xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision 92b14858b44dc4b3b57154a10e9de1b39d791e41)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Rick Macklem at The University of Guelph.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  */
35 
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 /*
40  * Functions that need to be different for different versions of BSD
41  * kernel should be kept here, along with any global storage specific
42  * to this BSD variant.
43  */
44 #include <fs/nfs/nfsport.h>
45 #include <sys/smp.h>
46 #include <sys/sysctl.h>
47 #include <sys/taskqueue.h>
48 #include <rpc/rpc_com.h>
49 #include <vm/vm.h>
50 #include <vm/vm_object.h>
51 #include <vm/vm_page.h>
52 #include <vm/vm_param.h>
53 #include <vm/vm_map.h>
54 #include <vm/vm_kern.h>
55 #include <vm/vm_extern.h>
56 #include <vm/uma.h>
57 
58 extern int nfscl_ticks;
59 extern nfsuserd_state nfsrv_nfsuserd;
60 extern struct nfssockreq nfsrv_nfsuserdsock;
61 extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
62     struct thread *);
63 extern int nfsrv_useacl;
64 struct mount nfsv4root_mnt;
65 int newnfs_numnfsd = 0;
66 struct nfsstatsv1 nfsstatsv1;
67 int nfs_numnfscbd = 0;
68 int nfscl_debuglevel = 0;
69 char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
70 struct callout newnfsd_callout;
71 int nfsrv_lughashsize = 100;
72 struct mtx nfsrv_dslock_mtx;
73 struct nfsdevicehead nfsrv_devidhead;
74 volatile int nfsrv_devidcnt = 0;
75 void (*nfsd_call_servertimer)(void) = NULL;
76 void (*ncl_call_invalcaches)(struct vnode *) = NULL;
77 
78 int nfs_pnfsio(task_fn_t *, void *);
79 
80 static int nfs_realign_test;
81 static int nfs_realign_count;
82 static struct ext_nfsstats oldnfsstats;
83 static struct nfsstatsov1 nfsstatsov1;
84 
85 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
86     "NFS filesystem");
87 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
88     0, "Number of realign tests done");
89 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
90     0, "Number of mbuf realignments done");
91 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
92     nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
93     "NFSv4 callback addr for server to use");
94 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
95     0, "Debug level for NFS client");
96 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
97     0, "Size of hash tables for uid/name mapping");
98 int nfs_pnfsiothreads = -1;
99 SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsiothreads, CTLFLAG_RW, &nfs_pnfsiothreads,
100     0, "Number of pNFS mirror I/O threads");
101 
102 /*
103  * Defines for malloc
104  * (Here for FreeBSD, since they allocate storage.)
105  */
106 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
107 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
108 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
109     "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
110 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
111 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
112 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
113 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
114 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
115 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
116 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
117 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
118 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
119 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
120 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
121 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
122 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
123 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer");
124 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
125     "NFS directory offset data");
126 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
127     "NFS local lock rollback");
128 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
129 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
130 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
131 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
132 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
133 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
134 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
135 
136 /*
137  * Definition of mutex locks.
138  * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
139  * and assorted other nfsd structures.
140  */
141 struct mtx newnfsd_mtx;
142 struct mtx nfs_sockl_mutex;
143 struct mtx nfs_state_mutex;
144 struct mtx nfs_nameid_mutex;
145 struct mtx nfs_req_mutex;
146 struct mtx nfs_slock_mutex;
147 struct mtx nfs_clstate_mutex;
148 
149 /* local functions */
150 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
151 
152 #ifdef __NO_STRICT_ALIGNMENT
153 /*
154  * These architectures don't need re-alignment, so just return.
155  */
156 int
157 newnfs_realign(struct mbuf **pm, int how)
158 {
159 
160 	return (0);
161 }
162 #else	/* !__NO_STRICT_ALIGNMENT */
163 /*
164  *	newnfs_realign:
165  *
166  *	Check for badly aligned mbuf data and realign by copying the unaligned
167  *	portion of the data into a new mbuf chain and freeing the portions
168  *	of the old chain that were replaced.
169  *
170  *	We cannot simply realign the data within the existing mbuf chain
171  *	because the underlying buffers may contain other rpc commands and
172  *	we cannot afford to overwrite them.
