nfs_srvsubs.c revision 76117
1/*
2 * Copyright (c) 1989, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 *    must display the following acknowledgement:
18 *	This product includes software developed by the University of
19 *	California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 *    may be used to endorse or promote products derived from this software
22 *    without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 *	@(#)nfs_subs.c  8.8 (Berkeley) 5/22/95
37 * $FreeBSD: head/sys/nfsserver/nfs_srvsubs.c 76117 2001-04-29 02:45:39Z grog $
38 */
39
40/*
41 * These functions support the macros and help fiddle mbuf chains for
42 * the nfs op functions. They do things like create the rpc header and
43 * copy data between mbuf chains and uio lists.
44 */
45#include <sys/param.h>
46#include <sys/systm.h>
47#include <sys/kernel.h>
48#include <sys/bio.h>
49#include <sys/buf.h>
50#include <sys/proc.h>
51#include <sys/mount.h>
52#include <sys/vnode.h>
53#include <sys/namei.h>
54#include <sys/mbuf.h>
55#include <sys/socket.h>
56#include <sys/stat.h>
57#include <sys/malloc.h>
58#include <sys/sysent.h>
59#include <sys/syscall.h>
60
61#include <vm/vm.h>
62#include <vm/vm_object.h>
63#include <vm/vm_extern.h>
64#include <vm/vm_zone.h>
65
66#include <nfs/rpcv2.h>
67#include <nfs/nfsproto.h>
68#include <nfs/nfs.h>
69#include <nfs/nfsnode.h>
70#include <nfs/xdr_subs.h>
71#include <nfs/nfsm_subs.h>
72#include <nfs/nfsmount.h>
73#include <nfs/nqnfs.h>
74#include <nfs/nfsrtt.h>
75
76#include <netinet/in.h>
77
78/*
79 * Data items converted to xdr at startup, since they are constant
80 * This is kinda hokey, but may save a little time doing byte swaps
81 */
82u_int32_t nfs_xdrneg1;
83u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
84	rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
85	rpc_auth_kerb;
86u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
87
88/* And other global data */
89static u_int32_t nfs_xid = 0;
90static enum vtype nv2tov_type[8]= {
91	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON,  VNON
92};
93enum vtype nv3tov_type[8]= {
94	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
95};
96
97int nfs_ticks;
98int nfs_pbuf_freecnt = -1;	/* start out unlimited */
99
100struct nfs_reqq nfs_reqq;
101struct nfssvc_sockhead nfssvc_sockhead;
102int nfssvc_sockhead_flag;
103struct nfsd_head nfsd_head;
104int nfsd_head_flag;
105struct nfs_bufq nfs_bufq;
106struct nqtimerhead nqtimerhead;
107struct nqfhhashhead *nqfhhashtbl;
108u_long nqfhhash;
109
110static void (*nfs_prev_lease_updatetime) __P((int));
111static int nfs_prev_nfssvc_sy_narg;
112static sy_call_t *nfs_prev_nfssvc_sy_call;
113
114#ifndef NFS_NOSERVER
115
116static vop_t *nfs_prev_vop_lease_check;
117
118/*
119 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
120 */
121int nfsv3_procid[NFS_NPROCS] = {
122	NFSPROC_NULL,
123	NFSPROC_GETATTR,
124	NFSPROC_SETATTR,
125	NFSPROC_NOOP,
126	NFSPROC_LOOKUP,
127	NFSPROC_READLINK,
128	NFSPROC_READ,
129	NFSPROC_NOOP,
130	NFSPROC_WRITE,
131	NFSPROC_CREATE,
132	NFSPROC_REMOVE,
133	NFSPROC_RENAME,
134	NFSPROC_LINK,
135	NFSPROC_SYMLINK,
136	NFSPROC_MKDIR,
137	NFSPROC_RMDIR,
138	NFSPROC_READDIR,
139	NFSPROC_FSSTAT,
140	NFSPROC_NOOP,
141	NFSPROC_NOOP,
142	NFSPROC_NOOP,
143	NFSPROC_NOOP,
144	NFSPROC_NOOP,
145	NFSPROC_NOOP,
146	NFSPROC_NOOP,
147	NFSPROC_NOOP
148};
149
150#endif /* NFS_NOSERVER */
151/*
152 * and the reverse mapping from generic to Version 2 procedure numbers
153 */
154int nfsv2_procid[NFS_NPROCS] = {
155	NFSV2PROC_NULL,
156	NFSV2PROC_GETATTR,
157	NFSV2PROC_SETATTR,
158	NFSV2PROC_LOOKUP,
159	NFSV2PROC_NOOP,
160	NFSV2PROC_READLINK,
161	NFSV2PROC_READ,
162	NFSV2PROC_WRITE,
163	NFSV2PROC_CREATE,
164	NFSV2PROC_MKDIR,
165	NFSV2PROC_SYMLINK,
166	NFSV2PROC_CREATE,
167	NFSV2PROC_REMOVE,
168	NFSV2PROC_RMDIR,
169	NFSV2PROC_RENAME,
170	NFSV2PROC_LINK,
171	NFSV2PROC_READDIR,
172	NFSV2PROC_NOOP,
173	NFSV2PROC_STATFS,
174	NFSV2PROC_NOOP,
175	NFSV2PROC_NOOP,
176	NFSV2PROC_NOOP,
177	NFSV2PROC_NOOP,
178	NFSV2PROC_NOOP,
179	NFSV2PROC_NOOP,
180	NFSV2PROC_NOOP,
181};
182
183#ifndef NFS_NOSERVER
184/*
185 * Maps errno values to nfs error numbers.
186 * Use NFSERR_IO as the catch all for ones not specifically defined in
187 * RFC 1094.
188 */
189static u_char nfsrv_v2errmap[ELAST] = {
190  NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
191  NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
192  NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
193  NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
194  NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
195  NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
196  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
197  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
198  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
199  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
200  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
201  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
202  NFSERR_IO,	NFSERR_IO,	NFSERR_NAMETOL,	NFSERR_IO,	NFSERR_IO,
203  NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
204  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
205  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
206  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
207  NFSERR_IO /* << Last is 86 */
208};
209
210/*
211 * Maps errno values to nfs error numbers.
212 * Although it is not obvious whether or not NFS clients really care if
213 * a returned error value is in the specified list for the procedure, the
214 * safest thing to do is filter them appropriately. For Version 2, the
215 * X/Open XNFS document is the only specification that defines error values
216 * for each RPC (The RFC simply lists all possible error values for all RPCs),
217 * so I have decided to not do this for Version 2.
218 * The first entry is the default error return and the rest are the valid
219 * errors for that RPC in increasing numeric order.
