nfs_common.c revision 51138
1226031Sstas/*
2226031Sstas * Copyright (c) 1989, 1993
3226031Sstas *	The Regents of the University of California.  All rights reserved.
4226031Sstas *
5226031Sstas * This code is derived from software contributed to Berkeley by
6226031Sstas * Rick Macklem at The University of Guelph.
7226031Sstas *
8226031Sstas * Redistribution and use in source and binary forms, with or without
9226031Sstas * modification, are permitted provided that the following conditions
10226031Sstas * are met:
11226031Sstas * 1. Redistributions of source code must retain the above copyright
12226031Sstas *    notice, this list of conditions and the following disclaimer.
13226031Sstas * 2. Redistributions in binary form must reproduce the above copyright
14226031Sstas *    notice, this list of conditions and the following disclaimer in the
15226031Sstas *    documentation and/or other materials provided with the distribution.
16226031Sstas * 3. All advertising materials mentioning features or use of this software
17226031Sstas *    must display the following acknowledgement:
18226031Sstas *	This product includes software developed by the University of
19226031Sstas *	California, Berkeley and its contributors.
20226031Sstas * 4. Neither the name of the University nor the names of its contributors
21226031Sstas *    may be used to endorse or promote products derived from this software
22226031Sstas *    without specific prior written permission.
23226031Sstas *
24226031Sstas * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25226031Sstas * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26226031Sstas * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27226031Sstas * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28226031Sstas * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29226031Sstas * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30226031Sstas * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31226031Sstas * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32226031Sstas * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33226031Sstas * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34226031Sstas * SUCH DAMAGE.
35226031Sstas *
36226031Sstas *	@(#)nfs_subs.c  8.8 (Berkeley) 5/22/95
37226031Sstas * $FreeBSD: head/sys/nfs/nfs_common.c 51138 1999-09-11 00:46:08Z alfred $
38226031Sstas */
39226031Sstas
40226031Sstas/*
41226031Sstas * These functions support the macros and help fiddle mbuf chains for
42226031Sstas * the nfs op functions. They do things like create the rpc header and
43226031Sstas * copy data between mbuf chains and uio lists.
44226031Sstas */
45226031Sstas#include <sys/param.h>
46226031Sstas#include <sys/systm.h>
47226031Sstas#include <sys/kernel.h>
48226031Sstas#include <sys/buf.h>
49226031Sstas#include <sys/proc.h>
50226031Sstas#include <sys/mount.h>
51226031Sstas#include <sys/vnode.h>
52226031Sstas#include <sys/namei.h>
53226031Sstas#include <sys/mbuf.h>
54226031Sstas#include <sys/socket.h>
55226031Sstas#include <sys/stat.h>
56226031Sstas#include <sys/malloc.h>
57226031Sstas#include <sys/sysent.h>
58226031Sstas#include <sys/syscall.h>
59226031Sstas#include <sys/conf.h>
60226031Sstas
61226031Sstas#include <vm/vm.h>
62226031Sstas#include <vm/vm_object.h>
63226031Sstas#include <vm/vm_extern.h>
64226031Sstas#include <vm/vm_zone.h>
65226031Sstas
66226031Sstas#include <nfs/rpcv2.h>
67226031Sstas#include <nfs/nfsproto.h>
68226031Sstas#include <nfs/nfs.h>
69226031Sstas#include <nfs/nfsnode.h>
70226031Sstas#include <nfs/xdr_subs.h>
71226031Sstas#include <nfs/nfsm_subs.h>
72226031Sstas#include <nfs/nfsmount.h>
73226031Sstas#include <nfs/nqnfs.h>
74226031Sstas#include <nfs/nfsrtt.h>
75226031Sstas
76226031Sstas#include <netinet/in.h>
77226031Sstas#ifdef ISO
78226031Sstas#include <netiso/iso.h>
79226031Sstas#endif
80226031Sstas
81226031Sstas/*
82226031Sstas * Data items converted to xdr at startup, since they are constant
83226031Sstas * This is kinda hokey, but may save a little time doing byte swaps
84226031Sstas */
85226031Sstasu_int32_t nfs_xdrneg1;
86226031Sstasu_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
87226031Sstas	rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
88226031Sstas	rpc_auth_kerb;
89226031Sstasu_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
90226031Sstas
91226031Sstas/* And other global data */
92226031Sstasstatic u_int32_t nfs_xid = 0;
93226031Sstasstatic enum vtype nv2tov_type[8]= {
94226031Sstas	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON,  VNON
95226031Sstas};
96226031Sstasenum vtype nv3tov_type[8]= {
97226031Sstas	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
98226031Sstas};
99226031Sstas
100226031Sstasint nfs_ticks;
101226031Sstasint nfs_pbuf_freecnt = -1;	/* start out unlimited */
102226031Sstas
103226031Sstasstruct nfs_reqq nfs_reqq;
104226031Sstasstruct nfssvc_sockhead nfssvc_sockhead;
105226031Sstasint nfssvc_sockhead_flag;
106226031Sstasstruct nfsd_head nfsd_head;
107226031Sstasint nfsd_head_flag;
108226031Sstasstruct nfs_bufq nfs_bufq;
109226031Sstasstruct nqtimerhead nqtimerhead;
110226031Sstasstruct nqfhhashhead *nqfhhashtbl;
111226031Sstasu_long nqfhhash;
112226031Sstas
113226031Sstasstatic void (*nfs_prev_lease_updatetime) __P((int));
114226031Sstasstatic int nfs_prev_nfssvc_sy_narg;
115226031Sstasstatic sy_call_t *nfs_prev_nfssvc_sy_call;
116226031Sstas
117226031Sstas#ifndef NFS_NOSERVER
118226031Sstas
119226031Sstasstatic vop_t *nfs_prev_vop_lease_check;
120226031Sstasstatic int nfs_prev_getfh_sy_narg;
121226031Sstasstatic sy_call_t *nfs_prev_getfh_sy_call;
122226031Sstas
123226031Sstas/*
124226031Sstas * Mapping of old NFS Version 2 RPC numbers to generic numbers.
125226031Sstas */
126226031Sstasint nfsv3_procid[NFS_NPROCS] = {
127226031Sstas	NFSPROC_NULL,
128226031Sstas	NFSPROC_GETATTR,
129226031Sstas	NFSPROC_SETATTR,
130226031Sstas	NFSPROC_NOOP,
131226031Sstas	NFSPROC_LOOKUP,
132226031Sstas	NFSPROC_READLINK,
133226031Sstas	NFSPROC_READ,
134226031Sstas	NFSPROC_NOOP,
135226031Sstas	NFSPROC_WRITE,
136226031Sstas	NFSPROC_CREATE,
137226031Sstas	NFSPROC_REMOVE,
138226031Sstas	NFSPROC_RENAME,
139226031Sstas	NFSPROC_LINK,
140226031Sstas	NFSPROC_SYMLINK,
141226031Sstas	NFSPROC_MKDIR,
142226031Sstas	NFSPROC_RMDIR,
143226031Sstas	NFSPROC_READDIR,
144226031Sstas	NFSPROC_FSSTAT,
145226031Sstas	NFSPROC_NOOP,
146226031Sstas	NFSPROC_NOOP,
147226031Sstas	NFSPROC_NOOP,
148226031Sstas	NFSPROC_NOOP,
149226031Sstas	NFSPROC_NOOP,
150226031Sstas	NFSPROC_NOOP,
151226031Sstas	NFSPROC_NOOP,
152226031Sstas	NFSPROC_NOOP
153226031Sstas};
154226031Sstas
155226031Sstas#endif /* NFS_NOSERVER */
156226031Sstas/*
157226031Sstas * and the reverse mapping from generic to Version 2 procedure numbers
158226031Sstas */
159226031Sstasint nfsv2_procid[NFS_NPROCS] = {
160226031Sstas	NFSV2PROC_NULL,
161226031Sstas	NFSV2PROC_GETATTR,
162226031Sstas	NFSV2PROC_SETATTR,
163226031Sstas	NFSV2PROC_LOOKUP,
164226031Sstas	NFSV2PROC_NOOP,
165226031Sstas	NFSV2PROC_READLINK,
166226031Sstas	NFSV2PROC_READ,
167226031Sstas	NFSV2PROC_WRITE,
168226031Sstas	NFSV2PROC_CREATE,
169226031Sstas	NFSV2PROC_MKDIR,
170226031Sstas	NFSV2PROC_SYMLINK,
171226031Sstas	NFSV2PROC_CREATE,
172226031Sstas	NFSV2PROC_REMOVE,
173226031Sstas	NFSV2PROC_RMDIR,
174226031Sstas	NFSV2PROC_RENAME,
175226031Sstas	NFSV2PROC_LINK,
176226031Sstas	NFSV2PROC_READDIR,
177226031Sstas	NFSV2PROC_NOOP,
178226031Sstas	NFSV2PROC_STATFS,
179226031Sstas	NFSV2PROC_NOOP,
180226031Sstas	NFSV2PROC_NOOP,
181226031Sstas	NFSV2PROC_NOOP,
182226031Sstas	NFSV2PROC_NOOP,
183226031Sstas	NFSV2PROC_NOOP,
184226031Sstas	NFSV2PROC_NOOP,
185226031Sstas	NFSV2PROC_NOOP,
186226031Sstas};
187226031Sstas
188226031Sstas#ifndef NFS_NOSERVER
189226031Sstas/*
190226031Sstas * Maps errno values to nfs error numbers.
191226031Sstas * Use NFSERR_IO as the catch all for ones not specifically defined in
192226031Sstas * RFC 1094.
193226031Sstas */
194226031Sstasstatic u_char nfsrv_v2errmap[ELAST] = {
195226031Sstas  NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
196226031Sstas  NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
197226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
198226031Sstas  NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
199226031Sstas  NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
200226031Sstas  NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
201226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
202226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
203226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
204226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
205226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
206226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
207226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_NAMETOL,	NFSERR_IO,	NFSERR_IO,
208226031Sstas  NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
209226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
210226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
211226031Sstas  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
212226031Sstas  NFSERR_IO /* << Last is 86 */
213226031Sstas};
214226031Sstas
215226031Sstas/*
216226031Sstas * Maps errno values to nfs error numbers.
