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
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
42 #include "opt_bootp.h"
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
48 #include <sys/malloc.h>
50 #include <sys/module.h>
51 #include <sys/mount.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sockio.h>
56 #include <sys/sysctl.h>
57 #include <sys/vnode.h>
60 #include <vm/vm_extern.h>
64 #include <net/route.h>
65 #include <netinet/in.h>
67 #include <nfs/rpcv2.h>
68 #include <nfs/nfsproto.h>
69 #include <nfsclient/nfs.h>
70 #include <nfsclient/nfsnode.h>
71 #include <nfsclient/nfsmount.h>
72 #include <nfs/xdr_subs.h>
73 #include <nfsclient/nfsm_subs.h>
74 #include <nfsclient/nfsdiskless.h>
76 MALLOC_DEFINE(M_NFSREQ, "NFS req", "NFS request header");
77 MALLOC_DEFINE(M_NFSBIGFH, "NFSV3 bigfh", "NFS version 3 file handle");
78 MALLOC_DEFINE(M_NFSDIROFF, "NFSV3 diroff", "NFS directory offset data");
79 MALLOC_DEFINE(M_NFSHASH, "NFS hash", "NFS hash tables");
81 uma_zone_t nfsmount_zone;
83 struct nfsstats nfsstats;
84 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem");
85 SYSCTL_STRUCT(_vfs_nfs, NFS_NFSSTATS, nfsstats, CTLFLAG_RD,
86 &nfsstats, nfsstats, "S,nfsstats");
89 SYSCTL_INT(_vfs_nfs, OID_AUTO, debug, CTLFLAG_RW, &nfs_debug, 0, "");
92 static int nfs_iosize(struct nfsmount *nmp);
93 static void nfs_decode_args(struct nfsmount *nmp, struct nfs_args *argp);
94 static int mountnfs(struct nfs_args *, struct mount *,
95 struct sockaddr *, char *, char *, struct vnode **,
97 static int nfs_mount(struct mount *mp, char *path, caddr_t data,
98 struct nameidata *ndp, struct thread *td);
99 static int nfs_unmount(struct mount *mp, int mntflags, struct thread *td);
100 static int nfs_root(struct mount *mp, struct vnode **vpp);
101 static int nfs_statfs(struct mount *mp, struct statfs *sbp,
103 static int nfs_sync(struct mount *mp, int waitfor, struct ucred *cred,
107 * nfs vfs operations.
109 static struct vfsops nfs_vfsops = {
118 vfs_stdfhtovp, /* shouldn't happen */
120 vfs_stdvptofh, /* shouldn't happen */
125 VFS_SET(nfs_vfsops, nfs, VFCF_NETWORK);
127 /* So that loader and kldload(2) can find us, wherever we are.. */
128 MODULE_VERSION(nfs, 1);
131 * This structure must be filled in by a primary bootstrap or bootstrap
132 * server for a diskless/dataless machine. It is initialized below just
133 * to ensure that it is allocated to initialized data (.data not .bss).
