&nfhp->nfh_xh;
3688 nfhp->nfh_xh.nxh_version = htonl(NFS_FH_VERSION);
3689 nfhp->nfh_xh.nxh_fsid = htonl(nx->nx_fs->nxfs_id);
3690 nfhp->nfh_xh.nxh_expid = htonl(nx->nx_id);
3691 nfhp->nfh_xh.nxh_flags = 0;
3692 nfhp->nfh_xh.nxh_reserved = 0;
3693
3694 if (nfsvers == NFS_VER2) {
3695 bzero(&nfhp->nfh_fid[0], NFSV2_MAX_FID_SIZE);
3696 }
3697
3698 /* if directory FH matches export root, return invalid FH */
3699 if (dnfhp && nfsrv_fhmatch(dnfhp, &nx->nx_fh)) {
3700 if (nfsvers == NFS_VER2) {
3701 nfhp->nfh_len = NFSX_V2FH;
3702 } else {
3703 nfhp->nfh_len = sizeof(nfhp->nfh_xh);
3704 }
3705 nfhp->nfh_xh.nxh_fidlen = 0;
3706 nfhp->nfh_xh.nxh_flags = htons(NXHF_INVALIDFH);
3707 return 0;
3708 }
3709
3710 if (nfsvers == NFS_VER2) {
3711 maxfidsize = NFSV2_MAX_FID_SIZE;
3712 } else {
3713 maxfidsize = NFSV3_MAX_FID_SIZE;
3714 }
3715 nfhp->nfh_len = maxfidsize;
3716
3717 error = VFS_VPTOFH(vp, (int*)&nfhp->nfh_len, &nfhp->nfh_fid[0], ctx);
3718 if (error) {
3719 return error;
3720 }
3721 if (nfhp->nfh_len > maxfidsize) {
3722 return EOVERFLOW;
3723 }
3724 nfhp->nfh_xh.nxh_fidlen = nfhp->nfh_len;
3725 nfhp->nfh_len += sizeof(nfhp->nfh_xh);
3726 if ((nfsvers == NFS_VER2) && (nfhp->nfh_len < NFSX_V2FH)) {
3727 nfhp->nfh_len = NFSX_V2FH;
3728 }
3729
3730 return 0;
3731}
3732
3733/*
3734 * Compare two file handles to see it they're the same.
3735 * Note that we don't use nfh_len because that may include
3736 * padding in an NFSv2 file handle.
3737 */
3738int
3739nfsrv_fhmatch(struct nfs_filehandle *fh1, struct nfs_filehandle *fh2)
3740{
3741 struct nfs_exphandle *nxh1, *nxh2;
3742 int len1, len2;
3743
3744 nxh1 = (struct nfs_exphandle *)fh1->nfh_fhp;
3745 nxh2 = (struct nfs_exphandle *)fh2->nfh_fhp;
3746 len1 = sizeof(fh1->nfh_xh) + nxh1->nxh_fidlen;
3747 len2 = sizeof(fh2->nfh_xh) + nxh2->nxh_fidlen;
3748 if (len1 != len2) {
3749 return 0;
3750 }
3751 if (bcmp(nxh1, nxh2, len1)) {
3752 return 0;
3753 }
3754 return 1;
3755}
3756
3757/*
3758 * Functions for dealing with active user lists
3759 */
3760
3761/*
3762 * Search the hash table for a user node with a matching IP address and uid field.
3763 * If found, the node's tm_last timestamp is updated and the node is returned.
3764 *
3765 * If not found, a new node is allocated (or reclaimed via LRU), initialized, and returned.
3766 * Returns NULL if a new node could not be allcoated.
3767 *
3768 * The list's user_mutex lock MUST be held.
