xref: /openbsd-src/sys/nfs/nfs_vfsops.c (revision 897fc685943471cf985a0fe38ba076ea6fe74fa5)
1 /*	$OpenBSD: nfs_vfsops.c,v 1.117 2018/04/09 09:39:53 mpi Exp $	*/
2 /*	$NetBSD: nfs_vfsops.c,v 1.46.4.1 1996/05/25 22:40:35 fvdl Exp $	*/
3 
4 /*
5  * Copyright (c) 1989, 1993, 1995
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Rick Macklem at The University of Guelph.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)nfs_vfsops.c	8.12 (Berkeley) 5/20/95
36  */
37 
38 #include <sys/param.h>
39 #include <sys/conf.h>
40 #include <sys/ioctl.h>
41 #include <sys/signal.h>
42 #include <sys/proc.h>
43 #include <sys/namei.h>
44 #include <sys/vnode.h>
45 #include <sys/lock.h>
46 #include <sys/kernel.h>
47 #include <sys/mount.h>
48 #include <sys/swap.h>
49 #include <sys/buf.h>
50 #include <sys/mbuf.h>
51 #include <sys/dirent.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/systm.h>
55 #include <sys/sysctl.h>
56 #include <sys/queue.h>
57 
58 #include <netinet/in.h>
59 
60 #include <nfs/rpcv2.h>
61 #include <nfs/nfsproto.h>
62 #include <nfs/nfsnode.h>
63 #include <nfs/nfs.h>
64 #include <nfs/nfsmount.h>
65 #include <nfs/xdr_subs.h>
66 #include <nfs/nfsm_subs.h>
67 #include <nfs/nfsdiskless.h>
68 #include <nfs/nfs_var.h>
69 
70 extern struct nfsstats nfsstats;
71 extern int nfs_ticks;
72 extern u_int32_t nfs_procids[NFS_NPROCS];
73 
74 int	nfs_sysctl(int *, u_int, void *, size_t *, void *, size_t,
75 	    struct proc *);
76 int	nfs_checkexp(struct mount *, struct mbuf *, int *, struct ucred **);
77 struct	mount *nfs_mount_diskless(struct nfs_dlmount *, char *, int,
78 	    struct vnode **, struct proc *p);
79 int	mountnfs(struct nfs_args *, struct mount *, struct mbuf *,
80 	    const char *, char *, struct vnode **, struct proc *p);
81 int	nfs_quotactl(struct mount *, int, uid_t, caddr_t, struct proc *);
82 int	nfs_root(struct mount *, struct vnode **);
83 int	nfs_start(struct mount *, int, struct proc *);
84 int	nfs_statfs(struct mount *, struct statfs *, struct proc *);
85 int	nfs_sync(struct mount *, int, int, struct ucred *, struct proc *);
86 int	nfs_unmount(struct mount *, int, struct proc *);
87 int	nfs_vget(struct mount *, ino_t, struct vnode **);
88 int	nfs_vptofh(struct vnode *, struct fid *);
89 int	nfs_mountroot(void);
90 void	nfs_decode_args(struct nfsmount *, struct nfs_args *,
91 	    struct nfs_args *);
92 int	nfs_fhtovp(struct mount *, struct fid *, struct vnode **);
93 
94 /*
95  * nfs vfs operations.
96  */
97 const struct vfsops nfs_vfsops = {
98 	nfs_mount,
99 	nfs_start,
100 	nfs_unmount,
101 	nfs_root,
102 	nfs_quotactl,
103 	nfs_statfs,
104 	nfs_sync,
105 	nfs_vget,
106 	nfs_fhtovp,
107 	nfs_vptofh,
108 	nfs_vfs_init,
109 	nfs_sysctl,
110 	nfs_checkexp
111 };
112 
113 /*
114  * nfs statfs call
115  */
116 int
117 nfs_statfs(struct mount *mp, struct statfs *sbp, struct proc *p)
118 {
119 	struct vnode *vp;
120 	struct nfs_statfs *sfp = NULL;
121 	struct nfsm_info	info;
122 	u_int32_t *tl;
123 	int32_t t1;
124 	caddr_t cp2;
125 	struct nfsmount *nmp = VFSTONFS(mp);
126 	int error = 0, retattr;
127 	struct ucred *cred;
128 	u_quad_t tquad;
129 
130 	info.nmi_v3 = (nmp->nm_flag & NFSMNT_NFSV3);
131 
132 	error = nfs_root(mp, &vp);
