xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision b07549f3c15637af6064d2f52b572f239f5ba870)
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 void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
60     struct thread *);
61 extern int nfsrv_useacl;
62 int newnfs_numnfsd = 0;
63 struct nfsstatsv1 nfsstatsv1;
64 int nfs_numnfscbd = 0;
65 int nfscl_debuglevel = 0;
66 char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
67 int nfsrv_lughashsize = 100;
68 struct mtx nfsrv_dslock_mtx;
69 struct nfsdevicehead nfsrv_devidhead;
70 volatile int nfsrv_devidcnt = 0;
71 void (*ncl_call_invalcaches)(struct vnode *) = NULL;
72 vop_advlock_t *nfs_advlock_p = NULL;
73 vop_reclaim_t *nfs_reclaim_p = NULL;
74 uint32_t nfs_srvmaxio = NFS_SRVMAXIO;
75 
76 NFSD_VNET_DEFINE(struct nfsstatsv1 *, nfsstatsv1_p);
77 
78 NFSD_VNET_DECLARE(struct nfssockreq, nfsrv_nfsuserdsock);
79 NFSD_VNET_DECLARE(nfsuserd_state, nfsrv_nfsuserd);
80 
81 int nfs_pnfsio(task_fn_t *, void *);
82 
83 static int nfs_realign_test;
84 static int nfs_realign_count;
85 static struct ext_nfsstats oldnfsstats;
86 static struct nfsstatsov1 nfsstatsov1;
87 
88 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
89     "NFS filesystem");
90 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
91     0, "Number of realign tests done");
92 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
93     0, "Number of mbuf realignments done");
94 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
95     nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
96     "NFSv4 callback addr for server to use");
97 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
98     0, "Debug level for NFS client");
99 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
100     0, "Size of hash tables for uid/name mapping");
101 int nfs_pnfsiothreads = -1;
102 SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsiothreads, CTLFLAG_RW, &nfs_pnfsiothreads,
103     0, "Number of pNFS mirror I/O threads");
104 
105 /*
106  * Defines for malloc
107  * (Here for FreeBSD, since they allocate storage.)
108  */
109 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
110 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
111 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
112     "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
113 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
114 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
115 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
116 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
117 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
118 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
119 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
120 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
121 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
122 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
123 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
124 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
125 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
126 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer");
127 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroff",
128     "NFS directory offset data");
129 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
130     "NFS local lock rollback");
131 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
132 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
133 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
134 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
135 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
136 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
137 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
138 
139 /*
140  * Definition of mutex locks.
141  * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
142  * and assorted other nfsd structures.
143  */
144 struct mtx newnfsd_mtx;
145 struct mtx nfs_sockl_mutex;
146 struct mtx nfs_state_mutex;
147 struct mtx nfs_nameid_mutex;
148 struct mtx nfs_req_mutex;
149 struct mtx nfs_slock_mutex;
150 struct mtx nfs_clstate_mutex;
151 
152 /* local functions */
153 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
154 
155 #ifdef __NO_STRICT_ALIGNMENT
156 /*
157  * These architectures don't need re-alignment, so just return.
158  */
159 int
160 newnfs_realign(struct mbuf **pm, int how)
161 {
162 
163 	return (0);
164 }
165 #else	/* !__NO_STRICT_ALIGNMENT */
166 /*
167  *	newnfs_realign:
168  *
169  *	Check for badly aligned mbuf data and realign by copying the unaligned
170  *	portion of the data into a new mbuf chain and freeing the portions
171  *	of the old chain that were replaced.
172  *
173  *	We cannot simply realign the data within the existing mbuf chain
174  *	because the underlying buffers may contain other rpc commands and
175  *	we cannot afford to overwrite them.
176  *
177  *	We would prefer to avoid this situation entirely.  The situation does
178  *	not occur with NFS/UDP and is supposed to only occasionally occur
179  *	with TCP.  Use vfs.nfs.realign_count and realign_test to check this.
