xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision 162ae9c834f6d9f9cb443bd62cceb23e0b5fef48)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Rick Macklem at The University of Guelph.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  */
35 
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 /*
40  * Functions that need to be different for different versions of BSD
41  * kernel should be kept here, along with any global storage specific
42  * to this BSD variant.
43  */
44 #include <fs/nfs/nfsport.h>
45 #include <sys/smp.h>
46 #include <sys/sysctl.h>
47 #include <sys/taskqueue.h>
48 #include <rpc/rpc_com.h>
49 #include <vm/vm.h>
50 #include <vm/vm_object.h>
51 #include <vm/vm_page.h>
52 #include <vm/vm_param.h>
53 #include <vm/vm_map.h>
54 #include <vm/vm_kern.h>
55 #include <vm/vm_extern.h>
56 #include <vm/uma.h>
57 
58 extern int nfscl_ticks;
59 extern nfsuserd_state nfsrv_nfsuserd;
60 extern struct nfssockreq nfsrv_nfsuserdsock;
61 extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
62     struct thread *);
63 extern int nfsrv_useacl;
64 struct mount nfsv4root_mnt;
65 int newnfs_numnfsd = 0;
66 struct nfsstatsv1 nfsstatsv1;
67 int nfs_numnfscbd = 0;
68 int nfscl_debuglevel = 0;
69 char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
70 struct callout newnfsd_callout;
71 int nfsrv_lughashsize = 100;
72 struct mtx nfsrv_dslock_mtx;
73 struct nfsdevicehead nfsrv_devidhead;
74 volatile int nfsrv_devidcnt = 0;
75 void (*nfsd_call_servertimer)(void) = NULL;
76 void (*ncl_call_invalcaches)(struct vnode *) = NULL;
77 
78 int nfs_pnfsio(task_fn_t *, void *);
79 
80 static int nfs_realign_test;
81 static int nfs_realign_count;
82 static struct ext_nfsstats oldnfsstats;
83 static struct nfsstatsov1 nfsstatsov1;
84 
85 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "NFS filesystem");
86 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
87     0, "Number of realign tests done");
88 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
89     0, "Number of mbuf realignments done");
90 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
91     nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
92     "NFSv4 callback addr for server to use");
93 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
94     0, "Debug level for NFS client");
95 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
96     0, "Size of hash tables for uid/name mapping");
97 int nfs_pnfsiothreads = -1;
98 SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsiothreads, CTLFLAG_RW, &nfs_pnfsiothreads,
99     0, "Number of pNFS mirror I/O threads");
100 
101 /*
102  * Defines for malloc
103  * (Here for FreeBSD, since they allocate storage.)
104  */
105 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
106 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
107 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
108     "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
109 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
110 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
111 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
112 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
113 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
114 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
115 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
116 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
117 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
118 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
119 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
120 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
121 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
122 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer");
123 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
124     "NFS directory offset data");
125 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
126     "NFS local lock rollback");
127 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
128 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
129 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
130 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
131 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
132 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
133 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
134 
135 /*
136  * Definition of mutex locks.
137  * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
138  * and assorted other nfsd structures.
139  */
140 struct mtx newnfsd_mtx;
141 struct mtx nfs_sockl_mutex;
142 struct mtx nfs_state_mutex;
143 struct mtx nfs_nameid_mutex;
144 struct mtx nfs_req_mutex;
145 struct mtx nfs_slock_mutex;
146 struct mtx nfs_clstate_mutex;
147 
148 /* local functions */
149 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
150 
151 #ifdef __NO_STRICT_ALIGNMENT
152 /*
153  * These architectures don't need re-alignment, so just return.
154  */
155 int
156 newnfs_realign(struct mbuf **pm, int how)
157 {
158 
159 	return (0);
160 }
161 #else	/* !__NO_STRICT_ALIGNMENT */
162 /*
163  *	newnfs_realign:
164  *
165  *	Check for badly aligned mbuf data and realign by copying the unaligned
166  *	portion of the data into a new mbuf chain and freeing the portions
167  *	of the old chain that were replaced.
168  *
169  *	We cannot simply realign the data within the existing mbuf chain
170  *	because the underlying buffers may contain other rpc commands and
171  *	we cannot afford to overwrite them.