133 	if (error)
134 		return (error);
135 	cred = crget();
136 	cred->cr_ngroups = 0;
137 	if (info.nmi_v3 && (nmp->nm_flag & NFSMNT_GOTFSINFO) == 0)
138 		(void)nfs_fsinfo(nmp, vp, cred, p);
139 	nfsstats.rpccnt[NFSPROC_FSSTAT]++;
140 	info.nmi_mb = info.nmi_mreq = nfsm_reqhead(NFSX_FH(info.nmi_v3));
141 	nfsm_fhtom(&info, vp, info.nmi_v3);
142 
143 	info.nmi_procp = p;
144 	info.nmi_cred = cred;
145 	error = nfs_request(vp, NFSPROC_FSSTAT, &info);
146 	if (info.nmi_v3)
147 		nfsm_postop_attr(vp, retattr);
148 	if (error) {
149 		m_freem(info.nmi_mrep);
150 		goto nfsmout;
151 	}
152 
153 	nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(info.nmi_v3));
154 	sbp->f_iosize = min(nmp->nm_rsize, nmp->nm_wsize);
155 	if (info.nmi_v3) {
156 		sbp->f_bsize = NFS_FABLKSIZE;
157 		tquad = fxdr_hyper(&sfp->sf_tbytes);
158 		sbp->f_blocks = tquad / (u_quad_t)NFS_FABLKSIZE;
159 		tquad = fxdr_hyper(&sfp->sf_fbytes);
160 		sbp->f_bfree = tquad / (u_quad_t)NFS_FABLKSIZE;
161 		tquad = fxdr_hyper(&sfp->sf_abytes);
162 		sbp->f_bavail = (quad_t)tquad / (quad_t)NFS_FABLKSIZE;
163 
164 		tquad = fxdr_hyper(&sfp->sf_tfiles);
165 		sbp->f_files = tquad;
166 		tquad = fxdr_hyper(&sfp->sf_ffiles);
167 		sbp->f_ffree = tquad;
168 		sbp->f_favail = tquad;
169 	} else {
170 		sbp->f_bsize = fxdr_unsigned(int32_t, sfp->sf_bsize);
171 		sbp->f_blocks = fxdr_unsigned(int32_t, sfp->sf_blocks);
172 		sbp->f_bfree = fxdr_unsigned(int32_t, sfp->sf_bfree);
173 		sbp->f_bavail = fxdr_unsigned(int32_t, sfp->sf_bavail);
174 		sbp->f_files = 0;
175 		sbp->f_ffree = 0;
176 		sbp->f_favail = 0;
177 	}
178 	copy_statfs_info(sbp, mp);
179 	m_freem(info.nmi_mrep);
180 nfsmout:
181 	vput(vp);
182 	crfree(cred);
183 	return (error);
184 }
185 
186 /*
187  * nfs version 3 fsinfo rpc call
188  */
189 int
190 nfs_fsinfo(struct nfsmount *nmp, struct vnode *vp, struct ucred *cred,
191     struct proc *p)
192 {
193 	struct nfsv3_fsinfo *fsp;
194 	struct nfsm_info	info;
195 	int32_t t1;
196 	u_int32_t *tl, pref, max;
197 	caddr_t cp2;
198 	int error = 0, retattr;
199 
200 	nfsstats.rpccnt[NFSPROC_FSINFO]++;
201 	info.nmi_mb = info.nmi_mreq = nfsm_reqhead(NFSX_FH(1));
202 	nfsm_fhtom(&info, vp, 1);
203 
204 	info.nmi_procp = p;
205 	info.nmi_cred = cred;
206 	error = nfs_request(vp, NFSPROC_FSINFO, &info);
207 
208 	nfsm_postop_attr(vp, retattr);
209 	if (error) {
210 		m_freem(info.nmi_mrep);
211 		goto nfsmout;
212 	}
213 
214 	nfsm_dissect(fsp, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
215 	pref = fxdr_unsigned(u_int32_t, fsp->fs_wtpref);
216 	if (pref < nmp->nm_wsize)
217 		nmp->nm_wsize = (pref + NFS_FABLKSIZE - 1) &
218 			~(NFS_FABLKSIZE - 1);
219 	max = fxdr_unsigned(u_int32_t, fsp->fs_wtmax);
220 	if (max < nmp->nm_wsize) {
221 		nmp->nm_wsize = max & ~(NFS_FABLKSIZE - 1);
222 		if (nmp->nm_wsize == 0)
223 			nmp->nm_wsize = max;
224 	}
225 	pref = fxdr_unsigned(u_int32_t, fsp->fs_rtpref);
226 	if (pref < nmp->nm_rsize)
227 		nmp->nm_rsize = (pref + NFS_FABLKSIZE - 1) &
228 			~(NFS_FABLKSIZE - 1);
229 	max = fxdr_unsigned(u_int32_t, fsp->fs_rtmax);
230 	if (max < nmp->nm_rsize) {
231 		nmp->nm_rsize = max & ~(NFS_FABLKSIZE - 1);
