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"
43 #include "opt_nfsroot.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/kernel.h>
49 #include <sys/malloc.h>
51 #include <sys/module.h>
52 #include <sys/mount.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/sockio.h>
57 #include <sys/sysctl.h>
58 #include <sys/vnode.h>
61 #include <vm/vm_extern.h>
65 #include <net/route.h>
66 #include <netinet/in.h>
68 #include <nfs/rpcv2.h>
69 #include <nfs/nfsproto.h>
70 #include <nfsclient/nfs.h>
71 #include <nfsclient/nfsnode.h>
72 #include <nfsclient/nfsmount.h>
73 #include <nfs/xdr_subs.h>
74 #include <nfsclient/nfsm_subs.h>
75 #include <nfsclient/nfsdiskless.h>
77 MALLOC_DEFINE(M_NFSREQ, "NFS req", "NFS request header");
78 MALLOC_DEFINE(M_NFSBIGFH, "NFSV3 bigfh", "NFS version 3 file handle");
79 MALLOC_DEFINE(M_NFSDIROFF, "NFSV3 diroff", "NFS directory offset data");
80 MALLOC_DEFINE(M_NFSHASH, "NFS hash", "NFS hash tables");
82 uma_zone_t nfsmount_zone;
84 struct nfsstats nfsstats;
85 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem");
86 SYSCTL_STRUCT(_vfs_nfs, NFS_NFSSTATS, nfsstats, CTLFLAG_RD,
87 &nfsstats, nfsstats, "S,nfsstats");
90 SYSCTL_INT(_vfs_nfs, OID_AUTO, debug, CTLFLAG_RW, &nfs_debug, 0, "");
93 static int nfs_iosize(struct nfsmount *nmp);
94 static void nfs_decode_args(struct nfsmount *nmp, struct nfs_args *argp);
95 static int mountnfs(struct nfs_args *, struct mount *,
96 struct sockaddr *, char *, char *, struct vnode **,
98 static vfs_mount_t nfs_mount;
99 static vfs_unmount_t nfs_unmount;
100 static vfs_root_t nfs_root;
101 static vfs_statfs_t nfs_statfs;
102 static vfs_sync_t nfs_sync;
105 * nfs vfs operations.
107 static struct vfsops nfs_vfsops = {
116 vfs_stdfhtovp, /* shouldn't happen */
118 vfs_stdvptofh, /* shouldn't happen */
123 VFS_SET(nfs_vfsops, nfs, VFCF_NETWORK);
125 /* So that loader and kldload(2) can find us, wherever we are.. */
126 MODULE_VERSION(nfs, 1);
129 * This structure must be filled in by a primary bootstrap or bootstrap
130 * server for a diskless/dataless machine. It is initialized below just
131 * to ensure that it is allocated to initialized data (.data not .bss).
