nfs_srvsubs.c revision 75858
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 75858 2001-04-23 09:05:15Z 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 <net/radix.h>
52#include <sys/socket.h>
53#include <sys/mount.h>
54#include <sys/vnode.h>
55#include <sys/namei.h>
56#include <sys/mbuf.h>
57#include <sys/socket.h>
58#include <sys/stat.h>
59#include <sys/malloc.h>
60#include <sys/sysent.h>
61#include <sys/syscall.h>
62
63#include <vm/vm.h>
64#include <vm/vm_object.h>
65#include <vm/vm_extern.h>
66#include <vm/vm_zone.h>
67
68#include <nfs/rpcv2.h>
69#include <nfs/nfsproto.h>
70#include <nfs/nfs.h>
71#include <nfs/nfsnode.h>
72#include <nfs/xdr_subs.h>
73#include <nfs/nfsm_subs.h>
74#include <nfs/nfsmount.h>
75#include <nfs/nqnfs.h>
76#include <nfs/nfsrtt.h>
77
78#include <netinet/in.h>
79
80/*
81 * Data items converted to xdr at startup, since they are constant
82 * This is kinda hokey, but may save a little time doing byte swaps
83 */
84u_int32_t nfs_xdrneg1;
85u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
86	rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
87	rpc_auth_kerb;
88u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
89
90/* And other global data */
91static u_int32_t nfs_xid = 0;
92static enum vtype nv2tov_type[8]= {
93	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON,  VNON
94};
95enum vtype nv3tov_type[8]= {
96	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
97};
98
99int nfs_ticks;
100int nfs_pbuf_freecnt = -1;	/* start out unlimited */
101
102struct nfs_reqq nfs_reqq;
103struct nfssvc_sockhead nfssvc_sockhead;
104int nfssvc_sockhead_flag;
105struct nfsd_head nfsd_head;
106int nfsd_head_flag;
107struct nfs_bufq nfs_bufq;
108struct nqtimerhead nqtimerhead;
109struct nqfhhashhead *nqfhhashtbl;
110u_long nqfhhash;
111
112static void (*nfs_prev_lease_updatetime) __P((int));
113static int nfs_prev_nfssvc_sy_narg;
114static sy_call_t *nfs_prev_nfssvc_sy_call;
115
116#ifndef NFS_NOSERVER
117
118static vop_t *nfs_prev_vop_lease_check;
119
120/*
121 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
122 */
123int nfsv3_procid[NFS_NPROCS] = {
124	NFSPROC_NULL,
125	NFSPROC_GETATTR,
126	NFSPROC_SETATTR,
127	NFSPROC_NOOP,
128	NFSPROC_LOOKUP,
129	NFSPROC_READLINK,
130	NFSPROC_READ,
131	NFSPROC_NOOP,
132	NFSPROC_WRITE,
133	NFSPROC_CREATE,
134	NFSPROC_REMOVE,
135	NFSPROC_RENAME,
136	NFSPROC_LINK,
137	NFSPROC_SYMLINK,
138	NFSPROC_MKDIR,
139	NFSPROC_RMDIR,
140	NFSPROC_READDIR,
141	NFSPROC_FSSTAT,
142	NFSPROC_NOOP,
143	NFSPROC_NOOP,
144	NFSPROC_NOOP,
145	NFSPROC_NOOP,
146	NFSPROC_NOOP,
147	NFSPROC_NOOP,
148	NFSPROC_NOOP,
149	NFSPROC_NOOP
150};
151
152#endif /* NFS_NOSERVER */
153/*
154 * and the reverse mapping from generic to Version 2 procedure numbers
155 */
156int nfsv2_procid[NFS_NPROCS] = {
157	NFSV2PROC_NULL,
158	NFSV2PROC_GETATTR,
159	NFSV2PROC_SETATTR,
160	NFSV2PROC_LOOKUP,
161	NFSV2PROC_NOOP,
162	NFSV2PROC_READLINK,
163	NFSV2PROC_READ,
164	NFSV2PROC_WRITE,
165	NFSV2PROC_CREATE,
166	NFSV2PROC_MKDIR,
167	NFSV2PROC_SYMLINK,
168	NFSV2PROC_CREATE,
169	NFSV2PROC_REMOVE,
170	NFSV2PROC_RMDIR,
171	NFSV2PROC_RENAME,
172	NFSV2PROC_LINK,
173	NFSV2PROC_READDIR,
174	NFSV2PROC_NOOP,
175	NFSV2PROC_STATFS,
176	NFSV2PROC_NOOP,
177	NFSV2PROC_NOOP,
178	NFSV2PROC_NOOP,
179	NFSV2PROC_NOOP,
180	NFSV2PROC_NOOP,
181	NFSV2PROC_NOOP,
182	NFSV2PROC_NOOP,
183};
184
185#ifndef NFS_NOSERVER
186/*
187 * Maps errno values to nfs error numbers.
188 * Use NFSERR_IO as the catch all for ones not specifically defined in
189 * RFC 1094.
190 */
191static u_char nfsrv_v2errmap[ELAST] = {
192  NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
193  NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
194  NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
195  NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
196  NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
197  NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
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_IO,	NFSERR_IO,	NFSERR_IO,
203  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
204  NFSERR_IO,	NFSERR_IO,	NFSERR_NAMETOL,	NFSERR_IO,	NFSERR_IO,
205  NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
206  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
207  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
208  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
209  NFSERR_IO /* << Last is 86 */
210};
211
212/*
213 * Maps errno values to nfs error numbers.
214 * Although it is not obvious whether or not NFS clients really care if
215 * a returned error value is in the specified list for the procedure, the
216 * safest thing to do is filter them appropriately. For Version 2, the
217 * X/Open XNFS document is the only specification that defines error values
218 * for each RPC (The RFC simply lists all possible error values for all RPCs),
219 * so I have decided to not do this for Version 2.
220 * The first entry is the default error return and the rest are the valid
221 * errors for that RPC in increasing numeric order.
222 */
223static short nfsv3err_null[] = {
224	0,
225	0,
226};
227
228static short nfsv3err_getattr[] = {
229	NFSERR_IO,
230	NFSERR_IO,
231	NFSERR_STALE,
232	NFSERR_BADHANDLE,
233	NFSERR_SERVERFAULT,
234	0,
235};
236
237static short nfsv3err_setattr[] = {
238	NFSERR_IO,
239	NFSERR_PERM,
240	NFSERR_IO,
241	NFSERR_ACCES,
242	NFSERR_INVAL,
243	NFSERR_NOSPC,
244	NFSERR_ROFS,
245	NFSERR_DQUOT,
246	NFSERR_STALE,
247	NFSERR_BADHANDLE,
248	NFSERR_NOT_SYNC,
249	NFSERR_SERVERFAULT,
250	0,
251};
252
253static short nfsv3err_lookup[] = {
254	NFSERR_IO,
255	NFSERR_NOENT,
256	NFSERR_IO,
257	NFSERR_ACCES,
258	NFSERR_NOTDIR,
259	NFSERR_NAMETOL,
260	NFSERR_STALE,
261	NFSERR_BADHANDLE,
262	NFSERR_SERVERFAULT,
263	0,
264};
265
266static short nfsv3err_access[] = {
267	NFSERR_IO,
268	NFSERR_IO,
269	NFSERR_STALE,
270	NFSERR_BADHANDLE,
271	NFSERR_SERVERFAULT,
272	0,
273};
274
275static short nfsv3err_readlink[] = {
276	NFSERR_IO,
277	NFSERR_IO,
278	NFSERR_ACCES,
279	NFSERR_INVAL,
280	NFSERR_STALE,
281	NFSERR_BADHANDLE,