232 		if (nmp->nm_rsize == 0)
233 			nmp->nm_rsize = max;
234 	}
235 	pref = fxdr_unsigned(u_int32_t, fsp->fs_dtpref);
236 	if (pref < nmp->nm_readdirsize)
237 		nmp->nm_readdirsize = (pref + NFS_DIRBLKSIZ - 1) &
238 			~(NFS_DIRBLKSIZ - 1);
239 	if (max < nmp->nm_readdirsize) {
240 		nmp->nm_readdirsize = max & ~(NFS_DIRBLKSIZ - 1);
241 		if (nmp->nm_readdirsize == 0)
242 			nmp->nm_readdirsize = max;
243 	}
244 	nmp->nm_flag |= NFSMNT_GOTFSINFO;
245 
246 	m_freem(info.nmi_mrep);
247 nfsmout:
248 	return (error);
249 }
250 
251 struct nfs_diskless nfs_diskless;
252 
253 /*
254  * Mount a remote root fs via. NFS.  It goes like this:
255  * - Call nfs_boot_init() to fill in the nfs_diskless struct
256  *   (using RARP, bootparam RPC, mountd RPC)
257  * - hand craft the swap nfs vnode hanging off a fake mount point
258  *	if swdevt[0].sw_dev == NODEV
259  * - build the rootfs mount point and call mountnfs() to do the rest.
260  */
261 int
262 nfs_mountroot(void)
263 {
264 	struct vattr attr;
265 	struct mount *mp;
266 	struct vnode *vp;
267 	struct proc *procp;
268 	long n;
269 	int error;
270 
271 	procp = curproc; /* XXX */
272 
273 	/*
274 	 * Call nfs_boot_init() to fill in the nfs_diskless struct.
275 	 * Side effect:	 Finds and configures a network interface.
276 	 */
277 	nfs_boot_init(&nfs_diskless, procp);
278 
279 	/*
280 	 * Create the root mount point.
281 	 */
282 	if (nfs_boot_getfh(&nfs_diskless.nd_boot, "root", &nfs_diskless.nd_root, -1))
283 		panic("nfs_mountroot: root");
284 	mp = nfs_mount_diskless(&nfs_diskless.nd_root, "/", 0, &vp, procp);
285 	printf("root on %s\n", nfs_diskless.nd_root.ndm_host);
286 
287 	/*
288 	 * Link it into the mount list.
289 	 */
290 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
291 	rootvp = vp;
292 	vfs_unbusy(mp);
293 
294 	/* Get root attributes (for the time). */
295 	error = VOP_GETATTR(rootvp, &attr, procp->p_ucred, procp);
296 	if (error) panic("nfs_mountroot: getattr for root");
297 	n = attr.va_atime.tv_sec;
298 #ifdef	DEBUG
299 	printf("root time: 0x%lx\n", n);
300 #endif
301 	inittodr(n);
302 
303 #ifdef notyet
304 	/* Set up swap credentials. */
305 	proc0.p_ucred->cr_uid = ntohl(nfs_diskless.swap_ucred.cr_uid);
306 	proc0.p_ucred->cr_gid = ntohl(nfs_diskless.swap_ucred.cr_gid);
307 	if ((proc0.p_ucred->cr_ngroups = ntohs(nfs_diskless.swap_ucred.cr_ngroups)) >
308 		NGROUPS_MAX)
309 		proc0.p_ucred->cr_ngroups = NGROUPS_MAX;
310 	for (i = 0; i < proc0.p_ucred->cr_ngroups; i++)
311 	    proc0.p_ucred->cr_groups[i] = ntohl(nfs_diskless.swap_ucred.cr_groups[i]);
312 #endif
313 
314 	/*
315 	 * "Mount" the swap device.
316 	 *
317 	 * On a "dataless" configuration (swap on disk) we will have:
318 	 *	(swdevt[0].sw_dev != NODEV) identifying the swap device.
319 	 */
320 	if (swdevt[0].sw_dev != NODEV) {
321 		if (bdevvp(swapdev, &swapdev_vp))
322 			panic("nfs_mountroot: can't setup swap vp");
323 		printf("swap on device 0x%x\n", swdevt[0].sw_dev);
324 		return (0);
325 	}
326 
327 	/*
328 	 * If swapping to an nfs node:	(swdevt[0].sw_dev == NODEV)
329 	 * Create a fake mount point just for the swap vnode so that the
330 	 * swap file can be on a different server from the rootfs.