133 struct nfs_diskless nfs_diskless = { { { 0 } } };
134 struct nfsv3_diskless nfsv3_diskless = { { { 0 } } };
135 int nfs_diskless_valid = 0;
137 SYSCTL_INT(_vfs_nfs, OID_AUTO, diskless_valid, CTLFLAG_RD,
138 &nfs_diskless_valid, 0, "");
140 SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_rootpath, CTLFLAG_RD,
141 nfsv3_diskless.root_hostnam, 0, "");
143 SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_rootaddr, CTLFLAG_RD,
144 &nfsv3_diskless.root_saddr, sizeof nfsv3_diskless.root_saddr,
145 "%Ssockaddr_in", "");
147 SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_swappath, CTLFLAG_RD,
148 nfsv3_diskless.swap_hostnam, 0, "");
150 SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_swapaddr, CTLFLAG_RD,
151 &nfsv3_diskless.swap_saddr, sizeof nfsv3_diskless.swap_saddr,
155 void nfsargs_ntoh(struct nfs_args *);
156 static int nfs_mountdiskless(char *, char *, int,
157 struct sockaddr_in *, struct nfs_args *,
158 struct thread *, struct vnode **, struct mount **);
159 static void nfs_convert_diskless(void);
160 static void nfs_convert_oargs(struct nfs_args *args,
161 struct onfs_args *oargs);
164 nfs_iosize(struct nfsmount *nmp)
169 * Calculate the size used for io buffers. Use the larger
170 * of the two sizes to minimise nfs requests but make sure
171 * that it is at least one VM page to avoid wasting buffer
174 iosize = max(nmp->nm_rsize, nmp->nm_wsize);
175 if (iosize < PAGE_SIZE) iosize = PAGE_SIZE;
180 nfs_convert_oargs(struct nfs_args *args, struct onfs_args *oargs)
183 args->version = NFS_ARGSVERSION;
184 args->addr = oargs->addr;
185 args->addrlen = oargs->addrlen;
186 args->sotype = oargs->sotype;
187 args->proto = oargs->proto;
188 args->fh = oargs->fh;
189 args->fhsize = oargs->fhsize;
190 args->flags = oargs->flags;
191 args->wsize = oargs->wsize;
192 args->rsize = oargs->rsize;
193 args->readdirsize = oargs->readdirsize;
194 args->timeo = oargs->timeo;
195 args->retrans = oargs->retrans;
196 args->maxgrouplist = oargs->maxgrouplist;
197 args->readahead = oargs->readahead;
198 args->deadthresh = oargs->deadthresh;
199 args->hostname = oargs->hostname;
203 nfs_convert_diskless(void)
206 bcopy(&nfs_diskless.myif, &nfsv3_diskless.myif,
207 sizeof(struct ifaliasreq));
208 bcopy(&nfs_diskless.mygateway, &nfsv3_diskless.mygateway,
209 sizeof(struct sockaddr_in));
210 nfs_convert_oargs(&nfsv3_diskless.swap_args,&nfs_diskless.swap_args);
211 nfsv3_diskless.swap_fhsize = NFSX_V2FH;
212 bcopy(nfs_diskless.swap_fh, nfsv3_diskless.swap_fh, NFSX_V2FH);
213 bcopy(&nfs_diskless.swap_saddr,&nfsv3_diskless.swap_saddr,
214 sizeof(struct sockaddr_in));
215 bcopy(nfs_diskless.swap_hostnam, nfsv3_diskless.swap_hostnam, MNAMELEN);
216 nfsv3_diskless.swap_nblks = nfs_diskless.swap_nblks;
217 bcopy(&nfs_diskless.swap_ucred, &nfsv3_diskless.swap_ucred,
218 sizeof(struct ucred));
219 nfs_convert_oargs(&nfsv3_diskless.root_args,&nfs_diskless.root_args);
220 nfsv3_diskless.root_fhsize = NFSX_V2FH;
221 bcopy(nfs_diskless.root_fh, nfsv3_diskless.root_fh, NFSX_V2FH);
222 bcopy(&nfs_diskless.root_saddr,&nfsv3_diskless.root_saddr,
223 sizeof(struct sockaddr_in));
224 bcopy(nfs_diskless.root_hostnam, nfsv3_diskless.root_hostnam, MNAMELEN);
225 nfsv3_diskless.root_time = nfs_diskless.root_time;
226 bcopy(nfs_diskless.my_hostnam, nfsv3_diskless.my_hostnam,
228 nfs_diskless_valid = 3;
235 nfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
238 struct nfs_statfs *sfp;
240 struct nfsmount *nmp = VFSTONFS(mp);
241 int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr;
242 struct mbuf *mreq, *mrep, *md, *mb;
249 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
