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
58extern int nfscl_ticks;
59extern nfsuserd_state nfsrv_nfsuserd;
60extern struct nfssockreq nfsrv_nfsuserdsock;
61extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
62    struct thread *);
63extern int nfsrv_useacl;
64struct mount nfsv4root_mnt;
65int newnfs_numnfsd = 0;
66struct nfsstatsv1 nfsstatsv1;
67int nfs_numnfscbd = 0;
68int nfscl_debuglevel = 0;
69char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
70struct callout newnfsd_callout;
71int nfsrv_lughashsize = 100;
72struct mtx nfsrv_dslock_mtx;
73struct nfsdevicehead nfsrv_devidhead;
74volatile int nfsrv_devidcnt = 0;
75void (*nfsd_call_servertimer)(void) = NULL;
76void (*ncl_call_invalcaches)(struct vnode *) = NULL;
77vop_advlock_t *nfs_advlock_p = NULL;
78vop_reclaim_t *nfs_reclaim_p = NULL;
79
80int nfs_pnfsio(task_fn_t *, void *);
81
82static int nfs_realign_test;
83static int nfs_realign_count;
84static struct ext_nfsstats oldnfsstats;
85static struct nfsstatsov1 nfsstatsov1;
86
87SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
88    "NFS filesystem");
89SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
90    0, "Number of realign tests done");
91SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
92    0, "Number of mbuf realignments done");
93SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
94    nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
95    "NFSv4 callback addr for server to use");
96SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
97    0, "Debug level for NFS client");
98SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
99    0, "Size of hash tables for uid/name mapping");
100int nfs_pnfsiothreads = -1;
101SYSCTL_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 */
108MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
109MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
110MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
111    "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
112MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
113MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
114MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
115MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
116MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
117MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
118MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
119MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
120MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
121MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
122MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
123MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
124MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
125MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer");
126MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroff",
127    "NFS directory offset data");
128MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
129    "NFS local lock rollback");
130MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
131MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
132MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
133MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
134MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
135MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
136MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
137
138/*
139 * Definition of mutex locks.
140 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
141 * and assorted other nfsd structures.
142 */
143struct mtx newnfsd_mtx;
144struct mtx nfs_sockl_mutex;
145struct mtx nfs_state_mutex;
146struct mtx nfs_nameid_mutex;
147struct mtx nfs_req_mutex;
148struct mtx nfs_slock_mutex;
149struct mtx nfs_clstate_mutex;
150
151/* local functions */
152static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
153
154#ifdef __NO_STRICT_ALIGNMENT
155/*
156 * These architectures don't need re-alignment, so just return.
157 */
158int
159newnfs_realign(struct mbuf **pm, int how)
160{
161
162	return (0);
163}
164#else	/* !__NO_STRICT_ALIGNMENT */
165/*
166 *	newnfs_realign:
167 *
168 *	Check for badly aligned mbuf data and realign by copying the unaligned
169 *	portion of the data into a new mbuf chain and freeing the portions
170 *	of the old chain that were replaced.
171 *
172 *	We cannot simply realign the data within the existing mbuf chain
173 *	because the underlying buffers may contain other rpc commands and
174 *	we cannot afford to overwrite them.
175 *
176 *	We would prefer to avoid this situation entirely.  The situation does
177 *	not occur with NFS/UDP and is supposed to only occasionally occur
178 *	with TCP.  Use vfs.nfs.realign_count and realign_test to check this.
179 *
180 */
181int
182newnfs_realign(struct mbuf **pm, int how)
183{
184	struct mbuf *m, *n;
185	int off, space;
186
187	++nfs_realign_test;
188	while ((m = *pm) != NULL) {
189		if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
190			/*
191			 * NB: we can't depend on m_pkthdr.len to help us
192			 * decide what to do here.  May not be worth doing
193			 * the m_length calculation as m_copyback will
194			 * expand the mbuf chain below as needed.
195			 */
196			space = m_length(m, NULL);
197			if (space >= MINCLSIZE) {
198				/* NB: m_copyback handles space > MCLBYTES */
199				n = m_getcl(how, MT_DATA, 0);
200			} else
201				n = m_get(how, MT_DATA);
202			if (n == NULL)
203				return (ENOMEM);
204			/*
205			 * Align the remainder of the mbuf chain.