331 	 *
332 	 * Wait 5 retries, finally no swap is cool. -mickey
333 	 */
334 	error = nfs_boot_getfh(&nfs_diskless.nd_boot, "swap", &nfs_diskless.nd_swap, 5);
335 	if (!error) {
336 		mp = nfs_mount_diskless(&nfs_diskless.nd_swap, "/swap", 0, &vp,
337 		    procp);
338 		vfs_unbusy(mp);
339 
340 		/*
341 		 * Since the swap file is not the root dir of a file system,
342 		 * hack it to a regular file.
343 		 */
344 		vp->v_type = VREG;
345 		vp->v_flag = 0;
346 
347 		/*
348 		 * Next line is a hack to make swapmount() work on NFS
349 		 * swap files.
350 		 */
351 		swdevt[0].sw_dev = NETDEV;
352 		/* end hack */
353 		nfs_diskless.sw_vp = vp;
354 
355 		/*
356 		 * Find out how large the swap file is.
357 		 */
358 		error = VOP_GETATTR(vp, &attr, procp->p_ucred, procp);
359 		if (error)
360 			printf("nfs_mountroot: getattr for swap\n");
361 		n = (long) (attr.va_size >> DEV_BSHIFT);
362 
363 		printf("swap on %s\n", nfs_diskless.nd_swap.ndm_host);
364 #ifdef	DEBUG
365 		printf("swap size: 0x%lx (blocks)\n", n);
366 #endif
367 		return (0);
368 	}
369 
370 	printf("WARNING: no swap\n");
371 	swdevt[0].sw_dev = NODEV;
372 	return (0);
373 }
374 
375 /*
376  * Internal version of mount system call for diskless setup.
377  */
378 struct mount *
379 nfs_mount_diskless(struct nfs_dlmount *ndmntp, char *mntname, int mntflag,
380     struct vnode **vpp, struct proc *p)
381 {
382 	struct mount *mp;
383 	struct mbuf *m;
384 	int error;
385 
386 	if (vfs_rootmountalloc("nfs", mntname, &mp))
387 		panic("nfs_mount_diskless: vfs_rootmountalloc failed");
388 	mp->mnt_flag |= mntflag;
389 
390 	/* Get mbuf for server sockaddr. */
391 	m = m_get(M_WAIT, MT_SONAME);
392 	bcopy(ndmntp->ndm_args.addr, mtod(m, caddr_t),
393 	    (m->m_len = ndmntp->ndm_args.addr->sa_len));
394 
395 	error = mountnfs(&ndmntp->ndm_args, mp, m, mntname,
396 	    ndmntp->ndm_args.hostname, vpp, p);
397 	if (error)
398 		panic("nfs_mountroot: mount %s failed: %d", mntname, error);
399 
400 	return (mp);
401 }
402 
403 void
404 nfs_decode_args(struct nfsmount *nmp, struct nfs_args *argp,
405     struct nfs_args *nargp)
406 {
407 	int adjsock = 0;
408 	int maxio;
409 
410 #if 0
411 	/* Re-bind if rsrvd port requested and wasn't on one */
412 	adjsock = !(nmp->nm_flag & NFSMNT_RESVPORT)
413 		  && (argp->flags & NFSMNT_RESVPORT);
414 #endif
415 	/* Also re-bind if we're switching to/from a connected UDP socket */
416 	adjsock |= ((nmp->nm_flag & NFSMNT_NOCONN) !=
417 	    (argp->flags & NFSMNT_NOCONN));
418 
419 	nmp->nm_flag =
420 	    (argp->flags & ~NFSMNT_INTERNAL) | (nmp->nm_flag & NFSMNT_INTERNAL);
421 
422 	if ((argp->flags & NFSMNT_TIMEO) && argp->timeo > 0) {
423 		nmp->nm_timeo = (argp->timeo * NFS_HZ + 5) / 10;
424 		if (nmp->nm_timeo < NFS_MINTIMEO)
425 			nmp->nm_timeo = NFS_MINTIMEO;
426 		else if (nmp->nm_timeo > NFS_MAXTIMEO)
427 			nmp->nm_timeo = NFS_MAXTIMEO;
428 	}
429 
430 	if ((argp->flags & NFSMNT_RETRANS) && argp->retrans > 1)
431 		nmp->nm_retry = MIN(argp->retrans, NFS_MAXREXMIT);
432 	if (!(nmp->nm_flag & NFSMNT_SOFT))
433 		nmp->nm_retry = NFS_MAXREXMIT + 1; /* past clip limit */
434 
435 	if (argp->flags & NFSMNT_NFSV3) {
436 		if (argp->sotype == SOCK_DGRAM)
437 			maxio = NFS_MAXDGRAMDATA;
438 		else
439 			maxio = NFS_MAXDATA;
440 	} else
441 		maxio = NFS_V2MAXDATA;
442 
443 	if ((argp->flags & NFSMNT_WSIZE) && argp->wsize > 0) {
444 		int osize = nmp->nm_wsize;