180  *
181  */
182 int
183 newnfs_realign(struct mbuf **pm, int how)
184 {
185 	struct mbuf *m, *n;
186 	int off, space;
187 
188 	++nfs_realign_test;
189 	while ((m = *pm) != NULL) {
190 		if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
191 			/*
192 			 * NB: we can't depend on m_pkthdr.len to help us
193 			 * decide what to do here.  May not be worth doing
194 			 * the m_length calculation as m_copyback will
195 			 * expand the mbuf chain below as needed.
196 			 */
197 			space = m_length(m, NULL);
198 			if (space >= MINCLSIZE) {
199 				/* NB: m_copyback handles space > MCLBYTES */
200 				n = m_getcl(how, MT_DATA, 0);
201 			} else
202 				n = m_get(how, MT_DATA);
203 			if (n == NULL)
204 				return (ENOMEM);
205 			/*
206 			 * Align the remainder of the mbuf chain.
207 			 */
208 			n->m_len = 0;
209 			off = 0;
210 			while (m != NULL) {
211 				m_copyback(n, off, m->m_len, mtod(m, caddr_t));
212 				off += m->m_len;
213 				m = m->m_next;
214 			}
215 			m_freem(*pm);
216 			*pm = n;
217 			++nfs_realign_count;
218 			break;
219 		}
220 		pm = &m->m_next;
221 	}
222 
223 	return (0);
224 }
225 #endif	/* __NO_STRICT_ALIGNMENT */
226 
227 #ifdef notdef
228 static void
229 nfsrv_object_create(struct vnode *vp, struct thread *td)
230 {
231 
232 	if (vp == NULL || vp->v_type != VREG)
233 		return;
234 	(void) vfs_object_create(vp, td, td->td_ucred);
235 }
236 #endif
237 
238 /*
239  * Look up a file name. Basically just initialize stuff and call namei().
240  */
241 int
242 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p __unused)
243 {
244 	int error;
245 
246 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname);
247 	error = namei(ndp);
248 	if (!error) {
249 		NDFREE_PNBUF(ndp);
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  * Sleep for a short period of time unless errval == NFSERR_GRACE, where
404  * the sleep should be for 5 seconds.
405  * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
406  * an event that never gets a wakeup. Only return EINTR or 0.
407  */
408 int
409 nfs_catnap(int prio, int errval, const char *wmesg)
410 {
411 	static int non_event;
412 	int ret;
413 
414 	if (errval == NFSERR_GRACE)
415 		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
416 	else
417 		ret = tsleep(&non_event, prio, wmesg, 1);
418 	if (ret != EINTR)
419 		ret = 0;
420 	return (ret);
421 }
422 
423 /*
424  * Get referral. For now, just fail.
425  */
426 struct nfsreferral *
427 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
428 {
429 
430 	return (NULL);
431 }
432 
433 static int
434 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
435 {
436 	int error;
437 
438 	NFSD_CURVNET_SET(NFSD_TD_TO_VNET(td));
439 	error = nfssvc_call(td, uap, td->td_ucred);
440 	NFSD_CURVNET_RESTORE();
441 	NFSEXITCODE(error);
442 	return (error);
443 }
444 
445 static int
446 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
447 {
448 	int error = EINVAL, i, j;
449 	struct nfsd_idargs nid;
450 	struct nfsd_oidargs onid;
451 	struct {
452 		int vers;	/* Just the first field of nfsstats. */
