xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision 7fdf597e96a02165cfe22ff357b857d5fa15ed8a)
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 
79 int nfs_pnfsio(task_fn_t *, void *);
80 
81 static int nfs_realign_test;
82 static int nfs_realign_count;
83 static struct ext_nfsstats oldnfsstats;
84 static struct nfsstatsov1 nfsstatsov1;
85 
86 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
87     "NFS filesystem");
88 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
89     0, "Number of realign tests done");
90 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
91     0, "Number of mbuf realignments done");
92 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
93     nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
94     "NFSv4 callback addr for server to use");
95 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
96     0, "Debug level for NFS client");
97 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
98     0, "Size of hash tables for uid/name mapping");
99 int nfs_pnfsiothreads = -1;
100 SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsiothreads, CTLFLAG_RW, &nfs_pnfsiothreads,
101     0, "Number of pNFS mirror I/O threads");
102 
103 /*
104  * Defines for malloc
105  * (Here for FreeBSD, since they allocate storage.)
106  */
107 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
108 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
109 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
110     "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
111 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
112 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
113 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
114 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
115 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
116 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
117 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
118 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
119 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
120 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
121 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
122 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
123 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
124 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroff",
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 __unused)
240 {
241 	int error;
242 
243 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname);
244 	error = namei(ndp);
245 	if (!error) {
246 		NDFREE_PNBUF(ndp);
247 	}
248 	return (error);
249 }
250 
251 /*
252  * Copy NFS uid, gids to the cred structure.
253  */
254 void
255 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
256 {
257 
258 	KASSERT(nfscr->nfsc_ngroups >= 0,
259 	    ("newnfs_copycred: negative nfsc_ngroups"));
260 	cr->cr_uid = nfscr->nfsc_uid;
261 	crsetgroups_fallback(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups,
262 	    GID_NOGROUP);
263 }
264 
265 /*
266  * Map args from nfsmsleep() to msleep().
267  */
268 int
269 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
270     struct timespec *ts)
271 {
272 	u_int64_t nsecval;
273 	int error, timeo;
274 
275 	if (ts) {
276 		timeo = hz * ts->tv_sec;
277 		nsecval = (u_int64_t)ts->tv_nsec;
278 		nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
279 		    1000000000;
280 		timeo += (int)nsecval;
281 	} else {
282 		timeo = 0;
283 	}
284 	error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
285 	return (error);
286 }
287 
288 /*
289  * Get the file system info for the server. For now, just assume FFS.
290  */
291 void
292 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
293 {
294 	int pref;
295 
296 	/*
297 	 * XXX
298 	 * There should be file system VFS OP(s) to get this information.
299 	 * For now, assume ufs.
300 	 */
301 	if (isdgram)
302 		pref = NFS_MAXDGRAMDATA;
303 	else
304 		pref = nfs_srvmaxio;
305 	sip->fs_rtmax = nfs_srvmaxio;
306 	sip->fs_rtpref = pref;
307 	sip->fs_rtmult = NFS_FABLKSIZE;
308 	sip->fs_wtmax = nfs_srvmaxio;
309 	sip->fs_wtpref = pref;
310 	sip->fs_wtmult = NFS_FABLKSIZE;
311 	sip->fs_dtpref = pref;
312 	sip->fs_maxfilesize = 0xffffffffffffffffull;
313 	sip->fs_timedelta.tv_sec = 0;
314 	sip->fs_timedelta.tv_nsec = 1;
315 	sip->fs_properties = (NFSV3FSINFO_LINK |
316 	    NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
317 	    NFSV3FSINFO_CANSETTIME);
318 }
319 
320 /*
321  * Do the pathconf vnode op.
322  */
323 int
324 nfsvno_pathconf(struct vnode *vp, int flag, long *retf,
325     struct ucred *cred, struct thread *p)
326 {
327 	int error;
328 
329 	error = VOP_PATHCONF(vp, flag, retf);
330 	if (error == EOPNOTSUPP || error == EINVAL) {
331 		/*
332 		 * Some file systems return EINVAL for name arguments not
333 		 * supported and some return EOPNOTSUPP for this case.