206			 */
207			n->m_len = 0;
208			off = 0;
209			while (m != NULL) {
210				m_copyback(n, off, m->m_len, mtod(m, caddr_t));
211				off += m->m_len;
212				m = m->m_next;
213			}
214			m_freem(*pm);
215			*pm = n;
216			++nfs_realign_count;
217			break;
218		}
219		pm = &m->m_next;
220	}
221
222	return (0);
223}
224#endif	/* __NO_STRICT_ALIGNMENT */
225
226#ifdef notdef
227static void
228nfsrv_object_create(struct vnode *vp, struct thread *td)
229{
230
231	if (vp == NULL || vp->v_type != VREG)
232		return;
233	(void) vfs_object_create(vp, td, td->td_ucred);
234}
235#endif
236
237/*
238 * Look up a file name. Basically just initialize stuff and call namei().
239 */
240int
241nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
242{
243	int error;
244
245	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
246	    p);
247	error = namei(ndp);
248	if (!error) {
249		NDFREE(ndp, NDF_ONLY_PNBUF);
250	}
251	return (error);
252}
253
254/*
255 * Copy NFS uid, gids to the cred structure.
256 */
257void
258newnfs_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 */
270int
271nfsmsleep(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 */
293void
294nfsvno_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 */
325int
326nfsvno_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 */
367int
368nfsrv_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 */
379void
380newnfs_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 */
391struct ucred *
392newnfs_getcred(void)
393{
394	struct ucred *cred;
395	struct thread *td = curthread;
396
397	cred = crdup(td->td_ucred);
398	newnfs_setroot(cred);
399	return (cred);
400}
401
402/*
403 * Nfs timer routine
404 * Call the nfsd's timer function once/sec.
405 */
406void
407newnfs_timer(void *arg)
408{
409	static time_t lasttime = 0;
410	/*
411	 * Call the server timer, if set up.
412	 * The argument indicates if it is the next second and therefore
413	 * leases should be checked.
414	 */
415	if (lasttime != NFSD_MONOSEC) {
416		lasttime = NFSD_MONOSEC;
417		if (nfsd_call_servertimer != NULL)
418			(*nfsd_call_servertimer)();
419	}
420	callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
421}
422
423/*
424 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
425 * the sleep should be for 5 seconds.
426 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
427 * an event that never gets a wakeup. Only return EINTR or 0.
428 */
429int
430nfs_catnap(int prio, int errval, const char *wmesg)
431{
432	static int non_event;
433	int ret;
434
435	if (errval == NFSERR_GRACE)
436		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
437	else
438		ret = tsleep(&non_event, prio, wmesg, 1);
439	if (ret != EINTR)
440		ret = 0;
441	return (ret);
442}
443
444/*
445 * Get referral. For now, just fail.