282	NFSERR_NOTSUPP,
283	NFSERR_SERVERFAULT,
284	0,
285};
286
287static short nfsv3err_read[] = {
288	NFSERR_IO,
289	NFSERR_IO,
290	NFSERR_NXIO,
291	NFSERR_ACCES,
292	NFSERR_INVAL,
293	NFSERR_STALE,
294	NFSERR_BADHANDLE,
295	NFSERR_SERVERFAULT,
296	0,
297};
298
299static short nfsv3err_write[] = {
300	NFSERR_IO,
301	NFSERR_IO,
302	NFSERR_ACCES,
303	NFSERR_INVAL,
304	NFSERR_FBIG,
305	NFSERR_NOSPC,
306	NFSERR_ROFS,
307	NFSERR_DQUOT,
308	NFSERR_STALE,
309	NFSERR_BADHANDLE,
310	NFSERR_SERVERFAULT,
311	0,
312};
313
314static short nfsv3err_create[] = {
315	NFSERR_IO,
316	NFSERR_IO,
317	NFSERR_ACCES,
318	NFSERR_EXIST,
319	NFSERR_NOTDIR,
320	NFSERR_NOSPC,
321	NFSERR_ROFS,
322	NFSERR_NAMETOL,
323	NFSERR_DQUOT,
324	NFSERR_STALE,
325	NFSERR_BADHANDLE,
326	NFSERR_NOTSUPP,
327	NFSERR_SERVERFAULT,
328	0,
329};
330
331static short nfsv3err_mkdir[] = {
332	NFSERR_IO,
333	NFSERR_IO,
334	NFSERR_ACCES,
335	NFSERR_EXIST,
336	NFSERR_NOTDIR,
337	NFSERR_NOSPC,
338	NFSERR_ROFS,
339	NFSERR_NAMETOL,
340	NFSERR_DQUOT,
341	NFSERR_STALE,
342	NFSERR_BADHANDLE,
343	NFSERR_NOTSUPP,
344	NFSERR_SERVERFAULT,
345	0,
346};
347
348static short nfsv3err_symlink[] = {
349	NFSERR_IO,
350	NFSERR_IO,
351	NFSERR_ACCES,
352	NFSERR_EXIST,
353	NFSERR_NOTDIR,
354	NFSERR_NOSPC,
355	NFSERR_ROFS,
356	NFSERR_NAMETOL,
357	NFSERR_DQUOT,
358	NFSERR_STALE,
359	NFSERR_BADHANDLE,
360	NFSERR_NOTSUPP,
361	NFSERR_SERVERFAULT,
362	0,
363};
364
365static short nfsv3err_mknod[] = {
366	NFSERR_IO,
367	NFSERR_IO,
368	NFSERR_ACCES,
369	NFSERR_EXIST,
370	NFSERR_NOTDIR,
371	NFSERR_NOSPC,
372	NFSERR_ROFS,
373	NFSERR_NAMETOL,
374	NFSERR_DQUOT,
375	NFSERR_STALE,
376	NFSERR_BADHANDLE,
377	NFSERR_NOTSUPP,
378	NFSERR_SERVERFAULT,
379	NFSERR_BADTYPE,
380	0,
381};
382
383static short nfsv3err_remove[] = {
384	NFSERR_IO,
385	NFSERR_NOENT,
386	NFSERR_IO,
387	NFSERR_ACCES,
388	NFSERR_NOTDIR,
389	NFSERR_ROFS,
390	NFSERR_NAMETOL,
391	NFSERR_STALE,
392	NFSERR_BADHANDLE,
393	NFSERR_SERVERFAULT,
394	0,
395};
396
397static short nfsv3err_rmdir[] = {
398	NFSERR_IO,
399	NFSERR_NOENT,
400	NFSERR_IO,
401	NFSERR_ACCES,
402	NFSERR_EXIST,
403	NFSERR_NOTDIR,
404	NFSERR_INVAL,
405	NFSERR_ROFS,
406	NFSERR_NAMETOL,
407	NFSERR_NOTEMPTY,
408	NFSERR_STALE,
409	NFSERR_BADHANDLE,
410	NFSERR_NOTSUPP,
411	NFSERR_SERVERFAULT,
412	0,
413};
414
415static short nfsv3err_rename[] = {
416	NFSERR_IO,
417	NFSERR_NOENT,
418	NFSERR_IO,
419	NFSERR_ACCES,
420	NFSERR_EXIST,
421	NFSERR_XDEV,
422	NFSERR_NOTDIR,
423	NFSERR_ISDIR,
424	NFSERR_INVAL,
425	NFSERR_NOSPC,
426	NFSERR_ROFS,
427	NFSERR_MLINK,
428	NFSERR_NAMETOL,
429	NFSERR_NOTEMPTY,
430	NFSERR_DQUOT,
431	NFSERR_STALE,
432	NFSERR_BADHANDLE,
433	NFSERR_NOTSUPP,
434	NFSERR_SERVERFAULT,
435	0,
436};
437
438static short nfsv3err_link[] = {
439	NFSERR_IO,
440	NFSERR_IO,
441	NFSERR_ACCES,
442	NFSERR_EXIST,
443	NFSERR_XDEV,
444	NFSERR_NOTDIR,
445	NFSERR_INVAL,
446	NFSERR_NOSPC,
447	NFSERR_ROFS,
448	NFSERR_MLINK,
449	NFSERR_NAMETOL,
450	NFSERR_DQUOT,
451	NFSERR_STALE,
452	NFSERR_BADHANDLE,
453	NFSERR_NOTSUPP,
454	NFSERR_SERVERFAULT,
455	0,
456};
457
458static short nfsv3err_readdir[] = {
459	NFSERR_IO,
460	NFSERR_IO,
461	NFSERR_ACCES,
462	NFSERR_NOTDIR,
463	NFSERR_STALE,
464	NFSERR_BADHANDLE,
465	NFSERR_BAD_COOKIE,
466	NFSERR_TOOSMALL,
467	NFSERR_SERVERFAULT,
468	0,
469};
470
471static short nfsv3err_readdirplus[] = {
472	NFSERR_IO,
473	NFSERR_IO,
474	NFSERR_ACCES,
475	NFSERR_NOTDIR,
476	NFSERR_STALE,
477	NFSERR_BADHANDLE,
478	NFSERR_BAD_COOKIE,
479	NFSERR_NOTSUPP,
480	NFSERR_TOOSMALL,
481	NFSERR_SERVERFAULT,
482	0,
483};
484
485static short nfsv3err_fsstat[] = {
486	NFSERR_IO,
487	NFSERR_IO,
488	NFSERR_STALE,
489	NFSERR_BADHANDLE,
490	NFSERR_SERVERFAULT,
491	0,
492};
493
494static short nfsv3err_fsinfo[] = {
495	NFSERR_STALE,
496	NFSERR_STALE,
497	NFSERR_BADHANDLE,
498	NFSERR_SERVERFAULT,
499	0,
500};
501
502static short nfsv3err_pathconf[] = {
503	NFSERR_STALE,
504	NFSERR_STALE,
505	NFSERR_BADHANDLE,
506	NFSERR_SERVERFAULT,
507	0,
508};
509
510static short nfsv3err_commit[] = {
511	NFSERR_IO,
512	NFSERR_IO,
513	NFSERR_STALE,
514	NFSERR_BADHANDLE,
515	NFSERR_SERVERFAULT,
516	0,
517};
518
519static short *nfsrv_v3errmap[] = {
520	nfsv3err_null,
521	nfsv3err_getattr,
522	nfsv3err_setattr,
523	nfsv3err_lookup,
524	nfsv3err_access,
525	nfsv3err_readlink,
526	nfsv3err_read,
527	nfsv3err_write,
528	nfsv3err_create,
529	nfsv3err_mkdir,
530	nfsv3err_symlink,
531	nfsv3err_mknod,
532	nfsv3err_remove,
533	nfsv3err_rmdir,
534	nfsv3err_rename,
535	nfsv3err_link,
536	nfsv3err_readdir,
537	nfsv3err_readdirplus,
538	nfsv3err_fsstat,
539	nfsv3err_fsinfo,
540	nfsv3err_pathconf,
541	nfsv3err_commit,
542};
543
544#endif /* NFS_NOSERVER */
545
546extern struct nfsrtt nfsrtt;
547extern time_t nqnfsstarttime;
548extern int nqsrv_clockskew;
549extern int nqsrv_writeslack;
550extern int nqsrv_maxlease;
551extern struct nfsstats nfsstats;
552extern int nqnfs_piggy[NFS_NPROCS];
553extern nfstype nfsv2_type[9];
554extern nfstype nfsv3_type[9];
555extern struct nfsnodehashhead *nfsnodehashtbl;
556extern u_long nfsnodehash;
557
558struct nfssvc_args;
559extern int nfssvc(struct proc *, struct nfssvc_args *, int *);
560
561LIST_HEAD(nfsnodehashhead, nfsnode);
562
563int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
564
565u_quad_t
566nfs_curusec()
567{
568	struct timeval tv;
569
570	getmicrotime(&tv);
571	return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
572}
573
574/*
575 * Create the header for an rpc request packet
576 * The hsiz is the size of the rest of the nfs request header.
577 * (just used to decide if a cluster is a good idea)
578 */
579struct mbuf *
580nfsm_reqh(vp, procid, hsiz, bposp)
581	struct vnode *vp;
582	u_long procid;
583	int hsiz;
584	caddr_t *bposp;
585{
586	register struct mbuf *mb;
587	register u_int32_t *tl;
588	register caddr_t bpos;
589	struct mbuf *mb2;
590	struct nfsmount *nmp;
591	int nqflag;
592
593	MGET(mb, M_TRYWAIT, MT_DATA);
594	if (hsiz >= MINCLSIZE)
595		MCLGET(mb, M_TRYWAIT);
596	mb->m_len = 0;
597	bpos = mtod(mb, caddr_t);
598
599	/*
600	 * For NQNFS, add lease request.