135 struct nfs_diskless nfs_diskless = { { { 0 } } };
136 struct nfsv3_diskless nfsv3_diskless = { { { 0 } } };
137 int nfs_diskless_valid = 0;
139 SYSCTL_INT(_vfs_nfs, OID_AUTO, diskless_valid, CTLFLAG_RD,
140 &nfs_diskless_valid, 0, "");
142 SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_rootpath, CTLFLAG_RD,
143 nfsv3_diskless.root_hostnam, 0, "");
145 SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_rootaddr, CTLFLAG_RD,
146 &nfsv3_diskless.root_saddr, sizeof nfsv3_diskless.root_saddr,
147 "%Ssockaddr_in", "");
149 SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_swappath, CTLFLAG_RD,
150 nfsv3_diskless.swap_hostnam, 0, "");
152 SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_swapaddr, CTLFLAG_RD,
153 &nfsv3_diskless.swap_saddr, sizeof nfsv3_diskless.swap_saddr,
157 void nfsargs_ntoh(struct nfs_args *);
158 static int nfs_mountdiskless(char *, char *, int,
159 struct sockaddr_in *, struct nfs_args *,
160 struct thread *, struct vnode **, struct mount **);
161 static void nfs_convert_diskless(void);
162 static void nfs_convert_oargs(struct nfs_args *args,
163 struct onfs_args *oargs);
166 nfs_iosize(struct nfsmount *nmp)
171 * Calculate the size used for io buffers. Use the larger
172 * of the two sizes to minimise nfs requests but make sure
173 * that it is at least one VM page to avoid wasting buffer
176 iosize = max(nmp->nm_rsize, nmp->nm_wsize);
177 if (iosize < PAGE_SIZE) iosize = PAGE_SIZE;
182 nfs_convert_oargs(struct nfs_args *args, struct onfs_args *oargs)
185 args->version = NFS_ARGSVERSION;
186 args->addr = oargs->addr;
187 args->addrlen = oargs->addrlen;
188 args->sotype = oargs->sotype;
189 args->proto = oargs->proto;
190 args->fh = oargs->fh;
191 args->fhsize = oargs->fhsize;
192 args->flags = oargs->flags;
193 args->wsize = oargs->wsize;
194 args->rsize = oargs->rsize;
195 args->readdirsize = oargs->readdirsize;
196 args->timeo = oargs->timeo;
197 args->retrans = oargs->retrans;
198 args->maxgrouplist = oargs->maxgrouplist;
199 args->readahead = oargs->readahead;
200 args->deadthresh = oargs->deadthresh;
201 args->hostname = oargs->hostname;
205 nfs_convert_diskless(void)
208 bcopy(&nfs_diskless.myif, &nfsv3_diskless.myif,
209 sizeof(struct ifaliasreq));
210 bcopy(&nfs_diskless.mygateway, &nfsv3_diskless.mygateway,
211 sizeof(struct sockaddr_in));
212 nfs_convert_oargs(&nfsv3_diskless.swap_args,&nfs_diskless.swap_args);
213 nfsv3_diskless.swap_fhsize = NFSX_V2FH;
214 bcopy(nfs_diskless.swap_fh, nfsv3_diskless.swap_fh, NFSX_V2FH);
215 bcopy(&nfs_diskless.swap_saddr,&nfsv3_diskless.swap_saddr,
216 sizeof(struct sockaddr_in));
217 bcopy(nfs_diskless.swap_hostnam, nfsv3_diskless.swap_hostnam, MNAMELEN);
218 nfsv3_diskless.swap_nblks = nfs_diskless.swap_nblks;
219 bcopy(&nfs_diskless.swap_ucred, &nfsv3_diskless.swap_ucred,
220 sizeof(struct ucred));
221 nfs_convert_oargs(&nfsv3_diskless.root_args,&nfs_diskless.root_args);
222 nfsv3_diskless.root_fhsize = NFSX_V2FH;
223 bcopy(nfs_diskless.root_fh, nfsv3_diskless.root_fh, NFSX_V2FH);
224 bcopy(&nfs_diskless.root_saddr,&nfsv3_diskless.root_saddr,
225 sizeof(struct sockaddr_in));
226 bcopy(nfs_diskless.root_hostnam, nfsv3_diskless.root_hostnam, MNAMELEN);
227 nfsv3_diskless.root_time = nfs_diskless.root_time;
228 bcopy(nfs_diskless.my_hostnam, nfsv3_diskless.my_hostnam,
230 nfs_diskless_valid = 3;
237 nfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
240 struct nfs_statfs *sfp;
242 struct nfsmount *nmp = VFSTONFS(mp);
243 int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr;
244 struct mbuf *mreq, *mrep, *md, *mb;
249 sfp = (struct nfs_statfs *)0;
251 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
