nfs_subs.c revision 60938
167754Smsmith/*
267754Smsmith * Copyright (c) 1989, 1993
367754Smsmith *	The Regents of the University of California.  All rights reserved.
4117521Snjl *
567754Smsmith * This code is derived from software contributed to Berkeley by
667754Smsmith * Rick Macklem at The University of Guelph.
767754Smsmith *
867754Smsmith * Redistribution and use in source and binary forms, with or without
967754Smsmith * modification, are permitted provided that the following conditions
1067754Smsmith * are met:
1167754Smsmith * 1. Redistributions of source code must retain the above copyright
12114237Snjl *    notice, this list of conditions and the following disclaimer.
1370243Smsmith * 2. Redistributions in binary form must reproduce the above copyright
1467754Smsmith *    notice, this list of conditions and the following disclaimer in the
1567754Smsmith *    documentation and/or other materials provided with the distribution.
1667754Smsmith * 3. All advertising materials mentioning features or use of this software
1767754Smsmith *    must display the following acknowledgement:
1867754Smsmith *	This product includes software developed by the University of
1967754Smsmith *	California, Berkeley and its contributors.
2067754Smsmith * 4. Neither the name of the University nor the names of its contributors
2167754Smsmith *    may be used to endorse or promote products derived from this software
2267754Smsmith *    without specific prior written permission.
2367754Smsmith *
2467754Smsmith * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
2567754Smsmith * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
2667754Smsmith * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
2767754Smsmith * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
2867754Smsmith * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
2967754Smsmith * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
3067754Smsmith * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
3167754Smsmith * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
3267754Smsmith * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
3367754Smsmith * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
3467754Smsmith * SUCH DAMAGE.
3567754Smsmith *
3667754Smsmith *	@(#)nfs_subs.c  8.8 (Berkeley) 5/22/95
3767754Smsmith * $FreeBSD: head/sys/nfsclient/nfs_subs.c 60938 2000-05-26 02:09:24Z jake $
3867754Smsmith */
3967754Smsmith
4067754Smsmith/*
4167754Smsmith * These functions support the macros and help fiddle mbuf chains for
4267754Smsmith * the nfs op functions. They do things like create the rpc header and
4367754Smsmith * copy data between mbuf chains and uio lists.
4467754Smsmith */
4567754Smsmith#include <sys/param.h>
4667754Smsmith#include <sys/systm.h>
4767754Smsmith#include <sys/kernel.h>
4867754Smsmith#include <sys/bio.h>
4967754Smsmith#include <sys/buf.h>
5067754Smsmith#include <sys/proc.h>
5167754Smsmith#include <sys/mount.h>
5267754Smsmith#include <sys/vnode.h>
5367754Smsmith#include <sys/namei.h>
5467754Smsmith#include <sys/mbuf.h>
5567754Smsmith#include <sys/socket.h>
5667754Smsmith#include <sys/stat.h>
5767754Smsmith#include <sys/malloc.h>
5867754Smsmith#include <sys/sysent.h>
5967754Smsmith#include <sys/syscall.h>
6067754Smsmith
6167754Smsmith#include <vm/vm.h>
6267754Smsmith#include <vm/vm_object.h>
6367754Smsmith#include <vm/vm_extern.h>
6467754Smsmith#include <vm/vm_zone.h>
6567754Smsmith
6667754Smsmith#include <nfs/rpcv2.h>
6767754Smsmith#include <nfs/nfsproto.h>
6867754Smsmith#include <nfs/nfs.h>
6967754Smsmith#include <nfs/nfsnode.h>
7067754Smsmith#include <nfs/xdr_subs.h>
7167754Smsmith#include <nfs/nfsm_subs.h>
7267754Smsmith#include <nfs/nfsmount.h>
7367754Smsmith#include <nfs/nqnfs.h>
7467754Smsmith#include <nfs/nfsrtt.h>
7567754Smsmith
7667754Smsmith#include <netinet/in.h>
7767754Smsmith
7867754Smsmith/*
7967754Smsmith * Data items converted to xdr at startup, since they are constant
8067754Smsmith * This is kinda hokey, but may save a little time doing byte swaps
8167754Smsmith */
8267754Smsmithu_int32_t nfs_xdrneg1;
8367754Smsmithu_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
8467754Smsmith	rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
8567754Smsmith	rpc_auth_kerb;
8667754Smsmithu_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
8767754Smsmith
8867754Smsmith/* And other global data */
8967754Smsmithstatic u_int32_t nfs_xid = 0;
9067754Smsmithstatic enum vtype nv2tov_type[8]= {
9167754Smsmith	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON,  VNON
9267754Smsmith};
9367754Smsmithenum vtype nv3tov_type[8]= {
9467754Smsmith	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
9567754Smsmith};
9667754Smsmith
9767754Smsmithint nfs_ticks;
9867754Smsmithint nfs_pbuf_freecnt = -1;	/* start out unlimited */
9967754Smsmith
10067754Smsmithstruct nfs_reqq nfs_reqq;
10167754Smsmithstruct nfssvc_sockhead nfssvc_sockhead;
10267754Smsmithint nfssvc_sockhead_flag;
10367754Smsmithstruct nfsd_head nfsd_head;
10467754Smsmithint nfsd_head_flag;
10567754Smsmithstruct nfs_bufq nfs_bufq;
10667754Smsmithstruct nqtimerhead nqtimerhead;
10767754Smsmithstruct nqfhhashhead *nqfhhashtbl;
10867754Smsmithu_long nqfhhash;
10967754Smsmith
11067754Smsmithstatic void (*nfs_prev_lease_updatetime) __P((int));
11167754Smsmithstatic int nfs_prev_nfssvc_sy_narg;
11267754Smsmithstatic sy_call_t *nfs_prev_nfssvc_sy_call;
11367754Smsmith
11467754Smsmith#ifndef NFS_NOSERVER
11567754Smsmith
11667754Smsmithstatic vop_t *nfs_prev_vop_lease_check;
11767754Smsmith
11867754Smsmith/*
11967754Smsmith * Mapping of old NFS Version 2 RPC numbers to generic numbers.
12067754Smsmith */
12167754Smsmithint nfsv3_procid[NFS_NPROCS] = {
12267754Smsmith	NFSPROC_NULL,
12367754Smsmith	NFSPROC_GETATTR,
12467754Smsmith	NFSPROC_SETATTR,
12567754Smsmith	NFSPROC_NOOP,
126102550Siwasaki	NFSPROC_LOOKUP,
12767754Smsmith	NFSPROC_READLINK,
128102550Siwasaki	NFSPROC_READ,
12991116Smsmith	NFSPROC_NOOP,
13067754Smsmith	NFSPROC_WRITE,
13167754Smsmith	NFSPROC_CREATE,
13267754Smsmith	NFSPROC_REMOVE,
13367754Smsmith	NFSPROC_RENAME,
134114237Snjl	NFSPROC_LINK,
135114237Snjl	NFSPROC_SYMLINK,
136114237Snjl	NFSPROC_MKDIR,
137114237Snjl	NFSPROC_RMDIR,
138114237Snjl	NFSPROC_READDIR,
139114237Snjl	NFSPROC_FSSTAT,
140114237Snjl	NFSPROC_NOOP,
141114237Snjl	NFSPROC_NOOP,
142114237Snjl	NFSPROC_NOOP,
143114237Snjl	NFSPROC_NOOP,
144114237Snjl	NFSPROC_NOOP,
145114237Snjl	NFSPROC_NOOP,
146114237Snjl	NFSPROC_NOOP,
147114237Snjl	NFSPROC_NOOP
148114237Snjl};
149114237Snjl
150114237Snjl#endif /* NFS_NOSERVER */
151114237Snjl/*
152114237Snjl * and the reverse mapping from generic to Version 2 procedure numbers
153114237Snjl */
154114237Snjlint nfsv2_procid[NFS_NPROCS] = {
155114237Snjl	NFSV2PROC_NULL,
156114237Snjl	NFSV2PROC_GETATTR,
157114237Snjl	NFSV2PROC_SETATTR,
158114237Snjl	NFSV2PROC_LOOKUP,
159114237Snjl	NFSV2PROC_NOOP,
160114237Snjl	NFSV2PROC_READLINK,
161114237Snjl	NFSV2PROC_READ,
162114237Snjl	NFSV2PROC_WRITE,
163114237Snjl	NFSV2PROC_CREATE,
164114237Snjl	NFSV2PROC_MKDIR,
165114237Snjl	NFSV2PROC_SYMLINK,
166114237Snjl	NFSV2PROC_CREATE,
167114237Snjl	NFSV2PROC_REMOVE,
168114237Snjl	NFSV2PROC_RMDIR,
169114237Snjl	NFSV2PROC_RENAME,
170114237Snjl	NFSV2PROC_LINK,
171114237Snjl	NFSV2PROC_READDIR,
172114237Snjl	NFSV2PROC_NOOP,
173114237Snjl	NFSV2PROC_STATFS,
17467754Smsmith	NFSV2PROC_NOOP,
17567754Smsmith	NFSV2PROC_NOOP,
17667754Smsmith	NFSV2PROC_NOOP,
17767754Smsmith	NFSV2PROC_NOOP,
17867754Smsmith	NFSV2PROC_NOOP,
17967754Smsmith	NFSV2PROC_NOOP,
18067754Smsmith	NFSV2PROC_NOOP,
18167754Smsmith};
18267754Smsmith
18367754Smsmith#ifndef NFS_NOSERVER
18467754Smsmith/*
18567754Smsmith * Maps errno values to nfs error numbers.
18667754Smsmith * Use NFSERR_IO as the catch all for ones not specifically defined in
18767754Smsmith * RFC 1094.
18867754Smsmith */
18967754Smsmithstatic u_char nfsrv_v2errmap[ELAST] = {
19067754Smsmith  NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
19167754Smsmith  NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
19291116Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
19367754Smsmith  NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
19491116Smsmith  NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
19567754Smsmith  NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
19667754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
19767754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
19867754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
19967754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
20067754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
20167754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
20267754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_NAMETOL,	NFSERR_IO,	NFSERR_IO,
20367754Smsmith  NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
20467754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
20567754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
20667754Smsmith  NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
20767754Smsmith  NFSERR_IO /* << Last is 86 */
20867754Smsmith};
20967754Smsmith
21067754Smsmith/*
21167754Smsmith * Maps errno values to nfs error numbers.