601	 */
602	if (vp) {
603		nmp = VFSTONFS(vp->v_mount);
604		if (nmp->nm_flag & NFSMNT_NQNFS) {
605			nqflag = NQNFS_NEEDLEASE(vp, procid);
606			if (nqflag) {
607				nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
608				*tl++ = txdr_unsigned(nqflag);
609				*tl = txdr_unsigned(nmp->nm_leaseterm);
610			} else {
611				nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
612				*tl = 0;
613			}
614		}
615	}
616	/* Finally, return values */
617	*bposp = bpos;
618	return (mb);
619}
620
621/*
622 * Build the RPC header and fill in the authorization info.
623 * The authorization string argument is only used when the credentials
624 * come from outside of the kernel.
625 * Returns the head of the mbuf list.
626 */
627struct mbuf *
628nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
629	verf_str, mrest, mrest_len, mbp, xidp)
630	register struct ucred *cr;
631	int nmflag;
632	int procid;
633	int auth_type;
634	int auth_len;
635	char *auth_str;
636	int verf_len;
637	char *verf_str;
638	struct mbuf *mrest;
639	int mrest_len;
640	struct mbuf **mbp;
641	u_int32_t *xidp;
642{
643	register struct mbuf *mb;
644	register u_int32_t *tl;
645	register caddr_t bpos;
646	register int i;
647	struct mbuf *mreq, *mb2;
648	int siz, grpsiz, authsiz;
649
650	authsiz = nfsm_rndup(auth_len);
651	MGETHDR(mb, M_TRYWAIT, MT_DATA);
652	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
653		MCLGET(mb, M_TRYWAIT);
654	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
655		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
656	} else {
657		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
658	}
659	mb->m_len = 0;
660	mreq = mb;
661	bpos = mtod(mb, caddr_t);
662
663	/*
664	 * First the RPC header.
665	 */
666	nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
667
668	/* Get a pretty random xid to start with */
669	if (!nfs_xid)
670		nfs_xid = random();
671	/*
672	 * Skip zero xid if it should ever happen.
673	 */
674	if (++nfs_xid == 0)
675		nfs_xid++;
676
677	*tl++ = *xidp = txdr_unsigned(nfs_xid);
678	*tl++ = rpc_call;
679	*tl++ = rpc_vers;
680	if (nmflag & NFSMNT_NQNFS) {
681		*tl++ = txdr_unsigned(NQNFS_PROG);
682		*tl++ = txdr_unsigned(NQNFS_VER3);
683	} else {
684		*tl++ = txdr_unsigned(NFS_PROG);
685		if (nmflag & NFSMNT_NFSV3)
686			*tl++ = txdr_unsigned(NFS_VER3);
687		else
688			*tl++ = txdr_unsigned(NFS_VER2);
689	}
690	if (nmflag & NFSMNT_NFSV3)
691		*tl++ = txdr_unsigned(procid);
692	else
693		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
694
695	/*
696	 * And then the authorization cred.
697	 */
698	*tl++ = txdr_unsigned(auth_type);
699	*tl = txdr_unsigned(authsiz);
700	switch (auth_type) {
701	case RPCAUTH_UNIX:
702		nfsm_build(tl, u_int32_t *, auth_len);
703		*tl++ = 0;		/* stamp ?? */
704		*tl++ = 0;		/* NULL hostname */
705		*tl++ = txdr_unsigned(cr->cr_uid);
706		*tl++ = txdr_unsigned(cr->cr_groups[0]);
707		grpsiz = (auth_len >> 2) - 5;
708		*tl++ = txdr_unsigned(grpsiz);
709		for (i = 1; i <= grpsiz; i++)
710			*tl++ = txdr_unsigned(cr->cr_groups[i]);
711		break;
712	case RPCAUTH_KERB4:
713		siz = auth_len;
714		while (siz > 0) {
715			if (M_TRAILINGSPACE(mb) == 0) {
716				MGET(mb2, M_TRYWAIT, MT_DATA);
717				if (siz >= MINCLSIZE)
718					MCLGET(mb2, M_TRYWAIT);
719				mb->m_next = mb2;
720				mb = mb2;
721				mb->m_len = 0;
722				bpos = mtod(mb, caddr_t);
723			}
724			i = min(siz, M_TRAILINGSPACE(mb));
725			bcopy(auth_str, bpos, i);
726			mb->m_len += i;
727			auth_str += i;
728			bpos += i;
729			siz -= i;
730		}
731		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
732			for (i = 0; i < siz; i++)
733				*bpos++ = '\0';
734			mb->m_len += siz;
735		}
736		break;
737	};
738
739	/*
740	 * And the verifier...
741	 */
742	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
743	if (verf_str) {
744		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
745		*tl = txdr_unsigned(verf_len);
746		siz = verf_len;
747		while (siz > 0) {
748			if (M_TRAILINGSPACE(mb) == 0) {
749				MGET(mb2, M_TRYWAIT, MT_DATA);
750				if (siz >= MINCLSIZE)
751					MCLGET(mb2, M_TRYWAIT);
752				mb->m_next = mb2;
753				mb = mb2;
754				mb->m_len = 0;
755				bpos = mtod(mb, caddr_t);
756			}
757			i = min(siz, M_TRAILINGSPACE(mb));
758			bcopy(verf_str, bpos, i);
759			mb->m_len += i;
760			verf_str += i;
761			bpos += i;
762			siz -= i;
763		}
764		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
765			for (i = 0; i < siz; i++)
766				*bpos++ = '\0';
767			mb->m_len += siz;
768		}
769	} else {
770		*tl++ = txdr_unsigned(RPCAUTH_NULL);
771		*tl = 0;
772	}
773	mb->m_next = mrest;
774	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
775	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
776	*mbp = mb;
777	return (mreq);
778}
779
780/*
781 * copies mbuf chain to the uio scatter/gather list
782 */
783int
784nfsm_mbuftouio(mrep, uiop, siz, dpos)
785	struct mbuf **mrep;
786	register struct uio *uiop;
787	int siz;
788	caddr_t *dpos;
789{
790	register char *mbufcp, *uiocp;
791	register int xfer, left, len;
792	register struct mbuf *mp;
793	long uiosiz, rem;
794	int error = 0;
795
796	mp = *mrep;
797	mbufcp = *dpos;
798	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
799	rem = nfsm_rndup(siz)-siz;
800	while (siz > 0) {
801		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
802			return (EFBIG);
803		left = uiop->uio_iov->iov_len;
804		uiocp = uiop->uio_iov->iov_base;
805		if (left > siz)
806			left = siz;
807		uiosiz = left;
808		while (left > 0) {
809			while (len == 0) {
810				mp = mp->m_next;
811				if (mp == NULL)
812					return (EBADRPC);
813				mbufcp = mtod(mp, caddr_t);
814				len = mp->m_len;
815			}
816			xfer = (left > len) ? len : left;
817#ifdef notdef
818			/* Not Yet.. */
819			if (uiop->uio_iov->iov_op != NULL)
820				(*(uiop->uio_iov->iov_op))
821				(mbufcp, uiocp, xfer);
822			else
823#endif
824			if (uiop->uio_segflg == UIO_SYSSPACE)
825				bcopy(mbufcp, uiocp, xfer);
826			else
827				copyout(mbufcp, uiocp, xfer);
828			left -= xfer;
829			len -= xfer;
830			mbufcp += xfer;
831			uiocp += xfer;
832			uiop->uio_offset += xfer;
833			uiop->uio_resid -= xfer;
834		}
835		if (uiop->uio_iov->iov_len <= siz) {
836			uiop->uio_iovcnt--;
837			uiop->uio_iov++;
838		} else {
839			uiop->uio_iov->iov_base += uiosiz;
840			uiop->uio_iov->iov_len -= uiosiz;
841		}
842		siz -= uiosiz;
843	}
844	*dpos = mbufcp;
845	*mrep = mp;
846	if (rem > 0) {
847		if (len < rem)
848			error = nfs_adv(mrep, dpos, rem, len);
849		else
850			*dpos += rem;
851	}
852	return (error);
853}
854
855/*
856 * copies a uio scatter/gather list to an mbuf chain.