173  *
174  *	We would prefer to avoid this situation entirely.  The situation does
175  *	not occur with NFS/UDP and is supposed to only occasionally occur
176  *	with TCP.  Use vfs.nfs.realign_count and realign_test to check this.
177  *
178  */
179 int
180 newnfs_realign(struct mbuf **pm, int how)
181 {
182 	struct mbuf *m, *n;
183 	int off, space;
184 
185 	++nfs_realign_test;
186 	while ((m = *pm) != NULL) {
187 		if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
188 			/*
189 			 * NB: we can't depend on m_pkthdr.len to help us
190 			 * decide what to do here.  May not be worth doing
191 			 * the m_length calculation as m_copyback will
192 			 * expand the mbuf chain below as needed.
193 			 */
194 			space = m_length(m, NULL);
195 			if (space >= MINCLSIZE) {
196 				/* NB: m_copyback handles space > MCLBYTES */
197 				n = m_getcl(how, MT_DATA, 0);
198 			} else
199 				n = m_get(how, MT_DATA);
200 			if (n == NULL)
201 				return (ENOMEM);
202 			/*
203 			 * Align the remainder of the mbuf chain.
204 			 */
205 			n->m_len = 0;
206 			off = 0;
207 			while (m != NULL) {
208 				m_copyback(n, off, m->m_len, mtod(m, caddr_t));
209 				off += m->m_len;
210 				m = m->m_next;
211 			}
212 			m_freem(*pm);
213 			*pm = n;
214 			++nfs_realign_count;
215 			break;
216 		}
217 		pm = &m->m_next;
218 	}
219 
220 	return (0);
221 }
222 #endif	/* __NO_STRICT_ALIGNMENT */
223 
224 #ifdef notdef
225 static void
226 nfsrv_object_create(struct vnode *vp, struct thread *td)
227 {
228 
229 	if (vp == NULL || vp->v_type != VREG)
230 		return;
231 	(void) vfs_object_create(vp, td, td->td_ucred);
232 }
233 #endif
234 
235 /*
236  * Look up a file name. Basically just initialize stuff and call namei().
237  */
238 int
239 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
240 {
241 	int error;
242 
243 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
244 	    p);
245 	error = namei(ndp);
246 	if (!error) {
247 		NDFREE(ndp, NDF_ONLY_PNBUF);
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  * Nfs timer routine
402  * Call the nfsd's timer function once/sec.
403  */
404 void
405 newnfs_timer(void *arg)
406 {
407 	static time_t lasttime = 0;
408 	/*
409 	 * Call the server timer, if set up.
410 	 * The argument indicates if it is the next second and therefore
411 	 * leases should be checked.
412 	 */
413 	if (lasttime != NFSD_MONOSEC) {
414 		lasttime = NFSD_MONOSEC;
415 		if (nfsd_call_servertimer != NULL)
416 			(*nfsd_call_servertimer)();
417 	}
418 	callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
419 }
420 
421 
422 /*
423  * Sleep for a short period of time unless errval == NFSERR_GRACE, where
424  * the sleep should be for 5 seconds.
425  * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
426  * an event that never gets a wakeup. Only return EINTR or 0.
427  */
428 int
429 nfs_catnap(int prio, int errval, const char *wmesg)
430 {
431 	static int non_event;
432 	int ret;
433 
434 	if (errval == NFSERR_GRACE)
435 		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
436 	else
437 		ret = tsleep(&non_event, prio, wmesg, 1);
438 	if (ret != EINTR)
439 		ret = 0;
440 	return (ret);
441 }
442 
443 /*
444  * Get referral. For now, just fail.
