Deleted Added
sdiff udiff text old ( 91406 ) new ( 92783 )
full compact
1/*
2 * Copyright (c) 1989, 1993, 1995
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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.
23 *
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
34 * SUCH DAMAGE.
35 *
36 * @(#)nfs_vfsops.c 8.12 (Berkeley) 5/20/95
37 */
38
39#include <sys/cdefs.h>
40__FBSDID("$FreeBSD: head/sys/nfsclient/nfs_vfsops.c 91406 2002-02-27 18:32:23Z jhb $");
40__FBSDID("$FreeBSD: head/sys/nfsclient/nfs_vfsops.c 92783 2002-03-20 10:07:52Z jeff $");
41
42#include "opt_bootp.h"
43
44#include <sys/param.h>
45#include <sys/systm.h>
46#include <sys/kernel.h>
47#include <sys/lock.h>
48#include <sys/malloc.h>
49#include <sys/mbuf.h>
50#include <sys/module.h>
51#include <sys/mount.h>
52#include <sys/proc.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>
58
59#include <vm/vm.h>
60#include <vm/vm_extern.h>
61#include <vm/vm_zone.h>
61#include <vm/uma.h>
62
63#include <net/if.h>
64#include <net/route.h>
65#include <netinet/in.h>
66
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>
75
76MALLOC_DEFINE(M_NFSREQ, "NFS req", "NFS request header");
77MALLOC_DEFINE(M_NFSBIGFH, "NFSV3 bigfh", "NFS version 3 file handle");
78MALLOC_DEFINE(M_NFSDIROFF, "NFSV3 diroff", "NFS directory offset data");
79MALLOC_DEFINE(M_NFSHASH, "NFS hash", "NFS hash tables");
80
81vm_zone_t nfsmount_zone;
81uma_zone_t nfsmount_zone;
82
83struct nfsstats nfsstats;
84SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem");
85SYSCTL_STRUCT(_vfs_nfs, NFS_NFSSTATS, nfsstats, CTLFLAG_RD,
86 &nfsstats, nfsstats, "S,nfsstats");
87#ifdef NFS_DEBUG
88int nfs_debug;
89SYSCTL_INT(_vfs_nfs, OID_AUTO, debug, CTLFLAG_RW, &nfs_debug, 0, "");
90#endif
91
92static int nfs_iosize(struct nfsmount *nmp);
93static void nfs_decode_args(struct nfsmount *nmp, struct nfs_args *argp);
94static int mountnfs(struct nfs_args *, struct mount *,
95 struct sockaddr *, char *, char *, struct vnode **,
96 struct ucred *cred);
97static int nfs_mount(struct mount *mp, char *path, caddr_t data,
98 struct nameidata *ndp, struct thread *td);
99static int nfs_unmount(struct mount *mp, int mntflags, struct thread *td);
100static int nfs_root(struct mount *mp, struct vnode **vpp);
101static int nfs_statfs(struct mount *mp, struct statfs *sbp,
102 struct thread *td);
103static int nfs_sync(struct mount *mp, int waitfor, struct ucred *cred,
104 struct thread *td);
105
106/*
107 * nfs vfs operations.
108 */
109static struct vfsops nfs_vfsops = {
110 nfs_mount,
111 vfs_stdstart,
112 nfs_unmount,
113 nfs_root,
114 vfs_stdquotactl,
115 nfs_statfs,
116 nfs_sync,
117 vfs_stdvget,
118 vfs_stdfhtovp, /* shouldn't happen */
119 vfs_stdcheckexp,
120 vfs_stdvptofh, /* shouldn't happen */
121 nfs_init,
122 nfs_uninit,
123 vfs_stdextattrctl,
124};
125VFS_SET(nfs_vfsops, nfs, VFCF_NETWORK);
126
127/* So that loader and kldload(2) can find us, wherever we are.. */
128MODULE_VERSION(nfs, 1);
129
130/*
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).