453 	} nfsstatver;
454 
455 	if (uap->flag & NFSSVC_IDNAME) {
456 		if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
457 			error = copyin(uap->argp, &nid, sizeof(nid));
458 		else {
459 			error = copyin(uap->argp, &onid, sizeof(onid));
460 			if (error == 0) {
461 				nid.nid_flag = onid.nid_flag;
462 				nid.nid_uid = onid.nid_uid;
463 				nid.nid_gid = onid.nid_gid;
464 				nid.nid_usermax = onid.nid_usermax;
465 				nid.nid_usertimeout = onid.nid_usertimeout;
466 				nid.nid_name = onid.nid_name;
467 				nid.nid_namelen = onid.nid_namelen;
468 				nid.nid_ngroup = 0;
469 				nid.nid_grps = NULL;
470 			}
471 		}
472 		if (error)
473 			goto out;
474 		error = nfssvc_idname(&nid);
475 		goto out;
476 	} else if (uap->flag & NFSSVC_GETSTATS) {
477 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
478 			/* Copy fields to the old ext_nfsstat structure. */
479 			oldnfsstats.attrcache_hits =
480 			    NFSD_VNET(nfsstatsv1_p)->attrcache_hits;
481 			oldnfsstats.attrcache_misses =
482 			    NFSD_VNET(nfsstatsv1_p)->attrcache_misses;
483 			oldnfsstats.lookupcache_hits =
484 			    NFSD_VNET(nfsstatsv1_p)->lookupcache_hits;
485 			oldnfsstats.lookupcache_misses =
486 			    NFSD_VNET(nfsstatsv1_p)->lookupcache_misses;
487 			oldnfsstats.direofcache_hits =
488 			    NFSD_VNET(nfsstatsv1_p)->direofcache_hits;
489 			oldnfsstats.direofcache_misses =
490 			    NFSD_VNET(nfsstatsv1_p)->direofcache_misses;
491 			oldnfsstats.accesscache_hits =
492 			    NFSD_VNET(nfsstatsv1_p)->accesscache_hits;
493 			oldnfsstats.accesscache_misses =
494 			    NFSD_VNET(nfsstatsv1_p)->accesscache_misses;
495 			oldnfsstats.biocache_reads =
496 			    NFSD_VNET(nfsstatsv1_p)->biocache_reads;
497 			oldnfsstats.read_bios =
498 			    NFSD_VNET(nfsstatsv1_p)->read_bios;
499 			oldnfsstats.read_physios =
500 			    NFSD_VNET(nfsstatsv1_p)->read_physios;
501 			oldnfsstats.biocache_writes =
502 			    NFSD_VNET(nfsstatsv1_p)->biocache_writes;
503 			oldnfsstats.write_bios =
504 			    NFSD_VNET(nfsstatsv1_p)->write_bios;
505 			oldnfsstats.write_physios =
506 			    NFSD_VNET(nfsstatsv1_p)->write_physios;
507 			oldnfsstats.biocache_readlinks =
508 			    NFSD_VNET(nfsstatsv1_p)->biocache_readlinks;
509 			oldnfsstats.readlink_bios =
510 			    NFSD_VNET(nfsstatsv1_p)->readlink_bios;
511 			oldnfsstats.biocache_readdirs =
512 			    NFSD_VNET(nfsstatsv1_p)->biocache_readdirs;
513 			oldnfsstats.readdir_bios =
514 			    NFSD_VNET(nfsstatsv1_p)->readdir_bios;
515 			for (i = 0; i < NFSV4_NPROCS; i++)
516 				oldnfsstats.rpccnt[i] =
517 				    NFSD_VNET(nfsstatsv1_p)->rpccnt[i];
518 			oldnfsstats.rpcretries =
519 			    NFSD_VNET(nfsstatsv1_p)->rpcretries;
520 			for (i = 0; i < NFSV4OP_NOPS; i++)
521 				oldnfsstats.srvrpccnt[i] =
522 				    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
523 			for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
524 			    i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
525 				oldnfsstats.srvrpccnt[j] =
526 				    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
527 			oldnfsstats.reserved_0 = 0;