3769 */
3770struct nfs_user_stat_node *
3771nfsrv_get_user_stat_node(struct nfs_active_user_list *list, struct sockaddr *saddr, uid_t uid)
3772{
3773 struct nfs_user_stat_node *unode;
3774 struct timeval now;
3775 struct nfs_user_stat_hashtbl_head *head;
3776
3777 /* seach the hash table */
3778 head = NFS_USER_STAT_HASH(list->user_hashtbl, uid);
3779 LIST_FOREACH(unode, head, hash_link) {
3780 if ((uid == unode->uid) && (nfs_sockaddr_cmp(saddr, (struct sockaddr*)&unode->sock) == 0)) {
3781 /* found matching node */
3782 break;
3783 }
3784 }
3785
3786 if (unode) {
3787 /* found node in the hash table, now update lru position */
3788 TAILQ_REMOVE(&list->user_lru, unode, lru_link);
3789 TAILQ_INSERT_TAIL(&list->user_lru, unode, lru_link);
3790
3791 /* update time stamp */
3792 microtime(&now);
3793 unode->tm_last = (uint32_t)now.tv_sec;
3794 return unode;
3795 }
3796
3797 if (list->node_count < nfsrv_user_stat_max_nodes) {
3798 /* Allocate a new node */
3799 MALLOC(unode, struct nfs_user_stat_node *, sizeof(struct nfs_user_stat_node),
3800 M_TEMP, M_WAITOK | M_ZERO);
3801
3802 if (!unode) {
3803 return NULL;
3804 }
3805
3806 /* increment node count */
3807 OSAddAtomic(1, &nfsrv_user_stat_node_count);
3808 list->node_count++;
3809 } else {
3810 /* reuse the oldest node in the lru list */
3811 unode = TAILQ_FIRST(&list->user_lru);
3812
3813 if (!unode) {
3814 return NULL;
3815 }
3816
3817 /* Remove the node */
3818 TAILQ_REMOVE(&list->user_lru, unode, lru_link);
3819 LIST_REMOVE(unode, hash_link);
3820 }
3821
3822 /* Initialize the node */
3823 unode->uid = uid;
3824 bcopy(saddr, &unode->sock, saddr->sa_len);
3825 microtime(&now);
3826 unode->ops = 0;
3827 unode->bytes_read = 0;
3828 unode->bytes_written = 0;
3829 unode->tm_start = (uint32_t)now.tv_sec;
3830 unode->tm_last = (uint32_t)now.tv_sec;
3831
3832 /* insert the node */
3833 TAILQ_INSERT_TAIL(&list->user_lru, unode, lru_link);
3834 LIST_INSERT_HEAD(head, unode, hash_link);
3835
3836 return unode;
3837}
3838
3839void
3840nfsrv_update_user_stat(struct nfs_export *nx, struct nfsrv_descript *nd, uid_t uid, u_int ops, u_int rd_bytes, u_int wr_bytes)
3841{
3842 struct nfs_user_stat_node *unode;
3843 struct nfs_active_user_list *ulist;
3844 struct sockaddr *saddr;
3845
3846 if ((!nfsrv_user_stat_enabled) || (!nx) || (!nd) || (!nd->nd_nam)) {
3847 return;
3848 }
3849
3850 saddr = (struct sockaddr *)mbuf_data(nd->nd_nam);
3851
3852 /* check address family before going any further */
3853 if ((saddr->sa_family != AF_INET) && (saddr->sa_family != AF_INET6)) {
3854 return;
3855 }
3856
3857 ulist = &nx->nx_user_list;
3858
3859 /* lock the active user list */
3860 lck_mtx_lock(&ulist->user_mutex);
3861
3862 /* get the user node */
3863 unode = nfsrv_get_user_stat_node(ulist, saddr, uid);
3864
3865 if (!unode) {
3866 lck_mtx_unlock(&ulist->user_mutex);
3867 return;
3868 }
3869
3870 /* update counters */
3871 unode->ops += ops;
3872 unode->bytes_read += rd_bytes;
3873 unode->bytes_written += wr_bytes;
3874
3875 /* done */
3876 lck_mtx_unlock(&ulist->user_mutex);
3877}
3878
3879/* initialize an active user list */
3880void
3881nfsrv_init_user_list(struct nfs_active_user_list *ulist)
3882{
3883 uint i;
3884
3885 /* initialize the lru */
3886 TAILQ_INIT(&ulist->user_lru);
3887
3888 /* initialize the hash table */
3889 for (i = 0; i < NFS_USER_STAT_HASH_SIZE; i++) {
3890 LIST_INIT(&ulist->user_hashtbl[i]);
3891 }
3892 ulist->node_count = 0;
3893
3894 lck_mtx_init(&ulist->user_mutex, nfsrv_active_user_mutex_group, LCK_ATTR_NULL);
3895}
3896
3897/* Free all nodes in an active user list */
3898void
3899nfsrv_free_user_list(struct nfs_active_user_list *ulist)
3900{
3901 struct nfs_user_stat_node *unode;
3902
3903 if (!ulist) {
3904 return;
3905 }