21267754Smsmith * Although it is not obvious whether or not NFS clients really care if
21367754Smsmith * a returned error value is in the specified list for the procedure, the
21467754Smsmith * safest thing to do is filter them appropriately. For Version 2, the
21567754Smsmith * X/Open XNFS document is the only specification that defines error values
21667754Smsmith * for each RPC (The RFC simply lists all possible error values for all RPCs),
21767754Smsmith * so I have decided to not do this for Version 2.
21867754Smsmith * The first entry is the default error return and the rest are the valid
21967754Smsmith * errors for that RPC in increasing numeric order.
22069746Smsmith */
22167754Smsmithstatic short nfsv3err_null[] = {
22282367Smsmith	0,
223117521Snjl	0,
224117521Snjl};
22567754Smsmith
22667754Smsmithstatic short nfsv3err_getattr[] = {
22767754Smsmith	NFSERR_IO,
22867754Smsmith	NFSERR_IO,
22967754Smsmith	NFSERR_STALE,
23067754Smsmith	NFSERR_BADHANDLE,
23167754Smsmith	NFSERR_SERVERFAULT,
23267754Smsmith	0,
23367754Smsmith};
23467754Smsmith
23567754Smsmithstatic short nfsv3err_setattr[] = {
23667754Smsmith	NFSERR_IO,
23767754Smsmith	NFSERR_PERM,
23867754Smsmith	NFSERR_IO,
23967754Smsmith	NFSERR_ACCES,
24067754Smsmith	NFSERR_INVAL,
24167754Smsmith	NFSERR_NOSPC,
24267754Smsmith	NFSERR_ROFS,
24367754Smsmith	NFSERR_DQUOT,
24467754Smsmith	NFSERR_STALE,
24567754Smsmith	NFSERR_BADHANDLE,
24667754Smsmith	NFSERR_NOT_SYNC,
24767754Smsmith	NFSERR_SERVERFAULT,
24867754Smsmith	0,
24967754Smsmith};
25067754Smsmith
25167754Smsmithstatic short nfsv3err_lookup[] = {
25267754Smsmith	NFSERR_IO,
25367754Smsmith	NFSERR_NOENT,
25467754Smsmith	NFSERR_IO,
25567754Smsmith	NFSERR_ACCES,
25667754Smsmith	NFSERR_NOTDIR,
25767754Smsmith	NFSERR_NAMETOL,
25867754Smsmith	NFSERR_STALE,
25967754Smsmith	NFSERR_BADHANDLE,
26067754Smsmith	NFSERR_SERVERFAULT,
26167754Smsmith	0,
26267754Smsmith};
26367754Smsmith
26482367Smsmithstatic short nfsv3err_access[] = {
26567754Smsmith	NFSERR_IO,
26667754Smsmith	NFSERR_IO,
26767754Smsmith	NFSERR_STALE,
26871867Smsmith	NFSERR_BADHANDLE,
26982367Smsmith	NFSERR_SERVERFAULT,
27091116Smsmith	0,
27191116Smsmith};
27291116Smsmith
27367754Smsmithstatic short nfsv3err_readlink[] = {
27467754Smsmith	NFSERR_IO,
27567754Smsmith	NFSERR_IO,
27667754Smsmith	NFSERR_ACCES,
27767754Smsmith	NFSERR_INVAL,
27867754Smsmith	NFSERR_STALE,
27967754Smsmith	NFSERR_BADHANDLE,
28067754Smsmith	NFSERR_NOTSUPP,
28167754Smsmith	NFSERR_SERVERFAULT,
28267754Smsmith	0,
28367754Smsmith};
28467754Smsmith
28567754Smsmithstatic short nfsv3err_read[] = {
28667754Smsmith	NFSERR_IO,
28767754Smsmith	NFSERR_IO,
28867754Smsmith	NFSERR_NXIO,
28999146Siwasaki	NFSERR_ACCES,
290114237Snjl	NFSERR_INVAL,
291114237Snjl	NFSERR_STALE,
292117521Snjl	NFSERR_BADHANDLE,
293117521Snjl	NFSERR_SERVERFAULT,
294114237Snjl	0,
295114237Snjl};
296114237Snjl
297114237Snjlstatic short nfsv3err_write[] = {
298114237Snjl	NFSERR_IO,
29967754Smsmith	NFSERR_IO,
30067754Smsmith	NFSERR_ACCES,
30167754Smsmith	NFSERR_INVAL,
30267754Smsmith	NFSERR_FBIG,
30367754Smsmith	NFSERR_NOSPC,
304117521Snjl	NFSERR_ROFS,
305117521Snjl	NFSERR_DQUOT,
30667754Smsmith	NFSERR_STALE,
30767754Smsmith	NFSERR_BADHANDLE,
30867754Smsmith	NFSERR_SERVERFAULT,
30967754Smsmith	0,
31067754Smsmith};
31167754Smsmith
31267754Smsmithstatic short nfsv3err_create[] = {
31367754Smsmith	NFSERR_IO,
314107325Siwasaki	NFSERR_IO,
31582367Smsmith	NFSERR_ACCES,
31682367Smsmith	NFSERR_EXIST,
31767754Smsmith	NFSERR_NOTDIR,
31867754Smsmith	NFSERR_NOSPC,
31982367Smsmith	NFSERR_ROFS,
32067754Smsmith	NFSERR_NAMETOL,
32182367Smsmith	NFSERR_DQUOT,
32267754Smsmith	NFSERR_STALE,
32367754Smsmith	NFSERR_BADHANDLE,
32467754Smsmith	NFSERR_NOTSUPP,
32582367Smsmith	NFSERR_SERVERFAULT,
32667754Smsmith	0,
32782367Smsmith};
32867754Smsmith
32967754Smsmithstatic short nfsv3err_mkdir[] = {
33067754Smsmith	NFSERR_IO,
33182367Smsmith	NFSERR_IO,
33267754Smsmith	NFSERR_ACCES,
33367754Smsmith	NFSERR_EXIST,
33469746Smsmith	NFSERR_NOTDIR,
33567754Smsmith	NFSERR_NOSPC,
33667754Smsmith	NFSERR_ROFS,
33767754Smsmith	NFSERR_NAMETOL,
33867754Smsmith	NFSERR_DQUOT,
33967754Smsmith	NFSERR_STALE,
34067754Smsmith	NFSERR_BADHANDLE,
34167754Smsmith	NFSERR_NOTSUPP,
34267754Smsmith	NFSERR_SERVERFAULT,
34367754Smsmith	0,
34467754Smsmith};
34567754Smsmith
34667754Smsmithstatic short nfsv3err_symlink[] = {
34767754Smsmith	NFSERR_IO,
34867754Smsmith	NFSERR_IO,
34967754Smsmith	NFSERR_ACCES,
35067754Smsmith	NFSERR_EXIST,
35167754Smsmith	NFSERR_NOTDIR,
35267754Smsmith	NFSERR_NOSPC,
35367754Smsmith	NFSERR_ROFS,
354114237Snjl	NFSERR_NAMETOL,
35567754Smsmith	NFSERR_DQUOT,
35667754Smsmith	NFSERR_STALE,
35767754Smsmith	NFSERR_BADHANDLE,
35867754Smsmith	NFSERR_NOTSUPP,
35967754Smsmith	NFSERR_SERVERFAULT,
36067754Smsmith	0,
36167754Smsmith};
36267754Smsmith
36391116Smsmithstatic short nfsv3err_mknod[] = {
36467754Smsmith	NFSERR_IO,
36567754Smsmith	NFSERR_IO,
36667754Smsmith	NFSERR_ACCES,
36767754Smsmith	NFSERR_EXIST,
36867754Smsmith	NFSERR_NOTDIR,
36967754Smsmith	NFSERR_NOSPC,
37067754Smsmith	NFSERR_ROFS,
37167754Smsmith	NFSERR_NAMETOL,
37267754Smsmith	NFSERR_DQUOT,
37367754Smsmith	NFSERR_STALE,
37467754Smsmith	NFSERR_BADHANDLE,
37567754Smsmith	NFSERR_NOTSUPP,
37667754Smsmith	NFSERR_SERVERFAULT,
37767754Smsmith	NFSERR_BADTYPE,
37867754Smsmith	0,
37967754Smsmith};
38067754Smsmith
38167754Smsmithstatic short nfsv3err_remove[] = {
38267754Smsmith	NFSERR_IO,
38367754Smsmith	NFSERR_NOENT,
38467754Smsmith	NFSERR_IO,
38567754Smsmith	NFSERR_ACCES,
38667754Smsmith	NFSERR_NOTDIR,
38767754Smsmith	NFSERR_ROFS,
38867754Smsmith	NFSERR_NAMETOL,
38967754Smsmith	NFSERR_STALE,
39067754Smsmith	NFSERR_BADHANDLE,
39167754Smsmith	NFSERR_SERVERFAULT,
39267754Smsmith	0,
39367754Smsmith};
39467754Smsmith
39567754Smsmithstatic short nfsv3err_rmdir[] = {
39667754Smsmith	NFSERR_IO,
39767754Smsmith	NFSERR_NOENT,
39867754Smsmith	NFSERR_IO,
39967754Smsmith	NFSERR_ACCES,
40067754Smsmith	NFSERR_EXIST,
40167754Smsmith	NFSERR_NOTDIR,
40267754Smsmith	NFSERR_INVAL,
40367754Smsmith	NFSERR_ROFS,
40467754Smsmith	NFSERR_NAMETOL,
40567754Smsmith	NFSERR_NOTEMPTY,
40667754Smsmith	NFSERR_STALE,
40767754Smsmith	NFSERR_BADHANDLE,
40867754Smsmith	NFSERR_NOTSUPP,
40967754Smsmith	NFSERR_SERVERFAULT,
41067754Smsmith	0,
41167754Smsmith};
41299679Siwasaki
41367754Smsmithstatic short nfsv3err_rename[] = {
41467754Smsmith	NFSERR_IO,
41567754Smsmith	NFSERR_NOENT,
41667754Smsmith	NFSERR_IO,
41784491Smsmith	NFSERR_ACCES,
41867754Smsmith	NFSERR_EXIST,
41967754Smsmith	NFSERR_XDEV,
42091116Smsmith	NFSERR_NOTDIR,
42184491Smsmith	NFSERR_ISDIR,
42284491Smsmith	NFSERR_INVAL,
42367754Smsmith	NFSERR_NOSPC,
42467754Smsmith	NFSERR_ROFS,
42567754Smsmith	NFSERR_MLINK,
42667754Smsmith	NFSERR_NAMETOL,
42799679Siwasaki	NFSERR_NOTEMPTY,
42867754Smsmith	NFSERR_DQUOT,
