xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision 8cc087a1eee9ec1ca9f7ac1e63ad51bdb5a682eb)
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 int nfsrv_lughashsize = 100;
71 struct mtx nfsrv_dslock_mtx;
72 struct nfsdevicehead nfsrv_devidhead;
73 volatile int nfsrv_devidcnt = 0;
74 void (*ncl_call_invalcaches)(struct vnode *) = NULL;
75 vop_advlock_t *nfs_advlock_p = NULL;
76 vop_reclaim_t *nfs_reclaim_p = NULL;
77 uint32_t nfs_srvmaxio = NFS_SRVMAXIO;
78 
79 int nfs_pnfsio(task_fn_t *, void *);
80 
81 static int nfs_realign_test;
82 static int nfs_realign_count;
83 static struct ext_nfsstats oldnfsstats;
84 static struct nfsstatsov1 nfsstatsov1;
85 
86 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
87     "NFS filesystem");
88 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
89     0, "Number of realign tests done");
90 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
91     0, "Number of mbuf realignments done");
92 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
93     nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
94     "NFSv4 callback addr for server to use");
95 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
96     0, "Debug level for NFS client");
97 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
98     0, "Size of hash tables for uid/name mapping");
99 int nfs_pnfsiothreads = -1;
100 SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsiothreads, CTLFLAG_RW, &nfs_pnfsiothreads,
101     0, "Number of pNFS mirror I/O threads");
102 
103 /*
104  * Defines for malloc
105  * (Here for FreeBSD, since they allocate storage.)
106  */
107 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
108 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
109 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
110     "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
111 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
112 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
113 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
114 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
115 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
116 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
117 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
118 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
119 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
120 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
121 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
122 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
123 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
124 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer");
125 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroff",
126     "NFS directory offset data");
127 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
128     "NFS local lock rollback");
129 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
130 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
131 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
132 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
133 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
134 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
135 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
136 
137 /*
138  * Definition of mutex locks.
139  * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
140  * and assorted other nfsd structures.
141  */
142 struct mtx newnfsd_mtx;
143 struct mtx nfs_sockl_mutex;
144 struct mtx nfs_state_mutex;
145 struct mtx nfs_nameid_mutex;
146 struct mtx nfs_req_mutex;
147 struct mtx nfs_slock_mutex;
148 struct mtx nfs_clstate_mutex;
149 
150 /* local functions */
151 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
152 
153 #ifdef __NO_STRICT_ALIGNMENT
154 /*
155  * These architectures don't need re-alignment, so just return.
156  */
157 int
158 newnfs_realign(struct mbuf **pm, int how)
159 {
160 
161 	return (0);
162 }
163 #else	/* !__NO_STRICT_ALIGNMENT */
164 /*
165  *	newnfs_realign:
166  *
167  *	Check for badly aligned mbuf data and realign by copying the unaligned
168  *	portion of the data into a new mbuf chain and freeing the portions
169  *	of the old chain that were replaced.
170  *
171  *	We cannot simply realign the data within the existing mbuf chain
172  *	because the underlying buffers may contain other rpc commands and
173  *	we cannot afford to overwrite them.
174  *
175  *	We would prefer to avoid this situation entirely.  The situation does
176  *	not occur with NFS/UDP and is supposed to only occasionally occur
177  *	with TCP.  Use vfs.nfs.realign_count and realign_test to check this.
178  *
179  */
180 int
181 newnfs_realign(struct mbuf **pm, int how)
182 {
183 	struct mbuf *m, *n;
184 	int off, space;
185 
186 	++nfs_realign_test;
187 	while ((m = *pm) != NULL) {
188 		if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
189 			/*
190 			 * NB: we can't depend on m_pkthdr.len to help us
191 			 * decide what to do here.  May not be worth doing
192 			 * the m_length calculation as m_copyback will
193 			 * expand the mbuf chain below as needed.
