2 * Copyright (c) 1989, 1993, 1995
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * @(#)nfs_vfsops.c 8.12 (Berkeley) 5/20/95
37 * $FreeBSD: src/sys/nfs/nfs_vfsops.c,v 1.91.2.7 2003/01/27 20:04:08 dillon Exp $
38 * $DragonFly: src/sys/vfs/nfs/nfs_vfsops.c,v 1.2 2003/06/17 04:28:54 dillon Exp $
41 #include "opt_bootp.h"
43 #include <sys/param.h>
44 #include <sys/sockio.h>
46 #include <sys/vnode.h>
47 #include <sys/kernel.h>
48 #include <sys/sysctl.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/systm.h>
57 #include <vm/vm_extern.h>
58 #include <vm/vm_zone.h>
61 #include <net/route.h>
62 #include <netinet/in.h>
64 #include <nfs/rpcv2.h>
65 #include <nfs/nfsproto.h>
67 #include <nfs/nfsnode.h>
68 #include <nfs/nfsmount.h>
69 #include <nfs/xdr_subs.h>
70 #include <nfs/nfsm_subs.h>
71 #include <nfs/nfsdiskless.h>
72 #include <nfs/nqnfs.h>
74 extern int nfs_mountroot __P((struct mount *mp));
78 MALLOC_DEFINE(M_NFSREQ, "NFS req", "NFS request header");
79 MALLOC_DEFINE(M_NFSBIGFH, "NFSV3 bigfh", "NFS version 3 file handle");
80 MALLOC_DEFINE(M_NFSD, "NFS daemon", "Nfs server daemon structure");
81 MALLOC_DEFINE(M_NFSDIROFF, "NFSV3 diroff", "NFS directory offset data");
82 MALLOC_DEFINE(M_NFSRVDESC, "NFSV3 srvdesc", "NFS server socket descriptor");
83 MALLOC_DEFINE(M_NFSUID, "NFS uid", "Nfs uid mapping structure");
84 MALLOC_DEFINE(M_NQLEASE, "NQNFS Lease", "Nqnfs lease");
85 MALLOC_DEFINE(M_NFSHASH, "NFS hash", "NFS hash tables");
87 vm_zone_t nfsmount_zone;
89 struct nfsstats nfsstats;
90 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem");
91 SYSCTL_STRUCT(_vfs_nfs, NFS_NFSSTATS, nfsstats, CTLFLAG_RD,
92 &nfsstats, nfsstats, "");
93 static int nfs_ip_paranoia = 1;
94 SYSCTL_INT(_vfs_nfs, OID_AUTO, nfs_ip_paranoia, CTLFLAG_RW,
95 &nfs_ip_paranoia, 0, "");
98 SYSCTL_INT(_vfs_nfs, OID_AUTO, debug, CTLFLAG_RW, &nfs_debug, 0, "");
101 static int nfs_iosize __P((struct nfsmount *nmp));
102 static void nfs_decode_args __P((struct nfsmount *nmp,
103 struct nfs_args *argp));
104 static int mountnfs __P((struct nfs_args *,struct mount *,
105 struct sockaddr *,char *,char *,struct vnode **));
106 static int nfs_mount __P(( struct mount *mp, char *path, caddr_t data,
107 struct nameidata *ndp, struct proc *p));
108 static int nfs_unmount __P(( struct mount *mp, int mntflags,
110 static int nfs_root __P(( struct mount *mp, struct vnode **vpp));
111 static int nfs_statfs __P(( struct mount *mp, struct statfs *sbp,
113 static int nfs_sync __P(( struct mount *mp, int waitfor,
114 struct ucred *cred, struct proc *p));
117 * nfs vfs operations.
119 static struct vfsops nfs_vfsops = {
128 vfs_stdfhtovp, /* shouldn't happen */
130 vfs_stdvptofh, /* shouldn't happen */
135 VFS_SET(nfs_vfsops, nfs, VFCF_NETWORK);
138 * This structure must be filled in by a primary bootstrap or bootstrap
139 * server for a diskless/dataless machine. It is initialized below just
140 * to ensure that it is allocated to initialized data (.data not .bss).
