xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision 62cfcf62f627e5093fb37026a6d8c98e4d2ef04c)
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 
80 int nfs_pnfsio(task_fn_t *, void *);
81 
82 static int nfs_realign_test;
83 static int nfs_realign_count;
84 static struct ext_nfsstats oldnfsstats;
85 static struct nfsstatsov1 nfsstatsov1;
86 
87 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
88     "NFS filesystem");
89 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
90     0, "Number of realign tests done");
91 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
92     0, "Number of mbuf realignments done");
93 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
94     nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
95     "NFSv4 callback addr for server to use");
96 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
97     0, "Debug level for NFS client");
98 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
99     0, "Size of hash tables for uid/name mapping");
100 int nfs_pnfsiothreads = -1;
101 SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsiothreads, CTLFLAG_RW, &nfs_pnfsiothreads,
102     0, "Number of pNFS mirror I/O threads");
103 
104 /*
105  * Defines for malloc
106  * (Here for FreeBSD, since they allocate storage.)
107  */
108 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
109 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
110 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
111     "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
112 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
113 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
114 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
115 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
116 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
117 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
118 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
119 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
120 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
121 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
122 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
123 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
124 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
125 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer");
126 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
127     "NFS directory offset data");
128 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
129     "NFS local lock rollback");
130 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
131 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
132 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
133 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
134 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
135 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
136 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
137 
138 /*
139  * Definition of mutex locks.
140  * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
141  * and assorted other nfsd structures.
142  */
143 struct mtx newnfsd_mtx;
144 struct mtx nfs_sockl_mutex;
145 struct mtx nfs_state_mutex;
146 struct mtx nfs_nameid_mutex;
147 struct mtx nfs_req_mutex;
148 struct mtx nfs_slock_mutex;
149 struct mtx nfs_clstate_mutex;
150 
151 /* local functions */
152 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
153 
154 #ifdef __NO_STRICT_ALIGNMENT
155 /*
156  * These architectures don't need re-alignment, so just return.
157  */
158 int
159 newnfs_realign(struct mbuf **pm, int how)
160 {
161 
162 	return (0);
163 }
164 #else	/* !__NO_STRICT_ALIGNMENT */
165 /*
166  *	newnfs_realign:
167  *
168  *	Check for badly aligned mbuf data and realign by copying the unaligned
169  *	portion of the data into a new mbuf chain and freeing the portions
170  *	of the old chain that were replaced.
171  *
172  *	We cannot simply realign the data within the existing mbuf chain
173  *	because the underlying buffers may contain other rpc commands and
174  *	we cannot afford to overwrite them.
175  *
176  *	We would prefer to avoid this situation entirely.  The situation does
177  *	not occur with NFS/UDP and is supposed to only occasionally occur
178  *	with TCP.  Use vfs.nfs.realign_count and realign_test to check this.
179  *
180  */
181 int
182 newnfs_realign(struct mbuf **pm, int how)
183 {
184 	struct mbuf *m, *n;
185 	int off, space;
186 
187 	++nfs_realign_test;
188 	while ((m = *pm) != NULL) {
189 		if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
190 			/*
191 			 * NB: we can't depend on m_pkthdr.len to help us
192 			 * decide what to do here.  May not be worth doing
193 			 * the m_length calculation as m_copyback will
194 			 * expand the mbuf chain below as needed.
195 			 */
196 			space = m_length(m, NULL);
197 			if (space >= MINCLSIZE) {
198 				/* NB: m_copyback handles space > MCLBYTES */
199 				n = m_getcl(how, MT_DATA, 0);
200 			} else
201 				n = m_get(how, MT_DATA);
202 			if (n == NULL)
203 				return (ENOMEM);
204 			/*
205 			 * Align the remainder of the mbuf chain.