445  */
446 struct nfsreferral *
447 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
448 {
449 
450 	return (NULL);
451 }
452 
453 static int
454 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
455 {
456 	int error;
457 
458 	error = nfssvc_call(td, uap, td->td_ucred);
459 	NFSEXITCODE(error);
460 	return (error);
461 }
462 
463 static int
464 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
465 {
466 	int error = EINVAL, i, j;
467 	struct nfsd_idargs nid;
468 	struct nfsd_oidargs onid;
469 	struct {
470 		int vers;	/* Just the first field of nfsstats. */
471 	} nfsstatver;
472 
473 	if (uap->flag & NFSSVC_IDNAME) {
474 		if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
475 			error = copyin(uap->argp, &nid, sizeof(nid));
476 		else {
477 			error = copyin(uap->argp, &onid, sizeof(onid));
478 			if (error == 0) {
479 				nid.nid_flag = onid.nid_flag;
480 				nid.nid_uid = onid.nid_uid;
481 				nid.nid_gid = onid.nid_gid;
482 				nid.nid_usermax = onid.nid_usermax;
483 				nid.nid_usertimeout = onid.nid_usertimeout;
484 				nid.nid_name = onid.nid_name;
485 				nid.nid_namelen = onid.nid_namelen;
486 				nid.nid_ngroup = 0;
487 				nid.nid_grps = NULL;
488 			}
489 		}
490 		if (error)
491 			goto out;
492 		error = nfssvc_idname(&nid);
493 		goto out;
494 	} else if (uap->flag & NFSSVC_GETSTATS) {
495 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
496 			/* Copy fields to the old ext_nfsstat structure. */
497 			oldnfsstats.attrcache_hits =
498 			    nfsstatsv1.attrcache_hits;
499 			oldnfsstats.attrcache_misses =
500 			    nfsstatsv1.attrcache_misses;
501 			oldnfsstats.lookupcache_hits =
502 			    nfsstatsv1.lookupcache_hits;
503 			oldnfsstats.lookupcache_misses =
504 			    nfsstatsv1.lookupcache_misses;
505 			oldnfsstats.direofcache_hits =
506 			    nfsstatsv1.direofcache_hits;
507 			oldnfsstats.direofcache_misses =
508 			    nfsstatsv1.direofcache_misses;
509 			oldnfsstats.accesscache_hits =
510 			    nfsstatsv1.accesscache_hits;
511 			oldnfsstats.accesscache_misses =
512 			    nfsstatsv1.accesscache_misses;
513 			oldnfsstats.biocache_reads =
514 			    nfsstatsv1.biocache_reads;
515 			oldnfsstats.read_bios =
516 			    nfsstatsv1.read_bios;
517 			oldnfsstats.read_physios =
518 			    nfsstatsv1.read_physios;
519 			oldnfsstats.biocache_writes =
520 			    nfsstatsv1.biocache_writes;
521 			oldnfsstats.write_bios =
522 			    nfsstatsv1.write_bios;
523 			oldnfsstats.write_physios =
524 			    nfsstatsv1.write_physios;
525 			oldnfsstats.biocache_readlinks =
526 			    nfsstatsv1.biocache_readlinks;
527 			oldnfsstats.readlink_bios =
528 			    nfsstatsv1.readlink_bios;
529 			oldnfsstats.biocache_readdirs =
530 			    nfsstatsv1.biocache_readdirs;
531 			oldnfsstats.readdir_bios =
532 			    nfsstatsv1.readdir_bios;
533 			for (i = 0; i < NFSV4_NPROCS; i++)
534 				oldnfsstats.rpccnt[i] = nfsstatsv1.rpccnt[i];
535 			oldnfsstats.rpcretries = nfsstatsv1.rpcretries;
536 			for (i = 0; i < NFSV4OP_NOPS; i++)
537 				oldnfsstats.srvrpccnt[i] =
538 				    nfsstatsv1.srvrpccnt[i];
539 			for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
540 			    i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
541 				oldnfsstats.srvrpccnt[j] =