142 struct nfs_diskless nfs_diskless = { { { 0 } } };
143 struct nfsv3_diskless nfsv3_diskless = { { { 0 } } };
144 int nfs_diskless_valid = 0;
146 SYSCTL_INT(_vfs_nfs, OID_AUTO, diskless_valid, CTLFLAG_RD,
147 &nfs_diskless_valid, 0, "");
149 SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_rootpath, CTLFLAG_RD,
150 nfsv3_diskless.root_hostnam, 0, "");
152 SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_rootaddr, CTLFLAG_RD,
153 &nfsv3_diskless.root_saddr, sizeof nfsv3_diskless.root_saddr,
154 "%Ssockaddr_in", "");
156 SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_swappath, CTLFLAG_RD,
157 nfsv3_diskless.swap_hostnam, 0, "");
159 SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_swapaddr, CTLFLAG_RD,
160 &nfsv3_diskless.swap_saddr, sizeof nfsv3_diskless.swap_saddr,
164 void nfsargs_ntoh __P((struct nfs_args *));
165 static int nfs_mountdiskless __P((char *, char *, int,
166 struct sockaddr_in *, struct nfs_args *,
167 struct proc *, struct vnode **,
169 static void nfs_convert_diskless __P((void));
170 static void nfs_convert_oargs __P((struct nfs_args *args,
171 struct onfs_args *oargs));
175 struct nfsmount* nmp;
180 * Calculate the size used for io buffers. Use the larger
181 * of the two sizes to minimise nfs requests but make sure
182 * that it is at least one VM page to avoid wasting buffer
185 iosize = max(nmp->nm_rsize, nmp->nm_wsize);
186 if (iosize < PAGE_SIZE) iosize = PAGE_SIZE;
191 nfs_convert_oargs(args, oargs)
192 struct nfs_args *args;
193 struct onfs_args *oargs;
195 args->version = NFS_ARGSVERSION;
196 args->addr = oargs->addr;
197 args->addrlen = oargs->addrlen;
198 args->sotype = oargs->sotype;
199 args->proto = oargs->proto;
200 args->fh = oargs->fh;
201 args->fhsize = oargs->fhsize;
202 args->flags = oargs->flags;
203 args->wsize = oargs->wsize;
204 args->rsize = oargs->rsize;
205 args->readdirsize = oargs->readdirsize;
206 args->timeo = oargs->timeo;
207 args->retrans = oargs->retrans;
208 args->maxgrouplist = oargs->maxgrouplist;
209 args->readahead = oargs->readahead;
210 args->leaseterm = oargs->leaseterm;
211 args->deadthresh = oargs->deadthresh;
212 args->hostname = oargs->hostname;
216 nfs_convert_diskless()
218 bcopy(&nfs_diskless.myif, &nfsv3_diskless.myif,
219 sizeof(struct ifaliasreq));
220 bcopy(&nfs_diskless.mygateway, &nfsv3_diskless.mygateway,
221 sizeof(struct sockaddr_in));
222 nfs_convert_oargs(&nfsv3_diskless.swap_args,&nfs_diskless.swap_args);
223 nfsv3_diskless.swap_fhsize = NFSX_V2FH;
224 bcopy(nfs_diskless.swap_fh,nfsv3_diskless.swap_fh,NFSX_V2FH);
225 bcopy(&nfs_diskless.swap_saddr,&nfsv3_diskless.swap_saddr,
226 sizeof(struct sockaddr_in));
227 bcopy(nfs_diskless.swap_hostnam,nfsv3_diskless.swap_hostnam, MNAMELEN);
228 nfsv3_diskless.swap_nblks = nfs_diskless.swap_nblks;
229 bcopy(&nfs_diskless.swap_ucred, &nfsv3_diskless.swap_ucred,
230 sizeof(struct ucred));
231 nfs_convert_oargs(&nfsv3_diskless.root_args,&nfs_diskless.root_args);
232 nfsv3_diskless.root_fhsize = NFSX_V2FH;
233 bcopy(nfs_diskless.root_fh,nfsv3_diskless.root_fh,NFSX_V2FH);
234 bcopy(&nfs_diskless.root_saddr,&nfsv3_diskless.root_saddr,
235 sizeof(struct sockaddr_in));
236 bcopy(nfs_diskless.root_hostnam,nfsv3_diskless.root_hostnam, MNAMELEN);
237 nfsv3_diskless.root_time = nfs_diskless.root_time;
238 bcopy(nfs_diskless.my_hostnam,nfsv3_diskless.my_hostnam,
240 nfs_diskless_valid = 3;
247 nfs_statfs(mp, sbp, p)
249 register struct statfs *sbp;
252 register struct vnode *vp;
253 register struct nfs_statfs *sfp;
255 register u_int32_t *tl;
256 register int32_t t1, t2;
257 caddr_t bpos, dpos, cp2;
258 struct nfsmount *nmp = VFSTONFS(mp);
259 int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr;
260 struct mbuf *mreq, *mrep, *md, *mb, *mb2;
266 sfp = (struct nfs_statfs *)0;