334 		 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
335 		 * just fake them.
336 		 */
337 		switch (flag) {
338 		case _PC_LINK_MAX:
339 			*retf = NFS_LINK_MAX;
340 			break;
341 		case _PC_NAME_MAX:
342 			*retf = NAME_MAX;
343 			break;
344 		case _PC_CHOWN_RESTRICTED:
345 			*retf = 1;
346 			break;
347 		case _PC_NO_TRUNC:
348 			*retf = 1;
349 			break;
350 		default:
351 			/*
352 			 * Only happens if a _PC_xxx is added to the server,
353 			 * but this isn't updated.
354 			 */
355 			*retf = 0;
356 			printf("nfsrvd pathconf flag=%d not supp\n", flag);
357 		}
358 		error = 0;
359 	}
360 	NFSEXITCODE(error);
361 	return (error);
362 }
363 
364 /* Fake nfsrv_atroot. Just return 0 */
365 int
366 nfsrv_atroot(struct vnode *vp, uint64_t *retp)
367 {
368 
369 	return (0);
370 }
371 
372 /*
373  * Set the credentials to refer to root.
374  * If only the various BSDen could agree on whether cr_gid is a separate
375  * field or cr_groups[0]...
376  */
377 void
378 newnfs_setroot(struct ucred *cred)
379 {
380 
381 	cred->cr_uid = 0;
382 	cred->cr_groups[0] = 0;
383 	cred->cr_ngroups = 1;
384 }
385 
386 /*
387  * Get the client credential. Used for Renew and recovery.
388  */
389 struct ucred *
390 newnfs_getcred(void)
391 {
392 	struct ucred *cred;
393 	struct thread *td = curthread;
394 
395 	cred = crdup(td->td_ucred);
396 	newnfs_setroot(cred);
397 	return (cred);
398 }
399 
400 /*
401  * Sleep for a short period of time unless errval == NFSERR_GRACE, where
402  * the sleep should be for 5 seconds.
403  * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
404  * an event that never gets a wakeup. Only return EINTR or 0.
405  */
406 int
407 nfs_catnap(int prio, int errval, const char *wmesg)
408 {
409 	static int non_event;
410 	int ret;
411 
412 	if (errval == NFSERR_GRACE)
413 		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
414 	else
415 		ret = tsleep(&non_event, prio, wmesg, 1);
416 	if (ret != EINTR)
417 		ret = 0;
418 	return (ret);
419 }
420 
421 /*
422  * Get referral. For now, just fail.
423  */
424 struct nfsreferral *
425 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
426 {
427 
428 	return (NULL);
429 }
430 
431 static int
432 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
433 {
434 	int error;
435 
436 	NFSD_CURVNET_SET(NFSD_TD_TO_VNET(td));
437 	error = nfssvc_call(td, uap, td->td_ucred);
438 	NFSD_CURVNET_RESTORE();
439 	NFSEXITCODE(error);
440 	return (error);
441 }
442 
443 static int
444 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
445 {
446 	int error = EINVAL, i, j;
447 	struct nfsd_idargs nid;
448 	struct nfsd_oidargs onid;
449 	struct {
450 		int vers;	/* Just the first field of nfsstats. */
451 	} nfsstatver;
452 
453 	if (uap->flag & NFSSVC_IDNAME) {
454 		if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
455 			error = copyin(uap->argp, &nid, sizeof(nid));
456 		else {
457 			error = copyin(uap->argp, &onid, sizeof(onid));
458 			if (error == 0) {
459 				nid.nid_flag = onid.nid_flag;