194 			 */
195 			space = m_length(m, NULL);
196 			if (space >= MINCLSIZE) {
197 				/* NB: m_copyback handles space > MCLBYTES */
198 				n = m_getcl(how, MT_DATA, 0);
199 			} else
200 				n = m_get(how, MT_DATA);
201 			if (n == NULL)
202 				return (ENOMEM);
203 			/*
204 			 * Align the remainder of the mbuf chain.
205 			 */
206 			n->m_len = 0;
207 			off = 0;
208 			while (m != NULL) {
209 				m_copyback(n, off, m->m_len, mtod(m, caddr_t));
210 				off += m->m_len;
211 				m = m->m_next;
212 			}
213 			m_freem(*pm);
214 			*pm = n;
215 			++nfs_realign_count;
216 			break;
217 		}
218 		pm = &m->m_next;
219 	}
220 
221 	return (0);
222 }
223 #endif	/* __NO_STRICT_ALIGNMENT */
224 
225 #ifdef notdef
226 static void
227 nfsrv_object_create(struct vnode *vp, struct thread *td)
228 {
229 
230 	if (vp == NULL || vp->v_type != VREG)
231 		return;
232 	(void) vfs_object_create(vp, td, td->td_ucred);
233 }
234 #endif
235 
236 /*
237  * Look up a file name. Basically just initialize stuff and call namei().
238  */
239 int
240 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p __unused)
241 {
242 	int error;
243 
244 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname);
245 	error = namei(ndp);
246 	if (!error) {
247 		NDFREE_PNBUF(ndp);
248 	}
249 	return (error);
250 }
251 
252 /*
253  * Copy NFS uid, gids to the cred structure.
254  */
255 void
256 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
257 {
258 
259 	KASSERT(nfscr->nfsc_ngroups >= 0,
260 	    ("newnfs_copycred: negative nfsc_ngroups"));
261 	cr->cr_uid = nfscr->nfsc_uid;
262 	crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
263 }
264 
265 /*
266  * Map args from nfsmsleep() to msleep().
267  */
268 int
269 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
270     struct timespec *ts)
271 {
272 	u_int64_t nsecval;
273 	int error, timeo;
274 
275 	if (ts) {
276 		timeo = hz * ts->tv_sec;
277 		nsecval = (u_int64_t)ts->tv_nsec;
278 		nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
279 		    1000000000;
280 		timeo += (int)nsecval;
281 	} else {
282 		timeo = 0;
283 	}
284 	error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
285 	return (error);
286 }
287 
288 /*
289  * Get the file system info for the server. For now, just assume FFS.
290  */
291 void
292 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
293 {
294 	int pref;
295 
296 	/*
297 	 * XXX
298 	 * There should be file system VFS OP(s) to get this information.
299 	 * For now, assume ufs.
300 	 */
301 	if (isdgram)
302 		pref = NFS_MAXDGRAMDATA;
303 	else
304 		pref = nfs_srvmaxio;
305 	sip->fs_rtmax = nfs_srvmaxio;
306 	sip->fs_rtpref = pref;
307 	sip->fs_rtmult = NFS_FABLKSIZE;
308 	sip->fs_wtmax = nfs_srvmaxio;
309 	sip->fs_wtpref = pref;
310 	sip->fs_wtmult = NFS_FABLKSIZE;
311 	sip->fs_dtpref = pref;
312 	sip->fs_maxfilesize = 0xffffffffffffffffull;
313 	sip->fs_timedelta.tv_sec = 0;
314 	sip->fs_timedelta.tv_nsec = 1;
315 	sip->fs_properties = (NFSV3FSINFO_LINK |
316 	    NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
317 	    NFSV3FSINFO_CANSETTIME);
318 }
319 
320 /*
321  * Do the pathconf vnode op.
322  */
323 int
324 nfsvno_pathconf(struct vnode *vp, int flag, long *retf,
325     struct ucred *cred, struct thread *p)
326 {
327 	int error;
328 
329 	error = VOP_PATHCONF(vp, flag, retf);
330 	if (error == EOPNOTSUPP || error == EINVAL) {
331 		/*
332 		 * Some file systems return EINVAL for name arguments not
333 		 * supported and some return EOPNOTSUPP for this case.