42999679Siwasaki	NFSERR_STALE,
43067754Smsmith	NFSERR_BADHANDLE,
43199679Siwasaki	NFSERR_NOTSUPP,
43299679Siwasaki	NFSERR_SERVERFAULT,
433117521Snjl	0,
43484491Smsmith};
43584491Smsmith
43684491Smsmithstatic short nfsv3err_link[] = {
43784491Smsmith	NFSERR_IO,
43867754Smsmith	NFSERR_IO,
43999679Siwasaki	NFSERR_ACCES,
44084491Smsmith	NFSERR_EXIST,
44184491Smsmith	NFSERR_XDEV,
44299679Siwasaki	NFSERR_NOTDIR,
44384491Smsmith	NFSERR_INVAL,
44499679Siwasaki	NFSERR_NOSPC,
44584491Smsmith	NFSERR_ROFS,
44684491Smsmith	NFSERR_MLINK,
44784491Smsmith	NFSERR_NAMETOL,
44884491Smsmith	NFSERR_DQUOT,
44984491Smsmith	NFSERR_STALE,
45099679Siwasaki	NFSERR_BADHANDLE,
45184491Smsmith	NFSERR_NOTSUPP,
45284491Smsmith	NFSERR_SERVERFAULT,
45384491Smsmith	0,
45499679Siwasaki};
45599679Siwasaki
45667754Smsmithstatic short nfsv3err_readdir[] = {
45767754Smsmith	NFSERR_IO,
45867754Smsmith	NFSERR_IO,
45967754Smsmith	NFSERR_ACCES,
46099679Siwasaki	NFSERR_NOTDIR,
46167754Smsmith	NFSERR_STALE,
46267754Smsmith	NFSERR_BADHANDLE,
46367754Smsmith	NFSERR_BAD_COOKIE,
46467754Smsmith	NFSERR_TOOSMALL,
46599679Siwasaki	NFSERR_SERVERFAULT,
46667754Smsmith	0,
46767754Smsmith};
46867754Smsmith
46967754Smsmithstatic short nfsv3err_readdirplus[] = {
47067754Smsmith	NFSERR_IO,
47167754Smsmith	NFSERR_IO,
47267754Smsmith	NFSERR_ACCES,
47367754Smsmith	NFSERR_NOTDIR,
47467754Smsmith	NFSERR_STALE,
47567754Smsmith	NFSERR_BADHANDLE,
47667754Smsmith	NFSERR_BAD_COOKIE,
47784491Smsmith	NFSERR_NOTSUPP,
47867754Smsmith	NFSERR_TOOSMALL,
47967754Smsmith	NFSERR_SERVERFAULT,
48067754Smsmith	0,
48167754Smsmith};
48267754Smsmith
48367754Smsmithstatic short nfsv3err_fsstat[] = {
48467754Smsmith	NFSERR_IO,
48567754Smsmith	NFSERR_IO,
48667754Smsmith	NFSERR_STALE,
48767754Smsmith	NFSERR_BADHANDLE,
48867754Smsmith	NFSERR_SERVERFAULT,
48967754Smsmith	0,
49067754Smsmith};
49167754Smsmith
49267754Smsmithstatic short nfsv3err_fsinfo[] = {
49367754Smsmith	NFSERR_STALE,
49467754Smsmith	NFSERR_STALE,
49567754Smsmith	NFSERR_BADHANDLE,
49667754Smsmith	NFSERR_SERVERFAULT,
49791116Smsmith	0,
49891116Smsmith};
49991116Smsmith
50067754Smsmithstatic short nfsv3err_pathconf[] = {
50167754Smsmith	NFSERR_STALE,
50267754Smsmith	NFSERR_STALE,
50367754Smsmith	NFSERR_BADHANDLE,
504114237Snjl	NFSERR_SERVERFAULT,
50567754Smsmith	0,
506114237Snjl};
50767754Smsmith
50867754Smsmithstatic short nfsv3err_commit[] = {
50967754Smsmith	NFSERR_IO,
51067754Smsmith	NFSERR_IO,
51167754Smsmith	NFSERR_STALE,
51267754Smsmith	NFSERR_BADHANDLE,
51367754Smsmith	NFSERR_SERVERFAULT,
51467754Smsmith	0,
51567754Smsmith};
51667754Smsmith
51767754Smsmithstatic short *nfsrv_v3errmap[] = {
51867754Smsmith	nfsv3err_null,
51967754Smsmith	nfsv3err_getattr,
52067754Smsmith	nfsv3err_setattr,
52167754Smsmith	nfsv3err_lookup,
52291116Smsmith	nfsv3err_access,
52391116Smsmith	nfsv3err_readlink,
52491116Smsmith	nfsv3err_read,
52591116Smsmith	nfsv3err_write,
52691116Smsmith	nfsv3err_create,
527117521Snjl	nfsv3err_mkdir,
52867754Smsmith	nfsv3err_symlink,
52967754Smsmith	nfsv3err_mknod,
530117521Snjl	nfsv3err_remove,
531117521Snjl	nfsv3err_rmdir,
53267754Smsmith	nfsv3err_rename,
53367754Smsmith	nfsv3err_link,
53480062Smsmith	nfsv3err_readdir,
53567754Smsmith	nfsv3err_readdirplus,
53667754Smsmith	nfsv3err_fsstat,
53767754Smsmith	nfsv3err_fsinfo,
53867754Smsmith	nfsv3err_pathconf,
539102550Siwasaki	nfsv3err_commit,
54067754Smsmith};
54167754Smsmith
542#endif /* NFS_NOSERVER */
543
544extern struct nfsrtt nfsrtt;
545extern time_t nqnfsstarttime;
546extern int nqsrv_clockskew;
547extern int nqsrv_writeslack;
548extern int nqsrv_maxlease;
549extern struct nfsstats nfsstats;
550extern int nqnfs_piggy[NFS_NPROCS];
551extern nfstype nfsv2_type[9];
552extern nfstype nfsv3_type[9];
553extern struct nfsnodehashhead *nfsnodehashtbl;
554extern u_long nfsnodehash;
555
556struct nfssvc_args;
557extern int nfssvc(struct proc *, struct nfssvc_args *, int *);
558
559LIST_HEAD(nfsnodehashhead, nfsnode);
560
561int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
562
563u_quad_t
564nfs_curusec()
565{
566	struct timeval tv;
567
568	getmicrotime(&tv);
569	return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
570}
571
572/*
573 * Create the header for an rpc request packet
574 * The hsiz is the size of the rest of the nfs request header.
575 * (just used to decide if a cluster is a good idea)
576 */
577struct mbuf *
578nfsm_reqh(vp, procid, hsiz, bposp)
579	struct vnode *vp;
580	u_long procid;
581	int hsiz;
582	caddr_t *bposp;
583{
584	register struct mbuf *mb;
585	register u_int32_t *tl;
586	register caddr_t bpos;
587	struct mbuf *mb2;
588	struct nfsmount *nmp;
589	int nqflag;
590
591	MGET(mb, M_WAIT, MT_DATA);
592	if (hsiz >= MINCLSIZE)
593		MCLGET(mb, M_WAIT);
594	mb->m_len = 0;
595	bpos = mtod(mb, caddr_t);
596
597	/*
598	 * For NQNFS, add lease request.
599	 */
600	if (vp) {
601		nmp = VFSTONFS(vp->v_mount);
602		if (nmp->nm_flag & NFSMNT_NQNFS) {
603			nqflag = NQNFS_NEEDLEASE(vp, procid);
604			if (nqflag) {
605				nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
606				*tl++ = txdr_unsigned(nqflag);
607				*tl = txdr_unsigned(nmp->nm_leaseterm);
608			} else {
609				nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
610				*tl = 0;
611			}
612		}
613	}
614	/* Finally, return values */
615	*bposp = bpos;
616	return (mb);
617}
618
619/*
620 * Build the RPC header and fill in the authorization info.
621 * The authorization string argument is only used when the credentials
622 * come from outside of the kernel.
623 * Returns the head of the mbuf list.
624 */
625struct mbuf *
626nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
627	verf_str, mrest, mrest_len, mbp, xidp)
628	register struct ucred *cr;
629	int nmflag;
630	int procid;
631	int auth_type;
632	int auth_len;
633	char *auth_str;
634	int verf_len;
635	char *verf_str;
636	struct mbuf *mrest;
637	int mrest_len;
638	struct mbuf **mbp;
639	u_int32_t *xidp;
640{
641	register struct mbuf *mb;
642	register u_int32_t *tl;
643	register caddr_t bpos;
644	register int i;
645	struct mbuf *mreq, *mb2;
646	int siz, grpsiz, authsiz;
647
648	authsiz = nfsm_rndup(auth_len);
649	MGETHDR(mb, M_WAIT, MT_DATA);
650	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
651		MCLGET(mb, M_WAIT);
652	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
653		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
654	} else {
655		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
656	}
657	mb->m_len = 0;
658	mreq = mb;
659	bpos = mtod(mb, caddr_t);
660
661	/*
662	 * First the RPC header.
663	 */
664	nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
665
666	/* Get a pretty random xid to start with */
667	if (!nfs_xid)
668		nfs_xid = random();
669	/*
670	 * Skip zero xid if it should ever happen.