542 				    nfsstatsv1.srvrpccnt[i];
543 			oldnfsstats.srvrpc_errs = nfsstatsv1.srvrpc_errs;
544 			oldnfsstats.srv_errs = nfsstatsv1.srv_errs;
545 			oldnfsstats.rpcrequests = nfsstatsv1.rpcrequests;
546 			oldnfsstats.rpctimeouts = nfsstatsv1.rpctimeouts;
547 			oldnfsstats.rpcunexpected = nfsstatsv1.rpcunexpected;
548 			oldnfsstats.rpcinvalid = nfsstatsv1.rpcinvalid;
549 			oldnfsstats.srvcache_inproghits =
550 			    nfsstatsv1.srvcache_inproghits;
551 			oldnfsstats.srvcache_idemdonehits =
552 			    nfsstatsv1.srvcache_idemdonehits;
553 			oldnfsstats.srvcache_nonidemdonehits =
554 			    nfsstatsv1.srvcache_nonidemdonehits;
555 			oldnfsstats.srvcache_misses =
556 			    nfsstatsv1.srvcache_misses;
557 			oldnfsstats.srvcache_tcppeak =
558 			    nfsstatsv1.srvcache_tcppeak;
559 			oldnfsstats.srvcache_size = nfsstatsv1.srvcache_size;
560 			oldnfsstats.srvclients = nfsstatsv1.srvclients;
561 			oldnfsstats.srvopenowners = nfsstatsv1.srvopenowners;
562 			oldnfsstats.srvopens = nfsstatsv1.srvopens;
563 			oldnfsstats.srvlockowners = nfsstatsv1.srvlockowners;
564 			oldnfsstats.srvlocks = nfsstatsv1.srvlocks;
565 			oldnfsstats.srvdelegates = nfsstatsv1.srvdelegates;
566 			for (i = 0; i < NFSV4OP_CBNOPS; i++)
567 				oldnfsstats.cbrpccnt[i] =
568 				    nfsstatsv1.cbrpccnt[i];
569 			oldnfsstats.clopenowners = nfsstatsv1.clopenowners;
570 			oldnfsstats.clopens = nfsstatsv1.clopens;
571 			oldnfsstats.cllockowners = nfsstatsv1.cllockowners;
572 			oldnfsstats.cllocks = nfsstatsv1.cllocks;
573 			oldnfsstats.cldelegates = nfsstatsv1.cldelegates;
574 			oldnfsstats.cllocalopenowners =
575 			    nfsstatsv1.cllocalopenowners;
576 			oldnfsstats.cllocalopens = nfsstatsv1.cllocalopens;
577 			oldnfsstats.cllocallockowners =
578 			    nfsstatsv1.cllocallockowners;
579 			oldnfsstats.cllocallocks = nfsstatsv1.cllocallocks;
580 			error = copyout(&oldnfsstats, uap->argp,
581 			    sizeof (oldnfsstats));
582 		} else {
583 			error = copyin(uap->argp, &nfsstatver,
584 			    sizeof(nfsstatver));
585 			if (error == 0) {
586 				if (nfsstatver.vers == NFSSTATS_OV1) {
587 					/* Copy nfsstatsv1 to nfsstatsov1. */
588 					nfsstatsov1.attrcache_hits =
589 					    nfsstatsv1.attrcache_hits;
590 					nfsstatsov1.attrcache_misses =
591 					    nfsstatsv1.attrcache_misses;
592 					nfsstatsov1.lookupcache_hits =
593 					    nfsstatsv1.lookupcache_hits;
594 					nfsstatsov1.lookupcache_misses =
595 					    nfsstatsv1.lookupcache_misses;
596 					nfsstatsov1.direofcache_hits =
597 					    nfsstatsv1.direofcache_hits;
598 					nfsstatsov1.direofcache_misses =
599 					    nfsstatsv1.direofcache_misses;
600 					nfsstatsov1.accesscache_hits =
601 					    nfsstatsv1.accesscache_hits;
602 					nfsstatsov1.accesscache_misses =
603 					    nfsstatsv1.accesscache_misses;
604 					nfsstatsov1.biocache_reads =
605 					    nfsstatsv1.biocache_reads;
606 					nfsstatsov1.read_bios =
607 					    nfsstatsv1.read_bios;
608 					nfsstatsov1.read_physios =
609 					    nfsstatsv1.read_physios;
610 					nfsstatsov1.biocache_writes =
611 					    nfsstatsv1.biocache_writes;
612 					nfsstatsov1.write_bios =
613 					    nfsstatsv1.write_bios;
614 					nfsstatsov1.write_physios =
615 					    nfsstatsv1.write_physios;
