xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision 105fd928b0b5b35ab529e5f6914788dc49582901)
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 vop_advlock_t *nfs_advlock_p = NULL;
78 vop_reclaim_t *nfs_reclaim_p = NULL;
79 uint32_t nfs_srvmaxio = NFS_SRVMAXIO;
80 
81 int nfs_pnfsio(task_fn_t *, void *);
82 
83 static int nfs_realign_test;
84 static int nfs_realign_count;
85 static struct ext_nfsstats oldnfsstats;
86 static struct nfsstatsov1 nfsstatsov1;
87 
88 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
89     "NFS filesystem");
90 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
91     0, "Number of realign tests done");
92 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
93     0, "Number of mbuf realignments done");
94 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
95     nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
96     "NFSv4 callback addr for server to use");
97 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
98     0, "Debug level for NFS client");
99 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
100     0, "Size of hash tables for uid/name mapping");
101 int nfs_pnfsiothreads = -1;
102 SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsiothreads, CTLFLAG_RW, &nfs_pnfsiothreads,
103     0, "Number of pNFS mirror I/O threads");
104 
105 /*
106  * Defines for malloc
107  * (Here for FreeBSD, since they allocate storage.)
108  */
109 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
110 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
111 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
112     "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
113 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
114 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
115 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
116 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
117 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
118 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
119 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
120 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
121 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
122 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
123 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
124 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
125 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
126 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer");
127 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroff",
128     "NFS directory offset data");
129 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
130     "NFS local lock rollback");
131 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
132 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
133 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
134 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
135 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
136 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
137 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
138 
139 /*
140  * Definition of mutex locks.
141  * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
142  * and assorted other nfsd structures.
143  */
144 struct mtx newnfsd_mtx;
145 struct mtx nfs_sockl_mutex;
146 struct mtx nfs_state_mutex;
147 struct mtx nfs_nameid_mutex;
148 struct mtx nfs_req_mutex;
149 struct mtx nfs_slock_mutex;
150 struct mtx nfs_clstate_mutex;
151 
152 /* local functions */
153 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
154 
155 #ifdef __NO_STRICT_ALIGNMENT
156 /*
157  * These architectures don't need re-alignment, so just return.
158  */
159 int
160 newnfs_realign(struct mbuf **pm, int how)
161 {
162 
163 	return (0);
164 }
165 #else	/* !__NO_STRICT_ALIGNMENT */
166 /*
167  *	newnfs_realign:
168  *
169  *	Check for badly aligned mbuf data and realign by copying the unaligned
170  *	portion of the data into a new mbuf chain and freeing the portions
171  *	of the old chain that were replaced.
172  *
173  *	We cannot simply realign the data within the existing mbuf chain
174  *	because the underlying buffers may contain other rpc commands and
175  *	we cannot afford to overwrite them.
176  *
177  *	We would prefer to avoid this situation entirely.  The situation does
178  *	not occur with NFS/UDP and is supposed to only occasionally occur
179  *	with TCP.  Use vfs.nfs.realign_count and realign_test to check this.
180  *
181  */
182 int
183 newnfs_realign(struct mbuf **pm, int how)
184 {
185 	struct mbuf *m, *n;
186 	int off, space;
187 
188 	++nfs_realign_test;
189 	while ((m = *pm) != NULL) {
190 		if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
191 			/*
192 			 * NB: we can't depend on m_pkthdr.len to help us
193 			 * decide what to do here.  May not be worth doing
194 			 * the m_length calculation as m_copyback will
195 			 * expand the mbuf chain below as needed.
196 			 */
197 			space = m_length(m, NULL);
198 			if (space >= MINCLSIZE) {
199 				/* NB: m_copyback handles space > MCLBYTES */
200 				n = m_getcl(how, MT_DATA, 0);
201 			} else
202 				n = m_get(how, MT_DATA);
203 			if (n == NULL)
204 				return (ENOMEM);
205 			/*
206 			 * Align the remainder of the mbuf chain.