460 				nid.nid_uid = onid.nid_uid;
461 				nid.nid_gid = onid.nid_gid;
462 				nid.nid_usermax = onid.nid_usermax;
463 				nid.nid_usertimeout = onid.nid_usertimeout;
464 				nid.nid_name = onid.nid_name;
465 				nid.nid_namelen = onid.nid_namelen;
466 				nid.nid_ngroup = 0;
467 				nid.nid_grps = NULL;
468 			}
469 		}
470 		if (error)
471 			goto out;
472 		error = nfssvc_idname(&nid);
473 		goto out;
474 	} else if (uap->flag & NFSSVC_GETSTATS) {
475 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
476 			/* Copy fields to the old ext_nfsstat structure. */
477 			oldnfsstats.attrcache_hits =
478 			    NFSD_VNET(nfsstatsv1_p)->attrcache_hits;
479 			oldnfsstats.attrcache_misses =
480 			    NFSD_VNET(nfsstatsv1_p)->attrcache_misses;
481 			oldnfsstats.lookupcache_hits =
482 			    NFSD_VNET(nfsstatsv1_p)->lookupcache_hits;
483 			oldnfsstats.lookupcache_misses =
484 			    NFSD_VNET(nfsstatsv1_p)->lookupcache_misses;
485 			oldnfsstats.direofcache_hits =
486 			    NFSD_VNET(nfsstatsv1_p)->direofcache_hits;
487 			oldnfsstats.direofcache_misses =
488 			    NFSD_VNET(nfsstatsv1_p)->direofcache_misses;
489 			oldnfsstats.accesscache_hits =
490 			    NFSD_VNET(nfsstatsv1_p)->accesscache_hits;
491 			oldnfsstats.accesscache_misses =
492 			    NFSD_VNET(nfsstatsv1_p)->accesscache_misses;
493 			oldnfsstats.biocache_reads =
494 			    NFSD_VNET(nfsstatsv1_p)->biocache_reads;
495 			oldnfsstats.read_bios =
496 			    NFSD_VNET(nfsstatsv1_p)->read_bios;
497 			oldnfsstats.read_physios =
498 			    NFSD_VNET(nfsstatsv1_p)->read_physios;
499 			oldnfsstats.biocache_writes =
500 			    NFSD_VNET(nfsstatsv1_p)->biocache_writes;
501 			oldnfsstats.write_bios =
502 			    NFSD_VNET(nfsstatsv1_p)->write_bios;
503 			oldnfsstats.write_physios =
504 			    NFSD_VNET(nfsstatsv1_p)->write_physios;
505 			oldnfsstats.biocache_readlinks =
506 			    NFSD_VNET(nfsstatsv1_p)->biocache_readlinks;
507 			oldnfsstats.readlink_bios =
508 			    NFSD_VNET(nfsstatsv1_p)->readlink_bios;
509 			oldnfsstats.biocache_readdirs =
510 			    NFSD_VNET(nfsstatsv1_p)->biocache_readdirs;
511 			oldnfsstats.readdir_bios =
512 			    NFSD_VNET(nfsstatsv1_p)->readdir_bios;
513 			for (i = 0; i < NFSV4_NPROCS; i++)
514 				oldnfsstats.rpccnt[i] =
515 				    NFSD_VNET(nfsstatsv1_p)->rpccnt[i];
516 			oldnfsstats.rpcretries =
517 			    NFSD_VNET(nfsstatsv1_p)->rpcretries;
518 			for (i = 0; i < NFSV4OP_NOPS; i++)
519 				oldnfsstats.srvrpccnt[i] =
520 				    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
521 			for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
522 			    i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
523 				oldnfsstats.srvrpccnt[j] =
524 				    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
525 			oldnfsstats.reserved_0 = 0;
526 			oldnfsstats.reserved_1 = 0;
527 			oldnfsstats.rpcrequests =
528 			    NFSD_VNET(nfsstatsv1_p)->rpcrequests;
529 			oldnfsstats.rpctimeouts =
530 			    NFSD_VNET(nfsstatsv1_p)->rpctimeouts;
531 			oldnfsstats.rpcunexpected =