253 if (v3 && (nmp->nm_state & NFSSTA_GOTFSINFO) == 0)
254 (void)nfs_fsinfo(nmp, vp, td->td_ucred, td);
255 nfsstats.rpccnt[NFSPROC_FSSTAT]++;
256 mreq = nfsm_reqhead(vp, NFSPROC_FSSTAT, NFSX_FH(v3));
258 bpos = mtod(mb, caddr_t);
260 nfsm_request(vp, NFSPROC_FSSTAT, td, td->td_ucred);
262 nfsm_postop_attr(vp, retattr);
268 sfp = nfsm_dissect(struct nfs_statfs *, NFSX_STATFS(v3));
269 sbp->f_flags = nmp->nm_flag;
270 sbp->f_iosize = nfs_iosize(nmp);
272 sbp->f_bsize = NFS_FABLKSIZE;
273 tquad = fxdr_hyper(&sfp->sf_tbytes);
274 sbp->f_blocks = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
275 tquad = fxdr_hyper(&sfp->sf_fbytes);
276 sbp->f_bfree = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
277 tquad = fxdr_hyper(&sfp->sf_abytes);
278 sbp->f_bavail = (long)(tquad / ((u_quad_t)NFS_FABLKSIZE));
279 sbp->f_files = (fxdr_unsigned(int32_t,
280 sfp->sf_tfiles.nfsuquad[1]) & 0x7fffffff);
281 sbp->f_ffree = (fxdr_unsigned(int32_t,
282 sfp->sf_ffiles.nfsuquad[1]) & 0x7fffffff);
284 sbp->f_bsize = fxdr_unsigned(int32_t, sfp->sf_bsize);
285 sbp->f_blocks = fxdr_unsigned(int32_t, sfp->sf_blocks);
286 sbp->f_bfree = fxdr_unsigned(int32_t, sfp->sf_bfree);
287 sbp->f_bavail = fxdr_unsigned(int32_t, sfp->sf_bavail);
291 if (sbp != &mp->mnt_stat) {
292 sbp->f_type = mp->mnt_vfc->vfc_typenum;
293 bcopy(mp->mnt_stat.f_mntonname, sbp->f_mntonname, MNAMELEN);
294 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
303 * nfs version 3 fsinfo rpc call
306 nfs_fsinfo(struct nfsmount *nmp, struct vnode *vp, struct ucred *cred,
309 struct nfsv3_fsinfo *fsp;
312 int error = 0, retattr;
313 struct mbuf *mreq, *mrep, *md, *mb;
316 nfsstats.rpccnt[NFSPROC_FSINFO]++;
317 mreq = nfsm_reqhead(vp, NFSPROC_FSINFO, NFSX_FH(1));
319 bpos = mtod(mb, caddr_t);
321 nfsm_request(vp, NFSPROC_FSINFO, td, cred);
322 nfsm_postop_attr(vp, retattr);
324 fsp = nfsm_dissect(struct nfsv3_fsinfo *, NFSX_V3FSINFO);
325 pref = fxdr_unsigned(u_int32_t, fsp->fs_wtpref);
326 if (pref < nmp->nm_wsize && pref >= NFS_FABLKSIZE)
327 nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) &
328 ~(NFS_FABLKSIZE - 1);
329 max = fxdr_unsigned(u_int32_t, fsp->fs_wtmax);
330 if (max < nmp->nm_wsize && max > 0) {
331 nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1);
332 if (nmp->nm_wsize == 0)
335 pref = fxdr_unsigned(u_int32_t, fsp->fs_rtpref);
336 if (pref < nmp->nm_rsize && pref >= NFS_FABLKSIZE)
337 nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) &
338 ~(NFS_FABLKSIZE - 1);
339 max = fxdr_unsigned(u_int32_t, fsp->fs_rtmax);
340 if (max < nmp->nm_rsize && max > 0) {
341 nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1);
342 if (nmp->nm_rsize == 0)
345 pref = fxdr_unsigned(u_int32_t, fsp->fs_dtpref);
346 if (pref < nmp->nm_readdirsize && pref >= NFS_DIRBLKSIZ)
347 nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) &
348 ~(NFS_DIRBLKSIZ - 1);
349 if (max < nmp->nm_readdirsize && max > 0) {
350 nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1);
351 if (nmp->nm_readdirsize == 0)
352 nmp->nm_readdirsize = max;
354 maxfsize = fxdr_hyper(&fsp->fs_maxfilesize);
355 if (maxfsize > 0 && maxfsize < nmp->nm_maxfilesize)
356 nmp->nm_maxfilesize = maxfsize;
357 nmp->nm_state |= NFSSTA_GOTFSINFO;
365 * Mount a remote root fs via. nfs. This depends on the info in the
366 * nfs_diskless structure that has been filled in properly by some primary
368 * It goes something like this:
369 * - do enough of "ifconfig" by calling ifioctl() so that the system
370 * can talk to the server
371 * - If nfs_diskless.mygateway is filled in, use that address as
373 * - build the rootfs mount point and call mountnfs() to do the rest.