217226031Sstas * Although it is not obvious whether or not NFS clients really care if
218226031Sstas * a returned error value is in the specified list for the procedure, the
219226031Sstas * safest thing to do is filter them appropriately. For Version 2, the
220226031Sstas * X/Open XNFS document is the only specification that defines error values
221226031Sstas * for each RPC (The RFC simply lists all possible error values for all RPCs),
222226031Sstas * so I have decided to not do this for Version 2.
223226031Sstas * The first entry is the default error return and the rest are the valid
224226031Sstas * errors for that RPC in increasing numeric order.
225226031Sstas */
226226031Sstasstatic short nfsv3err_null[] = {
227226031Sstas	0,
228226031Sstas	0,
229226031Sstas};
230226031Sstas
231226031Sstasstatic short nfsv3err_getattr[] = {
232226031Sstas	NFSERR_IO,
233226031Sstas	NFSERR_IO,
234226031Sstas	NFSERR_STALE,
235226031Sstas	NFSERR_BADHANDLE,
236226031Sstas	NFSERR_SERVERFAULT,
237226031Sstas	0,
238226031Sstas};
239226031Sstas
240226031Sstasstatic short nfsv3err_setattr[] = {
241226031Sstas	NFSERR_IO,
242226031Sstas	NFSERR_PERM,
243226031Sstas	NFSERR_IO,
244226031Sstas	NFSERR_ACCES,
245226031Sstas	NFSERR_INVAL,
246226031Sstas	NFSERR_NOSPC,
247226031Sstas	NFSERR_ROFS,
248226031Sstas	NFSERR_DQUOT,
249226031Sstas	NFSERR_STALE,
250226031Sstas	NFSERR_BADHANDLE,
251226031Sstas	NFSERR_NOT_SYNC,
252226031Sstas	NFSERR_SERVERFAULT,
253226031Sstas	0,
254226031Sstas};
255226031Sstas
256226031Sstasstatic short nfsv3err_lookup[] = {
257226031Sstas	NFSERR_IO,
258226031Sstas	NFSERR_NOENT,
259226031Sstas	NFSERR_IO,
260226031Sstas	NFSERR_ACCES,
261226031Sstas	NFSERR_NOTDIR,
262226031Sstas	NFSERR_NAMETOL,
263226031Sstas	NFSERR_STALE,
264226031Sstas	NFSERR_BADHANDLE,
265226031Sstas	NFSERR_SERVERFAULT,
266226031Sstas	0,
267226031Sstas};
268226031Sstas
269226031Sstasstatic short nfsv3err_access[] = {
270226031Sstas	NFSERR_IO,
271226031Sstas	NFSERR_IO,
272226031Sstas	NFSERR_STALE,
273226031Sstas	NFSERR_BADHANDLE,
274226031Sstas	NFSERR_SERVERFAULT,
275226031Sstas	0,
276226031Sstas};
277226031Sstas
278226031Sstasstatic short nfsv3err_readlink[] = {
279226031Sstas	NFSERR_IO,
280226031Sstas	NFSERR_IO,
281226031Sstas	NFSERR_ACCES,
282226031Sstas	NFSERR_INVAL,
283226031Sstas	NFSERR_STALE,
284226031Sstas	NFSERR_BADHANDLE,
285226031Sstas	NFSERR_NOTSUPP,
286226031Sstas	NFSERR_SERVERFAULT,
287226031Sstas	0,
288226031Sstas};
289226031Sstas
290226031Sstasstatic short nfsv3err_read[] = {
291226031Sstas	NFSERR_IO,
292226031Sstas	NFSERR_IO,
293226031Sstas	NFSERR_NXIO,
294226031Sstas	NFSERR_ACCES,
295226031Sstas	NFSERR_INVAL,
296226031Sstas	NFSERR_STALE,
297226031Sstas	NFSERR_BADHANDLE,
298226031Sstas	NFSERR_SERVERFAULT,
299226031Sstas	0,
300226031Sstas};
301226031Sstas
302226031Sstasstatic short nfsv3err_write[] = {
303226031Sstas	NFSERR_IO,
304226031Sstas	NFSERR_IO,
305226031Sstas	NFSERR_ACCES,
306226031Sstas	NFSERR_INVAL,
307226031Sstas	NFSERR_FBIG,
308226031Sstas	NFSERR_NOSPC,
309226031Sstas	NFSERR_ROFS,
310226031Sstas	NFSERR_DQUOT,
311226031Sstas	NFSERR_STALE,
312226031Sstas	NFSERR_BADHANDLE,
313226031Sstas	NFSERR_SERVERFAULT,
314226031Sstas	0,
315226031Sstas};
316226031Sstas
317226031Sstasstatic short nfsv3err_create[] = {
318226031Sstas	NFSERR_IO,
319226031Sstas	NFSERR_IO,
320226031Sstas	NFSERR_ACCES,
321226031Sstas	NFSERR_EXIST,
322226031Sstas	NFSERR_NOTDIR,
323226031Sstas	NFSERR_NOSPC,
324226031Sstas	NFSERR_ROFS,
325226031Sstas	NFSERR_NAMETOL,
326226031Sstas	NFSERR_DQUOT,
327226031Sstas	NFSERR_STALE,
328226031Sstas	NFSERR_BADHANDLE,
329226031Sstas	NFSERR_NOTSUPP,
330226031Sstas	NFSERR_SERVERFAULT,
331226031Sstas	0,
332226031Sstas};
333226031Sstas
334226031Sstasstatic short nfsv3err_mkdir[] = {
335226031Sstas	NFSERR_IO,
336226031Sstas	NFSERR_IO,
337226031Sstas	NFSERR_ACCES,
338226031Sstas	NFSERR_EXIST,
339226031Sstas	NFSERR_NOTDIR,
340226031Sstas	NFSERR_NOSPC,
341226031Sstas	NFSERR_ROFS,
342226031Sstas	NFSERR_NAMETOL,
343226031Sstas	NFSERR_DQUOT,
344226031Sstas	NFSERR_STALE,
345226031Sstas	NFSERR_BADHANDLE,
346226031Sstas	NFSERR_NOTSUPP,
347226031Sstas	NFSERR_SERVERFAULT,
348226031Sstas	0,
349226031Sstas};
350226031Sstas
351226031Sstasstatic short nfsv3err_symlink[] = {
352226031Sstas	NFSERR_IO,
353226031Sstas	NFSERR_IO,
354226031Sstas	NFSERR_ACCES,
355226031Sstas	NFSERR_EXIST,
356226031Sstas	NFSERR_NOTDIR,
357226031Sstas	NFSERR_NOSPC,
358226031Sstas	NFSERR_ROFS,
359226031Sstas	NFSERR_NAMETOL,
360226031Sstas	NFSERR_DQUOT,
361226031Sstas	NFSERR_STALE,
362226031Sstas	NFSERR_BADHANDLE,
363226031Sstas	NFSERR_NOTSUPP,
364226031Sstas	NFSERR_SERVERFAULT,
365226031Sstas	0,
366226031Sstas};
367226031Sstas
368226031Sstasstatic short nfsv3err_mknod[] = {
369226031Sstas	NFSERR_IO,
370226031Sstas	NFSERR_IO,
371226031Sstas	NFSERR_ACCES,
372226031Sstas	NFSERR_EXIST,
373226031Sstas	NFSERR_NOTDIR,
374226031Sstas	NFSERR_NOSPC,
375226031Sstas	NFSERR_ROFS,
376226031Sstas	NFSERR_NAMETOL,
377226031Sstas	NFSERR_DQUOT,
378226031Sstas	NFSERR_STALE,
379226031Sstas	NFSERR_BADHANDLE,
380226031Sstas	NFSERR_NOTSUPP,
381226031Sstas	NFSERR_SERVERFAULT,
382226031Sstas	NFSERR_BADTYPE,
383226031Sstas	0,
384226031Sstas};
385226031Sstas
386226031Sstasstatic short nfsv3err_remove[] = {
387226031Sstas	NFSERR_IO,
388226031Sstas	NFSERR_NOENT,
389226031Sstas	NFSERR_IO,
390226031Sstas	NFSERR_ACCES,
391226031Sstas	NFSERR_NOTDIR,
392226031Sstas	NFSERR_ROFS,
393226031Sstas	NFSERR_NAMETOL,
394226031Sstas	NFSERR_STALE,
395226031Sstas	NFSERR_BADHANDLE,
396226031Sstas	NFSERR_SERVERFAULT,
397226031Sstas	0,
398226031Sstas};
399226031Sstas
400226031Sstasstatic short nfsv3err_rmdir[] = {
401226031Sstas	NFSERR_IO,
402226031Sstas	NFSERR_NOENT,
403226031Sstas	NFSERR_IO,
404226031Sstas	NFSERR_ACCES,
405226031Sstas	NFSERR_EXIST,
406226031Sstas	NFSERR_NOTDIR,
407226031Sstas	NFSERR_INVAL,
408226031Sstas	NFSERR_ROFS,
409226031Sstas	NFSERR_NAMETOL,
410226031Sstas	NFSERR_NOTEMPTY,
411226031Sstas	NFSERR_STALE,
412226031Sstas	NFSERR_BADHANDLE,
413226031Sstas	NFSERR_NOTSUPP,
414226031Sstas	NFSERR_SERVERFAULT,
415226031Sstas	0,
416226031Sstas};
417226031Sstas
418226031Sstasstatic short nfsv3err_rename[] = {
419226031Sstas	NFSERR_IO,
420226031Sstas	NFSERR_NOENT,
421226031Sstas	NFSERR_IO,
422226031Sstas	NFSERR_ACCES,
423226031Sstas	NFSERR_EXIST,
424226031Sstas	NFSERR_XDEV,
425226031Sstas	NFSERR_NOTDIR,
426226031Sstas	NFSERR_ISDIR,
427226031Sstas	NFSERR_INVAL,
428226031Sstas	NFSERR_NOSPC,
429226031Sstas	NFSERR_ROFS,
430226031Sstas	NFSERR_MLINK,
431226031Sstas	NFSERR_NAMETOL,
432226031Sstas	NFSERR_NOTEMPTY,
433226031Sstas	NFSERR_DQUOT,
434226031Sstas	NFSERR_STALE,
435226031Sstas	NFSERR_BADHANDLE,
436226031Sstas	NFSERR_NOTSUPP,
437226031Sstas	NFSERR_SERVERFAULT,
438226031Sstas	0,
439226031Sstas};
440226031Sstas
441226031Sstasstatic short nfsv3err_link[] = {
442226031Sstas	NFSERR_IO,
443226031Sstas	NFSERR_IO,
444226031Sstas	NFSERR_ACCES,
445226031Sstas	NFSERR_EXIST,
446226031Sstas	NFSERR_XDEV,
447226031Sstas	NFSERR_NOTDIR,
448226031Sstas	NFSERR_INVAL,
449226031Sstas	NFSERR_NOSPC,