857 * NOTE: can ony handle iovcnt == 1
858 */
859int
860nfsm_uiotombuf(uiop, mq, siz, bpos)
861	register struct uio *uiop;
862	struct mbuf **mq;
863	int siz;
864	caddr_t *bpos;
865{
866	register char *uiocp;
867	register struct mbuf *mp, *mp2;
868	register int xfer, left, mlen;
869	int uiosiz, clflg, rem;
870	char *cp;
871
872#ifdef DIAGNOSTIC
873	if (uiop->uio_iovcnt != 1)
874		panic("nfsm_uiotombuf: iovcnt != 1");
875#endif
876
877	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
878		clflg = 1;
879	else
880		clflg = 0;
881	rem = nfsm_rndup(siz)-siz;
882	mp = mp2 = *mq;
883	while (siz > 0) {
884		left = uiop->uio_iov->iov_len;
885		uiocp = uiop->uio_iov->iov_base;
886		if (left > siz)
887			left = siz;
888		uiosiz = left;
889		while (left > 0) {
890			mlen = M_TRAILINGSPACE(mp);
891			if (mlen == 0) {
892				MGET(mp, M_TRYWAIT, MT_DATA);
893				if (clflg)
894					MCLGET(mp, M_TRYWAIT);
895				mp->m_len = 0;
896				mp2->m_next = mp;
897				mp2 = mp;
898				mlen = M_TRAILINGSPACE(mp);
899			}
900			xfer = (left > mlen) ? mlen : left;
901#ifdef notdef
902			/* Not Yet.. */
903			if (uiop->uio_iov->iov_op != NULL)
904				(*(uiop->uio_iov->iov_op))
905				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
906			else
907#endif
908			if (uiop->uio_segflg == UIO_SYSSPACE)
909				bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
910			else
911				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
912			mp->m_len += xfer;
913			left -= xfer;
914			uiocp += xfer;
915			uiop->uio_offset += xfer;
916			uiop->uio_resid -= xfer;
917		}
918		uiop->uio_iov->iov_base += uiosiz;
919		uiop->uio_iov->iov_len -= uiosiz;
920		siz -= uiosiz;
921	}
922	if (rem > 0) {
923		if (rem > M_TRAILINGSPACE(mp)) {
924			MGET(mp, M_TRYWAIT, MT_DATA);
925			mp->m_len = 0;
926			mp2->m_next = mp;
927		}
928		cp = mtod(mp, caddr_t)+mp->m_len;
929		for (left = 0; left < rem; left++)
930			*cp++ = '\0';
931		mp->m_len += rem;
932		*bpos = cp;
933	} else
934		*bpos = mtod(mp, caddr_t)+mp->m_len;
935	*mq = mp;
936	return (0);
937}
938
939/*
940 * Help break down an mbuf chain by setting the first siz bytes contiguous
941 * pointed to by returned val.
942 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
943 * cases. (The macros use the vars. dpos and dpos2)
944 */
945int
946nfsm_disct(mdp, dposp, siz, left, cp2)
947	struct mbuf **mdp;
948	caddr_t *dposp;
949	int siz;
950	int left;
951	caddr_t *cp2;
952{
953	register struct mbuf *mp, *mp2;
954	register int siz2, xfer;
955	register caddr_t p;
956
957	mp = *mdp;
958	while (left == 0) {
959		*mdp = mp = mp->m_next;
960		if (mp == NULL)
961			return (EBADRPC);
962		left = mp->m_len;
963		*dposp = mtod(mp, caddr_t);
964	}
965	if (left >= siz) {
966		*cp2 = *dposp;
967		*dposp += siz;
968	} else if (mp->m_next == NULL) {
969		return (EBADRPC);
970	} else if (siz > MHLEN) {
971		panic("nfs S too big");
972	} else {
973		MGET(mp2, M_TRYWAIT, MT_DATA);
974		mp2->m_next = mp->m_next;
975		mp->m_next = mp2;
976		mp->m_len -= left;
977		mp = mp2;
978		*cp2 = p = mtod(mp, caddr_t);
979		bcopy(*dposp, p, left);		/* Copy what was left */
980		siz2 = siz-left;
981		p += left;
982		mp2 = mp->m_next;
983		/* Loop around copying up the siz2 bytes */
984		while (siz2 > 0) {
985			if (mp2 == NULL)
986				return (EBADRPC);
987			xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
988			if (xfer > 0) {
989				bcopy(mtod(mp2, caddr_t), p, xfer);
990				NFSMADV(mp2, xfer);
991				mp2->m_len -= xfer;
992				p += xfer;
993				siz2 -= xfer;
994			}
995			if (siz2 > 0)
996				mp2 = mp2->m_next;
997		}
998		mp->m_len = siz;
999		*mdp = mp2;
1000		*dposp = mtod(mp2, caddr_t);
1001	}
1002	return (0);
1003}
1004
1005/*
1006 * Advance the position in the mbuf chain.
1007 */
1008int
1009nfs_adv(mdp, dposp, offs, left)
1010	struct mbuf **mdp;
1011	caddr_t *dposp;
1012	int offs;
1013	int left;
1014{
1015	register struct mbuf *m;
1016	register int s;
1017
1018	m = *mdp;
1019	s = left;
1020	while (s < offs) {
1021		offs -= s;
1022		m = m->m_next;
1023		if (m == NULL)
1024			return (EBADRPC);
1025		s = m->m_len;
1026	}
1027	*mdp = m;
1028	*dposp = mtod(m, caddr_t)+offs;
1029	return (0);
1030}
1031
1032/*
1033 * Copy a string into mbufs for the hard cases...
1034 */
1035int
1036nfsm_strtmbuf(mb, bpos, cp, siz)
1037	struct mbuf **mb;
1038	char **bpos;
1039	const char *cp;
1040	long siz;
1041{
1042	register struct mbuf *m1 = NULL, *m2;
1043	long left, xfer, len, tlen;
1044	u_int32_t *tl;
1045	int putsize;
1046
1047	putsize = 1;
1048	m2 = *mb;
1049	left = M_TRAILINGSPACE(m2);
1050	if (left > 0) {
1051		tl = ((u_int32_t *)(*bpos));
1052		*tl++ = txdr_unsigned(siz);
1053		putsize = 0;
1054		left -= NFSX_UNSIGNED;
1055		m2->m_len += NFSX_UNSIGNED;
1056		if (left > 0) {
1057			bcopy(cp, (caddr_t) tl, left);
1058			siz -= left;
1059			cp += left;
1060			m2->m_len += left;
1061			left = 0;
1062		}
1063	}
1064	/* Loop around adding mbufs */
1065	while (siz > 0) {
1066		MGET(m1, M_TRYWAIT, MT_DATA);
1067		if (siz > MLEN)
1068			MCLGET(m1, M_TRYWAIT);
1069		m1->m_len = NFSMSIZ(m1);
1070		m2->m_next = m1;
1071		m2 = m1;
1072		tl = mtod(m1, u_int32_t *);
1073		tlen = 0;
1074		if (putsize) {
1075			*tl++ = txdr_unsigned(siz);
1076			m1->m_len -= NFSX_UNSIGNED;
1077			tlen = NFSX_UNSIGNED;
1078			putsize = 0;
1079		}
1080		if (siz < m1->m_len) {
1081			len = nfsm_rndup(siz);
1082			xfer = siz;
1083			if (xfer < len)
1084				*(tl+(xfer>>2)) = 0;
1085		} else {
1086			xfer = len = m1->m_len;
1087		}
1088		bcopy(cp, (caddr_t) tl, xfer);
1089		m1->m_len = len+tlen;
1090		siz -= xfer;
1091		cp += xfer;
1092	}
1093	*mb = m1;
1094	*bpos = mtod(m1, caddr_t)+m1->m_len;
1095	return (0);
1096}
1097
1098/*
1099 * Called once to initialize data structures...
1100 */
1101int
1102nfs_init(vfsp)
1103	struct vfsconf *vfsp;
1104{
1105	register int i;
1106
1107	nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
1108
1109	nfs_mount_type = vfsp->vfc_typenum;
1110	nfsrtt.pos = 0;
1111	rpc_vers = txdr_unsigned(RPC_VER2);
1112	rpc_call = txdr_unsigned(RPC_CALL);
1113	rpc_reply = txdr_unsigned(RPC_REPLY);
1114	rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
1115	rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
1116	rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
1117	rpc_autherr = txdr_unsigned(RPC_AUTHERR);
1118	rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
1119	rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
1120	nfs_prog = txdr_unsigned(NFS_PROG);
1121	nqnfs_prog = txdr_unsigned(NQNFS_PROG);
1122	nfs_true = txdr_unsigned(TRUE);
1123	nfs_false = txdr_unsigned(FALSE);
1124	nfs_xdrneg1 = txdr_unsigned(-1);
1125	nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
1126	if (nfs_ticks < 1)
1127		nfs_ticks = 1;
1128	/* Ensure async daemons disabled */
1129	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1130		nfs_iodwant[i] = (struct proc *)0;
1131		nfs_iodmount[i] = (struct nfsmount *)0;
1132	}
1133	nfs_nhinit();			/* Init the nfsnode table */
1134#ifndef NFS_NOSERVER
1135	nfsrv_init(0);			/* Init server data structures */
1136	nfsrv_initcache();		/* Init the server request cache */
1137#endif
1138
1139	/*
1140	 * Initialize the nqnfs server stuff.
1141	 */
1142	if (nqnfsstarttime == 0) {
1143		nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
1144			+ nqsrv_clockskew + nqsrv_writeslack;
1145		NQLOADNOVRAM(nqnfsstarttime);
1146		TAILQ_INIT(&nqtimerhead);
1147		nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash);
1148	}
1149
1150	/*
1151	 * Initialize reply list and start timer
1152	 */
1153	TAILQ_INIT(&nfs_reqq);
1154
1155	nfs_timer(0);
1156
1157	/*
1158	 * Set up lease_check and lease_updatetime so that other parts
1159	 * of the system can call us, if we are loadable.