134 */
135struct nfs_diskless nfs_diskless = { { { 0 } } };
136struct nfsv3_diskless nfsv3_diskless = { { { 0 } } };
137int nfs_diskless_valid = 0;
138
139SYSCTL_INT(_vfs_nfs, OID_AUTO, diskless_valid, CTLFLAG_RD,
140 &nfs_diskless_valid, 0, "");
141
142SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_rootpath, CTLFLAG_RD,
143 nfsv3_diskless.root_hostnam, 0, "");
144
145SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_rootaddr, CTLFLAG_RD,
146 &nfsv3_diskless.root_saddr, sizeof nfsv3_diskless.root_saddr,
147 "%Ssockaddr_in", "");
148
149SYSCTL_STRING(_vfs_nfs, OID_AUTO, diskless_swappath, CTLFLAG_RD,
150 nfsv3_diskless.swap_hostnam, 0, "");
151
152SYSCTL_OPAQUE(_vfs_nfs, OID_AUTO, diskless_swapaddr, CTLFLAG_RD,
153 &nfsv3_diskless.swap_saddr, sizeof nfsv3_diskless.swap_saddr,
154 "%Ssockaddr_in","");
155
156
157void nfsargs_ntoh(struct nfs_args *);
158static int nfs_mountdiskless(char *, char *, int,
159 struct sockaddr_in *, struct nfs_args *,
160 struct thread *, struct vnode **, struct mount **);
161static void nfs_convert_diskless(void);
162static void nfs_convert_oargs(struct nfs_args *args,
163 struct onfs_args *oargs);
164
165static int
166nfs_iosize(struct nfsmount *nmp)
167{
168 int iosize;
169
170 /*
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
174 * space.
175 */
176 iosize = max(nmp->nm_rsize, nmp->nm_wsize);
177 if (iosize < PAGE_SIZE) iosize = PAGE_SIZE;
178 return iosize;
179}
180
181static void
182nfs_convert_oargs(struct nfs_args *args, struct onfs_args *oargs)
183{
184
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;
202}
203
204static void
205nfs_convert_diskless(void)
206{
207
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,
229 MAXHOSTNAMELEN);
230 nfs_diskless_valid = 3;
231}
232
233/*
234 * nfs statfs call
235 */
236int
237nfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
238{
239 struct vnode *vp;
240 struct nfs_statfs *sfp;
241 caddr_t bpos, dpos;
242 struct nfsmount *nmp = VFSTONFS(mp);
243 int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr;
244 struct mbuf *mreq, *mrep, *md, *mb;
245 struct nfsnode *np;
246 u_quad_t tquad;
247
248#ifndef nolint
249 sfp = (struct nfs_statfs *)0;
250#endif
251 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
252 if (error)
253 return (error);
254 vp = NFSTOV(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));
259 mb = mreq;
260 bpos = mtod(mb, caddr_t);
261 nfsm_fhtom(vp, v3);
262 nfsm_request(vp, NFSPROC_FSSTAT, td, td->td_ucred);
263 if (v3)
264 nfsm_postop_attr(vp, retattr);
265 if (error) {
266 if (mrep != NULL)
267 m_freem(mrep);
268 goto nfsmout;
269 }
270 sfp = nfsm_dissect(struct nfs_statfs *, NFSX_STATFS(v3));
271 sbp->f_flags = nmp->nm_flag;
272 sbp->f_iosize = nfs_iosize(nmp);
273 if (v3) {
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);
285 } else {
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);
290 sbp->f_files = 0;
291 sbp->f_ffree = 0;
292 }
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);
297 }
298 m_freem(mrep);
299nfsmout:
300 vput(vp);
301 return (error);
302}
303
304/*
305 * nfs version 3 fsinfo rpc call
306 */
307int
308nfs_fsinfo(struct nfsmount *nmp, struct vnode *vp, struct ucred *cred,
309 struct thread *td)
310{
311 struct nfsv3_fsinfo *fsp;
312 u_int32_t pref, max;
313 caddr_t bpos, dpos;
314 int error = 0, retattr;
315 struct mbuf *mreq, *mrep, *md, *mb;
316 u_int64_t maxfsize;
317
318 nfsstats.rpccnt[NFSPROC_FSINFO]++;
319 mreq = nfsm_reqhead(vp, NFSPROC_FSINFO, NFSX_FH(1));
320 mb = mreq;
321 bpos = mtod(mb, caddr_t);
322 nfsm_fhtom(vp, 1);
323 nfsm_request(vp, NFSPROC_FSINFO, td, cred);
324 nfsm_postop_attr(vp, retattr);
325 if (!error) {
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)
335 nmp->nm_wsize = max;
336 }
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)
345 nmp->nm_rsize = max;
346 }
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;
355 }
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;
360 }
361 m_freem(mrep);
362nfsmout:
363 return (error);
364}
365
366/*
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
369 * bootstrap.
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
374 * a default gateway.