376 nfs_mountroot(struct mount *mp, struct thread *td)
378 struct mount *swap_mp;
379 struct nfsv3_diskless *nd = &nfsv3_diskless;
386 #if defined(BOOTP_NFSROOT) && defined(BOOTP)
387 bootpc_init(); /* use bootp to get nfs_diskless filled in */
388 #elif defined(NFS_ROOT)
389 nfs_setup_diskless();
392 if (nfs_diskless_valid == 0)
394 if (nfs_diskless_valid == 1)
395 nfs_convert_diskless();
398 * XXX splnet, so networks will receive...
403 /* Set up swap credentials. */
404 proc0.p_ucred->cr_uid = ntohl(nd->swap_ucred.cr_uid);
405 proc0.p_ucred->cr_gid = ntohl(nd->swap_ucred.cr_gid);
406 if ((proc0.p_ucred->cr_ngroups = ntohs(nd->swap_ucred.cr_ngroups)) >
408 proc0.p_ucred->cr_ngroups = NGROUPS;
409 for (i = 0; i < proc0.p_ucred->cr_ngroups; i++)
410 proc0.p_ucred->cr_groups[i] = ntohl(nd->swap_ucred.cr_groups[i]);
414 * Do enough of ifconfig(8) so that the critical net interface can
415 * talk to the server.
417 error = socreate(nd->myif.ifra_addr.sa_family, &so, SOCK_DGRAM, 0,
420 panic("nfs_mountroot: socreate(%04x): %d",
421 nd->myif.ifra_addr.sa_family, error);
423 #if 0 /* XXX Bad idea */
425 * We might not have been told the right interface, so we pass
426 * over the first ten interfaces of the same kind, until we get
427 * one of them configured.
430 for (i = strlen(nd->myif.ifra_name) - 1;
431 nd->myif.ifra_name[i] >= '0' &&
432 nd->myif.ifra_name[i] <= '9';
433 nd->myif.ifra_name[i] ++) {
434 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td);
439 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td);
441 panic("nfs_mountroot: SIOCAIFADDR: %d", error);
445 * If the gateway field is filled in, set it as the default route.
446 * Note that pxeboot will set a default route of 0 if the route
447 * is not set by the DHCP server. Check also for a value of 0
448 * to avoid panicking inappropriately in that situation.
450 if (nd->mygateway.sin_len != 0 &&
451 nd->mygateway.sin_addr.s_addr != 0) {
452 struct sockaddr_in mask, sin;
454 bzero((caddr_t)&mask, sizeof(mask));
456 sin.sin_family = AF_INET;
457 sin.sin_len = sizeof(sin);
458 error = rtrequest(RTM_ADD, (struct sockaddr *)&sin,
459 (struct sockaddr *)&nd->mygateway,
460 (struct sockaddr *)&mask,
461 RTF_UP | RTF_GATEWAY, NULL);
463 panic("nfs_mountroot: RTM_ADD: %d", error);
467 * Create the rootfs mount point.