532 			    NFSD_VNET(nfsstatsv1_p)->rpcunexpected;
533 			oldnfsstats.rpcinvalid =
534 			    NFSD_VNET(nfsstatsv1_p)->rpcinvalid;
535 			oldnfsstats.srvcache_inproghits =
536 			    NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits;
537 			oldnfsstats.reserved_2 = 0;
538 			oldnfsstats.srvcache_nonidemdonehits =
539 			    NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
540 			oldnfsstats.srvcache_misses =
541 			    NFSD_VNET(nfsstatsv1_p)->srvcache_misses;
542 			oldnfsstats.srvcache_tcppeak =
543 			    NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak;
544 			oldnfsstats.srvcache_size =
545 			    NFSD_VNET(nfsstatsv1_p)->srvcache_size;
546 			oldnfsstats.srvclients =
547 			    NFSD_VNET(nfsstatsv1_p)->srvclients;
548 			oldnfsstats.srvopenowners =
549 			    NFSD_VNET(nfsstatsv1_p)->srvopenowners;
550 			oldnfsstats.srvopens =
551 			    NFSD_VNET(nfsstatsv1_p)->srvopens;
552 			oldnfsstats.srvlockowners =
553 			    NFSD_VNET(nfsstatsv1_p)->srvlockowners;
554 			oldnfsstats.srvlocks =
555 			    NFSD_VNET(nfsstatsv1_p)->srvlocks;
556 			oldnfsstats.srvdelegates =
557 			    NFSD_VNET(nfsstatsv1_p)->srvdelegates;
558 			for (i = 0; i < NFSV4OP_CBNOPS; i++)
559 				oldnfsstats.cbrpccnt[i] =
560 				    NFSD_VNET(nfsstatsv1_p)->cbrpccnt[i];
561 			oldnfsstats.clopenowners =
562 			    NFSD_VNET(nfsstatsv1_p)->clopenowners;
563 			oldnfsstats.clopens = NFSD_VNET(nfsstatsv1_p)->clopens;
564 			oldnfsstats.cllockowners =
565 			    NFSD_VNET(nfsstatsv1_p)->cllockowners;
566 			oldnfsstats.cllocks = NFSD_VNET(nfsstatsv1_p)->cllocks;
567 			oldnfsstats.cldelegates =
568 			    NFSD_VNET(nfsstatsv1_p)->cldelegates;
569 			oldnfsstats.cllocalopenowners =
570 			    NFSD_VNET(nfsstatsv1_p)->cllocalopenowners;
571 			oldnfsstats.cllocalopens =
572 			    NFSD_VNET(nfsstatsv1_p)->cllocalopens;
573 			oldnfsstats.cllocallockowners =
574 			    NFSD_VNET(nfsstatsv1_p)->cllocallockowners;
575 			oldnfsstats.cllocallocks =
576 			    NFSD_VNET(nfsstatsv1_p)->cllocallocks;
577 			error = copyout(&oldnfsstats, uap->argp,
578 			    sizeof (oldnfsstats));
579 		} else {
580 			error = copyin(uap->argp, &nfsstatver,
581 			    sizeof(nfsstatver));
582 			if (error == 0) {
583 				if (nfsstatver.vers == NFSSTATS_OV1) {
584 					/* Copy nfsstatsv1 to nfsstatsov1. */
585 					nfsstatsov1.attrcache_hits =
586 					    NFSD_VNET(nfsstatsv1_p)->attrcache_hits;
587 					nfsstatsov1.attrcache_misses =
588 					    NFSD_VNET(nfsstatsv1_p)->attrcache_misses;
589 					nfsstatsov1.lookupcache_hits =
590 					    NFSD_VNET(nfsstatsv1_p)->lookupcache_hits;
591 					nfsstatsov1.lookupcache_misses =
592 					    NFSD_VNET(nfsstatsv1_p)->lookupcache_misses;
593 					nfsstatsov1.direofcache_hits =
594 					    NFSD_VNET(nfsstatsv1_p)->direofcache_hits;
595 					nfsstatsov1.direofcache_misses =
596 					    NFSD_VNET(nfsstatsv1_p)->direofcache_misses;
597 					nfsstatsov1.accesscache_hits =
598 					    NFSD_VNET(nfsstatsv1_p)->accesscache_hits;
599 					nfsstatsov1.accesscache_misses =