375 * - build the rootfs mount point and call mountnfs() to do the rest.
376 */
377int
378nfs_mountroot(struct mount *mp, struct thread *td)
379{
380 struct mount *swap_mp;
381 struct nfsv3_diskless *nd = &nfsv3_diskless;
382 struct socket *so;
383 struct vnode *vp;
384 int error, i;
385 u_long l;
386 char buf[128];
387
388#if defined(BOOTP_NFSROOT) && defined(BOOTP)
389 bootpc_init(); /* use bootp to get nfs_diskless filled in */
390#endif
391
392 /*
393 * XXX time must be non-zero when we init the interface or else
394 * the arp code will wedge...
395 */
396 while (time_second == 0)
397 tsleep(&time_second, PZERO+8, "arpkludge", 10);
398
399 if (nfs_diskless_valid==1)
400 nfs_convert_diskless();
401
402 /*
403 * XXX splnet, so networks will receive...
404 */
405 splnet();
406
407#ifdef notyet
408 /* Set up swap credentials. */
409 proc0.p_ucred->cr_uid = ntohl(nd->swap_ucred.cr_uid);
410 proc0.p_ucred->cr_gid = ntohl(nd->swap_ucred.cr_gid);
411 if ((proc0.p_ucred->cr_ngroups = ntohs(nd->swap_ucred.cr_ngroups)) >
412 NGROUPS)
413 proc0.p_ucred->cr_ngroups = NGROUPS;
414 for (i = 0; i < proc0.p_ucred->cr_ngroups; i++)
415 proc0.p_ucred->cr_groups[i] = ntohl(nd->swap_ucred.cr_groups[i]);
416#endif
417
418 /*
419 * Do enough of ifconfig(8) so that the critical net interface can
420 * talk to the server.
421 */
422 error = socreate(nd->myif.ifra_addr.sa_family, &so, SOCK_DGRAM, 0,
423 td->td_ucred, td);
424 if (error)
425 panic("nfs_mountroot: socreate(%04x): %d",
426 nd->myif.ifra_addr.sa_family, error);
427
428#if 0 /* XXX Bad idea */
429 /*
430 * We might not have been told the right interface, so we pass
431 * over the first ten interfaces of the same kind, until we get
432 * one of them configured.
433 */
434
435 for (i = strlen(nd->myif.ifra_name) - 1;
436 nd->myif.ifra_name[i] >= '0' &&
437 nd->myif.ifra_name[i] <= '9';
438 nd->myif.ifra_name[i] ++) {
439 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td);
440 if(!error)
441 break;
442 }
443#endif
444 error = ifioctl(so, SIOCAIFADDR, (caddr_t)&nd->myif, td);
445 if (error)
446 panic("nfs_mountroot: SIOCAIFADDR: %d", error);
447 soclose(so);
448
449 /*
450 * If the gateway field is filled in, set it as the default route.
451 */
452 if (nd->mygateway.sin_len != 0) {
453 struct sockaddr_in mask, sin;
454
455 bzero((caddr_t)&mask, sizeof(mask));
456 sin = mask;
457 sin.sin_family = AF_INET;
458 sin.sin_len = sizeof(sin);
459 error = rtrequest(RTM_ADD, (struct sockaddr *)&sin,
460 (struct sockaddr *)&nd->mygateway,
461 (struct sockaddr *)&mask,
462 RTF_UP | RTF_GATEWAY, (struct rtentry **)0);
463 if (error)
464 panic("nfs_mountroot: RTM_ADD: %d", error);
465 }
466
467 /*
468 * Create the rootfs mount point.
469 */
470 nd->root_args.fh = nd->root_fh;
471 nd->root_args.fhsize = nd->root_fhsize;
472 l = ntohl(nd->root_saddr.sin_addr.s_addr);
473 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
474 (l >> 24) & 0xff, (l >> 16) & 0xff,
475 (l >> 8) & 0xff, (l >> 0) & 0xff, nd->root_hostnam);
476 printf("NFS ROOT: %s\n", buf);
477 if ((error = nfs_mountdiskless(buf, "/", MNT_RDONLY,
478 &nd->root_saddr, &nd->root_args, td, &vp, &mp)) != 0) {
479 if (swap_mp) {
480 mp->mnt_vfc->vfc_refcount--;
481 free(swap_mp, M_MOUNT);
482 }
483 return (error);
484 }
485
486 swap_mp = NULL;
487 if (nd->swap_nblks) {
488
489 /* Convert to DEV_BSIZE instead of Kilobyte */
490 nd->swap_nblks *= 2;
491
492 /*
493 * Create a fake mount point just for the swap vnode so that the
494 * swap file can be on a different server from the rootfs.