206 			 */
207 			n->m_len = 0;
208 			off = 0;
209 			while (m != NULL) {
210 				m_copyback(n, off, m->m_len, mtod(m, caddr_t));
211 				off += m->m_len;
212 				m = m->m_next;
213 			}
214 			m_freem(*pm);
215 			*pm = n;
216 			++nfs_realign_count;
217 			break;
218 		}
219 		pm = &m->m_next;
220 	}
221 
222 	return (0);
223 }
224 #endif	/* __NO_STRICT_ALIGNMENT */
225 
226 #ifdef notdef
227 static void
228 nfsrv_object_create(struct vnode *vp, struct thread *td)
229 {
230 
231 	if (vp == NULL || vp->v_type != VREG)
232 		return;
233 	(void) vfs_object_create(vp, td, td->td_ucred);
234 }
235 #endif
236 
237 /*
238  * Look up a file name. Basically just initialize stuff and call namei().
239  */
240 int
241 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
242 {
243 	int error;
244 
245 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
246 	    p);
247 	error = namei(ndp);
248 	if (!error) {
249 		NDFREE(ndp, NDF_ONLY_PNBUF);
250 	}
251 	return (error);
252 }
253 
254 /*
255  * Copy NFS uid, gids to the cred structure.
256  */
257 void
258 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
259 {
260 
261 	KASSERT(nfscr->nfsc_ngroups >= 0,
262 	    ("newnfs_copycred: negative nfsc_ngroups"));
263 	cr->cr_uid = nfscr->nfsc_uid;
264 	crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
265 }
266 
267 /*
268  * Map args from nfsmsleep() to msleep().
269  */
270 int
271 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
272     struct timespec *ts)
273 {
274 	u_int64_t nsecval;
275 	int error, timeo;
276 
277 	if (ts) {
278 		timeo = hz * ts->tv_sec;
279 		nsecval = (u_int64_t)ts->tv_nsec;
280 		nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
281 		    1000000000;
282 		timeo += (int)nsecval;
283 	} else {
284 		timeo = 0;
285 	}
286 	error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
287 	return (error);
288 }
289 
290 /*
291  * Get the file system info for the server. For now, just assume FFS.
292  */
293 void
294 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
295 {
296 	int pref;
297 
298 	/*
299 	 * XXX
300 	 * There should be file system VFS OP(s) to get this information.
301 	 * For now, assume ufs.
302 	 */
303 	if (isdgram)
304 		pref = NFS_MAXDGRAMDATA;
305 	else
306 		pref = NFS_SRVMAXIO;
307 	sip->fs_rtmax = NFS_SRVMAXIO;
308 	sip->fs_rtpref = pref;
309 	sip->fs_rtmult = NFS_FABLKSIZE;
310 	sip->fs_wtmax = NFS_SRVMAXIO;
311 	sip->fs_wtpref = pref;
312 	sip->fs_wtmult = NFS_FABLKSIZE;
313 	sip->fs_dtpref = pref;
314 	sip->fs_maxfilesize = 0xffffffffffffffffull;
315 	sip->fs_timedelta.tv_sec = 0;
316 	sip->fs_timedelta.tv_nsec = 1;
317 	sip->fs_properties = (NFSV3FSINFO_LINK |
318 	    NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
319 	    NFSV3FSINFO_CANSETTIME);
320 }
321 
322 /*
323  * Do the pathconf vnode op.
324  */
325 int
326 nfsvno_pathconf(struct vnode *vp, int flag, long *retf,
327     struct ucred *cred, struct thread *p)
328 {
329 	int error;
330 
331 	error = VOP_PATHCONF(vp, flag, retf);
332 	if (error == EOPNOTSUPP || error == EINVAL) {
333 		/*
334 		 * Some file systems return EINVAL for name arguments not
335 		 * supported and some return EOPNOTSUPP for this case.
336 		 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
337 		 * just fake them.