334 		 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
335 		 * just fake them.
336 		 */
337 		switch (flag) {
338 		case _PC_LINK_MAX:
339 			*retf = NFS_LINK_MAX;
340 			break;
341 		case _PC_NAME_MAX:
342 			*retf = NAME_MAX;
343 			break;
344 		case _PC_CHOWN_RESTRICTED:
345 			*retf = 1;
346 			break;
347 		case _PC_NO_TRUNC:
348 			*retf = 1;
349 			break;
350 		default:
351 			/*
352 			 * Only happens if a _PC_xxx is added to the server,
353 			 * but this isn't updated.
354 			 */
355 			*retf = 0;
356 			printf("nfsrvd pathconf flag=%d not supp\n", flag);
357 		}
358 		error = 0;
359 	}
360 	NFSEXITCODE(error);
361 	return (error);
362 }
363 
364 /* Fake nfsrv_atroot. Just return 0 */
365 int
366 nfsrv_atroot(struct vnode *vp, uint64_t *retp)
367 {
368 
369 	return (0);
370 }
371 
372 /*
373  * Set the credentials to refer to root.
374  * If only the various BSDen could agree on whether cr_gid is a separate
375  * field or cr_groups[0]...
376  */
377 void
378 newnfs_setroot(struct ucred *cred)
379 {
380 
381 	cred->cr_uid = 0;
382 	cred->cr_groups[0] = 0;
383 	cred->cr_ngroups = 1;
384 }
385 
386 /*
387  * Get the client credential. Used for Renew and recovery.
388  */
389 struct ucred *
390 newnfs_getcred(void)
391 {
392 	struct ucred *cred;
393 	struct thread *td = curthread;
394 
395 	cred = crdup(td->td_ucred);
396 	newnfs_setroot(cred);
397 	return (cred);
398 }
399 
400 /*
401  * Sleep for a short period of time unless errval == NFSERR_GRACE, where
402  * the sleep should be for 5 seconds.
403  * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
404  * an event that never gets a wakeup. Only return EINTR or 0.
405  */
406 int
407 nfs_catnap(int prio, int errval, const char *wmesg)
408 {
409 	static int non_event;
410 	int ret;
411 
412 	if (errval == NFSERR_GRACE)
413 		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
414 	else
415 		ret = tsleep(&non_event, prio, wmesg, 1);
416 	if (ret != EINTR)
417 		ret = 0;
418 	return (ret);
419 }
420 
421 /*
422  * Get referral. For now, just fail.
423  */
424 struct nfsreferral *
425 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
426 {
427 
428 	return (NULL);
429 }
430 
431 static int
432 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
433 {
434 	int error;
435 
436 	error = nfssvc_call(td, uap, td->td_ucred);
437 	NFSEXITCODE(error);
438 	return (error);
439 }
440 
441 static int
442 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
443 {
444 	int error = EINVAL, i, j;
445 	struct nfsd_idargs nid;
446 	struct nfsd_oidargs onid;
447 	struct {
448 		int vers;	/* Just the first field of nfsstats. */
449 	} nfsstatver;
450 
451 	if (uap->flag & NFSSVC_IDNAME) {
452 		if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
453 			error = copyin(uap->argp, &nid, sizeof(nid));
454 		else {
455 			error = copyin(uap->argp, &onid, sizeof(onid));
456 			if (error == 0) {
457 				nid.nid_flag = onid.nid_flag;
458 				nid.nid_uid = onid.nid_uid;
459 				nid.nid_gid = onid.nid_gid;