495 */
496 nd->swap_args.fh = nd->swap_fh;
497 nd->swap_args.fhsize = nd->swap_fhsize;
498 l = ntohl(nd->swap_saddr.sin_addr.s_addr);
499 snprintf(buf, sizeof(buf), "%ld.%ld.%ld.%ld:%s",
500 (l >> 24) & 0xff, (l >> 16) & 0xff,
501 (l >> 8) & 0xff, (l >> 0) & 0xff, nd->swap_hostnam);
502 printf("NFS SWAP: %s\n", buf);
503 if ((error = nfs_mountdiskless(buf, "/swap", 0,
504 &nd->swap_saddr, &nd->swap_args, td, &vp, &swap_mp)) != 0)
505 return (error);
506 vfs_unbusy(swap_mp, td);
507
508 VTONFS(vp)->n_size = VTONFS(vp)->n_vattr.va_size =
509 nd->swap_nblks * DEV_BSIZE ;
510 /*
511 * Since the swap file is not the root dir of a file system,
512 * hack it to a regular file.
513 */
514 vp->v_type = VREG;
515 vp->v_flag = 0;
516 VREF(vp);
517 swaponvp(td, vp, NODEV, nd->swap_nblks);
518 }
519
520 mp->mnt_flag |= MNT_ROOTFS;
521 mp->mnt_vnodecovered = NULLVP;
522 rootvp = vp;
523 vfs_unbusy(mp, td);
524
525 /*
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.
529 */
530 bcopy(nd->my_hostnam, hostname, MAXHOSTNAMELEN);
531 hostname[MAXHOSTNAMELEN - 1] = '\0';
532 for (i = 0; i < MAXHOSTNAMELEN; i++)
533 if (hostname[i] == '\0')
534 break;
535 inittodr(ntohl(nd->root_time));
536 return (0);
537}
538
539/*
540 * Internal version of mount system call for diskless setup.
541 */
542static int
543nfs_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)
546{
547 struct mount *mp;
548 struct sockaddr *nam;
549 int error;
550
551 mp = *mpp;
552
553 if (!mp && (error = vfs_rootmountalloc("nfs", path, &mp))) {
554 printf("nfs_mountroot: NFS not configured");
555 return (error);
556 }
557
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--;
565 vfs_unbusy(mp, td);
566 free(mp, M_MOUNT);
567 FREE(nam, M_SONAME);
568 return (error);
569 }
570 (void) copystr(which, mp->mnt_stat.f_mntonname, MNAMELEN - 1, 0);
571 *mpp = mp;
572 return (0);
573}
574
575static void
576nfs_decode_args(struct nfsmount *nmp, struct nfs_args *argp)
577{
578 int s;
579 int adjsock;
580 int maxio;
581
582 s = splnet();
583 /*
584 * Silently clear NFSMNT_NOCONN if it's a TCP mount, it makes
585 * no sense in that context.