172  *
173  *	We would prefer to avoid this situation entirely.  The situation does
174  *	not occur with NFS/UDP and is supposed to only occasionally occur
175  *	with TCP.  Use vfs.nfs.realign_count and realign_test to check this.
176  *
177  */
178 int
179 newnfs_realign(struct mbuf **pm, int how)
180 {
181 	struct mbuf *m, *n;
182 	int off, space;
183 
184 	++nfs_realign_test;
185 	while ((m = *pm) != NULL) {
186 		if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
187 			/*
188 			 * NB: we can't depend on m_pkthdr.len to help us
189 			 * decide what to do here.  May not be worth doing
190 			 * the m_length calculation as m_copyback will
191 			 * expand the mbuf chain below as needed.
192 			 */
193 			space = m_length(m, NULL);
194 			if (space >= MINCLSIZE) {
195 				/* NB: m_copyback handles space > MCLBYTES */
196 				n = m_getcl(how, MT_DATA, 0);
197 			} else
198 				n = m_get(how, MT_DATA);
199 			if (n == NULL)
200 				return (ENOMEM);
201 			/*
202 			 * Align the remainder of the mbuf chain.
203 			 */
204 			n->m_len = 0;
205 			off = 0;
206 			while (m != NULL) {
207 				m_copyback(n, off, m->m_len, mtod(m, caddr_t));
208 				off += m->m_len;
209 				m = m->m_next;
210 			}
211 			m_freem(*pm);
212 			*pm = n;
213 			++nfs_realign_count;
214 			break;
215 		}
216 		pm = &m->m_next;
217 	}
218 
219 	return (0);
220 }
221 #endif	/* __NO_STRICT_ALIGNMENT */
222 
223 #ifdef notdef
224 static void
225 nfsrv_object_create(struct vnode *vp, struct thread *td)
226 {
227 
228 	if (vp == NULL || vp->v_type != VREG)
229 		return;
230 	(void) vfs_object_create(vp, td, td->td_ucred);
231 }
232 #endif
233 
234 /*
235  * Look up a file name. Basically just initialize stuff and call namei().
236  */
237 int
238 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
239 {
240 	int error;
241 
242 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
243 	    p);
244 	error = namei(ndp);
245 	if (!error) {
246 		NDFREE(ndp, NDF_ONLY_PNBUF);
247 	}
248 	return (error);
249 }
250 
251 /*
252  * Copy NFS uid, gids to the cred structure.
253  */
254 void
255 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
256 {
257 
258 	KASSERT(nfscr->nfsc_ngroups >= 0,
259 	    ("newnfs_copycred: negative nfsc_ngroups"));
260 	cr->cr_uid = nfscr->nfsc_uid;
261 	crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
262 }
263 
264 /*
265  * Map args from nfsmsleep() to msleep().
266  */
267 int
268 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
269     struct timespec *ts)
270 {
271 	u_int64_t nsecval;
272 	int error, timeo;
273 
274 	if (ts) {
275 		timeo = hz * ts->tv_sec;
276 		nsecval = (u_int64_t)ts->tv_nsec;
277 		nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
278 		    1000000000;
279 		timeo += (int)nsecval;
280 	} else {
281 		timeo = 0;
282 	}
283 	error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
284 	return (error);
285 }
286 
287 /*
288  * Get the file system info for the server. For now, just assume FFS.
289  */
290 void
291 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
292 {
293 	int pref;
294 
295 	/*
296 	 * XXX
297 	 * There should be file system VFS OP(s) to get this information.
298 	 * For now, assume ufs.
299 	 */
300 	if (isdgram)
301 		pref = NFS_MAXDGRAMDATA;
302 	else
303 		pref = NFS_SRVMAXIO;
304 	sip->fs_rtmax = NFS_SRVMAXIO;
305 	sip->fs_rtpref = pref;
306 	sip->fs_rtmult = NFS_FABLKSIZE;
307 	sip->fs_wtmax = NFS_SRVMAXIO;
308 	sip->fs_wtpref = pref;
309 	sip->fs_wtmult = NFS_FABLKSIZE;
310 	sip->fs_dtpref = pref;
311 	sip->fs_maxfilesize = 0xffffffffffffffffull;
312 	sip->fs_timedelta.tv_sec = 0;
313 	sip->fs_timedelta.tv_nsec = 1;
314 	sip->fs_properties = (NFSV3FSINFO_LINK |
315 	    NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
316 	    NFSV3FSINFO_CANSETTIME);
317 }
318 
319 /*
320  * Do the pathconf vnode op.