460 				nid.nid_usermax = onid.nid_usermax;
461 				nid.nid_usertimeout = onid.nid_usertimeout;
462 				nid.nid_name = onid.nid_name;
463 				nid.nid_namelen = onid.nid_namelen;
464 				nid.nid_ngroup = 0;
465 				nid.nid_grps = NULL;
466 			}
467 		}
468 		if (error)
469 			goto out;
470 		error = nfssvc_idname(&nid);
471 		goto out;
472 	} else if (uap->flag & NFSSVC_GETSTATS) {
473 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
474 			/* Copy fields to the old ext_nfsstat structure. */
475 			oldnfsstats.attrcache_hits =
476 			    nfsstatsv1.attrcache_hits;
477 			oldnfsstats.attrcache_misses =
478 			    nfsstatsv1.attrcache_misses;
479 			oldnfsstats.lookupcache_hits =
480 			    nfsstatsv1.lookupcache_hits;
481 			oldnfsstats.lookupcache_misses =
482 			    nfsstatsv1.lookupcache_misses;
483 			oldnfsstats.direofcache_hits =
484 			    nfsstatsv1.direofcache_hits;
485 			oldnfsstats.direofcache_misses =
486 			    nfsstatsv1.direofcache_misses;
487 			oldnfsstats.accesscache_hits =
488 			    nfsstatsv1.accesscache_hits;
489 			oldnfsstats.accesscache_misses =
490 			    nfsstatsv1.accesscache_misses;
491 			oldnfsstats.biocache_reads =
492 			    nfsstatsv1.biocache_reads;
493 			oldnfsstats.read_bios =
494 			    nfsstatsv1.read_bios;
495 			oldnfsstats.read_physios =
496 			    nfsstatsv1.read_physios;
497 			oldnfsstats.biocache_writes =
498 			    nfsstatsv1.biocache_writes;
499 			oldnfsstats.write_bios =
500 			    nfsstatsv1.write_bios;
501 			oldnfsstats.write_physios =
502 			    nfsstatsv1.write_physios;
503 			oldnfsstats.biocache_readlinks =
504 			    nfsstatsv1.biocache_readlinks;
505 			oldnfsstats.readlink_bios =
506 			    nfsstatsv1.readlink_bios;
507 			oldnfsstats.biocache_readdirs =
508 			    nfsstatsv1.biocache_readdirs;
509 			oldnfsstats.readdir_bios =
510 			    nfsstatsv1.readdir_bios;
511 			for (i = 0; i < NFSV4_NPROCS; i++)
512 				oldnfsstats.rpccnt[i] = nfsstatsv1.rpccnt[i];
513 			oldnfsstats.rpcretries = nfsstatsv1.rpcretries;
514 			for (i = 0; i < NFSV4OP_NOPS; i++)
515 				oldnfsstats.srvrpccnt[i] =
516 				    nfsstatsv1.srvrpccnt[i];
517 			for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
518 			    i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
519 				oldnfsstats.srvrpccnt[j] =
520 				    nfsstatsv1.srvrpccnt[i];
521 			oldnfsstats.reserved_0 = 0;
522 			oldnfsstats.reserved_1 = 0;
523 			oldnfsstats.rpcrequests = nfsstatsv1.rpcrequests;
524 			oldnfsstats.rpctimeouts = nfsstatsv1.rpctimeouts;
525 			oldnfsstats.rpcunexpected = nfsstatsv1.rpcunexpected;
526 			oldnfsstats.rpcinvalid = nfsstatsv1.rpcinvalid;
527 			oldnfsstats.srvcache_inproghits =
528 			    nfsstatsv1.srvcache_inproghits;
529 			oldnfsstats.reserved_2 = 0;
530 			oldnfsstats.srvcache_nonidemdonehits =
531 			    nfsstatsv1.srvcache_nonidemdonehits;
532 			oldnfsstats.srvcache_misses =
533 			    nfsstatsv1.srvcache_misses;
534 			oldnfsstats.srvcache_tcppeak =
535 			    nfsstatsv1.srvcache_tcppeak;
536 			oldnfsstats.srvcache_size = nfsstatsv1.srvcache_size;