586 */
587 if (argp->sotype == SOCK_STREAM)
588 nmp->nm_flag &= ~NFSMNT_NOCONN;
589
590 /* Also clear RDIRPLUS if not NFSv3, it crashes some servers */
591 if ((argp->flags & NFSMNT_NFSV3) == 0)
592 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
593
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));
600
601 /* Update flags atomically. Don't change the lock bits. */
602 nmp->nm_flag = argp->flags | nmp->nm_flag;
603 splx(s);
604
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;
611 }
612
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;
617 }
618
619 if (argp->flags & NFSMNT_NFSV3) {
620 if (argp->sotype == SOCK_DGRAM)
621 maxio = NFS_MAXDGRAMDATA;
622 else
623 maxio = NFS_MAXDATA;
624 } else
625 maxio = NFS_V2MAXDATA;
626
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;
633 }
634 if (nmp->nm_wsize > maxio)
635 nmp->nm_wsize = maxio;
636 if (nmp->nm_wsize > MAXBSIZE)
637 nmp->nm_wsize = MAXBSIZE;
638
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;
645 }
646 if (nmp->nm_rsize > maxio)
647 nmp->nm_rsize = maxio;
648 if (nmp->nm_rsize > MAXBSIZE)
649 nmp->nm_rsize = MAXBSIZE;
650
651 if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) {
652 nmp->nm_readdirsize = argp->readdirsize;
653 }
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;
658
659 if ((argp->flags & NFSMNT_ACREGMIN) && argp->acregmin >= 0)
660 nmp->nm_acregmin = argp->acregmin;
661 else
662 nmp->nm_acregmin = NFS_MINATTRTIMO;
663 if ((argp->flags & NFSMNT_ACREGMAX) && argp->acregmax >= 0)
664 nmp->nm_acregmax = argp->acregmax;
665 else
666 nmp->nm_acregmax = NFS_MAXATTRTIMO;
667 if ((argp->flags & NFSMNT_ACDIRMIN) && argp->acdirmin >= 0)
668 nmp->nm_acdirmin = argp->acdirmin;
669 else
670 nmp->nm_acdirmin = NFS_MINDIRATTRTIMO;
671 if ((argp->flags & NFSMNT_ACDIRMAX) && argp->acdirmax >= 0)
672 nmp->nm_acdirmax = argp->acdirmax;
673 else
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;
679
680 if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0) {
681 if (argp->maxgrouplist <= NFS_MAXGRPS)
682 nmp->nm_numgrps = argp->maxgrouplist;
683 else
684 nmp->nm_numgrps = NFS_MAXGRPS;
685 }
686 if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0) {
687 if (argp->readahead <= NFS_MAXRAHEAD)
688 nmp->nm_readahead = argp->readahead;
689 else
690 nmp->nm_readahead = NFS_MAXRAHEAD;
691 }
692 if ((argp->flags & NFSMNT_DEADTHRESH) && argp->deadthresh >= 0) {
693 if (argp->deadthresh <= NFS_MAXDEADTHRESH)
694 nmp->nm_deadthresh = argp->deadthresh;
695 else
696 nmp->nm_deadthresh = NFS_MAXDEADTHRESH;
697 }
698
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;
703
704 if (nmp->nm_so && adjsock) {
705 nfs_safedisconnect(nmp);
706 if (nmp->nm_sotype == SOCK_DGRAM)
707 while (nfs_connect(nmp, (struct nfsreq *)0)) {
708 printf("nfs_args: retrying connect\n");
709 (void) tsleep((caddr_t)&lbolt,
710 PSOCK, "nfscon", 0);
711 }
712 }
713}
714
715/*
716 * VFS Operations.
717 *
718 * mount system call
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.
723 */
724/* ARGSUSED */
725static int
726nfs_mount(struct mount *mp, char *path, caddr_t data, struct nameidata *ndp,
727 struct thread *td)
728{
729 int error;
730 struct nfs_args args;
731 struct sockaddr *nam;
732 struct vnode *vp;
733 char hst[MNAMELEN];
734 size_t len;
735 u_char nfh[NFSX_V3FHMAX];
736
737 if (path == NULL) {
738 nfs_mountroot(mp, td);
739 return (0);
740 }
741 error = copyin(data, (caddr_t)&args, sizeof (struct nfs_args));
742 if (error)
743 return (error);
744 if (args.version != NFS_ARGSVERSION) {
745#ifdef COMPAT_PRELITE2
746 /*
747 * If the argument version is unknown, then assume the
748 * caller is a pre-lite2 4.4BSD client and convert its
749 * arguments.