255 if (v3 && (nmp->nm_state & NFSSTA_GOTFSINFO) == 0)
256 (void)nfs_fsinfo(nmp, vp, td->td_ucred, td);
257 nfsstats.rpccnt[NFSPROC_FSSTAT]++;
258 mreq = nfsm_reqhead(vp, NFSPROC_FSSTAT, NFSX_FH(v3));
260 bpos = mtod(mb, caddr_t);
262 nfsm_request(vp, NFSPROC_FSSTAT, td, td->td_ucred);
264 nfsm_postop_attr(vp, retattr);
270 sfp = nfsm_dissect(struct nfs_statfs *, NFSX_STATFS(v3));
271 sbp->f_flags = nmp->nm_flag;
272 sbp->f_iosize = nfs_iosize(nmp);
274 sbp->f_bsize = NFS_FABLKSIZE;
275 tquad = fxdr_hyper(&sfp->sf_tbytes);
276 sbp->f_blocks = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
277 tquad = fxdr_hyper(&sfp->sf_fbytes);
278 sbp->f_bfree = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
279 tquad = fxdr_hyper(&sfp->sf_abytes);
280 sbp->f_bavail = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
281 sbp->f_files = (fxdr_unsigned(int32_t,
282 sfp->sf_tfiles.nfsuquad[1]) & 0x7fffffff);
283 sbp->f_ffree = (fxdr_unsigned(int32_t,
284 sfp->sf_ffiles.nfsuquad[1]) & 0x7fffffff);
286 sbp->f_bsize = fxdr_unsigned(int32_t, sfp->sf_bsize);
287 sbp->f_blocks = fxdr_unsigned(int32_t, sfp->sf_blocks);
288 sbp->f_bfree = fxdr_unsigned(int32_t, sfp->sf_bfree);
289 sbp->f_bavail = fxdr_unsigned(int32_t, sfp->sf_bavail);
293 if (sbp != &mp->mnt_stat) {
294 sbp->f_type = mp->mnt_vfc->vfc_typenum;
295 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
296 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
305 * nfs version 3 fsinfo rpc call
308 nfs_fsinfo(struct nfsmount *nmp, struct vnode *vp, struct ucred *cred,
311 struct nfsv3_fsinfo *fsp;
314 int error = 0, retattr;
315 struct mbuf *mreq, *mrep, *md, *mb;
318 nfsstats.rpccnt[NFSPROC_FSINFO]++;
319 mreq = nfsm_reqhead(vp, NFSPROC_FSINFO, NFSX_FH(1));
321 bpos = mtod(mb, caddr_t);
323 nfsm_request(vp, NFSPROC_FSINFO, td, cred);
324 nfsm_postop_attr(vp, retattr);
326 fsp = nfsm_dissect(struct nfsv3_fsinfo *, NFSX_V3FSINFO);
327 pref = fxdr_unsigned(u_int32_t, fsp->fs_wtpref);
328 if (pref < nmp->nm_wsize && pref >= NFS_FABLKSIZE)
329 nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) &
330 ~(NFS_FABLKSIZE - 1);
331 max = fxdr_unsigned(u_int32_t, fsp->fs_wtmax);
332 if (max < nmp->nm_wsize && max > 0) {
333 nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1);
334 if (nmp->nm_wsize == 0)
337 pref = fxdr_unsigned(u_int32_t, fsp->fs_rtpref);
338 if (pref < nmp->nm_rsize && pref >= NFS_FABLKSIZE)
339 nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) &
340 ~(NFS_FABLKSIZE - 1);
341 max = fxdr_unsigned(u_int32_t, fsp->fs_rtmax);
342 if (max < nmp->nm_rsize && max > 0) {
343 nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1);
344 if (nmp->nm_rsize == 0)
347 pref = fxdr_unsigned(u_int32_t, fsp->fs_dtpref);
348 if (pref < nmp->nm_readdirsize && pref >= NFS_DIRBLKSIZ)
349 nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) &
350 ~(NFS_DIRBLKSIZ - 1);
351 if (max < nmp->nm_readdirsize && max > 0) {
352 nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1);
353 if (nmp->nm_readdirsize == 0)
354 nmp->nm_readdirsize = max;
356 maxfsize = fxdr_hyper(&fsp->fs_maxfilesize);
357 if (maxfsize > 0 && maxfsize < nmp->nm_maxfilesize)
358 nmp->nm_maxfilesize = maxfsize;
359 nmp->nm_state |= NFSSTA_GOTFSINFO;
367 * Mount a remote root fs via. nfs. This depends on the info in the
368 * nfs_diskless structure that has been filled in properly by some primary
370 * It goes something like this:
371 * - do enough of "ifconfig" by calling ifioctl() so that the system
372 * can talk to the server
373 * - If nfs_diskless.mygateway is filled in, use that address as
375 * - build the rootfs mount point and call mountnfs() to do the rest.