220 */
221static short nfsv3err_null[] = {
222	0,
223	0,
224};
225
226static short nfsv3err_getattr[] = {
227	NFSERR_IO,
228	NFSERR_IO,
229	NFSERR_STALE,
230	NFSERR_BADHANDLE,
231	NFSERR_SERVERFAULT,
232	0,
233};
234
235static short nfsv3err_setattr[] = {
236	NFSERR_IO,
237	NFSERR_PERM,
238	NFSERR_IO,
239	NFSERR_ACCES,
240	NFSERR_INVAL,
241	NFSERR_NOSPC,
242	NFSERR_ROFS,
243	NFSERR_DQUOT,
244	NFSERR_STALE,
245	NFSERR_BADHANDLE,
246	NFSERR_NOT_SYNC,
247	NFSERR_SERVERFAULT,
248	0,
249};
250
251static short nfsv3err_lookup[] = {
252	NFSERR_IO,
253	NFSERR_NOENT,
254	NFSERR_IO,
255	NFSERR_ACCES,
256	NFSERR_NOTDIR,
257	NFSERR_NAMETOL,
258	NFSERR_STALE,
259	NFSERR_BADHANDLE,
260	NFSERR_SERVERFAULT,
261	0,
262};
263
264static short nfsv3err_access[] = {
265	NFSERR_IO,
266	NFSERR_IO,
267	NFSERR_STALE,
268	NFSERR_BADHANDLE,
269	NFSERR_SERVERFAULT,
270	0,
271};
272
273static short nfsv3err_readlink[] = {
274	NFSERR_IO,
275	NFSERR_IO,
276	NFSERR_ACCES,
277	NFSERR_INVAL,
278	NFSERR_STALE,
279	NFSERR_BADHANDLE,
280	NFSERR_NOTSUPP,
281	NFSERR_SERVERFAULT,
282	0,
283};
284
285static short nfsv3err_read[] = {
286	NFSERR_IO,
287	NFSERR_IO,
288	NFSERR_NXIO,
289	NFSERR_ACCES,
290	NFSERR_INVAL,
291	NFSERR_STALE,
292	NFSERR_BADHANDLE,
293	NFSERR_SERVERFAULT,
294	0,
295};
296
297static short nfsv3err_write[] = {
298	NFSERR_IO,
299	NFSERR_IO,
300	NFSERR_ACCES,
301	NFSERR_INVAL,
302	NFSERR_FBIG,
303	NFSERR_NOSPC,
304	NFSERR_ROFS,
305	NFSERR_DQUOT,
306	NFSERR_STALE,
307	NFSERR_BADHANDLE,
308	NFSERR_SERVERFAULT,
309	0,
310};
311
312static short nfsv3err_create[] = {
313	NFSERR_IO,
314	NFSERR_IO,
315	NFSERR_ACCES,
316	NFSERR_EXIST,
317	NFSERR_NOTDIR,
318	NFSERR_NOSPC,
319	NFSERR_ROFS,
320	NFSERR_NAMETOL,
321	NFSERR_DQUOT,
322	NFSERR_STALE,
323	NFSERR_BADHANDLE,
324	NFSERR_NOTSUPP,
325	NFSERR_SERVERFAULT,
326	0,
327};
328
329static short nfsv3err_mkdir[] = {
330	NFSERR_IO,
331	NFSERR_IO,
332	NFSERR_ACCES,
333	NFSERR_EXIST,
334	NFSERR_NOTDIR,
335	NFSERR_NOSPC,
336	NFSERR_ROFS,
337	NFSERR_NAMETOL,
338	NFSERR_DQUOT,
339	NFSERR_STALE,
340	NFSERR_BADHANDLE,
341	NFSERR_NOTSUPP,
342	NFSERR_SERVERFAULT,
343	0,
344};
345
346static short nfsv3err_symlink[] = {
347	NFSERR_IO,
348	NFSERR_IO,
349	NFSERR_ACCES,
350	NFSERR_EXIST,
351	NFSERR_NOTDIR,
352	NFSERR_NOSPC,
353	NFSERR_ROFS,
354	NFSERR_NAMETOL,
355	NFSERR_DQUOT,
356	NFSERR_STALE,
357	NFSERR_BADHANDLE,
358	NFSERR_NOTSUPP,
359	NFSERR_SERVERFAULT,
360	0,
361};
362
363static short nfsv3err_mknod[] = {
364	NFSERR_IO,
365	NFSERR_IO,
366	NFSERR_ACCES,
367	NFSERR_EXIST,
368	NFSERR_NOTDIR,
369	NFSERR_NOSPC,
370	NFSERR_ROFS,
371	NFSERR_NAMETOL,
372	NFSERR_DQUOT,
373	NFSERR_STALE,
374	NFSERR_BADHANDLE,
375	NFSERR_NOTSUPP,
376	NFSERR_SERVERFAULT,
377	NFSERR_BADTYPE,
378	0,
379};
380
381static short nfsv3err_remove[] = {
382	NFSERR_IO,
383	NFSERR_NOENT,
384	NFSERR_IO,
385	NFSERR_ACCES,
386	NFSERR_NOTDIR,
387	NFSERR_ROFS,
388	NFSERR_NAMETOL,
389	NFSERR_STALE,
390	NFSERR_BADHANDLE,
391	NFSERR_SERVERFAULT,
392	0,
393};
394
395static short nfsv3err_rmdir[] = {
396	NFSERR_IO,
397	NFSERR_NOENT,
398	NFSERR_IO,
399	NFSERR_ACCES,
400	NFSERR_EXIST,
401	NFSERR_NOTDIR,
402	NFSERR_INVAL,
403	NFSERR_ROFS,
404	NFSERR_NAMETOL,
405	NFSERR_NOTEMPTY,
406	NFSERR_STALE,
407	NFSERR_BADHANDLE,
408	NFSERR_NOTSUPP,
409	NFSERR_SERVERFAULT,
410	0,
411};
412
413static short nfsv3err_rename[] = {
414	NFSERR_IO,
415	NFSERR_NOENT,
416	NFSERR_IO,
417	NFSERR_ACCES,
418	NFSERR_EXIST,
419	NFSERR_XDEV,
420	NFSERR_NOTDIR,
421	NFSERR_ISDIR,
422	NFSERR_INVAL,
423	NFSERR_NOSPC,
424	NFSERR_ROFS,
425	NFSERR_MLINK,
426	NFSERR_NAMETOL,
427	NFSERR_NOTEMPTY,
428	NFSERR_DQUOT,
429	NFSERR_STALE,
430	NFSERR_BADHANDLE,
431	NFSERR_NOTSUPP,
432	NFSERR_SERVERFAULT,
433	0,
434};
435
436static short nfsv3err_link[] = {
437	NFSERR_IO,
438	NFSERR_IO,
439	NFSERR_ACCES,
440	NFSERR_EXIST,
441	NFSERR_XDEV,
442	NFSERR_NOTDIR,
443	NFSERR_INVAL,
444	NFSERR_NOSPC,
445	NFSERR_ROFS,
446	NFSERR_MLINK,
447	NFSERR_NAMETOL,
448	NFSERR_DQUOT,
449	NFSERR_STALE,
450	NFSERR_BADHANDLE,
451	NFSERR_NOTSUPP,
452	NFSERR_SERVERFAULT,
453	0,
454};
455
456static short nfsv3err_readdir[] = {
457	NFSERR_IO,
458	NFSERR_IO,
459	NFSERR_ACCES,
460	NFSERR_NOTDIR,
461	NFSERR_STALE,
462	NFSERR_BADHANDLE,
463	NFSERR_BAD_COOKIE,
464	NFSERR_TOOSMALL,
465	NFSERR_SERVERFAULT,
466	0,
467};
468
469static short nfsv3err_readdirplus[] = {
470	NFSERR_IO,
471	NFSERR_IO,
472	NFSERR_ACCES,
473	NFSERR_NOTDIR,
474	NFSERR_STALE,
475	NFSERR_BADHANDLE,
476	NFSERR_BAD_COOKIE,
477	NFSERR_NOTSUPP,
478	NFSERR_TOOSMALL,
479	NFSERR_SERVERFAULT,
480	0,
481};
482
483static short nfsv3err_fsstat[] = {
484	NFSERR_IO,
485	NFSERR_IO,
486	NFSERR_STALE,
487	NFSERR_BADHANDLE,
488	NFSERR_SERVERFAULT,
489	0,
490};
491
492static short nfsv3err_fsinfo[] = {
493	NFSERR_STALE,
494	NFSERR_STALE,
495	NFSERR_BADHANDLE,
496	NFSERR_SERVERFAULT,
497	0,
498};
499
500static short nfsv3err_pathconf[] = {
501	NFSERR_STALE,
502	NFSERR_STALE,
503	NFSERR_BADHANDLE,
504	NFSERR_SERVERFAULT,
505	0,
506};
507
508static short nfsv3err_commit[] = {
509	NFSERR_IO,
510	NFSERR_IO,
511	NFSERR_STALE,
512	NFSERR_BADHANDLE,
513	NFSERR_SERVERFAULT,
514	0,
515};
516
517static short *nfsrv_v3errmap[] = {
518	nfsv3err_null,
519	nfsv3err_getattr,
520	nfsv3err_setattr,
521	nfsv3err_lookup,
522	nfsv3err_access,
523	nfsv3err_readlink,
524	nfsv3err_read,
525	nfsv3err_write,
526	nfsv3err_create,
527	nfsv3err_mkdir,
528	nfsv3err_symlink,
529	nfsv3err_mknod,
530	nfsv3err_remove,
531	nfsv3err_rmdir,
532	nfsv3err_rename,
533	nfsv3err_link,
534	nfsv3err_readdir,
535	nfsv3err_readdirplus,
536	nfsv3err_fsstat,
537	nfsv3err_fsinfo,
538	nfsv3err_pathconf,
539	nfsv3err_commit,
540};
541
542#endif /* NFS_NOSERVER */
543
544extern struct nfsrtt nfsrtt;
545extern time_t nqnfsstarttime;
546extern int nqsrv_clockskew;
547extern int nqsrv_writeslack;
548extern int nqsrv_maxlease;
549extern struct nfsstats nfsstats;
550extern int nqnfs_piggy[NFS_NPROCS];
551extern nfstype nfsv2_type[9];
552extern nfstype nfsv3_type[9];
553extern struct nfsnodehashhead *nfsnodehashtbl;
554extern u_long nfsnodehash;
555
556struct nfssvc_args;
557extern int nfssvc(struct proc *, struct nfssvc_args *, int *);
558
559LIST_HEAD(nfsnodehashhead, nfsnode);
560
561int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
562
563u_quad_t
564nfs_curusec()
565{
566	struct timeval tv;
567
568	getmicrotime(&tv);
569	return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
570}
571
572/*
573 * Create the header for an rpc request packet
574 * The hsiz is the size of the rest of the nfs request header.
575 * (just used to decide if a cluster is a good idea)
576 */
577struct mbuf *
578nfsm_reqh(vp, procid, hsiz, bposp)
579	struct vnode *vp;
580	u_long procid;
581	int hsiz;
582	caddr_t *bposp;
583{
584	register struct mbuf *mb;
585	register u_int32_t *tl;
586	register caddr_t bpos;
587	struct mbuf *mb2;
588	struct nfsmount *nmp;
589	int nqflag;
590
591	MGET(mb, M_TRYWAIT, MT_DATA);
592	if (hsiz >= MINCLSIZE)
593		MCLGET(mb, M_TRYWAIT);
594	mb->m_len = 0;
595	bpos = mtod(mb, caddr_t);
596
597	/*
598	 * For NQNFS, add lease request.