3906
3907 while ((unode = TAILQ_FIRST(&ulist->user_lru))) {
3908 /* Remove node and free */
3909 TAILQ_REMOVE(&ulist->user_lru, unode, lru_link);
3910 LIST_REMOVE(unode, hash_link);
3911 FREE(unode, M_TEMP);
3912
3913 /* decrement node count */
3914 OSAddAtomic(-1, &nfsrv_user_stat_node_count);
3915 }
3916 ulist->node_count = 0;
3917
3918 lck_mtx_destroy(&ulist->user_mutex, nfsrv_active_user_mutex_group);
3919}
3920
3921/* Reclaim old expired user nodes from active user lists. */
3922void
3923nfsrv_active_user_list_reclaim(void)
3924{
3925 struct nfs_exportfs *nxfs;
3926 struct nfs_export *nx;
3927 struct nfs_active_user_list *ulist;
3928 struct nfs_user_stat_hashtbl_head oldlist;
3929 struct nfs_user_stat_node *unode, *unode_next;
3930 struct timeval now;
3931 uint32_t tstale;
3932
3933 LIST_INIT(&oldlist);
3934
3935 lck_rw_lock_shared(&nfsrv_export_rwlock);
3936 microtime(&now);
3937 tstale = now.tv_sec - nfsrv_user_stat_max_idle_sec;
3938 LIST_FOREACH(nxfs, &nfsrv_exports, nxfs_next) {
3939 LIST_FOREACH(nx, &nxfs->nxfs_exports, nx_next) {
3940 /* Scan through all user nodes of this export */
3941 ulist = &nx->nx_user_list;
3942 lck_mtx_lock(&ulist->user_mutex);
3943 for (unode = TAILQ_FIRST(&ulist->user_lru); unode; unode = unode_next) {
3944 unode_next = TAILQ_NEXT(unode, lru_link);
3945
3946 /* check if this node has expired */
3947 if (unode->tm_last >= tstale) {
3948 break;
3949 }
3950
3951 /* Remove node from the active user list */
3952 TAILQ_REMOVE(&ulist->user_lru, unode, lru_link);
3953 LIST_REMOVE(unode, hash_link);
3954
3955 /* Add node to temp list */
3956 LIST_INSERT_HEAD(&oldlist, unode, hash_link);
3957
3958 /* decrement node count */
3959 OSAddAtomic(-1, &nfsrv_user_stat_node_count);
3960 ulist->node_count--;
3961 }
3962 /* can unlock this export's list now */
3963 lck_mtx_unlock(&ulist->user_mutex);
3964 }
3965 }
3966 lck_rw_done(&nfsrv_export_rwlock);
3967
3968 /* Free expired nodes */
3969 while ((unode = LIST_FIRST(&oldlist))) {
3970 LIST_REMOVE(unode, hash_link);
3971 FREE(unode, M_TEMP);
3972 }
3973}
3974
3975/*
3976 * Maps errno values to nfs error numbers.
3977 * Use NFSERR_IO as the catch all for ones not specifically defined in
3978 * RFC 1094.
3979 */
3980static u_char nfsrv_v2errmap[] = {
3981 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO,
3982 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
3983 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO,
3984 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR,
3985 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
3986 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS,
3987 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
3988 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
3989 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
3990 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
3991 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
3992 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
3993 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO,
3994 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE,
3995};
3996
3997/*
3998 * Maps errno values to nfs error numbers.
3999 * Although it is not obvious whether or not NFS clients really care if
4000 * a returned error value is in the specified list for the procedure, the
4001 * safest thing to do is filter them appropriately. For Version 2, the
4002 * X/Open XNFS document is the only specification that defines error values
4003 * for each RPC (The RFC simply lists all possible error values for all RPCs),
4004 * so I have decided to not do this for Version 2.
4005 * The first entry is the default error return and the rest are the valid
4006 * errors for that RPC in increasing numeric order.