600 					    NFSD_VNET(nfsstatsv1_p)->accesscache_misses;
601 					nfsstatsov1.biocache_reads =
602 					    NFSD_VNET(nfsstatsv1_p)->biocache_reads;
603 					nfsstatsov1.read_bios =
604 					    NFSD_VNET(nfsstatsv1_p)->read_bios;
605 					nfsstatsov1.read_physios =
606 					    NFSD_VNET(nfsstatsv1_p)->read_physios;
607 					nfsstatsov1.biocache_writes =
608 					    NFSD_VNET(nfsstatsv1_p)->biocache_writes;
609 					nfsstatsov1.write_bios =
610 					    NFSD_VNET(nfsstatsv1_p)->write_bios;
611 					nfsstatsov1.write_physios =
612 					    NFSD_VNET(nfsstatsv1_p)->write_physios;
613 					nfsstatsov1.biocache_readlinks =
614 					    NFSD_VNET(nfsstatsv1_p)->biocache_readlinks;
615 					nfsstatsov1.readlink_bios =
616 					    NFSD_VNET(nfsstatsv1_p)->readlink_bios;
617 					nfsstatsov1.biocache_readdirs =
618 					    NFSD_VNET(nfsstatsv1_p)->biocache_readdirs;
619 					nfsstatsov1.readdir_bios =
620 					    NFSD_VNET(nfsstatsv1_p)->readdir_bios;
621 					for (i = 0; i < NFSV42_OLDNPROCS; i++)
622 						nfsstatsov1.rpccnt[i] =
623 						    NFSD_VNET(nfsstatsv1_p)->rpccnt[i];
624 					nfsstatsov1.rpcretries =
625 					    NFSD_VNET(nfsstatsv1_p)->rpcretries;
626 					for (i = 0; i < NFSV42_PURENOPS; i++)
627 						nfsstatsov1.srvrpccnt[i] =
628 						    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
629 					for (i = NFSV42_NOPS,
630 					     j = NFSV42_PURENOPS;
631 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
632 					     i++, j++)
633 						nfsstatsov1.srvrpccnt[j] =
634 						    NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
635 					nfsstatsov1.reserved_0 = 0;
636 					nfsstatsov1.reserved_1 = 0;
637 					nfsstatsov1.rpcrequests =
638 					    NFSD_VNET(nfsstatsv1_p)->rpcrequests;
639 					nfsstatsov1.rpctimeouts =
640 					    NFSD_VNET(nfsstatsv1_p)->rpctimeouts;
641 					nfsstatsov1.rpcunexpected =
642 					    NFSD_VNET(nfsstatsv1_p)->rpcunexpected;
643 					nfsstatsov1.rpcinvalid =
644 					    NFSD_VNET(nfsstatsv1_p)->rpcinvalid;
645 					nfsstatsov1.srvcache_inproghits =
646 					    NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits;
647 					nfsstatsov1.reserved_2 = 0;
648 					nfsstatsov1.srvcache_nonidemdonehits =
649 					    NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
650 					nfsstatsov1.srvcache_misses =
651 					    NFSD_VNET(nfsstatsv1_p)->srvcache_misses;
652 					nfsstatsov1.srvcache_tcppeak =
653 					    NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak;
654 					nfsstatsov1.srvcache_size =
655 					    NFSD_VNET(nfsstatsv1_p)->srvcache_size;
656 					nfsstatsov1.srvclients =
657 					    NFSD_VNET(nfsstatsv1_p)->srvclients;
658 					nfsstatsov1.srvopenowners =
659 					    NFSD_VNET(nfsstatsv1_p)->srvopenowners;
660 					nfsstatsov1.srvopens =
661 					    NFSD_VNET(nfsstatsv1_p)->srvopens;
662 					nfsstatsov1.srvlockowners =
663 					    NFSD_VNET(nfsstatsv1_p)->srvlockowners;
664 					nfsstatsov1.srvlocks =
665 					    NFSD_VNET(nfsstatsv1_p)->srvlocks;
666 					nfsstatsov1.srvdelegates =