446 */
447struct nfsreferral *
448nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
449{
450
451	return (NULL);
452}
453
454static int
455nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
456{
457	int error;
458
459	error = nfssvc_call(td, uap, td->td_ucred);
460	NFSEXITCODE(error);
461	return (error);
462}
463
464static int
465nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
466{
467	int error = EINVAL, i, j;
468	struct nfsd_idargs nid;
469	struct nfsd_oidargs onid;
470	struct {
471		int vers;	/* Just the first field of nfsstats. */
472	} nfsstatver;
473
474	if (uap->flag & NFSSVC_IDNAME) {
475		if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
476			error = copyin(uap->argp, &nid, sizeof(nid));
477		else {
478			error = copyin(uap->argp, &onid, sizeof(onid));
479			if (error == 0) {
480				nid.nid_flag = onid.nid_flag;
481				nid.nid_uid = onid.nid_uid;
482				nid.nid_gid = onid.nid_gid;
483				nid.nid_usermax = onid.nid_usermax;
484				nid.nid_usertimeout = onid.nid_usertimeout;
485				nid.nid_name = onid.nid_name;
486				nid.nid_namelen = onid.nid_namelen;
487				nid.nid_ngroup = 0;
488				nid.nid_grps = NULL;
489			}
490		}
491		if (error)
492			goto out;
493		error = nfssvc_idname(&nid);
494		goto out;
495	} else if (uap->flag & NFSSVC_GETSTATS) {
496		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
497			/* Copy fields to the old ext_nfsstat structure. */
498			oldnfsstats.attrcache_hits =
499			    nfsstatsv1.attrcache_hits;
500			oldnfsstats.attrcache_misses =
501			    nfsstatsv1.attrcache_misses;
502			oldnfsstats.lookupcache_hits =
503			    nfsstatsv1.lookupcache_hits;
504			oldnfsstats.lookupcache_misses =
505			    nfsstatsv1.lookupcache_misses;
506			oldnfsstats.direofcache_hits =
507			    nfsstatsv1.direofcache_hits;
508			oldnfsstats.direofcache_misses =
509			    nfsstatsv1.direofcache_misses;
510			oldnfsstats.accesscache_hits =
511			    nfsstatsv1.accesscache_hits;
512			oldnfsstats.accesscache_misses =
513			    nfsstatsv1.accesscache_misses;
514			oldnfsstats.biocache_reads =
515			    nfsstatsv1.biocache_reads;
516			oldnfsstats.read_bios =
517			    nfsstatsv1.read_bios;
518			oldnfsstats.read_physios =
519			    nfsstatsv1.read_physios;
520			oldnfsstats.biocache_writes =
521			    nfsstatsv1.biocache_writes;
522			oldnfsstats.write_bios =
523			    nfsstatsv1.write_bios;
524			oldnfsstats.write_physios =
525			    nfsstatsv1.write_physios;
526			oldnfsstats.biocache_readlinks =
527			    nfsstatsv1.biocache_readlinks;
528			oldnfsstats.readlink_bios =
529			    nfsstatsv1.readlink_bios;
530			oldnfsstats.biocache_readdirs =
531			    nfsstatsv1.biocache_readdirs;
532			oldnfsstats.readdir_bios =
533			    nfsstatsv1.readdir_bios;
534			for (i = 0; i < NFSV4_NPROCS; i++)
535				oldnfsstats.rpccnt[i] = nfsstatsv1.rpccnt[i];
536			oldnfsstats.rpcretries = nfsstatsv1.rpcretries;
537			for (i = 0; i < NFSV4OP_NOPS; i++)
538				oldnfsstats.srvrpccnt[i] =
539				    nfsstatsv1.srvrpccnt[i];
540			for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
541			    i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
542				oldnfsstats.srvrpccnt[j] =
543				    nfsstatsv1.srvrpccnt[i];
544			oldnfsstats.reserved_0 = 0;
545			oldnfsstats.reserved_1 = 0;
546			oldnfsstats.rpcrequests = nfsstatsv1.rpcrequests;
547			oldnfsstats.rpctimeouts = nfsstatsv1.rpctimeouts;
548			oldnfsstats.rpcunexpected = nfsstatsv1.rpcunexpected;
549			oldnfsstats.rpcinvalid = nfsstatsv1.rpcinvalid;