750 */
751 struct onfs_args oargs;
752 error = copyin(data, (caddr_t)&oargs, sizeof (struct onfs_args));
753 if (error)
754 return (error);
755 nfs_convert_oargs(&args,&oargs);
756#else /* !COMPAT_PRELITE2 */
757 return (EPROGMISMATCH);
758#endif /* COMPAT_PRELITE2 */
759 }
760 if (mp->mnt_flag & MNT_UPDATE) {
761 struct nfsmount *nmp = VFSTONFS(mp);
762
763 if (nmp == NULL)
764 return (EIO);
765 /*
766 * When doing an update, we can't change from or to
767 * v3, switch lockd strategies or change cookie translation
768 */
769 args.flags = (args.flags &
770 ~(NFSMNT_NFSV3 | NFSMNT_NOLOCKD /*|NFSMNT_XLATECOOKIE*/)) |
771 (nmp->nm_flag &
772 (NFSMNT_NFSV3 | NFSMNT_NOLOCKD /*|NFSMNT_XLATECOOKIE*/));
773 nfs_decode_args(nmp, &args);
774 return (0);
775 }
776 if (args.fhsize < 0 || args.fhsize > NFSX_V3FHMAX)
777 return (EINVAL);
778 error = copyin((caddr_t)args.fh, (caddr_t)nfh, args.fhsize);
779 if (error)
780 return (error);
781 error = copyinstr(args.hostname, hst, MNAMELEN-1, &len);
782 if (error)
783 return (error);
784 bzero(&hst[len], MNAMELEN - len);
785 /* sockargs() call must be after above copyin() calls */
786 error = getsockaddr(&nam, (caddr_t)args.addr, args.addrlen);
787 if (error)
788 return (error);
789 args.fh = nfh;
790 error = mountnfs(&args, mp, nam, path, hst, &vp, td->td_ucred);
791 return (error);
792}
793
794/*
795 * Common code for mount and mountroot
796 */
797static int
798mountnfs(struct nfs_args *argp, struct mount *mp, struct sockaddr *nam,
799 char *pth, char *hst, struct vnode **vpp, struct ucred *cred)
800{
801 struct nfsmount *nmp;
802 struct nfsnode *np;
803 int error;
804 struct vattr attrs;
805
806 if (mp->mnt_flag & MNT_UPDATE) {
807 nmp = VFSTONFS(mp);
808 /* update paths, file handles, etc, here XXX */
809 FREE(nam, M_SONAME);
810 return (0);
811 } else {
812 nmp = zalloc(nfsmount_zone);
812 nmp = uma_zalloc(nfsmount_zone, M_WAITOK);
813 bzero((caddr_t)nmp, sizeof (struct nfsmount));
814 TAILQ_INIT(&nmp->nm_bufq);
815 mp->mnt_data = (qaddr_t)nmp;
816 }
817 vfs_getnewfsid(mp);
818 nmp->nm_mountp = mp;
819 nmp->nm_cred = crhold(cred);
820
821 /*
822 * V2 can only handle 32 bit filesizes. A 4GB-1 limit may be too
823 * high, depending on whether we end up with negative offsets in
824 * the client or server somewhere. 2GB-1 may be safer.
825 *
826 * For V3, nfs_fsinfo will adjust this as necessary. Assume maximum
827 * that we can handle until we find out otherwise.
828 * XXX Our "safe" limit on the client is what we can store in our
829 * buffer cache using signed(!) block numbers.
830 */
831 if ((argp->flags & NFSMNT_NFSV3) == 0)
832 nmp->nm_maxfilesize = 0xffffffffLL;
833 else
834 nmp->nm_maxfilesize = (u_int64_t)0x80000000 * DEV_BSIZE - 1;
835
836 nmp->nm_timeo = NFS_TIMEO;
837 nmp->nm_retry = NFS_RETRANS;
838 nmp->nm_wsize = NFS_WSIZE;
839 nmp->nm_rsize = NFS_RSIZE;
840 nmp->nm_readdirsize = NFS_READDIRSIZE;
841 nmp->nm_numgrps = NFS_MAXGRPS;
842 nmp->nm_readahead = NFS_DEFRAHEAD;
843 nmp->nm_deadthresh = NFS_MAXDEADTHRESH;
844 nmp->nm_fhsize = argp->fhsize;
845 bcopy((caddr_t)argp->fh, (caddr_t)nmp->nm_fh, argp->fhsize);
846 bcopy(hst, mp->mnt_stat.f_mntfromname, MNAMELEN);
847 bcopy(pth, mp->mnt_stat.f_mntonname, MNAMELEN);
848 nmp->nm_nam = nam;
849 /* Set up the sockets and per-host congestion */
850 nmp->nm_sotype = argp->sotype;
851 nmp->nm_soproto = argp->proto;
852
853 nfs_decode_args(nmp, argp);
854
855 /*
856 * For Connection based sockets (TCP,...) defer the connect until
857 * the first request, in case the server is not responding.
858 */
859 if (nmp->nm_sotype == SOCK_DGRAM &&
860 (error = nfs_connect(nmp, (struct nfsreq *)0)))
861 goto bad;
862
863 /*
864 * This is silly, but it has to be set so that vinifod() works.
865 * We do not want to do an nfs_statfs() here since we can get
866 * stuck on a dead server and we are holding a lock on the mount
867 * point.