338 		 */
339 		switch (flag) {
340 		case _PC_LINK_MAX:
341 			*retf = NFS_LINK_MAX;
342 			break;
343 		case _PC_NAME_MAX:
344 			*retf = NAME_MAX;
345 			break;
346 		case _PC_CHOWN_RESTRICTED:
347 			*retf = 1;
348 			break;
349 		case _PC_NO_TRUNC:
350 			*retf = 1;
351 			break;
352 		default:
353 			/*
354 			 * Only happens if a _PC_xxx is added to the server,
355 			 * but this isn't updated.
356 			 */
357 			*retf = 0;
358 			printf("nfsrvd pathconf flag=%d not supp\n", flag);
359 		}
360 		error = 0;
361 	}
362 	NFSEXITCODE(error);
363 	return (error);
364 }
365 
366 /* Fake nfsrv_atroot. Just return 0 */
367 int
368 nfsrv_atroot(struct vnode *vp, uint64_t *retp)
369 {
370 
371 	return (0);
372 }
373 
374 /*
375  * Set the credentials to refer to root.
376  * If only the various BSDen could agree on whether cr_gid is a separate
377  * field or cr_groups[0]...
378  */
379 void
380 newnfs_setroot(struct ucred *cred)
381 {
382 
383 	cred->cr_uid = 0;
384 	cred->cr_groups[0] = 0;
385 	cred->cr_ngroups = 1;
386 }
387 
388 /*
389  * Get the client credential. Used for Renew and recovery.
390  */
391 struct ucred *
392 newnfs_getcred(void)
393 {
394 	struct ucred *cred;
395 	struct thread *td = curthread;
396 
397 	cred = crdup(td->td_ucred);
398 	newnfs_setroot(cred);
399 	return (cred);
400 }
401 
402 /*
403  * Nfs timer routine
404  * Call the nfsd's timer function once/sec.
405  */
406 void
407 newnfs_timer(void *arg)
408 {
409 	static time_t lasttime = 0;
410 	/*
411 	 * Call the server timer, if set up.
412 	 * The argument indicates if it is the next second and therefore
413 	 * leases should be checked.
414 	 */
415 	if (lasttime != NFSD_MONOSEC) {
416 		lasttime = NFSD_MONOSEC;
417 		if (nfsd_call_servertimer != NULL)
418 			(*nfsd_call_servertimer)();
419 	}
420 	callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
421 }
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.srvrpc_errs = nfsstatsv1.srvrpc_errs;
546 			oldnfsstats.srv_errs = nfsstatsv1.srv_errs;
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.srvcache_idemdonehits =
554 			    nfsstatsv1.srvcache_idemdonehits;
555 			oldnfsstats.srvcache_nonidemdonehits =
556 			    nfsstatsv1.srvcache_nonidemdonehits;
557 			oldnfsstats.srvcache_misses =
558 			    nfsstatsv1.srvcache_misses;
559 			oldnfsstats.srvcache_tcppeak =
560 			    nfsstatsv1.srvcache_tcppeak;
561 			oldnfsstats.srvcache_size = nfsstatsv1.srvcache_size;
562 			oldnfsstats.srvclients = nfsstatsv1.srvclients;
563 			oldnfsstats.srvopenowners = nfsstatsv1.srvopenowners;
564 			oldnfsstats.srvopens = nfsstatsv1.srvopens;
565 			oldnfsstats.srvlockowners = nfsstatsv1.srvlockowners;
566 			oldnfsstats.srvlocks = nfsstatsv1.srvlocks;
567 			oldnfsstats.srvdelegates = nfsstatsv1.srvdelegates;
568 			for (i = 0; i < NFSV4OP_CBNOPS; i++)
569 				oldnfsstats.cbrpccnt[i] =
570 				    nfsstatsv1.cbrpccnt[i];
571 			oldnfsstats.clopenowners = nfsstatsv1.clopenowners;