321  */
322 int
323 nfsvno_pathconf(struct vnode *vp, int flag, long *retf,
324     struct ucred *cred, struct thread *p)
325 {
326 	int error;
327 
328 	error = VOP_PATHCONF(vp, flag, retf);
329 	if (error == EOPNOTSUPP || error == EINVAL) {
330 		/*
331 		 * Some file systems return EINVAL for name arguments not
332 		 * supported and some return EOPNOTSUPP for this case.
333 		 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
334 		 * just fake them.
335 		 */
336 		switch (flag) {
337 		case _PC_LINK_MAX:
338 			*retf = NFS_LINK_MAX;
339 			break;
340 		case _PC_NAME_MAX:
341 			*retf = NAME_MAX;
342 			break;
343 		case _PC_CHOWN_RESTRICTED:
344 			*retf = 1;
345 			break;
346 		case _PC_NO_TRUNC:
347 			*retf = 1;
348 			break;
349 		default:
350 			/*
351 			 * Only happens if a _PC_xxx is added to the server,
352 			 * but this isn't updated.
353 			 */
354 			*retf = 0;
355 			printf("nfsrvd pathconf flag=%d not supp\n", flag);
356 		}
357 		error = 0;
358 	}
359 	NFSEXITCODE(error);
360 	return (error);
361 }
362 
363 /* Fake nfsrv_atroot. Just return 0 */
364 int
365 nfsrv_atroot(struct vnode *vp, uint64_t *retp)
366 {
367 
368 	return (0);
369 }
370 
371 /*
372  * Set the credentials to refer to root.
373  * If only the various BSDen could agree on whether cr_gid is a separate
374  * field or cr_groups[0]...
375  */
376 void
377 newnfs_setroot(struct ucred *cred)
378 {
379 
380 	cred->cr_uid = 0;
381 	cred->cr_groups[0] = 0;
382 	cred->cr_ngroups = 1;
383 }
384 
385 /*
386  * Get the client credential. Used for Renew and recovery.
387  */
388 struct ucred *
389 newnfs_getcred(void)
390 {
391 	struct ucred *cred;
392 	struct thread *td = curthread;
393 
394 	cred = crdup(td->td_ucred);
395 	newnfs_setroot(cred);
396 	return (cred);
397 }
398 
399 /*
400  * Nfs timer routine
401  * Call the nfsd's timer function once/sec.
402  */
403 void
404 newnfs_timer(void *arg)
405 {
406 	static time_t lasttime = 0;
407 	/*
408 	 * Call the server timer, if set up.
409 	 * The argument indicates if it is the next second and therefore
410 	 * leases should be checked.
411 	 */
412 	if (lasttime != NFSD_MONOSEC) {
413 		lasttime = NFSD_MONOSEC;
414 		if (nfsd_call_servertimer != NULL)
415 			(*nfsd_call_servertimer)();
416 	}
417 	callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
418 }
419 
420 
421 /*
422  * Sleep for a short period of time unless errval == NFSERR_GRACE, where
423  * the sleep should be for 5 seconds.
424  * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
425  * an event that never gets a wakeup. Only return EINTR or 0.
426  */
427 int
428 nfs_catnap(int prio, int errval, const char *wmesg)
429 {
430 	static int non_event;
431 	int ret;
432 
433 	if (errval == NFSERR_GRACE)
434 		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
435 	else
436 		ret = tsleep(&non_event, prio, wmesg, 1);
437 	if (ret != EINTR)
438 		ret = 0;
439 	return (ret);
440 }
441 
442 /*
443  * Get referral. For now, just fail.