868 */
869 mp->mnt_stat.f_iosize = nfs_iosize(nmp);
870 /*
871 * A reference count is needed on the nfsnode representing the
872 * remote root. If this object is not persistent, then backward
873 * traversals of the mount point (i.e. "..") will not work if
874 * the nfsnode gets flushed out of the cache. Ufs does not have
875 * this problem, because one can identify root inodes by their
876 * number == ROOTINO (2).
877 */
878 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
879 if (error)
880 goto bad;
881 *vpp = NFSTOV(np);
882
883 /*
884 * Get file attributes for the mountpoint. This has the side
885 * effect of filling in (*vpp)->v_type with the correct value.
886 */
887 VOP_GETATTR(*vpp, &attrs, curthread->td_ucred, curthread);
888
889 /*
890 * Lose the lock but keep the ref.
891 */
892 VOP_UNLOCK(*vpp, 0, curthread);
893
894 return (0);
895bad:
896 nfs_disconnect(nmp);
897 crfree(nmp->nm_cred);
898 zfree(nfsmount_zone, nmp);
898 uma_zfree(nfsmount_zone, nmp);
899 FREE(nam, M_SONAME);
900 return (error);
901}
902
903/*
904 * unmount system call
905 */
906static int
907nfs_unmount(struct mount *mp, int mntflags, struct thread *td)
908{
909 struct nfsmount *nmp;
910 int error, flags = 0;
911
912 if (mntflags & MNT_FORCE)
913 flags |= FORCECLOSE;
914 nmp = VFSTONFS(mp);
915 /*
916 * Goes something like this..
917 * - Call vflush() to clear out vnodes for this file system
918 * - Close the socket
919 * - Free up the data structures
920 */
921 /* In the forced case, cancel any outstanding requests. */
922 if (flags & FORCECLOSE) {
923 error = nfs_nmcancelreqs(nmp);
924 if (error)
925 return (error);
926 }
927 /* We hold 1 extra ref on the root vnode; see comment in mountnfs(). */
928 error = vflush(mp, 1, flags);
929 if (error)
930 return (error);
931
932 /*
933 * We are now committed to the unmount.
934 */
935 nfs_disconnect(nmp);
936 FREE(nmp->nm_nam, M_SONAME);
937
938 crfree(nmp->nm_cred);
939 zfree(nfsmount_zone, nmp);
939 uma_zfree(nfsmount_zone, nmp);
940 return (0);
941}
942
943/*
944 * Return root of a filesystem
945 */
946static int
947nfs_root(struct mount *mp, struct vnode **vpp)
948{
949 struct vnode *vp;
950 struct nfsmount *nmp;
951 struct nfsnode *np;
952 int error;
953
954 nmp = VFSTONFS(mp);
955 error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np);
956 if (error)
957 return (error);
958 vp = NFSTOV(np);
959 if (vp->v_type == VNON)
960 vp->v_type = VDIR;
961 vp->v_flag = VROOT;
962 *vpp = vp;
963 return (0);
964}
965
966/*
967 * Flush out the buffer cache
968 */
969/* ARGSUSED */
970static int
971nfs_sync(struct mount *mp, int waitfor, struct ucred *cred, struct thread *td)
972{
973 struct vnode *vp, *vnp;
974 int error, allerror = 0;
975
976 /*
977 * Force stale buffer cache information to be flushed.
978 */
979 mtx_lock(&mntvnode_mtx);
980loop:
981 for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
982 vp != NULL;
983 vp = vnp) {
984 /*
985 * If the vnode that we are about to sync is no longer
986 * associated with this mount point, start over.
987 */
988 if (vp->v_mount != mp)
989 goto loop;
990 vnp = TAILQ_NEXT(vp, v_nmntvnodes);
991 mtx_unlock(&mntvnode_mtx);
992 mtx_lock(&vp->v_interlock);
993 if (VOP_ISLOCKED(vp, NULL) || TAILQ_EMPTY(&vp->v_dirtyblkhd) ||
994 waitfor == MNT_LAZY) {
995 mtx_unlock(&vp->v_interlock);
996 mtx_lock(&mntvnode_mtx);
997 continue;
998 }
999 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
1000 mtx_lock(&mntvnode_mtx);
1001 goto loop;
1002 }
1003 error = VOP_FSYNC(vp, cred, waitfor, td);
1004 if (error)
1005 allerror = error;
1006 vput(vp);
1007 mtx_lock(&mntvnode_mtx);
1008 }
1009 mtx_unlock(&mntvnode_mtx);
1010 return (allerror);
1011}