537 			oldnfsstats.srvclients = nfsstatsv1.srvclients;
538 			oldnfsstats.srvopenowners = nfsstatsv1.srvopenowners;
539 			oldnfsstats.srvopens = nfsstatsv1.srvopens;
540 			oldnfsstats.srvlockowners = nfsstatsv1.srvlockowners;
541 			oldnfsstats.srvlocks = nfsstatsv1.srvlocks;
542 			oldnfsstats.srvdelegates = nfsstatsv1.srvdelegates;
543 			for (i = 0; i < NFSV4OP_CBNOPS; i++)
544 				oldnfsstats.cbrpccnt[i] =
545 				    nfsstatsv1.cbrpccnt[i];
546 			oldnfsstats.clopenowners = nfsstatsv1.clopenowners;
547 			oldnfsstats.clopens = nfsstatsv1.clopens;
548 			oldnfsstats.cllockowners = nfsstatsv1.cllockowners;
549 			oldnfsstats.cllocks = nfsstatsv1.cllocks;
550 			oldnfsstats.cldelegates = nfsstatsv1.cldelegates;
551 			oldnfsstats.cllocalopenowners =
552 			    nfsstatsv1.cllocalopenowners;
553 			oldnfsstats.cllocalopens = nfsstatsv1.cllocalopens;
554 			oldnfsstats.cllocallockowners =
555 			    nfsstatsv1.cllocallockowners;
556 			oldnfsstats.cllocallocks = nfsstatsv1.cllocallocks;
557 			error = copyout(&oldnfsstats, uap->argp,
558 			    sizeof (oldnfsstats));
559 		} else {
560 			error = copyin(uap->argp, &nfsstatver,
561 			    sizeof(nfsstatver));
562 			if (error == 0) {
563 				if (nfsstatver.vers == NFSSTATS_OV1) {
564 					/* Copy nfsstatsv1 to nfsstatsov1. */
565 					nfsstatsov1.attrcache_hits =
566 					    nfsstatsv1.attrcache_hits;
567 					nfsstatsov1.attrcache_misses =
568 					    nfsstatsv1.attrcache_misses;
569 					nfsstatsov1.lookupcache_hits =
570 					    nfsstatsv1.lookupcache_hits;
571 					nfsstatsov1.lookupcache_misses =
572 					    nfsstatsv1.lookupcache_misses;
573 					nfsstatsov1.direofcache_hits =
574 					    nfsstatsv1.direofcache_hits;
575 					nfsstatsov1.direofcache_misses =
576 					    nfsstatsv1.direofcache_misses;
577 					nfsstatsov1.accesscache_hits =
578 					    nfsstatsv1.accesscache_hits;
579 					nfsstatsov1.accesscache_misses =
580 					    nfsstatsv1.accesscache_misses;
581 					nfsstatsov1.biocache_reads =
582 					    nfsstatsv1.biocache_reads;
583 					nfsstatsov1.read_bios =
584 					    nfsstatsv1.read_bios;
585 					nfsstatsov1.read_physios =
586 					    nfsstatsv1.read_physios;
587 					nfsstatsov1.biocache_writes =
588 					    nfsstatsv1.biocache_writes;
589 					nfsstatsov1.write_bios =
590 					    nfsstatsv1.write_bios;
591 					nfsstatsov1.write_physios =
592 					    nfsstatsv1.write_physios;
593 					nfsstatsov1.biocache_readlinks =
594 					    nfsstatsv1.biocache_readlinks;
595 					nfsstatsov1.readlink_bios =
596 					    nfsstatsv1.readlink_bios;
597 					nfsstatsov1.biocache_readdirs =
598 					    nfsstatsv1.biocache_readdirs;
599 					nfsstatsov1.readdir_bios =
600 					    nfsstatsv1.readdir_bios;
601 					for (i = 0; i < NFSV42_OLDNPROCS; i++)
602 						nfsstatsov1.rpccnt[i] =
603 						    nfsstatsv1.rpccnt[i];
604 					nfsstatsov1.rpcretries =
605 					    nfsstatsv1.rpcretries;
606 					for (i = 0; i < NFSV42_PURENOPS; i++)
607 						nfsstatsov1.srvrpccnt[i] =
608 						    nfsstatsv1.srvrpccnt[i];