444  */
445 struct nfsreferral *
446 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
447 {
448 
449 	return (NULL);
450 }
451 
452 static int
453 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
454 {
455 	int error;
456 
457 	error = nfssvc_call(td, uap, td->td_ucred);
458 	NFSEXITCODE(error);
459 	return (error);
460 }
461 
462 static int
463 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
464 {
465 	int error = EINVAL, i, j;
466 	struct nfsd_idargs nid;
467 	struct nfsd_oidargs onid;
468 	struct {
469 		int vers;	/* Just the first field of nfsstats. */
470 	} nfsstatver;
471 
472 	if (uap->flag & NFSSVC_IDNAME) {
473 		if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
474 			error = copyin(uap->argp, &nid, sizeof(nid));
475 		else {
476 			error = copyin(uap->argp, &onid, sizeof(onid));
477 			if (error == 0) {
478 				nid.nid_flag = onid.nid_flag;
479 				nid.nid_uid = onid.nid_uid;
480 				nid.nid_gid = onid.nid_gid;
481 				nid.nid_usermax = onid.nid_usermax;
482 				nid.nid_usertimeout = onid.nid_usertimeout;
483 				nid.nid_name = onid.nid_name;
484 				nid.nid_namelen = onid.nid_namelen;
485 				nid.nid_ngroup = 0;
486 				nid.nid_grps = NULL;
487 			}
488 		}
489 		if (error)
490 			goto out;
491 		error = nfssvc_idname(&nid);
492 		goto out;
493 	} else if (uap->flag & NFSSVC_GETSTATS) {
494 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
495 			/* Copy fields to the old ext_nfsstat structure. */
496 			oldnfsstats.attrcache_hits =
497 			    nfsstatsv1.attrcache_hits;
498 			oldnfsstats.attrcache_misses =
499 			    nfsstatsv1.attrcache_misses;
500 			oldnfsstats.lookupcache_hits =
501 			    nfsstatsv1.lookupcache_hits;
502 			oldnfsstats.lookupcache_misses =
503 			    nfsstatsv1.lookupcache_misses;
504 			oldnfsstats.direofcache_hits =
505 			    nfsstatsv1.direofcache_hits;
506 			oldnfsstats.direofcache_misses =
507 			    nfsstatsv1.direofcache_misses;
508 			oldnfsstats.accesscache_hits =
509 			    nfsstatsv1.accesscache_hits;
510 			oldnfsstats.accesscache_misses =
511 			    nfsstatsv1.accesscache_misses;
512 			oldnfsstats.biocache_reads =
513 			    nfsstatsv1.biocache_reads;
514 			oldnfsstats.read_bios =
515 			    nfsstatsv1.read_bios;
516 			oldnfsstats.read_physios =
517 			    nfsstatsv1.read_physios;
518 			oldnfsstats.biocache_writes =
519 			    nfsstatsv1.biocache_writes;
520 			oldnfsstats.write_bios =
521 			    nfsstatsv1.write_bios;
522 			oldnfsstats.write_physios =
523 			    nfsstatsv1.write_physios;
524 			oldnfsstats.biocache_readlinks =
525 			    nfsstatsv1.biocache_readlinks;
526 			oldnfsstats.readlink_bios =
527 			    nfsstatsv1.readlink_bios;
528 			oldnfsstats.biocache_readdirs =
529 			    nfsstatsv1.biocache_readdirs;
530 			oldnfsstats.readdir_bios =
531 			    nfsstatsv1.readdir_bios;
532 			for (i = 0; i < NFSV4_NPROCS; i++)
533 				oldnfsstats.rpccnt[i] = nfsstatsv1.rpccnt[i];
534 			oldnfsstats.rpcretries = nfsstatsv1.rpcretries;
535 			for (i = 0; i < NFSV4OP_NOPS; i++)
536 				oldnfsstats.srvrpccnt[i] =
537 				    nfsstatsv1.srvrpccnt[i];
538 			for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
539 			    i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
540 				oldnfsstats.srvrpccnt[j] =