4007 */
4008static short nfsv3err_null[] = {
4009 0,
4010 0,
4011};
4012
4013static short nfsv3err_getattr[] = {
4014 NFSERR_IO,
4015 NFSERR_IO,
4016 NFSERR_STALE,
4017 NFSERR_BADHANDLE,
4018 NFSERR_SERVERFAULT,
4019 NFSERR_TRYLATER,
4020 0,
4021};
4022
4023static short nfsv3err_setattr[] = {
4024 NFSERR_IO,
4025 NFSERR_PERM,
4026 NFSERR_IO,
4027 NFSERR_ACCES,
4028 NFSERR_INVAL,
4029 NFSERR_NOSPC,
4030 NFSERR_ROFS,
4031 NFSERR_DQUOT,
4032 NFSERR_STALE,
4033 NFSERR_BADHANDLE,
4034 NFSERR_NOT_SYNC,
4035 NFSERR_SERVERFAULT,
4036 NFSERR_TRYLATER,
4037 0,
4038};
4039
4040static short nfsv3err_lookup[] = {
4041 NFSERR_IO,
4042 NFSERR_NOENT,
4043 NFSERR_IO,
4044 NFSERR_ACCES,
4045 NFSERR_NOTDIR,
4046 NFSERR_NAMETOL,
4047 NFSERR_STALE,
4048 NFSERR_BADHANDLE,
4049 NFSERR_SERVERFAULT,
4050 NFSERR_TRYLATER,
4051 0,
4052};
4053
4054static short nfsv3err_access[] = {
4055 NFSERR_IO,
4056 NFSERR_IO,
4057 NFSERR_STALE,
4058 NFSERR_BADHANDLE,
4059 NFSERR_SERVERFAULT,
4060 NFSERR_TRYLATER,
4061 0,
4062};
4063
4064static short nfsv3err_readlink[] = {
4065 NFSERR_IO,
4066 NFSERR_IO,
4067 NFSERR_ACCES,
4068 NFSERR_INVAL,
4069 NFSERR_STALE,
4070 NFSERR_BADHANDLE,
4071 NFSERR_NOTSUPP,
4072 NFSERR_SERVERFAULT,
4073 NFSERR_TRYLATER,
4074 0,
4075};
4076
4077static short nfsv3err_read[] = {
4078 NFSERR_IO,
4079 NFSERR_IO,
4080 NFSERR_NXIO,
4081 NFSERR_ACCES,
4082 NFSERR_INVAL,
4083 NFSERR_STALE,
4084 NFSERR_BADHANDLE,
4085 NFSERR_SERVERFAULT,
4086 NFSERR_TRYLATER,
4087 0,
4088};
4089
4090static short nfsv3err_write[] = {
4091 NFSERR_IO,
4092 NFSERR_IO,
4093 NFSERR_ACCES,
4094 NFSERR_INVAL,
4095 NFSERR_FBIG,
4096 NFSERR_NOSPC,
4097 NFSERR_ROFS,
4098 NFSERR_DQUOT,
4099 NFSERR_STALE,
4100 NFSERR_BADHANDLE,
4101 NFSERR_SERVERFAULT,
4102 NFSERR_TRYLATER,
4103 0,
4104};
4105
4106static short nfsv3err_create[] = {
4107 NFSERR_IO,
4108 NFSERR_IO,
4109 NFSERR_ACCES,
4110 NFSERR_EXIST,
4111 NFSERR_NOTDIR,
4112 NFSERR_NOSPC,
4113 NFSERR_ROFS,
4114 NFSERR_NAMETOL,
4115 NFSERR_DQUOT,
4116 NFSERR_STALE,
4117 NFSERR_BADHANDLE,
4118 NFSERR_NOTSUPP,
4119 NFSERR_SERVERFAULT,
4120 NFSERR_TRYLATER,
4121 0,
4122};
4123
4124static short nfsv3err_mkdir[] = {
4125 NFSERR_IO,
4126 NFSERR_IO,
4127 NFSERR_ACCES,
4128 NFSERR_EXIST,
4129 NFSERR_NOTDIR,
4130 NFSERR_NOSPC,
4131 NFSERR_ROFS,
4132 NFSERR_NAMETOL,
4133 NFSERR_DQUOT,
4134 NFSERR_STALE,
4135 NFSERR_BADHANDLE,
4136 NFSERR_NOTSUPP,