572 			oldnfsstats.clopens = nfsstatsv1.clopens;
573 			oldnfsstats.cllockowners = nfsstatsv1.cllockowners;
574 			oldnfsstats.cllocks = nfsstatsv1.cllocks;
575 			oldnfsstats.cldelegates = nfsstatsv1.cldelegates;
576 			oldnfsstats.cllocalopenowners =
577 			    nfsstatsv1.cllocalopenowners;
578 			oldnfsstats.cllocalopens = nfsstatsv1.cllocalopens;
579 			oldnfsstats.cllocallockowners =
580 			    nfsstatsv1.cllocallockowners;
581 			oldnfsstats.cllocallocks = nfsstatsv1.cllocallocks;
582 			error = copyout(&oldnfsstats, uap->argp,
583 			    sizeof (oldnfsstats));
584 		} else {
585 			error = copyin(uap->argp, &nfsstatver,
586 			    sizeof(nfsstatver));
587 			if (error == 0) {
588 				if (nfsstatver.vers == NFSSTATS_OV1) {
589 					/* Copy nfsstatsv1 to nfsstatsov1. */
590 					nfsstatsov1.attrcache_hits =
591 					    nfsstatsv1.attrcache_hits;
592 					nfsstatsov1.attrcache_misses =
593 					    nfsstatsv1.attrcache_misses;
594 					nfsstatsov1.lookupcache_hits =
595 					    nfsstatsv1.lookupcache_hits;
596 					nfsstatsov1.lookupcache_misses =
597 					    nfsstatsv1.lookupcache_misses;
598 					nfsstatsov1.direofcache_hits =
599 					    nfsstatsv1.direofcache_hits;
600 					nfsstatsov1.direofcache_misses =
601 					    nfsstatsv1.direofcache_misses;
602 					nfsstatsov1.accesscache_hits =
603 					    nfsstatsv1.accesscache_hits;
604 					nfsstatsov1.accesscache_misses =
605 					    nfsstatsv1.accesscache_misses;
606 					nfsstatsov1.biocache_reads =
607 					    nfsstatsv1.biocache_reads;
608 					nfsstatsov1.read_bios =
609 					    nfsstatsv1.read_bios;
610 					nfsstatsov1.read_physios =
611 					    nfsstatsv1.read_physios;
612 					nfsstatsov1.biocache_writes =
613 					    nfsstatsv1.biocache_writes;
614 					nfsstatsov1.write_bios =
615 					    nfsstatsv1.write_bios;
616 					nfsstatsov1.write_physios =
617 					    nfsstatsv1.write_physios;
618 					nfsstatsov1.biocache_readlinks =
619 					    nfsstatsv1.biocache_readlinks;
620 					nfsstatsov1.readlink_bios =
621 					    nfsstatsv1.readlink_bios;
622 					nfsstatsov1.biocache_readdirs =
623 					    nfsstatsv1.biocache_readdirs;
624 					nfsstatsov1.readdir_bios =
625 					    nfsstatsv1.readdir_bios;
626 					for (i = 0; i < NFSV42_NPROCS; i++)
627 						nfsstatsov1.rpccnt[i] =
628 						    nfsstatsv1.rpccnt[i];
629 					nfsstatsov1.rpcretries =
630 					    nfsstatsv1.rpcretries;
631 					for (i = 0; i < NFSV42_PURENOPS; i++)
632 						nfsstatsov1.srvrpccnt[i] =
633 						    nfsstatsv1.srvrpccnt[i];
634 					for (i = NFSV42_NOPS,
635 					     j = NFSV42_PURENOPS;
636 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
637 					     i++, j++)
638 						nfsstatsov1.srvrpccnt[j] =
639 						    nfsstatsv1.srvrpccnt[i];
640 					nfsstatsov1.srvrpc_errs =
641 					    nfsstatsv1.srvrpc_errs;
642 					nfsstatsov1.srv_errs =
643 					    nfsstatsv1.srv_errs;
644 					nfsstatsov1.rpcrequests =
645 					    nfsstatsv1.rpcrequests;