378 nfs_mountroot(struct mount *mp, struct thread *td)
380 struct mount *swap_mp;
381 struct nfsv3_diskless *nd = &nfsv3_diskless;
388 #if defined(BOOTP_NFSROOT) && defined(BOOTP)
389 bootpc_init(); /* use bootp to get nfs_diskless filled in */
392 if (nfs_diskless_valid==1)
393 nfs_convert_diskless();
396 * XXX splnet, so networks will receive...
401 /* Set up swap credentials. */
402 proc0.p_ucred->cr_uid = ntohl(nd->swap_ucred.cr_uid);
403 proc0.p_ucred->cr_gid = ntohl(nd->swap_ucred.cr_gid);
404 if ((proc0.p_ucred->cr_ngroups = ntohs(nd->swap_ucred.cr_ngroups)) >
406 proc0.p_ucred->cr_ngroups = NGROUPS;
407 for (i = 0; i < proc0.p_ucred->cr_ngroups; i++)
408 proc0.p_ucred->cr_groups[i] = ntohl(nd->swap_ucred.cr_groups[i]);
412 * Do enough of ifconfig(8) so that the critical net interface can
413 * talk to the server.
415 error = socreate(nd->myif.ifra_addr.sa_family, &so, SOCK_DGRAM, 0,
418 panic("nfs_mountroot: socreate(%04x): %d",
419 nd->myif.ifra_addr.sa_family, error);
421 #if 0 /* XXX Bad idea */
423 * We might not have been told the right interface, so we pass
424 * over the first ten interfaces of the same kind, until we get
425 * one of them configured.
428 for (i = strlen(nd->myif.ifra_name) - 1;
429 nd->myif.ifra_name[i] >= '0' &&
430 nd->myif.ifra_name[i] <= '9';
431 nd->myif.ifra_name[i] ++) {
432 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td);
437 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td);
439 panic("nfs_mountroot: SIOCAIFADDR: %d", error);
443 * If the gateway field is filled in, set it as the default route.
445 if (nd->mygateway.sin_len != 0) {
446 struct sockaddr_in mask, sin;
448 bzero((caddr_t)&mask, sizeof(mask));
450 sin.sin_family = AF_INET;
451 sin.sin_len = sizeof(sin);
452 error = rtrequest(RTM_ADD, (struct sockaddr *)&sin,
453 (struct sockaddr *)&nd->mygateway,
454 (struct sockaddr *)&mask,
455 RTF_UP | RTF_GATEWAY, (struct rtentry **)0);
457 panic("nfs_mountroot: RTM_ADD: %d", error);
461 * Create the rootfs mount point.
463 nd->root_args.fh = nd->root_fh;
464 nd->root_args.fhsize = nd->root_fhsize;
465 l = ntohl(nd->root_saddr.sin_addr.s_addr);
466 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
467 (l >> 24) & 0xff, (l >> 16) & 0xff,
468 (l >> 8) & 0xff, (l >> 0) & 0xff, nd->root_hostnam);
469 printf("NFS ROOT: %s\n", buf);
470 if ((error = nfs_mountdiskless(buf, "/", MNT_RDONLY,
471 &nd->root_saddr, &nd->root_args, td, &vp, &mp)) != 0) {
473 mp->mnt_vfc->vfc_refcount--;
474 free(swap_mp, M_MOUNT);
480 if (nd->swap_nblks) {
482 /* Convert to DEV_BSIZE instead of Kilobyte */
486 * Create a fake mount point just for the swap vnode so that the
487 * swap file can be on a different server from the rootfs.