1160	 */
1161#ifndef NFS_NOSERVER
1162	nfs_prev_vop_lease_check = default_vnodeop_p[VOFFSET(vop_lease)];
1163	default_vnodeop_p[VOFFSET(vop_lease)] = (vop_t *)nqnfs_vop_lease_check;
1164#endif
1165	nfs_prev_lease_updatetime = lease_updatetime;
1166	lease_updatetime = nfs_lease_updatetime;
1167	nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
1168	sysent[SYS_nfssvc].sy_narg = 2;
1169	nfs_prev_nfssvc_sy_call = sysent[SYS_nfssvc].sy_call;
1170	sysent[SYS_nfssvc].sy_call = (sy_call_t *)nfssvc;
1171
1172	nfs_pbuf_freecnt = nswbuf / 2 + 1;
1173
1174	return (0);
1175}
1176
1177int
1178nfs_uninit(vfsp)
1179	struct vfsconf *vfsp;
1180{
1181
1182	untimeout(nfs_timer, (void *)NULL, nfs_timer_handle);
1183	nfs_mount_type = -1;
1184#ifndef NFS_NOSERVER
1185	default_vnodeop_p[VOFFSET(vop_lease)] = nfs_prev_vop_lease_check;
1186#endif
1187	lease_updatetime = nfs_prev_lease_updatetime;
1188	sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
1189	sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
1190	return (0);
1191}
1192
1193/*
1194 * Attribute cache routines.
1195 * nfs_loadattrcache() - loads or updates the cache contents from attributes
1196 *	that are on the mbuf list
1197 * nfs_getattrcache() - returns valid attributes if found in cache, returns
1198 *	error otherwise
1199 */
1200
1201/*
1202 * Load the attribute cache (that lives in the nfsnode entry) with
1203 * the values on the mbuf list and
1204 * Iff vap not NULL
1205 *    copy the attributes to *vaper
1206 */
1207int
1208nfs_loadattrcache(vpp, mdp, dposp, vaper, dontshrink)
1209	struct vnode **vpp;
1210	struct mbuf **mdp;
1211	caddr_t *dposp;
1212	struct vattr *vaper;
1213	int dontshrink;
1214{
1215	register struct vnode *vp = *vpp;
1216	register struct vattr *vap;
1217	register struct nfs_fattr *fp;
1218	register struct nfsnode *np;
1219	register int32_t t1;
1220	caddr_t cp2;
1221	int error = 0, rdev;
1222	struct mbuf *md;
1223	enum vtype vtyp;
1224	u_short vmode;
1225	struct timespec mtime;
1226	int v3 = NFS_ISV3(vp);
1227
1228	md = *mdp;
1229	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1230	if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
1231		return (error);
1232	fp = (struct nfs_fattr *)cp2;
1233	if (v3) {
1234		vtyp = nfsv3tov_type(fp->fa_type);
1235		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1236		rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
1237			fxdr_unsigned(int, fp->fa3_rdev.specdata2));
1238		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1239	} else {
1240		vtyp = nfsv2tov_type(fp->fa_type);
1241		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1242		/*
1243		 * XXX
1244		 *
1245		 * The duplicate information returned in fa_type and fa_mode
1246		 * is an ambiguity in the NFS version 2 protocol.
1247		 *
1248		 * VREG should be taken literally as a regular file.  If a
1249		 * server intents to return some type information differently
1250		 * in the upper bits of the mode field (e.g. for sockets, or
1251		 * FIFOs), NFSv2 mandates fa_type to be VNON.  Anyway, we
1252		 * leave the examination of the mode bits even in the VREG
1253		 * case to avoid breakage for bogus servers, but we make sure
1254		 * that there are actually type bits set in the upper part of
1255		 * fa_mode (and failing that, trust the va_type field).
1256		 *
1257		 * NFSv3 cleared the issue, and requires fa_mode to not
1258		 * contain any type information (while also introduing sockets
1259		 * and FIFOs for fa_type).
1260		 */
1261		if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
1262			vtyp = IFTOVT(vmode);
1263		rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1264		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1265
1266		/*
1267		 * Really ugly NFSv2 kludge.
1268		 */
1269		if (vtyp == VCHR && rdev == 0xffffffff)
1270			vtyp = VFIFO;
1271	}
1272
1273	/*
1274	 * If v_type == VNON it is a new node, so fill in the v_type,
1275	 * n_mtime fields. Check to see if it represents a special
1276	 * device, and if so, check for a possible alias. Once the
1277	 * correct vnode has been obtained, fill in the rest of the
1278	 * information.
1279	 */
1280	np = VTONFS(vp);
1281	if (vp->v_type != vtyp) {
1282		vp->v_type = vtyp;
1283		if (vp->v_type == VFIFO) {
1284			vp->v_op = fifo_nfsv2nodeop_p;
1285		}
1286		if (vp->v_type == VCHR || vp->v_type == VBLK) {
1287			vp->v_op = spec_nfsv2nodeop_p;
1288			vp = addaliasu(vp, rdev);
1289			np->n_vnode = vp;
1290		}
1291		np->n_mtime = mtime.tv_sec;
1292	}
1293	vap = &np->n_vattr;
1294	vap->va_type = vtyp;
1295	vap->va_mode = (vmode & 07777);
1296	vap->va_rdev = rdev;
1297	vap->va_mtime = mtime;
1298	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1299	if (v3) {
1300		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1301		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1302		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1303		vap->va_size = fxdr_hyper(&fp->fa3_size);
1304		vap->va_blocksize = NFS_FABLKSIZE;
1305		vap->va_bytes = fxdr_hyper(&fp->fa3_used);
1306		vap->va_fileid = fxdr_unsigned(int32_t,
1307		    fp->fa3_fileid.nfsuquad[1]);
1308		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1309		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1310		vap->va_flags = 0;
1311		vap->va_filerev = 0;
1312	} else {
1313		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1314		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1315		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1316		vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1317		vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
1318		vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
1319		    * NFS_FABLKSIZE;
1320		vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1321		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1322		vap->va_flags = 0;
1323		vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1324		    fp->fa2_ctime.nfsv2_sec);
1325		vap->va_ctime.tv_nsec = 0;
1326		vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1327		vap->va_filerev = 0;
1328	}
1329	np->n_attrstamp = time_second;
1330	if (vap->va_size != np->n_size) {
1331		if (vap->va_type == VREG) {
1332			if (dontshrink && vap->va_size < np->n_size) {
1333				/*
1334				 * We've been told not to shrink the file;
1335				 * zero np->n_attrstamp to indicate that
1336				 * the attributes are stale.
1337				 */
1338				vap->va_size = np->n_size;
1339				np->n_attrstamp = 0;
1340			} else if (np->n_flag & NMODIFIED) {
1341				if (vap->va_size < np->n_size)
1342					vap->va_size = np->n_size;
1343				else
1344					np->n_size = vap->va_size;
1345			} else {
1346				np->n_size = vap->va_size;
1347			}
1348			vnode_pager_setsize(vp, np->n_size);
1349		} else {
1350			np->n_size = vap->va_size;
1351		}
1352	}
1353	if (vaper != NULL) {
1354		bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
1355		if (np->n_flag & NCHG) {
1356			if (np->n_flag & NACC)
1357				vaper->va_atime = np->n_atim;
1358			if (np->n_flag & NUPD)
1359				vaper->va_mtime = np->n_mtim;
1360		}
1361	}
1362	return (0);
1363}
1364
1365#ifdef NFS_ACDEBUG
1366#include <sys/sysctl.h>
1367SYSCTL_DECL(_vfs_nfs);
1368static int nfs_acdebug;
1369SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
1370#endif
1371
1372/*
1373 * Check the time stamp
1374 * If the cache is valid, copy contents to *vap and return 0
1375 * otherwise return an error
1376 */
1377int
1378nfs_getattrcache(vp, vaper)
1379	register struct vnode *vp;
1380	struct vattr *vaper;
1381{
1382	register struct nfsnode *np;
1383	register struct vattr *vap;
1384	struct nfsmount *nmp;
1385	int timeo;
1386
1387	np = VTONFS(vp);
1388	vap = &np->n_vattr;
1389	nmp = VFSTONFS(vp->v_mount);
1390	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
1391	timeo = (time_second - np->n_mtime) / 10;
1392
1393#ifdef NFS_ACDEBUG
1394	if (nfs_acdebug>1)
1395		printf("nfs_getattrcache: initial timeo = %d\n", timeo);
1396#endif
1397
1398	if (vap->va_type == VDIR) {
1399		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
1400			timeo = nmp->nm_acdirmin;
1401		else if (timeo > nmp->nm_acdirmax)
1402			timeo = nmp->nm_acdirmax;
1403	} else {
1404		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
1405			timeo = nmp->nm_acregmin;
1406		else if (timeo > nmp->nm_acregmax)
1407			timeo = nmp->nm_acregmax;
1408	}
1409
1410#ifdef NFS_ACDEBUG
1411	if (nfs_acdebug > 2)
1412		printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
1413			nmp->nm_acregmin, nmp->nm_acregmax,
1414			nmp->nm_acdirmin, nmp->nm_acdirmax);
1415
1416	if (nfs_acdebug)
1417		printf("nfs_getattrcache: age = %d; final timeo = %d\n",
1418			(time_second - np->n_attrstamp), timeo);
1419#endif
1420
1421	if ((time_second - np->n_attrstamp) >= timeo) {
1422		nfsstats.attrcache_misses++;
1423		return (ENOENT);
1424	}
1425	nfsstats.attrcache_hits++;
1426	if (vap->va_size != np->n_size) {
1427		if (vap->va_type == VREG) {
1428			if (np->n_flag & NMODIFIED) {
1429				if (vap->va_size < np->n_size)
1430					vap->va_size = np->n_size;
1431				else
1432					np->n_size = vap->va_size;
1433			} else {
1434				np->n_size = vap->va_size;
1435			}
1436			vnode_pager_setsize(vp, np->n_size);
1437		} else {
1438			np->n_size = vap->va_size;
1439		}
1440	}
1441	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
1442	if (np->n_flag & NCHG) {
1443		if (np->n_flag & NACC)
1444			vaper->va_atime = np->n_atim;
1445		if (np->n_flag & NUPD)
1446			vaper->va_mtime = np->n_mtim;
1447	}
1448	return (0);
1449}
1450
1451#ifndef NFS_NOSERVER
1452/*
1453 * Set up nameidata for a lookup() call and do it.