609 					for (i = NFSV42_NOPS,
610 					     j = NFSV42_PURENOPS;
611 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
612 					     i++, j++)
613 						nfsstatsov1.srvrpccnt[j] =
614 						    nfsstatsv1.srvrpccnt[i];
615 					nfsstatsov1.reserved_0 = 0;
616 					nfsstatsov1.reserved_1 = 0;
617 					nfsstatsov1.rpcrequests =
618 					    nfsstatsv1.rpcrequests;
619 					nfsstatsov1.rpctimeouts =
620 					    nfsstatsv1.rpctimeouts;
621 					nfsstatsov1.rpcunexpected =
622 					    nfsstatsv1.rpcunexpected;
623 					nfsstatsov1.rpcinvalid =
624 					    nfsstatsv1.rpcinvalid;
625 					nfsstatsov1.srvcache_inproghits =
626 					    nfsstatsv1.srvcache_inproghits;
627 					nfsstatsov1.reserved_2 = 0;
628 					nfsstatsov1.srvcache_nonidemdonehits =
629 					    nfsstatsv1.srvcache_nonidemdonehits;
630 					nfsstatsov1.srvcache_misses =
631 					    nfsstatsv1.srvcache_misses;
632 					nfsstatsov1.srvcache_tcppeak =
633 					    nfsstatsv1.srvcache_tcppeak;
634 					nfsstatsov1.srvcache_size =
635 					    nfsstatsv1.srvcache_size;
636 					nfsstatsov1.srvclients =
637 					    nfsstatsv1.srvclients;
638 					nfsstatsov1.srvopenowners =
639 					    nfsstatsv1.srvopenowners;
640 					nfsstatsov1.srvopens =
641 					    nfsstatsv1.srvopens;
642 					nfsstatsov1.srvlockowners =
643 					    nfsstatsv1.srvlockowners;
644 					nfsstatsov1.srvlocks =
645 					    nfsstatsv1.srvlocks;
646 					nfsstatsov1.srvdelegates =
647 					    nfsstatsv1.srvdelegates;
648 					for (i = 0; i < NFSV42_CBNOPS; i++)
649 						nfsstatsov1.cbrpccnt[i] =
650 						    nfsstatsv1.cbrpccnt[i];
651 					nfsstatsov1.clopenowners =
652 					    nfsstatsv1.clopenowners;
653 					nfsstatsov1.clopens =
654 					    nfsstatsv1.clopens;
655 					nfsstatsov1.cllockowners =
656 					    nfsstatsv1.cllockowners;
657 					nfsstatsov1.cllocks =
658 					    nfsstatsv1.cllocks;
659 					nfsstatsov1.cldelegates =
660 					    nfsstatsv1.cldelegates;
661 					nfsstatsov1.cllocalopenowners =
662 					    nfsstatsv1.cllocalopenowners;
663 					nfsstatsov1.cllocalopens =
664 					    nfsstatsv1.cllocalopens;
665 					nfsstatsov1.cllocallockowners =
666 					    nfsstatsv1.cllocallockowners;
667 					nfsstatsov1.cllocallocks =
668 					    nfsstatsv1.cllocallocks;
669 					nfsstatsov1.srvstartcnt =
670 					    nfsstatsv1.srvstartcnt;
671 					nfsstatsov1.srvdonecnt =
672 					    nfsstatsv1.srvdonecnt;
673 					for (i = NFSV42_NOPS,
674 					     j = NFSV42_PURENOPS;
675 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
676 					     i++, j++) {
677 						nfsstatsov1.srvbytes[j] =
678 						    nfsstatsv1.srvbytes[i];
679 						nfsstatsov1.srvops[j] =
680 						    nfsstatsv1.srvops[i];
681 						nfsstatsov1.srvduration[j] =
682 						    nfsstatsv1.srvduration[i];
683 					}
684 					nfsstatsov1.busyfrom =
685 					    nfsstatsv1.busyfrom;
686 					nfsstatsov1.busyfrom =
687 					    nfsstatsv1.busyfrom;
688 					error = copyout(&nfsstatsov1, uap->argp,
689 					    sizeof(nfsstatsov1));