671	 */
672	if (++nfs_xid == 0)
673		nfs_xid++;
674
675	*tl++ = *xidp = txdr_unsigned(nfs_xid);
676	*tl++ = rpc_call;
677	*tl++ = rpc_vers;
678	if (nmflag & NFSMNT_NQNFS) {
679		*tl++ = txdr_unsigned(NQNFS_PROG);
680		*tl++ = txdr_unsigned(NQNFS_VER3);
681	} else {
682		*tl++ = txdr_unsigned(NFS_PROG);
683		if (nmflag & NFSMNT_NFSV3)
684			*tl++ = txdr_unsigned(NFS_VER3);
685		else
686			*tl++ = txdr_unsigned(NFS_VER2);
687	}
688	if (nmflag & NFSMNT_NFSV3)
689		*tl++ = txdr_unsigned(procid);
690	else
691		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
692
693	/*
694	 * And then the authorization cred.
695	 */
696	*tl++ = txdr_unsigned(auth_type);
697	*tl = txdr_unsigned(authsiz);
698	switch (auth_type) {
699	case RPCAUTH_UNIX:
700		nfsm_build(tl, u_int32_t *, auth_len);
701		*tl++ = 0;		/* stamp ?? */
702		*tl++ = 0;		/* NULL hostname */
703		*tl++ = txdr_unsigned(cr->cr_uid);
704		*tl++ = txdr_unsigned(cr->cr_groups[0]);
705		grpsiz = (auth_len >> 2) - 5;
706		*tl++ = txdr_unsigned(grpsiz);
707		for (i = 1; i <= grpsiz; i++)
708			*tl++ = txdr_unsigned(cr->cr_groups[i]);
709		break;
710	case RPCAUTH_KERB4:
711		siz = auth_len;
712		while (siz > 0) {
713			if (M_TRAILINGSPACE(mb) == 0) {
714				MGET(mb2, M_WAIT, MT_DATA);
715				if (siz >= MINCLSIZE)
716					MCLGET(mb2, M_WAIT);
717				mb->m_next = mb2;
718				mb = mb2;
719				mb->m_len = 0;
720				bpos = mtod(mb, caddr_t);
721			}
722			i = min(siz, M_TRAILINGSPACE(mb));
723			bcopy(auth_str, bpos, i);
724			mb->m_len += i;
725			auth_str += i;
726			bpos += i;
727			siz -= i;
728		}
729		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
730			for (i = 0; i < siz; i++)
731				*bpos++ = '\0';
732			mb->m_len += siz;
733		}
734		break;
735	};
736
737	/*
738	 * And the verifier...
739	 */
740	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
741	if (verf_str) {
742		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
743		*tl = txdr_unsigned(verf_len);
744		siz = verf_len;
745		while (siz > 0) {
746			if (M_TRAILINGSPACE(mb) == 0) {
747				MGET(mb2, M_WAIT, MT_DATA);
748				if (siz >= MINCLSIZE)
749					MCLGET(mb2, M_WAIT);
750				mb->m_next = mb2;
751				mb = mb2;
752				mb->m_len = 0;
753				bpos = mtod(mb, caddr_t);
754			}
755			i = min(siz, M_TRAILINGSPACE(mb));
756			bcopy(verf_str, bpos, i);
757			mb->m_len += i;
758			verf_str += i;
759			bpos += i;
760			siz -= i;
761		}
762		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
763			for (i = 0; i < siz; i++)
764				*bpos++ = '\0';
765			mb->m_len += siz;
766		}
767	} else {
768		*tl++ = txdr_unsigned(RPCAUTH_NULL);
769		*tl = 0;
770	}
771	mb->m_next = mrest;
772	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
773	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
774	*mbp = mb;
775	return (mreq);
776}
777
778/*
779 * copies mbuf chain to the uio scatter/gather list
780 */
781int
782nfsm_mbuftouio(mrep, uiop, siz, dpos)
783	struct mbuf **mrep;
784	register struct uio *uiop;
785	int siz;
786	caddr_t *dpos;
787{
788	register char *mbufcp, *uiocp;
789	register int xfer, left, len;
790	register struct mbuf *mp;
791	long uiosiz, rem;
792	int error = 0;
793
794	mp = *mrep;
795	mbufcp = *dpos;
796	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
797	rem = nfsm_rndup(siz)-siz;
798	while (siz > 0) {
799		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
800			return (EFBIG);
801		left = uiop->uio_iov->iov_len;
802		uiocp = uiop->uio_iov->iov_base;
803		if (left > siz)
804			left = siz;
805		uiosiz = left;
806		while (left > 0) {
807			while (len == 0) {
808				mp = mp->m_next;
809				if (mp == NULL)
810					return (EBADRPC);
811				mbufcp = mtod(mp, caddr_t);
812				len = mp->m_len;
813			}
814			xfer = (left > len) ? len : left;
815#ifdef notdef
816			/* Not Yet.. */
817			if (uiop->uio_iov->iov_op != NULL)
818				(*(uiop->uio_iov->iov_op))
819				(mbufcp, uiocp, xfer);
820			else
821#endif
822			if (uiop->uio_segflg == UIO_SYSSPACE)
823				bcopy(mbufcp, uiocp, xfer);
824			else
825				copyout(mbufcp, uiocp, xfer);
826			left -= xfer;
827			len -= xfer;
828			mbufcp += xfer;
829			uiocp += xfer;
830			uiop->uio_offset += xfer;
831			uiop->uio_resid -= xfer;
832		}
833		if (uiop->uio_iov->iov_len <= siz) {
834			uiop->uio_iovcnt--;
835			uiop->uio_iov++;
836		} else {
837			uiop->uio_iov->iov_base += uiosiz;
838			uiop->uio_iov->iov_len -= uiosiz;
839		}
840		siz -= uiosiz;
841	}
842	*dpos = mbufcp;
843	*mrep = mp;
844	if (rem > 0) {
845		if (len < rem)
846			error = nfs_adv(mrep, dpos, rem, len);
847		else
848			*dpos += rem;
849	}
850	return (error);
851}
852
853/*
854 * copies a uio scatter/gather list to an mbuf chain.
855 * NOTE: can ony handle iovcnt == 1
856 */
857int
858nfsm_uiotombuf(uiop, mq, siz, bpos)
859	register struct uio *uiop;
860	struct mbuf **mq;
861	int siz;
862	caddr_t *bpos;
863{
864	register char *uiocp;
865	register struct mbuf *mp, *mp2;
866	register int xfer, left, mlen;
867	int uiosiz, clflg, rem;
868	char *cp;
869
870#ifdef DIAGNOSTIC
871	if (uiop->uio_iovcnt != 1)
872		panic("nfsm_uiotombuf: iovcnt != 1");
873#endif
874
875	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
876		clflg = 1;
877	else
878		clflg = 0;
879	rem = nfsm_rndup(siz)-siz;
880	mp = mp2 = *mq;
881	while (siz > 0) {
882		left = uiop->uio_iov->iov_len;
883		uiocp = uiop->uio_iov->iov_base;
884		if (left > siz)
885			left = siz;
886		uiosiz = left;
887		while (left > 0) {
888			mlen = M_TRAILINGSPACE(mp);
889			if (mlen == 0) {
890				MGET(mp, M_WAIT, MT_DATA);
891				if (clflg)
892					MCLGET(mp, M_WAIT);
893				mp->m_len = 0;
894				mp2->m_next = mp;
895				mp2 = mp;
896				mlen = M_TRAILINGSPACE(mp);
897			}
898			xfer = (left > mlen) ? mlen : left;
899#ifdef notdef
900			/* Not Yet.. */
901			if (uiop->uio_iov->iov_op != NULL)
902				(*(uiop->uio_iov->iov_op))
903				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
904			else
905#endif
906			if (uiop->uio_segflg == UIO_SYSSPACE)
907				bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
908			else
909				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
910			mp->m_len += xfer;
911			left -= xfer;
912			uiocp += xfer;
913			uiop->uio_offset += xfer;
914			uiop->uio_resid -= xfer;
915		}
916		uiop->uio_iov->iov_base += uiosiz;
917		uiop->uio_iov->iov_len -= uiosiz;
918		siz -= uiosiz;
919	}
920	if (rem > 0) {
921		if (rem > M_TRAILINGSPACE(mp)) {
922			MGET(mp, M_WAIT, MT_DATA);
923			mp->m_len = 0;
924			mp2->m_next = mp;
925		}
926		cp = mtod(mp, caddr_t)+mp->m_len;
927		for (left = 0; left < rem; left++)
928			*cp++ = '\0';
929		mp->m_len += rem;
930		*bpos = cp;
931	} else
932		*bpos = mtod(mp, caddr_t)+mp->m_len;
933	*mq = mp;
934	return (0);
935}
936
937/*
938 * Help break down an mbuf chain by setting the first siz bytes contiguous
939 * pointed to by returned val.
940 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
941 * cases. (The macros use the vars. dpos and dpos2)
942 */
943int
944nfsm_disct(mdp, dposp, siz, left, cp2)
945	struct mbuf **mdp;
946	caddr_t *dposp;
947	int siz;
948	int left;
949	caddr_t *cp2;
950{
951	register struct mbuf *mp, *mp2;
952	register int siz2, xfer;
953	register caddr_t p;
954
955	mp = *mdp;
956	while (left == 0) {
957		*mdp = mp = mp->m_next;
958		if (mp == NULL)
959			return (EBADRPC);
960		left = mp->m_len;
961		*dposp = mtod(mp, caddr_t);
962	}
963	if (left >= siz) {
964		*cp2 = *dposp;
965		*dposp += siz;
966	} else if (mp->m_next == NULL) {
967		return (EBADRPC);
968	} else if (siz > MHLEN) {
969		panic("nfs S too big");
970	} else {
971		MGET(mp2, M_WAIT, MT_DATA);
972		mp2->m_next = mp->m_next;
973		mp->m_next = mp2;
974		mp->m_len -= left;
975		mp = mp2;
976		*cp2 = p = mtod(mp, caddr_t);
977		bcopy(*dposp, p, left);		/* Copy what was left */
978		siz2 = siz-left;
979		p += left;
980		mp2 = mp->m_next;
981		/* Loop around copying up the siz2 bytes */
982		while (siz2 > 0) {
983			if (mp2 == NULL)
984				return (EBADRPC);
985			xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
986			if (xfer > 0) {
987				bcopy(mtod(mp2, caddr_t), p, xfer);
988				NFSMADV(mp2, xfer);
989				mp2->m_len -= xfer;
990				p += xfer;
991				siz2 -= xfer;
992			}
993			if (siz2 > 0)
994				mp2 = mp2->m_next;
995		}
996		mp->m_len = siz;
997		*mdp = mp2;
998		*dposp = mtod(mp2, caddr_t);
999	}
1000	return (0);
1001}
1002
1003/*
1004 * Advance the position in the mbuf chain.