4137 NFSERR_SERVERFAULT,
4138 NFSERR_TRYLATER,
4139 0,
4140};
4141
4142static short nfsv3err_symlink[] = {
4143 NFSERR_IO,
4144 NFSERR_IO,
4145 NFSERR_ACCES,
4146 NFSERR_EXIST,
4147 NFSERR_NOTDIR,
4148 NFSERR_NOSPC,
4149 NFSERR_ROFS,
4150 NFSERR_NAMETOL,
4151 NFSERR_DQUOT,
4152 NFSERR_STALE,
4153 NFSERR_BADHANDLE,
4154 NFSERR_NOTSUPP,
4155 NFSERR_SERVERFAULT,
4156 NFSERR_TRYLATER,
4157 0,
4158};
4159
4160static short nfsv3err_mknod[] = {
4161 NFSERR_IO,
4162 NFSERR_IO,
4163 NFSERR_ACCES,
4164 NFSERR_EXIST,
4165 NFSERR_NOTDIR,
4166 NFSERR_NOSPC,
4167 NFSERR_ROFS,
4168 NFSERR_NAMETOL,
4169 NFSERR_DQUOT,
4170 NFSERR_STALE,
4171 NFSERR_BADHANDLE,
4172 NFSERR_NOTSUPP,
4173 NFSERR_SERVERFAULT,
4174 NFSERR_BADTYPE,
4175 NFSERR_TRYLATER,
4176 0,
4177};
4178
4179static short nfsv3err_remove[] = {
4180 NFSERR_IO,
4181 NFSERR_NOENT,
4182 NFSERR_IO,
4183 NFSERR_ACCES,
4184 NFSERR_NOTDIR,
4185 NFSERR_ROFS,
4186 NFSERR_NAMETOL,
4187 NFSERR_STALE,
4188 NFSERR_BADHANDLE,
4189 NFSERR_SERVERFAULT,
4190 NFSERR_TRYLATER,
4191 0,
4192};
4193
4194static short nfsv3err_rmdir[] = {
4195 NFSERR_IO,
4196 NFSERR_NOENT,
4197 NFSERR_IO,
4198 NFSERR_ACCES,
4199 NFSERR_EXIST,
4200 NFSERR_NOTDIR,
4201 NFSERR_INVAL,
4202 NFSERR_ROFS,
4203 NFSERR_NAMETOL,
4204 NFSERR_NOTEMPTY,
4205 NFSERR_STALE,
4206 NFSERR_BADHANDLE,
4207 NFSERR_NOTSUPP,
4208 NFSERR_SERVERFAULT,
4209 NFSERR_TRYLATER,
4210 0,
4211};
4212
4213static short nfsv3err_rename[] = {
4214 NFSERR_IO,
4215 NFSERR_NOENT,
4216 NFSERR_IO,
4217 NFSERR_ACCES,
4218 NFSERR_EXIST,
4219 NFSERR_XDEV,
4220 NFSERR_NOTDIR,
4221 NFSERR_ISDIR,
4222 NFSERR_INVAL,
4223 NFSERR_NOSPC,
4224 NFSERR_ROFS,
4225 NFSERR_MLINK,
4226 NFSERR_NAMETOL,
4227 NFSERR_NOTEMPTY,
4228 NFSERR_DQUOT,
4229 NFSERR_STALE,
4230 NFSERR_BADHANDLE,
4231 NFSERR_NOTSUPP,
4232 NFSERR_SERVERFAULT,
4233 NFSERR_TRYLATER,
4234 0,
4235};
4236
4237static short nfsv3err_link[] = {
4238 NFSERR_IO,
4239 NFSERR_IO,
4240 NFSERR_ACCES,
4241 NFSERR_EXIST,
4242 NFSERR_XDEV,
4243 NFSERR_NOTDIR,
4244 NFSERR_INVAL,
4245 NFSERR_NOSPC,
4246 NFSERR_ROFS,
4247 NFSERR_MLINK,
4248 NFSERR_NAMETOL,
4249 NFSERR_DQUOT,
4250 NFSERR_STALE,
4251 NFSERR_BADHANDLE,
4252 NFSERR_NOTSUPP,
4253 NFSERR_SERVERFAULT,
4254 NFSERR_TRYLATER,
4255 0,
4256};
4257
4258static short nfsv3err_readdir[] = {
4259 NFSERR_IO,
4260 NFSERR_IO,
4261 NFSERR_ACCES,
4262 NFSERR_NOTDIR,
4263 NFSERR_STALE,
4264 NFSERR_BADHANDLE,