690 				} else if (nfsstatver.vers != NFSSTATS_V1)
691 					error = EPERM;
692 				else
693 					error = copyout(&nfsstatsv1, uap->argp,
694 					    sizeof(nfsstatsv1));
695 			}
696 		}
697 		if (error == 0) {
698 			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
699 				nfsstatsv1.attrcache_hits = 0;
700 				nfsstatsv1.attrcache_misses = 0;
701 				nfsstatsv1.lookupcache_hits = 0;
702 				nfsstatsv1.lookupcache_misses = 0;
703 				nfsstatsv1.direofcache_hits = 0;
704 				nfsstatsv1.direofcache_misses = 0;
705 				nfsstatsv1.accesscache_hits = 0;
706 				nfsstatsv1.accesscache_misses = 0;
707 				nfsstatsv1.biocache_reads = 0;
708 				nfsstatsv1.read_bios = 0;
709 				nfsstatsv1.read_physios = 0;
710 				nfsstatsv1.biocache_writes = 0;
711 				nfsstatsv1.write_bios = 0;
712 				nfsstatsv1.write_physios = 0;
713 				nfsstatsv1.biocache_readlinks = 0;
714 				nfsstatsv1.readlink_bios = 0;
715 				nfsstatsv1.biocache_readdirs = 0;
716 				nfsstatsv1.readdir_bios = 0;
717 				nfsstatsv1.rpcretries = 0;
718 				nfsstatsv1.rpcrequests = 0;
719 				nfsstatsv1.rpctimeouts = 0;
720 				nfsstatsv1.rpcunexpected = 0;
721 				nfsstatsv1.rpcinvalid = 0;
722 				bzero(nfsstatsv1.rpccnt,
723 				    sizeof(nfsstatsv1.rpccnt));
724 			}
725 			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
726 				nfsstatsv1.srvcache_inproghits = 0;
727 				nfsstatsv1.srvcache_nonidemdonehits = 0;
728 				nfsstatsv1.srvcache_misses = 0;
729 				nfsstatsv1.srvcache_tcppeak = 0;
730 				bzero(nfsstatsv1.srvrpccnt,
731 				    sizeof(nfsstatsv1.srvrpccnt));
732 				bzero(nfsstatsv1.cbrpccnt,
733 				    sizeof(nfsstatsv1.cbrpccnt));
734 			}
735 		}
736 		goto out;
737 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
738 		u_short sockport;
739 		struct nfsuserd_args nargs;
740 
741 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
742 			error = copyin(uap->argp, (caddr_t)&sockport,
743 			    sizeof (u_short));
744 			if (error == 0) {
745 				nargs.nuserd_family = AF_INET;
746 				nargs.nuserd_port = sockport;
747 			}
748 		} else {
749 			/*
750 			 * New nfsuserd_args structure, which indicates
751 			 * which IP version to use along with the port#.
752 			 */
753 			error = copyin(uap->argp, &nargs, sizeof(nargs));
754 		}
755 		if (!error)
756 			error = nfsrv_nfsuserdport(&nargs, p);
757 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
758 		nfsrv_nfsuserddelport();
759 		error = 0;
760 	}
761 
762 out:
763 	NFSEXITCODE(error);
764 	return (error);
765 }
766 
767 /*
768  * called by all three modevent routines, so that it gets things
769  * initialized soon enough.
770  */
771 void
772 newnfs_portinit(void)
773 {
774 	static int inited = 0;
775 
776 	if (inited)
777 		return;
778 	inited = 1;
779 	/* Initialize SMP locks used by both client and server. */
780 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
781 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
782 	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
783 }
784 
785 /*
786  * Determine if the file system supports NFSv4 ACLs.