445 		nmp->nm_wsize = argp->wsize;
446 		/* Round down to multiple of blocksize */
447 		nmp->nm_wsize &= ~(NFS_FABLKSIZE - 1);
448 		if (nmp->nm_wsize <= 0)
449 			nmp->nm_wsize = NFS_FABLKSIZE;
450 		adjsock |= (nmp->nm_wsize != osize);
451 	}
452 	if (nmp->nm_wsize > maxio)
453 		nmp->nm_wsize = maxio;
454 	if (nmp->nm_wsize > MAXBSIZE)
455 		nmp->nm_wsize = MAXBSIZE;
456 
457 	if ((argp->flags & NFSMNT_RSIZE) && argp->rsize > 0) {
458 		int osize = nmp->nm_rsize;
459 		nmp->nm_rsize = argp->rsize;
460 		/* Round down to multiple of blocksize */
461 		nmp->nm_rsize &= ~(NFS_FABLKSIZE - 1);
462 		if (nmp->nm_rsize <= 0)
463 			nmp->nm_rsize = NFS_FABLKSIZE;
464 		adjsock |= (nmp->nm_rsize != osize);
465 	}
466 	if (nmp->nm_rsize > maxio)
467 		nmp->nm_rsize = maxio;
468 	if (nmp->nm_rsize > MAXBSIZE)
469 		nmp->nm_rsize = MAXBSIZE;
470 
471 	if ((argp->flags & NFSMNT_READDIRSIZE) && argp->readdirsize > 0) {
472 		nmp->nm_readdirsize = argp->readdirsize;
473 		/* Round down to multiple of blocksize */
474 		nmp->nm_readdirsize &= ~(NFS_DIRBLKSIZ - 1);
475 		if (nmp->nm_readdirsize < NFS_DIRBLKSIZ)
476 			nmp->nm_readdirsize = NFS_DIRBLKSIZ;
477 	} else if (argp->flags & NFSMNT_RSIZE)
478 		nmp->nm_readdirsize = nmp->nm_rsize;
479 
480 	if (nmp->nm_readdirsize > maxio)
481 		nmp->nm_readdirsize = maxio;
482 
483 	if ((argp->flags & NFSMNT_MAXGRPS) && argp->maxgrouplist >= 0 &&
484 		argp->maxgrouplist <= NFS_MAXGRPS)
485 		nmp->nm_numgrps = argp->maxgrouplist;
486 	if ((argp->flags & NFSMNT_READAHEAD) && argp->readahead >= 0 &&
487 		argp->readahead <= NFS_MAXRAHEAD)
488 		nmp->nm_readahead = argp->readahead;
489 	if (argp->flags & NFSMNT_ACREGMIN && argp->acregmin >= 0) {
490 		if (argp->acregmin > 0xffff)
491 			nmp->nm_acregmin = 0xffff;
492 		else
493 			nmp->nm_acregmin = argp->acregmin;
494 	}
495 	if (argp->flags & NFSMNT_ACREGMAX && argp->acregmax >= 0) {
496 		if (argp->acregmax > 0xffff)
497 			nmp->nm_acregmax = 0xffff;
498 		else
499 			nmp->nm_acregmax = argp->acregmax;
500 	}
501 	if (nmp->nm_acregmin > nmp->nm_acregmax)
502 	  nmp->nm_acregmin = nmp->nm_acregmax;
503 
504 	if (argp->flags & NFSMNT_ACDIRMIN && argp->acdirmin >= 0) {
505 		if (argp->acdirmin > 0xffff)
506 			nmp->nm_acdirmin = 0xffff;
507 		else
508 			nmp->nm_acdirmin = argp->acdirmin;
509 	}
510 	if (argp->flags & NFSMNT_ACDIRMAX && argp->acdirmax >= 0) {
511 		if (argp->acdirmax > 0xffff)
512 			nmp->nm_acdirmax = 0xffff;
513 		else
514 			nmp->nm_acdirmax = argp->acdirmax;
515 	}
516 	if (nmp->nm_acdirmin > nmp->nm_acdirmax)
517 	  nmp->nm_acdirmin = nmp->nm_acdirmax;
518 
519 	if (nmp->nm_so && adjsock) {
520 		nfs_disconnect(nmp);
521 		if (nmp->nm_sotype == SOCK_DGRAM)
522 			while (nfs_connect(nmp, NULL)) {
523 				printf("nfs_args: retrying connect\n");
524 				(void) tsleep(&lbolt,
525 					      PSOCK, "nfscon", 0);
526 			}
527 	}
528 
529 	/* Update nargp based on nmp */
530 	nargp->wsize = nmp->nm_wsize;
531 	nargp->rsize = nmp->nm_rsize;
532 	nargp->readdirsize = nmp->nm_readdirsize;
533 	nargp->timeo = nmp->nm_timeo;
534 	nargp->retrans = nmp->nm_retry;
535 	nargp->maxgrouplist = nmp->nm_numgrps;
536 	nargp->readahead = nmp->nm_readahead;
537 	nargp->acregmin = nmp->nm_acregmin;
538 	nargp->acregmax = nmp->nm_acregmax;
539 	nargp->acdirmin = nmp->nm_acdirmin;
540 	nargp->acdirmax = nmp->nm_acdirmax;
541 }
542 
543 /*
544  * VFS Operations.