207 			 */
208 			n->m_len = 0;
209 			off = 0;
210 			while (m != NULL) {
211 				m_copyback(n, off, m->m_len, mtod(m, caddr_t));
212 				off += m->m_len;
213 				m = m->m_next;
214 			}
215 			m_freem(*pm);
216 			*pm = n;
217 			++nfs_realign_count;
218 			break;
219 		}
220 		pm = &m->m_next;
221 	}
222 
223 	return (0);
224 }
225 #endif	/* __NO_STRICT_ALIGNMENT */
226 
227 #ifdef notdef
228 static void
229 nfsrv_object_create(struct vnode *vp, struct thread *td)
230 {
231 
232 	if (vp == NULL || vp->v_type != VREG)
233 		return;
234 	(void) vfs_object_create(vp, td, td->td_ucred);
235 }
236 #endif
237 
238 /*
239  * Look up a file name. Basically just initialize stuff and call namei().
240  */
241 int
242 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
243 {
244 	int error;
245 
246 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
247 	    p);
248 	error = namei(ndp);
249 	if (!error) {
250 		NDFREE(ndp, NDF_ONLY_PNBUF);
251 	}
252 	return (error);
253 }
254 
255 /*
256  * Copy NFS uid, gids to the cred structure.
257  */
258 void
259 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
260 {
261 
262 	KASSERT(nfscr->nfsc_ngroups >= 0,
263 	    ("newnfs_copycred: negative nfsc_ngroups"));
264 	cr->cr_uid = nfscr->nfsc_uid;
265 	crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
266 }
267 
268 /*
269  * Map args from nfsmsleep() to msleep().
270  */
271 int
272 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
273     struct timespec *ts)
274 {
275 	u_int64_t nsecval;
276 	int error, timeo;
277 
278 	if (ts) {
279 		timeo = hz * ts->tv_sec;
280 		nsecval = (u_int64_t)ts->tv_nsec;
281 		nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
282 		    1000000000;
283 		timeo += (int)nsecval;
284 	} else {
285 		timeo = 0;
286 	}
287 	error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
288 	return (error);
289 }
290 
291 /*
292  * Get the file system info for the server. For now, just assume FFS.
293  */
294 void
295 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
296 {
297 	int pref;
298 
299 	/*
300 	 * XXX
301 	 * There should be file system VFS OP(s) to get this information.
302 	 * For now, assume ufs.
303 	 */
304 	if (isdgram)
305 		pref = NFS_MAXDGRAMDATA;
306 	else
307 		pref = nfs_srvmaxio;
308 	sip->fs_rtmax = nfs_srvmaxio;
309 	sip->fs_rtpref = pref;
310 	sip->fs_rtmult = NFS_FABLKSIZE;
311 	sip->fs_wtmax = nfs_srvmaxio;
312 	sip->fs_wtpref = pref;
313 	sip->fs_wtmult = NFS_FABLKSIZE;
314 	sip->fs_dtpref = pref;
315 	sip->fs_maxfilesize = 0xffffffffffffffffull;
316 	sip->fs_timedelta.tv_sec = 0;
317 	sip->fs_timedelta.tv_nsec = 1;
318 	sip->fs_properties = (NFSV3FSINFO_LINK |
319 	    NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
320 	    NFSV3FSINFO_CANSETTIME);
321 }
322 
323 /*
324  * Do the pathconf vnode op.
325  */
326 int
327 nfsvno_pathconf(struct vnode *vp, int flag, long *retf,
328     struct ucred *cred, struct thread *p)
329 {
330 	int error;
331 
332 	error = VOP_PATHCONF(vp, flag, retf);
333 	if (error == EOPNOTSUPP || error == EINVAL) {
334 		/*
335 		 * Some file systems return EINVAL for name arguments not
336 		 * supported and some return EOPNOTSUPP for this case.
337 		 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
338 		 * just fake them.