787  * Return 1 if it does, 0 otherwise.
788  */
789 int
790 nfs_supportsnfsv4acls(struct vnode *vp)
791 {
792 	int error;
793 	long retval;
794 
795 	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
796 
797 	if (nfsrv_useacl == 0)
798 		return (0);
799 	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
800 	if (error == 0 && retval != 0)
801 		return (1);
802 	return (0);
803 }
804 
805 /*
806  * These are the first fields of all the context structures passed into
807  * nfs_pnfsio().
808  */
809 struct pnfsio {
810 	int		done;
811 	int		inprog;
812 	struct task	tsk;
813 };
814 
815 /*
816  * Do a mirror I/O on a pNFS thread.
817  */
818 int
819 nfs_pnfsio(task_fn_t *func, void *context)
820 {
821 	struct pnfsio *pio;
822 	int ret;
823 	static struct taskqueue *pnfsioq = NULL;
824 
825 	pio = (struct pnfsio *)context;
826 	if (pnfsioq == NULL) {
827 		if (nfs_pnfsiothreads == 0)
828 			return (EPERM);
829 		if (nfs_pnfsiothreads < 0)
830 			nfs_pnfsiothreads = mp_ncpus * 4;
831 		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
832 		    taskqueue_thread_enqueue, &pnfsioq);
833 		if (pnfsioq == NULL)
834 			return (ENOMEM);
835 		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
836 		    0, "pnfsiot");
837 		if (ret != 0) {
838 			taskqueue_free(pnfsioq);
839 			pnfsioq = NULL;
840 			return (ret);
841 		}
842 	}
843 	pio->inprog = 1;
844 	TASK_INIT(&pio->tsk, 0, func, context);
845 	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
846 	if (ret != 0)
847 		pio->inprog = 0;
848 	return (ret);
849 }
850 
851 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
852 
853 /*
854  * Called once to initialize data structures...
855  */
856 static int
857 nfscommon_modevent(module_t mod, int type, void *data)
858 {
859 	int error = 0;
860 	static int loaded = 0;
861 
862 	switch (type) {
863 	case MOD_LOAD:
864 		if (loaded)
865 			goto out;
866 		newnfs_portinit();
867 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
868 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
869 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
870 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
871 		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
872 		    MTX_DEF);
873 		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
874 		TAILQ_INIT(&nfsrv_devidhead);
875 		newnfs_init();
876 		nfsd_call_nfscommon = nfssvc_nfscommon;
877 		loaded = 1;
878 		break;
879 
880 	case MOD_UNLOAD:
881 		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != NOTRUNNING ||
882 		    nfs_numnfscbd != 0) {
883 			error = EBUSY;
884 			break;
885 		}
886 
887 		nfsd_call_nfscommon = NULL;
888 		/* Clean out the name<-->id cache. */
889 		nfsrv_cleanusergroup();
890 		/* and get rid of the mutexes */
891 		mtx_destroy(&nfs_nameid_mutex);
892 		mtx_destroy(&newnfsd_mtx);
893 		mtx_destroy(&nfs_state_mutex);
894 		mtx_destroy(&nfs_clstate_mutex);
895 		mtx_destroy(&nfs_sockl_mutex);
896 		mtx_destroy(&nfs_slock_mutex);
897 		mtx_destroy(&nfs_req_mutex);
898 		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
899 		mtx_destroy(&nfsrv_dslock_mtx);
900 		loaded = 0;
901 		break;
902 	default:
903 		error = EOPNOTSUPP;
904 		break;
905 	}
906 
907 out:
908 	NFSEXITCODE(error);
909 	return error;
910 }
911 static moduledata_t nfscommon_mod = {
912 	"nfscommon",
913 	nfscommon_modevent,
914 	NULL,
915 };
916 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
917 
918 /* So that loader and kldload(2) can find us, wherever we are.. */
919 MODULE_VERSION(nfscommon, 1);
920 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
921 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
922