541 				    nfsstatsv1.srvrpccnt[i];
542 			oldnfsstats.srvrpc_errs = nfsstatsv1.srvrpc_errs;
543 			oldnfsstats.srv_errs = nfsstatsv1.srv_errs;
544 			oldnfsstats.rpcrequests = nfsstatsv1.rpcrequests;
545 			oldnfsstats.rpctimeouts = nfsstatsv1.rpctimeouts;
546 			oldnfsstats.rpcunexpected = nfsstatsv1.rpcunexpected;
547 			oldnfsstats.rpcinvalid = nfsstatsv1.rpcinvalid;
548 			oldnfsstats.srvcache_inproghits =
549 			    nfsstatsv1.srvcache_inproghits;
550 			oldnfsstats.srvcache_idemdonehits =
551 			    nfsstatsv1.srvcache_idemdonehits;
552 			oldnfsstats.srvcache_nonidemdonehits =
553 			    nfsstatsv1.srvcache_nonidemdonehits;
554 			oldnfsstats.srvcache_misses =
555 			    nfsstatsv1.srvcache_misses;
556 			oldnfsstats.srvcache_tcppeak =
557 			    nfsstatsv1.srvcache_tcppeak;
558 			oldnfsstats.srvcache_size = nfsstatsv1.srvcache_size;
559 			oldnfsstats.srvclients = nfsstatsv1.srvclients;
560 			oldnfsstats.srvopenowners = nfsstatsv1.srvopenowners;
561 			oldnfsstats.srvopens = nfsstatsv1.srvopens;
562 			oldnfsstats.srvlockowners = nfsstatsv1.srvlockowners;
563 			oldnfsstats.srvlocks = nfsstatsv1.srvlocks;
564 			oldnfsstats.srvdelegates = nfsstatsv1.srvdelegates;
565 			for (i = 0; i < NFSV4OP_CBNOPS; i++)
566 				oldnfsstats.cbrpccnt[i] =
567 				    nfsstatsv1.cbrpccnt[i];
568 			oldnfsstats.clopenowners = nfsstatsv1.clopenowners;
569 			oldnfsstats.clopens = nfsstatsv1.clopens;
570 			oldnfsstats.cllockowners = nfsstatsv1.cllockowners;
571 			oldnfsstats.cllocks = nfsstatsv1.cllocks;
572 			oldnfsstats.cldelegates = nfsstatsv1.cldelegates;
573 			oldnfsstats.cllocalopenowners =
574 			    nfsstatsv1.cllocalopenowners;
575 			oldnfsstats.cllocalopens = nfsstatsv1.cllocalopens;
576 			oldnfsstats.cllocallockowners =
577 			    nfsstatsv1.cllocallockowners;
578 			oldnfsstats.cllocallocks = nfsstatsv1.cllocallocks;
579 			error = copyout(&oldnfsstats, uap->argp,
580 			    sizeof (oldnfsstats));
581 		} else {
582 			error = copyin(uap->argp, &nfsstatver,
583 			    sizeof(nfsstatver));
584 			if (error == 0) {
585 				if (nfsstatver.vers == NFSSTATS_OV1) {
586 					/* Copy nfsstatsv1 to nfsstatsov1. */
587 					nfsstatsov1.attrcache_hits =
588 					    nfsstatsv1.attrcache_hits;
589 					nfsstatsov1.attrcache_misses =
590 					    nfsstatsv1.attrcache_misses;
591 					nfsstatsov1.lookupcache_hits =
592 					    nfsstatsv1.lookupcache_hits;
593 					nfsstatsov1.lookupcache_misses =
594 					    nfsstatsv1.lookupcache_misses;
595 					nfsstatsov1.direofcache_hits =
596 					    nfsstatsv1.direofcache_hits;
597 					nfsstatsov1.direofcache_misses =
598 					    nfsstatsv1.direofcache_misses;
599 					nfsstatsov1.accesscache_hits =
600 					    nfsstatsv1.accesscache_hits;
601 					nfsstatsov1.accesscache_misses =
602 					    nfsstatsv1.accesscache_misses;
603 					nfsstatsov1.biocache_reads =
604 					    nfsstatsv1.biocache_reads;
605 					nfsstatsov1.read_bios =
606 					    nfsstatsv1.read_bios;
607 					nfsstatsov1.read_physios =
608 					    nfsstatsv1.read_physios;
609 					nfsstatsov1.biocache_writes =
610 					    nfsstatsv1.biocache_writes;
611 					nfsstatsov1.write_bios =
612 					    nfsstatsv1.write_bios;
613 					nfsstatsov1.write_physios =
614 					    nfsstatsv1.write_physios;