528 			oldnfsstats.reserved_1 = 0;
529 			oldnfsstats.rpcrequests =
530 			    NFSD_VNET(nfsstatsv1_p)->rpcrequests;
531 			oldnfsstats.rpctimeouts =
532 			    NFSD_VNET(nfsstatsv1_p)->rpctimeouts;
533 			oldnfsstats.rpcunexpected =
534 			    NFSD_VNET(nfsstatsv1_p)->rpcunexpected;
535 			oldnfsstats.rpcinvalid =
536 			    NFSD_VNET(nfsstatsv1_p)->rpcinvalid;
537 			oldnfsstats.srvcache_inproghits =
538 			    NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits;
539 			oldnfsstats.reserved_2 = 0;
540 			oldnfsstats.srvcache_nonidemdonehits =
541 			    NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
542 			oldnfsstats.srvcache_misses =
543 			    NFSD_VNET(nfsstatsv1_p)->srvcache_misses;
544 			oldnfsstats.srvcache_tcppeak =
545 			    NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak;
546 			oldnfsstats.srvcache_size =
547 			    NFSD_VNET(nfsstatsv1_p)->srvcache_size;
548 			oldnfsstats.srvclients =
549 			    NFSD_VNET(nfsstatsv1_p)->srvclients;
550 			oldnfsstats.srvopenowners =
551 			    NFSD_VNET(nfsstatsv1_p)->srvopenowners;
552 			oldnfsstats.srvopens =
553 			    NFSD_VNET(nfsstatsv1_p)->srvopens;
554 			oldnfsstats.srvlockowners =
555 			    NFSD_VNET(nfsstatsv1_p)->srvlockowners;
556 			oldnfsstats.srvlocks =
557 			    NFSD_VNET(nfsstatsv1_p)->srvlocks;
558 			oldnfsstats.srvdelegates =
559 			    NFSD_VNET(nfsstatsv1_p)->srvdelegates;
560 			for (i = 0; i < NFSV4OP_CBNOPS; i++)
561 				oldnfsstats.cbrpccnt[i] =
562 				    NFSD_VNET(nfsstatsv1_p)->cbrpccnt[i];
563 			oldnfsstats.clopenowners =
564 			    NFSD_VNET(nfsstatsv1_p)->clopenowners;
565 			oldnfsstats.clopens = NFSD_VNET(nfsstatsv1_p)->clopens;
566 			oldnfsstats.cllockowners =
567 			    NFSD_VNET(nfsstatsv1_p)->cllockowners;
568 			oldnfsstats.cllocks = NFSD_VNET(nfsstatsv1_p)->cllocks;
569 			oldnfsstats.cldelegates =
570 			    NFSD_VNET(nfsstatsv1_p)->cldelegates;
571 			oldnfsstats.cllocalopenowners =
572 			    NFSD_VNET(nfsstatsv1_p)->cllocalopenowners;
573 			oldnfsstats.cllocalopens =
574 			    NFSD_VNET(nfsstatsv1_p)->cllocalopens;
575 			oldnfsstats.cllocallockowners =
576 			    NFSD_VNET(nfsstatsv1_p)->cllocallockowners;
577 			oldnfsstats.cllocallocks =
578 			    NFSD_VNET(nfsstatsv1_p)->cllocallocks;
579 			error = copyout(&oldnfsstats, uap->argp,
580 			    sizeof (oldnfsstats));
581 		} else {
582 			error = copyin(uap->argp, &nfsstatver,
583 			    sizeof(nfsstatver));
584 			if (error == 0) {
585 				if (nfsstatver.vers == NFSSTATS_OV1) {
586 					/* Copy nfsstatsv1 to nfsstatsov1. */
587 					nfsstatsov1.attrcache_hits =
588 					    NFSD_VNET(nfsstatsv1_p)->attrcache_hits;
589 					nfsstatsov1.attrcache_misses =
590 					    NFSD_VNET(nfsstatsv1_p)->attrcache_misses;
591 					nfsstatsov1.lookupcache_hits =
592 					    NFSD_VNET(nfsstatsv1_p)->lookupcache_hits;
593 					nfsstatsov1.lookupcache_misses =
594 					    NFSD_VNET(nfsstatsv1_p)->lookupcache_misses;
595 					nfsstatsov1.direofcache_hits =
596 					    NFSD_VNET(nfsstatsv1_p)->direofcache_hits;
597 					nfsstatsov1.direofcache_misses =