599	 */
600	if (vp) {
601		nmp = VFSTONFS(vp->v_mount);
602		if (nmp->nm_flag & NFSMNT_NQNFS) {
603			nqflag = NQNFS_NEEDLEASE(vp, procid);
604			if (nqflag) {
605				nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
606				*tl++ = txdr_unsigned(nqflag);
607				*tl = txdr_unsigned(nmp->nm_leaseterm);
608			} else {
609				nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
610				*tl = 0;
611			}
612		}
613	}
614	/* Finally, return values */
615	*bposp = bpos;
616	return (mb);
617}
618
619/*
620 * Build the RPC header and fill in the authorization info.
621 * The authorization string argument is only used when the credentials
622 * come from outside of the kernel.
623 * Returns the head of the mbuf list.
624 */
625struct mbuf *
626nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
627	verf_str, mrest, mrest_len, mbp, xidp)
628	register struct ucred *cr;
629	int nmflag;
630	int procid;
631	int auth_type;
632	int auth_len;
633	char *auth_str;
634	int verf_len;
635	char *verf_str;
636	struct mbuf *mrest;
637	int mrest_len;
638	struct mbuf **mbp;
639	u_int32_t *xidp;
640{
641	register struct mbuf *mb;
642	register u_int32_t *tl;
643	register caddr_t bpos;
644	register int i;
645	struct mbuf *mreq, *mb2;
646	int siz, grpsiz, authsiz;
647
648	authsiz = nfsm_rndup(auth_len);
649	MGETHDR(mb, M_TRYWAIT, MT_DATA);
650	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
651		MCLGET(mb, M_TRYWAIT);
652	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
653		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
654	} else {
655		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
656	}
657	mb->m_len = 0;
658	mreq = mb;
659	bpos = mtod(mb, caddr_t);
660
661	/*
662	 * First the RPC header.
663	 */
664	nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
665
666	/* Get a pretty random xid to start with */
667	if (!nfs_xid)
668		nfs_xid = random();
669	/*
670	 * Skip zero xid if it should ever happen.
671	 */
672	if (++nfs_xid == 0)
673		nfs_xid++;
674
675	*tl++ = *xidp = txdr_unsigned(nfs_xid);
676	*tl++ = rpc_call;
677	*tl++ = rpc_vers;
678	if (nmflag & NFSMNT_NQNFS) {
679		*tl++ = txdr_unsigned(NQNFS_PROG);
680		*tl++ = txdr_unsigned(NQNFS_VER3);
681	} else {
682		*tl++ = txdr_unsigned(NFS_PROG);
683		if (nmflag & NFSMNT_NFSV3)
684			*tl++ = txdr_unsigned(NFS_VER3);
685		else
686			*tl++ = txdr_unsigned(NFS_VER2);
687	}
688	if (nmflag & NFSMNT_NFSV3)
689		*tl++ = txdr_unsigned(procid);
690	else
691		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
692
693	/*
694	 * And then the authorization cred.
695	 */
696	*tl++ = txdr_unsigned(auth_type);
697	*tl = txdr_unsigned(authsiz);
698	switch (auth_type) {
699	case RPCAUTH_UNIX:
700		nfsm_build(tl, u_int32_t *, auth_len);
701		*tl++ = 0;		/* stamp ?? */
702		*tl++ = 0;		/* NULL hostname */
703		*tl++ = txdr_unsigned(cr->cr_uid);
704		*tl++ = txdr_unsigned(cr->cr_groups[0]);
705		grpsiz = (auth_len >> 2) - 5;
706		*tl++ = txdr_unsigned(grpsiz);
707		for (i = 1; i <= grpsiz; i++)
708			*tl++ = txdr_unsigned(cr->cr_groups[i]);
709		break;
710	case RPCAUTH_KERB4:
711		siz = auth_len;
712		while (siz > 0) {
713			if (M_TRAILINGSPACE(mb) == 0) {
714				MGET(mb2, M_TRYWAIT, MT_DATA);
715				if (siz >= MINCLSIZE)
716					MCLGET(mb2, M_TRYWAIT);
717				mb->m_next = mb2;
718				mb = mb2;
719				mb->m_len = 0;
720				bpos = mtod(mb, caddr_t);
721			}
722			i = min(siz, M_TRAILINGSPACE(mb));
723			bcopy(auth_str, bpos, i);
724			mb->m_len += i;
725			auth_str += i;
726			bpos += i;
727			siz -= i;
728		}
729		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
730			for (i = 0; i < siz; i++)
731				*bpos++ = '\0';
732			mb->m_len += siz;
733		}
734		break;
735	};
736
737	/*
738	 * And the verifier...
739	 */
740	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
741	if (verf_str) {
742		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
743		*tl = txdr_unsigned(verf_len);
744		siz = verf_len;
745		while (siz > 0) {
746			if (M_TRAILINGSPACE(mb) == 0) {
747				MGET(mb2, M_TRYWAIT, MT_DATA);
748				if (siz >= MINCLSIZE)
749					MCLGET(mb2, M_TRYWAIT);
750				mb->m_next = mb2;
751				mb = mb2;
752				mb->m_len = 0;
753				bpos = mtod(mb, caddr_t);
754			}
755			i = min(siz, M_TRAILINGSPACE(mb));
756			bcopy(verf_str, bpos, i);
757			mb->m_len += i;
758			verf_str += i;
759			bpos += i;
760			siz -= i;
761		}
762		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
763			for (i = 0; i < siz; i++)
764				*bpos++ = '\0';
765			mb->m_len += siz;
766		}
767	} else {
768		*tl++ = txdr_unsigned(RPCAUTH_NULL);
769		*tl = 0;
770	}
771	mb->m_next = mrest;
772	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
773	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
774	*mbp = mb;
775	return (mreq);
776}
777
778/*
779 * copies mbuf chain to the uio scatter/gather list
780 */
781int
782nfsm_mbuftouio(mrep, uiop, siz, dpos)
783	struct mbuf **mrep;
784	register struct uio *uiop;
785	int siz;
786	caddr_t *dpos;
787{
788	register char *mbufcp, *uiocp;
789	register int xfer, left, len;
790	register struct mbuf *mp;
791	long uiosiz, rem;
792	int error = 0;
793
794	mp = *mrep;
795	mbufcp = *dpos;
796	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
797	rem = nfsm_rndup(siz)-siz;
798	while (siz > 0) {
799		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
800			return (EFBIG);
801		left = uiop->uio_iov->iov_len;
802		uiocp = uiop->uio_iov->iov_base;
803		if (left > siz)
804			left = siz;
805		uiosiz = left;
806		while (left > 0) {
807			while (len == 0) {
808				mp = mp->m_next;
809				if (mp == NULL)
810					return (EBADRPC);
811				mbufcp = mtod(mp, caddr_t);
812				len = mp->m_len;
813			}
814			xfer = (left > len) ? len : left;
815#ifdef notdef
816			/* Not Yet.. */
817			if (uiop->uio_iov->iov_op != NULL)
818				(*(uiop->uio_iov->iov_op))
819				(mbufcp, uiocp, xfer);
820			else
821#endif
822			if (uiop->uio_segflg == UIO_SYSSPACE)
823				bcopy(mbufcp, uiocp, xfer);
824			else
825				copyout(mbufcp, uiocp, xfer);
826			left -= xfer;
827			len -= xfer;
828			mbufcp += xfer;
829			uiocp += xfer;
830			uiop->uio_offset += xfer;
831			uiop->uio_resid -= xfer;
832		}
833		if (uiop->uio_iov->iov_len <= siz) {
834			uiop->uio_iovcnt--;
835			uiop->uio_iov++;
836		} else {
837			uiop->uio_iov->iov_base += uiosiz;
838			uiop->uio_iov->iov_len -= uiosiz;
839		}
840		siz -= uiosiz;
841	}
842	*dpos = mbufcp;
843	*mrep = mp;
844	if (rem > 0) {
845		if (len < rem)
846			error = nfs_adv(mrep, dpos, rem, len);
847		else
848			*dpos += rem;
849	}
850	return (error);
851}
852
853/*
854 * copies a uio scatter/gather list to an mbuf chain.