469 nd->root_args.fh = nd->root_fh;
470 nd->root_args.fhsize = nd->root_fhsize;
471 l = ntohl(nd->root_saddr.sin_addr.s_addr);
472 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
473 (l >> 24) & 0xff, (l >> 16) & 0xff,
474 (l >> 8) & 0xff, (l >> 0) & 0xff, nd->root_hostnam);
475 printf("NFS ROOT: %s\n", buf);
476 if ((error = nfs_mountdiskless(buf, "/", MNT_RDONLY,
477 &nd->root_saddr, &nd->root_args, td, &vp, &mp)) != 0) {
479 mp->mnt_vfc->vfc_refcount--;
480 free(swap_mp, M_MOUNT);
486 if (nd->swap_nblks) {
488 /* Convert to DEV_BSIZE instead of Kilobyte */
492 * Create a fake mount point just for the swap vnode so that the
493 * swap file can be on a different server from the rootfs.
495 nd->swap_args.fh = nd->swap_fh;
496 nd->swap_args.fhsize = nd->swap_fhsize;
497 l = ntohl(nd->swap_saddr.sin_addr.s_addr);
498 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
499 (l >> 24) & 0xff, (l >> 16) & 0xff,
500 (l >> 8) & 0xff, (l >> 0) & 0xff, nd->swap_hostnam);
501 printf("NFS SWAP: %s\n", buf);
502 if ((error = nfs_mountdiskless(buf, "/swap", 0,
503 &nd->swap_saddr, &nd->swap_args, td, &vp, &swap_mp)) != 0)
505 vfs_unbusy(swap_mp, td);
507 VTONFS(vp)->n_size = VTONFS(vp)->n_vattr.va_size =
508 nd->swap_nblks * DEV_BSIZE ;
510 * Since the swap file is not the root dir of a filesystem,
511 * hack it to a regular file.
517 swaponvp(td, vp, NODEV, nd->swap_nblks);
520 mp->mnt_flag |= MNT_ROOTFS;
521 mp->mnt_vnodecovered = NULLVP;
526 * This is not really an nfs issue, but it is much easier to
527 * set hostname here and then let the "/etc/rc.xxx" files
528 * mount the right /var based upon its preset value.
530 bcopy(nd->my_hostnam, hostname, MAXHOSTNAMELEN);
531 hostname[MAXHOSTNAMELEN - 1] = '\0';
532 for (i = 0; i < MAXHOSTNAMELEN; i++)
533 if (hostname[i] == '\0')
535 inittodr(ntohl(nd->root_time));
540 * Internal version of mount system call for diskless setup.
543 nfs_mountdiskless(char *path, char *which, int mountflag,
544 struct sockaddr_in *sin, struct nfs_args *args, struct thread *td,
545 struct vnode **vpp, struct mount **mpp)
548 struct sockaddr *nam;
553 if (!mp && (error = vfs_rootmountalloc("nfs", path, &mp))) {
554 printf("nfs_mountroot: NFS not configured");
558 mp->mnt_kern_flag = 0;
559 mp->mnt_flag = mountflag;
560 nam = dup_sockaddr((struct sockaddr *)sin, 1);
561 if ((error = mountnfs(args, mp, nam, which, path, vpp,
562 td->td_ucred)) != 0) {
563 printf("nfs_mountroot: mount %s on %s: %d", path, which, error);
564 mp->mnt_vfc->vfc_refcount--;
570 (void) copystr(which, mp->mnt_stat.f_mntonname, MNAMELEN - 1, 0);
576 nfs_decode_args(struct nfsmount *nmp, struct nfs_args *argp)
584 * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes
585 * no sense in that context.