667 					    NFSD_VNET(nfsstatsv1_p)->srvdelegates;
668 					for (i = 0; i < NFSV42_CBNOPS; i++)
669 						nfsstatsov1.cbrpccnt[i] =
670 						    NFSD_VNET(nfsstatsv1_p)->cbrpccnt[i];
671 					nfsstatsov1.clopenowners =
672 					    NFSD_VNET(nfsstatsv1_p)->clopenowners;
673 					nfsstatsov1.clopens =
674 					    NFSD_VNET(nfsstatsv1_p)->clopens;
675 					nfsstatsov1.cllockowners =
676 					    NFSD_VNET(nfsstatsv1_p)->cllockowners;
677 					nfsstatsov1.cllocks =
678 					    NFSD_VNET(nfsstatsv1_p)->cllocks;
679 					nfsstatsov1.cldelegates =
680 					    NFSD_VNET(nfsstatsv1_p)->cldelegates;
681 					nfsstatsov1.cllocalopenowners =
682 					    NFSD_VNET(nfsstatsv1_p)->cllocalopenowners;
683 					nfsstatsov1.cllocalopens =
684 					    NFSD_VNET(nfsstatsv1_p)->cllocalopens;
685 					nfsstatsov1.cllocallockowners =
686 					    NFSD_VNET(nfsstatsv1_p)->cllocallockowners;
687 					nfsstatsov1.cllocallocks =
688 					    NFSD_VNET(nfsstatsv1_p)->cllocallocks;
689 					nfsstatsov1.srvstartcnt =
690 					    NFSD_VNET(nfsstatsv1_p)->srvstartcnt;
691 					nfsstatsov1.srvdonecnt =
692 					    NFSD_VNET(nfsstatsv1_p)->srvdonecnt;
693 					for (i = NFSV42_NOPS,
694 					     j = NFSV42_PURENOPS;
695 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
696 					     i++, j++) {
697 						nfsstatsov1.srvbytes[j] =
698 						    NFSD_VNET(nfsstatsv1_p)->srvbytes[i];
699 						nfsstatsov1.srvops[j] =
700 						    NFSD_VNET(nfsstatsv1_p)->srvops[i];
701 						nfsstatsov1.srvduration[j] =
702 						    NFSD_VNET(nfsstatsv1_p)->srvduration[i];
703 					}
704 					nfsstatsov1.busyfrom =
705 					    NFSD_VNET(nfsstatsv1_p)->busyfrom;
706 					nfsstatsov1.busyfrom =
707 					    NFSD_VNET(nfsstatsv1_p)->busyfrom;
708 					error = copyout(&nfsstatsov1, uap->argp,
709 					    sizeof(nfsstatsov1));
710 				} else if (nfsstatver.vers != NFSSTATS_V1)
711 					error = EPERM;
712 				else
713 					error = copyout(NFSD_VNET(nfsstatsv1_p),
714 					    uap->argp, sizeof(nfsstatsv1));
715 			}
716 		}
717 		if (error == 0) {
718 			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
719 				NFSD_VNET(nfsstatsv1_p)->attrcache_hits = 0;
720 				NFSD_VNET(nfsstatsv1_p)->attrcache_misses = 0;
721 				NFSD_VNET(nfsstatsv1_p)->lookupcache_hits = 0;
722 				NFSD_VNET(nfsstatsv1_p)->lookupcache_misses = 0;
723 				NFSD_VNET(nfsstatsv1_p)->direofcache_hits = 0;
724 				NFSD_VNET(nfsstatsv1_p)->direofcache_misses = 0;
725 				NFSD_VNET(nfsstatsv1_p)->accesscache_hits = 0;
726 				NFSD_VNET(nfsstatsv1_p)->accesscache_misses = 0;
727 				NFSD_VNET(nfsstatsv1_p)->biocache_reads = 0;
728 				NFSD_VNET(nfsstatsv1_p)->read_bios = 0;
729 				NFSD_VNET(nfsstatsv1_p)->read_physios = 0;
730 				NFSD_VNET(nfsstatsv1_p)->biocache_writes = 0;
731 				NFSD_VNET(nfsstatsv1_p)->write_bios = 0;
732 				NFSD_VNET(nfsstatsv1_p)->write_physios = 0;
733 				NFSD_VNET(nfsstatsv1_p)->biocache_readlinks = 0;
734 				NFSD_VNET(nfsstatsv1_p)->readlink_bios = 0;
735 				NFSD_VNET(nfsstatsv1_p)->biocache_readdirs = 0;
