xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision 47ef2a131091508e049ab10cad7f91a3c1342cd9)
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 /*
38  * Functions that need to be different for different versions of BSD
39  * kernel should be kept here, along with any global storage specific
40  * to this BSD variant.
41  */
42 #include <fs/nfs/nfsport.h>
43 #include <sys/smp.h>
44 #include <sys/sysctl.h>
45 #include <sys/taskqueue.h>
46 #include <rpc/rpc_com.h>
47 #include <vm/vm.h>
48 #include <vm/vm_object.h>
49 #include <vm/vm_page.h>
50 #include <vm/vm_param.h>
51 #include <vm/vm_map.h>
52 #include <vm/vm_kern.h>
53 #include <vm/vm_extern.h>
54 #include <vm/uma.h>
55 
56 extern int nfscl_ticks;
57 extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
58     struct thread *);
59 extern int nfsrv_useacl;
60 int newnfs_numnfsd = 0;
61 struct nfsstatsv1 nfsstatsv1;
62 int nfs_numnfscbd = 0;
63 int nfscl_debuglevel = 0;
64 char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
65 int nfsrv_lughashsize = 100;
66 struct mtx nfsrv_dslock_mtx;
67 struct nfsdevicehead nfsrv_devidhead;
68 volatile int nfsrv_devidcnt = 0;
69 void (*ncl_call_invalcaches)(struct vnode *) = NULL;
70 vop_advlock_t *nfs_advlock_p = NULL;
71 vop_reclaim_t *nfs_reclaim_p = NULL;
72 uint32_t nfs_srvmaxio = NFS_SRVMAXIO;
73 
74 NFSD_VNET_DEFINE(struct nfsstatsv1 *, nfsstatsv1_p);
75 
76 NFSD_VNET_DECLARE(struct nfssockreq, nfsrv_nfsuserdsock);
77 NFSD_VNET_DECLARE(nfsuserd_state, nfsrv_nfsuserd);
78 NFSD_VNET_DECLARE(gid_t, nfsrv_defaultgid);
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_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_fallback(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups,
263 	    NFSD_VNET(nfsrv_defaultgid));
264 }
265 
266 /*
267  * Map args from nfsmsleep() to msleep().
268  */
269 int
270 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
271     struct timespec *ts)
272 {
273 	u_int64_t nsecval;
274 	int error, timeo;
275 
276 	if (ts) {
277 		timeo = hz * ts->tv_sec;
278 		nsecval = (u_int64_t)ts->tv_nsec;
279 		nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
280 		    1000000000;
281 		timeo += (int)nsecval;
282 	} else {
283 		timeo = 0;
284 	}
285 	error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
286 	return (error);
287 }
288 
289 /*
290  * Get the file system info for the server. For now, just assume FFS.
291  */
292 void
293 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
294 {
295 	int pref;
296 
297 	/*
298 	 * XXX
299 	 * There should be file system VFS OP(s) to get this information.
300 	 * For now, assume ufs.
301 	 */
302 	if (isdgram)
303 		pref = NFS_MAXDGRAMDATA;
304 	else
305 		pref = nfs_srvmaxio;
306 	sip->fs_rtmax = nfs_srvmaxio;
307 	sip->fs_rtpref = pref;
308 	sip->fs_rtmult = NFS_FABLKSIZE;
309 	sip->fs_wtmax = nfs_srvmaxio;
310 	sip->fs_wtpref = pref;
311 	sip->fs_wtmult = NFS_FABLKSIZE;
312 	sip->fs_dtpref = pref;
313 	sip->fs_maxfilesize = 0xffffffffffffffffull;
314 	sip->fs_timedelta.tv_sec = 0;
315 	sip->fs_timedelta.tv_nsec = 1;
316 	sip->fs_properties = (NFSV3FSINFO_LINK |
317 	    NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
318 	    NFSV3FSINFO_CANSETTIME);
319 }
320 
321 /*
322  * Do the pathconf vnode op.