587 if (argp->sotype == SOCK_STREAM)
588 nmp->nm_flag &= ~NFSMNT_NOCONN;
590 /* Also clear RDIRPLUS if not NFSv3, it crashes some servers */
591 if ((argp->flags & NFSMNT_NFSV3) == 0)
592 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
594 /* Re-bind if rsrvd port requested and wasn't on one */
595 adjsock = !(nmp->nm_flag & NFSMNT_RESVPORT)
596 && (argp->flags & NFSMNT_RESVPORT);
597 /* Also re-bind if we're switching to/from a connected UDP socket */
598 adjsock |= ((nmp->nm_flag & NFSMNT_NOCONN) !=
599 (argp->flags & NFSMNT_NOCONN));
601 /* Update flags atomically. Don't change the lock bits. */
602 nmp->nm_flag = argp->flags | nmp->nm_flag;
605 if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) {
606 nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10;
607 if (nmp->nm_timeo < NFS_MINTIMEO)
608 nmp->nm_timeo = NFS_MINTIMEO;
609 else if (nmp->nm_timeo > NFS_MAXTIMEO)
610 nmp->nm_timeo = NFS_MAXTIMEO;
613 if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1) {
614 nmp->nm_retry = argp->retrans;
615 if (nmp->nm_retry > NFS_MAXREXMIT)
616 nmp->nm_retry = NFS_MAXREXMIT;
619 if (argp->flags & NFSMNT_NFSV3) {
620 if (argp->sotype == SOCK_DGRAM)
621 maxio = NFS_MAXDGRAMDATA;
625 maxio = NFS_V2MAXDATA;
627 if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
628 nmp->nm_wsize = argp->wsize;
629 /* Round down to multiple of blocksize */
630 nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1);
631 if (nmp->nm_wsize <= 0)
632 nmp->nm_wsize = NFS_FABLKSIZE;
634 if (nmp->nm_wsize > maxio)
635 nmp->nm_wsize = maxio;
636 if (nmp->nm_wsize > MAXBSIZE)
637 nmp->nm_wsize = MAXBSIZE;
639 if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
640 nmp->nm_rsize = argp->rsize;
641 /* Round down to multiple of blocksize */
642 nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1);
643 if (nmp->nm_rsize <= 0)
644 nmp->nm_rsize = NFS_FABLKSIZE;
646 if (nmp->nm_rsize > maxio)
647 nmp->nm_rsize = maxio;
648 if (nmp->nm_rsize > MAXBSIZE)
649 nmp->nm_rsize = MAXBSIZE;
651 if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) {
652 nmp->nm_readdirsize = argp->readdirsize;
654 if (nmp->nm_readdirsize > maxio)
655 nmp->nm_readdirsize = maxio;
656 if (nmp->nm_readdirsize > nmp->nm_rsize)
657 nmp->nm_readdirsize = nmp->nm_rsize;
659 if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0)
660 nmp->nm_acregmin = argp->acregmin;
662 nmp->nm_acregmin = NFS_MINATTRTIMO;
663 if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0)
664 nmp->nm_acregmax = argp->acregmax;
666 nmp->nm_acregmax = NFS_MAXATTRTIMO;
667 if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0)
668 nmp->nm_acdirmin = argp->acdirmin;
670 nmp->nm_acdirmin = NFS_MINDIRATTRTIMO;
671 if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0)
672 nmp->nm_acdirmax = argp->acdirmax;
674 nmp->nm_acdirmax = NFS_MAXDIRATTRTIMO;
675 if (nmp->nm_acdirmin > nmp->nm_acdirmax)
676 nmp->nm_acdirmin = nmp->nm_acdirmax;
677 if (nmp->nm_acregmin > nmp->nm_acregmax)
678 nmp->nm_acregmin = nmp->nm_acregmax;
680 if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0) {
681 if (argp->maxgrouplist <= NFS_MAXGRPS)
682 nmp->nm_numgrps = argp->maxgrouplist;
684 nmp->nm_numgrps = NFS_MAXGRPS;
686 if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0) {
687 if (argp->readahead <= NFS_MAXRAHEAD)
688 nmp->nm_readahead = argp->readahead;
690 nmp->nm_readahead = NFS_MAXRAHEAD;
692 if ((argp->flags & NFSMNT_DEADTHRESH) && argp->deadthresh >= 0) {
693 if (argp->deadthresh <= NFS_MAXDEADTHRESH)
694 nmp->nm_deadthresh = argp->deadthresh;
696 nmp->nm_deadthresh = NFS_MAXDEADTHRESH;
699 adjsock |= ((nmp->nm_sotype != argp->sotype) ||
700 (nmp->nm_soproto != argp->proto));
701 nmp->nm_sotype = argp->sotype;
702 nmp->nm_soproto = argp->proto;
704 if (nmp->nm_so && adjsock) {
705 nfs_safedisconnect(nmp);
706 if (nmp->nm_sotype == SOCK_DGRAM)
707 while (nfs_connect(nmp, NULL)) {
708 printf("nfs_args: retrying connect\n");
709 (void) tsleep((caddr_t)&lbolt,
719 * It seems a bit dumb to copyinstr() the host and path here and then
720 * bcopy() them in mountnfs(), but I wanted to detect errors before
721 * doing the sockargs() call because sockargs() allocates an mbuf and
722 * an error after that means that I have to release the mbuf.