268 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
273 cred->cr_ngroups = 1;
274 if (v3 && (nmp->nm_state & NFSSTA_GOTFSINFO) == 0)
275 (void)nfs_fsinfo(nmp, vp, cred, p);
276 nfsstats.rpccnt[NFSPROC_FSSTAT]++;
277 nfsm_reqhead(vp, NFSPROC_FSSTAT, NFSX_FH(v3));
279 nfsm_request(vp, NFSPROC_FSSTAT, p, cred);
281 nfsm_postop_attr(vp, retattr);
287 nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
288 sbp->f_flags = nmp->nm_flag;
289 sbp->f_iosize = nfs_iosize(nmp);
291 sbp->f_bsize = NFS_FABLKSIZE;
292 tquad = fxdr_hyper(&sfp->sf_tbytes);
293 sbp->f_blocks = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
294 tquad = fxdr_hyper(&sfp->sf_fbytes);
295 sbp->f_bfree = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
296 tquad = fxdr_hyper(&sfp->sf_abytes);
297 sbp->f_bavail = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
298 sbp->f_files = (fxdr_unsigned(int32_t,
299 sfp->sf_tfiles.nfsuquad[1]) & 0x7fffffff);
300 sbp->f_ffree = (fxdr_unsigned(int32_t,
301 sfp->sf_ffiles.nfsuquad[1]) & 0x7fffffff);
303 sbp->f_bsize = fxdr_unsigned(int32_t, sfp->sf_bsize);
304 sbp->f_blocks = fxdr_unsigned(int32_t, sfp->sf_blocks);
305 sbp->f_bfree = fxdr_unsigned(int32_t, sfp->sf_bfree);
306 sbp->f_bavail = fxdr_unsigned(int32_t, sfp->sf_bavail);
310 if (sbp != &mp->mnt_stat) {
311 sbp->f_type = mp->mnt_vfc->vfc_typenum;
312 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
313 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
322 * nfs version 3 fsinfo rpc call
325 nfs_fsinfo(nmp, vp, cred, p)
326 register struct nfsmount *nmp;
327 register struct vnode *vp;
331 register struct nfsv3_fsinfo *fsp;
333 register int32_t t1, t2;
334 register u_int32_t *tl, pref, max;
335 caddr_t bpos, dpos, cp2;
336 int error = 0, retattr;
337 struct mbuf *mreq, *mrep, *md, *mb, *mb2;
340 nfsstats.rpccnt[NFSPROC_FSINFO]++;
341 nfsm_reqhead(vp, NFSPROC_FSINFO, NFSX_FH(1));
343 nfsm_request(vp, NFSPROC_FSINFO, p, cred);
344 nfsm_postop_attr(vp, retattr);
346 nfsm_dissect(fsp, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
347 pref = fxdr_unsigned(u_int32_t, fsp->fs_wtpref);
348 if (pref < nmp->nm_wsize && pref >= NFS_FABLKSIZE)
349 nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) &
350 ~(NFS_FABLKSIZE - 1);
351 max = fxdr_unsigned(u_int32_t, fsp->fs_wtmax);
352 if (max < nmp->nm_wsize && max > 0) {
353 nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1);
354 if (nmp->nm_wsize == 0)
357 pref = fxdr_unsigned(u_int32_t, fsp->fs_rtpref);
358 if (pref < nmp->nm_rsize && pref >= NFS_FABLKSIZE)
359 nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) &
360 ~(NFS_FABLKSIZE - 1);
361 max = fxdr_unsigned(u_int32_t, fsp->fs_rtmax);
362 if (max < nmp->nm_rsize && max > 0) {
363 nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1);
364 if (nmp->nm_rsize == 0)
367 pref = fxdr_unsigned(u_int32_t, fsp->fs_dtpref);
368 if (pref < nmp->nm_readdirsize && pref >= NFS_DIRBLKSIZ)
369 nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) &
370 ~(NFS_DIRBLKSIZ - 1);
371 if (max < nmp->nm_readdirsize && max > 0) {
372 nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1);
373 if (nmp->nm_readdirsize == 0)
374 nmp->nm_readdirsize = max;
376 maxfsize = fxdr_hyper(&fsp->fs_maxfilesize);
377 if (maxfsize > 0 && maxfsize < nmp->nm_maxfilesize)
378 nmp->nm_maxfilesize = maxfsize;
379 nmp->nm_state |= NFSSTA_GOTFSINFO;
386 * Mount a remote root fs via. nfs. This depends on the info in the
387 * nfs_diskless structure that has been filled in properly by some primary
389 * It goes something like this:
390 * - do enough of "ifconfig" by calling ifioctl() so that the system
391 * can talk to the server
392 * - If nfs_diskless.mygateway is filled in, use that address as
394 * - build the rootfs mount point and call mountnfs() to do the rest.