736 				NFSD_VNET(nfsstatsv1_p)->readdir_bios = 0;
737 				NFSD_VNET(nfsstatsv1_p)->rpcretries = 0;
738 				NFSD_VNET(nfsstatsv1_p)->rpcrequests = 0;
739 				NFSD_VNET(nfsstatsv1_p)->rpctimeouts = 0;
740 				NFSD_VNET(nfsstatsv1_p)->rpcunexpected = 0;
741 				NFSD_VNET(nfsstatsv1_p)->rpcinvalid = 0;
742 				bzero(NFSD_VNET(nfsstatsv1_p)->rpccnt,
743 				    sizeof(NFSD_VNET(nfsstatsv1_p)->rpccnt));
744 			}
745 			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
746 				NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits = 0;
747 				NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits = 0;
748 				NFSD_VNET(nfsstatsv1_p)->srvcache_misses = 0;
749 				NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak = 0;
750 				bzero(NFSD_VNET(nfsstatsv1_p)->srvrpccnt,
751 				    sizeof(NFSD_VNET(nfsstatsv1_p)->srvrpccnt));
752 				bzero(NFSD_VNET(nfsstatsv1_p)->cbrpccnt,
753 				    sizeof(NFSD_VNET(nfsstatsv1_p)->cbrpccnt));
754 			}
755 		}
756 		goto out;
757 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
758 		u_short sockport;
759 		struct nfsuserd_args nargs;
760 
761 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
762 			error = copyin(uap->argp, (caddr_t)&sockport,
763 			    sizeof (u_short));
764 			if (error == 0) {
765 				nargs.nuserd_family = AF_INET;
766 				nargs.nuserd_port = sockport;
767 			}
768 		} else {
769 			/*
770 			 * New nfsuserd_args structure, which indicates
771 			 * which IP version to use along with the port#.
772 			 */
773 			error = copyin(uap->argp, &nargs, sizeof(nargs));
774 		}
775 		if (!error)
776 			error = nfsrv_nfsuserdport(&nargs, p);
777 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
778 		nfsrv_nfsuserddelport();
779 		error = 0;
780 	}
781 
782 out:
783 	NFSEXITCODE(error);
784 	return (error);
785 }
786 
787 /*
788  * called by all three modevent routines, so that it gets things
789  * initialized soon enough.
790  */
791 void
792 newnfs_portinit(void)
793 {
794 	static int inited = 0;
795 
796 	if (inited)
797 		return;
798 	inited = 1;
799 	/* Initialize SMP locks used by both client and server. */
800 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
801 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
802 	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
803 }
804 
805 /*
806  * Determine if the file system supports NFSv4 ACLs.
807  * Return 1 if it does, 0 otherwise.
808  */
809 int
810 nfs_supportsnfsv4acls(struct vnode *vp)
811 {
812 	int error;
813 	long retval;
814 
815 	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
816 
817 	if (nfsrv_useacl == 0)
818 		return (0);
819 	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
820 	if (error == 0 && retval != 0)
821 		return (1);
822 	return (0);
823 }
824 
825 /*
826  * These are the first fields of all the context structures passed into
827  * nfs_pnfsio().
828  */
829 struct pnfsio {
830 	int		done;
831 	int		inprog;
832 	struct task	tsk;
833 };
834 
835 /*
836  * Do a mirror I/O on a pNFS thread.