726 nfs_mount(struct mount *mp, char *path, caddr_t data, struct nameidata *ndp,
730 struct nfs_args args;
731 struct sockaddr *nam;
735 u_char nfh[NFSX_V3FHMAX];
738 return (nfs_mountroot(mp, td));
739 error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args));
742 if (args.version != NFS_ARGSVERSION) {
743 #ifdef COMPAT_PRELITE2
745 * If the argument version is unknown, then assume the
746 * caller is a pre-lite2 4.4BSD client and convert its
749 struct onfs_args oargs;
750 error = copyin(data, (caddr_t)&oargs, sizeof (struct onfs_args));
753 nfs_convert_oargs(&args,&oargs);
754 #else /* !COMPAT_PRELITE2 */
755 return (EPROGMISMATCH);
756 #endif /* COMPAT_PRELITE2 */
758 if (mp->mnt_flag & MNT_UPDATE) {
759 struct nfsmount *nmp = VFSTONFS(mp);
764 * When doing an update, we can't change from or to
765 * v3, switch lockd strategies or change cookie translation
767 args.flags = (args.flags &
768 ~(NFSMNT_NFSV3 | NFSMNT_NOLOCKD /*|NFSMNT_XLATECOOKIE*/)) |
770 (NFSMNT_NFSV3 | NFSMNT_NOLOCKD /*|NFSMNT_XLATECOOKIE*/));
771 nfs_decode_args(nmp, &args);
774 if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX)
776 error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize);
779 error = copyinstr(args.hostname, hst, MNAMELEN-1, &len);
782 bzero(&hst[len], MNAMELEN - len);
783 /* sockargs() call must be after above copyin() calls */
784 error = getsockaddr(&nam, (caddr_t)args.addr, args.addrlen);
788 error = mountnfs(&args, mp, nam, path, hst, &vp, td->td_ucred);
793 * Common code for mount and mountroot
796 mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam,
797 char *pth, char *hst, struct vnode **vpp, struct ucred *cred)
799 struct nfsmount *nmp;
804 if (mp->mnt_flag & MNT_UPDATE) {
806 /* update paths, file handles, etc, here XXX */
810 nmp = uma_zalloc(nfsmount_zone, M_WAITOK);
811 bzero((caddr_t)nmp, sizeof (struct nfsmount));
812 TAILQ_INIT(&nmp->nm_bufq);
813 mp->mnt_data = (qaddr_t)nmp;
819 * V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too
820 * high, depending on whether we end up with negative offsets in
821 * the client or server somewhere. 2GB-1 may be safer.
823 * For V3, nfs_fsinfo will adjust this as necessary. Assume maximum
824 * that we can handle until we find out otherwise.
825 * XXX Our "safe" limit on the client is what we can store in our
826 * buffer cache using signed(!) block numbers.