400 struct mount *swap_mp;
401 struct nfsv3_diskless *nd = &nfsv3_diskless;
404 struct proc *p = curproc; /* XXX */
409 #if defined(BOOTP_NFSROOT) && defined(BOOTP)
410 bootpc_init(); /* use bootp to get nfs_diskless filled in */
414 * XXX time must be non-zero when we init the interface or else
415 * the arp code will wedge...
417 while (time_second == 0)
418 tsleep(&time_second, PZERO+8, "arpkludge", 10);
420 if (nfs_diskless_valid==1)
421 nfs_convert_diskless();
424 * XXX splnet, so networks will receive...
429 /* Set up swap credentials. */
430 proc0.p_ucred->cr_uid = ntohl(nd->swap_ucred.cr_uid);
431 proc0.p_ucred->cr_gid = ntohl(nd->swap_ucred.cr_gid);
432 if ((proc0.p_ucred->cr_ngroups = ntohs(nd->swap_ucred.cr_ngroups)) >
434 proc0.p_ucred->cr_ngroups = NGROUPS;
435 for (i = 0; i < proc0.p_ucred->cr_ngroups; i++)
436 proc0.p_ucred->cr_groups[i] = ntohl(nd->swap_ucred.cr_groups[i]);
440 * Do enough of ifconfig(8) so that the critical net interface can
441 * talk to the server.
443 error = socreate(nd->myif.ifra_addr.sa_family, &so, SOCK_DGRAM, 0, p);
445 panic("nfs_mountroot: socreate(%04x): %d",
446 nd->myif.ifra_addr.sa_family, error);
448 #if 0 /* XXX Bad idea */
450 * We might not have been told the right interface, so we pass
451 * over the first ten interfaces of the same kind, until we get
452 * one of them configured.
455 for (i = strlen(nd->myif.ifra_name) - 1;
456 nd->myif.ifra_name[i] >= '0' &&
457 nd->myif.ifra_name[i] <= '9';
458 nd->myif.ifra_name[i] ++) {
459 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, p);
464 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, p);
466 panic("nfs_mountroot: SIOCAIFADDR: %d", error);
470 * If the gateway field is filled in, set it as the default route.
472 if (nd->mygateway.sin_len != 0) {
473 struct sockaddr_in mask, sin;
475 bzero((caddr_t)&mask, sizeof(mask));
477 sin.sin_family = AF_INET;
478 sin.sin_len = sizeof(sin);
479 error = rtrequest(RTM_ADD, (struct sockaddr *)&sin,
480 (struct sockaddr *)&nd->mygateway,
481 (struct sockaddr *)&mask,
482 RTF_UP | RTF_GATEWAY, (struct rtentry **)0);
484 panic("nfs_mountroot: RTM_ADD: %d", error);
488 * Create the rootfs mount point.
490 nd->root_args.fh = nd->root_fh;
491 nd->root_args.fhsize = nd->root_fhsize;
492 l = ntohl(nd->root_saddr.sin_addr.s_addr);
493 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
494 (l >> 24) & 0xff, (l >> 16) & 0xff,
495 (l >> 8) & 0xff, (l >> 0) & 0xff,nd->root_hostnam);
496 printf("NFS ROOT: %s\n",buf);
497 if ((error = nfs_mountdiskless(buf, "/", MNT_RDONLY,
498 &nd->root_saddr, &nd->root_args, p, &vp, &mp)) != 0) {
500 mp->mnt_vfc->vfc_refcount--;
501 free(swap_mp, M_MOUNT);
507 if (nd->swap_nblks) {
509 /* Convert to DEV_BSIZE instead of Kilobyte */
513 * Create a fake mount point just for the swap vnode so that the
514 * swap file can be on a different server from the rootfs.