1005 */
1006int
1007nfs_adv(mdp, dposp, offs, left)
1008	struct mbuf **mdp;
1009	caddr_t *dposp;
1010	int offs;
1011	int left;
1012{
1013	register struct mbuf *m;
1014	register int s;
1015
1016	m = *mdp;
1017	s = left;
1018	while (s < offs) {
1019		offs -= s;
1020		m = m->m_next;
1021		if (m == NULL)
1022			return (EBADRPC);
1023		s = m->m_len;
1024	}
1025	*mdp = m;
1026	*dposp = mtod(m, caddr_t)+offs;
1027	return (0);
1028}
1029
1030/*
1031 * Copy a string into mbufs for the hard cases...
1032 */
1033int
1034nfsm_strtmbuf(mb, bpos, cp, siz)
1035	struct mbuf **mb;
1036	char **bpos;
1037	const char *cp;
1038	long siz;
1039{
1040	register struct mbuf *m1 = NULL, *m2;
1041	long left, xfer, len, tlen;
1042	u_int32_t *tl;
1043	int putsize;
1044
1045	putsize = 1;
1046	m2 = *mb;
1047	left = M_TRAILINGSPACE(m2);
1048	if (left > 0) {
1049		tl = ((u_int32_t *)(*bpos));
1050		*tl++ = txdr_unsigned(siz);
1051		putsize = 0;
1052		left -= NFSX_UNSIGNED;
1053		m2->m_len += NFSX_UNSIGNED;
1054		if (left > 0) {
1055			bcopy(cp, (caddr_t) tl, left);
1056			siz -= left;
1057			cp += left;
1058			m2->m_len += left;
1059			left = 0;
1060		}
1061	}
1062	/* Loop around adding mbufs */
1063	while (siz > 0) {
1064		MGET(m1, M_WAIT, MT_DATA);
1065		if (siz > MLEN)
1066			MCLGET(m1, M_WAIT);
1067		m1->m_len = NFSMSIZ(m1);
1068		m2->m_next = m1;
1069		m2 = m1;
1070		tl = mtod(m1, u_int32_t *);
1071		tlen = 0;
1072		if (putsize) {
1073			*tl++ = txdr_unsigned(siz);
1074			m1->m_len -= NFSX_UNSIGNED;
1075			tlen = NFSX_UNSIGNED;
1076			putsize = 0;
1077		}
1078		if (siz < m1->m_len) {
1079			len = nfsm_rndup(siz);
1080			xfer = siz;
1081			if (xfer < len)
1082				*(tl+(xfer>>2)) = 0;
1083		} else {
1084			xfer = len = m1->m_len;
1085		}
1086		bcopy(cp, (caddr_t) tl, xfer);
1087		m1->m_len = len+tlen;
1088		siz -= xfer;
1089		cp += xfer;
1090	}
1091	*mb = m1;
1092	*bpos = mtod(m1, caddr_t)+m1->m_len;
1093	return (0);
1094}
1095
1096/*
1097 * Called once to initialize data structures...
1098 */
1099int
1100nfs_init(vfsp)
1101	struct vfsconf *vfsp;
1102{
1103	register int i;
1104
1105	nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
1106
1107	nfs_mount_type = vfsp->vfc_typenum;
1108	nfsrtt.pos = 0;
1109	rpc_vers = txdr_unsigned(RPC_VER2);
1110	rpc_call = txdr_unsigned(RPC_CALL);
1111	rpc_reply = txdr_unsigned(RPC_REPLY);
1112	rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
1113	rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
1114	rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
1115	rpc_autherr = txdr_unsigned(RPC_AUTHERR);
1116	rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
1117	rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
1118	nfs_prog = txdr_unsigned(NFS_PROG);
1119	nqnfs_prog = txdr_unsigned(NQNFS_PROG);
1120	nfs_true = txdr_unsigned(TRUE);
1121	nfs_false = txdr_unsigned(FALSE);
1122	nfs_xdrneg1 = txdr_unsigned(-1);
1123	nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
1124	if (nfs_ticks < 1)
1125		nfs_ticks = 1;
1126	/* Ensure async daemons disabled */
1127	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1128		nfs_iodwant[i] = (struct proc *)0;
1129		nfs_iodmount[i] = (struct nfsmount *)0;
1130	}
1131	nfs_nhinit();			/* Init the nfsnode table */
1132#ifndef NFS_NOSERVER
1133	nfsrv_init(0);			/* Init server data structures */
1134	nfsrv_initcache();		/* Init the server request cache */
1135#endif
1136
1137	/*
1138	 * Initialize the nqnfs server stuff.
1139	 */
1140	if (nqnfsstarttime == 0) {
1141		nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
1142			+ nqsrv_clockskew + nqsrv_writeslack;
1143		NQLOADNOVRAM(nqnfsstarttime);
1144		CIRCLEQ_INIT(&nqtimerhead);
1145		nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash);
1146	}
1147
1148	/*
1149	 * Initialize reply list and start timer
1150	 */
1151	TAILQ_INIT(&nfs_reqq);
1152
1153	nfs_timer(0);
1154
1155	/*
1156	 * Set up lease_check and lease_updatetime so that other parts
1157	 * of the system can call us, if we are loadable.
1158	 */
1159#ifndef NFS_NOSERVER
1160	nfs_prev_vop_lease_check = default_vnodeop_p[VOFFSET(vop_lease)];
1161	default_vnodeop_p[VOFFSET(vop_lease)] = (vop_t *)nqnfs_vop_lease_check;
1162#endif
1163	nfs_prev_lease_updatetime = lease_updatetime;
1164	lease_updatetime = nfs_lease_updatetime;
1165	nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
1166	sysent[SYS_nfssvc].sy_narg = 2;
1167	nfs_prev_nfssvc_sy_call = sysent[SYS_nfssvc].sy_call;
1168	sysent[SYS_nfssvc].sy_call = (sy_call_t *)nfssvc;
1169
1170	nfs_pbuf_freecnt = nswbuf / 2 + 1;
1171
1172	return (0);
1173}
1174
1175int
1176nfs_uninit(vfsp)
1177	struct vfsconf *vfsp;
1178{
1179
1180	untimeout(nfs_timer, (void *)NULL, nfs_timer_handle);
1181	nfs_mount_type = -1;
1182#ifndef NFS_NOSERVER
1183	default_vnodeop_p[VOFFSET(vop_lease)] = nfs_prev_vop_lease_check;
1184#endif
1185	lease_updatetime = nfs_prev_lease_updatetime;
1186	sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
1187	sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
1188	return (0);
1189}
1190
1191/*
1192 * Attribute cache routines.
1193 * nfs_loadattrcache() - loads or updates the cache contents from attributes
1194 *	that are on the mbuf list
1195 * nfs_getattrcache() - returns valid attributes if found in cache, returns
1196 *	error otherwise
1197 */
1198
1199/*
1200 * Load the attribute cache (that lives in the nfsnode entry) with
1201 * the values on the mbuf list and
1202 * Iff vap not NULL
1203 *    copy the attributes to *vaper
1204 */
1205int
1206nfs_loadattrcache(vpp, mdp, dposp, vaper)
1207	struct vnode **vpp;
1208	struct mbuf **mdp;
1209	caddr_t *dposp;
1210	struct vattr *vaper;
1211{
1212	register struct vnode *vp = *vpp;
1213	register struct vattr *vap;
1214	register struct nfs_fattr *fp;
1215	register struct nfsnode *np;
1216	register int32_t t1;
1217	caddr_t cp2;
1218	int error = 0, rdev;
1219	struct mbuf *md;
1220	enum vtype vtyp;
1221	u_short vmode;
1222	struct timespec mtime;
1223	int v3 = NFS_ISV3(vp);
1224
1225	md = *mdp;
1226	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1227	if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
1228		return (error);
1229	fp = (struct nfs_fattr *)cp2;
1230	if (v3) {
1231		vtyp = nfsv3tov_type(fp->fa_type);
1232		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1233		rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
1234			fxdr_unsigned(int, fp->fa3_rdev.specdata2));
1235		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1236	} else {
1237		vtyp = nfsv2tov_type(fp->fa_type);
1238		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1239		/*
1240		 * XXX
1241		 *
1242		 * The duplicate information returned in fa_type and fa_mode
1243		 * is an ambiguity in the NFS version 2 protocol.
1244		 *
1245		 * VREG should be taken literally as a regular file.  If a
1246		 * server intents to return some type information differently
1247		 * in the upper bits of the mode field (e.g. for sockets, or
1248		 * FIFOs), NFSv2 mandates fa_type to be VNON.  Anyway, we
1249		 * leave the examination of the mode bits even in the VREG
1250		 * case to avoid breakage for bogus servers, but we make sure
1251		 * that there are actually type bits set in the upper part of
1252		 * fa_mode (and failing that, trust the va_type field).
1253		 *
1254		 * NFSv3 cleared the issue, and requires fa_mode to not
1255		 * contain any type information (while also introduing sockets
1256		 * and FIFOs for fa_type).
1257		 */
1258		if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
1259			vtyp = IFTOVT(vmode);
1260		rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1261		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1262
1263		/*
1264		 * Really ugly NFSv2 kludge.
1265		 */
1266		if (vtyp == VCHR && rdev == 0xffffffff)
1267			vtyp = VFIFO;
1268	}
1269
1270	/*
1271	 * If v_type == VNON it is a new node, so fill in the v_type,
1272	 * n_mtime fields. Check to see if it represents a special
1273	 * device, and if so, check for a possible alias. Once the
1274	 * correct vnode has been obtained, fill in the rest of the
1275	 * information.