339 		 */
340 		switch (flag) {
341 		case _PC_LINK_MAX:
342 			*retf = NFS_LINK_MAX;
343 			break;
344 		case _PC_NAME_MAX:
345 			*retf = NAME_MAX;
346 			break;
347 		case _PC_CHOWN_RESTRICTED:
348 			*retf = 1;
349 			break;
350 		case _PC_NO_TRUNC:
351 			*retf = 1;
352 			break;
353 		default:
354 			/*
355 			 * Only happens if a _PC_xxx is added to the server,
356 			 * but this isn't updated.
357 			 */
358 			*retf = 0;
359 			printf("nfsrvd pathconf flag=%d not supp\n", flag);
360 		}
361 		error = 0;
362 	}
363 	NFSEXITCODE(error);
364 	return (error);
365 }
366 
367 /* Fake nfsrv_atroot. Just return 0 */
368 int
369 nfsrv_atroot(struct vnode *vp, uint64_t *retp)
370 {
371 
372 	return (0);
373 }
374 
375 /*
376  * Set the credentials to refer to root.
377  * If only the various BSDen could agree on whether cr_gid is a separate
378  * field or cr_groups[0]...
379  */
380 void
381 newnfs_setroot(struct ucred *cred)
382 {
383 
384 	cred->cr_uid = 0;
385 	cred->cr_groups[0] = 0;
386 	cred->cr_ngroups = 1;
387 }
388 
389 /*
390  * Get the client credential. Used for Renew and recovery.
391  */
392 struct ucred *
393 newnfs_getcred(void)
394 {
395 	struct ucred *cred;
396 	struct thread *td = curthread;
397 
398 	cred = crdup(td->td_ucred);
399 	newnfs_setroot(cred);
400 	return (cred);
401 }
402 
403 /*
404  * Nfs timer routine
405  * Call the nfsd's timer function once/sec.
406  */
407 void
408 newnfs_timer(void *arg)
409 {
410 	static time_t lasttime = 0;
411 	/*
412 	 * Call the server timer, if set up.
413 	 * The argument indicates if it is the next second and therefore
414 	 * leases should be checked.
415 	 */
416 	if (lasttime != NFSD_MONOSEC) {
417 		lasttime = NFSD_MONOSEC;
418 		if (nfsd_call_servertimer != NULL)
419 			(*nfsd_call_servertimer)();
420 	}
421 	callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
422 }
423 
424 /*
425  * Sleep for a short period of time unless errval == NFSERR_GRACE, where
426  * the sleep should be for 5 seconds.
427  * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
428  * an event that never gets a wakeup. Only return EINTR or 0.
429  */
430 int
431 nfs_catnap(int prio, int errval, const char *wmesg)
432 {
433 	static int non_event;
434 	int ret;
435 
436 	if (errval == NFSERR_GRACE)
437 		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
438 	else
439 		ret = tsleep(&non_event, prio, wmesg, 1);
440 	if (ret != EINTR)
441 		ret = 0;
442 	return (ret);
443 }
444 
445 /*
446  * Get referral. For now, just fail.