598 					    NFSD_VNET(nfsstatsv1_p)->direofcache_misses;
599 					nfsstatsov1.accesscache_hits =
600 					    NFSD_VNET(nfsstatsv1_p)->accesscache_hits;
601 					nfsstatsov1.accesscache_misses =
602 					    NFSD_VNET(nfsstatsv1_p)->accesscache_misses;
603 					nfsstatsov1.biocache_reads =
604 					    NFSD_VNET(nfsstatsv1_p)->biocache_reads;
605 					nfsstatsov1.read_bios =
606 					    NFSD_VNET(nfsstatsv1_p)->read_bios;
607 					nfsstatsov1.read_physios =
608 					    NFSD_VNET(nfsstatsv1_p)->read_physios;
609 					nfsstatsov1.biocache_writes =
610 					    NFSD_VNET(nfsstatsv1_p)->biocache_writes;
611 					nfsstatsov1.write_bios =
612 					    NFSD_VNET(nfsstatsv1_p)->write_bios;
613 					nfsstatsov1.write_physios =
614 					    NFSD_VNET(nfsstatsv1_p)->write_physios;
615 					nfsstatsov1.biocache_readlinks =
616 					    NFSD_VNET(nfsstatsv1_p)->biocache_readlinks;
617 					nfsstatsov1.readlink_bios =
618 					    NFSD_VNET(nfsstatsv1_p)->readlink_bios;
619 					nfsstatsov1.biocache_readdirs =
620 					    NFSD_VNET(nfsstatsv1_p)->biocache_readdirs;
621 					nfsstatsov1.readdir_bios =
622 					    NFSD_VNET(nfsstatsv1_p)->readdir_bios;
623 					for (i = 0; i < NFSV42_OLDNPROCS; i++)
624 						nfsstatsov1.rpccnt[i] =
625 						    NFSD_VNET(nfsstatsv1_p)->rpccnt[i];
626 					nfsstatsov1.rpcretries =
627 					    NFSD_VNET(nfsstatsv1_p)->rpcretries;
628 					for (i = 0; i < NFSV42_PURENOPS; i++)
629 						nfsstatsov1.srvrpccnt[i] =
630 						    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
631 					for (i = NFSV42_NOPS,
632 					     j = NFSV42_PURENOPS;
633 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
634 					     i++, j++)
635 						nfsstatsov1.srvrpccnt[j] =
636 						    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
637 					nfsstatsov1.reserved_0 = 0;
638 					nfsstatsov1.reserved_1 = 0;
639 					nfsstatsov1.rpcrequests =
640 					    NFSD_VNET(nfsstatsv1_p)->rpcrequests;
641 					nfsstatsov1.rpctimeouts =
642 					    NFSD_VNET(nfsstatsv1_p)->rpctimeouts;
643 					nfsstatsov1.rpcunexpected =
644 					    NFSD_VNET(nfsstatsv1_p)->rpcunexpected;
645 					nfsstatsov1.rpcinvalid =
646 					    NFSD_VNET(nfsstatsv1_p)->rpcinvalid;
647 					nfsstatsov1.srvcache_inproghits =
648 					    NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits;
649 					nfsstatsov1.reserved_2 = 0;
650 					nfsstatsov1.srvcache_nonidemdonehits =
651 					    NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
652 					nfsstatsov1.srvcache_misses =
653 					    NFSD_VNET(nfsstatsv1_p)->srvcache_misses;
654 					nfsstatsov1.srvcache_tcppeak =
655 					    NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak;
656 					nfsstatsov1.srvcache_size =
657 					    NFSD_VNET(nfsstatsv1_p)->srvcache_size;
658 					nfsstatsov1.srvclients =
659 					    NFSD_VNET(nfsstatsv1_p)->srvclients;
660 					nfsstatsov1.srvopenowners =
661 					    NFSD_VNET(nfsstatsv1_p)->srvopenowners;
662 					nfsstatsov1.srvopens =
663 					    NFSD_VNET(nfsstatsv1_p)->srvopens;
664 					nfsstatsov1.srvlockowners =
665 					    NFSD_VNET(nfsstatsv1_p)->srvlockowners;