550			oldnfsstats.srvcache_inproghits =
551			    nfsstatsv1.srvcache_inproghits;
552			oldnfsstats.reserved_2 = 0;
553			oldnfsstats.srvcache_nonidemdonehits =
554			    nfsstatsv1.srvcache_nonidemdonehits;
555			oldnfsstats.srvcache_misses =
556			    nfsstatsv1.srvcache_misses;
557			oldnfsstats.srvcache_tcppeak =
558			    nfsstatsv1.srvcache_tcppeak;
559			oldnfsstats.srvcache_size = nfsstatsv1.srvcache_size;
560			oldnfsstats.srvclients = nfsstatsv1.srvclients;
561			oldnfsstats.srvopenowners = nfsstatsv1.srvopenowners;
562			oldnfsstats.srvopens = nfsstatsv1.srvopens;
563			oldnfsstats.srvlockowners = nfsstatsv1.srvlockowners;
564			oldnfsstats.srvlocks = nfsstatsv1.srvlocks;
565			oldnfsstats.srvdelegates = nfsstatsv1.srvdelegates;
566			for (i = 0; i < NFSV4OP_CBNOPS; i++)
567				oldnfsstats.cbrpccnt[i] =
568				    nfsstatsv1.cbrpccnt[i];
569			oldnfsstats.clopenowners = nfsstatsv1.clopenowners;
570			oldnfsstats.clopens = nfsstatsv1.clopens;
571			oldnfsstats.cllockowners = nfsstatsv1.cllockowners;
572			oldnfsstats.cllocks = nfsstatsv1.cllocks;
573			oldnfsstats.cldelegates = nfsstatsv1.cldelegates;
574			oldnfsstats.cllocalopenowners =
575			    nfsstatsv1.cllocalopenowners;
576			oldnfsstats.cllocalopens = nfsstatsv1.cllocalopens;
577			oldnfsstats.cllocallockowners =
578			    nfsstatsv1.cllocallockowners;
579			oldnfsstats.cllocallocks = nfsstatsv1.cllocallocks;
580			error = copyout(&oldnfsstats, uap->argp,
581			    sizeof (oldnfsstats));
582		} else {
583			error = copyin(uap->argp, &nfsstatver,
584			    sizeof(nfsstatver));
585			if (error == 0) {
586				if (nfsstatver.vers == NFSSTATS_OV1) {
587					/* Copy nfsstatsv1 to nfsstatsov1. */
588					nfsstatsov1.attrcache_hits =
589					    nfsstatsv1.attrcache_hits;
590					nfsstatsov1.attrcache_misses =
591					    nfsstatsv1.attrcache_misses;
592					nfsstatsov1.lookupcache_hits =
593					    nfsstatsv1.lookupcache_hits;
594					nfsstatsov1.lookupcache_misses =
595					    nfsstatsv1.lookupcache_misses;
596					nfsstatsov1.direofcache_hits =
597					    nfsstatsv1.direofcache_hits;
598					nfsstatsov1.direofcache_misses =
599					    nfsstatsv1.direofcache_misses;
600					nfsstatsov1.accesscache_hits =
601					    nfsstatsv1.accesscache_hits;
602					nfsstatsov1.accesscache_misses =
603					    nfsstatsv1.accesscache_misses;
604					nfsstatsov1.biocache_reads =
605					    nfsstatsv1.biocache_reads;
606					nfsstatsov1.read_bios =
607					    nfsstatsv1.read_bios;
608					nfsstatsov1.read_physios =
609					    nfsstatsv1.read_physios;
610					nfsstatsov1.biocache_writes =
611					    nfsstatsv1.biocache_writes;
612					nfsstatsov1.write_bios =
613					    nfsstatsv1.write_bios;
614					nfsstatsov1.write_physios =
615					    nfsstatsv1.write_physios;
616					nfsstatsov1.biocache_readlinks =
617					    nfsstatsv1.biocache_readlinks;
618					nfsstatsov1.readlink_bios =
619					    nfsstatsv1.readlink_bios;
620					nfsstatsov1.biocache_readdirs =
621					    nfsstatsv1.biocache_readdirs;
622					nfsstatsov1.readdir_bios =
623					    nfsstatsv1.readdir_bios;
624					for (i = 0; i < NFSV42_NPROCS; i++)
625						nfsstatsov1.rpccnt[i] =
626						    nfsstatsv1.rpccnt[i];
627					nfsstatsov1.rpcretries =
628					    nfsstatsv1.rpcretries;
629					for (i = 0; i < NFSV42_PURENOPS; i++)
630						nfsstatsov1.srvrpccnt[i] =
631						    nfsstatsv1.srvrpccnt[i];
632					for (i = NFSV42_NOPS,
633					     j = NFSV42_PURENOPS;
634					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