450226031Sstas	NFSERR_ROFS,
451226031Sstas	NFSERR_MLINK,
452226031Sstas	NFSERR_NAMETOL,
453226031Sstas	NFSERR_DQUOT,
454226031Sstas	NFSERR_STALE,
455226031Sstas	NFSERR_BADHANDLE,
456226031Sstas	NFSERR_NOTSUPP,
457226031Sstas	NFSERR_SERVERFAULT,
458226031Sstas	0,
459226031Sstas};
460226031Sstas
461226031Sstasstatic short nfsv3err_readdir[] = {
462226031Sstas	NFSERR_IO,
463226031Sstas	NFSERR_IO,
464226031Sstas	NFSERR_ACCES,
465226031Sstas	NFSERR_NOTDIR,
466226031Sstas	NFSERR_STALE,
467226031Sstas	NFSERR_BADHANDLE,
468226031Sstas	NFSERR_BAD_COOKIE,
469226031Sstas	NFSERR_TOOSMALL,
470226031Sstas	NFSERR_SERVERFAULT,
471226031Sstas	0,
472226031Sstas};
473226031Sstas
474226031Sstasstatic short nfsv3err_readdirplus[] = {
475226031Sstas	NFSERR_IO,
476226031Sstas	NFSERR_IO,
477226031Sstas	NFSERR_ACCES,
478226031Sstas	NFSERR_NOTDIR,
479226031Sstas	NFSERR_STALE,
480226031Sstas	NFSERR_BADHANDLE,
481226031Sstas	NFSERR_BAD_COOKIE,
482226031Sstas	NFSERR_NOTSUPP,
483226031Sstas	NFSERR_TOOSMALL,
484226031Sstas	NFSERR_SERVERFAULT,
485226031Sstas	0,
486226031Sstas};
487226031Sstas
488226031Sstasstatic short nfsv3err_fsstat[] = {
489226031Sstas	NFSERR_IO,
490226031Sstas	NFSERR_IO,
491226031Sstas	NFSERR_STALE,
492226031Sstas	NFSERR_BADHANDLE,
493226031Sstas	NFSERR_SERVERFAULT,
494226031Sstas	0,
495226031Sstas};
496226031Sstas
497226031Sstasstatic short nfsv3err_fsinfo[] = {
498226031Sstas	NFSERR_STALE,
499226031Sstas	NFSERR_STALE,
500226031Sstas	NFSERR_BADHANDLE,
501226031Sstas	NFSERR_SERVERFAULT,
502226031Sstas	0,
503226031Sstas};
504226031Sstas
505226031Sstasstatic short nfsv3err_pathconf[] = {
506226031Sstas	NFSERR_STALE,
507226031Sstas	NFSERR_STALE,
508226031Sstas	NFSERR_BADHANDLE,
509226031Sstas	NFSERR_SERVERFAULT,
510226031Sstas	0,
511226031Sstas};
512226031Sstas
513226031Sstasstatic short nfsv3err_commit[] = {
514226031Sstas	NFSERR_IO,
515226031Sstas	NFSERR_IO,
516226031Sstas	NFSERR_STALE,
517226031Sstas	NFSERR_BADHANDLE,
518226031Sstas	NFSERR_SERVERFAULT,
519226031Sstas	0,
520226031Sstas};
521226031Sstas
522226031Sstasstatic short *nfsrv_v3errmap[] = {
523226031Sstas	nfsv3err_null,
524226031Sstas	nfsv3err_getattr,
525226031Sstas	nfsv3err_setattr,
526226031Sstas	nfsv3err_lookup,
527226031Sstas	nfsv3err_access,
528226031Sstas	nfsv3err_readlink,
529226031Sstas	nfsv3err_read,
530226031Sstas	nfsv3err_write,
531226031Sstas	nfsv3err_create,
532226031Sstas	nfsv3err_mkdir,
533226031Sstas	nfsv3err_symlink,
534226031Sstas	nfsv3err_mknod,
535226031Sstas	nfsv3err_remove,
536226031Sstas	nfsv3err_rmdir,
537226031Sstas	nfsv3err_rename,
538226031Sstas	nfsv3err_link,
539226031Sstas	nfsv3err_readdir,
540226031Sstas	nfsv3err_readdirplus,
541226031Sstas	nfsv3err_fsstat,
542226031Sstas	nfsv3err_fsinfo,
543226031Sstas	nfsv3err_pathconf,
544226031Sstas	nfsv3err_commit,
545226031Sstas};
546226031Sstas
547226031Sstas#endif /* NFS_NOSERVER */
548226031Sstas
549226031Sstasextern struct nfsrtt nfsrtt;
550226031Sstasextern time_t nqnfsstarttime;
551226031Sstasextern int nqsrv_clockskew;
552226031Sstasextern int nqsrv_writeslack;
553226031Sstasextern int nqsrv_maxlease;
554226031Sstasextern struct nfsstats nfsstats;
555226031Sstasextern int nqnfs_piggy[NFS_NPROCS];
556226031Sstasextern nfstype nfsv2_type[9];
557226031Sstasextern nfstype nfsv3_type[9];
558226031Sstasextern struct nfsnodehashhead *nfsnodehashtbl;
559226031Sstasextern u_long nfsnodehash;
560
561struct getfh_args;
562extern int getfh(struct proc *, struct getfh_args *, int *);
563struct nfssvc_args;
564extern int nfssvc(struct proc *, struct nfssvc_args *, int *);
565
566LIST_HEAD(nfsnodehashhead, nfsnode);
567
568int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
569
570u_quad_t
571nfs_curusec()
572{
573	struct timeval tv;
574
575	getmicrotime(&tv);
576	return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
577}
578
579/*
580 * Create the header for an rpc request packet
581 * The hsiz is the size of the rest of the nfs request header.
582 * (just used to decide if a cluster is a good idea)
583 */
584struct mbuf *
585nfsm_reqh(vp, procid, hsiz, bposp)
586	struct vnode *vp;
587	u_long procid;
588	int hsiz;
589	caddr_t *bposp;
590{
591	register struct mbuf *mb;
592	register u_int32_t *tl;
593	register caddr_t bpos;
594	struct mbuf *mb2;
595	struct nfsmount *nmp;
596	int nqflag;
597
598	MGET(mb, M_WAIT, MT_DATA);
599	if (hsiz >= MINCLSIZE)
600		MCLGET(mb, M_WAIT);
601	mb->m_len = 0;
602	bpos = mtod(mb, caddr_t);
603
604	/*
605	 * For NQNFS, add lease request.
606	 */
607	if (vp) {
608		nmp = VFSTONFS(vp->v_mount);
609		if (nmp->nm_flag & NFSMNT_NQNFS) {
610			nqflag = NQNFS_NEEDLEASE(vp, procid);
611			if (nqflag) {
612				nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
613				*tl++ = txdr_unsigned(nqflag);
614				*tl = txdr_unsigned(nmp->nm_leaseterm);
615			} else {
616				nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
617				*tl = 0;
618			}
619		}
620	}
621	/* Finally, return values */
622	*bposp = bpos;
623	return (mb);
624}
625
626/*
627 * Build the RPC header and fill in the authorization info.
628 * The authorization string argument is only used when the credentials
629 * come from outside of the kernel.
630 * Returns the head of the mbuf list.
631 */
632struct mbuf *
633nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
634	verf_str, mrest, mrest_len, mbp, xidp)
635	register struct ucred *cr;
636	int nmflag;
637	int procid;
638	int auth_type;
639	int auth_len;
640	char *auth_str;
641	int verf_len;
642	char *verf_str;
643	struct mbuf *mrest;
644	int mrest_len;
645	struct mbuf **mbp;
646	u_int32_t *xidp;
647{
648	register struct mbuf *mb;
649	register u_int32_t *tl;
650	register caddr_t bpos;
651	register int i;
652	struct mbuf *mreq, *mb2;
653	int siz, grpsiz, authsiz;
654
655	authsiz = nfsm_rndup(auth_len);
656	MGETHDR(mb, M_WAIT, MT_DATA);
657	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
658		MCLGET(mb, M_WAIT);
659	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
660		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
661	} else {
662		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
663	}
664	mb->m_len = 0;
665	mreq = mb;
666	bpos = mtod(mb, caddr_t);
667
668	/*
669	 * First the RPC header.
670	 */
671	nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
672
673	/* Get a pretty random xid to start with */
674	if (!nfs_xid)
675		nfs_xid = random();
676	/*
677	 * Skip zero xid if it should ever happen.
678	 */
679	if (++nfs_xid == 0)
680		nfs_xid++;
681
682	*tl++ = *xidp = txdr_unsigned(nfs_xid);
683	*tl++ = rpc_call;
684	*tl++ = rpc_vers;
685	if (nmflag & NFSMNT_NQNFS) {
686		*tl++ = txdr_unsigned(NQNFS_PROG);
687		*tl++ = txdr_unsigned(NQNFS_VER3);
688	} else {
689		*tl++ = txdr_unsigned(NFS_PROG);
690		if (nmflag & NFSMNT_NFSV3)
691			*tl++ = txdr_unsigned(NFS_VER3);
692		else
693			*tl++ = txdr_unsigned(NFS_VER2);
694	}
695	if (nmflag & NFSMNT_NFSV3)
696		*tl++ = txdr_unsigned(procid);
697	else
698		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
699
700	/*
701	 * And then the authorization cred.