323  */
324 int
325 nfsvno_pathconf(struct vnode *vp, int flag, long *retf,
326     struct ucred *cred, struct thread *p)
327 {
328 	int error;
329 
330 	error = VOP_PATHCONF(vp, flag, retf);
331 	if (error == EOPNOTSUPP || error == EINVAL) {
332 		/*
333 		 * Some file systems return EINVAL for name arguments not
334 		 * supported and some return EOPNOTSUPP for this case.
335 		 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
336 		 * just fake them.
337 		 */
338 		switch (flag) {
339 		case _PC_LINK_MAX:
340 			*retf = NFS_LINK_MAX;
341 			break;
342 		case _PC_NAME_MAX:
343 			*retf = NAME_MAX;
344 			break;
345 		case _PC_CHOWN_RESTRICTED:
346 			*retf = 1;
347 			break;
348 		case _PC_NO_TRUNC:
349 			*retf = 1;
350 			break;
351 		default:
352 			/*
353 			 * Only happens if a _PC_xxx is added to the server,
354 			 * but this isn't updated.
355 			 */
356 			*retf = 0;
357 			printf("nfsrvd pathconf flag=%d not supp\n", flag);
358 		}
359 		error = 0;
360 	}
361 	NFSEXITCODE(error);
362 	return (error);
363 }
364 
365 /* Fake nfsrv_atroot. Just return 0 */
366 int
367 nfsrv_atroot(struct vnode *vp, uint64_t *retp)
368 {
369 
370 	return (0);
371 }
372 
373 /*
374  * Set the credentials to refer to root.
375  * If only the various BSDen could agree on whether cr_gid is a separate
376  * field or cr_groups[0]...
377  */
378 void
379 newnfs_setroot(struct ucred *cred)
380 {
381 
382 	cred->cr_uid = 0;
383 	cred->cr_groups[0] = 0;
384 	cred->cr_ngroups = 1;
385 }
386 
387 /*
388  * Get the client credential. Used for Renew and recovery.
389  */
390 struct ucred *
391 newnfs_getcred(void)
392 {
393 	struct ucred *cred;
394 	struct thread *td = curthread;
395 
396 	cred = crdup(td->td_ucred);
397 	newnfs_setroot(cred);
398 	return (cred);
399 }
400 
401 /*
402  * Sleep for a short period of time unless errval == NFSERR_GRACE, where
403  * the sleep should be for 5 seconds.
404  * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
405  * an event that never gets a wakeup. Only return EINTR or 0.
406  */
407 int
408 nfs_catnap(int prio, int errval, const char *wmesg)
409 {
410 	static int non_event;
411 	int ret;
412 
413 	if (errval == NFSERR_GRACE)
414 		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
415 	else
416 		ret = tsleep(&non_event, prio, wmesg, 1);
417 	if (ret != EINTR)
418 		ret = 0;
419 	return (ret);
420 }
421 
422 /*
423  * Get referral. For now, just fail.
424  */
425 struct nfsreferral *
426 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
427 {
428 
429 	return (NULL);
430 }
431 
432 static int
433 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
434 {
435 	int error;
436 
437 	NFSD_CURVNET_SET(NFSD_TD_TO_VNET(td));
438 	error = nfssvc_call(td, uap, td->td_ucred);
439 	NFSD_CURVNET_RESTORE();
440 	NFSEXITCODE(error);
441 	return (error);
442 }
443 
444 static int
445 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
446 {
447 	int error = EINVAL, i, j;
448 	struct nfsd_idargs nid;
449 	struct nfsd_oidargs onid;
450 	struct {
451 		int vers;	/* Just the first field of nfsstats. */