616 					nfsstatsov1.biocache_readlinks =
617 					    nfsstatsv1.biocache_readlinks;
618 					nfsstatsov1.readlink_bios =
619 					    nfsstatsv1.readlink_bios;
620 					nfsstatsov1.biocache_readdirs =
621 					    nfsstatsv1.biocache_readdirs;
622 					nfsstatsov1.readdir_bios =
623 					    nfsstatsv1.readdir_bios;
624 					for (i = 0; i < NFSV42_NPROCS; i++)
625 						nfsstatsov1.rpccnt[i] =
626 						    nfsstatsv1.rpccnt[i];
627 					nfsstatsov1.rpcretries =
628 					    nfsstatsv1.rpcretries;
629 					for (i = 0; i < NFSV42_PURENOPS; i++)
630 						nfsstatsov1.srvrpccnt[i] =
631 						    nfsstatsv1.srvrpccnt[i];
632 					for (i = NFSV42_NOPS,
633 					     j = NFSV42_PURENOPS;
634 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
635 					     i++, j++)
636 						nfsstatsov1.srvrpccnt[j] =
637 						    nfsstatsv1.srvrpccnt[i];
638 					nfsstatsov1.srvrpc_errs =
639 					    nfsstatsv1.srvrpc_errs;
640 					nfsstatsov1.srv_errs =
641 					    nfsstatsv1.srv_errs;
642 					nfsstatsov1.rpcrequests =
643 					    nfsstatsv1.rpcrequests;
644 					nfsstatsov1.rpctimeouts =
645 					    nfsstatsv1.rpctimeouts;
646 					nfsstatsov1.rpcunexpected =
647 					    nfsstatsv1.rpcunexpected;
648 					nfsstatsov1.rpcinvalid =
649 					    nfsstatsv1.rpcinvalid;
650 					nfsstatsov1.srvcache_inproghits =
651 					    nfsstatsv1.srvcache_inproghits;
652 					nfsstatsov1.srvcache_idemdonehits =
653 					    nfsstatsv1.srvcache_idemdonehits;
654 					nfsstatsov1.srvcache_nonidemdonehits =
655 					    nfsstatsv1.srvcache_nonidemdonehits;
656 					nfsstatsov1.srvcache_misses =
657 					    nfsstatsv1.srvcache_misses;
658 					nfsstatsov1.srvcache_tcppeak =
659 					    nfsstatsv1.srvcache_tcppeak;
660 					nfsstatsov1.srvcache_size =
661 					    nfsstatsv1.srvcache_size;
662 					nfsstatsov1.srvclients =
663 					    nfsstatsv1.srvclients;
664 					nfsstatsov1.srvopenowners =
665 					    nfsstatsv1.srvopenowners;
666 					nfsstatsov1.srvopens =
667 					    nfsstatsv1.srvopens;
668 					nfsstatsov1.srvlockowners =
669 					    nfsstatsv1.srvlockowners;
670 					nfsstatsov1.srvlocks =
671 					    nfsstatsv1.srvlocks;
672 					nfsstatsov1.srvdelegates =
673 					    nfsstatsv1.srvdelegates;
674 					for (i = 0; i < NFSV42_CBNOPS; i++)
675 						nfsstatsov1.cbrpccnt[i] =
676 						    nfsstatsv1.cbrpccnt[i];
677 					nfsstatsov1.clopenowners =
678 					    nfsstatsv1.clopenowners;
679 					nfsstatsov1.clopens =
680 					    nfsstatsv1.clopens;
681 					nfsstatsov1.cllockowners =
682 					    nfsstatsv1.cllockowners;
683 					nfsstatsov1.cllocks =
684 					    nfsstatsv1.cllocks;
685 					nfsstatsov1.cldelegates =
686 					    nfsstatsv1.cldelegates;
687 					nfsstatsov1.cllocalopenowners =
688 					    nfsstatsv1.cllocalopenowners;
689 					nfsstatsov1.cllocalopens =
690 					    nfsstatsv1.cllocalopens;
691 					nfsstatsov1.cllocallockowners =
692 					    nfsstatsv1.cllocallockowners;
693 					nfsstatsov1.cllocallocks =
694 					    nfsstatsv1.cllocallocks;
695 					nfsstatsov1.srvstartcnt =
696 					    nfsstatsv1.srvstartcnt;
697 					nfsstatsov1.srvdonecnt =
698 					    nfsstatsv1.srvdonecnt;
699 					for (i = NFSV42_NOPS,
700 					     j = NFSV42_PURENOPS;