489 nd->swap_args.fh = nd->swap_fh;
490 nd->swap_args.fhsize = nd->swap_fhsize;
491 l = ntohl(nd->swap_saddr.sin_addr.s_addr);
492 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
493 (l >> 24) & 0xff, (l >> 16) & 0xff,
494 (l >> 8) & 0xff, (l >> 0) & 0xff, nd->swap_hostnam);
495 printf("NFS SWAP: %s\n", buf);
496 if ((error = nfs_mountdiskless(buf, "/swap", 0,
497 &nd->swap_saddr, &nd->swap_args, td, &vp, &swap_mp)) != 0)
499 vfs_unbusy(swap_mp, td);
501 VTONFS(vp)->n_size = VTONFS(vp)->n_vattr.va_size =
502 nd->swap_nblks * DEV_BSIZE ;
504 * Since the swap file is not the root dir of a filesystem,
505 * hack it to a regular file.
510 swaponvp(td, vp, NODEV, nd->swap_nblks);
513 mp->mnt_flag |= MNT_ROOTFS;
514 mp->mnt_vnodecovered = NULLVP;
519 * This is not really an nfs issue, but it is much easier to
520 * set hostname here and then let the "/etc/rc.xxx" files
521 * mount the right /var based upon its preset value.
523 bcopy(nd->my_hostnam, hostname, MAXHOSTNAMELEN);
524 hostname[MAXHOSTNAMELEN - 1] = '\0';
525 for (i = 0; i < MAXHOSTNAMELEN; i++)
526 if (hostname[i] == '\0')
528 inittodr(ntohl(nd->root_time));
533 * Internal version of mount system call for diskless setup.
536 nfs_mountdiskless(char *path, char *which, int mountflag,
537 struct sockaddr_in *sin, struct nfs_args *args, struct thread *td,
538 struct vnode **vpp, struct mount **mpp)
541 struct sockaddr *nam;
546 if (!mp && (error = vfs_rootmountalloc("nfs", path, &mp))) {
547 printf("nfs_mountroot: NFS not configured");
551 mp->mnt_kern_flag = 0;
552 mp->mnt_flag = mountflag;
553 nam = dup_sockaddr((struct sockaddr *)sin, 1);
554 if ((error = mountnfs(args, mp, nam, which, path, vpp,
555 td->td_ucred)) != 0) {
556 printf("nfs_mountroot: mount %s on %s: %d", path, which, error);
557 mp->mnt_vfc->vfc_refcount--;
563 (void) copystr(which, mp->mnt_stat.f_mntonname, MNAMELEN - 1, 0);
569 nfs_decode_args(struct nfsmount *nmp, struct nfs_args *argp)
577 * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes
578 * no sense in that context.