545  *
546  * mount system call
547  * It seems a bit dumb to copyinstr() the host here and then
548  * bcopy() it in mountnfs(), but I wanted to detect errors before
549  * doing the sockargs() call because sockargs() allocates an mbuf and
550  * an error after that means that I have to release the mbuf.
551  */
552 /* ARGSUSED */
553 int
554 nfs_mount(struct mount *mp, const char *path, void *data,
555     struct nameidata *ndp, struct proc *p)
556 {
557 	int error;
558 	struct nfs_args *args = data;
559 	struct mbuf *nam;
560 	struct vnode *vp;
561 	char hst[MNAMELEN];
562 	size_t len;
563 	u_char nfh[NFSX_V3FHMAX];
564 
565 	if (args &&
566 	    (args->flags & (NFSMNT_NFSV3|NFSMNT_RDIRPLUS)) == NFSMNT_RDIRPLUS)
567 		return (EINVAL);
568 
569 	if (nfs_niothreads < 0) {
570 		nfs_niothreads = 4;
571 		nfs_getset_niothreads(1);
572 	}
573 
574 	if (mp->mnt_flag & MNT_UPDATE) {
575 		struct nfsmount *nmp = VFSTONFS(mp);
576 
577 		if (nmp == NULL)
578 			return (EIO);
579 		/*
580 		 * When doing an update, we can't change from or to
581 		 * v3.
582 		 */
583 		if (args) {
584 			args->flags = (args->flags & ~(NFSMNT_NFSV3)) |
585 			    (nmp->nm_flag & (NFSMNT_NFSV3));
586 			nfs_decode_args(nmp, args, &mp->mnt_stat.mount_info.nfs_args);
587 		}
588 		return (0);
589 	}
590 	if (args->fhsize < 0 || args->fhsize > NFSX_V3FHMAX)
591 		return (EINVAL);
592 	error = copyin(args->fh, nfh, args->fhsize);
593 	if (error)
594 		return (error);
595 	error = copyinstr(args->hostname, hst, MNAMELEN-1, &len);
596 	if (error)
597 		return (error);
598 	memset(&hst[len], 0, MNAMELEN - len);
599 	/* sockargs() call must be after above copyin() calls */
600 	error = sockargs(&nam, args->addr, args->addrlen, MT_SONAME);
601 	if (error)
602 		return (error);
603 	args->fh = nfh;
604 	error = mountnfs(args, mp, nam, path, hst, &vp, p);
605 	return (error);
606 }
607 
608 /*
609  * Common code for mount and mountroot
610  */
611 int
612 mountnfs(struct nfs_args *argp, struct mount *mp, struct mbuf *nam,
613     const char *pth, char *hst, struct vnode **vpp, struct proc *p)
614 {
615 	struct nfsmount *nmp;
616 	struct nfsnode *np;
617 	struct vnode *vp;
618 	struct vattr attr;
619 	int error;
620 
621 	if (mp->mnt_flag & MNT_UPDATE) {
622 		nmp = VFSTONFS(mp);
623 		/* update paths, file handles, etc, here	XXX */
624 		m_freem(nam);
625 		return (0);
626 	} else {
627 		nmp = malloc(sizeof(*nmp), M_NFSMNT,
628 		    M_WAITOK|M_ZERO);
629 		mp->mnt_data = nmp;
630 	}
631 
632 	vfs_getnewfsid(mp);
633 	nmp->nm_mountp = mp;
634 	nmp->nm_timeo = NFS_TIMEO;
635 	nmp->nm_retry = NFS_RETRANS;
636 	nmp->nm_wsize = NFS_WSIZE;
637 	nmp->nm_rsize = NFS_RSIZE;
638 	nmp->nm_readdirsize = NFS_READDIRSIZE;
639 	nmp->nm_numgrps = NFS_MAXGRPS;
640 	nmp->nm_readahead = NFS_DEFRAHEAD;
641 	nmp->nm_acregmin = NFS_MINATTRTIMO;
642 	nmp->nm_acregmax = NFS_MAXATTRTIMO;
643 	nmp->nm_acdirmin = NFS_MINATTRTIMO;
644 	nmp->nm_acdirmax = NFS_MAXATTRTIMO;