452 	} nfsstatver;
453 
454 	if (uap->flag & NFSSVC_IDNAME) {
455 		if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
456 			error = copyin(uap->argp, &nid, sizeof(nid));
457 		else {
458 			error = copyin(uap->argp, &onid, sizeof(onid));
459 			if (error == 0) {
460 				nid.nid_flag = onid.nid_flag;
461 				nid.nid_uid = onid.nid_uid;
462 				nid.nid_gid = onid.nid_gid;
463 				nid.nid_usermax = onid.nid_usermax;
464 				nid.nid_usertimeout = onid.nid_usertimeout;
465 				nid.nid_name = onid.nid_name;
466 				nid.nid_namelen = onid.nid_namelen;
467 				nid.nid_ngroup = 0;
468 				nid.nid_grps = NULL;
469 			}
470 		}
471 		if (error)
472 			goto out;
473 		error = nfssvc_idname(&nid);
474 		goto out;
475 	} else if (uap->flag & NFSSVC_GETSTATS) {
476 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
477 			/* Copy fields to the old ext_nfsstat structure. */
478 			oldnfsstats.attrcache_hits =
479 			    NFSD_VNET(nfsstatsv1_p)->attrcache_hits;
480 			oldnfsstats.attrcache_misses =
481 			    NFSD_VNET(nfsstatsv1_p)->attrcache_misses;
482 			oldnfsstats.lookupcache_hits =
483 			    NFSD_VNET(nfsstatsv1_p)->lookupcache_hits;
484 			oldnfsstats.lookupcache_misses =
485 			    NFSD_VNET(nfsstatsv1_p)->lookupcache_misses;
486 			oldnfsstats.direofcache_hits =
487 			    NFSD_VNET(nfsstatsv1_p)->direofcache_hits;
488 			oldnfsstats.direofcache_misses =
489 			    NFSD_VNET(nfsstatsv1_p)->direofcache_misses;
490 			oldnfsstats.accesscache_hits =
491 			    NFSD_VNET(nfsstatsv1_p)->accesscache_hits;
492 			oldnfsstats.accesscache_misses =
493 			    NFSD_VNET(nfsstatsv1_p)->accesscache_misses;
494 			oldnfsstats.biocache_reads =
495 			    NFSD_VNET(nfsstatsv1_p)->biocache_reads;
496 			oldnfsstats.read_bios =
497 			    NFSD_VNET(nfsstatsv1_p)->read_bios;
498 			oldnfsstats.read_physios =
499 			    NFSD_VNET(nfsstatsv1_p)->read_physios;
500 			oldnfsstats.biocache_writes =
501 			    NFSD_VNET(nfsstatsv1_p)->biocache_writes;
502 			oldnfsstats.write_bios =
503 			    NFSD_VNET(nfsstatsv1_p)->write_bios;
504 			oldnfsstats.write_physios =
505 			    NFSD_VNET(nfsstatsv1_p)->write_physios;
506 			oldnfsstats.biocache_readlinks =
507 			    NFSD_VNET(nfsstatsv1_p)->biocache_readlinks;
508 			oldnfsstats.readlink_bios =
509 			    NFSD_VNET(nfsstatsv1_p)->readlink_bios;
510 			oldnfsstats.biocache_readdirs =
511 			    NFSD_VNET(nfsstatsv1_p)->biocache_readdirs;
512 			oldnfsstats.readdir_bios =
513 			    NFSD_VNET(nfsstatsv1_p)->readdir_bios;
514 			for (i = 0; i < NFSV4_NPROCS; i++)
515 				oldnfsstats.rpccnt[i] =
516 				    NFSD_VNET(nfsstatsv1_p)->rpccnt[i];
517 			oldnfsstats.rpcretries =
518 			    NFSD_VNET(nfsstatsv1_p)->rpcretries;
519 			for (i = 0; i < NFSV4OP_NOPS; i++)
520 				oldnfsstats.srvrpccnt[i] =
521 				    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
522 			for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
523 			    i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
524 				oldnfsstats.srvrpccnt[j] =
525 				    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
526 			oldnfsstats.reserved_0 = 0;