1454 *
1455 * If pubflag is set, this call is done for a lookup operation on the
1456 * public filehandle. In that case we allow crossing mountpoints and
1457 * absolute pathnames. However, the caller is expected to check that
1458 * the lookup result is within the public fs, and deny access if
1459 * it is not.
1460 *
1461 * nfs_namei() clears out garbage fields that namei() might leave garbage.
1462 * This is mainly ni_vp and ni_dvp when an error occurs, and ni_dvp when no
1463 * error occurs but the parent was not requested.
1464 *
1465 * dirp may be set whether an error is returned or not, and must be
1466 * released by the caller.
1467 */
1468int
1469nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
1470	register struct nameidata *ndp;
1471	fhandle_t *fhp;
1472	int len;
1473	struct nfssvc_sock *slp;
1474	struct sockaddr *nam;
1475	struct mbuf **mdp;
1476	caddr_t *dposp;
1477	struct vnode **retdirp;
1478	struct proc *p;
1479	int kerbflag, pubflag;
1480{
1481	register int i, rem;
1482	register struct mbuf *md;
1483	register char *fromcp, *tocp, *cp;
1484	struct iovec aiov;
1485	struct uio auio;
1486	struct vnode *dp;
1487	int error, rdonly, linklen;
1488	struct componentname *cnp = &ndp->ni_cnd;
1489
1490	*retdirp = (struct vnode *)0;
1491	cnp->cn_pnbuf = zalloc(namei_zone);
1492
1493	/*
1494	 * Copy the name from the mbuf list to ndp->ni_pnbuf
1495	 * and set the various ndp fields appropriately.
1496	 */
1497	fromcp = *dposp;
1498	tocp = cnp->cn_pnbuf;
1499	md = *mdp;
1500	rem = mtod(md, caddr_t) + md->m_len - fromcp;
1501	for (i = 0; i < len; i++) {
1502		while (rem == 0) {
1503			md = md->m_next;
1504			if (md == NULL) {
1505				error = EBADRPC;
1506				goto out;
1507			}
1508			fromcp = mtod(md, caddr_t);
1509			rem = md->m_len;
1510		}
1511		if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1512			error = EACCES;
1513			goto out;
1514		}
1515		*tocp++ = *fromcp++;
1516		rem--;
1517	}
1518	*tocp = '\0';
1519	*mdp = md;
1520	*dposp = fromcp;
1521	len = nfsm_rndup(len)-len;
1522	if (len > 0) {
1523		if (rem >= len)
1524			*dposp += len;
1525		else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1526			goto out;
1527	}
1528
1529	/*
1530	 * Extract and set starting directory.
1531	 */
1532	error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
1533	    nam, &rdonly, kerbflag, pubflag);
1534	if (error)
1535		goto out;
1536	if (dp->v_type != VDIR) {
1537		vrele(dp);
1538		error = ENOTDIR;
1539		goto out;
1540	}
1541
1542	if (rdonly)
1543		cnp->cn_flags |= RDONLY;
1544
1545	/*
1546	 * Set return directory.  Reference to dp is implicitly transfered
1547	 * to the returned pointer
1548	 */
1549	*retdirp = dp;
1550
1551	if (pubflag) {
1552		/*
1553		 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1554		 * and the 'native path' indicator.
1555		 */
1556		cp = zalloc(namei_zone);
1557		fromcp = cnp->cn_pnbuf;
1558		tocp = cp;
1559		if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1560			switch ((unsigned char)*fromcp) {
1561			case WEBNFS_NATIVE_CHAR:
1562				/*
1563				 * 'Native' path for us is the same
1564				 * as a path according to the NFS spec,
1565				 * just skip the escape char.
1566				 */
1567				fromcp++;
1568				break;
1569			/*
1570			 * More may be added in the future, range 0x80-0xff
1571			 */
1572			default:
1573				error = EIO;
1574				zfree(namei_zone, cp);
1575				goto out;
1576			}
1577		}
1578		/*
1579		 * Translate the '%' escapes, URL-style.
1580		 */
1581		while (*fromcp != '\0') {
1582			if (*fromcp == WEBNFS_ESC_CHAR) {
1583				if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1584					fromcp++;
1585					*tocp++ = HEXSTRTOI(fromcp);
1586					fromcp += 2;
1587					continue;
1588				} else {
1589					error = ENOENT;
1590					zfree(namei_zone, cp);
1591					goto out;
1592				}
1593			} else
1594				*tocp++ = *fromcp++;
1595		}
1596		*tocp = '\0';
1597		zfree(namei_zone, cnp->cn_pnbuf);
1598		cnp->cn_pnbuf = cp;
1599	}
1600
1601	ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
1602	ndp->ni_segflg = UIO_SYSSPACE;
1603
1604	if (pubflag) {
1605		ndp->ni_rootdir = rootvnode;
1606		ndp->ni_loopcnt = 0;
1607		if (cnp->cn_pnbuf[0] == '/')
1608			dp = rootvnode;
1609	} else {
1610		cnp->cn_flags |= NOCROSSMOUNT;
1611	}
1612
1613	/*
1614	 * Initialize for scan, set ni_startdir and bump ref on dp again
1615	 * becuase lookup() will dereference ni_startdir.
1616	 */
1617
1618	cnp->cn_proc = p;
1619	VREF(dp);
1620	ndp->ni_startdir = dp;
1621
1622	for (;;) {
1623		cnp->cn_nameptr = cnp->cn_pnbuf;
1624		/*
1625		 * Call lookup() to do the real work.  If an error occurs,
1626		 * ndp->ni_vp and ni_dvp are left uninitialized or NULL and
1627		 * we do not have to dereference anything before returning.
1628		 * In either case ni_startdir will be dereferenced and NULLed
1629		 * out.
1630		 */
1631		error = lookup(ndp);
1632		if (error)
1633			break;
1634
1635		/*
1636		 * Check for encountering a symbolic link.  Trivial
1637		 * termination occurs if no symlink encountered.
1638		 * Note: zfree is safe because error is 0, so we will
1639		 * not zfree it again when we break.