1276	 */
1277	np = VTONFS(vp);
1278	if (vp->v_type != vtyp) {
1279		vp->v_type = vtyp;
1280		if (vp->v_type == VFIFO) {
1281			vp->v_op = fifo_nfsv2nodeop_p;
1282		}
1283		if (vp->v_type == VCHR || vp->v_type == VBLK) {
1284			vp->v_op = spec_nfsv2nodeop_p;
1285			addaliasu(vp, rdev);
1286		}
1287		np->n_mtime = mtime.tv_sec;
1288	}
1289	vap = &np->n_vattr;
1290	vap->va_type = vtyp;
1291	vap->va_mode = (vmode & 07777);
1292	vap->va_rdev = rdev;
1293	vap->va_mtime = mtime;
1294	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1295	if (v3) {
1296		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1297		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1298		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1299		vap->va_size = fxdr_hyper(&fp->fa3_size);
1300		vap->va_blocksize = NFS_FABLKSIZE;
1301		vap->va_bytes = fxdr_hyper(&fp->fa3_used);
1302		vap->va_fileid = fxdr_unsigned(int32_t,
1303		    fp->fa3_fileid.nfsuquad[1]);
1304		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1305		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1306		vap->va_flags = 0;
1307		vap->va_filerev = 0;
1308	} else {
1309		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1310		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1311		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1312		vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1313		vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
1314		vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
1315		    * NFS_FABLKSIZE;
1316		vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1317		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1318		vap->va_flags = 0;
1319		vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1320		    fp->fa2_ctime.nfsv2_sec);
1321		vap->va_ctime.tv_nsec = 0;
1322		vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1323		vap->va_filerev = 0;
1324	}
1325	if (vap->va_size != np->n_size) {
1326		if (vap->va_type == VREG) {
1327			if (np->n_flag & NMODIFIED) {
1328				if (vap->va_size < np->n_size)
1329					vap->va_size = np->n_size;
1330				else
1331					np->n_size = vap->va_size;
1332			} else {
1333				np->n_size = vap->va_size;
1334			}
1335			vnode_pager_setsize(vp, np->n_size);
1336		} else {
1337			np->n_size = vap->va_size;
1338		}
1339	}
1340	np->n_attrstamp = time_second;
1341	if (vaper != NULL) {
1342		bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
1343		if (np->n_flag & NCHG) {
1344			if (np->n_flag & NACC)
1345				vaper->va_atime = np->n_atim;
1346			if (np->n_flag & NUPD)
1347				vaper->va_mtime = np->n_mtim;
1348		}
1349	}
1350	return (0);
1351}
1352
1353#ifdef NFS_ACDEBUG
1354#include <sys/sysctl.h>
1355SYSCTL_DECL(_vfs_nfs);
1356static int nfs_acdebug;
1357SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
1358#endif
1359
1360/*
1361 * Check the time stamp
1362 * If the cache is valid, copy contents to *vap and return 0
1363 * otherwise return an error
1364 */
1365int
1366nfs_getattrcache(vp, vaper)
1367	register struct vnode *vp;
1368	struct vattr *vaper;
1369{
1370	register struct nfsnode *np;
1371	register struct vattr *vap;
1372	struct nfsmount *nmp;
1373	int timeo;
1374
1375	np = VTONFS(vp);
1376	vap = &np->n_vattr;
1377	nmp = VFSTONFS(vp->v_mount);
1378	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
1379	timeo = (time_second - np->n_mtime) / 10;
1380
1381#ifdef NFS_ACDEBUG
1382	if (nfs_acdebug>1)
1383		printf("nfs_getattrcache: initial timeo = %d\n", timeo);
1384#endif
1385
1386	if (vap->va_type == VDIR) {
1387		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
1388			timeo = nmp->nm_acdirmin;
1389		else if (timeo > nmp->nm_acdirmax)
1390			timeo = nmp->nm_acdirmax;
1391	} else {
1392		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
1393			timeo = nmp->nm_acregmin;
1394		else if (timeo > nmp->nm_acregmax)
1395			timeo = nmp->nm_acregmax;
1396	}
1397
1398#ifdef NFS_ACDEBUG
1399	if (nfs_acdebug > 2)
1400		printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
1401			nmp->nm_acregmin, nmp->nm_acregmax,
1402			nmp->nm_acdirmin, nmp->nm_acdirmax);
1403
1404	if (nfs_acdebug)
1405		printf("nfs_getattrcache: age = %d; final timeo = %d\n",
1406			(time_second - np->n_attrstamp), timeo);
1407#endif
1408
1409	if ((time_second - np->n_attrstamp) >= timeo) {
1410		nfsstats.attrcache_misses++;
1411		return (ENOENT);
1412	}
1413	nfsstats.attrcache_hits++;
1414	if (vap->va_size != np->n_size) {
1415		if (vap->va_type == VREG) {
1416			if (np->n_flag & NMODIFIED) {
1417				if (vap->va_size < np->n_size)
1418					vap->va_size = np->n_size;
1419				else
1420					np->n_size = vap->va_size;
1421			} else {
1422				np->n_size = vap->va_size;
1423			}
1424			vnode_pager_setsize(vp, np->n_size);
1425		} else {
1426			np->n_size = vap->va_size;
1427		}
1428	}
1429	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
1430	if (np->n_flag & NCHG) {
1431		if (np->n_flag & NACC)
1432			vaper->va_atime = np->n_atim;
1433		if (np->n_flag & NUPD)
1434			vaper->va_mtime = np->n_mtim;
1435	}
1436	return (0);
1437}
1438
1439#ifndef NFS_NOSERVER
1440/*
1441 * Set up nameidata for a lookup() call and do it.
1442 *
1443 * If pubflag is set, this call is done for a lookup operation on the
1444 * public filehandle. In that case we allow crossing mountpoints and
1445 * absolute pathnames. However, the caller is expected to check that
1446 * the lookup result is within the public fs, and deny access if
1447 * it is not.
1448 *
1449 * nfs_namei() clears out garbage fields that namei() might leave garbage.
1450 * This is mainly ni_vp and ni_dvp when an error occurs, and ni_dvp when no
1451 * error occurs but the parent was not requested.
1452 *
1453 * dirp may be set whether an error is returned or not, and must be
1454 * released by the caller.
1455 */
1456int
1457nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
1458	register struct nameidata *ndp;
1459	fhandle_t *fhp;
1460	int len;
1461	struct nfssvc_sock *slp;
1462	struct sockaddr *nam;
1463	struct mbuf **mdp;
1464	caddr_t *dposp;
1465	struct vnode **retdirp;
1466	struct proc *p;
1467	int kerbflag, pubflag;
1468{
1469	register int i, rem;
1470	register struct mbuf *md;
1471	register char *fromcp, *tocp, *cp;
1472	struct iovec aiov;
1473	struct uio auio;
1474	struct vnode *dp;
1475	int error, rdonly, linklen;
1476	struct componentname *cnp = &ndp->ni_cnd;
1477
1478	*retdirp = (struct vnode *)0;
1479	cnp->cn_pnbuf = zalloc(namei_zone);
1480
1481	/*
1482	 * Copy the name from the mbuf list to ndp->ni_pnbuf
1483	 * and set the various ndp fields appropriately.
1484	 */
1485	fromcp = *dposp;
1486	tocp = cnp->cn_pnbuf;
1487	md = *mdp;
1488	rem = mtod(md, caddr_t) + md->m_len - fromcp;
1489	for (i = 0; i < len; i++) {
1490		while (rem == 0) {
1491			md = md->m_next;
1492			if (md == NULL) {
1493				error = EBADRPC;
1494				goto out;
1495			}
1496			fromcp = mtod(md, caddr_t);
1497			rem = md->m_len;
1498		}
1499		if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1500			error = EACCES;
1501			goto out;
1502		}
1503		*tocp++ = *fromcp++;
1504		rem--;
1505	}
1506	*tocp = '\0';
1507	*mdp = md;
1508	*dposp = fromcp;
1509	len = nfsm_rndup(len)-len;
1510	if (len > 0) {
1511		if (rem >= len)
1512			*dposp += len;
1513		else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1514			goto out;
1515	}
1516
1517	/*
1518	 * Extract and set starting directory.
1519	 */
1520	error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
1521	    nam, &rdonly, kerbflag, pubflag);
1522	if (error)
1523		goto out;
1524	if (dp->v_type != VDIR) {
1525		vrele(dp);
1526		error = ENOTDIR;
1527		goto out;
1528	}
1529
1530	if (rdonly)
1531		cnp->cn_flags |= RDONLY;
1532
1533	/*
1534	 * Set return directory.  Reference to dp is implicitly transfered
1535	 * to the returned pointer
1536	 */
1537	*retdirp = dp;
1538
1539	if (pubflag) {
1540		/*
1541		 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1542		 * and the 'native path' indicator.
1543		 */
1544		cp = zalloc(namei_zone);
1545		fromcp = cnp->cn_pnbuf;
1546		tocp = cp;
1547		if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1548			switch ((unsigned char)*fromcp) {
1549			case WEBNFS_NATIVE_CHAR:
1550				/*
1551				 * 'Native' path for us is the same
1552				 * as a path according to the NFS spec,
1553				 * just skip the escape char.
1554				 */
1555				fromcp++;
1556				break;
1557			/*
1558			 * More may be added in the future, range 0x80-0xff
1559			 */
1560			default:
1561				error = EIO;
1562				zfree(namei_zone, cp);
1563				goto out;
1564			}
1565		}
1566		/*
1567		 * Translate the '%' escapes, URL-style.
1568		 */
1569		while (*fromcp != '\0') {
1570			if (*fromcp == WEBNFS_ESC_CHAR) {
1571				if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1572					fromcp++;
1573					*tocp++ = HEXSTRTOI(fromcp);
1574					fromcp += 2;
1575					continue;
1576				} else {
1577					error = ENOENT;
1578					zfree(namei_zone, cp);
1579					goto out;
1580				}
1581			} else
1582				*tocp++ = *fromcp++;
1583		}
1584		*tocp = '\0';
1585		zfree(namei_zone, cnp->cn_pnbuf);
1586		cnp->cn_pnbuf = cp;
1587	}
1588
1589	ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
1590	ndp->ni_segflg = UIO_SYSSPACE;
1591
1592	if (pubflag) {
1593		ndp->ni_rootdir = rootvnode;
1594		ndp->ni_loopcnt = 0;
1595		if (cnp->cn_pnbuf[0] == '/')
1596			dp = rootvnode;
1597	} else {
1598		cnp->cn_flags |= NOCROSSMOUNT;
1599	}
1600
1601	/*
1602	 * Initialize for scan, set ni_startdir and bump ref on dp again
1603	 * becuase lookup() will dereference ni_startdir.