702	 */
703	*tl++ = txdr_unsigned(auth_type);
704	*tl = txdr_unsigned(authsiz);
705	switch (auth_type) {
706	case RPCAUTH_UNIX:
707		nfsm_build(tl, u_int32_t *, auth_len);
708		*tl++ = 0;		/* stamp ?? */
709		*tl++ = 0;		/* NULL hostname */
710		*tl++ = txdr_unsigned(cr->cr_uid);
711		*tl++ = txdr_unsigned(cr->cr_groups[0]);
712		grpsiz = (auth_len >> 2) - 5;
713		*tl++ = txdr_unsigned(grpsiz);
714		for (i = 1; i <= grpsiz; i++)
715			*tl++ = txdr_unsigned(cr->cr_groups[i]);
716		break;
717	case RPCAUTH_KERB4:
718		siz = auth_len;
719		while (siz > 0) {
720			if (M_TRAILINGSPACE(mb) == 0) {
721				MGET(mb2, M_WAIT, MT_DATA);
722				if (siz >= MINCLSIZE)
723					MCLGET(mb2, M_WAIT);
724				mb->m_next = mb2;
725				mb = mb2;
726				mb->m_len = 0;
727				bpos = mtod(mb, caddr_t);
728			}
729			i = min(siz, M_TRAILINGSPACE(mb));
730			bcopy(auth_str, bpos, i);
731			mb->m_len += i;
732			auth_str += i;
733			bpos += i;
734			siz -= i;
735		}
736		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
737			for (i = 0; i < siz; i++)
738				*bpos++ = '\0';
739			mb->m_len += siz;
740		}
741		break;
742	};
743
744	/*
745	 * And the verifier...
746	 */
747	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
748	if (verf_str) {
749		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
750		*tl = txdr_unsigned(verf_len);
751		siz = verf_len;
752		while (siz > 0) {
753			if (M_TRAILINGSPACE(mb) == 0) {
754				MGET(mb2, M_WAIT, MT_DATA);
755				if (siz >= MINCLSIZE)
756					MCLGET(mb2, M_WAIT);
757				mb->m_next = mb2;
758				mb = mb2;
759				mb->m_len = 0;
760				bpos = mtod(mb, caddr_t);
761			}
762			i = min(siz, M_TRAILINGSPACE(mb));
763			bcopy(verf_str, bpos, i);
764			mb->m_len += i;
765			verf_str += i;
766			bpos += i;
767			siz -= i;
768		}
769		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
770			for (i = 0; i < siz; i++)
771				*bpos++ = '\0';
772			mb->m_len += siz;
773		}
774	} else {
775		*tl++ = txdr_unsigned(RPCAUTH_NULL);
776		*tl = 0;
777	}
778	mb->m_next = mrest;
779	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
780	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
781	*mbp = mb;
782	return (mreq);
783}
784
785/*
786 * copies mbuf chain to the uio scatter/gather list
787 */
788int
789nfsm_mbuftouio(mrep, uiop, siz, dpos)
790	struct mbuf **mrep;
791	register struct uio *uiop;
792	int siz;
793	caddr_t *dpos;
794{
795	register char *mbufcp, *uiocp;
796	register int xfer, left, len;
797	register struct mbuf *mp;
798	long uiosiz, rem;
799	int error = 0;
800
801	mp = *mrep;
802	mbufcp = *dpos;
803	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
804	rem = nfsm_rndup(siz)-siz;
805	while (siz > 0) {
806		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
807			return (EFBIG);
808		left = uiop->uio_iov->iov_len;
809		uiocp = uiop->uio_iov->iov_base;
810		if (left > siz)
811			left = siz;
812		uiosiz = left;
813		while (left > 0) {
814			while (len == 0) {
815				mp = mp->m_next;
816				if (mp == NULL)
817					return (EBADRPC);
818				mbufcp = mtod(mp, caddr_t);
819				len = mp->m_len;
820			}
821			xfer = (left > len) ? len : left;
822#ifdef notdef
823			/* Not Yet.. */
824			if (uiop->uio_iov->iov_op != NULL)
825				(*(uiop->uio_iov->iov_op))
826				(mbufcp, uiocp, xfer);
827			else
828#endif
829			if (uiop->uio_segflg == UIO_SYSSPACE)
830				bcopy(mbufcp, uiocp, xfer);
831			else
832				copyout(mbufcp, uiocp, xfer);
833			left -= xfer;
834			len -= xfer;
835			mbufcp += xfer;
836			uiocp += xfer;
837			uiop->uio_offset += xfer;
838			uiop->uio_resid -= xfer;
839		}
840		if (uiop->uio_iov->iov_len <= siz) {
841			uiop->uio_iovcnt--;
842			uiop->uio_iov++;
843		} else {
844			uiop->uio_iov->iov_base += uiosiz;
845			uiop->uio_iov->iov_len -= uiosiz;
846		}
847		siz -= uiosiz;
848	}
849	*dpos = mbufcp;
850	*mrep = mp;
851	if (rem > 0) {
852		if (len < rem)
853			error = nfs_adv(mrep, dpos, rem, len);
854		else
855			*dpos += rem;
856	}
857	return (error);
858}
859
860/*
861 * copies a uio scatter/gather list to an mbuf chain.
862 * NOTE: can ony handle iovcnt == 1
863 */
864int
865nfsm_uiotombuf(uiop, mq, siz, bpos)
866	register struct uio *uiop;
867	struct mbuf **mq;
868	int siz;
869	caddr_t *bpos;
870{
871	register char *uiocp;
872	register struct mbuf *mp, *mp2;
873	register int xfer, left, mlen;
874	int uiosiz, clflg, rem;
875	char *cp;
876
877#ifdef DIAGNOSTIC
878	if (uiop->uio_iovcnt != 1)
879		panic("nfsm_uiotombuf: iovcnt != 1");
880#endif
881
882	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
883		clflg = 1;
884	else
885		clflg = 0;
886	rem = nfsm_rndup(siz)-siz;
887	mp = mp2 = *mq;
888	while (siz > 0) {
889		left = uiop->uio_iov->iov_len;
890		uiocp = uiop->uio_iov->iov_base;
891		if (left > siz)
892			left = siz;
893		uiosiz = left;
894		while (left > 0) {
895			mlen = M_TRAILINGSPACE(mp);
896			if (mlen == 0) {
897				MGET(mp, M_WAIT, MT_DATA);
898				if (clflg)
899					MCLGET(mp, M_WAIT);
900				mp->m_len = 0;
901				mp2->m_next = mp;
902				mp2 = mp;
903				mlen = M_TRAILINGSPACE(mp);
904			}
905			xfer = (left > mlen) ? mlen : left;
906#ifdef notdef
907			/* Not Yet.. */
908			if (uiop->uio_iov->iov_op != NULL)
909				(*(uiop->uio_iov->iov_op))
910				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
911			else
912#endif
913			if (uiop->uio_segflg == UIO_SYSSPACE)
914				bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
915			else
916				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
917			mp->m_len += xfer;
918			left -= xfer;
919			uiocp += xfer;
920			uiop->uio_offset += xfer;
921			uiop->uio_resid -= xfer;
922		}
923		uiop->uio_iov->iov_base += uiosiz;
924		uiop->uio_iov->iov_len -= uiosiz;
925		siz -= uiosiz;
926	}
927	if (rem > 0) {
928		if (rem > M_TRAILINGSPACE(mp)) {
929			MGET(mp, M_WAIT, MT_DATA);
930			mp->m_len = 0;
931			mp2->m_next = mp;
932		}
933		cp = mtod(mp, caddr_t)+mp->m_len;
934		for (left = 0; left < rem; left++)
935			*cp++ = '\0';
936		mp->m_len += rem;
937		*bpos = cp;
938	} else
939		*bpos = mtod(mp, caddr_t)+mp->m_len;
940	*mq = mp;
941	return (0);
942}
943
944/*
945 * Help break down an mbuf chain by setting the first siz bytes contiguous
946 * pointed to by returned val.
947 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
948 * cases. (The macros use the vars. dpos and dpos2)
949 */
950int
951nfsm_disct(mdp, dposp, siz, left, cp2)
952	struct mbuf **mdp;
953	caddr_t *dposp;
954	int siz;
955	int left;
956	caddr_t *cp2;
957{
958	register struct mbuf *mp, *mp2;
959	register int siz2, xfer;
960	register caddr_t p;
961
962	mp = *mdp;
963	while (left == 0) {
964		*mdp = mp = mp->m_next;
965		if (mp == NULL)
966			return (EBADRPC);
967		left = mp->m_len;
968		*dposp = mtod(mp, caddr_t);
969	}
970	if (left >= siz) {
971		*cp2 = *dposp;
972		*dposp += siz;
973	} else if (mp->m_next == NULL) {
974		return (EBADRPC);
975	} else if (siz > MHLEN) {
976		panic("nfs S too big");
977	} else {
978		MGET(mp2, M_WAIT, MT_DATA);
979		mp2->m_next = mp->m_next;
980		mp->m_next = mp2;
981		mp->m_len -= left;
982		mp = mp2;
983		*cp2 = p = mtod(mp, caddr_t);
984		bcopy(*dposp, p, left);		/* Copy what was left */
985		siz2 = siz-left;
986		p += left;
987		mp2 = mp->m_next;
988		/* Loop around copying up the siz2 bytes */
989		while (siz2 > 0) {
990			if (mp2 == NULL)
991				return (EBADRPC);
992			xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
993			if (xfer > 0) {
994				bcopy(mtod(mp2, caddr_t), p, xfer);
995				NFSMADV(mp2, xfer);
996				mp2->m_len -= xfer;
997				p += xfer;
998				siz2 -= xfer;
999			}
1000			if (siz2 > 0)
1001				mp2 = mp2->m_next;
1002		}
1003		mp->m_len = siz;
1004		*mdp = mp2;
1005		*dposp = mtod(mp2, caddr_t);
1006	}
1007	return (0);
1008}
1009
1010/*
1011 * Advance the position in the mbuf chain.
1012 */
1013int
1014nfs_adv(mdp, dposp, offs, left)
1015	struct mbuf **mdp;
1016	caddr_t *dposp;
1017	int offs;
1018	int left;
1019{
1020	register struct mbuf *m;
1021	register int s;
1022
1023	m = *mdp;
1024	s = left;
1025	while (s < offs) {
1026		offs -= s;
1027		m = m->m_next;
1028		if (m == NULL)
1029			return (EBADRPC);
1030		s = m->m_len;
1031	}
1032	*mdp = m;
1033	*dposp = mtod(m, caddr_t)+offs;
1034	return (0);
1035}
1036
1037/*
1038 * Copy a string into mbufs for the hard cases...