1640		 */
1641		if ((cnp->cn_flags & ISSYMLINK) == 0) {
1642			nfsrv_object_create(ndp->ni_vp);
1643			if (cnp->cn_flags & (SAVENAME | SAVESTART))
1644				cnp->cn_flags |= HASBUF;
1645			else
1646				zfree(namei_zone, cnp->cn_pnbuf);
1647			break;
1648		}
1649
1650		/*
1651		 * Validate symlink
1652		 */
1653		if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
1654			VOP_UNLOCK(ndp->ni_dvp, 0, p);
1655		if (!pubflag) {
1656			error = EINVAL;
1657			goto badlink2;
1658		}
1659
1660		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
1661			error = ELOOP;
1662			goto badlink2;
1663		}
1664		if (ndp->ni_pathlen > 1)
1665			cp = zalloc(namei_zone);
1666		else
1667			cp = cnp->cn_pnbuf;
1668		aiov.iov_base = cp;
1669		aiov.iov_len = MAXPATHLEN;
1670		auio.uio_iov = &aiov;
1671		auio.uio_iovcnt = 1;
1672		auio.uio_offset = 0;
1673		auio.uio_rw = UIO_READ;
1674		auio.uio_segflg = UIO_SYSSPACE;
1675		auio.uio_procp = (struct proc *)0;
1676		auio.uio_resid = MAXPATHLEN;
1677		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
1678		if (error) {
1679		badlink1:
1680			if (ndp->ni_pathlen > 1)
1681				zfree(namei_zone, cp);
1682		badlink2:
1683			vrele(ndp->ni_dvp);
1684			vput(ndp->ni_vp);
1685			break;
1686		}
1687		linklen = MAXPATHLEN - auio.uio_resid;
1688		if (linklen == 0) {
1689			error = ENOENT;
1690			goto badlink1;
1691		}
1692		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
1693			error = ENAMETOOLONG;
1694			goto badlink1;
1695		}
1696
1697		/*
1698		 * Adjust or replace path
1699		 */
1700		if (ndp->ni_pathlen > 1) {
1701			bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
1702			zfree(namei_zone, cnp->cn_pnbuf);
1703			cnp->cn_pnbuf = cp;
1704		} else
1705			cnp->cn_pnbuf[linklen] = '\0';
1706		ndp->ni_pathlen += linklen;
1707
1708		/*
1709		 * Cleanup refs for next loop and check if root directory
1710		 * should replace current directory.  Normally ni_dvp
1711		 * becomes the new base directory and is cleaned up when
1712		 * we loop.  Explicitly null pointers after invalidation
1713		 * to clarify operation.
1714		 */
1715		vput(ndp->ni_vp);
1716		ndp->ni_vp = NULL;
1717
1718		if (cnp->cn_pnbuf[0] == '/') {
1719			vrele(ndp->ni_dvp);
1720			ndp->ni_dvp = ndp->ni_rootdir;
1721			VREF(ndp->ni_dvp);
1722		}
1723		ndp->ni_startdir = ndp->ni_dvp;
1724		ndp->ni_dvp = NULL;
1725	}
1726
1727	/*
1728	 * nfs_namei() guarentees that fields will not contain garbage
1729	 * whether an error occurs or not.  This allows the caller to track
1730	 * cleanup state trivially.
1731	 */
1732out:
1733	if (error) {
1734		zfree(namei_zone, cnp->cn_pnbuf);
1735		ndp->ni_vp = NULL;
1736		ndp->ni_dvp = NULL;
1737		ndp->ni_startdir = NULL;
1738		cnp->cn_flags &= ~HASBUF;
1739	} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
1740		ndp->ni_dvp = NULL;
1741	}
1742	return (error);
1743}
1744
1745/*
1746 * A fiddled version of m_adj() that ensures null fill to a long
1747 * boundary and only trims off the back end
1748 */
1749void
1750nfsm_adj(mp, len, nul)
1751	struct mbuf *mp;
1752	register int len;
1753	int nul;
1754{
1755	register struct mbuf *m;
1756	register int count, i;
1757	register char *cp;
1758
1759	/*
1760	 * Trim from tail.  Scan the mbuf chain,
1761	 * calculating its length and finding the last mbuf.
1762	 * If the adjustment only affects this mbuf, then just
1763	 * adjust and return.  Otherwise, rescan and truncate
1764	 * after the remaining size.
1765	 */
1766	count = 0;
1767	m = mp;
1768	for (;;) {
1769		count += m->m_len;
1770		if (m->m_next == (struct mbuf *)0)
1771			break;
1772		m = m->m_next;
1773	}
1774	if (m->m_len > len) {
1775		m->m_len -= len;
1776		if (nul > 0) {
1777			cp = mtod(m, caddr_t)+m->m_len-nul;
1778			for (i = 0; i < nul; i++)
1779				*cp++ = '\0';
1780		}
1781		return;
1782	}
1783	count -= len;
1784	if (count < 0)
1785		count = 0;
1786	/*
1787	 * Correct length for chain is "count".
1788	 * Find the mbuf with last data, adjust its length,
1789	 * and toss data from remaining mbufs on chain.
1790	 */
1791	for (m = mp; m; m = m->m_next) {
1792		if (m->m_len >= count) {
1793			m->m_len = count;
1794			if (nul > 0) {
1795				cp = mtod(m, caddr_t)+m->m_len-nul;
1796				for (i = 0; i < nul; i++)
1797					*cp++ = '\0';
1798			}
1799			break;
1800		}
1801		count -= m->m_len;
1802	}
1803	for (m = m->m_next;m;m = m->m_next)
1804		m->m_len = 0;
1805}
1806
1807/*
1808 * Make these functions instead of macros, so that the kernel text size
1809 * doesn't get too big...
1810 */
1811void
1812nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
1813	struct nfsrv_descript *nfsd;
1814	int before_ret;
1815	register struct vattr *before_vap;
1816	int after_ret;
1817	struct vattr *after_vap;
1818	struct mbuf **mbp;
1819	char **bposp;
1820{
1821	register struct mbuf *mb = *mbp, *mb2;
1822	register char *bpos = *bposp;
1823	register u_int32_t *tl;
1824
1825	if (before_ret) {
1826		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1827		*tl = nfs_false;
1828	} else {
1829		nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
1830		*tl++ = nfs_true;
1831		txdr_hyper(before_vap->va_size, tl);
1832		tl += 2;
1833		txdr_nfsv3time(&(before_vap->va_mtime), tl);
1834		tl += 2;
1835		txdr_nfsv3time(&(before_vap->va_ctime), tl);
1836	}
1837	*bposp = bpos;
1838	*mbp = mb;
1839	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
1840}
1841
1842void
1843nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
1844	struct nfsrv_descript *nfsd;
1845	int after_ret;
1846	struct vattr *after_vap;
1847	struct mbuf **mbp;
1848	char **bposp;
1849{
1850	register struct mbuf *mb = *mbp, *mb2;
1851	register char *bpos = *bposp;
1852	register u_int32_t *tl;
1853	register struct nfs_fattr *fp;
1854
1855	if (after_ret) {
1856		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1857		*tl = nfs_false;
1858	} else {
1859		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
1860		*tl++ = nfs_true;
1861		fp = (struct nfs_fattr *)tl;
1862		nfsm_srvfattr(nfsd, after_vap, fp);
1863	}
1864	*mbp = mb;
1865	*bposp = bpos;
1866}
1867
1868void
1869nfsm_srvfattr(nfsd, vap, fp)
1870	register struct nfsrv_descript *nfsd;
1871	register struct vattr *vap;
1872	register struct nfs_fattr *fp;
1873{
1874
1875	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
1876	fp->fa_uid = txdr_unsigned(vap->va_uid);
1877	fp->fa_gid = txdr_unsigned(vap->va_gid);
1878	if (nfsd->nd_flag & ND_NFSV3) {
1879		fp->fa_type = vtonfsv3_type(vap->va_type);
1880		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
1881		txdr_hyper(vap->va_size, &fp->fa3_size);
1882		txdr_hyper(vap->va_bytes, &fp->fa3_used);
1883		fp->fa3_rdev.specdata1 = txdr_unsigned(umajor(vap->va_rdev));
1884		fp->fa3_rdev.specdata2 = txdr_unsigned(uminor(vap->va_rdev));
1885		fp->fa3_fsid.nfsuquad[0] = 0;
1886		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
1887		fp->fa3_fileid.nfsuquad[0] = 0;
1888		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
1889		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
1890		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
1891		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
1892	} else {
1893		fp->fa_type = vtonfsv2_type(vap->va_type);
1894		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
1895		fp->fa2_size = txdr_unsigned(vap->va_size);
1896		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
1897		if (vap->va_type == VFIFO)
1898			fp->fa2_rdev = 0xffffffff;
1899		else
1900			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
1901		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
1902		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
1903		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
1904		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
1905		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
1906		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
1907	}
1908}
1909
1910/*
1911 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
1912 * 	- look up fsid in mount list (if not found ret error)
1913 *	- get vp and export rights by calling VFS_FHTOVP()
1914 *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
1915 *	- if not lockflag unlock it with VOP_UNLOCK()
1916 */
1917int
1918nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag)
1919	fhandle_t *fhp;
1920	int lockflag;
1921	struct vnode **vpp;
1922	struct ucred *cred;
1923	struct nfssvc_sock *slp;
1924	struct sockaddr *nam;
1925	int *rdonlyp;
1926	int kerbflag;
1927	int pubflag;
1928{
1929	struct proc *p = curproc; /* XXX */
1930	register struct mount *mp;
1931	register int i;
1932	struct ucred *credanon;
1933	int error, exflags;
1934#ifdef MNT_EXNORESPORT		/* XXX needs mountd and /etc/exports help yet */
1935	struct sockaddr_int *saddr;
1936#endif
1937
1938	*vpp = (struct vnode *)0;
1939
1940	if (nfs_ispublicfh(fhp)) {
1941		if (!pubflag || !nfs_pub.np_valid)
1942			return (ESTALE);
1943		fhp = &nfs_pub.np_handle;
1944	}
1945
1946	mp = vfs_getvfs(&fhp->fh_fsid);
1947	if (!mp)
1948		return (ESTALE);
1949	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
1950	if (error)
1951		return (error);
1952	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
1953	if (error)
1954		return (error);
1955#ifdef MNT_EXNORESPORT
1956	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
1957		saddr = (struct sockaddr_in *)nam;
1958		if (saddr->sin_family == AF_INET &&
1959		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
1960			vput(*vpp);
1961			*vpp = NULL;
1962			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1963		}
1964	}
1965#endif
1966	/*
1967	 * Check/setup credentials.