1604	 */
1605
1606	cnp->cn_proc = p;
1607	VREF(dp);
1608	ndp->ni_startdir = dp;
1609
1610	for (;;) {
1611		cnp->cn_nameptr = cnp->cn_pnbuf;
1612		/*
1613		 * Call lookup() to do the real work.  If an error occurs,
1614		 * ndp->ni_vp and ni_dvp are left uninitialized or NULL and
1615		 * we do not have to dereference anything before returning.
1616		 * In either case ni_startdir will be dereferenced and NULLed
1617		 * out.
1618		 */
1619		error = lookup(ndp);
1620		if (error)
1621			break;
1622
1623		/*
1624		 * Check for encountering a symbolic link.  Trivial
1625		 * termination occurs if no symlink encountered.
1626		 * Note: zfree is safe because error is 0, so we will
1627		 * not zfree it again when we break.
1628		 */
1629		if ((cnp->cn_flags & ISSYMLINK) == 0) {
1630			nfsrv_object_create(ndp->ni_vp);
1631			if (cnp->cn_flags & (SAVENAME | SAVESTART))
1632				cnp->cn_flags |= HASBUF;
1633			else
1634				zfree(namei_zone, cnp->cn_pnbuf);
1635			break;
1636		}
1637
1638		/*
1639		 * Validate symlink
1640		 */
1641		if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
1642			VOP_UNLOCK(ndp->ni_dvp, 0, p);
1643		if (!pubflag) {
1644			error = EINVAL;
1645			goto badlink2;
1646		}
1647
1648		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
1649			error = ELOOP;
1650			goto badlink2;
1651		}
1652		if (ndp->ni_pathlen > 1)
1653			cp = zalloc(namei_zone);
1654		else
1655			cp = cnp->cn_pnbuf;
1656		aiov.iov_base = cp;
1657		aiov.iov_len = MAXPATHLEN;
1658		auio.uio_iov = &aiov;
1659		auio.uio_iovcnt = 1;
1660		auio.uio_offset = 0;
1661		auio.uio_rw = UIO_READ;
1662		auio.uio_segflg = UIO_SYSSPACE;
1663		auio.uio_procp = (struct proc *)0;
1664		auio.uio_resid = MAXPATHLEN;
1665		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
1666		if (error) {
1667		badlink1:
1668			if (ndp->ni_pathlen > 1)
1669				zfree(namei_zone, cp);
1670		badlink2:
1671			vrele(ndp->ni_dvp);
1672			vput(ndp->ni_vp);
1673			break;
1674		}
1675		linklen = MAXPATHLEN - auio.uio_resid;
1676		if (linklen == 0) {
1677			error = ENOENT;
1678			goto badlink1;
1679		}
1680		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
1681			error = ENAMETOOLONG;
1682			goto badlink1;
1683		}
1684
1685		/*
1686		 * Adjust or replace path
1687		 */
1688		if (ndp->ni_pathlen > 1) {
1689			bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
1690			zfree(namei_zone, cnp->cn_pnbuf);
1691			cnp->cn_pnbuf = cp;
1692		} else
1693			cnp->cn_pnbuf[linklen] = '\0';
1694		ndp->ni_pathlen += linklen;
1695
1696		/*
1697		 * Cleanup refs for next loop and check if root directory
1698		 * should replace current directory.  Normally ni_dvp
1699		 * becomes the new base directory and is cleaned up when
1700		 * we loop.  Explicitly null pointers after invalidation
1701		 * to clarify operation.
1702		 */
1703		vput(ndp->ni_vp);
1704		ndp->ni_vp = NULL;
1705
1706		if (cnp->cn_pnbuf[0] == '/') {
1707			vrele(ndp->ni_dvp);
1708			ndp->ni_dvp = ndp->ni_rootdir;
1709			VREF(ndp->ni_dvp);
1710		}
1711		ndp->ni_startdir = ndp->ni_dvp;
1712		ndp->ni_dvp = NULL;
1713	}
1714
1715	/*
1716	 * nfs_namei() guarentees that fields will not contain garbage
1717	 * whether an error occurs or not.  This allows the caller to track
1718	 * cleanup state trivially.
1719	 */
1720out:
1721	if (error) {
1722		zfree(namei_zone, cnp->cn_pnbuf);
1723		ndp->ni_vp = NULL;
1724		ndp->ni_dvp = NULL;
1725		ndp->ni_startdir = NULL;
1726		cnp->cn_flags &= ~HASBUF;
1727	} else if ((ndp->ni_cnd.cn_flags & (WANTPARENT|LOCKPARENT)) == 0) {
1728		ndp->ni_dvp = NULL;
1729	}
1730	return (error);
1731}
1732
1733/*
1734 * A fiddled version of m_adj() that ensures null fill to a long
1735 * boundary and only trims off the back end
1736 */
1737void
1738nfsm_adj(mp, len, nul)
1739	struct mbuf *mp;
1740	register int len;
1741	int nul;
1742{
1743	register struct mbuf *m;
1744	register int count, i;
1745	register char *cp;
1746
1747	/*
1748	 * Trim from tail.  Scan the mbuf chain,
1749	 * calculating its length and finding the last mbuf.
1750	 * If the adjustment only affects this mbuf, then just
1751	 * adjust and return.  Otherwise, rescan and truncate
1752	 * after the remaining size.
1753	 */
1754	count = 0;
1755	m = mp;
1756	for (;;) {
1757		count += m->m_len;
1758		if (m->m_next == (struct mbuf *)0)
1759			break;
1760		m = m->m_next;
1761	}
1762	if (m->m_len > len) {
1763		m->m_len -= len;
1764		if (nul > 0) {
1765			cp = mtod(m, caddr_t)+m->m_len-nul;
1766			for (i = 0; i < nul; i++)
1767				*cp++ = '\0';
1768		}
1769		return;
1770	}
1771	count -= len;
1772	if (count < 0)
1773		count = 0;
1774	/*
1775	 * Correct length for chain is "count".
1776	 * Find the mbuf with last data, adjust its length,
1777	 * and toss data from remaining mbufs on chain.
1778	 */
1779	for (m = mp; m; m = m->m_next) {
1780		if (m->m_len >= count) {
1781			m->m_len = count;
1782			if (nul > 0) {
1783				cp = mtod(m, caddr_t)+m->m_len-nul;
1784				for (i = 0; i < nul; i++)
1785					*cp++ = '\0';
1786			}
1787			break;
1788		}
1789		count -= m->m_len;
1790	}
1791	for (m = m->m_next;m;m = m->m_next)
1792		m->m_len = 0;
1793}
1794
1795/*
1796 * Make these functions instead of macros, so that the kernel text size
1797 * doesn't get too big...
1798 */
1799void
1800nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
1801	struct nfsrv_descript *nfsd;
1802	int before_ret;
1803	register struct vattr *before_vap;
1804	int after_ret;
1805	struct vattr *after_vap;
1806	struct mbuf **mbp;
1807	char **bposp;
1808{
1809	register struct mbuf *mb = *mbp, *mb2;
1810	register char *bpos = *bposp;
1811	register u_int32_t *tl;
1812
1813	if (before_ret) {
1814		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1815		*tl = nfs_false;
1816	} else {
1817		nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
1818		*tl++ = nfs_true;
1819		txdr_hyper(before_vap->va_size, tl);
1820		tl += 2;
1821		txdr_nfsv3time(&(before_vap->va_mtime), tl);
1822		tl += 2;
1823		txdr_nfsv3time(&(before_vap->va_ctime), tl);
1824	}
1825	*bposp = bpos;
1826	*mbp = mb;
1827	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
1828}
1829
1830void
1831nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
1832	struct nfsrv_descript *nfsd;
1833	int after_ret;
1834	struct vattr *after_vap;
1835	struct mbuf **mbp;
1836	char **bposp;
1837{
1838	register struct mbuf *mb = *mbp, *mb2;
1839	register char *bpos = *bposp;
1840	register u_int32_t *tl;
1841	register struct nfs_fattr *fp;
1842
1843	if (after_ret) {
1844		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1845		*tl = nfs_false;
1846	} else {
1847		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
1848		*tl++ = nfs_true;
1849		fp = (struct nfs_fattr *)tl;
1850		nfsm_srvfattr(nfsd, after_vap, fp);
1851	}
1852	*mbp = mb;
1853	*bposp = bpos;
1854}
1855
1856void
1857nfsm_srvfattr(nfsd, vap, fp)
1858	register struct nfsrv_descript *nfsd;
1859	register struct vattr *vap;
1860	register struct nfs_fattr *fp;
1861{
1862
1863	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
1864	fp->fa_uid = txdr_unsigned(vap->va_uid);
1865	fp->fa_gid = txdr_unsigned(vap->va_gid);
1866	if (nfsd->nd_flag & ND_NFSV3) {
1867		fp->fa_type = vtonfsv3_type(vap->va_type);
1868		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
1869		txdr_hyper(vap->va_size, &fp->fa3_size);
1870		txdr_hyper(vap->va_bytes, &fp->fa3_used);
1871		fp->fa3_rdev.specdata1 = txdr_unsigned(umajor(vap->va_rdev));
1872		fp->fa3_rdev.specdata2 = txdr_unsigned(uminor(vap->va_rdev));
1873		fp->fa3_fsid.nfsuquad[0] = 0;
1874		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
1875		fp->fa3_fileid.nfsuquad[0] = 0;
1876		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
1877		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
1878		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
1879		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
1880	} else {
1881		fp->fa_type = vtonfsv2_type(vap->va_type);
1882		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
1883		fp->fa2_size = txdr_unsigned(vap->va_size);
1884		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
1885		if (vap->va_type == VFIFO)
1886			fp->fa2_rdev = 0xffffffff;
1887		else
1888			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
1889		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
1890		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
1891		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
1892		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
1893		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
1894		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
1895	}
1896}
1897
1898/*
1899 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
1900 * 	- look up fsid in mount list (if not found ret error)
1901 *	- get vp and export rights by calling VFS_FHTOVP()
1902 *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
1903 *	- if not lockflag unlock it with VOP_UNLOCK()
1904 */
1905int
1906nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag)
1907	fhandle_t *fhp;
1908	int lockflag;
1909	struct vnode **vpp;
1910	struct ucred *cred;
1911	struct nfssvc_sock *slp;
1912	struct sockaddr *nam;
1913	int *rdonlyp;
1914	int kerbflag;
1915	int pubflag;
1916{
1917	struct proc *p = curproc; /* XXX */
1918	register struct mount *mp;
1919	register int i;
1920	struct ucred *credanon;
1921	int error, exflags;
1922#ifdef MNT_EXNORESPORT		/* XXX needs mountd and /etc/exports help yet */
1923	struct sockaddr_int *saddr;
1924#endif
1925
1926	*vpp = (struct vnode *)0;
1927
1928	if (nfs_ispublicfh(fhp)) {
1929		if (!pubflag || !nfs_pub.np_valid)
1930			return (ESTALE);
1931		fhp = &nfs_pub.np_handle;
1932	}
1933
1934	mp = vfs_getvfs(&fhp->fh_fsid);
1935	if (!mp)
1936		return (ESTALE);
1937	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
1938	if (error)
1939		return (error);
1940	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
1941	if (error)
1942		return (error);
1943#ifdef MNT_EXNORESPORT
1944	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
1945		saddr = (struct sockaddr_in *)nam;
1946		if (saddr->sin_family == AF_INET &&
1947		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
1948			vput(*vpp);
1949			*vpp = NULL;
1950			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1951		}
1952	}
1953#endif
1954	/*
1955	 * Check/setup credentials.