828 if ((argp->flags & NFSMNT_NFSV3) == 0)
829 nmp->nm_maxfilesize = 0xffffffffLL;
831 nmp->nm_maxfilesize = (u_int64_t)0x80000000 * DEV_BSIZE - 1;
833 nmp->nm_timeo = NFS_TIMEO;
834 nmp->nm_retry = NFS_RETRANS;
835 nmp->nm_wsize = NFS_WSIZE;
836 nmp->nm_rsize = NFS_RSIZE;
837 nmp->nm_readdirsize = NFS_READDIRSIZE;
838 nmp->nm_numgrps = NFS_MAXGRPS;
839 nmp->nm_readahead = NFS_DEFRAHEAD;
840 nmp->nm_deadthresh = NFS_MAXDEADTHRESH;
841 nmp->nm_fhsize = argp->fhsize;
842 bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize);
843 bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN);
844 bcopy(pth, mp->mnt_stat.f_mntonname, MNAMELEN);
846 /* Set up the sockets and per-host congestion */
847 nmp->nm_sotype = argp->sotype;
848 nmp->nm_soproto = argp->proto;
850 nfs_decode_args(nmp, argp);
853 * For Connection based sockets (TCP,...) defer the connect until
854 * the first request, in case the server is not responding.
856 if (nmp->nm_sotype == SOCK_DGRAM &&
857 (error = nfs_connect(nmp, NULL)))
861 * This is silly, but it has to be set so that vinifod() works.
862 * We do not want to do an nfs_statfs() here since we can get
863 * stuck on a dead server and we are holding a lock on the mount
866 mp->mnt_stat.f_iosize = nfs_iosize(nmp);
868 * A reference count is needed on the nfsnode representing the
869 * remote root. If this object is not persistent, then backward
870 * traversals of the mount point (i.e. "..") will not work if
871 * the nfsnode gets flushed out of the cache. Ufs does not have
872 * this problem, because one can identify root inodes by their
873 * number == ROOTINO (2).
875 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
881 * Get file attributes for the mountpoint. This has the side
882 * effect of filling in (*vpp)->v_type with the correct value.
884 VOP_GETATTR(*vpp, &attrs, curthread->td_ucred, curthread);
887 * Lose the lock but keep the ref.
889 VOP_UNLOCK(*vpp, 0, curthread);
894 uma_zfree(nfsmount_zone, nmp);
900 * unmount system call
903 nfs_unmount(struct mount *mp, int mntflags, struct thread *td)
905 struct nfsmount *nmp;
906 int error, flags = 0;
908 if (mntflags & MNT_FORCE)
912 * Goes something like this..
913 * - Call vflush() to clear out vnodes for this filesystem
915 * - Free up the data structures
917 /* In the forced case, cancel any outstanding requests. */
918 if (flags & FORCECLOSE) {
919 error = nfs_nmcancelreqs(nmp);
923 /* We hold 1 extra ref on the root vnode; see comment in mountnfs(). */
924 error = vflush(mp, 1, flags);
929 * We are now committed to the unmount.
932 FREE(nmp->nm_nam, M_SONAME);
934 uma_zfree(nfsmount_zone, nmp);
939 * Return root of a filesystem
942 nfs_root(struct mount *mp, struct vnode **vpp)
945 struct nfsmount *nmp;
950 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
954 if (vp->v_type == VNON)
956 vp->v_vflag |= VV_ROOT;
962 * Flush out the buffer cache
966 nfs_sync(struct mount *mp, int waitfor, struct ucred *cred, struct thread *td)
968 struct vnode *vp, *vnp;
969 int error, allerror = 0;
972 * Force stale buffer cache information to be flushed.
974 mtx_lock(&mntvnode_mtx);
976 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
980 * If the vnode that we are about to sync is no longer
981 * associated with this mount point, start over.
983 if (vp->v_mount != mp)
985 vnp = TAILQ_NEXT(vp, v_nmntvnodes);
986 mtx_unlock(&mntvnode_mtx);
988 if (VOP_ISLOCKED(vp, NULL) || TAILQ_EMPTY(&vp->v_dirtyblkhd) ||
989 waitfor == MNT_LAZY) {
991 mtx_lock(&mntvnode_mtx);
994 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
995 mtx_lock(&mntvnode_mtx);
998 error = VOP_FSYNC(vp, cred, waitfor, td);
1002 mtx_lock(&mntvnode_mtx);
1004 mtx_unlock(&mntvnode_mtx);