646 					nfsstatsov1.rpctimeouts =
647 					    nfsstatsv1.rpctimeouts;
648 					nfsstatsov1.rpcunexpected =
649 					    nfsstatsv1.rpcunexpected;
650 					nfsstatsov1.rpcinvalid =
651 					    nfsstatsv1.rpcinvalid;
652 					nfsstatsov1.srvcache_inproghits =
653 					    nfsstatsv1.srvcache_inproghits;
654 					nfsstatsov1.srvcache_idemdonehits =
655 					    nfsstatsv1.srvcache_idemdonehits;
656 					nfsstatsov1.srvcache_nonidemdonehits =
657 					    nfsstatsv1.srvcache_nonidemdonehits;
658 					nfsstatsov1.srvcache_misses =
659 					    nfsstatsv1.srvcache_misses;
660 					nfsstatsov1.srvcache_tcppeak =
661 					    nfsstatsv1.srvcache_tcppeak;
662 					nfsstatsov1.srvcache_size =
663 					    nfsstatsv1.srvcache_size;
664 					nfsstatsov1.srvclients =
665 					    nfsstatsv1.srvclients;
666 					nfsstatsov1.srvopenowners =
667 					    nfsstatsv1.srvopenowners;
668 					nfsstatsov1.srvopens =
669 					    nfsstatsv1.srvopens;
670 					nfsstatsov1.srvlockowners =
671 					    nfsstatsv1.srvlockowners;
672 					nfsstatsov1.srvlocks =
673 					    nfsstatsv1.srvlocks;
674 					nfsstatsov1.srvdelegates =
675 					    nfsstatsv1.srvdelegates;
676 					for (i = 0; i < NFSV42_CBNOPS; i++)
677 						nfsstatsov1.cbrpccnt[i] =
678 						    nfsstatsv1.cbrpccnt[i];
679 					nfsstatsov1.clopenowners =
680 					    nfsstatsv1.clopenowners;
681 					nfsstatsov1.clopens =
682 					    nfsstatsv1.clopens;
683 					nfsstatsov1.cllockowners =
684 					    nfsstatsv1.cllockowners;
685 					nfsstatsov1.cllocks =
686 					    nfsstatsv1.cllocks;
687 					nfsstatsov1.cldelegates =
688 					    nfsstatsv1.cldelegates;
689 					nfsstatsov1.cllocalopenowners =
690 					    nfsstatsv1.cllocalopenowners;
691 					nfsstatsov1.cllocalopens =
692 					    nfsstatsv1.cllocalopens;
693 					nfsstatsov1.cllocallockowners =
694 					    nfsstatsv1.cllocallockowners;
695 					nfsstatsov1.cllocallocks =
696 					    nfsstatsv1.cllocallocks;
697 					nfsstatsov1.srvstartcnt =
698 					    nfsstatsv1.srvstartcnt;
699 					nfsstatsov1.srvdonecnt =
700 					    nfsstatsv1.srvdonecnt;
701 					for (i = NFSV42_NOPS,
702 					     j = NFSV42_PURENOPS;
703 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
704 					     i++, j++) {
705 						nfsstatsov1.srvbytes[j] =
706 						    nfsstatsv1.srvbytes[i];
707 						nfsstatsov1.srvops[j] =
708 						    nfsstatsv1.srvops[i];
709 						nfsstatsov1.srvduration[j] =
710 						    nfsstatsv1.srvduration[i];
711 					}
712 					nfsstatsov1.busyfrom =
713 					    nfsstatsv1.busyfrom;
714 					nfsstatsov1.busyfrom =
715 					    nfsstatsv1.busyfrom;
716 					error = copyout(&nfsstatsov1, uap->argp,
717 					    sizeof(nfsstatsov1));
718 				} else if (nfsstatver.vers != NFSSTATS_V1)
719 					error = EPERM;
720 				else
721 					error = copyout(&nfsstatsv1, uap->argp,
722 					    sizeof(nfsstatsv1));
723 			}
724 		}
725 		if (error == 0) {
726 			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