615 					nfsstatsov1.biocache_readlinks =
616 					    nfsstatsv1.biocache_readlinks;
617 					nfsstatsov1.readlink_bios =
618 					    nfsstatsv1.readlink_bios;
619 					nfsstatsov1.biocache_readdirs =
620 					    nfsstatsv1.biocache_readdirs;
621 					nfsstatsov1.readdir_bios =
622 					    nfsstatsv1.readdir_bios;
623 					for (i = 0; i < NFSV42_NPROCS; i++)
624 						nfsstatsov1.rpccnt[i] =
625 						    nfsstatsv1.rpccnt[i];
626 					nfsstatsov1.rpcretries =
627 					    nfsstatsv1.rpcretries;
628 					for (i = 0; i < NFSV42_PURENOPS; i++)
629 						nfsstatsov1.srvrpccnt[i] =
630 						    nfsstatsv1.srvrpccnt[i];
631 					for (i = NFSV42_NOPS,
632 					     j = NFSV42_PURENOPS;
633 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
634 					     i++, j++)
635 						nfsstatsov1.srvrpccnt[j] =
636 						    nfsstatsv1.srvrpccnt[i];
637 					nfsstatsov1.srvrpc_errs =
638 					    nfsstatsv1.srvrpc_errs;
639 					nfsstatsov1.srv_errs =
640 					    nfsstatsv1.srv_errs;
641 					nfsstatsov1.rpcrequests =
642 					    nfsstatsv1.rpcrequests;
643 					nfsstatsov1.rpctimeouts =
644 					    nfsstatsv1.rpctimeouts;
645 					nfsstatsov1.rpcunexpected =
646 					    nfsstatsv1.rpcunexpected;
647 					nfsstatsov1.rpcinvalid =
648 					    nfsstatsv1.rpcinvalid;
649 					nfsstatsov1.srvcache_inproghits =
650 					    nfsstatsv1.srvcache_inproghits;
651 					nfsstatsov1.srvcache_idemdonehits =
652 					    nfsstatsv1.srvcache_idemdonehits;
653 					nfsstatsov1.srvcache_nonidemdonehits =
654 					    nfsstatsv1.srvcache_nonidemdonehits;
655 					nfsstatsov1.srvcache_misses =
656 					    nfsstatsv1.srvcache_misses;
657 					nfsstatsov1.srvcache_tcppeak =
658 					    nfsstatsv1.srvcache_tcppeak;
659 					nfsstatsov1.srvcache_size =
660 					    nfsstatsv1.srvcache_size;
661 					nfsstatsov1.srvclients =
662 					    nfsstatsv1.srvclients;
663 					nfsstatsov1.srvopenowners =
664 					    nfsstatsv1.srvopenowners;
665 					nfsstatsov1.srvopens =
666 					    nfsstatsv1.srvopens;
667 					nfsstatsov1.srvlockowners =
668 					    nfsstatsv1.srvlockowners;
669 					nfsstatsov1.srvlocks =
670 					    nfsstatsv1.srvlocks;
671 					nfsstatsov1.srvdelegates =
672 					    nfsstatsv1.srvdelegates;
673 					for (i = 0; i < NFSV42_CBNOPS; i++)
674 						nfsstatsov1.cbrpccnt[i] =
675 						    nfsstatsv1.cbrpccnt[i];
676 					nfsstatsov1.clopenowners =
677 					    nfsstatsv1.clopenowners;
678 					nfsstatsov1.clopens =
679 					    nfsstatsv1.clopens;
680 					nfsstatsov1.cllockowners =
681 					    nfsstatsv1.cllockowners;
682 					nfsstatsov1.cllocks =
683 					    nfsstatsv1.cllocks;
684 					nfsstatsov1.cldelegates =
685 					    nfsstatsv1.cldelegates;
686 					nfsstatsov1.cllocalopenowners =
687 					    nfsstatsv1.cllocalopenowners;
688 					nfsstatsov1.cllocalopens =
689 					    nfsstatsv1.cllocalopens;
690 					nfsstatsov1.cllocallockowners =
691 					    nfsstatsv1.cllocallockowners;
692 					nfsstatsov1.cllocallocks =
693 					    nfsstatsv1.cllocallocks;
694 					nfsstatsov1.srvstartcnt =
695 					    nfsstatsv1.srvstartcnt;
696 					nfsstatsov1.srvdonecnt =
697 					    nfsstatsv1.srvdonecnt;
698 					for (i = NFSV42_NOPS,
699 					     j = NFSV42_PURENOPS;