701 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
702 					     i++, j++) {
703 						nfsstatsov1.srvbytes[j] =
704 						    nfsstatsv1.srvbytes[i];
705 						nfsstatsov1.srvops[j] =
706 						    nfsstatsv1.srvops[i];
707 						nfsstatsov1.srvduration[j] =
708 						    nfsstatsv1.srvduration[i];
709 					}
710 					nfsstatsov1.busyfrom =
711 					    nfsstatsv1.busyfrom;
712 					nfsstatsov1.busyfrom =
713 					    nfsstatsv1.busyfrom;
714 					error = copyout(&nfsstatsov1, uap->argp,
715 					    sizeof(nfsstatsov1));
716 				} else if (nfsstatver.vers != NFSSTATS_V1)
717 					error = EPERM;
718 				else
719 					error = copyout(&nfsstatsv1, uap->argp,
720 					    sizeof(nfsstatsv1));
721 			}
722 		}
723 		if (error == 0) {
724 			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
725 				nfsstatsv1.attrcache_hits = 0;
726 				nfsstatsv1.attrcache_misses = 0;
727 				nfsstatsv1.lookupcache_hits = 0;
728 				nfsstatsv1.lookupcache_misses = 0;
729 				nfsstatsv1.direofcache_hits = 0;
730 				nfsstatsv1.direofcache_misses = 0;
731 				nfsstatsv1.accesscache_hits = 0;
732 				nfsstatsv1.accesscache_misses = 0;
733 				nfsstatsv1.biocache_reads = 0;
734 				nfsstatsv1.read_bios = 0;
735 				nfsstatsv1.read_physios = 0;
736 				nfsstatsv1.biocache_writes = 0;
737 				nfsstatsv1.write_bios = 0;
738 				nfsstatsv1.write_physios = 0;
739 				nfsstatsv1.biocache_readlinks = 0;
740 				nfsstatsv1.readlink_bios = 0;
741 				nfsstatsv1.biocache_readdirs = 0;
742 				nfsstatsv1.readdir_bios = 0;
743 				nfsstatsv1.rpcretries = 0;
744 				nfsstatsv1.rpcrequests = 0;
745 				nfsstatsv1.rpctimeouts = 0;
746 				nfsstatsv1.rpcunexpected = 0;
747 				nfsstatsv1.rpcinvalid = 0;
748 				bzero(nfsstatsv1.rpccnt,
749 				    sizeof(nfsstatsv1.rpccnt));
750 			}
751 			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
752 				nfsstatsv1.srvrpc_errs = 0;
753 				nfsstatsv1.srv_errs = 0;
754 				nfsstatsv1.srvcache_inproghits = 0;
755 				nfsstatsv1.srvcache_idemdonehits = 0;
756 				nfsstatsv1.srvcache_nonidemdonehits = 0;
757 				nfsstatsv1.srvcache_misses = 0;
758 				nfsstatsv1.srvcache_tcppeak = 0;
759 				bzero(nfsstatsv1.srvrpccnt,
760 				    sizeof(nfsstatsv1.srvrpccnt));
761 				bzero(nfsstatsv1.cbrpccnt,
762 				    sizeof(nfsstatsv1.cbrpccnt));
763 			}
764 		}
765 		goto out;
766 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
767 		u_short sockport;
768 		struct nfsuserd_args nargs;
769 
770 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
771 			error = copyin(uap->argp, (caddr_t)&sockport,
772 			    sizeof (u_short));
773 			if (error == 0) {
774 				nargs.nuserd_family = AF_INET;
775 				nargs.nuserd_port = sockport;
776 			}
777 		} else {
778 			/*
779 			 * New nfsuserd_args structure, which indicates
780 			 * which IP version to use along with the port#.
781 			 */
782 			error = copyin(uap->argp, &nargs, sizeof(nargs));
783 		}
784 		if (!error)
785 			error = nfsrv_nfsuserdport(&nargs, p);
786 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
787 		nfsrv_nfsuserddelport();
788 		error = 0;
789 	}
790 
791 out:
792 	NFSEXITCODE(error);
793 	return (error);
794 }
795 
796 /*
797  * called by all three modevent routines, so that it gets things
798  * initialized soon enough.