645 	mp->mnt_stat.f_namemax = MAXNAMLEN;
646 	memset(mp->mnt_stat.f_mntonname, 0, MNAMELEN);
647 	strlcpy(mp->mnt_stat.f_mntonname, pth, MNAMELEN);
648 	memset(mp->mnt_stat.f_mntfromname, 0, MNAMELEN);
649 	strlcpy(mp->mnt_stat.f_mntfromname, hst, MNAMELEN);
650 	memset(mp->mnt_stat.f_mntfromspec, 0, MNAMELEN);
651 	strlcpy(mp->mnt_stat.f_mntfromspec, hst, MNAMELEN);
652 	bcopy(argp, &mp->mnt_stat.mount_info.nfs_args, sizeof(*argp));
653 	nmp->nm_nam = nam;
654 	nfs_decode_args(nmp, argp, &mp->mnt_stat.mount_info.nfs_args);
655 
656 	nfs_ninit(nmp);
657 	TAILQ_INIT(&nmp->nm_reqsq);
658 	timeout_set_proc(&nmp->nm_rtimeout, nfs_timer, nmp);
659 
660 	/* Set up the sockets and per-host congestion */
661 	nmp->nm_sotype = argp->sotype;
662 	nmp->nm_soproto = argp->proto;
663 
664 	/*
665 	 * For Connection based sockets (TCP,...) defer the connect until
666 	 * the first request, in case the server is not responding.
667 	 */
668 	if (nmp->nm_sotype == SOCK_DGRAM &&
669 	    (error = nfs_connect(nmp, NULL)))
670 		goto bad;
671 
672 	/*
673 	 * This is silly, but it has to be set so that vinifod() works.
674 	 * We do not want to do an nfs_statfs() here since we can get
675 	 * stuck on a dead server and we are holding a lock on the mount
676 	 * point.
677 	 */
678 	mp->mnt_stat.f_iosize = NFS_MAXDGRAMDATA;
679 	error = nfs_nget(mp, (nfsfh_t *)argp->fh, argp->fhsize, &np);
680 	if (error)
681 		goto bad;
682 	vp = NFSTOV(np);
683 	error = VOP_GETATTR(vp, &attr, p->p_ucred, p);
684 	if (error) {
685 		vput(vp);
686 		goto bad;
687 	}
688 
689 	/*
690 	 * A reference count is needed on the nfsnode representing the
691 	 * remote root.  If this object is not persistent, then backward
692 	 * traversals of the mount point (i.e. "..") will not work if
693 	 * the nfsnode gets flushed out of the cache. Ufs does not have
694 	 * this problem, because one can identify root inodes by their
695 	 * number == ROOTINO (2).  So, just unlock, but no rele.
696 	 */
697 	nmp->nm_vnode = vp;
698 	if (vp->v_type == VNON)
699 		vp->v_type = VDIR;
700 	vp->v_flag = VROOT;
701 	VOP_UNLOCK(vp, curproc);
702 	*vpp = vp;
703 
704 	return (0);
705 bad:
706 	nfs_disconnect(nmp);
707 	free(nmp, M_NFSMNT, sizeof(*nmp));
708 	m_freem(nam);
709 	return (error);
710 }
711 
712 /* unmount system call */
713 int
714 nfs_unmount(struct mount *mp, int mntflags, struct proc *p)
715 {
716 	struct nfsmount *nmp;
717 	struct vnode *vp;
718 	int error, flags = 0;
719 
720 	nmp = VFSTONFS(mp);
721 	error = nfs_root(mp, &vp);
722 	if (error)
723 		return (error);
724 
725 	if ((mntflags & MNT_FORCE) == 0 && vp->v_usecount > 2) {
726 		vput(vp);
727 		return (EBUSY);
728 	}
729 
730 	if (mntflags & MNT_FORCE)
731 		flags |= FORCECLOSE;
732 
733 	error = vflush(mp, vp, flags);
734 	if (error) {
735 		vput(vp);
736 		return (error);
737 	}
738 
739 	/*
740 	 * There are two references count to get rid of here: one
741 	 * from mountnfs() and one from nfs_root() above.