447  */
448 struct nfsreferral *
449 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
450 {
451 
452 	return (NULL);
453 }
454 
455 static int
456 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
457 {
458 	int error;
459 
460 	error = nfssvc_call(td, uap, td->td_ucred);
461 	NFSEXITCODE(error);
462 	return (error);
463 }
464 
465 static int
466 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
467 {
468 	int error = EINVAL, i, j;
469 	struct nfsd_idargs nid;
470 	struct nfsd_oidargs onid;
471 	struct {
472 		int vers;	/* Just the first field of nfsstats. */
473 	} nfsstatver;
474 
475 	if (uap->flag & NFSSVC_IDNAME) {
476 		if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
477 			error = copyin(uap->argp, &nid, sizeof(nid));
478 		else {
479 			error = copyin(uap->argp, &onid, sizeof(onid));
480 			if (error == 0) {
481 				nid.nid_flag = onid.nid_flag;
482 				nid.nid_uid = onid.nid_uid;
483 				nid.nid_gid = onid.nid_gid;
484 				nid.nid_usermax = onid.nid_usermax;
485 				nid.nid_usertimeout = onid.nid_usertimeout;
486 				nid.nid_name = onid.nid_name;
487 				nid.nid_namelen = onid.nid_namelen;
488 				nid.nid_ngroup = 0;
489 				nid.nid_grps = NULL;
490 			}
491 		}
492 		if (error)
493 			goto out;
494 		error = nfssvc_idname(&nid);
495 		goto out;
496 	} else if (uap->flag & NFSSVC_GETSTATS) {
497 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
498 			/* Copy fields to the old ext_nfsstat structure. */
499 			oldnfsstats.attrcache_hits =
500 			    nfsstatsv1.attrcache_hits;
501 			oldnfsstats.attrcache_misses =
502 			    nfsstatsv1.attrcache_misses;
503 			oldnfsstats.lookupcache_hits =
504 			    nfsstatsv1.lookupcache_hits;
505 			oldnfsstats.lookupcache_misses =
506 			    nfsstatsv1.lookupcache_misses;
507 			oldnfsstats.direofcache_hits =
508 			    nfsstatsv1.direofcache_hits;
509 			oldnfsstats.direofcache_misses =
510 			    nfsstatsv1.direofcache_misses;
511 			oldnfsstats.accesscache_hits =
512 			    nfsstatsv1.accesscache_hits;
513 			oldnfsstats.accesscache_misses =
514 			    nfsstatsv1.accesscache_misses;
515 			oldnfsstats.biocache_reads =
516 			    nfsstatsv1.biocache_reads;
517 			oldnfsstats.read_bios =
518 			    nfsstatsv1.read_bios;
519 			oldnfsstats.read_physios =
520 			    nfsstatsv1.read_physios;
521 			oldnfsstats.biocache_writes =
522 			    nfsstatsv1.biocache_writes;
523 			oldnfsstats.write_bios =
524 			    nfsstatsv1.write_bios;
525 			oldnfsstats.write_physios =
526 			    nfsstatsv1.write_physios;
527 			oldnfsstats.biocache_readlinks =
528 			    nfsstatsv1.biocache_readlinks;
529 			oldnfsstats.readlink_bios =
530 			    nfsstatsv1.readlink_bios;
531 			oldnfsstats.biocache_readdirs =
532 			    nfsstatsv1.biocache_readdirs;
533 			oldnfsstats.readdir_bios =
534 			    nfsstatsv1.readdir_bios;
535 			for (i = 0; i < NFSV4_NPROCS; i++)
536 				oldnfsstats.rpccnt[i] = nfsstatsv1.rpccnt[i];
537 			oldnfsstats.rpcretries = nfsstatsv1.rpcretries;
538 			for (i = 0; i < NFSV4OP_NOPS; i++)
539 				oldnfsstats.srvrpccnt[i] =
540 				    nfsstatsv1.srvrpccnt[i];
541 			for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
542 			    i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