837  */
838 int
839 nfs_pnfsio(task_fn_t *func, void *context)
840 {
841 	struct pnfsio *pio;
842 	int ret;
843 	static struct taskqueue *pnfsioq = NULL;
844 
845 	pio = (struct pnfsio *)context;
846 	if (pnfsioq == NULL) {
847 		if (nfs_pnfsiothreads == 0)
848 			return (EPERM);
849 		if (nfs_pnfsiothreads < 0)
850 			nfs_pnfsiothreads = mp_ncpus * 4;
851 		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
852 		    taskqueue_thread_enqueue, &pnfsioq);
853 		if (pnfsioq == NULL)
854 			return (ENOMEM);
855 		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
856 		    0, "pnfsiot");
857 		if (ret != 0) {
858 			taskqueue_free(pnfsioq);
859 			pnfsioq = NULL;
860 			return (ret);
861 		}
862 	}
863 	pio->inprog = 1;
864 	TASK_INIT(&pio->tsk, 0, func, context);
865 	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
866 	if (ret != 0)
867 		pio->inprog = 0;
868 	return (ret);
869 }
870 
871 /*
872  * Initialize everything that needs to be initialized for a vnet.
873  */
874 static void
875 nfs_vnetinit(const void *unused __unused)
876 {
877 
878 	if (IS_DEFAULT_VNET(curvnet))
879 		NFSD_VNET(nfsstatsv1_p) = &nfsstatsv1;
880 	else
881 		NFSD_VNET(nfsstatsv1_p) = malloc(sizeof(struct nfsstatsv1),
882 		    M_TEMP, M_WAITOK | M_ZERO);
883 	mtx_init(&NFSD_VNET(nfsrv_nfsuserdsock).nr_mtx, "nfsuserd",
884 	    NULL, MTX_DEF);
885 }
886 VNET_SYSINIT(nfs_vnetinit, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
887     nfs_vnetinit, NULL);
888 
889 static void
890 nfs_cleanup(void *unused __unused)
891 {
892 
893 	mtx_destroy(&NFSD_VNET(nfsrv_nfsuserdsock).nr_mtx);
894 	if (!IS_DEFAULT_VNET(curvnet)) {
895 		free(NFSD_VNET(nfsstatsv1_p), M_TEMP);
896 		NFSD_VNET(nfsstatsv1_p) = NULL;
897 	}
898 	/* Clean out the name<-->id cache. */
899 	nfsrv_cleanusergroup();
900 }
901 VNET_SYSUNINIT(nfs_cleanup, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
902     nfs_cleanup, NULL);
903 
904 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
905 
906 /*
907  * Called once to initialize data structures...
908  */
909 static int
910 nfscommon_modevent(module_t mod, int type, void *data)
911 {
912 	int error = 0;
913 	static int loaded = 0;
914 
915 	switch (type) {
916 	case MOD_LOAD:
917 		if (loaded)
918 			goto out;
919 		newnfs_portinit();
920 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
921 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
922 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
923 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
924 		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
925 		TAILQ_INIT(&nfsrv_devidhead);
926 		newnfs_init();
927 		nfsd_call_nfscommon = nfssvc_nfscommon;
928 		loaded = 1;
929 		break;
930 
931 	case MOD_UNLOAD:
932 		if (newnfs_numnfsd != 0 ||
933 		    NFSD_VNET(nfsrv_nfsuserd) != NOTRUNNING ||
934 		    nfs_numnfscbd != 0) {
935 			error = EBUSY;
936 			break;
937 		}
938 
939 		nfsd_call_nfscommon = NULL;
940 		/* and get rid of the mutexes */
941 		mtx_destroy(&nfs_nameid_mutex);
942 		mtx_destroy(&newnfsd_mtx);
943 		mtx_destroy(&nfs_state_mutex);
944 		mtx_destroy(&nfs_clstate_mutex);
945 		mtx_destroy(&nfs_sockl_mutex);
946 		mtx_destroy(&nfs_slock_mutex);
947 		mtx_destroy(&nfs_req_mutex);
948 		mtx_destroy(&nfsrv_dslock_mtx);
949 		loaded = 0;
950 		break;
951 	default:
952 		error = EOPNOTSUPP;
953 		break;
954 	}
955 
956 out:
957 	NFSEXITCODE(error);
958 	return error;
959 }
960 static moduledata_t nfscommon_mod = {
961 	"nfscommon",
962 	nfscommon_modevent,
963 	NULL,
964 };
965 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
966 
967 /* So that loader and kldload(2) can find us, wherever we are.. */
968 MODULE_VERSION(nfscommon, 1);
969 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
970 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
971