635					     i++, j++)
636						nfsstatsov1.srvrpccnt[j] =
637						    nfsstatsv1.srvrpccnt[i];
638					nfsstatsov1.reserved_0 = 0;
639					nfsstatsov1.reserved_1 = 0;
640					nfsstatsov1.rpcrequests =
641					    nfsstatsv1.rpcrequests;
642					nfsstatsov1.rpctimeouts =
643					    nfsstatsv1.rpctimeouts;
644					nfsstatsov1.rpcunexpected =
645					    nfsstatsv1.rpcunexpected;
646					nfsstatsov1.rpcinvalid =
647					    nfsstatsv1.rpcinvalid;
648					nfsstatsov1.srvcache_inproghits =
649					    nfsstatsv1.srvcache_inproghits;
650					nfsstatsov1.reserved_2 = 0;
651					nfsstatsov1.srvcache_nonidemdonehits =
652					    nfsstatsv1.srvcache_nonidemdonehits;
653					nfsstatsov1.srvcache_misses =
654					    nfsstatsv1.srvcache_misses;
655					nfsstatsov1.srvcache_tcppeak =
656					    nfsstatsv1.srvcache_tcppeak;
657					nfsstatsov1.srvcache_size =
658					    nfsstatsv1.srvcache_size;
659					nfsstatsov1.srvclients =
660					    nfsstatsv1.srvclients;
661					nfsstatsov1.srvopenowners =
662					    nfsstatsv1.srvopenowners;
663					nfsstatsov1.srvopens =
664					    nfsstatsv1.srvopens;
665					nfsstatsov1.srvlockowners =
666					    nfsstatsv1.srvlockowners;
667					nfsstatsov1.srvlocks =
668					    nfsstatsv1.srvlocks;
669					nfsstatsov1.srvdelegates =
670					    nfsstatsv1.srvdelegates;
671					for (i = 0; i < NFSV42_CBNOPS; i++)
672						nfsstatsov1.cbrpccnt[i] =
673						    nfsstatsv1.cbrpccnt[i];
674					nfsstatsov1.clopenowners =
675					    nfsstatsv1.clopenowners;
676					nfsstatsov1.clopens =
677					    nfsstatsv1.clopens;
678					nfsstatsov1.cllockowners =
679					    nfsstatsv1.cllockowners;
680					nfsstatsov1.cllocks =
681					    nfsstatsv1.cllocks;
682					nfsstatsov1.cldelegates =
683					    nfsstatsv1.cldelegates;
684					nfsstatsov1.cllocalopenowners =
685					    nfsstatsv1.cllocalopenowners;
686					nfsstatsov1.cllocalopens =
687					    nfsstatsv1.cllocalopens;
688					nfsstatsov1.cllocallockowners =
689					    nfsstatsv1.cllocallockowners;
690					nfsstatsov1.cllocallocks =
691					    nfsstatsv1.cllocallocks;
692					nfsstatsov1.srvstartcnt =
693					    nfsstatsv1.srvstartcnt;
694					nfsstatsov1.srvdonecnt =
695					    nfsstatsv1.srvdonecnt;
696					for (i = NFSV42_NOPS,
697					     j = NFSV42_PURENOPS;
698					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
699					     i++, j++) {
700						nfsstatsov1.srvbytes[j] =
701						    nfsstatsv1.srvbytes[i];
702						nfsstatsov1.srvops[j] =
703						    nfsstatsv1.srvops[i];
704						nfsstatsov1.srvduration[j] =
705						    nfsstatsv1.srvduration[i];
706					}
707					nfsstatsov1.busyfrom =
708					    nfsstatsv1.busyfrom;
709					nfsstatsov1.busyfrom =
710					    nfsstatsv1.busyfrom;
711					error = copyout(&nfsstatsov1, uap->argp,
712					    sizeof(nfsstatsov1));
713				} else if (nfsstatver.vers != NFSSTATS_V1)
714					error = EPERM;
715				else
716					error = copyout(&nfsstatsv1, uap->argp,
717					    sizeof(nfsstatsv1));
718			}
719		}
720		if (error == 0) {
721			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
722				nfsstatsv1.attrcache_hits = 0;
723				nfsstatsv1.attrcache_misses = 0;
724				nfsstatsv1.lookupcache_hits = 0;
725				nfsstatsv1.lookupcache_misses = 0;
726				nfsstatsv1.direofcache_hits = 0;
727				nfsstatsv1.direofcache_misses = 0;
728				nfsstatsv1.accesscache_hits = 0;
729				nfsstatsv1.accesscache_misses = 0;
730				nfsstatsv1.biocache_reads = 0;
731				nfsstatsv1.read_bios = 0;
732				nfsstatsv1.read_physios = 0;