543 				oldnfsstats.srvrpccnt[j] =
544 				    nfsstatsv1.srvrpccnt[i];
545 			oldnfsstats.reserved_0 = 0;
546 			oldnfsstats.reserved_1 = 0;
547 			oldnfsstats.rpcrequests = nfsstatsv1.rpcrequests;
548 			oldnfsstats.rpctimeouts = nfsstatsv1.rpctimeouts;
549 			oldnfsstats.rpcunexpected = nfsstatsv1.rpcunexpected;
550 			oldnfsstats.rpcinvalid = nfsstatsv1.rpcinvalid;
551 			oldnfsstats.srvcache_inproghits =
552 			    nfsstatsv1.srvcache_inproghits;
553 			oldnfsstats.reserved_2 = 0;
554 			oldnfsstats.srvcache_nonidemdonehits =
555 			    nfsstatsv1.srvcache_nonidemdonehits;
556 			oldnfsstats.srvcache_misses =
557 			    nfsstatsv1.srvcache_misses;
558 			oldnfsstats.srvcache_tcppeak =
559 			    nfsstatsv1.srvcache_tcppeak;
560 			oldnfsstats.srvcache_size = nfsstatsv1.srvcache_size;
561 			oldnfsstats.srvclients = nfsstatsv1.srvclients;
562 			oldnfsstats.srvopenowners = nfsstatsv1.srvopenowners;
563 			oldnfsstats.srvopens = nfsstatsv1.srvopens;
564 			oldnfsstats.srvlockowners = nfsstatsv1.srvlockowners;
565 			oldnfsstats.srvlocks = nfsstatsv1.srvlocks;
566 			oldnfsstats.srvdelegates = nfsstatsv1.srvdelegates;
567 			for (i = 0; i < NFSV4OP_CBNOPS; i++)
568 				oldnfsstats.cbrpccnt[i] =
569 				    nfsstatsv1.cbrpccnt[i];
570 			oldnfsstats.clopenowners = nfsstatsv1.clopenowners;
571 			oldnfsstats.clopens = nfsstatsv1.clopens;
572 			oldnfsstats.cllockowners = nfsstatsv1.cllockowners;
573 			oldnfsstats.cllocks = nfsstatsv1.cllocks;
574 			oldnfsstats.cldelegates = nfsstatsv1.cldelegates;
575 			oldnfsstats.cllocalopenowners =
576 			    nfsstatsv1.cllocalopenowners;
577 			oldnfsstats.cllocalopens = nfsstatsv1.cllocalopens;
578 			oldnfsstats.cllocallockowners =
579 			    nfsstatsv1.cllocallockowners;
580 			oldnfsstats.cllocallocks = nfsstatsv1.cllocallocks;
581 			error = copyout(&oldnfsstats, uap->argp,
582 			    sizeof (oldnfsstats));
583 		} else {
584 			error = copyin(uap->argp, &nfsstatver,
585 			    sizeof(nfsstatver));
586 			if (error == 0) {
587 				if (nfsstatver.vers == NFSSTATS_OV1) {
588 					/* Copy nfsstatsv1 to nfsstatsov1. */
589 					nfsstatsov1.attrcache_hits =
590 					    nfsstatsv1.attrcache_hits;
591 					nfsstatsov1.attrcache_misses =
592 					    nfsstatsv1.attrcache_misses;
593 					nfsstatsov1.lookupcache_hits =
594 					    nfsstatsv1.lookupcache_hits;
595 					nfsstatsov1.lookupcache_misses =
596 					    nfsstatsv1.lookupcache_misses;
597 					nfsstatsov1.direofcache_hits =
598 					    nfsstatsv1.direofcache_hits;
599 					nfsstatsov1.direofcache_misses =
600 					    nfsstatsv1.direofcache_misses;
601 					nfsstatsov1.accesscache_hits =
602 					    nfsstatsv1.accesscache_hits;
603 					nfsstatsov1.accesscache_misses =
604 					    nfsstatsv1.accesscache_misses;
605 					nfsstatsov1.biocache_reads =
606 					    nfsstatsv1.biocache_reads;
607 					nfsstatsov1.read_bios =
608 					    nfsstatsv1.read_bios;
609 					nfsstatsov1.read_physios =
610 					    nfsstatsv1.read_physios;
611 					nfsstatsov1.biocache_writes =
612 					    nfsstatsv1.biocache_writes;
613 					nfsstatsov1.write_bios =
614 					    nfsstatsv1.write_bios;
615 					nfsstatsov1.write_physios =
616 					    nfsstatsv1.write_physios;