4265 NFSERR_BAD_COOKIE,
4266 NFSERR_TOOSMALL,
4267 NFSERR_SERVERFAULT,
4268 NFSERR_TRYLATER,
4269 0,
4270};
4271
4272static short nfsv3err_readdirplus[] = {
4273 NFSERR_IO,
4274 NFSERR_IO,
4275 NFSERR_ACCES,
4276 NFSERR_NOTDIR,
4277 NFSERR_STALE,
4278 NFSERR_BADHANDLE,
4279 NFSERR_BAD_COOKIE,
4280 NFSERR_NOTSUPP,
4281 NFSERR_TOOSMALL,
4282 NFSERR_SERVERFAULT,
4283 NFSERR_TRYLATER,
4284 0,
4285};
4286
4287static short nfsv3err_fsstat[] = {
4288 NFSERR_IO,
4289 NFSERR_IO,
4290 NFSERR_STALE,
4291 NFSERR_BADHANDLE,
4292 NFSERR_SERVERFAULT,
4293 NFSERR_TRYLATER,
4294 0,
4295};
4296
4297static short nfsv3err_fsinfo[] = {
4298 NFSERR_STALE,
4299 NFSERR_STALE,
4300 NFSERR_BADHANDLE,
4301 NFSERR_SERVERFAULT,
4302 NFSERR_TRYLATER,
4303 0,
4304};
4305
4306static short nfsv3err_pathconf[] = {
4307 NFSERR_STALE,
4308 NFSERR_STALE,
4309 NFSERR_BADHANDLE,
4310 NFSERR_SERVERFAULT,
4311 NFSERR_TRYLATER,
4312 0,
4313};
4314
4315static short nfsv3err_commit[] = {
4316 NFSERR_IO,
4317 NFSERR_IO,
4318 NFSERR_STALE,
4319 NFSERR_BADHANDLE,
4320 NFSERR_SERVERFAULT,
4321 NFSERR_TRYLATER,
4322 0,
4323};
4324
4325static short *nfsrv_v3errmap[] = {
4326 nfsv3err_null,
4327 nfsv3err_getattr,
4328 nfsv3err_setattr,
4329 nfsv3err_lookup,
4330 nfsv3err_access,
4331 nfsv3err_readlink,
4332 nfsv3err_read,
4333 nfsv3err_write,
4334 nfsv3err_create,
4335 nfsv3err_mkdir,
4336 nfsv3err_symlink,
4337 nfsv3err_mknod,
4338 nfsv3err_remove,
4339 nfsv3err_rmdir,
4340 nfsv3err_rename,
4341 nfsv3err_link,
4342 nfsv3err_readdir,
4343 nfsv3err_readdirplus,
4344 nfsv3err_fsstat,
4345 nfsv3err_fsinfo,
4346 nfsv3err_pathconf,
4347 nfsv3err_commit,
4348};
4349
4350/*
4351 * Map errnos to NFS error numbers. For Version 3 also filter out error
4352 * numbers not specified for the associated procedure.
4353 */
4354int
4355nfsrv_errmap(struct nfsrv_descript *nd, int err)
4356{
4357 short *defaulterrp, *errp;
4358
4359 if (nd->nd_vers == NFS_VER2) {
4360 if (err <= (int)sizeof(nfsrv_v2errmap)) {
4361 return (int)nfsrv_v2errmap[err - 1];
4362 }
4363 return NFSERR_IO;
4364 }
4365 /* NFSv3 */
4366 if (nd->nd_procnum > NFSPROC_COMMIT) {
4367 return err & 0xffff;
4368 }
4369 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
4370 while (*++errp) {
4371 if (*errp == err) {
4372 return err;
4373 } else if (*errp > err) {
4374 break;
4375 }
4376 }
4377 return (int)*defaulterrp;
4378}
4379
4380#endif /* NFSSERVER */