666 					nfsstatsov1.srvlocks =
667 					    NFSD_VNET(nfsstatsv1_p)->srvlocks;
668 					nfsstatsov1.srvdelegates =
669 					    NFSD_VNET(nfsstatsv1_p)->srvdelegates;
670 					for (i = 0; i < NFSV42_CBNOPS; i++)
671 						nfsstatsov1.cbrpccnt[i] =
672 						    NFSD_VNET(nfsstatsv1_p)->cbrpccnt[i];
673 					nfsstatsov1.clopenowners =
674 					    NFSD_VNET(nfsstatsv1_p)->clopenowners;
675 					nfsstatsov1.clopens =
676 					    NFSD_VNET(nfsstatsv1_p)->clopens;
677 					nfsstatsov1.cllockowners =
678 					    NFSD_VNET(nfsstatsv1_p)->cllockowners;
679 					nfsstatsov1.cllocks =
680 					    NFSD_VNET(nfsstatsv1_p)->cllocks;
681 					nfsstatsov1.cldelegates =
682 					    NFSD_VNET(nfsstatsv1_p)->cldelegates;
683 					nfsstatsov1.cllocalopenowners =
684 					    NFSD_VNET(nfsstatsv1_p)->cllocalopenowners;
685 					nfsstatsov1.cllocalopens =
686 					    NFSD_VNET(nfsstatsv1_p)->cllocalopens;
687 					nfsstatsov1.cllocallockowners =
688 					    NFSD_VNET(nfsstatsv1_p)->cllocallockowners;
689 					nfsstatsov1.cllocallocks =
690 					    NFSD_VNET(nfsstatsv1_p)->cllocallocks;
691 					nfsstatsov1.srvstartcnt =
692 					    NFSD_VNET(nfsstatsv1_p)->srvstartcnt;
693 					nfsstatsov1.srvdonecnt =
694 					    NFSD_VNET(nfsstatsv1_p)->srvdonecnt;
695 					for (i = NFSV42_NOPS,
696 					     j = NFSV42_PURENOPS;
697 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
698 					     i++, j++) {
699 						nfsstatsov1.srvbytes[j] =
700 						    NFSD_VNET(nfsstatsv1_p)->srvbytes[i];
701 						nfsstatsov1.srvops[j] =
702 						    NFSD_VNET(nfsstatsv1_p)->srvops[i];
703 						nfsstatsov1.srvduration[j] =
704 						    NFSD_VNET(nfsstatsv1_p)->srvduration[i];
705 					}
706 					nfsstatsov1.busyfrom =
707 					    NFSD_VNET(nfsstatsv1_p)->busyfrom;
708 					nfsstatsov1.busyfrom =
709 					    NFSD_VNET(nfsstatsv1_p)->busyfrom;
710 					error = copyout(&nfsstatsov1, uap->argp,
711 					    sizeof(nfsstatsov1));
712 				} else if (nfsstatver.vers != NFSSTATS_V1)
713 					error = EPERM;
714 				else
715 					error = copyout(NFSD_VNET(nfsstatsv1_p),
716 					    uap->argp, sizeof(nfsstatsv1));
717 			}
718 		}
719 		if (error == 0) {
720 			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
721 				NFSD_VNET(nfsstatsv1_p)->attrcache_hits = 0;
722 				NFSD_VNET(nfsstatsv1_p)->attrcache_misses = 0;
723 				NFSD_VNET(nfsstatsv1_p)->lookupcache_hits = 0;
724 				NFSD_VNET(nfsstatsv1_p)->lookupcache_misses = 0;
725 				NFSD_VNET(nfsstatsv1_p)->direofcache_hits = 0;
726 				NFSD_VNET(nfsstatsv1_p)->direofcache_misses = 0;
727 				NFSD_VNET(nfsstatsv1_p)->accesscache_hits = 0;
728 				NFSD_VNET(nfsstatsv1_p)->accesscache_misses = 0;
729 				NFSD_VNET(nfsstatsv1_p)->biocache_reads = 0;
730 				NFSD_VNET(nfsstatsv1_p)->read_bios = 0;
731 				NFSD_VNET(nfsstatsv1_p)->read_physios = 0;
732 				NFSD_VNET(nfsstatsv1_p)->biocache_writes = 0;
733 				NFSD_VNET(nfsstatsv1_p)->write_bios = 0;
734 				NFSD_VNET(nfsstatsv1_p)->write_physios = 0;
735 				NFSD_VNET(nfsstatsv1_p)->biocache_readlinks = 0;
736 				NFSD_VNET(nfsstatsv1_p)->readlink_bios = 0;