527 			oldnfsstats.reserved_1 = 0;
528 			oldnfsstats.rpcrequests =
529 			    NFSD_VNET(nfsstatsv1_p)->rpcrequests;
530 			oldnfsstats.rpctimeouts =
531 			    NFSD_VNET(nfsstatsv1_p)->rpctimeouts;
532 			oldnfsstats.rpcunexpected =
533 			    NFSD_VNET(nfsstatsv1_p)->rpcunexpected;
534 			oldnfsstats.rpcinvalid =
535 			    NFSD_VNET(nfsstatsv1_p)->rpcinvalid;
536 			oldnfsstats.srvcache_inproghits =
537 			    NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits;
538 			oldnfsstats.reserved_2 = 0;
539 			oldnfsstats.srvcache_nonidemdonehits =
540 			    NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
541 			oldnfsstats.srvcache_misses =
542 			    NFSD_VNET(nfsstatsv1_p)->srvcache_misses;
543 			oldnfsstats.srvcache_tcppeak =
544 			    NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak;
545 			oldnfsstats.srvcache_size =
546 			    NFSD_VNET(nfsstatsv1_p)->srvcache_size;
547 			oldnfsstats.srvclients =
548 			    NFSD_VNET(nfsstatsv1_p)->srvclients;
549 			oldnfsstats.srvopenowners =
550 			    NFSD_VNET(nfsstatsv1_p)->srvopenowners;
551 			oldnfsstats.srvopens =
552 			    NFSD_VNET(nfsstatsv1_p)->srvopens;
553 			oldnfsstats.srvlockowners =
554 			    NFSD_VNET(nfsstatsv1_p)->srvlockowners;
555 			oldnfsstats.srvlocks =
556 			    NFSD_VNET(nfsstatsv1_p)->srvlocks;
557 			oldnfsstats.srvdelegates =
558 			    NFSD_VNET(nfsstatsv1_p)->srvdelegates;
559 			for (i = 0; i < NFSV4OP_CBNOPS; i++)
560 				oldnfsstats.cbrpccnt[i] =
561 				    NFSD_VNET(nfsstatsv1_p)->cbrpccnt[i];
562 			oldnfsstats.clopenowners =
563 			    NFSD_VNET(nfsstatsv1_p)->clopenowners;
564 			oldnfsstats.clopens = NFSD_VNET(nfsstatsv1_p)->clopens;
565 			oldnfsstats.cllockowners =
566 			    NFSD_VNET(nfsstatsv1_p)->cllockowners;
567 			oldnfsstats.cllocks = NFSD_VNET(nfsstatsv1_p)->cllocks;
568 			oldnfsstats.cldelegates =
569 			    NFSD_VNET(nfsstatsv1_p)->cldelegates;
570 			oldnfsstats.cllocalopenowners =
571 			    NFSD_VNET(nfsstatsv1_p)->cllocalopenowners;
572 			oldnfsstats.cllocalopens =
573 			    NFSD_VNET(nfsstatsv1_p)->cllocalopens;
574 			oldnfsstats.cllocallockowners =
575 			    NFSD_VNET(nfsstatsv1_p)->cllocallockowners;
576 			oldnfsstats.cllocallocks =
577 			    NFSD_VNET(nfsstatsv1_p)->cllocallocks;
578 			error = copyout(&oldnfsstats, uap->argp,
579 			    sizeof (oldnfsstats));
580 		} else {
581 			error = copyin(uap->argp, &nfsstatver,
582 			    sizeof(nfsstatver));
583 			if (error == 0) {
584 				if (nfsstatver.vers == NFSSTATS_OV1) {
585 					/* Copy nfsstatsv1 to nfsstatsov1. */
586 					nfsstatsov1.attrcache_hits =
587 					    NFSD_VNET(nfsstatsv1_p)->attrcache_hits;
588 					nfsstatsov1.attrcache_misses =
589 					    NFSD_VNET(nfsstatsv1_p)->attrcache_misses;
590 					nfsstatsov1.lookupcache_hits =
591 					    NFSD_VNET(nfsstatsv1_p)->lookupcache_hits;
592 					nfsstatsov1.lookupcache_misses =
593 					    NFSD_VNET(nfsstatsv1_p)->lookupcache_misses;
594 					nfsstatsov1.direofcache_hits =
595 					    NFSD_VNET(nfsstatsv1_p)->direofcache_hits;
596 					nfsstatsov1.direofcache_misses =