727 				nfsstatsv1.attrcache_hits = 0;
728 				nfsstatsv1.attrcache_misses = 0;
729 				nfsstatsv1.lookupcache_hits = 0;
730 				nfsstatsv1.lookupcache_misses = 0;
731 				nfsstatsv1.direofcache_hits = 0;
732 				nfsstatsv1.direofcache_misses = 0;
733 				nfsstatsv1.accesscache_hits = 0;
734 				nfsstatsv1.accesscache_misses = 0;
735 				nfsstatsv1.biocache_reads = 0;
736 				nfsstatsv1.read_bios = 0;
737 				nfsstatsv1.read_physios = 0;
738 				nfsstatsv1.biocache_writes = 0;
739 				nfsstatsv1.write_bios = 0;
740 				nfsstatsv1.write_physios = 0;
741 				nfsstatsv1.biocache_readlinks = 0;
742 				nfsstatsv1.readlink_bios = 0;
743 				nfsstatsv1.biocache_readdirs = 0;
744 				nfsstatsv1.readdir_bios = 0;
745 				nfsstatsv1.rpcretries = 0;
746 				nfsstatsv1.rpcrequests = 0;
747 				nfsstatsv1.rpctimeouts = 0;
748 				nfsstatsv1.rpcunexpected = 0;
749 				nfsstatsv1.rpcinvalid = 0;
750 				bzero(nfsstatsv1.rpccnt,
751 				    sizeof(nfsstatsv1.rpccnt));
752 			}
753 			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
754 				nfsstatsv1.srvrpc_errs = 0;
755 				nfsstatsv1.srv_errs = 0;
756 				nfsstatsv1.srvcache_inproghits = 0;
757 				nfsstatsv1.srvcache_idemdonehits = 0;
758 				nfsstatsv1.srvcache_nonidemdonehits = 0;
759 				nfsstatsv1.srvcache_misses = 0;
760 				nfsstatsv1.srvcache_tcppeak = 0;
761 				bzero(nfsstatsv1.srvrpccnt,
762 				    sizeof(nfsstatsv1.srvrpccnt));
763 				bzero(nfsstatsv1.cbrpccnt,
764 				    sizeof(nfsstatsv1.cbrpccnt));
765 			}
766 		}
767 		goto out;
768 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
769 		u_short sockport;
770 		struct nfsuserd_args nargs;
771 
772 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
773 			error = copyin(uap->argp, (caddr_t)&sockport,
774 			    sizeof (u_short));
775 			if (error == 0) {
776 				nargs.nuserd_family = AF_INET;
777 				nargs.nuserd_port = sockport;
778 			}
779 		} else {
780 			/*
781 			 * New nfsuserd_args structure, which indicates
782 			 * which IP version to use along with the port#.
783 			 */
784 			error = copyin(uap->argp, &nargs, sizeof(nargs));
785 		}
786 		if (!error)
787 			error = nfsrv_nfsuserdport(&nargs, p);
788 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
789 		nfsrv_nfsuserddelport();
790 		error = 0;
791 	}
792 
793 out:
794 	NFSEXITCODE(error);
795 	return (error);
796 }
797 
798 /*
799  * called by all three modevent routines, so that it gets things
800  * initialized soon enough.
801  */
802 void
803 newnfs_portinit(void)
804 {
805 	static int inited = 0;
806 
807 	if (inited)
808 		return;
809 	inited = 1;
810 	/* Initialize SMP locks used by both client and server. */
811 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
812 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
813 	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
814 }
815 
816 /*
817  * Determine if the file system supports NFSv4 ACLs.
818  * Return 1 if it does, 0 otherwise.