855 * NOTE: can ony handle iovcnt == 1
856 */
857int
858nfsm_uiotombuf(uiop, mq, siz, bpos)
859	register struct uio *uiop;
860	struct mbuf **mq;
861	int siz;
862	caddr_t *bpos;
863{
864	register char *uiocp;
865	register struct mbuf *mp, *mp2;
866	register int xfer, left, mlen;
867	int uiosiz, clflg, rem;
868	char *cp;
869
870#ifdef DIAGNOSTIC
871	if (uiop->uio_iovcnt != 1)
872		panic("nfsm_uiotombuf: iovcnt != 1");
873#endif
874
875	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
876		clflg = 1;
877	else
878		clflg = 0;
879	rem = nfsm_rndup(siz)-siz;
880	mp = mp2 = *mq;
881	while (siz > 0) {
882		left = uiop->uio_iov->iov_len;
883		uiocp = uiop->uio_iov->iov_base;
884		if (left > siz)
885			left = siz;
886		uiosiz = left;
887		while (left > 0) {
888			mlen = M_TRAILINGSPACE(mp);
889			if (mlen == 0) {
890				MGET(mp, M_TRYWAIT, MT_DATA);
891				if (clflg)
892					MCLGET(mp, M_TRYWAIT);
893				mp->m_len = 0;
894				mp2->m_next = mp;
895				mp2 = mp;
896				mlen = M_TRAILINGSPACE(mp);
897			}
898			xfer = (left > mlen) ? mlen : left;
899#ifdef notdef
900			/* Not Yet.. */
901			if (uiop->uio_iov->iov_op != NULL)
902				(*(uiop->uio_iov->iov_op))
903				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
904			else
905#endif
906			if (uiop->uio_segflg == UIO_SYSSPACE)
907				bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
908			else
909				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
910			mp->m_len += xfer;
911			left -= xfer;
912			uiocp += xfer;
913			uiop->uio_offset += xfer;
914			uiop->uio_resid -= xfer;
915		}
916		uiop->uio_iov->iov_base += uiosiz;
917		uiop->uio_iov->iov_len -= uiosiz;
918		siz -= uiosiz;
919	}
920	if (rem > 0) {
921		if (rem > M_TRAILINGSPACE(mp)) {
922			MGET(mp, M_TRYWAIT, MT_DATA);
923			mp->m_len = 0;
924			mp2->m_next = mp;
925		}
926		cp = mtod(mp, caddr_t)+mp->m_len;
927		for (left = 0; left < rem; left++)
928			*cp++ = '\0';
929		mp->m_len += rem;
930		*bpos = cp;
931	} else
932		*bpos = mtod(mp, caddr_t)+mp->m_len;
933	*mq = mp;
934	return (0);
935}
936
937/*
938 * Help break down an mbuf chain by setting the first siz bytes contiguous
939 * pointed to by returned val.
940 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
941 * cases. (The macros use the vars. dpos and dpos2)
942 */
943int
944nfsm_disct(mdp, dposp, siz, left, cp2)
945	struct mbuf **mdp;
946	caddr_t *dposp;
947	int siz;
948	int left;
949	caddr_t *cp2;
950{
951	register struct mbuf *mp, *mp2;
952	register int siz2, xfer;
953	register caddr_t p;
954
955	mp = *mdp;
956	while (left == 0) {
957		*mdp = mp = mp->m_next;
958		if (mp == NULL)
959			return (EBADRPC);
960		left = mp->m_len;
961		*dposp = mtod(mp, caddr_t);
962	}
963	if (left >= siz) {
964		*cp2 = *dposp;
965		*dposp += siz;
966	} else if (mp->m_next == NULL) {
967		return (EBADRPC);
968	} else if (siz > MHLEN) {
969		panic("nfs S too big");
970	} else {
971		MGET(mp2, M_TRYWAIT, MT_DATA);
972		mp2->m_next = mp->m_next;
973		mp->m_next = mp2;
974		mp->m_len -= left;
975		mp = mp2;
976		*cp2 = p = mtod(mp, caddr_t);
977		bcopy(*dposp, p, left);		/* Copy what was left */
978		siz2 = siz-left;
979		p += left;
980		mp2 = mp->m_next;
981		/* Loop around copying up the siz2 bytes */
982		while (siz2 > 0) {
983			if (mp2 == NULL)
984				return (EBADRPC);
985			xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
986			if (xfer > 0) {
987				bcopy(mtod(mp2, caddr_t), p, xfer);
988				NFSMADV(mp2, xfer);
989				mp2->m_len -= xfer;
990				p += xfer;
991				siz2 -= xfer;
992			}
993			if (siz2 > 0)
994				mp2 = mp2->m_next;
995		}
996		mp->m_len = siz;
997		*mdp = mp2;
998		*dposp = mtod(mp2, caddr_t);
999	}
1000	return (0);
1001}
1002
1003/*
1004 * Advance the position in the mbuf chain.
1005 */
1006int
1007nfs_adv(mdp, dposp, offs, left)
1008	struct mbuf **mdp;
1009	caddr_t *dposp;
1010	int offs;
1011	int left;
1012{
1013	register struct mbuf *m;
1014	register int s;
1015
1016	m = *mdp;
1017	s = left;
1018	while (s < offs) {
1019		offs -= s;
1020		m = m->m_next;
1021		if (m == NULL)
1022			return (EBADRPC);
1023		s = m->m_len;
1024	}
1025	*mdp = m;
1026	*dposp = mtod(m, caddr_t)+offs;
1027	return (0);
1028}
1029
1030/*
1031 * Copy a string into mbufs for the hard cases...
1032 */
1033int
1034nfsm_strtmbuf(mb, bpos, cp, siz)
1035	struct mbuf **mb;
1036	char **bpos;
1037	const char *cp;
1038	long siz;
1039{
1040	register struct mbuf *m1 = NULL, *m2;
1041	long left, xfer, len, tlen;
1042	u_int32_t *tl;
1043	int putsize;
1044
1045	putsize = 1;
1046	m2 = *mb;
1047	left = M_TRAILINGSPACE(m2);
1048	if (left > 0) {
1049		tl = ((u_int32_t *)(*bpos));
1050		*tl++ = txdr_unsigned(siz);
1051		putsize = 0;
1052		left -= NFSX_UNSIGNED;
1053		m2->m_len += NFSX_UNSIGNED;
1054		if (left > 0) {
1055			bcopy(cp, (caddr_t) tl, left);
1056			siz -= left;
1057			cp += left;
1058			m2->m_len += left;
1059			left = 0;
1060		}
1061	}
1062	/* Loop around adding mbufs */
1063	while (siz > 0) {
1064		MGET(m1, M_TRYWAIT, MT_DATA);
1065		if (siz > MLEN)
1066			MCLGET(m1, M_TRYWAIT);
1067		m1->m_len = NFSMSIZ(m1);
1068		m2->m_next = m1;
1069		m2 = m1;
1070		tl = mtod(m1, u_int32_t *);
1071		tlen = 0;
1072		if (putsize) {
1073			*tl++ = txdr_unsigned(siz);
1074			m1->m_len -= NFSX_UNSIGNED;
1075			tlen = NFSX_UNSIGNED;
1076			putsize = 0;
1077		}
1078		if (siz < m1->m_len) {
1079			len = nfsm_rndup(siz);
1080			xfer = siz;
1081			if (xfer < len)
1082				*(tl+(xfer>>2)) = 0;
1083		} else {
1084			xfer = len = m1->m_len;
1085		}
1086		bcopy(cp, (caddr_t) tl, xfer);
1087		m1->m_len = len+tlen;
1088		siz -= xfer;
1089		cp += xfer;
1090	}
1091	*mb = m1;
1092	*bpos = mtod(m1, caddr_t)+m1->m_len;
1093	return (0);
1094}
1095
1096/*
1097 * Called once to initialize data structures...
1098 */
1099int
1100nfs_init(vfsp)
1101	struct vfsconf *vfsp;
1102{
1103	register int i;
1104
1105	nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
1106
1107	nfs_mount_type = vfsp->vfc_typenum;
1108	nfsrtt.pos = 0;
1109	rpc_vers = txdr_unsigned(RPC_VER2);
1110	rpc_call = txdr_unsigned(RPC_CALL);
1111	rpc_reply = txdr_unsigned(RPC_REPLY);
1112	rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
1113	rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
1114	rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
1115	rpc_autherr = txdr_unsigned(RPC_AUTHERR);
1116	rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
1117	rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
1118	nfs_prog = txdr_unsigned(NFS_PROG);
1119	nqnfs_prog = txdr_unsigned(NQNFS_PROG);
1120	nfs_true = txdr_unsigned(TRUE);
1121	nfs_false = txdr_unsigned(FALSE);
1122	nfs_xdrneg1 = txdr_unsigned(-1);
1123	nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
1124	if (nfs_ticks < 1)
1125		nfs_ticks = 1;
1126	/* Ensure async daemons disabled */
1127	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1128		nfs_iodwant[i] = (struct proc *)0;
1129		nfs_iodmount[i] = (struct nfsmount *)0;
1130	}
1131	nfs_nhinit();			/* Init the nfsnode table */
1132#ifndef NFS_NOSERVER
1133	nfsrv_init(0);			/* Init server data structures */
1134	nfsrv_initcache();		/* Init the server request cache */
1135#endif
1136
1137	/*
1138	 * Initialize the nqnfs server stuff.