597 					    NFSD_VNET(nfsstatsv1_p)->direofcache_misses;
598 					nfsstatsov1.accesscache_hits =
599 					    NFSD_VNET(nfsstatsv1_p)->accesscache_hits;
600 					nfsstatsov1.accesscache_misses =
601 					    NFSD_VNET(nfsstatsv1_p)->accesscache_misses;
602 					nfsstatsov1.biocache_reads =
603 					    NFSD_VNET(nfsstatsv1_p)->biocache_reads;
604 					nfsstatsov1.read_bios =
605 					    NFSD_VNET(nfsstatsv1_p)->read_bios;
606 					nfsstatsov1.read_physios =
607 					    NFSD_VNET(nfsstatsv1_p)->read_physios;
608 					nfsstatsov1.biocache_writes =
609 					    NFSD_VNET(nfsstatsv1_p)->biocache_writes;
610 					nfsstatsov1.write_bios =
611 					    NFSD_VNET(nfsstatsv1_p)->write_bios;
612 					nfsstatsov1.write_physios =
613 					    NFSD_VNET(nfsstatsv1_p)->write_physios;
614 					nfsstatsov1.biocache_readlinks =
615 					    NFSD_VNET(nfsstatsv1_p)->biocache_readlinks;
616 					nfsstatsov1.readlink_bios =
617 					    NFSD_VNET(nfsstatsv1_p)->readlink_bios;
618 					nfsstatsov1.biocache_readdirs =
619 					    NFSD_VNET(nfsstatsv1_p)->biocache_readdirs;
620 					nfsstatsov1.readdir_bios =
621 					    NFSD_VNET(nfsstatsv1_p)->readdir_bios;
622 					for (i = 0; i < NFSV42_OLDNPROCS; i++)
623 						nfsstatsov1.rpccnt[i] =
624 						    NFSD_VNET(nfsstatsv1_p)->rpccnt[i];
625 					nfsstatsov1.rpcretries =
626 					    NFSD_VNET(nfsstatsv1_p)->rpcretries;
627 					for (i = 0; i < NFSV42_PURENOPS; i++)
628 						nfsstatsov1.srvrpccnt[i] =
629 						    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
630 					for (i = NFSV42_NOPS,
631 					     j = NFSV42_PURENOPS;
632 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
633 					     i++, j++)
634 						nfsstatsov1.srvrpccnt[j] =
635 						    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
636 					nfsstatsov1.reserved_0 = 0;
637 					nfsstatsov1.reserved_1 = 0;
638 					nfsstatsov1.rpcrequests =
639 					    NFSD_VNET(nfsstatsv1_p)->rpcrequests;
640 					nfsstatsov1.rpctimeouts =
641 					    NFSD_VNET(nfsstatsv1_p)->rpctimeouts;
642 					nfsstatsov1.rpcunexpected =
643 					    NFSD_VNET(nfsstatsv1_p)->rpcunexpected;
644 					nfsstatsov1.rpcinvalid =
645 					    NFSD_VNET(nfsstatsv1_p)->rpcinvalid;
646 					nfsstatsov1.srvcache_inproghits =
647 					    NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits;
648 					nfsstatsov1.reserved_2 = 0;
649 					nfsstatsov1.srvcache_nonidemdonehits =
650 					    NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
651 					nfsstatsov1.srvcache_misses =
652 					    NFSD_VNET(nfsstatsv1_p)->srvcache_misses;
653 					nfsstatsov1.srvcache_tcppeak =
654 					    NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak;
655 					nfsstatsov1.srvcache_size =
656 					    NFSD_VNET(nfsstatsv1_p)->srvcache_size;
657 					nfsstatsov1.srvclients =
658 					    NFSD_VNET(nfsstatsv1_p)->srvclients;
659 					nfsstatsov1.srvopenowners =
660 					    NFSD_VNET(nfsstatsv1_p)->srvopenowners;
661 					nfsstatsov1.srvopens =
662 					    NFSD_VNET(nfsstatsv1_p)->srvopens;
663 					nfsstatsov1.srvlockowners =
664 					    NFSD_VNET(nfsstatsv1_p)->srvlockowners;