617 					nfsstatsov1.biocache_readlinks =
618 					    nfsstatsv1.biocache_readlinks;
619 					nfsstatsov1.readlink_bios =
620 					    nfsstatsv1.readlink_bios;
621 					nfsstatsov1.biocache_readdirs =
622 					    nfsstatsv1.biocache_readdirs;
623 					nfsstatsov1.readdir_bios =
624 					    nfsstatsv1.readdir_bios;
625 					for (i = 0; i < NFSV42_NPROCS; i++)
626 						nfsstatsov1.rpccnt[i] =
627 						    nfsstatsv1.rpccnt[i];
628 					nfsstatsov1.rpcretries =
629 					    nfsstatsv1.rpcretries;
630 					for (i = 0; i < NFSV42_PURENOPS; i++)
631 						nfsstatsov1.srvrpccnt[i] =
632 						    nfsstatsv1.srvrpccnt[i];
633 					for (i = NFSV42_NOPS,
634 					     j = NFSV42_PURENOPS;
635 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
636 					     i++, j++)
637 						nfsstatsov1.srvrpccnt[j] =
638 						    nfsstatsv1.srvrpccnt[i];
639 					nfsstatsov1.reserved_0 = 0;
640 					nfsstatsov1.reserved_1 = 0;
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.reserved_2 = 0;
652 					nfsstatsov1.srvcache_nonidemdonehits =
653 					    nfsstatsv1.srvcache_nonidemdonehits;
654 					nfsstatsov1.srvcache_misses =
655 					    nfsstatsv1.srvcache_misses;
656 					nfsstatsov1.srvcache_tcppeak =
657 					    nfsstatsv1.srvcache_tcppeak;
658 					nfsstatsov1.srvcache_size =
659 					    nfsstatsv1.srvcache_size;
660 					nfsstatsov1.srvclients =
661 					    nfsstatsv1.srvclients;
662 					nfsstatsov1.srvopenowners =
663 					    nfsstatsv1.srvopenowners;
664 					nfsstatsov1.srvopens =
665 					    nfsstatsv1.srvopens;
666 					nfsstatsov1.srvlockowners =
667 					    nfsstatsv1.srvlockowners;
668 					nfsstatsov1.srvlocks =
669 					    nfsstatsv1.srvlocks;
670 					nfsstatsov1.srvdelegates =
671 					    nfsstatsv1.srvdelegates;
672 					for (i = 0; i < NFSV42_CBNOPS; i++)
673 						nfsstatsov1.cbrpccnt[i] =
674 						    nfsstatsv1.cbrpccnt[i];
675 					nfsstatsov1.clopenowners =
676 					    nfsstatsv1.clopenowners;
677 					nfsstatsov1.clopens =
678 					    nfsstatsv1.clopens;
679 					nfsstatsov1.cllockowners =
680 					    nfsstatsv1.cllockowners;
681 					nfsstatsov1.cllocks =
682 					    nfsstatsv1.cllocks;
683 					nfsstatsov1.cldelegates =
684 					    nfsstatsv1.cldelegates;
685 					nfsstatsov1.cllocalopenowners =
686 					    nfsstatsv1.cllocalopenowners;
687 					nfsstatsov1.cllocalopens =
688 					    nfsstatsv1.cllocalopens;
689 					nfsstatsov1.cllocallockowners =
690 					    nfsstatsv1.cllocallockowners;
691 					nfsstatsov1.cllocallocks =
692 					    nfsstatsv1.cllocallocks;
693 					nfsstatsov1.srvstartcnt =
694 					    nfsstatsv1.srvstartcnt;
695 					nfsstatsov1.srvdonecnt =
696 					    nfsstatsv1.srvdonecnt;
697 					for (i = NFSV42_NOPS,
698 					     j = NFSV42_PURENOPS;
699 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
700 					     i++, j++) {
701 						nfsstatsov1.srvbytes[j] =
702 						    nfsstatsv1.srvbytes[i];
703 						nfsstatsov1.srvops[j] =
704 						    nfsstatsv1.srvops[i];
705 						nfsstatsov1.srvduration[j] =
706 						    nfsstatsv1.srvduration[i];
707 					}
708 					nfsstatsov1.busyfrom =
709 					    nfsstatsv1.busyfrom;
710 					nfsstatsov1.busyfrom =
711 					    nfsstatsv1.busyfrom;