1139	 */
1140	if (nqnfsstarttime == 0) {
1141		nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
1142			+ nqsrv_clockskew + nqsrv_writeslack;
1143		NQLOADNOVRAM(nqnfsstarttime);
1144		TAILQ_INIT(&nqtimerhead);
1145		nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash);
1146	}
1147
1148	/*
1149	 * Initialize reply list and start timer
1150	 */
1151	TAILQ_INIT(&nfs_reqq);
1152
1153	nfs_timer(0);
1154
1155	/*
1156	 * Set up lease_check and lease_updatetime so that other parts
1157	 * of the system can call us, if we are loadable.
1158	 */
1159#ifndef NFS_NOSERVER
1160	nfs_prev_vop_lease_check = default_vnodeop_p[VOFFSET(vop_lease)];
1161	default_vnodeop_p[VOFFSET(vop_lease)] = (vop_t *)nqnfs_vop_lease_check;
1162#endif
1163	nfs_prev_lease_updatetime = lease_updatetime;
1164	lease_updatetime = nfs_lease_updatetime;
1165	nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
1166	sysent[SYS_nfssvc].sy_narg = 2;
1167	nfs_prev_nfssvc_sy_call = sysent[SYS_nfssvc].sy_call;
1168	sysent[SYS_nfssvc].sy_call = (sy_call_t *)nfssvc;
1169
1170	nfs_pbuf_freecnt = nswbuf / 2 + 1;
1171
1172	return (0);
1173}
1174
1175int
1176nfs_uninit(vfsp)
1177	struct vfsconf *vfsp;
1178{
1179
1180	untimeout(nfs_timer, (void *)NULL, nfs_timer_handle);
1181	nfs_mount_type = -1;
1182#ifndef NFS_NOSERVER
1183	default_vnodeop_p[VOFFSET(vop_lease)] = nfs_prev_vop_lease_check;
1184#endif
1185	lease_updatetime = nfs_prev_lease_updatetime;
1186	sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
1187	sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
1188	return (0);
1189}
1190
1191/*
1192 * Attribute cache routines.
1193 * nfs_loadattrcache() - loads or updates the cache contents from attributes
1194 *	that are on the mbuf list
1195 * nfs_getattrcache() - returns valid attributes if found in cache, returns
1196 *	error otherwise
1197 */
1198
1199/*
1200 * Load the attribute cache (that lives in the nfsnode entry) with
1201 * the values on the mbuf list and
1202 * Iff vap not NULL
1203 *    copy the attributes to *vaper
1204 */
1205int
1206nfs_loadattrcache(vpp, mdp, dposp, vaper, dontshrink)
1207	struct vnode **vpp;
1208	struct mbuf **mdp;
1209	caddr_t *dposp;
1210	struct vattr *vaper;
1211	int dontshrink;
1212{
1213	register struct vnode *vp = *vpp;
1214	register struct vattr *vap;
1215	register struct nfs_fattr *fp;
1216	register struct nfsnode *np;
1217	register int32_t t1;
1218	caddr_t cp2;
1219	int error = 0, rdev;
1220	struct mbuf *md;
1221	enum vtype vtyp;
1222	u_short vmode;
1223	struct timespec mtime;
1224	int v3 = NFS_ISV3(vp);
1225
1226	md = *mdp;
1227	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1228	if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
1229		return (error);
1230	fp = (struct nfs_fattr *)cp2;
1231	if (v3) {
1232		vtyp = nfsv3tov_type(fp->fa_type);
1233		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1234		rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
1235			fxdr_unsigned(int, fp->fa3_rdev.specdata2));
1236		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1237	} else {
1238		vtyp = nfsv2tov_type(fp->fa_type);
1239		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1240		/*
1241		 * XXX
1242		 *
1243		 * The duplicate information returned in fa_type and fa_mode
1244		 * is an ambiguity in the NFS version 2 protocol.
1245		 *
1246		 * VREG should be taken literally as a regular file.  If a
1247		 * server intents to return some type information differently
1248		 * in the upper bits of the mode field (e.g. for sockets, or
1249		 * FIFOs), NFSv2 mandates fa_type to be VNON.  Anyway, we
1250		 * leave the examination of the mode bits even in the VREG
1251		 * case to avoid breakage for bogus servers, but we make sure
1252		 * that there are actually type bits set in the upper part of
1253		 * fa_mode (and failing that, trust the va_type field).
1254		 *
1255		 * NFSv3 cleared the issue, and requires fa_mode to not
1256		 * contain any type information (while also introduing sockets
1257		 * and FIFOs for fa_type).
1258		 */
1259		if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
1260			vtyp = IFTOVT(vmode);
1261		rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1262		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1263
1264		/*
1265		 * Really ugly NFSv2 kludge.
1266		 */
1267		if (vtyp == VCHR && rdev == 0xffffffff)
1268			vtyp = VFIFO;
1269	}
1270
1271	/*
1272	 * If v_type == VNON it is a new node, so fill in the v_type,
1273	 * n_mtime fields. Check to see if it represents a special
1274	 * device, and if so, check for a possible alias. Once the
1275	 * correct vnode has been obtained, fill in the rest of the
1276	 * information.
1277	 */
1278	np = VTONFS(vp);
1279	if (vp->v_type != vtyp) {
1280		vp->v_type = vtyp;
1281		if (vp->v_type == VFIFO) {
1282			vp->v_op = fifo_nfsv2nodeop_p;
1283		}
1284		if (vp->v_type == VCHR || vp->v_type == VBLK) {
1285			vp->v_op = spec_nfsv2nodeop_p;
1286			vp = addaliasu(vp, rdev);
1287			np->n_vnode = vp;
1288		}
1289		np->n_mtime = mtime.tv_sec;
1290	}
1291	vap = &np->n_vattr;
1292	vap->va_type = vtyp;
1293	vap->va_mode = (vmode & 07777);
1294	vap->va_rdev = rdev;
1295	vap->va_mtime = mtime;
1296	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1297	if (v3) {
1298		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1299		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1300		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1301		vap->va_size = fxdr_hyper(&fp->fa3_size);
1302		vap->va_blocksize = NFS_FABLKSIZE;
1303		vap->va_bytes = fxdr_hyper(&fp->fa3_used);
1304		vap->va_fileid = fxdr_unsigned(int32_t,
1305		    fp->fa3_fileid.nfsuquad[1]);
1306		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1307		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1308		vap->va_flags = 0;
1309		vap->va_filerev = 0;
1310	} else {
1311		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1312		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1313		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1314		vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1315		vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
1316		vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
1317		    * NFS_FABLKSIZE;
1318		vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1319		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1320		vap->va_flags = 0;
1321		vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1322		    fp->fa2_ctime.nfsv2_sec);
1323		vap->va_ctime.tv_nsec = 0;
1324		vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1325		vap->va_filerev = 0;
1326	}
1327	np->n_attrstamp = time_second;
1328	if (vap->va_size != np->n_size) {
1329		if (vap->va_type == VREG) {
1330			if (dontshrink && vap->va_size < np->n_size) {
1331				/*
1332				 * We've been told not to shrink the file;
1333				 * zero np->n_attrstamp to indicate that
1334				 * the attributes are stale.
1335				 */
1336				vap->va_size = np->n_size;
1337				np->n_attrstamp = 0;
1338			} else if (np->n_flag & NMODIFIED) {
1339				if (vap->va_size < np->n_size)
1340					vap->va_size = np->n_size;
1341				else
1342					np->n_size = vap->va_size;
1343			} else {
1344				np->n_size = vap->va_size;
1345			}
1346			vnode_pager_setsize(vp, np->n_size);
1347		} else {
1348			np->n_size = vap->va_size;
1349		}
1350	}
1351	if (vaper != NULL) {
1352		bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
1353		if (np->n_flag & NCHG) {
1354			if (np->n_flag & NACC)
1355				vaper->va_atime = np->n_atim;
1356			if (np->n_flag & NUPD)
1357				vaper->va_mtime = np->n_mtim;
1358		}
1359	}
1360	return (0);
1361}
1362
1363#ifdef NFS_ACDEBUG
1364#include <sys/sysctl.h>
1365SYSCTL_DECL(_vfs_nfs);
1366static int nfs_acdebug;
1367SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
1368#endif
1369
1370/*
1371 * Check the time stamp
1372 * If the cache is valid, copy contents to *vap and return 0
1373 * otherwise return an error
1374 */
1375int
1376nfs_getattrcache(vp, vaper)
1377	register struct vnode *vp;
1378	struct vattr *vaper;
1379{
1380	register struct nfsnode *np;
1381	register struct vattr *vap;
1382	struct nfsmount *nmp;
1383	int timeo;
1384
1385	np = VTONFS(vp);
1386	vap = &np->n_vattr;
1387	nmp = VFSTONFS(vp->v_mount);
1388	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
1389	timeo = (time_second - np->n_mtime) / 10;
1390
1391#ifdef NFS_ACDEBUG
1392	if (nfs_acdebug>1)
1393		printf("nfs_getattrcache: initial timeo = %d\n", timeo);
1394#endif
1395
1396	if (vap->va_type == VDIR) {
1397		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
1398			timeo = nmp->nm_acdirmin;
1399		else if (timeo > nmp->nm_acdirmax)
1400			timeo = nmp->nm_acdirmax;
1401	} else {
1402		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
1403			timeo = nmp->nm_acregmin;
1404		else if (timeo > nmp->nm_acregmax)
1405			timeo = nmp->nm_acregmax;
1406	}
1407
1408#ifdef NFS_ACDEBUG
1409	if (nfs_acdebug > 2)
1410		printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
1411			nmp->nm_acregmin, nmp->nm_acregmax,
1412			nmp->nm_acdirmin, nmp->nm_acdirmax);
1413
1414	if (nfs_acdebug)
1415		printf("nfs_getattrcache: age = %d; final timeo = %d\n",
1416			(time_second - np->n_attrstamp), timeo);
1417#endif
1418
1419	if ((time_second - np->n_attrstamp) >= timeo) {
1420		nfsstats.attrcache_misses++;
1421		return (ENOENT);
1422	}
1423	nfsstats.attrcache_hits++;
1424	if (vap->va_size != np->n_size) {
1425		if (vap->va_type == VREG) {
1426			if (np->n_flag & NMODIFIED) {
1427				if (vap->va_size < np->n_size)
1428					vap->va_size = np->n_size;
1429				else
1430					np->n_size = vap->va_size;
1431			} else {
1432				np->n_size = vap->va_size;
1433			}
1434			vnode_pager_setsize(vp, np->n_size);
1435		} else {
1436			np->n_size = vap->va_size;
1437		}
1438	}
1439	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
1440	if (np->n_flag & NCHG) {
1441		if (np->n_flag & NACC)
1442			vaper->va_atime = np->n_atim;
1443		if (np->n_flag & NUPD)
1444			vaper->va_mtime = np->n_mtim;
1445	}
1446	return (0);
1447}
1448
1449#ifndef NFS_NOSERVER
1450/*
1451 * Set up nameidata for a lookup() call and do it.