1039 */
1040int
1041nfsm_strtmbuf(mb, bpos, cp, siz)
1042	struct mbuf **mb;
1043	char **bpos;
1044	const char *cp;
1045	long siz;
1046{
1047	register struct mbuf *m1 = NULL, *m2;
1048	long left, xfer, len, tlen;
1049	u_int32_t *tl;
1050	int putsize;
1051
1052	putsize = 1;
1053	m2 = *mb;
1054	left = M_TRAILINGSPACE(m2);
1055	if (left > 0) {
1056		tl = ((u_int32_t *)(*bpos));
1057		*tl++ = txdr_unsigned(siz);
1058		putsize = 0;
1059		left -= NFSX_UNSIGNED;
1060		m2->m_len += NFSX_UNSIGNED;
1061		if (left > 0) {
1062			bcopy(cp, (caddr_t) tl, left);
1063			siz -= left;
1064			cp += left;
1065			m2->m_len += left;
1066			left = 0;
1067		}
1068	}
1069	/* Loop around adding mbufs */
1070	while (siz > 0) {
1071		MGET(m1, M_WAIT, MT_DATA);
1072		if (siz > MLEN)
1073			MCLGET(m1, M_WAIT);
1074		m1->m_len = NFSMSIZ(m1);
1075		m2->m_next = m1;
1076		m2 = m1;
1077		tl = mtod(m1, u_int32_t *);
1078		tlen = 0;
1079		if (putsize) {
1080			*tl++ = txdr_unsigned(siz);
1081			m1->m_len -= NFSX_UNSIGNED;
1082			tlen = NFSX_UNSIGNED;
1083			putsize = 0;
1084		}
1085		if (siz < m1->m_len) {
1086			len = nfsm_rndup(siz);
1087			xfer = siz;
1088			if (xfer < len)
1089				*(tl+(xfer>>2)) = 0;
1090		} else {
1091			xfer = len = m1->m_len;
1092		}
1093		bcopy(cp, (caddr_t) tl, xfer);
1094		m1->m_len = len+tlen;
1095		siz -= xfer;
1096		cp += xfer;
1097	}
1098	*mb = m1;
1099	*bpos = mtod(m1, caddr_t)+m1->m_len;
1100	return (0);
1101}
1102
1103/*
1104 * Called once to initialize data structures...
1105 */
1106int
1107nfs_init(vfsp)
1108	struct vfsconf *vfsp;
1109{
1110	register int i;
1111
1112	nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
1113
1114	/*
1115	 * Check to see if major data structures haven't bloated.
1116	 */
1117	if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) {
1118		printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC);
1119		printf("Try reducing NFS_UIDHASHSIZ\n");
1120	}
1121	if (sizeof (struct nfsuid) > NFS_UIDALLOC) {
1122		printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC);
1123		printf("Try unionizing the nu_nickname and nu_flag fields\n");
1124	}
1125	nfs_mount_type = vfsp->vfc_typenum;
1126	nfsrtt.pos = 0;
1127	rpc_vers = txdr_unsigned(RPC_VER2);
1128	rpc_call = txdr_unsigned(RPC_CALL);
1129	rpc_reply = txdr_unsigned(RPC_REPLY);
1130	rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
1131	rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
1132	rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
1133	rpc_autherr = txdr_unsigned(RPC_AUTHERR);
1134	rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
1135	rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
1136	nfs_prog = txdr_unsigned(NFS_PROG);
1137	nqnfs_prog = txdr_unsigned(NQNFS_PROG);
1138	nfs_true = txdr_unsigned(TRUE);
1139	nfs_false = txdr_unsigned(FALSE);
1140	nfs_xdrneg1 = txdr_unsigned(-1);
1141	nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
1142	if (nfs_ticks < 1)
1143		nfs_ticks = 1;
1144	/* Ensure async daemons disabled */
1145	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1146		nfs_iodwant[i] = (struct proc *)0;
1147		nfs_iodmount[i] = (struct nfsmount *)0;
1148	}
1149	nfs_nhinit();			/* Init the nfsnode table */
1150#ifndef NFS_NOSERVER
1151	nfsrv_init(0);			/* Init server data structures */
1152	nfsrv_initcache();		/* Init the server request cache */
1153#endif
1154
1155	/*
1156	 * Initialize the nqnfs server stuff.
1157	 */
1158	if (nqnfsstarttime == 0) {
1159		nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
1160			+ nqsrv_clockskew + nqsrv_writeslack;
1161		NQLOADNOVRAM(nqnfsstarttime);
1162		CIRCLEQ_INIT(&nqtimerhead);
1163		nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash);
1164	}
1165
1166	/*
1167	 * Initialize reply list and start timer
1168	 */
1169	TAILQ_INIT(&nfs_reqq);
1170
1171	nfs_timer(0);
1172
1173	/*
1174	 * Set up lease_check and lease_updatetime so that other parts
1175	 * of the system can call us, if we are loadable.
1176	 */
1177#ifndef NFS_NOSERVER
1178	nfs_prev_vop_lease_check = default_vnodeop_p[VOFFSET(vop_lease)];
1179	default_vnodeop_p[VOFFSET(vop_lease)] = (vop_t *)nqnfs_vop_lease_check;
1180#endif
1181	nfs_prev_lease_updatetime = lease_updatetime;
1182	lease_updatetime = nfs_lease_updatetime;
1183	nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
1184	sysent[SYS_nfssvc].sy_narg = 2;
1185	nfs_prev_nfssvc_sy_call = sysent[SYS_nfssvc].sy_call;
1186	sysent[SYS_nfssvc].sy_call = (sy_call_t *)nfssvc;
1187#ifndef NFS_NOSERVER
1188	nfs_prev_getfh_sy_narg = sysent[SYS_getfh].sy_narg;
1189	sysent[SYS_getfh].sy_narg = 2;
1190	nfs_prev_getfh_sy_call = sysent[SYS_getfh].sy_call;
1191	sysent[SYS_getfh].sy_call = (sy_call_t *)getfh;
1192#endif
1193
1194	nfs_pbuf_freecnt = nswbuf / 2 + 1;
1195
1196	return (0);
1197}
1198
1199int
1200nfs_uninit(vfsp)
1201	struct vfsconf *vfsp;
1202{
1203
1204	untimeout(nfs_timer, (void *)NULL, nfs_timer_handle);
1205	nfs_mount_type = -1;
1206#ifndef NFS_NOSERVER
1207	default_vnodeop_p[VOFFSET(vop_lease)] = nfs_prev_vop_lease_check;
1208#endif
1209	lease_updatetime = nfs_prev_lease_updatetime;
1210	sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
1211	sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
1212#ifndef NFS_NOSERVER
1213	sysent[SYS_getfh].sy_narg = nfs_prev_getfh_sy_narg;
1214	sysent[SYS_getfh].sy_call = nfs_prev_getfh_sy_call;
1215#endif
1216	return (0);
1217}
1218
1219/*
1220 * Attribute cache routines.
1221 * nfs_loadattrcache() - loads or updates the cache contents from attributes
1222 *	that are on the mbuf list
1223 * nfs_getattrcache() - returns valid attributes if found in cache, returns
1224 *	error otherwise
1225 */
1226
1227/*
1228 * Load the attribute cache (that lives in the nfsnode entry) with
1229 * the values on the mbuf list and
1230 * Iff vap not NULL
1231 *    copy the attributes to *vaper
1232 */
1233int
1234nfs_loadattrcache(vpp, mdp, dposp, vaper)
1235	struct vnode **vpp;
1236	struct mbuf **mdp;
1237	caddr_t *dposp;
1238	struct vattr *vaper;
1239{
1240	register struct vnode *vp = *vpp;
1241	register struct vattr *vap;
1242	register struct nfs_fattr *fp;
1243	register struct nfsnode *np;
1244	register int32_t t1;
1245	caddr_t cp2;
1246	int error = 0, rdev;
1247	struct mbuf *md;
1248	enum vtype vtyp;
1249	u_short vmode;
1250	struct timespec mtime;
1251	int v3 = NFS_ISV3(vp);
1252
1253	md = *mdp;
1254	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1255	if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
1256		return (error);
1257	fp = (struct nfs_fattr *)cp2;
1258	if (v3) {
1259		vtyp = nfsv3tov_type(fp->fa_type);
1260		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1261		rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
1262			fxdr_unsigned(int, fp->fa3_rdev.specdata2));
1263		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1264	} else {
1265		vtyp = nfsv2tov_type(fp->fa_type);
1266		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1267		/*
1268		 * XXX
1269		 *
1270		 * The duplicate information returned in fa_type and fa_mode
1271		 * is an ambiguity in the NFS version 2 protocol.
1272		 *
1273		 * VREG should be taken literally as a regular file.  If a
1274		 * server intents to return some type information differently
1275		 * in the upper bits of the mode field (e.g. for sockets, or
1276		 * FIFOs), NFSv2 mandates fa_type to be VNON.  Anyway, we
1277		 * leave the examination of the mode bits even in the VREG
1278		 * case to avoid breakage for bogus servers, but we make sure
1279		 * that there are actually type bits set in the upper part of
1280		 * fa_mode (and failing that, trust the va_type field).
1281		 *
1282		 * NFSv3 cleared the issue, and requires fa_mode to not
1283		 * contain any type information (while also introduing sockets
1284		 * and FIFOs for fa_type).
1285		 */
1286		if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
1287			vtyp = IFTOVT(vmode);
1288		rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1289		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1290
1291		/*
1292		 * Really ugly NFSv2 kludge.
1293		 */
1294		if (vtyp == VCHR && rdev == 0xffffffff)
1295			vtyp = VFIFO;
1296	}
1297
1298	/*
1299	 * If v_type == VNON it is a new node, so fill in the v_type,
1300	 * n_mtime fields. Check to see if it represents a special
1301	 * device, and if so, check for a possible alias. Once the
1302	 * correct vnode has been obtained, fill in the rest of the
1303	 * information.