665 					nfsstatsov1.srvlocks =
666 					    NFSD_VNET(nfsstatsv1_p)->srvlocks;
667 					nfsstatsov1.srvdelegates =
668 					    NFSD_VNET(nfsstatsv1_p)->srvdelegates;
669 					for (i = 0; i < NFSV42_CBNOPS; i++)
670 						nfsstatsov1.cbrpccnt[i] =
671 						    NFSD_VNET(nfsstatsv1_p)->cbrpccnt[i];
672 					nfsstatsov1.clopenowners =
673 					    NFSD_VNET(nfsstatsv1_p)->clopenowners;
674 					nfsstatsov1.clopens =
675 					    NFSD_VNET(nfsstatsv1_p)->clopens;
676 					nfsstatsov1.cllockowners =
677 					    NFSD_VNET(nfsstatsv1_p)->cllockowners;
678 					nfsstatsov1.cllocks =
679 					    NFSD_VNET(nfsstatsv1_p)->cllocks;
680 					nfsstatsov1.cldelegates =
681 					    NFSD_VNET(nfsstatsv1_p)->cldelegates;
682 					nfsstatsov1.cllocalopenowners =
683 					    NFSD_VNET(nfsstatsv1_p)->cllocalopenowners;
684 					nfsstatsov1.cllocalopens =
685 					    NFSD_VNET(nfsstatsv1_p)->cllocalopens;
686 					nfsstatsov1.cllocallockowners =
687 					    NFSD_VNET(nfsstatsv1_p)->cllocallockowners;
688 					nfsstatsov1.cllocallocks =
689 					    NFSD_VNET(nfsstatsv1_p)->cllocallocks;
690 					nfsstatsov1.srvstartcnt =
691 					    NFSD_VNET(nfsstatsv1_p)->srvstartcnt;
692 					nfsstatsov1.srvdonecnt =
693 					    NFSD_VNET(nfsstatsv1_p)->srvdonecnt;
694 					for (i = NFSV42_NOPS,
695 					     j = NFSV42_PURENOPS;
696 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
697 					     i++, j++) {
698 						nfsstatsov1.srvbytes[j] =
699 						    NFSD_VNET(nfsstatsv1_p)->srvbytes[i];
700 						nfsstatsov1.srvops[j] =
701 						    NFSD_VNET(nfsstatsv1_p)->srvops[i];
702 						nfsstatsov1.srvduration[j] =
703 						    NFSD_VNET(nfsstatsv1_p)->srvduration[i];
704 					}
705 					nfsstatsov1.busyfrom =
706 					    NFSD_VNET(nfsstatsv1_p)->busyfrom;
707 					nfsstatsov1.busyfrom =
708 					    NFSD_VNET(nfsstatsv1_p)->busyfrom;
709 					error = copyout(&nfsstatsov1, uap->argp,
710 					    sizeof(nfsstatsov1));
711 				} else if (nfsstatver.vers != NFSSTATS_V1)
712 					error = EPERM;
713 				else
714 					error = copyout(NFSD_VNET(nfsstatsv1_p),
715 					    uap->argp, sizeof(nfsstatsv1));
716 			}
717 		}
718 		if (error == 0) {
719 			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
720 				NFSD_VNET(nfsstatsv1_p)->attrcache_hits = 0;
721 				NFSD_VNET(nfsstatsv1_p)->attrcache_misses = 0;
722 				NFSD_VNET(nfsstatsv1_p)->lookupcache_hits = 0;
723 				NFSD_VNET(nfsstatsv1_p)->lookupcache_misses = 0;
724 				NFSD_VNET(nfsstatsv1_p)->direofcache_hits = 0;
725 				NFSD_VNET(nfsstatsv1_p)->direofcache_misses = 0;
726 				NFSD_VNET(nfsstatsv1_p)->accesscache_hits = 0;
727 				NFSD_VNET(nfsstatsv1_p)->accesscache_misses = 0;
728 				NFSD_VNET(nfsstatsv1_p)->biocache_reads = 0;
729 				NFSD_VNET(nfsstatsv1_p)->read_bios = 0;
730 				NFSD_VNET(nfsstatsv1_p)->read_physios = 0;
731 				NFSD_VNET(nfsstatsv1_p)->biocache_writes = 0;
732 				NFSD_VNET(nfsstatsv1_p)->write_bios = 0;
733 				NFSD_VNET(nfsstatsv1_p)->write_physios = 0;
734 				NFSD_VNET(nfsstatsv1_p)->biocache_readlinks = 0;
735 				NFSD_VNET(nfsstatsv1_p)->readlink_bios = 0;