1452 *
1453 * If pubflag is set, this call is done for a lookup operation on the
1454 * public filehandle. In that case we allow crossing mountpoints and
1455 * absolute pathnames. However, the caller is expected to check that
1456 * the lookup result is within the public fs, and deny access if
1457 * it is not.
1458 *
1459 * nfs_namei() clears out garbage fields that namei() might leave garbage.
1460 * This is mainly ni_vp and ni_dvp when an error occurs, and ni_dvp when no
1461 * error occurs but the parent was not requested.
1462 *
1463 * dirp may be set whether an error is returned or not, and must be
1464 * released by the caller.
1465 */
1466int
1467nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
1468	register struct nameidata *ndp;
1469	fhandle_t *fhp;
1470	int len;
1471	struct nfssvc_sock *slp;
1472	struct sockaddr *nam;
1473	struct mbuf **mdp;
1474	caddr_t *dposp;
1475	struct vnode **retdirp;
1476	struct proc *p;
1477	int kerbflag, pubflag;
1478{
1479	register int i, rem;
1480	register struct mbuf *md;
1481	register char *fromcp, *tocp, *cp;
1482	struct iovec aiov;
1483	struct uio auio;
1484	struct vnode *dp;
1485	int error, rdonly, linklen;
1486	struct componentname *cnp = &ndp->ni_cnd;
1487
1488	*retdirp = (struct vnode *)0;
1489	cnp->cn_pnbuf = zalloc(namei_zone);
1490
1491	/*
1492	 * Copy the name from the mbuf list to ndp->ni_pnbuf
1493	 * and set the various ndp fields appropriately.
1494	 */
1495	fromcp = *dposp;
1496	tocp = cnp->cn_pnbuf;
1497	md = *mdp;
1498	rem = mtod(md, caddr_t) + md->m_len - fromcp;
1499	for (i = 0; i < len; i++) {
1500		while (rem == 0) {
1501			md = md->m_next;
1502			if (md == NULL) {
1503				error = EBADRPC;
1504				goto out;
1505			}
1506			fromcp = mtod(md, caddr_t);
1507			rem = md->m_len;
1508		}
1509		if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1510			error = EACCES;
1511			goto out;
1512		}
1513		*tocp++ = *fromcp++;
1514		rem--;
1515	}
1516	*tocp = '\0';
1517	*mdp = md;
1518	*dposp = fromcp;
1519	len = nfsm_rndup(len)-len;
1520	if (len > 0) {
1521		if (rem >= len)
1522			*dposp += len;
1523		else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1524			goto out;
1525	}
1526
1527	/*
1528	 * Extract and set starting directory.
1529	 */
1530	error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
1531	    nam, &rdonly, kerbflag, pubflag);
1532	if (error)
1533		goto out;
1534	if (dp->v_type != VDIR) {
1535		vrele(dp);
1536		error = ENOTDIR;
1537		goto out;
1538	}
1539
1540	if (rdonly)
1541		cnp->cn_flags |= RDONLY;
1542
1543	/*
1544	 * Set return directory.  Reference to dp is implicitly transfered
1545	 * to the returned pointer
1546	 */
1547	*retdirp = dp;
1548
1549	if (pubflag) {
1550		/*
1551		 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1552		 * and the 'native path' indicator.
1553		 */
1554		cp = zalloc(namei_zone);
1555		fromcp = cnp->cn_pnbuf;
1556		tocp = cp;
1557		if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1558			switch ((unsigned char)*fromcp) {
1559			case WEBNFS_NATIVE_CHAR:
1560				/*
1561				 * 'Native' path for us is the same
1562				 * as a path according to the NFS spec,
1563				 * just skip the escape char.
1564				 */
1565				fromcp++;
1566				break;
1567			/*
1568			 * More may be added in the future, range 0x80-0xff
1569			 */
1570			default:
1571				error = EIO;
1572				zfree(namei_zone, cp);
1573				goto out;
1574			}
1575		}
1576		/*
1577		 * Translate the '%' escapes, URL-style.
1578		 */
1579		while (*fromcp != '\0') {
1580			if (*fromcp == WEBNFS_ESC_CHAR) {
1581				if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1582					fromcp++;
1583					*tocp++ = HEXSTRTOI(fromcp);
1584					fromcp += 2;
1585					continue;
1586				} else {
1587					error = ENOENT;
1588					zfree(namei_zone, cp);
1589					goto out;
1590				}
1591			} else
1592				*tocp++ = *fromcp++;
1593		}
1594		*tocp = '\0';
1595		zfree(namei_zone, cnp->cn_pnbuf);
1596		cnp->cn_pnbuf = cp;
1597	}
1598
1599	ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
1600	ndp->ni_segflg = UIO_SYSSPACE;
1601
1602	if (pubflag) {
1603		ndp->ni_rootdir = rootvnode;
1604		ndp->ni_loopcnt = 0;
1605		if (cnp->cn_pnbuf[0] == '/')
1606			dp = rootvnode;
1607	} else {
1608		cnp->cn_flags |= NOCROSSMOUNT;
1609	}
1610
1611	/*
1612	 * Initialize for scan, set ni_startdir and bump ref on dp again
1613	 * becuase lookup() will dereference ni_startdir.
1614	 */
1615
1616	cnp->cn_proc = p;
1617	VREF(dp);
1618	ndp->ni_startdir = dp;
1619
1620	for (;;) {
1621		cnp->cn_nameptr = cnp->cn_pnbuf;
1622		/*
1623		 * Call lookup() to do the real work.  If an error occurs,
1624		 * ndp->ni_vp and ni_dvp are left uninitialized or NULL and
1625		 * we do not have to dereference anything before returning.
1626		 * In either case ni_startdir will be dereferenced and NULLed
1627		 * out.
1628		 */
1629		error = lookup(ndp);
1630		if (error)
1631			break;
1632
1633		/*
1634		 * Check for encountering a symbolic link.  Trivial
1635		 * termination occurs if no symlink encountered.
1636		 * Note: zfree is safe because error is 0, so we will
1637		 * not zfree it again when we break.