737 				NFSD_VNET(nfsstatsv1_p)->biocache_readdirs = 0;
738 				NFSD_VNET(nfsstatsv1_p)->readdir_bios = 0;
739 				NFSD_VNET(nfsstatsv1_p)->rpcretries = 0;
740 				NFSD_VNET(nfsstatsv1_p)->rpcrequests = 0;
741 				NFSD_VNET(nfsstatsv1_p)->rpctimeouts = 0;
742 				NFSD_VNET(nfsstatsv1_p)->rpcunexpected = 0;
743 				NFSD_VNET(nfsstatsv1_p)->rpcinvalid = 0;
744 				bzero(NFSD_VNET(nfsstatsv1_p)->rpccnt,
745 				    sizeof(NFSD_VNET(nfsstatsv1_p)->rpccnt));
746 			}
747 			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
748 				NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits = 0;
749 				NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits = 0;
750 				NFSD_VNET(nfsstatsv1_p)->srvcache_misses = 0;
751 				NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak = 0;
752 				bzero(NFSD_VNET(nfsstatsv1_p)->srvrpccnt,
753 				    sizeof(NFSD_VNET(nfsstatsv1_p)->srvrpccnt));
754 				bzero(NFSD_VNET(nfsstatsv1_p)->cbrpccnt,
755 				    sizeof(NFSD_VNET(nfsstatsv1_p)->cbrpccnt));
756 			}
757 		}
758 		goto out;
759 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
760 		u_short sockport;
761 		struct nfsuserd_args nargs;
762 
763 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
764 			error = copyin(uap->argp, (caddr_t)&sockport,
765 			    sizeof (u_short));
766 			if (error == 0) {
767 				nargs.nuserd_family = AF_INET;
768 				nargs.nuserd_port = sockport;
769 			}
770 		} else {
771 			/*
772 			 * New nfsuserd_args structure, which indicates
773 			 * which IP version to use along with the port#.
774 			 */
775 			error = copyin(uap->argp, &nargs, sizeof(nargs));
776 		}
777 		if (!error)
778 			error = nfsrv_nfsuserdport(&nargs, p);
779 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
780 		nfsrv_nfsuserddelport();
781 		error = 0;
782 	}
783 
784 out:
785 	NFSEXITCODE(error);
786 	return (error);
787 }
788 
789 /*
790  * called by all three modevent routines, so that it gets things
791  * initialized soon enough.
792  */
793 void
794 newnfs_portinit(void)
795 {
796 	static int inited = 0;
797 
798 	if (inited)
799 		return;
800 	inited = 1;
801 	/* Initialize SMP locks used by both client and server. */
802 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
803 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
804 	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
805 }
806 
807 /*
808  * Determine if the file system supports NFSv4 ACLs.
809  * Return 1 if it does, 0 otherwise.
810  */
811 int
812 nfs_supportsnfsv4acls(struct vnode *vp)
813 {
814 	int error;
815 	long retval;
816 
817 	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
818 
819 	if (nfsrv_useacl == 0)
820 		return (0);
821 	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
822 	if (error == 0 && retval != 0)
823 		return (1);
824 	return (0);
825 }
826 
827 /*
828  * These are the first fields of all the context structures passed into
829  * nfs_pnfsio().
830  */
831 struct pnfsio {
832 	int		done;
833 	int		inprog;
834 	struct task	tsk;
835 };
836 
837 /*
838  * Do a mirror I/O on a pNFS thread.