580 if (argp->sotype == SOCK_STREAM)
581 nmp->nm_flag &= ~NFSMNT_NOCONN;
583 /* Also clear RDIRPLUS if not NFSv3, it crashes some servers */
584 if ((argp->flags & NFSMNT_NFSV3) == 0)
585 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
587 /* Re-bind if rsrvd port requested and wasn't on one */
588 adjsock = !(nmp->nm_flag & NFSMNT_RESVPORT)
589 && (argp->flags & NFSMNT_RESVPORT);
590 /* Also re-bind if we're switching to/from a connected UDP socket */
591 adjsock |= ((nmp->nm_flag & NFSMNT_NOCONN) !=
592 (argp->flags & NFSMNT_NOCONN));
594 /* Update flags atomically. Don't change the lock bits. */
595 nmp->nm_flag = argp->flags | nmp->nm_flag;
598 if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) {
599 nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10;
600 if (nmp->nm_timeo < NFS_MINTIMEO)
601 nmp->nm_timeo = NFS_MINTIMEO;
602 else if (nmp->nm_timeo > NFS_MAXTIMEO)
603 nmp->nm_timeo = NFS_MAXTIMEO;
606 if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) {
607 nmp->nm_retry = argp->retrans;
608 if (nmp->nm_retry > NFS_MAXREXMIT)
609 nmp->nm_retry = NFS_MAXREXMIT;
612 if (argp->flags & NFSMNT_NFSV3) {
613 if (argp->sotype == SOCK_DGRAM)
614 maxio = NFS_MAXDGRAMDATA;
618 maxio = NFS_V2MAXDATA;
620 if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
621 nmp->nm_wsize = argp->wsize;
622 /* Round down to multiple of blocksize */
623 nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1);
624 if (nmp->nm_wsize <= 0)
625 nmp->nm_wsize = NFS_FABLKSIZE;
627 if (nmp->nm_wsize > maxio)
628 nmp->nm_wsize = maxio;
629 if (nmp->nm_wsize > MAXBSIZE)
630 nmp->nm_wsize = MAXBSIZE;
632 if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
633 nmp->nm_rsize = argp->rsize;
634 /* Round down to multiple of blocksize */
635 nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1);
636 if (nmp->nm_rsize <= 0)
637 nmp->nm_rsize = NFS_FABLKSIZE;
639 if (nmp->nm_rsize > maxio)
640 nmp->nm_rsize = maxio;
641 if (nmp->nm_rsize > MAXBSIZE)
642 nmp->nm_rsize = MAXBSIZE;
644 if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) {
645 nmp->nm_readdirsize = argp->readdirsize;
647 if (nmp->nm_readdirsize > maxio)
648 nmp->nm_readdirsize = maxio;
649 if (nmp->nm_readdirsize > nmp->nm_rsize)
650 nmp->nm_readdirsize = nmp->nm_rsize;
652 if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0)
653 nmp->nm_acregmin = argp->acregmin;
655 nmp->nm_acregmin = NFS_MINATTRTIMO;
656 if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0)
657 nmp->nm_acregmax = argp->acregmax;
659 nmp->nm_acregmax = NFS_MAXATTRTIMO;
660 if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0)
661 nmp->nm_acdirmin = argp->acdirmin;
663 nmp->nm_acdirmin = NFS_MINDIRATTRTIMO;
664 if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0)
665 nmp->nm_acdirmax = argp->acdirmax;
667 nmp->nm_acdirmax = NFS_MAXDIRATTRTIMO;
668 if (nmp->nm_acdirmin > nmp->nm_acdirmax)
669 nmp->nm_acdirmin = nmp->nm_acdirmax;
670 if (nmp->nm_acregmin > nmp->nm_acregmax)
671 nmp->nm_acregmin = nmp->nm_acregmax;
673 if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0) {
674 if (argp->maxgrouplist <= NFS_MAXGRPS)
675 nmp->nm_numgrps = argp->maxgrouplist;
677 nmp->nm_numgrps = NFS_MAXGRPS;
679 if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0) {
680 if (argp->readahead <= NFS_MAXRAHEAD)
681 nmp->nm_readahead = argp->readahead;
683 nmp->nm_readahead = NFS_MAXRAHEAD;
685 if ((argp->flags & NFSMNT_DEADTHRESH) && argp->deadthresh >= 0) {
686 if (argp->deadthresh <= NFS_MAXDEADTHRESH)
687 nmp->nm_deadthresh = argp->deadthresh;
689 nmp->nm_deadthresh = NFS_MAXDEADTHRESH;
692 adjsock |= ((nmp->nm_sotype != argp->sotype) ||
693 (nmp->nm_soproto != argp->proto));
694 nmp->nm_sotype = argp->sotype;
695 nmp->nm_soproto = argp->proto;
697 if (nmp->nm_so && adjsock) {
698 nfs_safedisconnect(nmp);
699 if (nmp->nm_sotype == SOCK_DGRAM)
700 while (nfs_connect(nmp, (struct nfsreq *)0)) {
701 printf("nfs_args: retrying connect\n");
702 (void) tsleep((caddr_t)&lbolt,
712 * It seems a bit dumb to copyinstr() the host and path here and then
713 * bcopy() them in mountnfs(), but I wanted to detect errors before
714 * doing the sockargs() call because sockargs() allocates an mbuf and
715 * an error after that means that I have to release the mbuf.