1304	 */
1305	np = VTONFS(vp);
1306	if (vp->v_type != vtyp) {
1307		vp->v_type = vtyp;
1308		if (vp->v_type == VFIFO) {
1309			vp->v_op = fifo_nfsv2nodeop_p;
1310		}
1311		if (vp->v_type == VCHR || vp->v_type == VBLK) {
1312			vp->v_op = spec_nfsv2nodeop_p;
1313			addaliasu(vp, rdev);
1314		}
1315		np->n_mtime = mtime.tv_sec;
1316	}
1317	vap = &np->n_vattr;
1318	vap->va_type = vtyp;
1319	vap->va_mode = (vmode & 07777);
1320	vap->va_rdev = rdev;
1321	vap->va_mtime = mtime;
1322	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1323	if (v3) {
1324		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1325		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1326		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1327		vap->va_size = fxdr_hyper(&fp->fa3_size);
1328		vap->va_blocksize = NFS_FABLKSIZE;
1329		vap->va_bytes = fxdr_hyper(&fp->fa3_used);
1330		vap->va_fileid = fxdr_unsigned(int32_t,
1331		    fp->fa3_fileid.nfsuquad[1]);
1332		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1333		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1334		vap->va_flags = 0;
1335		vap->va_filerev = 0;
1336	} else {
1337		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1338		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1339		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1340		vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1341		vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
1342		vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
1343		    * NFS_FABLKSIZE;
1344		vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1345		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1346		vap->va_flags = 0;
1347		vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1348		    fp->fa2_ctime.nfsv2_sec);
1349		vap->va_ctime.tv_nsec = 0;
1350		vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1351		vap->va_filerev = 0;
1352	}
1353	if (vap->va_size != np->n_size) {
1354		if (vap->va_type == VREG) {
1355			if (np->n_flag & NMODIFIED) {
1356				if (vap->va_size < np->n_size)
1357					vap->va_size = np->n_size;
1358				else
1359					np->n_size = vap->va_size;
1360			} else
1361				np->n_size = vap->va_size;
1362			vnode_pager_setsize(vp, np->n_size);
1363		} else
1364			np->n_size = vap->va_size;
1365	}
1366	np->n_attrstamp = time_second;
1367	if (vaper != NULL) {
1368		bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
1369		if (np->n_flag & NCHG) {
1370			if (np->n_flag & NACC)
1371				vaper->va_atime = np->n_atim;
1372			if (np->n_flag & NUPD)
1373				vaper->va_mtime = np->n_mtim;
1374		}
1375	}
1376	return (0);
1377}
1378
1379#ifdef NFS_ACDEBUG
1380#include <sys/sysctl.h>
1381SYSCTL_DECL(_vfs_nfs);
1382static int nfs_acdebug;
1383SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
1384#endif
1385
1386/*
1387 * Check the time stamp
1388 * If the cache is valid, copy contents to *vap and return 0
1389 * otherwise return an error
1390 */
1391int
1392nfs_getattrcache(vp, vaper)
1393	register struct vnode *vp;
1394	struct vattr *vaper;
1395{
1396	register struct nfsnode *np;
1397	register struct vattr *vap;
1398	struct nfsmount *nmp;
1399	int timeo;
1400
1401	np = VTONFS(vp);
1402	vap = &np->n_vattr;
1403	nmp = VFSTONFS(vp->v_mount);
1404	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
1405	timeo = (time_second - np->n_mtime) / 10;
1406
1407#ifdef NFS_ACDEBUG
1408	if (nfs_acdebug>1)
1409		printf("nfs_getattrcache: initial timeo = %d\n", timeo);
1410#endif
1411
1412	if (vap->va_type == VDIR) {
1413		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
1414			timeo = nmp->nm_acdirmin;
1415		else if (timeo > nmp->nm_acdirmax)
1416			timeo = nmp->nm_acdirmax;
1417	} else {
1418		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
1419			timeo = nmp->nm_acregmin;
1420		else if (timeo > nmp->nm_acregmax)
1421			timeo = nmp->nm_acregmax;
1422	}
1423
1424#ifdef NFS_ACDEBUG
1425	if (nfs_acdebug > 2)
1426		printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
1427			nmp->nm_acregmin, nmp->nm_acregmax,
1428			nmp->nm_acdirmin, nmp->nm_acdirmax);
1429
1430	if (nfs_acdebug)
1431		printf("nfs_getattrcache: age = %d; final timeo = %d\n",
1432			(time_second - np->n_attrstamp), timeo);
1433#endif
1434
1435	if ((time_second - np->n_attrstamp) >= timeo) {
1436		nfsstats.attrcache_misses++;
1437		return (ENOENT);
1438	}
1439	nfsstats.attrcache_hits++;
1440	if (vap->va_size != np->n_size) {
1441		if (vap->va_type == VREG) {
1442			if (np->n_flag & NMODIFIED) {
1443				if (vap->va_size < np->n_size)
1444					vap->va_size = np->n_size;
1445				else
1446					np->n_size = vap->va_size;
1447			} else
1448				np->n_size = vap->va_size;
1449			vnode_pager_setsize(vp, np->n_size);
1450		} else
1451			np->n_size = vap->va_size;
1452	}
1453	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
1454	if (np->n_flag & NCHG) {
1455		if (np->n_flag & NACC)
1456			vaper->va_atime = np->n_atim;
1457		if (np->n_flag & NUPD)
1458			vaper->va_mtime = np->n_mtim;
1459	}
1460	return (0);
1461}
1462
1463#ifndef NFS_NOSERVER
1464/*
1465 * Set up nameidata for a lookup() call and do it.
1466 *
1467 * If pubflag is set, this call is done for a lookup operation on the
1468 * public filehandle. In that case we allow crossing mountpoints and
1469 * absolute pathnames. However, the caller is expected to check that
1470 * the lookup result is within the public fs, and deny access if
1471 * it is not.
1472 *
1473 * nfs_namei() clears out garbage fields that namei() might leave garbage.
1474 * This is mainly ni_vp and ni_dvp when an error occurs, and ni_dvp when no
1475 * error occurs but the parent was not requested.
1476 *
1477 * dirp may be set whether an error is returned or not, and must be
1478 * released by the caller.
1479 */
1480int
1481nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
1482	register struct nameidata *ndp;
1483	fhandle_t *fhp;
1484	int len;
1485	struct nfssvc_sock *slp;
1486	struct sockaddr *nam;
1487	struct mbuf **mdp;
1488	caddr_t *dposp;
1489	struct vnode **retdirp;
1490	struct proc *p;
1491	int kerbflag, pubflag;
1492{
1493	register int i, rem;
1494	register struct mbuf *md;
1495	register char *fromcp, *tocp, *cp;
1496	struct iovec aiov;
1497	struct uio auio;
1498	struct vnode *dp;
1499	int error, rdonly, linklen;
1500	struct componentname *cnp = &ndp->ni_cnd;
1501
1502	*retdirp = (struct vnode *)0;
1503	cnp->cn_pnbuf = zalloc(namei_zone);
1504
1505	/*
1506	 * Copy the name from the mbuf list to ndp->ni_pnbuf
1507	 * and set the various ndp fields appropriately.
1508	 */
1509	fromcp = *dposp;
1510	tocp = cnp->cn_pnbuf;
1511	md = *mdp;
1512	rem = mtod(md, caddr_t) + md->m_len - fromcp;
1513	cnp->cn_hash = 0;
1514	for (i = 0; i < len; i++) {
1515		while (rem == 0) {
1516			md = md->m_next;
1517			if (md == NULL) {
1518				error = EBADRPC;
1519				goto out;
1520			}
1521			fromcp = mtod(md, caddr_t);
1522			rem = md->m_len;
1523		}
1524		if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1525			error = EACCES;
1526			goto out;
1527		}
1528		cnp->cn_hash += (unsigned char)*fromcp;
1529		*tocp++ = *fromcp++;
1530		rem--;
1531	}
1532	*tocp = '\0';
1533	*mdp = md;
1534	*dposp = fromcp;
1535	len = nfsm_rndup(len)-len;
1536	if (len > 0) {
1537		if (rem >= len)
1538			*dposp += len;
1539		else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1540			goto out;
1541	}
1542
1543	/*
1544	 * Extract and set starting directory.
1545	 */
1546	error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
1547	    nam, &rdonly, kerbflag, pubflag);
1548	if (error)
1549		goto out;
1550	if (dp->v_type != VDIR) {
1551		vrele(dp);
1552		error = ENOTDIR;
1553		goto out;
1554	}
1555
1556	if (rdonly)
1557		cnp->cn_flags |= RDONLY;
1558
1559	/*
1560	 * Set return directory.  Reference to dp is implicitly transfered
1561	 * to the returned pointer
1562	 */
1563	*retdirp = dp;
1564
1565	if (pubflag) {
1566		/*
1567		 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1568		 * and the 'native path' indicator.
1569		 */
1570		cp = zalloc(namei_zone);
1571		fromcp = cnp->cn_pnbuf;
1572		tocp = cp;
1573		if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1574			switch ((unsigned char)*fromcp) {
1575			case WEBNFS_NATIVE_CHAR:
1576				/*
1577				 * 'Native' path for us is the same
1578				 * as a path according to the NFS spec,
1579				 * just skip the escape char.
1580				 */
1581				fromcp++;
1582				break;
1583			/*
1584			 * More may be added in the future, range 0x80-0xff
1585			 */
1586			default:
1587				error = EIO;
1588				zfree(namei_zone, cp);
1589				goto out;
1590			}
1591		}
1592		/*
1593		 * Translate the '%' escapes, URL-style.
1594		 */
1595		while (*fromcp != '\0') {
1596			if (*fromcp == WEBNFS_ESC_CHAR) {
1597				if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1598					fromcp++;
1599					*tocp++ = HEXSTRTOI(fromcp);
1600					fromcp += 2;
1601					continue;
1602				} else {
1603					error = ENOENT;
1604					zfree(namei_zone, cp);
1605					goto out;
1606				}
1607			} else
1608				*tocp++ = *fromcp++;
1609		}
1610		*tocp = '\0';
1611		zfree(namei_zone, cnp->cn_pnbuf);
1612		cnp->cn_pnbuf = cp;
1613	}
1614
1615	ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
1616	ndp->ni_segflg = UIO_SYSSPACE;
1617
1618	if (pubflag) {
1619		ndp->ni_rootdir = rootvnode;
1620		ndp->ni_loopcnt = 0;
1621		if (cnp->cn_pnbuf[0] == '/')
1622			dp = rootvnode;
1623	} else {
1624		cnp->cn_flags |= NOCROSSMOUNT;
1625	}
1626
1627	/*
1628	 * Initialize for scan, set ni_startdir and bump ref on dp again
1629	 * becuase lookup() will dereference ni_startdir.