1638		 */
1639		if ((cnp->cn_flags & ISSYMLINK) == 0) {
1640			nfsrv_object_create(ndp->ni_vp);
1641			if (cnp->cn_flags & (SAVENAME | SAVESTART))
1642				cnp->cn_flags |= HASBUF;
1643			else
1644				zfree(namei_zone, cnp->cn_pnbuf);
1645			break;
1646		}
1647
1648		/*
1649		 * Validate symlink
1650		 */
1651		if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
1652			VOP_UNLOCK(ndp->ni_dvp, 0, p);
1653		if (!pubflag) {
1654			error = EINVAL;
1655			goto badlink2;
1656		}
1657
1658		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
1659			error = ELOOP;
1660			goto badlink2;
1661		}
1662		if (ndp->ni_pathlen > 1)
1663			cp = zalloc(namei_zone);
1664		else
1665			cp = cnp->cn_pnbuf;
1666		aiov.iov_base = cp;
1667		aiov.iov_len = MAXPATHLEN;
1668		auio.uio_iov = &aiov;
1669		auio.uio_iovcnt = 1;
1670		auio.uio_offset = 0;
1671		auio.uio_rw = UIO_READ;
1672		auio.uio_segflg = UIO_SYSSPACE;
1673		auio.uio_procp = (struct proc *)0;
1674		auio.uio_resid = MAXPATHLEN;
1675		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
1676		if (error) {
1677		badlink1:
1678			if (ndp->ni_pathlen > 1)
1679				zfree(namei_zone, cp);
1680		badlink2:
1681			vrele(ndp->ni_dvp);
1682			vput(ndp->ni_vp);
1683			break;
1684		}
1685		linklen = MAXPATHLEN - auio.uio_resid;
1686		if (linklen == 0) {
1687			error = ENOENT;
1688			goto badlink1;
1689		}
1690		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
1691			error = ENAMETOOLONG;
1692			goto badlink1;
1693		}
1694
1695		/*
1696		 * Adjust or replace path
1697		 */
1698		if (ndp->ni_pathlen > 1) {
1699			bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
1700			zfree(namei_zone, cnp->cn_pnbuf);
1701			cnp->cn_pnbuf = cp;
1702		} else
1703			cnp->cn_pnbuf[linklen] = '\0';
1704		ndp->ni_pathlen += linklen;
1705
1706		/*
1707		 * Cleanup refs for next loop and check if root directory
1708		 * should replace current directory.  Normally ni_dvp
1709		 * becomes the new base directory and is cleaned up when
1710		 * we loop.  Explicitly null pointers after invalidation
1711		 * to clarify operation.
1712		 */
1713		vput(ndp->ni_vp);
1714		ndp->ni_vp = NULL;
1715
1716		if (cnp->cn_pnbuf[0] == '/') {
1717			vrele(ndp->ni_dvp);
1718			ndp->ni_dvp = ndp->ni_rootdir;
1719			VREF(ndp->ni_dvp);
1720		}
1721		ndp->ni_startdir = ndp->ni_dvp;
1722		ndp->ni_dvp = NULL;
1723	}
1724
1725	/*
1726	 * nfs_namei() guarentees that fields will not contain garbage
1727	 * whether an error occurs or not.  This allows the caller to track
1728	 * cleanup state trivially.
1729	 */
1730out:
1731	if (error) {
1732		zfree(namei_zone, cnp->cn_pnbuf);
1733		ndp->ni_vp = NULL;
1734		ndp->ni_dvp = NULL;
1735		ndp->ni_startdir = NULL;
1736		cnp->cn_flags &= ~HASBUF;
1737	} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
1738		ndp->ni_dvp = NULL;
1739	}
1740	return (error);
1741}
1742
1743/*
1744 * A fiddled version of m_adj() that ensures null fill to a long
1745 * boundary and only trims off the back end
1746 */
1747void
1748nfsm_adj(mp, len, nul)
1749	struct mbuf *mp;
1750	register int len;
1751	int nul;
1752{
1753	register struct mbuf *m;
1754	register int count, i;
1755	register char *cp;
1756
1757	/*
1758	 * Trim from tail.  Scan the mbuf chain,
1759	 * calculating its length and finding the last mbuf.
1760	 * If the adjustment only affects this mbuf, then just
1761	 * adjust and return.  Otherwise, rescan and truncate
1762	 * after the remaining size.
1763	 */
1764	count = 0;
1765	m = mp;
1766	for (;;) {
1767		count += m->m_len;
1768		if (m->m_next == (struct mbuf *)0)
1769			break;
1770		m = m->m_next;
1771	}
1772	if (m->m_len > len) {
1773		m->m_len -= len;
1774		if (nul > 0) {
1775			cp = mtod(m, caddr_t)+m->m_len-nul;
1776			for (i = 0; i < nul; i++)
1777				*cp++ = '\0';
1778		}
1779		return;
1780	}
1781	count -= len;
1782	if (count < 0)
1783		count = 0;
1784	/*
1785	 * Correct length for chain is "count".
1786	 * Find the mbuf with last data, adjust its length,
1787	 * and toss data from remaining mbufs on chain.
1788	 */
1789	for (m = mp; m; m = m->m_next) {
1790		if (m->m_len >= count) {
1791			m->m_len = count;
1792			if (nul > 0) {
1793				cp = mtod(m, caddr_t)+m->m_len-nul;
1794				for (i = 0; i < nul; i++)
1795					*cp++ = '\0';
1796			}
1797			break;
1798		}
1799		count -= m->m_len;
1800	}
1801	for (m = m->m_next;m;m = m->m_next)
1802		m->m_len = 0;
1803}
1804
1805/*
1806 * Make these functions instead of macros, so that the kernel text size
1807 * doesn't get too big...
1808 */
1809void
1810nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
1811	struct nfsrv_descript *nfsd;
1812	int before_ret;
1813	register struct vattr *before_vap;
1814	int after_ret;
1815	struct vattr *after_vap;
1816	struct mbuf **mbp;
1817	char **bposp;
1818{
1819	register struct mbuf *mb = *mbp, *mb2;
1820	register char *bpos = *bposp;
1821	register u_int32_t *tl;
1822
1823	if (before_ret) {
1824		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1825		*tl = nfs_false;
1826	} else {
1827		nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
1828		*tl++ = nfs_true;
1829		txdr_hyper(before_vap->va_size, tl);
1830		tl += 2;
1831		txdr_nfsv3time(&(before_vap->va_mtime), tl);
1832		tl += 2;
1833		txdr_nfsv3time(&(before_vap->va_ctime), tl);
1834	}
1835	*bposp = bpos;
1836	*mbp = mb;
1837	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
1838}
1839
1840void
1841nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
1842	struct nfsrv_descript *nfsd;
1843	int after_ret;
1844	struct vattr *after_vap;
1845	struct mbuf **mbp;
1846	char **bposp;
1847{
1848	register struct mbuf *mb = *mbp, *mb2;
1849	register char *bpos = *bposp;
1850	register u_int32_t *tl;
1851	register struct nfs_fattr *fp;
1852
1853	if (after_ret) {
1854		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1855		*tl = nfs_false;
1856	} else {
1857		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
1858		*tl++ = nfs_true;
1859		fp = (struct nfs_fattr *)tl;
1860		nfsm_srvfattr(nfsd, after_vap, fp);
1861	}
1862	*mbp = mb;
1863	*bposp = bpos;
1864}
1865
1866void
1867nfsm_srvfattr(nfsd, vap, fp)
1868	register struct nfsrv_descript *nfsd;
1869	register struct vattr *vap;
1870	register struct nfs_fattr *fp;
1871{
1872
1873	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
1874	fp->fa_uid = txdr_unsigned(vap->va_uid);
1875	fp->fa_gid = txdr_unsigned(vap->va_gid);
1876	if (nfsd->nd_flag & ND_NFSV3) {
1877		fp->fa_type = vtonfsv3_type(vap->va_type);
1878		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
1879		txdr_hyper(vap->va_size, &fp->fa3_size);
1880		txdr_hyper(vap->va_bytes, &fp->fa3_used);
1881		fp->fa3_rdev.specdata1 = txdr_unsigned(umajor(vap->va_rdev));
1882		fp->fa3_rdev.specdata2 = txdr_unsigned(uminor(vap->va_rdev));
1883		fp->fa3_fsid.nfsuquad[0] = 0;
1884		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
1885		fp->fa3_fileid.nfsuquad[0] = 0;
1886		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
1887		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
1888		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
1889		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
1890	} else {
1891		fp->fa_type = vtonfsv2_type(vap->va_type);
1892		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
1893		fp->fa2_size = txdr_unsigned(vap->va_size);
1894		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
1895		if (vap->va_type == VFIFO)
1896			fp->fa2_rdev = 0xffffffff;
1897		else
1898			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
1899		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
1900		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
1901		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
1902		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
1903		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
1904		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
1905	}
1906}
1907
1908/*
1909 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
1910 * 	- look up fsid in mount list (if not found ret error)
1911 *	- get vp and export rights by calling VFS_FHTOVP()
1912 *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
1913 *	- if not lockflag unlock it with VOP_UNLOCK()
1914 */
1915int
1916nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag)
1917	fhandle_t *fhp;
1918	int lockflag;
1919	struct vnode **vpp;
1920	struct ucred *cred;
1921	struct nfssvc_sock *slp;
1922	struct sockaddr *nam;
1923	int *rdonlyp;
1924	int kerbflag;
1925	int pubflag;
1926{
1927	struct proc *p = curproc; /* XXX */
1928	register struct mount *mp;
1929	register int i;
1930	struct ucred *credanon;
1931	int error, exflags;
1932#ifdef MNT_EXNORESPORT		/* XXX needs mountd and /etc/exports help yet */
1933	struct sockaddr_int *saddr;
1934#endif
1935
1936	*vpp = (struct vnode *)0;
1937
1938	if (nfs_ispublicfh(fhp)) {
1939		if (!pubflag || !nfs_pub.np_valid)
1940			return (ESTALE);
1941		fhp = &nfs_pub.np_handle;
1942	}
1943
1944	mp = vfs_getvfs(&fhp->fh_fsid);
1945	if (!mp)
1946		return (ESTALE);
1947	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
1948	if (error)
1949		return (error);
1950	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
1951	if (error)
1952		return (error);
1953#ifdef MNT_EXNORESPORT
1954	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
1955		saddr = (struct sockaddr_in *)nam;
1956		if (saddr->sin_family == AF_INET &&
1957		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
1958			vput(*vpp);
1959			*vpp = NULL;
1960			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1961		}
1962	}
1963#endif
1964	/*
1965	 * Check/setup credentials.