819  */
820 int
821 nfs_supportsnfsv4acls(struct vnode *vp)
822 {
823 	int error;
824 	long retval;
825 
826 	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
827 
828 	if (nfsrv_useacl == 0)
829 		return (0);
830 	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
831 	if (error == 0 && retval != 0)
832 		return (1);
833 	return (0);
834 }
835 
836 /*
837  * These are the first fields of all the context structures passed into
838  * nfs_pnfsio().
839  */
840 struct pnfsio {
841 	int		done;
842 	int		inprog;
843 	struct task	tsk;
844 };
845 
846 /*
847  * Do a mirror I/O on a pNFS thread.
848  */
849 int
850 nfs_pnfsio(task_fn_t *func, void *context)
851 {
852 	struct pnfsio *pio;
853 	int ret;
854 	static struct taskqueue *pnfsioq = NULL;
855 
856 	pio = (struct pnfsio *)context;
857 	if (pnfsioq == NULL) {
858 		if (nfs_pnfsiothreads == 0)
859 			return (EPERM);
860 		if (nfs_pnfsiothreads < 0)
861 			nfs_pnfsiothreads = mp_ncpus * 4;
862 		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
863 		    taskqueue_thread_enqueue, &pnfsioq);
864 		if (pnfsioq == NULL)
865 			return (ENOMEM);
866 		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
867 		    0, "pnfsiot");
868 		if (ret != 0) {
869 			taskqueue_free(pnfsioq);
870 			pnfsioq = NULL;
871 			return (ret);
872 		}
873 	}
874 	pio->inprog = 1;
875 	TASK_INIT(&pio->tsk, 0, func, context);
876 	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
877 	if (ret != 0)
878 		pio->inprog = 0;
879 	return (ret);
880 }
881 
882 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
883 
884 /*
885  * Called once to initialize data structures...
886  */
887 static int
888 nfscommon_modevent(module_t mod, int type, void *data)
889 {
890 	int error = 0;
891 	static int loaded = 0;
892 
893 	switch (type) {
894 	case MOD_LOAD:
895 		if (loaded)
896 			goto out;
897 		newnfs_portinit();
898 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
899 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
900 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
901 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
902 		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
903 		    MTX_DEF);
904 		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
905 		TAILQ_INIT(&nfsrv_devidhead);
906 		callout_init(&newnfsd_callout, 1);
907 		newnfs_init();
908 		nfsd_call_nfscommon = nfssvc_nfscommon;
909 		loaded = 1;
910 		break;
911 
912 	case MOD_UNLOAD:
913 		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != NOTRUNNING ||
914 		    nfs_numnfscbd != 0) {
915 			error = EBUSY;
916 			break;
917 		}
918 
919 		nfsd_call_nfscommon = NULL;
920 		callout_drain(&newnfsd_callout);
921 		/* Clean out the name<-->id cache. */
922 		nfsrv_cleanusergroup();
923 		/* and get rid of the mutexes */
924 		mtx_destroy(&nfs_nameid_mutex);
925 		mtx_destroy(&newnfsd_mtx);
926 		mtx_destroy(&nfs_state_mutex);
927 		mtx_destroy(&nfs_clstate_mutex);
928 		mtx_destroy(&nfs_sockl_mutex);
929 		mtx_destroy(&nfs_slock_mutex);
930 		mtx_destroy(&nfs_req_mutex);
931 		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
932 		mtx_destroy(&nfsrv_dslock_mtx);
933 		loaded = 0;
934 		break;
935 	default:
936 		error = EOPNOTSUPP;
937 		break;
938 	}
939 
940 out:
941 	NFSEXITCODE(error);
942 	return error;
943 }
944 static moduledata_t nfscommon_mod = {
945 	"nfscommon",
946 	nfscommon_modevent,
947 	NULL,
948 };
949 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
950 
951 /* So that loader and kldload(2) can find us, wherever we are.. */
952 MODULE_VERSION(nfscommon, 1);
953 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
954 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
955 
956