1630	 */
1631
1632	cnp->cn_proc = p;
1633	VREF(dp);
1634	ndp->ni_startdir = dp;
1635
1636	for (;;) {
1637		cnp->cn_nameptr = cnp->cn_pnbuf;
1638		/*
1639		 * Call lookup() to do the real work.  If an error occurs,
1640		 * ndp->ni_vp and ni_dvp are left uninitialized or NULL and
1641		 * we do not have to dereference anything before returning.
1642		 * In either case ni_startdir will be dereferenced and NULLed
1643		 * out.
1644		 */
1645		error = lookup(ndp);
1646		if (error)
1647			break;
1648
1649		/*
1650		 * Check for encountering a symbolic link.  Trivial
1651		 * termination occurs if no symlink encountered.
1652		 * Note: zfree is safe because error is 0, so we will
1653		 * not zfree it again when we break.
1654		 */
1655		if ((cnp->cn_flags & ISSYMLINK) == 0) {
1656			nfsrv_object_create(ndp->ni_vp);
1657			if (cnp->cn_flags & (SAVENAME | SAVESTART))
1658				cnp->cn_flags |= HASBUF;
1659			else
1660				zfree(namei_zone, cnp->cn_pnbuf);
1661			break;
1662		}
1663
1664		/*
1665		 * Validate symlink
1666		 */
1667		if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
1668			VOP_UNLOCK(ndp->ni_dvp, 0, p);
1669		if (!pubflag) {
1670			error = EINVAL;
1671			goto badlink2;
1672		}
1673
1674		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
1675			error = ELOOP;
1676			goto badlink2;
1677		}
1678		if (ndp->ni_pathlen > 1)
1679			cp = zalloc(namei_zone);
1680		else
1681			cp = cnp->cn_pnbuf;
1682		aiov.iov_base = cp;
1683		aiov.iov_len = MAXPATHLEN;
1684		auio.uio_iov = &aiov;
1685		auio.uio_iovcnt = 1;
1686		auio.uio_offset = 0;
1687		auio.uio_rw = UIO_READ;
1688		auio.uio_segflg = UIO_SYSSPACE;
1689		auio.uio_procp = (struct proc *)0;
1690		auio.uio_resid = MAXPATHLEN;
1691		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
1692		if (error) {
1693		badlink1:
1694			if (ndp->ni_pathlen > 1)
1695				zfree(namei_zone, cp);
1696		badlink2:
1697			vrele(ndp->ni_dvp);
1698			vput(ndp->ni_vp);
1699			break;
1700		}
1701		linklen = MAXPATHLEN - auio.uio_resid;
1702		if (linklen == 0) {
1703			error = ENOENT;
1704			goto badlink1;
1705		}
1706		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
1707			error = ENAMETOOLONG;
1708			goto badlink1;
1709		}
1710
1711		/*
1712		 * Adjust or replace path
1713		 */
1714		if (ndp->ni_pathlen > 1) {
1715			bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
1716			zfree(namei_zone, cnp->cn_pnbuf);
1717			cnp->cn_pnbuf = cp;
1718		} else
1719			cnp->cn_pnbuf[linklen] = '\0';
1720		ndp->ni_pathlen += linklen;
1721
1722		/*
1723		 * Cleanup refs for next loop and check if root directory
1724		 * should replace current directory.  Normally ni_dvp
1725		 * becomes the new base directory and is cleaned up when
1726		 * we loop.  Explicitly null pointers after invalidation
1727		 * to clarify operation.
1728		 */
1729		vput(ndp->ni_vp);
1730		ndp->ni_vp = NULL;
1731
1732		if (cnp->cn_pnbuf[0] == '/') {
1733			vrele(ndp->ni_dvp);
1734			ndp->ni_dvp = ndp->ni_rootdir;
1735			VREF(ndp->ni_dvp);
1736		}
1737		ndp->ni_startdir = ndp->ni_dvp;
1738		ndp->ni_dvp = NULL;
1739	}
1740
1741	/*
1742	 * nfs_namei() guarentees that fields will not contain garbage
1743	 * whether an error occurs or not.  This allows the caller to track
1744	 * cleanup state trivially.
1745	 */
1746out:
1747	if (error) {
1748		zfree(namei_zone, cnp->cn_pnbuf);
1749		ndp->ni_vp = NULL;
1750		ndp->ni_dvp = NULL;
1751		ndp->ni_startdir = NULL;
1752		cnp->cn_flags &= ~HASBUF;
1753	} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
1754		ndp->ni_dvp = NULL;
1755	}
1756	return (error);
1757}
1758
1759/*
1760 * A fiddled version of m_adj() that ensures null fill to a long
1761 * boundary and only trims off the back end
1762 */
1763void
1764nfsm_adj(mp, len, nul)
1765	struct mbuf *mp;
1766	register int len;
1767	int nul;
1768{
1769	register struct mbuf *m;
1770	register int count, i;
1771	register char *cp;
1772
1773	/*
1774	 * Trim from tail.  Scan the mbuf chain,
1775	 * calculating its length and finding the last mbuf.
1776	 * If the adjustment only affects this mbuf, then just
1777	 * adjust and return.  Otherwise, rescan and truncate
1778	 * after the remaining size.
1779	 */
1780	count = 0;
1781	m = mp;
1782	for (;;) {
1783		count += m->m_len;
1784		if (m->m_next == (struct mbuf *)0)
1785			break;
1786		m = m->m_next;
1787	}
1788	if (m->m_len > len) {
1789		m->m_len -= len;
1790		if (nul > 0) {
1791			cp = mtod(m, caddr_t)+m->m_len-nul;
1792			for (i = 0; i < nul; i++)
1793				*cp++ = '\0';
1794		}
1795		return;
1796	}
1797	count -= len;
1798	if (count < 0)
1799		count = 0;
1800	/*
1801	 * Correct length for chain is "count".
1802	 * Find the mbuf with last data, adjust its length,
1803	 * and toss data from remaining mbufs on chain.
1804	 */
1805	for (m = mp; m; m = m->m_next) {
1806		if (m->m_len >= count) {
1807			m->m_len = count;
1808			if (nul > 0) {
1809				cp = mtod(m, caddr_t)+m->m_len-nul;
1810				for (i = 0; i < nul; i++)
1811					*cp++ = '\0';
1812			}
1813			break;
1814		}
1815		count -= m->m_len;
1816	}
1817	for (m = m->m_next;m;m = m->m_next)
1818		m->m_len = 0;
1819}
1820
1821/*
1822 * Make these functions instead of macros, so that the kernel text size
1823 * doesn't get too big...
1824 */
1825void
1826nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
1827	struct nfsrv_descript *nfsd;
1828	int before_ret;
1829	register struct vattr *before_vap;
1830	int after_ret;
1831	struct vattr *after_vap;
1832	struct mbuf **mbp;
1833	char **bposp;
1834{
1835	register struct mbuf *mb = *mbp, *mb2;
1836	register char *bpos = *bposp;
1837	register u_int32_t *tl;
1838
1839	if (before_ret) {
1840		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1841		*tl = nfs_false;
1842	} else {
1843		nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
1844		*tl++ = nfs_true;
1845		txdr_hyper(before_vap->va_size, tl);
1846		tl += 2;
1847		txdr_nfsv3time(&(before_vap->va_mtime), tl);
1848		tl += 2;
1849		txdr_nfsv3time(&(before_vap->va_ctime), tl);
1850	}
1851	*bposp = bpos;
1852	*mbp = mb;
1853	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
1854}
1855
1856void
1857nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
1858	struct nfsrv_descript *nfsd;
1859	int after_ret;
1860	struct vattr *after_vap;
1861	struct mbuf **mbp;
1862	char **bposp;
1863{
1864	register struct mbuf *mb = *mbp, *mb2;
1865	register char *bpos = *bposp;
1866	register u_int32_t *tl;
1867	register struct nfs_fattr *fp;
1868
1869	if (after_ret) {
1870		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1871		*tl = nfs_false;
1872	} else {
1873		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
1874		*tl++ = nfs_true;
1875		fp = (struct nfs_fattr *)tl;
1876		nfsm_srvfattr(nfsd, after_vap, fp);
1877	}
1878	*mbp = mb;
1879	*bposp = bpos;
1880}
1881
1882void
1883nfsm_srvfattr(nfsd, vap, fp)
1884	register struct nfsrv_descript *nfsd;
1885	register struct vattr *vap;
1886	register struct nfs_fattr *fp;
1887{
1888
1889	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
1890	fp->fa_uid = txdr_unsigned(vap->va_uid);
1891	fp->fa_gid = txdr_unsigned(vap->va_gid);
1892	if (nfsd->nd_flag & ND_NFSV3) {
1893		fp->fa_type = vtonfsv3_type(vap->va_type);
1894		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
1895		txdr_hyper(vap->va_size, &fp->fa3_size);
1896		txdr_hyper(vap->va_bytes, &fp->fa3_used);
1897		fp->fa3_rdev.specdata1 = txdr_unsigned(umajor(vap->va_rdev));
1898		fp->fa3_rdev.specdata2 = txdr_unsigned(uminor(vap->va_rdev));
1899		fp->fa3_fsid.nfsuquad[0] = 0;
1900		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
1901		fp->fa3_fileid.nfsuquad[0] = 0;
1902		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
1903		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
1904		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
1905		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
1906	} else {
1907		fp->fa_type = vtonfsv2_type(vap->va_type);
1908		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
1909		fp->fa2_size = txdr_unsigned(vap->va_size);
1910		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
1911		if (vap->va_type == VFIFO)
1912			fp->fa2_rdev = 0xffffffff;
1913		else
1914			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
1915		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
1916		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
1917		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
1918		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
1919		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
1920		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
1921	}
1922}
1923
1924/*
1925 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
1926 * 	- look up fsid in mount list (if not found ret error)
1927 *	- get vp and export rights by calling VFS_FHTOVP()
1928 *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
1929 *	- if not lockflag unlock it with VOP_UNLOCK()
1930 */
1931int
1932nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag)
1933	fhandle_t *fhp;
1934	int lockflag;
1935	struct vnode **vpp;
1936	struct ucred *cred;
1937	struct nfssvc_sock *slp;
1938	struct sockaddr *nam;
1939	int *rdonlyp;
1940	int kerbflag;
1941	int pubflag;
1942{
1943	struct proc *p = curproc; /* XXX */
1944	register struct mount *mp;
1945	register int i;
1946	struct ucred *credanon;
1947	int error, exflags;
1948#ifdef MNT_EXNORESPORT		/* XXX needs mountd and /etc/exports help yet */
1949	struct sockaddr_int *saddr;
1950#endif
1951
1952	*vpp = (struct vnode *)0;
1953
1954	if (nfs_ispublicfh(fhp)) {
1955		if (!pubflag || !nfs_pub.np_valid)
1956			return (ESTALE);
1957		fhp = &nfs_pub.np_handle;
1958	}
1959
1960	mp = vfs_getvfs(&fhp->fh_fsid);
1961	if (!mp)
1962		return (ESTALE);
1963	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
1964	if (error)
1965		return (error);
1966	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
1967	if (error)
1968		return (error);
1969#ifdef MNT_EXNORESPORT
1970	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
1971		saddr = (struct sockaddr_in *)nam;
1972		if (saddr->sin_family == AF_INET &&
1973		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
1974			vput(*vpp);
1975			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1976		}
1977	}
1978#endif
1979	/*
1980	 * Check/setup credentials.