516 nd->swap_args.fh = nd->swap_fh;
517 nd->swap_args.fhsize = nd->swap_fhsize;
518 l = ntohl(nd->swap_saddr.sin_addr.s_addr);
519 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
520 (l >> 24) & 0xff, (l >> 16) & 0xff,
521 (l >> 8) & 0xff, (l >> 0) & 0xff,nd->swap_hostnam);
522 printf("NFS SWAP: %s\n",buf);
523 if ((error = nfs_mountdiskless(buf, "/swap", 0,
524 &nd->swap_saddr, &nd->swap_args, p, &vp, &swap_mp)) != 0)
526 vfs_unbusy(swap_mp, p);
528 VTONFS(vp)->n_size = VTONFS(vp)->n_vattr.va_size =
529 nd->swap_nblks * DEV_BSIZE ;
532 * Since the swap file is not the root dir of a file system,
533 * hack it to a regular file.
538 swaponvp(p, vp, NODEV, nd->swap_nblks);
541 mp->mnt_flag |= MNT_ROOTFS;
542 mp->mnt_vnodecovered = NULLVP;
547 * This is not really an nfs issue, but it is much easier to
548 * set hostname here and then let the "/etc/rc.xxx" files
549 * mount the right /var based upon its preset value.
551 bcopy(nd->my_hostnam, hostname, MAXHOSTNAMELEN);
552 hostname[MAXHOSTNAMELEN - 1] = '\0';
553 for (i = 0; i < MAXHOSTNAMELEN; i++)
554 if (hostname[i] == '\0')
556 inittodr(ntohl(nd->root_time));
561 * Internal version of mount system call for diskless setup.
564 nfs_mountdiskless(path, which, mountflag, sin, args, p, vpp, mpp)
568 struct sockaddr_in *sin;
569 struct nfs_args *args;
575 struct sockaddr *nam;
582 if ((error = vfs_rootmountalloc("nfs", path, &mp)) != 0) {
583 printf("nfs_mountroot: NFS not configured");
589 mp->mnt_kern_flag = 0;
590 mp->mnt_flag = mountflag;
591 nam = dup_sockaddr((struct sockaddr *)sin, 1);
592 if ((error = mountnfs(args, mp, nam, which, path, vpp)) != 0) {
593 printf("nfs_mountroot: mount %s on %s: %d", path, which, error);
594 mp->mnt_vfc->vfc_refcount--;
601 (void) copystr(which, mp->mnt_stat.f_mntonname, MNAMELEN - 1, 0);
607 nfs_decode_args(nmp, argp)
608 struct nfsmount *nmp;
609 struct nfs_args *argp;
617 * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes
618 * no sense in that context.
620 if (argp->sotype == SOCK_STREAM)
621 nmp->nm_flag &= ~NFSMNT_NOCONN;
623 /* Also clear RDIRPLUS if not NFSv3, it crashes some servers */
624 if ((argp->flags & NFSMNT_NFSV3) == 0)
625 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
627 /* Re-bind if rsrvd port requested and wasn't on one */
628 adjsock = !(nmp->nm_flag & NFSMNT_RESVPORT)
629 && (argp->flags & NFSMNT_RESVPORT);
630 /* Also re-bind if we're switching to/from a connected UDP socket */
631 adjsock |= ((nmp->nm_flag & NFSMNT_NOCONN) !=
632 (argp->flags & NFSMNT_NOCONN));
634 /* Update flags atomically. Don't change the lock bits. */
635 nmp->nm_flag = argp->flags | nmp->nm_flag;
638 if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) {
639 nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10;
640 if (nmp->nm_timeo < NFS_MINTIMEO)
641 nmp->nm_timeo = NFS_MINTIMEO;
642 else if (nmp->nm_timeo > NFS_MAXTIMEO)
643 nmp->nm_timeo = NFS_MAXTIMEO;
646 if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) {
647 nmp->nm_retry = argp->retrans;
648 if (nmp->nm_retry > NFS_MAXREXMIT)
649 nmp->nm_retry = NFS_MAXREXMIT;
652 if (argp->flags & NFSMNT_NFSV3) {
653 if (argp->sotype == SOCK_DGRAM)
654 maxio = NFS_MAXDGRAMDATA;
658 maxio = NFS_V2MAXDATA;