719 nfs_mount(struct mount *mp, char *path, caddr_t data, struct nameidata *ndp,
723 struct nfs_args args;
724 struct sockaddr *nam;
728 u_char nfh[NFSX_V3FHMAX];
731 nfs_mountroot(mp, td);
734 error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args));
737 if (args.version != NFS_ARGSVERSION) {
738 #ifdef COMPAT_PRELITE2
740 * If the argument version is unknown, then assume the
741 * caller is a pre-lite2 4.4BSD client and convert its
744 struct onfs_args oargs;
745 error = copyin(data, (caddr_t)&oargs, sizeof (struct onfs_args));
748 nfs_convert_oargs(&args,&oargs);
749 #else /* !COMPAT_PRELITE2 */
750 return (EPROGMISMATCH);
751 #endif /* COMPAT_PRELITE2 */
753 if (mp->mnt_flag & MNT_UPDATE) {
754 struct nfsmount *nmp = VFSTONFS(mp);
759 * When doing an update, we can't change from or to
760 * v3, switch lockd strategies or change cookie translation
762 args.flags = (args.flags &
763 ~(NFSMNT_NFSV3 | NFSMNT_NOLOCKD /*|NFSMNT_XLATECOOKIE*/)) |
765 (NFSMNT_NFSV3 | NFSMNT_NOLOCKD /*|NFSMNT_XLATECOOKIE*/));
766 nfs_decode_args(nmp, &args);
769 if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX)
771 error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize);
774 error = copyinstr(args.hostname, hst, MNAMELEN-1, &len);
777 bzero(&hst[len], MNAMELEN - len);
778 /* sockargs() call must be after above copyin() calls */
779 error = getsockaddr(&nam, (caddr_t)args.addr, args.addrlen);
783 error = mountnfs(&args, mp, nam, path, hst, &vp, td->td_ucred);
788 * Common code for mount and mountroot
791 mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam,
792 char *pth, char *hst, struct vnode **vpp, struct ucred *cred)
794 struct nfsmount *nmp;
799 if (mp->mnt_flag & MNT_UPDATE) {
801 /* update paths, file handles, etc, here XXX */
805 nmp = uma_zalloc(nfsmount_zone, M_WAITOK);
806 bzero((caddr_t)nmp, sizeof (struct nfsmount));
807 TAILQ_INIT(&nmp->nm_bufq);
808 mp->mnt_data = (qaddr_t)nmp;
812 nmp->nm_cred = crhold(cred);
815 * V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too
816 * high, depending on whether we end up with negative offsets in
817 * the client or server somewhere. 2GB-1 may be safer.
819 * For V3, nfs_fsinfo will adjust this as necessary. Assume maximum
820 * that we can handle until we find out otherwise.
821 * XXX Our "safe" limit on the client is what we can store in our
822 * buffer cache using signed(!) block numbers.
824 if ((argp->flags & NFSMNT_NFSV3) == 0)
825 nmp->nm_maxfilesize = 0xffffffffLL;
827 nmp->nm_maxfilesize = (u_int64_t)0x80000000 * DEV_BSIZE - 1;
829 nmp->nm_timeo = NFS_TIMEO;
830 nmp->nm_retry = NFS_RETRANS;
831 nmp->nm_wsize = NFS_WSIZE;
832 nmp->nm_rsize = NFS_RSIZE;
833 nmp->nm_readdirsize = NFS_READDIRSIZE;
834 nmp->nm_numgrps = NFS_MAXGRPS;
835 nmp->nm_readahead = NFS_DEFRAHEAD;
836 nmp->nm_deadthresh = NFS_MAXDEADTHRESH;
837 nmp->nm_fhsize = argp->fhsize;
838 bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize);
839 bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN);
840 bcopy(pth, mp->mnt_stat.f_mntonname, MNAMELEN);
842 /* Set up the sockets and per-host congestion */
843 nmp->nm_sotype = argp->sotype;
844 nmp->nm_soproto = argp->proto;
846 nfs_decode_args(nmp, argp);
849 * For Connection based sockets (TCP,...) defer the connect until
850 * the first request, in case the server is not responding.