839  */
840 int
841 nfs_pnfsio(task_fn_t *func, void *context)
842 {
843 	struct pnfsio *pio;
844 	int ret;
845 	static struct taskqueue *pnfsioq = NULL;
846 
847 	pio = (struct pnfsio *)context;
848 	if (pnfsioq == NULL) {
849 		if (nfs_pnfsiothreads == 0)
850 			return (EPERM);
851 		if (nfs_pnfsiothreads < 0)
852 			nfs_pnfsiothreads = mp_ncpus * 4;
853 		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
854 		    taskqueue_thread_enqueue, &pnfsioq);
855 		if (pnfsioq == NULL)
856 			return (ENOMEM);
857 		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
858 		    0, "pnfsiot");
859 		if (ret != 0) {
860 			taskqueue_free(pnfsioq);
861 			pnfsioq = NULL;
862 			return (ret);
863 		}
864 	}
865 	pio->inprog = 1;
866 	TASK_INIT(&pio->tsk, 0, func, context);
867 	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
868 	if (ret != 0)
869 		pio->inprog = 0;
870 	return (ret);
871 }
872 
873 /*
874  * Initialize everything that needs to be initialized for a vnet.
875  */
876 static void
877 nfs_vnetinit(const void *unused __unused)
878 {
879 
880 	if (IS_DEFAULT_VNET(curvnet))
881 		NFSD_VNET(nfsstatsv1_p) = &nfsstatsv1;
882 	else
883 		NFSD_VNET(nfsstatsv1_p) = malloc(sizeof(struct nfsstatsv1),
884 		    M_TEMP, M_WAITOK | M_ZERO);
885 	mtx_init(&NFSD_VNET(nfsrv_nfsuserdsock).nr_mtx, "nfsuserd",
886 	    NULL, MTX_DEF);
887 }
888 VNET_SYSINIT(nfs_vnetinit, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
889     nfs_vnetinit, NULL);
890 
891 static void
892 nfs_cleanup(void *unused __unused)
893 {
894 
895 	mtx_destroy(&NFSD_VNET(nfsrv_nfsuserdsock).nr_mtx);
896 	if (!IS_DEFAULT_VNET(curvnet)) {
897 		free(NFSD_VNET(nfsstatsv1_p), M_TEMP);
898 		NFSD_VNET(nfsstatsv1_p) = NULL;
899 	}
900 	/* Clean out the name<-->id cache. */
901 	nfsrv_cleanusergroup();
902 }
903 VNET_SYSUNINIT(nfs_cleanup, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
904     nfs_cleanup, NULL);
905 
906 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
907 
908 /*
909  * Called once to initialize data structures...
910  */
911 static int
912 nfscommon_modevent(module_t mod, int type, void *data)
913 {
914 	int error = 0;
915 	static int loaded = 0;
916 
917 	switch (type) {
918 	case MOD_LOAD:
919 		if (loaded)
920 			goto out;
921 		newnfs_portinit();
922 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
923 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
924 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
925 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
926 		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
927 		TAILQ_INIT(&nfsrv_devidhead);
928 		newnfs_init();
929 		nfsd_call_nfscommon = nfssvc_nfscommon;
930 		loaded = 1;
931 		break;
932 
933 	case MOD_UNLOAD:
934 		if (newnfs_numnfsd != 0 ||
935 		    NFSD_VNET(nfsrv_nfsuserd) != NOTRUNNING ||
936 		    nfs_numnfscbd != 0) {
937 			error = EBUSY;
938 			break;
939 		}
940 
941 		nfsd_call_nfscommon = NULL;
942 		/* and get rid of the mutexes */
943 		mtx_destroy(&nfs_nameid_mutex);
944 		mtx_destroy(&newnfsd_mtx);
945 		mtx_destroy(&nfs_state_mutex);
946 		mtx_destroy(&nfs_clstate_mutex);
947 		mtx_destroy(&nfs_sockl_mutex);
948 		mtx_destroy(&nfs_slock_mutex);
949 		mtx_destroy(&nfs_req_mutex);
950 		mtx_destroy(&nfsrv_dslock_mtx);
951 		loaded = 0;
952 		break;
953 	default:
954 		error = EOPNOTSUPP;
955 		break;
956 	}
957 
958 out:
959 	NFSEXITCODE(error);
960 	return error;
961 }
962 static moduledata_t nfscommon_mod = {
963 	"nfscommon",
964 	nfscommon_modevent,
965 	NULL,
966 };
967 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
968 
969 /* So that loader and kldload(2) can find us, wherever we are.. */
970 MODULE_VERSION(nfscommon, 1);
971 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
972 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
973