660 if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
661 nmp->nm_wsize = argp->wsize;
662 /* Round down to multiple of blocksize */
663 nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1);
664 if (nmp->nm_wsize <= 0)
665 nmp->nm_wsize = NFS_FABLKSIZE;
667 if (nmp->nm_wsize > maxio)
668 nmp->nm_wsize = maxio;
669 if (nmp->nm_wsize > MAXBSIZE)
670 nmp->nm_wsize = MAXBSIZE;
672 if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
673 nmp->nm_rsize = argp->rsize;
674 /* Round down to multiple of blocksize */
675 nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1);
676 if (nmp->nm_rsize <= 0)
677 nmp->nm_rsize = NFS_FABLKSIZE;
679 if (nmp->nm_rsize > maxio)
680 nmp->nm_rsize = maxio;
681 if (nmp->nm_rsize > MAXBSIZE)
682 nmp->nm_rsize = MAXBSIZE;
684 if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) {
685 nmp->nm_readdirsize = argp->readdirsize;
687 if (nmp->nm_readdirsize > maxio)
688 nmp->nm_readdirsize = maxio;
689 if (nmp->nm_readdirsize > nmp->nm_rsize)
690 nmp->nm_readdirsize = nmp->nm_rsize;
692 if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0)
693 nmp->nm_acregmin = argp->acregmin;
695 nmp->nm_acregmin = NFS_MINATTRTIMO;
696 if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0)
697 nmp->nm_acregmax = argp->acregmax;
699 nmp->nm_acregmax = NFS_MAXATTRTIMO;
700 if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0)
701 nmp->nm_acdirmin = argp->acdirmin;
703 nmp->nm_acdirmin = NFS_MINDIRATTRTIMO;
704 if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0)
705 nmp->nm_acdirmax = argp->acdirmax;
707 nmp->nm_acdirmax = NFS_MAXDIRATTRTIMO;
708 if (nmp->nm_acdirmin > nmp->nm_acdirmax)
709 nmp->nm_acdirmin = nmp->nm_acdirmax;
710 if (nmp->nm_acregmin > nmp->nm_acregmax)
711 nmp->nm_acregmin = nmp->nm_acregmax;
713 if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0) {
714 if (argp->maxgrouplist <= NFS_MAXGRPS)
715 nmp->nm_numgrps = argp->maxgrouplist;
717 nmp->nm_numgrps = NFS_MAXGRPS;
719 if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0) {
720 if (argp->readahead <= NFS_MAXRAHEAD)
721 nmp->nm_readahead = argp->readahead;
723 nmp->nm_readahead = NFS_MAXRAHEAD;
725 if ((argp->flags & NFSMNT_LEASETERM) && argp->leaseterm >= 2) {
726 if (argp->leaseterm <= NQ_MAXLEASE)
727 nmp->nm_leaseterm = argp->leaseterm;
729 nmp->nm_leaseterm = NQ_MAXLEASE;
731 if ((argp->flags & NFSMNT_DEADTHRESH) && argp->deadthresh >= 1) {
732 if (argp->deadthresh <= NQ_NEVERDEAD)
733 nmp->nm_deadthresh = argp->deadthresh;
735 nmp->nm_deadthresh = NQ_NEVERDEAD;
738 adjsock |= ((nmp->nm_sotype != argp->sotype) ||
739 (nmp->nm_soproto != argp->proto));
740 nmp->nm_sotype = argp->sotype;
741 nmp->nm_soproto = argp->proto;
743 if (nmp->nm_so && adjsock) {
744 nfs_safedisconnect(nmp);
745 if (nmp->nm_sotype == SOCK_DGRAM)
746 while (nfs_connect(nmp, (struct nfsreq *)0)) {
747 printf("nfs_args: retrying connect\n");
748 (void) tsleep((caddr_t)&lbolt,
758 * It seems a bit dumb to copyinstr() the host and path here and then
759 * bcopy() them in mountnfs(), but I wanted to detect errors before
760 * doing the sockargs() call because sockargs() allocates an mbuf and
761 * an error after that means that I have to release the mbuf.