700 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
701 					     i++, j++) {
702 						nfsstatsov1.srvbytes[j] =
703 						    nfsstatsv1.srvbytes[i];
704 						nfsstatsov1.srvops[j] =
705 						    nfsstatsv1.srvops[i];
706 						nfsstatsov1.srvduration[j] =
707 						    nfsstatsv1.srvduration[i];
708 					}
709 					nfsstatsov1.busyfrom =
710 					    nfsstatsv1.busyfrom;
711 					nfsstatsov1.busyfrom =
712 					    nfsstatsv1.busyfrom;
713 					error = copyout(&nfsstatsov1, uap->argp,
714 					    sizeof(nfsstatsov1));
715 				} else if (nfsstatver.vers != NFSSTATS_V1)
716 					error = EPERM;
717 				else
718 					error = copyout(&nfsstatsv1, uap->argp,
719 					    sizeof(nfsstatsv1));
720 			}
721 		}
722 		if (error == 0) {
723 			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
724 				nfsstatsv1.attrcache_hits = 0;
725 				nfsstatsv1.attrcache_misses = 0;
726 				nfsstatsv1.lookupcache_hits = 0;
727 				nfsstatsv1.lookupcache_misses = 0;
728 				nfsstatsv1.direofcache_hits = 0;
729 				nfsstatsv1.direofcache_misses = 0;
730 				nfsstatsv1.accesscache_hits = 0;
731 				nfsstatsv1.accesscache_misses = 0;
732 				nfsstatsv1.biocache_reads = 0;
733 				nfsstatsv1.read_bios = 0;
734 				nfsstatsv1.read_physios = 0;
735 				nfsstatsv1.biocache_writes = 0;
736 				nfsstatsv1.write_bios = 0;
737 				nfsstatsv1.write_physios = 0;
738 				nfsstatsv1.biocache_readlinks = 0;
739 				nfsstatsv1.readlink_bios = 0;
740 				nfsstatsv1.biocache_readdirs = 0;
741 				nfsstatsv1.readdir_bios = 0;
742 				nfsstatsv1.rpcretries = 0;
743 				nfsstatsv1.rpcrequests = 0;
744 				nfsstatsv1.rpctimeouts = 0;
745 				nfsstatsv1.rpcunexpected = 0;
746 				nfsstatsv1.rpcinvalid = 0;
747 				bzero(nfsstatsv1.rpccnt,
748 				    sizeof(nfsstatsv1.rpccnt));
749 			}
750 			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
751 				nfsstatsv1.srvrpc_errs = 0;
752 				nfsstatsv1.srv_errs = 0;
753 				nfsstatsv1.srvcache_inproghits = 0;
754 				nfsstatsv1.srvcache_idemdonehits = 0;
755 				nfsstatsv1.srvcache_nonidemdonehits = 0;
756 				nfsstatsv1.srvcache_misses = 0;
757 				nfsstatsv1.srvcache_tcppeak = 0;
758 				bzero(nfsstatsv1.srvrpccnt,
759 				    sizeof(nfsstatsv1.srvrpccnt));
760 				bzero(nfsstatsv1.cbrpccnt,
761 				    sizeof(nfsstatsv1.cbrpccnt));
762 			}
763 		}
764 		goto out;
765 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
766 		u_short sockport;
767 		struct nfsuserd_args nargs;
768 
769 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
770 			error = copyin(uap->argp, (caddr_t)&sockport,
771 			    sizeof (u_short));
772 			if (error == 0) {
773 				nargs.nuserd_family = AF_INET;
774 				nargs.nuserd_port = sockport;
775 			}
776 		} else {
777 			/*
778 			 * New nfsuserd_args structure, which indicates
779 			 * which IP version to use along with the port#.
780 			 */
781 			error = copyin(uap->argp, &nargs, sizeof(nargs));
782 		}
783 		if (!error)
784 			error = nfsrv_nfsuserdport(&nargs, p);
785 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
786 		nfsrv_nfsuserddelport();
787 		error = 0;
788 	}
789 
790 out:
791 	NFSEXITCODE(error);
792 	return (error);
793 }
794 
795 /*
796  * called by all three modevent routines, so that it gets things
797  * initialized soon enough.