712 					error = copyout(&nfsstatsov1, uap->argp,
713 					    sizeof(nfsstatsov1));
714 				} else if (nfsstatver.vers != NFSSTATS_V1)
715 					error = EPERM;
716 				else
717 					error = copyout(&nfsstatsv1, uap->argp,
718 					    sizeof(nfsstatsv1));
719 			}
720 		}
721 		if (error == 0) {
722 			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
723 				nfsstatsv1.attrcache_hits = 0;
724 				nfsstatsv1.attrcache_misses = 0;
725 				nfsstatsv1.lookupcache_hits = 0;
726 				nfsstatsv1.lookupcache_misses = 0;
727 				nfsstatsv1.direofcache_hits = 0;
728 				nfsstatsv1.direofcache_misses = 0;
729 				nfsstatsv1.accesscache_hits = 0;
730 				nfsstatsv1.accesscache_misses = 0;
731 				nfsstatsv1.biocache_reads = 0;
732 				nfsstatsv1.read_bios = 0;
733 				nfsstatsv1.read_physios = 0;
734 				nfsstatsv1.biocache_writes = 0;
735 				nfsstatsv1.write_bios = 0;
736 				nfsstatsv1.write_physios = 0;
737 				nfsstatsv1.biocache_readlinks = 0;
738 				nfsstatsv1.readlink_bios = 0;
739 				nfsstatsv1.biocache_readdirs = 0;
740 				nfsstatsv1.readdir_bios = 0;
741 				nfsstatsv1.rpcretries = 0;
742 				nfsstatsv1.rpcrequests = 0;
743 				nfsstatsv1.rpctimeouts = 0;
744 				nfsstatsv1.rpcunexpected = 0;
745 				nfsstatsv1.rpcinvalid = 0;
746 				bzero(nfsstatsv1.rpccnt,
747 				    sizeof(nfsstatsv1.rpccnt));
748 			}
749 			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
750 				nfsstatsv1.srvcache_inproghits = 0;
751 				nfsstatsv1.srvcache_nonidemdonehits = 0;
752 				nfsstatsv1.srvcache_misses = 0;
753 				nfsstatsv1.srvcache_tcppeak = 0;
754 				bzero(nfsstatsv1.srvrpccnt,
755 				    sizeof(nfsstatsv1.srvrpccnt));
756 				bzero(nfsstatsv1.cbrpccnt,
757 				    sizeof(nfsstatsv1.cbrpccnt));
758 			}
759 		}
760 		goto out;
761 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
762 		u_short sockport;
763 		struct nfsuserd_args nargs;
764 
765 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
766 			error = copyin(uap->argp, (caddr_t)&sockport,
767 			    sizeof (u_short));
768 			if (error == 0) {
769 				nargs.nuserd_family = AF_INET;
770 				nargs.nuserd_port = sockport;
771 			}
772 		} else {
773 			/*
774 			 * New nfsuserd_args structure, which indicates
775 			 * which IP version to use along with the port#.
776 			 */
777 			error = copyin(uap->argp, &nargs, sizeof(nargs));
778 		}
779 		if (!error)
780 			error = nfsrv_nfsuserdport(&nargs, p);
781 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
782 		nfsrv_nfsuserddelport();
783 		error = 0;
784 	}
785 
786 out:
787 	NFSEXITCODE(error);
788 	return (error);
789 }
790 
791 /*
792  * called by all three modevent routines, so that it gets things
793  * initialized soon enough.
794  */
795 void
796 newnfs_portinit(void)
797 {
798 	static int inited = 0;
799 
800 	if (inited)
801 		return;
802 	inited = 1;
803 	/* Initialize SMP locks used by both client and server. */
804 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
805 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
806 	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
807 }
808 
809 /*
810  * Determine if the file system supports NFSv4 ACLs.
811  * Return 1 if it does, 0 otherwise.