742 	 */
743 	vrele(vp);
744 	vput(vp);
745 	vgone(vp);
746 	nfs_disconnect(nmp);
747 	m_freem(nmp->nm_nam);
748 	timeout_del(&nmp->nm_rtimeout);
749 	free(nmp, M_NFSMNT, sizeof(*nmp));
750 	mp->mnt_data = NULL;
751 	return (0);
752 }
753 
754 /*
755  * Return root of a filesystem
756  */
757 int
758 nfs_root(struct mount *mp, struct vnode **vpp)
759 {
760 	struct vnode *vp;
761 	struct nfsmount *nmp;
762 	int error;
763 
764 	nmp = VFSTONFS(mp);
765 	vp = nmp->nm_vnode;
766 	vref(vp);
767 	error = vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, curproc);
768 	if (error) {
769 		vrele(vp);
770 		return (error);
771 	}
772 	*vpp = vp;
773 	return (0);
774 }
775 
776 /*
777  * Flush out the buffer cache
778  */
779 int
780 nfs_sync(struct mount *mp, int waitfor, int stall, struct ucred *cred, struct proc *p)
781 {
782 	struct vnode *vp;
783 	int error, allerror = 0;
784 
785 	/*
786 	 * Don't traverse the vnode list if we want to skip all of them.
787 	 */
788 	if (waitfor == MNT_LAZY)
789 		return (allerror);
790 
791 	/*
792 	 * Force stale buffer cache information to be flushed.
793 	 */
794 loop:
795 	LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) {
796 		/*
797 		 * If the vnode that we are about to sync is no longer
798 		 * associated with this mount point, start over.
799 		 */
800 		if (vp->v_mount != mp)
801 			goto loop;
802 		if (VOP_ISLOCKED(vp) || LIST_FIRST(&vp->v_dirtyblkhd) == NULL)
803 			continue;
804 		if (vget(vp, LK_EXCLUSIVE, p))
805 			goto loop;
806 		error = VOP_FSYNC(vp, cred, waitfor, p);
807 		if (error)
808 			allerror = error;
809 		vput(vp);
810 	}
811 
812 	return (allerror);
813 }
814 
815 /*
816  * NFS flat namespace lookup.
817  * Currently unsupported.
818  */
819 /* ARGSUSED */
820 int
821 nfs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
822 {
823 
824 	return (EOPNOTSUPP);
825 }
826 
827 /*
828  * Do that sysctl thang...
829  */
830 int
831 nfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
832     size_t newlen, struct proc *p)
833 {
834 	int rv;
835 
836 	/*
837 	 * All names at this level are terminal.
838 	 */
839 	if(namelen > 1)
840 		return ENOTDIR;	/* overloaded */
841 
842 	switch(name[0]) {
843 	case NFS_NFSSTATS:
844 		if(!oldp) {
845 			*oldlenp = sizeof nfsstats;
846 			return 0;
847 		}
848 
849 		if(*oldlenp < sizeof nfsstats) {
850 			*oldlenp = sizeof nfsstats;
851 			return ENOMEM;
852 		}
853 
854 		rv = copyout(&nfsstats, oldp, sizeof nfsstats);
855 		if(rv) return rv;
856 
857 		if(newp && newlen != sizeof nfsstats)
858 			return EINVAL;
859 
860 		if(newp) {
861 			return copyin(newp, &nfsstats, sizeof nfsstats);
862 		}
863 		return 0;
864 
865 	case NFS_NIOTHREADS:
866 		nfs_getset_niothreads(0);
867 
868 		rv = sysctl_int(oldp, oldlenp, newp, newlen, &nfs_niothreads);
869 		if (newp)
870 			nfs_getset_niothreads(1);
871 
872 		return rv;
873 
874 	default:
875 		return EOPNOTSUPP;
876 	}
877 }
878 
879 
880 /*
881  * At this point, this should never happen
882  */
883 /* ARGSUSED */
884 int
885 nfs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
886 {
887 	return (EINVAL);
888 }
889 
890 /*
891  * Vnode pointer to File handle, should never happen either
892  */
893 /* ARGSUSED */
894 int
895 nfs_vptofh(struct vnode *vp, struct fid *fhp)
896 {
897 	return (EINVAL);
898 }
899 
900 /*
901  * Vfs start routine, a no-op.
902  */
903 /* ARGSUSED */
904 int
905 nfs_start(struct mount *mp, int flags, struct proc *p)
906 {
907 	return (0);
908 }
909 
910 /*
911  * Do operations associated with quotas, not supported
912  */
913 /* ARGSUSED */
914 int
915 nfs_quotactl(struct mount *mp, int cmd, uid_t uid, caddr_t arg, struct proc *p)
916 {
917 	return (EOPNOTSUPP);
918 }
919 
920 /*
921  * check export permission, not supported
922  */
923 /* ARGUSED */
924 int
925 nfs_checkexp(struct mount *mp, struct mbuf *nam, int *exflagsp,
926     struct ucred **credanonp)
927 {
928 	return (EOPNOTSUPP);
929 }
930 
931