852 if (nmp->nm_sotype == SOCK_DGRAM &&
853 (error = nfs_connect(nmp, (struct nfsreq *)0)))
857 * This is silly, but it has to be set so that vinifod() works.
858 * We do not want to do an nfs_statfs() here since we can get
859 * stuck on a dead server and we are holding a lock on the mount
862 mp->mnt_stat.f_iosize = nfs_iosize(nmp);
864 * A reference count is needed on the nfsnode representing the
865 * remote root. If this object is not persistent, then backward
866 * traversals of the mount point (i.e. "..") will not work if
867 * the nfsnode gets flushed out of the cache. Ufs does not have
868 * this problem, because one can identify root inodes by their
869 * number == ROOTINO (2).
871 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
877 * Get file attributes for the mountpoint. This has the side
878 * effect of filling in (*vpp)->v_type with the correct value.
880 VOP_GETATTR(*vpp, &attrs, curthread->td_ucred, curthread);
883 * Lose the lock but keep the ref.
885 VOP_UNLOCK(*vpp, 0, curthread);
890 crfree(nmp->nm_cred);
891 uma_zfree(nfsmount_zone, nmp);
897 * unmount system call
900 nfs_unmount(struct mount *mp, int mntflags, struct thread *td)
902 struct nfsmount *nmp;
903 int error, flags = 0;
905 if (mntflags & MNT_FORCE)
909 * Goes something like this..
910 * - Call vflush() to clear out vnodes for this filesystem
912 * - Free up the data structures
914 /* In the forced case, cancel any outstanding requests. */
915 if (flags & FORCECLOSE) {
916 error = nfs_nmcancelreqs(nmp);
920 /* We hold 1 extra ref on the root vnode; see comment in mountnfs(). */
921 error = vflush(mp, 1, flags);
926 * We are now committed to the unmount.
929 FREE(nmp->nm_nam, M_SONAME);
931 crfree(nmp->nm_cred);
932 uma_zfree(nfsmount_zone, nmp);
937 * Return root of a filesystem
940 nfs_root(struct mount *mp, struct vnode **vpp)
943 struct nfsmount *nmp;
948 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
952 if (vp->v_type == VNON)
960 * Flush out the buffer cache
964 nfs_sync(struct mount *mp, int waitfor, struct ucred *cred, struct thread *td)
966 struct vnode *vp, *vnp;
967 int error, allerror = 0;
970 * Force stale buffer cache information to be flushed.
972 mtx_lock(&mntvnode_mtx);
974 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
978 * If the vnode that we are about to sync is no longer
979 * associated with this mount point, start over.
981 if (vp->v_mount != mp)
983 vnp = TAILQ_NEXT(vp, v_nmntvnodes);
984 mtx_unlock(&mntvnode_mtx);
985 mtx_lock(&vp->v_interlock);
986 if (VOP_ISLOCKED(vp, NULL) || TAILQ_EMPTY(&vp->v_dirtyblkhd) ||
987 waitfor == MNT_LAZY) {
988 mtx_unlock(&vp->v_interlock);
989 mtx_lock(&mntvnode_mtx);
992 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
993 mtx_lock(&mntvnode_mtx);
996 error = VOP_FSYNC(vp, cred, waitfor, td);
1000 mtx_lock(&mntvnode_mtx);
1002 mtx_unlock(&mntvnode_mtx);