799  */
800 void
801 newnfs_portinit(void)
802 {
803 	static int inited = 0;
804 
805 	if (inited)
806 		return;
807 	inited = 1;
808 	/* Initialize SMP locks used by both client and server. */
809 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
810 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
811 	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
812 }
813 
814 /*
815  * Determine if the file system supports NFSv4 ACLs.
816  * Return 1 if it does, 0 otherwise.
817  */
818 int
819 nfs_supportsnfsv4acls(struct vnode *vp)
820 {
821 	int error;
822 	long retval;
823 
824 	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
825 
826 	if (nfsrv_useacl == 0)
827 		return (0);
828 	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
829 	if (error == 0 && retval != 0)
830 		return (1);
831 	return (0);
832 }
833 
834 /*
835  * These are the first fields of all the context structures passed into
836  * nfs_pnfsio().
837  */
838 struct pnfsio {
839 	int		done;
840 	int		inprog;
841 	struct task	tsk;
842 };
843 
844 /*
845  * Do a mirror I/O on a pNFS thread.
846  */
847 int
848 nfs_pnfsio(task_fn_t *func, void *context)
849 {
850 	struct pnfsio *pio;
851 	int ret;
852 	static struct taskqueue *pnfsioq = NULL;
853 
854 	pio = (struct pnfsio *)context;
855 	if (pnfsioq == NULL) {
856 		if (nfs_pnfsiothreads == 0)
857 			return (EPERM);
858 		if (nfs_pnfsiothreads < 0)
859 			nfs_pnfsiothreads = mp_ncpus * 4;
860 		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
861 		    taskqueue_thread_enqueue, &pnfsioq);
862 		if (pnfsioq == NULL)
863 			return (ENOMEM);
864 		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
865 		    0, "pnfsiot");
866 		if (ret != 0) {
867 			taskqueue_free(pnfsioq);
868 			pnfsioq = NULL;
869 			return (ret);
870 		}
871 	}
872 	pio->inprog = 1;
873 	TASK_INIT(&pio->tsk, 0, func, context);
874 	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
875 	if (ret != 0)
876 		pio->inprog = 0;
877 	return (ret);
878 }
879 
880 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
881 
882 /*
883  * Called once to initialize data structures...
884  */
885 static int
886 nfscommon_modevent(module_t mod, int type, void *data)
887 {
888 	int error = 0;
889 	static int loaded = 0;
890 
891 	switch (type) {
892 	case MOD_LOAD:
893 		if (loaded)
894 			goto out;
895 		newnfs_portinit();
896 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
897 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
898 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
899 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
900 		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
901 		    MTX_DEF);
902 		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
903 		TAILQ_INIT(&nfsrv_devidhead);
904 		callout_init(&newnfsd_callout, 1);
905 		newnfs_init();
906 		nfsd_call_nfscommon = nfssvc_nfscommon;
907 		loaded = 1;
908 		break;
909 
910 	case MOD_UNLOAD:
911 		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != NOTRUNNING ||
912 		    nfs_numnfscbd != 0) {
913 			error = EBUSY;
914 			break;
915 		}
916 
917 		nfsd_call_nfscommon = NULL;
918 		callout_drain(&newnfsd_callout);
919 		/* Clean out the name<-->id cache. */
920 		nfsrv_cleanusergroup();
921 		/* and get rid of the mutexes */
922 		mtx_destroy(&nfs_nameid_mutex);
923 		mtx_destroy(&newnfsd_mtx);
924 		mtx_destroy(&nfs_state_mutex);
925 		mtx_destroy(&nfs_clstate_mutex);
926 		mtx_destroy(&nfs_sockl_mutex);
927 		mtx_destroy(&nfs_slock_mutex);
928 		mtx_destroy(&nfs_req_mutex);
929 		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
930 		mtx_destroy(&nfsrv_dslock_mtx);
931 		loaded = 0;
932 		break;
933 	default:
934 		error = EOPNOTSUPP;
935 		break;
936 	}
937 
938 out:
939 	NFSEXITCODE(error);
940 	return error;
941 }
942 static moduledata_t nfscommon_mod = {
943 	"nfscommon",
944 	nfscommon_modevent,
945 	NULL,
946 };
947 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
948 
949 /* So that loader and kldload(2) can find us, wherever we are.. */
950 MODULE_VERSION(nfscommon, 1);
951 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
952 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
953 
954