798  */
799 void
800 newnfs_portinit(void)
801 {
802 	static int inited = 0;
803 
804 	if (inited)
805 		return;
806 	inited = 1;
807 	/* Initialize SMP locks used by both client and server. */
808 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
809 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
810 	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
811 }
812 
813 /*
814  * Determine if the file system supports NFSv4 ACLs.
815  * Return 1 if it does, 0 otherwise.
816  */
817 int
818 nfs_supportsnfsv4acls(struct vnode *vp)
819 {
820 	int error;
821 	long retval;
822 
823 	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
824 
825 	if (nfsrv_useacl == 0)
826 		return (0);
827 	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
828 	if (error == 0 && retval != 0)
829 		return (1);
830 	return (0);
831 }
832 
833 /*
834  * These are the first fields of all the context structures passed into
835  * nfs_pnfsio().
836  */
837 struct pnfsio {
838 	int		done;
839 	int		inprog;
840 	struct task	tsk;
841 };
842 
843 /*
844  * Do a mirror I/O on a pNFS thread.
845  */
846 int
847 nfs_pnfsio(task_fn_t *func, void *context)
848 {
849 	struct pnfsio *pio;
850 	int ret;
851 	static struct taskqueue *pnfsioq = NULL;
852 
853 	pio = (struct pnfsio *)context;
854 	if (pnfsioq == NULL) {
855 		if (nfs_pnfsiothreads == 0)
856 			return (EPERM);
857 		if (nfs_pnfsiothreads < 0)
858 			nfs_pnfsiothreads = mp_ncpus * 4;
859 		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
860 		    taskqueue_thread_enqueue, &pnfsioq);
861 		if (pnfsioq == NULL)
862 			return (ENOMEM);
863 		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
864 		    0, "pnfsiot");
865 		if (ret != 0) {
866 			taskqueue_free(pnfsioq);
867 			pnfsioq = NULL;
868 			return (ret);
869 		}
870 	}
871 	pio->inprog = 1;
872 	TASK_INIT(&pio->tsk, 0, func, context);
873 	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
874 	if (ret != 0)
875 		pio->inprog = 0;
876 	return (ret);
877 }
878 
879 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
880 
881 /*
882  * Called once to initialize data structures...
883  */
884 static int
885 nfscommon_modevent(module_t mod, int type, void *data)
886 {
887 	int error = 0;
888 	static int loaded = 0;
889 
890 	switch (type) {
891 	case MOD_LOAD:
892 		if (loaded)
893 			goto out;
894 		newnfs_portinit();
895 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
896 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
897 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
898 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
899 		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
900 		    MTX_DEF);
901 		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
902 		TAILQ_INIT(&nfsrv_devidhead);
903 		callout_init(&newnfsd_callout, 1);
904 		newnfs_init();
905 		nfsd_call_nfscommon = nfssvc_nfscommon;
906 		loaded = 1;
907 		break;
908 
909 	case MOD_UNLOAD:
910 		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != NOTRUNNING ||
911 		    nfs_numnfscbd != 0) {
912 			error = EBUSY;
913 			break;
914 		}
915 
916 		nfsd_call_nfscommon = NULL;
917 		callout_drain(&newnfsd_callout);
918 		/* Clean out the name<-->id cache. */
919 		nfsrv_cleanusergroup();
920 		/* and get rid of the mutexes */
921 		mtx_destroy(&nfs_nameid_mutex);
922 		mtx_destroy(&newnfsd_mtx);
923 		mtx_destroy(&nfs_state_mutex);
924 		mtx_destroy(&nfs_clstate_mutex);
925 		mtx_destroy(&nfs_sockl_mutex);
926 		mtx_destroy(&nfs_slock_mutex);
927 		mtx_destroy(&nfs_req_mutex);
928 		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
929 		mtx_destroy(&nfsrv_dslock_mtx);
930 		loaded = 0;
931 		break;
932 	default:
933 		error = EOPNOTSUPP;
934 		break;
935 	}
936 
937 out:
938 	NFSEXITCODE(error);
939 	return error;
940 }
941 static moduledata_t nfscommon_mod = {
942 	"nfscommon",
943 	nfscommon_modevent,
944 	NULL,
945 };
946 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
947 
948 /* So that loader and kldload(2) can find us, wherever we are.. */
949 MODULE_VERSION(nfscommon, 1);
950 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
951 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
952 
953