1968	 */
1969	if (exflags & MNT_EXKERB) {
1970		if (!kerbflag) {
1971			vput(*vpp);
1972			*vpp = NULL;
1973			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1974		}
1975	} else if (kerbflag) {
1976		vput(*vpp);
1977		*vpp = NULL;
1978		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1979	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1980		cred->cr_uid = credanon->cr_uid;
1981		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
1982			cred->cr_groups[i] = credanon->cr_groups[i];
1983		cred->cr_ngroups = i;
1984	}
1985	if (exflags & MNT_EXRDONLY)
1986		*rdonlyp = 1;
1987	else
1988		*rdonlyp = 0;
1989
1990	nfsrv_object_create(*vpp);
1991
1992	if (!lockflag)
1993		VOP_UNLOCK(*vpp, 0, p);
1994	return (0);
1995}
1996
1997
1998/*
1999 * WebNFS: check if a filehandle is a public filehandle. For v3, this
2000 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
2001 * transformed this to all zeroes in both cases, so check for it.
2002 */
2003int
2004nfs_ispublicfh(fhp)
2005	fhandle_t *fhp;
2006{
2007	char *cp = (char *)fhp;
2008	int i;
2009
2010	for (i = 0; i < NFSX_V3FH; i++)
2011		if (*cp++ != 0)
2012			return (FALSE);
2013	return (TRUE);
2014}
2015
2016#endif /* NFS_NOSERVER */
2017/*
2018 * This function compares two net addresses by family and returns TRUE
2019 * if they are the same host.
2020 * If there is any doubt, return FALSE.
2021 * The AF_INET family is handled as a special case so that address mbufs
2022 * don't need to be saved to store "struct in_addr", which is only 4 bytes.
2023 */
2024int
2025netaddr_match(family, haddr, nam)
2026	int family;
2027	union nethostaddr *haddr;
2028	struct sockaddr *nam;
2029{
2030	register struct sockaddr_in *inetaddr;
2031
2032	switch (family) {
2033	case AF_INET:
2034		inetaddr = (struct sockaddr_in *)nam;
2035		if (inetaddr->sin_family == AF_INET &&
2036		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
2037			return (1);
2038		break;
2039	default:
2040		break;
2041	};
2042	return (0);
2043}
2044
2045static nfsuint64 nfs_nullcookie = { { 0, 0 } };
2046/*
2047 * This function finds the directory cookie that corresponds to the
2048 * logical byte offset given.
2049 */
2050nfsuint64 *
2051nfs_getcookie(np, off, add)
2052	register struct nfsnode *np;
2053	off_t off;
2054	int add;
2055{
2056	register struct nfsdmap *dp, *dp2;
2057	register int pos;
2058
2059	pos = (uoff_t)off / NFS_DIRBLKSIZ;
2060	if (pos == 0 || off < 0) {
2061#ifdef DIAGNOSTIC
2062		if (add)
2063			panic("nfs getcookie add at <= 0");
2064#endif
2065		return (&nfs_nullcookie);
2066	}
2067	pos--;
2068	dp = np->n_cookies.lh_first;
2069	if (!dp) {
2070		if (add) {
2071			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
2072				M_NFSDIROFF, M_WAITOK);
2073			dp->ndm_eocookie = 0;
2074			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
2075		} else
2076			return ((nfsuint64 *)0);
2077	}
2078	while (pos >= NFSNUMCOOKIES) {
2079		pos -= NFSNUMCOOKIES;
2080		if (dp->ndm_list.le_next) {
2081			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
2082				pos >= dp->ndm_eocookie)
2083				return ((nfsuint64 *)0);
2084			dp = dp->ndm_list.le_next;
2085		} else if (add) {
2086			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
2087				M_NFSDIROFF, M_WAITOK);
2088			dp2->ndm_eocookie = 0;
2089			LIST_INSERT_AFTER(dp, dp2, ndm_list);
2090			dp = dp2;
2091		} else
2092			return ((nfsuint64 *)0);
2093	}
2094	if (pos >= dp->ndm_eocookie) {
2095		if (add)
2096			dp->ndm_eocookie = pos + 1;
2097		else
2098			return ((nfsuint64 *)0);
2099	}
2100	return (&dp->ndm_cookies[pos]);
2101}
2102
2103/*
2104 * Invalidate cached directory information, except for the actual directory
2105 * blocks (which are invalidated separately).
2106 * Done mainly to avoid the use of stale offset cookies.
2107 */
2108void
2109nfs_invaldir(vp)
2110	register struct vnode *vp;
2111{
2112	register struct nfsnode *np = VTONFS(vp);
2113
2114#ifdef DIAGNOSTIC
2115	if (vp->v_type != VDIR)
2116		panic("nfs: invaldir not dir");
2117#endif
2118	np->n_direofoffset = 0;
2119	np->n_cookieverf.nfsuquad[0] = 0;
2120	np->n_cookieverf.nfsuquad[1] = 0;
2121	if (np->n_cookies.lh_first)
2122		np->n_cookies.lh_first->ndm_eocookie = 0;
2123}
2124
2125/*
2126 * The write verifier has changed (probably due to a server reboot), so all
2127 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
2128 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2129 * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
2130 * mount point.
2131 *
2132 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
2133 * writes are not clusterable.
2134 */
2135void
2136nfs_clearcommit(mp)
2137	struct mount *mp;
2138{
2139	register struct vnode *vp, *nvp;
2140	register struct buf *bp, *nbp;
2141	int s;
2142
2143	s = splbio();
2144loop:
2145	for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) {
2146		if (vp->v_mount != mp)	/* Paranoia */
2147			goto loop;
2148		nvp = vp->v_mntvnodes.le_next;
2149		for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
2150			nbp = TAILQ_NEXT(bp, b_vnbufs);
2151			if (BUF_REFCNT(bp) == 0 &&
2152			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
2153				== (B_DELWRI | B_NEEDCOMMIT))
2154				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
2155		}
2156	}
2157	splx(s);
2158}
2159
2160#ifndef NFS_NOSERVER
2161/*
2162 * Map errnos to NFS error numbers. For Version 3 also filter out error
2163 * numbers not specified for the associated procedure.
2164 */
2165int
2166nfsrv_errmap(nd, err)
2167	struct nfsrv_descript *nd;
2168	register int err;
2169{
2170	register short *defaulterrp, *errp;
2171
2172	if (nd->nd_flag & ND_NFSV3) {
2173	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
2174		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2175		while (*++errp) {
2176			if (*errp == err)
2177				return (err);
2178			else if (*errp > err)
2179				break;
2180		}
2181		return ((int)*defaulterrp);
2182	    } else
2183		return (err & 0xffff);
2184	}
2185	if (err <= ELAST)
2186		return ((int)nfsrv_v2errmap[err - 1]);
2187	return (NFSERR_IO);
2188}
2189
2190int
2191nfsrv_object_create(vp)
2192	struct vnode *vp;
2193{
2194
2195	if (vp == NULL || vp->v_type != VREG)
2196		return (1);
2197	return (vfs_object_create(vp, curproc,
2198				  curproc ? curproc->p_ucred : NULL));
2199}
2200
2201/*
2202 * Sort the group list in increasing numerical order.
2203 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2204 *  that used to be here.)
2205 */
2206void
2207nfsrvw_sort(list, num)
2208        register gid_t *list;
2209        register int num;
2210{
2211	register int i, j;
2212	gid_t v;
2213
2214	/* Insertion sort. */
2215	for (i = 1; i < num; i++) {
2216		v = list[i];
2217		/* find correct slot for value v, moving others up */
2218		for (j = i; --j >= 0 && v < list[j];)
2219			list[j + 1] = list[j];
2220		list[j + 1] = v;
2221	}
2222}
2223
2224/*
2225 * copy credentials making sure that the result can be compared with bcmp().
2226 */
2227void
2228nfsrv_setcred(incred, outcred)
2229	register struct ucred *incred, *outcred;
2230{
2231	register int i;
2232
2233	bzero((caddr_t)outcred, sizeof (struct ucred));
2234	outcred->cr_ref = 1;
2235	outcred->cr_uid = incred->cr_uid;
2236	outcred->cr_ngroups = incred->cr_ngroups;
2237	for (i = 0; i < incred->cr_ngroups; i++)
2238		outcred->cr_groups[i] = incred->cr_groups[i];
2239	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2240}
2241#endif /* NFS_NOSERVER */
2242