765 nfs_mount(mp, path, data, ndp, p)
769 struct nameidata *ndp;
773 struct nfs_args args;
774 struct sockaddr *nam;
776 char pth[MNAMELEN], hst[MNAMELEN];
778 u_char nfh[NFSX_V3FHMAX];
784 error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args));
787 if (args.version != NFS_ARGSVERSION) {
788 #ifdef COMPAT_PRELITE2
790 * If the argument version is unknown, then assume the
791 * caller is a pre-lite2 4.4BSD client and convert its
794 struct onfs_args oargs;
795 error = copyin(data, (caddr_t)&oargs, sizeof (struct onfs_args));
798 nfs_convert_oargs(&args,&oargs);
799 #else /* !COMPAT_PRELITE2 */
800 return (EPROGMISMATCH);
801 #endif /* COMPAT_PRELITE2 */
803 if (mp->mnt_flag & MNT_UPDATE) {
804 register struct nfsmount *nmp = VFSTONFS(mp);
809 * When doing an update, we can't change from or to
810 * v3 and/or nqnfs, or change cookie translation
812 args.flags = (args.flags &
813 ~(NFSMNT_NFSV3|NFSMNT_NQNFS /*|NFSMNT_XLATECOOKIE*/)) |
815 (NFSMNT_NFSV3|NFSMNT_NQNFS /*|NFSMNT_XLATECOOKIE*/));
816 nfs_decode_args(nmp, &args);
821 * Make the nfs_ip_paranoia sysctl serve as the default connection
822 * or no-connection mode for those protocols that support
823 * no-connection mode (the flag will be cleared later for protocols
824 * that do not support no-connection mode). This will allow a client
825 * to receive replies from a different IP then the request was
826 * sent to. Note: default value for nfs_ip_paranoia is 1 (paranoid),
829 if (nfs_ip_paranoia == 0)
830 args.flags |= NFSMNT_NOCONN;
831 if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX)
833 error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize);
836 error = copyinstr(path, pth, MNAMELEN-1, &len);
839 bzero(&pth[len], MNAMELEN - len);
840 error = copyinstr(args.hostname, hst, MNAMELEN-1, &len);
843 bzero(&hst[len], MNAMELEN - len);
844 /* sockargs() call must be after above copyin() calls */
845 error = getsockaddr(&nam, (caddr_t)args.addr, args.addrlen);
849 error = mountnfs(&args, mp, nam, pth, hst, &vp);
854 * Common code for mount and mountroot
857 mountnfs(argp, mp, nam, pth, hst, vpp)
858 register struct nfs_args *argp;
859 register struct mount *mp;
860 struct sockaddr *nam;
864 register struct nfsmount *nmp;
869 if (mp->mnt_flag & MNT_UPDATE) {
871 /* update paths, file handles, etc, here XXX */
875 nmp = zalloc(nfsmount_zone);
876 bzero((caddr_t)nmp, sizeof (struct nfsmount));
877 TAILQ_INIT(&nmp->nm_uidlruhead);
878 TAILQ_INIT(&nmp->nm_bufq);
879 mp->mnt_data = (qaddr_t)nmp;
883 if (argp->flags & NFSMNT_NQNFS)
885 * We have to set mnt_maxsymlink to a non-zero value so
886 * that COMPAT_43 routines will know that we are setting
887 * the d_type field in directories (and can zero it for
888 * unsuspecting binaries).
890 mp->mnt_maxsymlinklen = 1;
893 * V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too
894 * high, depending on whether we end up with negative offsets in
895 * the client or server somewhere. 2GB-1 may be safer.
897 * For V3, nfs_fsinfo will adjust this as necessary. Assume maximum
898 * that we can handle until we find out otherwise.
899 * XXX Our "safe" limit on the client is what we can store in our
900 * buffer cache using signed(!) block numbers.