1981	 */
1982	if (exflags & MNT_EXKERB) {
1983		if (!kerbflag) {
1984			vput(*vpp);
1985			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1986		}
1987	} else if (kerbflag) {
1988		vput(*vpp);
1989		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1990	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1991		cred->cr_uid = credanon->cr_uid;
1992		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
1993			cred->cr_groups[i] = credanon->cr_groups[i];
1994		cred->cr_ngroups = i;
1995	}
1996	if (exflags & MNT_EXRDONLY)
1997		*rdonlyp = 1;
1998	else
1999		*rdonlyp = 0;
2000
2001	nfsrv_object_create(*vpp);
2002
2003	if (!lockflag)
2004		VOP_UNLOCK(*vpp, 0, p);
2005	return (0);
2006}
2007
2008
2009/*
2010 * WebNFS: check if a filehandle is a public filehandle. For v3, this
2011 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
2012 * transformed this to all zeroes in both cases, so check for it.
2013 */
2014int
2015nfs_ispublicfh(fhp)
2016	fhandle_t *fhp;
2017{
2018	char *cp = (char *)fhp;
2019	int i;
2020
2021	for (i = 0; i < NFSX_V3FH; i++)
2022		if (*cp++ != 0)
2023			return (FALSE);
2024	return (TRUE);
2025}
2026
2027#endif /* NFS_NOSERVER */
2028/*
2029 * This function compares two net addresses by family and returns TRUE
2030 * if they are the same host.
2031 * If there is any doubt, return FALSE.
2032 * The AF_INET family is handled as a special case so that address mbufs
2033 * don't need to be saved to store "struct in_addr", which is only 4 bytes.
2034 */
2035int
2036netaddr_match(family, haddr, nam)
2037	int family;
2038	union nethostaddr *haddr;
2039	struct sockaddr *nam;
2040{
2041	register struct sockaddr_in *inetaddr;
2042
2043	switch (family) {
2044	case AF_INET:
2045		inetaddr = (struct sockaddr_in *)nam;
2046		if (inetaddr->sin_family == AF_INET &&
2047		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
2048			return (1);
2049		break;
2050#ifdef ISO
2051	case AF_ISO:
2052	    {
2053		register struct sockaddr_iso *isoaddr1, *isoaddr2;
2054
2055		isoaddr1 = (struct sockaddr_iso *)nam;
2056		isoaddr2 = (struct sockaddr_iso *)haddr->had_nam;
2057		if (isoaddr1->siso_family == AF_ISO &&
2058		    isoaddr1->siso_nlen > 0 &&
2059		    isoaddr1->siso_nlen == isoaddr2->siso_nlen &&
2060		    SAME_ISOADDR(isoaddr1, isoaddr2))
2061			return (1);
2062		break;
2063	    }
2064#endif	/* ISO */
2065	default:
2066		break;
2067	};
2068	return (0);
2069}
2070
2071static nfsuint64 nfs_nullcookie = { { 0, 0 } };
2072/*
2073 * This function finds the directory cookie that corresponds to the
2074 * logical byte offset given.
2075 */
2076nfsuint64 *
2077nfs_getcookie(np, off, add)
2078	register struct nfsnode *np;
2079	off_t off;
2080	int add;
2081{
2082	register struct nfsdmap *dp, *dp2;
2083	register int pos;
2084
2085	pos = (uoff_t)off / NFS_DIRBLKSIZ;
2086	if (pos == 0 || off < 0) {
2087#ifdef DIAGNOSTIC
2088		if (add)
2089			panic("nfs getcookie add at <= 0");
2090#endif
2091		return (&nfs_nullcookie);
2092	}
2093	pos--;
2094	dp = np->n_cookies.lh_first;
2095	if (!dp) {
2096		if (add) {
2097			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
2098				M_NFSDIROFF, M_WAITOK);
2099			dp->ndm_eocookie = 0;
2100			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
2101		} else
2102			return ((nfsuint64 *)0);
2103	}
2104	while (pos >= NFSNUMCOOKIES) {
2105		pos -= NFSNUMCOOKIES;
2106		if (dp->ndm_list.le_next) {
2107			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
2108				pos >= dp->ndm_eocookie)
2109				return ((nfsuint64 *)0);
2110			dp = dp->ndm_list.le_next;
2111		} else if (add) {
2112			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
2113				M_NFSDIROFF, M_WAITOK);
2114			dp2->ndm_eocookie = 0;
2115			LIST_INSERT_AFTER(dp, dp2, ndm_list);
2116			dp = dp2;
2117		} else
2118			return ((nfsuint64 *)0);
2119	}
2120	if (pos >= dp->ndm_eocookie) {
2121		if (add)
2122			dp->ndm_eocookie = pos + 1;
2123		else
2124			return ((nfsuint64 *)0);
2125	}
2126	return (&dp->ndm_cookies[pos]);
2127}
2128
2129/*
2130 * Invalidate cached directory information, except for the actual directory
2131 * blocks (which are invalidated separately).
2132 * Done mainly to avoid the use of stale offset cookies.
2133 */
2134void
2135nfs_invaldir(vp)
2136	register struct vnode *vp;
2137{
2138	register struct nfsnode *np = VTONFS(vp);
2139
2140#ifdef DIAGNOSTIC
2141	if (vp->v_type != VDIR)
2142		panic("nfs: invaldir not dir");
2143#endif
2144	np->n_direofoffset = 0;
2145	np->n_cookieverf.nfsuquad[0] = 0;
2146	np->n_cookieverf.nfsuquad[1] = 0;
2147	if (np->n_cookies.lh_first)
2148		np->n_cookies.lh_first->ndm_eocookie = 0;
2149}
2150
2151/*
2152 * The write verifier has changed (probably due to a server reboot), so all
2153 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
2154 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2155 * flag. Once done the new write verifier can be set for the mount point.
2156 */
2157void
2158nfs_clearcommit(mp)
2159	struct mount *mp;
2160{
2161	register struct vnode *vp, *nvp;
2162	register struct buf *bp, *nbp;
2163	int s;
2164
2165	s = splbio();
2166loop:
2167	for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) {
2168		if (vp->v_mount != mp)	/* Paranoia */
2169			goto loop;
2170		nvp = vp->v_mntvnodes.le_next;
2171		for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
2172			nbp = TAILQ_NEXT(bp, b_vnbufs);
2173			if (BUF_REFCNT(bp) == 0 &&
2174			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
2175				== (B_DELWRI | B_NEEDCOMMIT))
2176				bp->b_flags &= ~B_NEEDCOMMIT;
2177		}
2178	}
2179	splx(s);
2180}
2181
2182#ifndef NFS_NOSERVER
2183/*
2184 * Map errnos to NFS error numbers. For Version 3 also filter out error
2185 * numbers not specified for the associated procedure.
2186 */
2187int
2188nfsrv_errmap(nd, err)
2189	struct nfsrv_descript *nd;
2190	register int err;
2191{
2192	register short *defaulterrp, *errp;
2193
2194	if (nd->nd_flag & ND_NFSV3) {
2195	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
2196		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2197		while (*++errp) {
2198			if (*errp == err)
2199				return (err);
2200			else if (*errp > err)
2201				break;
2202		}
2203		return ((int)*defaulterrp);
2204	    } else
2205		return (err & 0xffff);
2206	}
2207	if (err <= ELAST)
2208		return ((int)nfsrv_v2errmap[err - 1]);
2209	return (NFSERR_IO);
2210}
2211
2212int
2213nfsrv_object_create(vp)
2214	struct vnode *vp;
2215{
2216
2217	if (vp == NULL || vp->v_type != VREG)
2218		return (1);
2219	return (vfs_object_create(vp, curproc,
2220				  curproc ? curproc->p_ucred : NULL));
2221}
2222
2223/*
2224 * Sort the group list in increasing numerical order.
2225 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2226 *  that used to be here.)
2227 */
2228void
2229nfsrvw_sort(list, num)
2230        register gid_t *list;
2231        register int num;
2232{
2233	register int i, j;
2234	gid_t v;
2235
2236	/* Insertion sort. */
2237	for (i = 1; i < num; i++) {
2238		v = list[i];
2239		/* find correct slot for value v, moving others up */
2240		for (j = i; --j >= 0 && v < list[j];)
2241			list[j + 1] = list[j];
2242		list[j + 1] = v;
2243	}
2244}
2245
2246/*
2247 * copy credentials making sure that the result can be compared with bcmp().
2248 */
2249void
2250nfsrv_setcred(incred, outcred)
2251	register struct ucred *incred, *outcred;
2252{
2253	register int i;
2254
2255	bzero((caddr_t)outcred, sizeof (struct ucred));
2256	outcred->cr_ref = 1;
2257	outcred->cr_uid = incred->cr_uid;
2258	outcred->cr_ngroups = incred->cr_ngroups;
2259	for (i = 0; i < incred->cr_ngroups; i++)
2260		outcred->cr_groups[i] = incred->cr_groups[i];
2261	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2262}
2263#endif /* NFS_NOSERVER */
2264