736 				NFSD_VNET(nfsstatsv1_p)->biocache_readdirs = 0;
737 				NFSD_VNET(nfsstatsv1_p)->readdir_bios = 0;
738 				NFSD_VNET(nfsstatsv1_p)->rpcretries = 0;
739 				NFSD_VNET(nfsstatsv1_p)->rpcrequests = 0;
740 				NFSD_VNET(nfsstatsv1_p)->rpctimeouts = 0;
741 				NFSD_VNET(nfsstatsv1_p)->rpcunexpected = 0;
742 				NFSD_VNET(nfsstatsv1_p)->rpcinvalid = 0;
743 				bzero(NFSD_VNET(nfsstatsv1_p)->rpccnt,
744 				    sizeof(NFSD_VNET(nfsstatsv1_p)->rpccnt));
745 			}
746 			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
747 				NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits = 0;
748 				NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits = 0;
749 				NFSD_VNET(nfsstatsv1_p)->srvcache_misses = 0;
750 				NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak = 0;
751 				bzero(NFSD_VNET(nfsstatsv1_p)->srvrpccnt,
752 				    sizeof(NFSD_VNET(nfsstatsv1_p)->srvrpccnt));
753 				bzero(NFSD_VNET(nfsstatsv1_p)->cbrpccnt,
754 				    sizeof(NFSD_VNET(nfsstatsv1_p)->cbrpccnt));
755 			}
756 		}
757 		goto out;
758 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
759 		u_short sockport;
760 		struct nfsuserd_args nargs;
761 
762 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
763 			error = copyin(uap->argp, (caddr_t)&sockport,
764 			    sizeof (u_short));
765 			if (error == 0) {
766 				nargs.nuserd_family = AF_INET;
767 				nargs.nuserd_port = sockport;
768 			}
769 		} else {
770 			/*
771 			 * New nfsuserd_args structure, which indicates
772 			 * which IP version to use along with the port#.
773 			 */
774 			error = copyin(uap->argp, &nargs, sizeof(nargs));
775 		}
776 		if (!error)
777 			error = nfsrv_nfsuserdport(&nargs, p);
778 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
779 		nfsrv_nfsuserddelport();
780 		error = 0;
781 	}
782 
783 out:
784 	NFSEXITCODE(error);
785 	return (error);
786 }
787 
788 /*
789  * called by all three modevent routines, so that it gets things
790  * initialized soon enough.
791  */
792 void
793 newnfs_portinit(void)
794 {
795 	static int inited = 0;
796 
797 	if (inited)
798 		return;
799 	inited = 1;
800 	/* Initialize SMP locks used by both client and server. */
801 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
802 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
803 	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
804 }
805 
806 /*
807  * Determine if the file system supports NFSv4 ACLs.
808  * Return 1 if it does, 0 otherwise.
809  */
810 int
811 nfs_supportsnfsv4acls(struct vnode *vp)
812 {
813 	int error;
814 	long retval;
815 
816 	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
817 
818 	if (nfsrv_useacl == 0)
819 		return (0);
820 	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
821 	if (error == 0 && retval != 0)
822 		return (1);
823 	return (0);
824 }
825 
826 /*
827  * These are the first fields of all the context structures passed into
828  * nfs_pnfsio().
829  */
830 struct pnfsio {
831 	int		done;
832 	int		inprog;
833 	struct task	tsk;
834 };
835 
836 /*
837  * Do a mirror I/O on a pNFS thread.