812  */
813 int
814 nfs_supportsnfsv4acls(struct vnode *vp)
815 {
816 	int error;
817 	long retval;
818 
819 	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
820 
821 	if (nfsrv_useacl == 0)
822 		return (0);
823 	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
824 	if (error == 0 && retval != 0)
825 		return (1);
826 	return (0);
827 }
828 
829 /*
830  * These are the first fields of all the context structures passed into
831  * nfs_pnfsio().
832  */
833 struct pnfsio {
834 	int		done;
835 	int		inprog;
836 	struct task	tsk;
837 };
838 
839 /*
840  * Do a mirror I/O on a pNFS thread.
841  */
842 int
843 nfs_pnfsio(task_fn_t *func, void *context)
844 {
845 	struct pnfsio *pio;
846 	int ret;
847 	static struct taskqueue *pnfsioq = NULL;
848 
849 	pio = (struct pnfsio *)context;
850 	if (pnfsioq == NULL) {
851 		if (nfs_pnfsiothreads == 0)
852 			return (EPERM);
853 		if (nfs_pnfsiothreads < 0)
854 			nfs_pnfsiothreads = mp_ncpus * 4;
855 		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
856 		    taskqueue_thread_enqueue, &pnfsioq);
857 		if (pnfsioq == NULL)
858 			return (ENOMEM);
859 		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
860 		    0, "pnfsiot");
861 		if (ret != 0) {
862 			taskqueue_free(pnfsioq);
863 			pnfsioq = NULL;
864 			return (ret);
865 		}
866 	}
867 	pio->inprog = 1;
868 	TASK_INIT(&pio->tsk, 0, func, context);
869 	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
870 	if (ret != 0)
871 		pio->inprog = 0;
872 	return (ret);
873 }
874 
875 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
876 
877 /*
878  * Called once to initialize data structures...
879  */
880 static int
881 nfscommon_modevent(module_t mod, int type, void *data)
882 {
883 	int error = 0;
884 	static int loaded = 0;
885 
886 	switch (type) {
887 	case MOD_LOAD:
888 		if (loaded)
889 			goto out;
890 		newnfs_portinit();
891 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
892 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
893 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
894 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
895 		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
896 		    MTX_DEF);
897 		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
898 		TAILQ_INIT(&nfsrv_devidhead);
899 		callout_init(&newnfsd_callout, 1);
900 		newnfs_init();
901 		nfsd_call_nfscommon = nfssvc_nfscommon;
902 		loaded = 1;
903 		break;
904 
905 	case MOD_UNLOAD:
906 		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != NOTRUNNING ||
907 		    nfs_numnfscbd != 0) {
908 			error = EBUSY;
909 			break;
910 		}
911 
912 		nfsd_call_nfscommon = NULL;
913 		callout_drain(&newnfsd_callout);
914 		/* Clean out the name<-->id cache. */
915 		nfsrv_cleanusergroup();
916 		/* and get rid of the mutexes */
917 		mtx_destroy(&nfs_nameid_mutex);
918 		mtx_destroy(&newnfsd_mtx);
919 		mtx_destroy(&nfs_state_mutex);
920 		mtx_destroy(&nfs_clstate_mutex);
921 		mtx_destroy(&nfs_sockl_mutex);
922 		mtx_destroy(&nfs_slock_mutex);
923 		mtx_destroy(&nfs_req_mutex);
924 		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
925 		mtx_destroy(&nfsrv_dslock_mtx);
926 		loaded = 0;
927 		break;
928 	default:
929 		error = EOPNOTSUPP;
930 		break;
931 	}
932 
933 out:
934 	NFSEXITCODE(error);
935 	return error;
936 }
937 static moduledata_t nfscommon_mod = {
938 	"nfscommon",
939 	nfscommon_modevent,
940 	NULL,
941 };
942 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
943 
944 /* So that loader and kldload(2) can find us, wherever we are.. */
945 MODULE_VERSION(nfscommon, 1);
946 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
947 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
948