1966	 */
1967	if (exflags & MNT_EXKERB) {
1968		if (!kerbflag) {
1969			vput(*vpp);
1970			*vpp = NULL;
1971			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1972		}
1973	} else if (kerbflag) {
1974		vput(*vpp);
1975		*vpp = NULL;
1976		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1977	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1978		cred->cr_uid = credanon->cr_uid;
1979		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
1980			cred->cr_groups[i] = credanon->cr_groups[i];
1981		cred->cr_ngroups = i;
1982	}
1983	if (exflags & MNT_EXRDONLY)
1984		*rdonlyp = 1;
1985	else
1986		*rdonlyp = 0;
1987
1988	nfsrv_object_create(*vpp);
1989
1990	if (!lockflag)
1991		VOP_UNLOCK(*vpp, 0, p);
1992	return (0);
1993}
1994
1995
1996/*
1997 * WebNFS: check if a filehandle is a public filehandle. For v3, this
1998 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
1999 * transformed this to all zeroes in both cases, so check for it.
2000 */
2001int
2002nfs_ispublicfh(fhp)
2003	fhandle_t *fhp;
2004{
2005	char *cp = (char *)fhp;
2006	int i;
2007
2008	for (i = 0; i < NFSX_V3FH; i++)
2009		if (*cp++ != 0)
2010			return (FALSE);
2011	return (TRUE);
2012}
2013
2014#endif /* NFS_NOSERVER */
2015/*
2016 * This function compares two net addresses by family and returns TRUE
2017 * if they are the same host.
2018 * If there is any doubt, return FALSE.
2019 * The AF_INET family is handled as a special case so that address mbufs
2020 * don't need to be saved to store "struct in_addr", which is only 4 bytes.
2021 */
2022int
2023netaddr_match(family, haddr, nam)
2024	int family;
2025	union nethostaddr *haddr;
2026	struct sockaddr *nam;
2027{
2028	register struct sockaddr_in *inetaddr;
2029
2030	switch (family) {
2031	case AF_INET:
2032		inetaddr = (struct sockaddr_in *)nam;
2033		if (inetaddr->sin_family == AF_INET &&
2034		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
2035			return (1);
2036		break;
2037	default:
2038		break;
2039	};
2040	return (0);
2041}
2042
2043static nfsuint64 nfs_nullcookie = { { 0, 0 } };
2044/*
2045 * This function finds the directory cookie that corresponds to the
2046 * logical byte offset given.
2047 */
2048nfsuint64 *
2049nfs_getcookie(np, off, add)
2050	register struct nfsnode *np;
2051	off_t off;
2052	int add;
2053{
2054	register struct nfsdmap *dp, *dp2;
2055	register int pos;
2056
2057	pos = (uoff_t)off / NFS_DIRBLKSIZ;
2058	if (pos == 0 || off < 0) {
2059#ifdef DIAGNOSTIC
2060		if (add)
2061			panic("nfs getcookie add at <= 0");
2062#endif
2063		return (&nfs_nullcookie);
2064	}
2065	pos--;
2066	dp = np->n_cookies.lh_first;
2067	if (!dp) {
2068		if (add) {
2069			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
2070				M_NFSDIROFF, M_WAITOK);
2071			dp->ndm_eocookie = 0;
2072			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
2073		} else
2074			return ((nfsuint64 *)0);
2075	}
2076	while (pos >= NFSNUMCOOKIES) {
2077		pos -= NFSNUMCOOKIES;
2078		if (dp->ndm_list.le_next) {
2079			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
2080				pos >= dp->ndm_eocookie)
2081				return ((nfsuint64 *)0);
2082			dp = dp->ndm_list.le_next;
2083		} else if (add) {
2084			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
2085				M_NFSDIROFF, M_WAITOK);
2086			dp2->ndm_eocookie = 0;
2087			LIST_INSERT_AFTER(dp, dp2, ndm_list);
2088			dp = dp2;
2089		} else
2090			return ((nfsuint64 *)0);
2091	}
2092	if (pos >= dp->ndm_eocookie) {
2093		if (add)
2094			dp->ndm_eocookie = pos + 1;
2095		else
2096			return ((nfsuint64 *)0);
2097	}
2098	return (&dp->ndm_cookies[pos]);
2099}
2100
2101/*
2102 * Invalidate cached directory information, except for the actual directory
2103 * blocks (which are invalidated separately).
2104 * Done mainly to avoid the use of stale offset cookies.
2105 */
2106void
2107nfs_invaldir(vp)
2108	register struct vnode *vp;
2109{
2110	register struct nfsnode *np = VTONFS(vp);
2111
2112#ifdef DIAGNOSTIC
2113	if (vp->v_type != VDIR)
2114		panic("nfs: invaldir not dir");
2115#endif
2116	np->n_direofoffset = 0;
2117	np->n_cookieverf.nfsuquad[0] = 0;
2118	np->n_cookieverf.nfsuquad[1] = 0;
2119	if (np->n_cookies.lh_first)
2120		np->n_cookies.lh_first->ndm_eocookie = 0;
2121}
2122
2123/*
2124 * The write verifier has changed (probably due to a server reboot), so all
2125 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
2126 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2127 * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
2128 * mount point.
2129 *
2130 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
2131 * writes are not clusterable.
2132 */
2133void
2134nfs_clearcommit(mp)
2135	struct mount *mp;
2136{
2137	register struct vnode *vp, *nvp;
2138	register struct buf *bp, *nbp;
2139	int s;
2140
2141	s = splbio();
2142loop:
2143	for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) {
2144		if (vp->v_mount != mp)	/* Paranoia */
2145			goto loop;
2146		nvp = vp->v_mntvnodes.le_next;
2147		for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
2148			nbp = TAILQ_NEXT(bp, b_vnbufs);
2149			if (BUF_REFCNT(bp) == 0 &&
2150			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
2151				== (B_DELWRI | B_NEEDCOMMIT))
2152				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
2153		}
2154	}
2155	splx(s);
2156}
2157
2158#ifndef NFS_NOSERVER
2159/*
2160 * Map errnos to NFS error numbers. For Version 3 also filter out error
2161 * numbers not specified for the associated procedure.
2162 */
2163int
2164nfsrv_errmap(nd, err)
2165	struct nfsrv_descript *nd;
2166	register int err;
2167{
2168	register short *defaulterrp, *errp;
2169
2170	if (nd->nd_flag & ND_NFSV3) {
2171	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
2172		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2173		while (*++errp) {
2174			if (*errp == err)
2175				return (err);
2176			else if (*errp > err)
2177				break;
2178		}
2179		return ((int)*defaulterrp);
2180	    } else
2181		return (err & 0xffff);
2182	}
2183	if (err <= ELAST)
2184		return ((int)nfsrv_v2errmap[err - 1]);
2185	return (NFSERR_IO);
2186}
2187
2188int
2189nfsrv_object_create(vp)
2190	struct vnode *vp;
2191{
2192
2193	if (vp == NULL || vp->v_type != VREG)
2194		return (1);
2195	return (vfs_object_create(vp, curproc,
2196				  curproc ? curproc->p_ucred : NULL));
2197}
2198
2199/*
2200 * Sort the group list in increasing numerical order.
2201 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2202 *  that used to be here.)
2203 */
2204void
2205nfsrvw_sort(list, num)
2206        register gid_t *list;
2207        register int num;
2208{
2209	register int i, j;
2210	gid_t v;
2211
2212	/* Insertion sort. */
2213	for (i = 1; i < num; i++) {
2214		v = list[i];
2215		/* find correct slot for value v, moving others up */
2216		for (j = i; --j >= 0 && v < list[j];)
2217			list[j + 1] = list[j];
2218		list[j + 1] = v;
2219	}
2220}
2221
2222/*
2223 * copy credentials making sure that the result can be compared with bcmp().
2224 */
2225void
2226nfsrv_setcred(incred, outcred)
2227	register struct ucred *incred, *outcred;
2228{
2229	register int i;
2230
2231	bzero((caddr_t)outcred, sizeof (struct ucred));
2232	outcred->cr_ref = 1;
2233	outcred->cr_uid = incred->cr_uid;
2234	outcred->cr_ngroups = incred->cr_ngroups;
2235	for (i = 0; i < incred->cr_ngroups; i++)
2236		outcred->cr_groups[i] = incred->cr_groups[i];
2237	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2238}
2239#endif /* NFS_NOSERVER */
2240