838  */
839 int
840 nfs_pnfsio(task_fn_t *func, void *context)
841 {
842 	struct pnfsio *pio;
843 	int ret;
844 	static struct taskqueue *pnfsioq = NULL;
845 
846 	pio = (struct pnfsio *)context;
847 	if (pnfsioq == NULL) {
848 		if (nfs_pnfsiothreads == 0)
849 			return (EPERM);
850 		if (nfs_pnfsiothreads < 0)
851 			nfs_pnfsiothreads = mp_ncpus * 4;
852 		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
853 		    taskqueue_thread_enqueue, &pnfsioq);
854 		if (pnfsioq == NULL)
855 			return (ENOMEM);
856 		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
857 		    0, "pnfsiot");
858 		if (ret != 0) {
859 			taskqueue_free(pnfsioq);
860 			pnfsioq = NULL;
861 			return (ret);
862 		}
863 	}
864 	pio->inprog = 1;
865 	TASK_INIT(&pio->tsk, 0, func, context);
866 	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
867 	if (ret != 0)
868 		pio->inprog = 0;
869 	return (ret);
870 }
871 
872 /*
873  * Initialize everything that needs to be initialized for a vnet.
874  */
875 static void
876 nfs_vnetinit(const void *unused __unused)
877 {
878 
879 	if (IS_DEFAULT_VNET(curvnet))
880 		NFSD_VNET(nfsstatsv1_p) = &nfsstatsv1;
881 	else
882 		NFSD_VNET(nfsstatsv1_p) = malloc(sizeof(struct nfsstatsv1),
883 		    M_TEMP, M_WAITOK | M_ZERO);
884 	mtx_init(&NFSD_VNET(nfsrv_nfsuserdsock).nr_mtx, "nfsuserd",
885 	    NULL, MTX_DEF);
886 }
887 VNET_SYSINIT(nfs_vnetinit, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
888     nfs_vnetinit, NULL);
889 
890 static void
891 nfs_cleanup(void *unused __unused)
892 {
893 
894 	mtx_destroy(&NFSD_VNET(nfsrv_nfsuserdsock).nr_mtx);
895 	if (!IS_DEFAULT_VNET(curvnet)) {
896 		free(NFSD_VNET(nfsstatsv1_p), M_TEMP);
897 		NFSD_VNET(nfsstatsv1_p) = NULL;
898 	}
899 	/* Clean out the name<-->id cache. */
900 	nfsrv_cleanusergroup();
901 }
902 VNET_SYSUNINIT(nfs_cleanup, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
903     nfs_cleanup, NULL);
904 
905 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
906 
907 /*
908  * Called once to initialize data structures...
909  */
910 static int
911 nfscommon_modevent(module_t mod, int type, void *data)
912 {
913 	int error = 0;
914 	static int loaded = 0;
915 
916 	switch (type) {
917 	case MOD_LOAD:
918 		if (loaded)
919 			goto out;
920 		newnfs_portinit();
921 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
922 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
923 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
924 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
925 		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
926 		TAILQ_INIT(&nfsrv_devidhead);
927 		newnfs_init();
928 		nfsd_call_nfscommon = nfssvc_nfscommon;
929 		loaded = 1;
930 		break;
931 
932 	case MOD_UNLOAD:
933 		if (newnfs_numnfsd != 0 ||
934 		    NFSD_VNET(nfsrv_nfsuserd) != NOTRUNNING ||
935 		    nfs_numnfscbd != 0) {
936 			error = EBUSY;
937 			break;
938 		}
939 
940 		nfsd_call_nfscommon = NULL;
941 		/* and get rid of the mutexes */
942 		mtx_destroy(&nfs_nameid_mutex);
943 		mtx_destroy(&newnfsd_mtx);
944 		mtx_destroy(&nfs_state_mutex);
945 		mtx_destroy(&nfs_clstate_mutex);
946 		mtx_destroy(&nfs_sockl_mutex);
947 		mtx_destroy(&nfs_slock_mutex);
948 		mtx_destroy(&nfs_req_mutex);
949 		mtx_destroy(&nfsrv_dslock_mtx);
950 		loaded = 0;
951 		break;
952 	default:
953 		error = EOPNOTSUPP;
954 		break;
955 	}
956 
957 out:
958 	NFSEXITCODE(error);
959 	return error;
960 }
961 static moduledata_t nfscommon_mod = {
962 	"nfscommon",
963 	nfscommon_modevent,
964 	NULL,
965 };
966 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
967 
968 /* So that loader and kldload(2) can find us, wherever we are.. */
969 MODULE_VERSION(nfscommon, 1);
970 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
971 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
972