902 if ((argp->flags & NFSMNT_NFSV3) == 0)
903 nmp->nm_maxfilesize = 0xffffffffLL;
905 nmp->nm_maxfilesize = (u_int64_t)0x80000000 * DEV_BSIZE - 1;
907 nmp->nm_timeo = NFS_TIMEO;
908 nmp->nm_retry = NFS_RETRANS;
909 nmp->nm_wsize = NFS_WSIZE;
910 nmp->nm_rsize = NFS_RSIZE;
911 nmp->nm_readdirsize = NFS_READDIRSIZE;
912 nmp->nm_numgrps = NFS_MAXGRPS;
913 nmp->nm_readahead = NFS_DEFRAHEAD;
914 nmp->nm_leaseterm = NQ_DEFLEASE;
915 nmp->nm_deadthresh = NQ_DEADTHRESH;
916 CIRCLEQ_INIT(&nmp->nm_timerhead);
917 nmp->nm_inprog = NULLVP;
918 nmp->nm_fhsize = argp->fhsize;
919 bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize);
920 bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN);
921 bcopy(pth, mp->mnt_stat.f_mntonname, MNAMELEN);
923 /* Set up the sockets and per-host congestion */
924 nmp->nm_sotype = argp->sotype;
925 nmp->nm_soproto = argp->proto;
927 nfs_decode_args(nmp, argp);
930 * For Connection based sockets (TCP,...) defer the connect until
931 * the first request, in case the server is not responding.
933 if (nmp->nm_sotype == SOCK_DGRAM &&
934 (error = nfs_connect(nmp, (struct nfsreq *)0)))
938 * This is silly, but it has to be set so that vinifod() works.
939 * We do not want to do an nfs_statfs() here since we can get
940 * stuck on a dead server and we are holding a lock on the mount
943 mp->mnt_stat.f_iosize = nfs_iosize(nmp);
945 * A reference count is needed on the nfsnode representing the
946 * remote root. If this object is not persistent, then backward
947 * traversals of the mount point (i.e. "..") will not work if
948 * the nfsnode gets flushed out of the cache. Ufs does not have
949 * this problem, because one can identify root inodes by their
950 * number == ROOTINO (2).
952 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
958 * Get file attributes for the mountpoint. This has the side
959 * effect of filling in (*vpp)->v_type with the correct value.
961 VOP_GETATTR(*vpp, &attrs, curproc->p_ucred, curproc);
964 * Lose the lock but keep the ref.
966 VOP_UNLOCK(*vpp, 0, curproc);
971 zfree(nfsmount_zone, nmp);
977 * unmount system call
980 nfs_unmount(mp, mntflags, p)
985 register struct nfsmount *nmp;
986 int error, flags = 0;
988 if (mntflags & MNT_FORCE)
992 * Goes something like this..
993 * - Call vflush() to clear out vnodes for this file system
995 * - Free up the data structures
997 /* In the forced case, cancel any outstanding requests. */
998 if (flags & FORCECLOSE) {
999 error = nfs_nmcancelreqs(nmp);
1004 * Must handshake with nqnfs_clientd() if it is active.
1006 nmp->nm_state |= NFSSTA_DISMINPROG;
1007 while (nmp->nm_inprog != NULLVP)
1008 (void) tsleep((caddr_t)&lbolt, PSOCK, "nfsdism", 0);
1010 /* We hold 1 extra ref on the root vnode; see comment in mountnfs(). */
1011 error = vflush(mp, 1, flags);
1013 nmp->nm_state &= ~NFSSTA_DISMINPROG;
1018 * We are now committed to the unmount.
1019 * For NQNFS, let the server daemon free the nfsmount structure.
1021 if (nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_KERB))
1022 nmp->nm_state |= NFSSTA_DISMNT;
1024 nfs_disconnect(nmp);
1025 FREE(nmp->nm_nam, M_SONAME);
1027 if ((nmp->nm_flag & (NFSMNT_NQNFS | NFSMNT_KERB)) == 0)
1028 zfree(nfsmount_zone, nmp);
1033 * Return root of a filesystem
1040 register struct vnode *vp;
1041 struct nfsmount *nmp;
1046 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
1050 if (vp->v_type == VNON)
1060 * Flush out the buffer cache
1064 nfs_sync(mp, waitfor, cred, p)
1070 register struct vnode *vp;
1071 int error, allerror = 0;
1074 * Force stale buffer cache information to be flushed.
1077 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1079 vp = TAILQ_NEXT(vp, v_nmntvnodes)) {
1081 * If the vnode that we are about to sync is no longer
1082 * associated with this mount point, start over.
1084 if (vp->v_mount != mp)
1086 if (VOP_ISLOCKED(vp, NULL) || TAILQ_EMPTY(&vp->v_dirtyblkhd) ||
1087 waitfor == MNT_LAZY)
1089 if (vget(vp, LK_EXCLUSIVE, p))
1091 error = VOP_FSYNC(vp, cred, waitfor, p);