1956	 */
1957	if (exflags & MNT_EXKERB) {
1958		if (!kerbflag) {
1959			vput(*vpp);
1960			*vpp = NULL;
1961			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1962		}
1963	} else if (kerbflag) {
1964		vput(*vpp);
1965		*vpp = NULL;
1966		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1967	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1968		cred->cr_uid = credanon->cr_uid;
1969		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
1970			cred->cr_groups[i] = credanon->cr_groups[i];
1971		cred->cr_ngroups = i;
1972	}
1973	if (exflags & MNT_EXRDONLY)
1974		*rdonlyp = 1;
1975	else
1976		*rdonlyp = 0;
1977
1978	nfsrv_object_create(*vpp);
1979
1980	if (!lockflag)
1981		VOP_UNLOCK(*vpp, 0, p);
1982	return (0);
1983}
1984
1985
1986/*
1987 * WebNFS: check if a filehandle is a public filehandle. For v3, this
1988 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
1989 * transformed this to all zeroes in both cases, so check for it.
1990 */
1991int
1992nfs_ispublicfh(fhp)
1993	fhandle_t *fhp;
1994{
1995	char *cp = (char *)fhp;
1996	int i;
1997
1998	for (i = 0; i < NFSX_V3FH; i++)
1999		if (*cp++ != 0)
2000			return (FALSE);
2001	return (TRUE);
2002}
2003
2004#endif /* NFS_NOSERVER */
2005/*
2006 * This function compares two net addresses by family and returns TRUE
2007 * if they are the same host.
2008 * If there is any doubt, return FALSE.
2009 * The AF_INET family is handled as a special case so that address mbufs
2010 * don't need to be saved to store "struct in_addr", which is only 4 bytes.
2011 */
2012int
2013netaddr_match(family, haddr, nam)
2014	int family;
2015	union nethostaddr *haddr;
2016	struct sockaddr *nam;
2017{
2018	register struct sockaddr_in *inetaddr;
2019
2020	switch (family) {
2021	case AF_INET:
2022		inetaddr = (struct sockaddr_in *)nam;
2023		if (inetaddr->sin_family == AF_INET &&
2024		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
2025			return (1);
2026		break;
2027	default:
2028		break;
2029	};
2030	return (0);
2031}
2032
2033static nfsuint64 nfs_nullcookie = { { 0, 0 } };
2034/*
2035 * This function finds the directory cookie that corresponds to the
2036 * logical byte offset given.
2037 */
2038nfsuint64 *
2039nfs_getcookie(np, off, add)
2040	register struct nfsnode *np;
2041	off_t off;
2042	int add;
2043{
2044	register struct nfsdmap *dp, *dp2;
2045	register int pos;
2046
2047	pos = (uoff_t)off / NFS_DIRBLKSIZ;
2048	if (pos == 0 || off < 0) {
2049#ifdef DIAGNOSTIC
2050		if (add)
2051			panic("nfs getcookie add at <= 0");
2052#endif
2053		return (&nfs_nullcookie);
2054	}
2055	pos--;
2056	dp = np->n_cookies.lh_first;
2057	if (!dp) {
2058		if (add) {
2059			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
2060				M_NFSDIROFF, M_WAITOK);
2061			dp->ndm_eocookie = 0;
2062			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
2063		} else
2064			return ((nfsuint64 *)0);
2065	}
2066	while (pos >= NFSNUMCOOKIES) {
2067		pos -= NFSNUMCOOKIES;
2068		if (dp->ndm_list.le_next) {
2069			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
2070				pos >= dp->ndm_eocookie)
2071				return ((nfsuint64 *)0);
2072			dp = dp->ndm_list.le_next;
2073		} else if (add) {
2074			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
2075				M_NFSDIROFF, M_WAITOK);
2076			dp2->ndm_eocookie = 0;
2077			LIST_INSERT_AFTER(dp, dp2, ndm_list);
2078			dp = dp2;
2079		} else
2080			return ((nfsuint64 *)0);
2081	}
2082	if (pos >= dp->ndm_eocookie) {
2083		if (add)
2084			dp->ndm_eocookie = pos + 1;
2085		else
2086			return ((nfsuint64 *)0);
2087	}
2088	return (&dp->ndm_cookies[pos]);
2089}
2090
2091/*
2092 * Invalidate cached directory information, except for the actual directory
2093 * blocks (which are invalidated separately).
2094 * Done mainly to avoid the use of stale offset cookies.
2095 */
2096void
2097nfs_invaldir(vp)
2098	register struct vnode *vp;
2099{
2100	register struct nfsnode *np = VTONFS(vp);
2101
2102#ifdef DIAGNOSTIC
2103	if (vp->v_type != VDIR)
2104		panic("nfs: invaldir not dir");
2105#endif
2106	np->n_direofoffset = 0;
2107	np->n_cookieverf.nfsuquad[0] = 0;
2108	np->n_cookieverf.nfsuquad[1] = 0;
2109	if (np->n_cookies.lh_first)
2110		np->n_cookies.lh_first->ndm_eocookie = 0;
2111}
2112
2113/*
2114 * The write verifier has changed (probably due to a server reboot), so all
2115 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
2116 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2117 * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
2118 * mount point.
2119 *
2120 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
2121 * writes are not clusterable.
2122 */
2123void
2124nfs_clearcommit(mp)
2125	struct mount *mp;
2126{
2127	register struct vnode *vp, *nvp;
2128	register struct buf *bp, *nbp;
2129	int s;
2130
2131	s = splbio();
2132loop:
2133	for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) {
2134		if (vp->v_mount != mp)	/* Paranoia */
2135			goto loop;
2136		nvp = vp->v_mntvnodes.le_next;
2137		for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
2138			nbp = TAILQ_NEXT(bp, b_vnbufs);
2139			if (BUF_REFCNT(bp) == 0 &&
2140			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
2141				== (B_DELWRI | B_NEEDCOMMIT))
2142				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
2143		}
2144	}
2145	splx(s);
2146}
2147
2148#ifndef NFS_NOSERVER
2149/*
2150 * Map errnos to NFS error numbers. For Version 3 also filter out error
2151 * numbers not specified for the associated procedure.
2152 */
2153int
2154nfsrv_errmap(nd, err)
2155	struct nfsrv_descript *nd;
2156	register int err;
2157{
2158	register short *defaulterrp, *errp;
2159
2160	if (nd->nd_flag & ND_NFSV3) {
2161	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
2162		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2163		while (*++errp) {
2164			if (*errp == err)
2165				return (err);
2166			else if (*errp > err)
2167				break;
2168		}
2169		return ((int)*defaulterrp);
2170	    } else
2171		return (err & 0xffff);
2172	}
2173	if (err <= ELAST)
2174		return ((int)nfsrv_v2errmap[err - 1]);
2175	return (NFSERR_IO);
2176}
2177
2178int
2179nfsrv_object_create(vp)
2180	struct vnode *vp;
2181{
2182
2183	if (vp == NULL || vp->v_type != VREG)
2184		return (1);
2185	return (vfs_object_create(vp, curproc,
2186				  curproc ? curproc->p_ucred : NULL));
2187}
2188
2189/*
2190 * Sort the group list in increasing numerical order.
2191 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2192 *  that used to be here.)
2193 */
2194void
2195nfsrvw_sort(list, num)
2196        register gid_t *list;
2197        register int num;
2198{
2199	register int i, j;
2200	gid_t v;
2201
2202	/* Insertion sort. */
2203	for (i = 1; i < num; i++) {
2204		v = list[i];
2205		/* find correct slot for value v, moving others up */
2206		for (j = i; --j >= 0 && v < list[j];)
2207			list[j + 1] = list[j];
2208		list[j + 1] = v;
2209	}
2210}
2211
2212/*
2213 * copy credentials making sure that the result can be compared with bcmp().
2214 */
2215void
2216nfsrv_setcred(incred, outcred)
2217	register struct ucred *incred, *outcred;
2218{
2219	register int i;
2220
2221	bzero((caddr_t)outcred, sizeof (struct ucred));
2222	outcred->cr_ref = 1;
2223	outcred->cr_uid = incred->cr_uid;
2224	outcred->cr_ngroups = incred->cr_ngroups;
2225	for (i = 0; i < incred->cr_ngroups; i++)
2226		outcred->cr_groups[i] = incred->cr_groups[i];
2227	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2228}
2229#endif /* NFS_NOSERVER */
2230