733				nfsstatsv1.biocache_writes = 0;
734				nfsstatsv1.write_bios = 0;
735				nfsstatsv1.write_physios = 0;
736				nfsstatsv1.biocache_readlinks = 0;
737				nfsstatsv1.readlink_bios = 0;
738				nfsstatsv1.biocache_readdirs = 0;
739				nfsstatsv1.readdir_bios = 0;
740				nfsstatsv1.rpcretries = 0;
741				nfsstatsv1.rpcrequests = 0;
742				nfsstatsv1.rpctimeouts = 0;
743				nfsstatsv1.rpcunexpected = 0;
744				nfsstatsv1.rpcinvalid = 0;
745				bzero(nfsstatsv1.rpccnt,
746				    sizeof(nfsstatsv1.rpccnt));
747			}
748			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
749				nfsstatsv1.srvcache_inproghits = 0;
750				nfsstatsv1.srvcache_nonidemdonehits = 0;
751				nfsstatsv1.srvcache_misses = 0;
752				nfsstatsv1.srvcache_tcppeak = 0;
753				bzero(nfsstatsv1.srvrpccnt,
754				    sizeof(nfsstatsv1.srvrpccnt));
755				bzero(nfsstatsv1.cbrpccnt,
756				    sizeof(nfsstatsv1.cbrpccnt));
757			}
758		}
759		goto out;
760	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
761		u_short sockport;
762		struct nfsuserd_args nargs;
763
764		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
765			error = copyin(uap->argp, (caddr_t)&sockport,
766			    sizeof (u_short));
767			if (error == 0) {
768				nargs.nuserd_family = AF_INET;
769				nargs.nuserd_port = sockport;
770			}
771		} else {
772			/*
773			 * New nfsuserd_args structure, which indicates
774			 * which IP version to use along with the port#.
775			 */
776			error = copyin(uap->argp, &nargs, sizeof(nargs));
777		}
778		if (!error)
779			error = nfsrv_nfsuserdport(&nargs, p);
780	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
781		nfsrv_nfsuserddelport();
782		error = 0;
783	}
784
785out:
786	NFSEXITCODE(error);
787	return (error);
788}
789
790/*
791 * called by all three modevent routines, so that it gets things
792 * initialized soon enough.
793 */
794void
795newnfs_portinit(void)
796{
797	static int inited = 0;
798
799	if (inited)
800		return;
801	inited = 1;
802	/* Initialize SMP locks used by both client and server. */
803	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
804	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
805	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
806}
807
808/*
809 * Determine if the file system supports NFSv4 ACLs.
810 * Return 1 if it does, 0 otherwise.
811 */
812int
813nfs_supportsnfsv4acls(struct vnode *vp)
814{
815	int error;
816	long retval;
817
818	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
819
820	if (nfsrv_useacl == 0)
821		return (0);
822	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
823	if (error == 0 && retval != 0)
824		return (1);
825	return (0);
826}
827
828/*
829 * These are the first fields of all the context structures passed into
830 * nfs_pnfsio().
831 */
832struct pnfsio {
833	int		done;
834	int		inprog;
835	struct task	tsk;
836};
837
838/*
839 * Do a mirror I/O on a pNFS thread.
840 */
841int
842nfs_pnfsio(task_fn_t *func, void *context)
843{
844	struct pnfsio *pio;
845	int ret;
846	static struct taskqueue *pnfsioq = NULL;
847
848	pio = (struct pnfsio *)context;
849	if (pnfsioq == NULL) {
850		if (nfs_pnfsiothreads == 0)
851			return (EPERM);
852		if (nfs_pnfsiothreads < 0)
853			nfs_pnfsiothreads = mp_ncpus * 4;
854		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
855		    taskqueue_thread_enqueue, &pnfsioq);
856		if (pnfsioq == NULL)
857			return (ENOMEM);
858		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
859		    0, "pnfsiot");
860		if (ret != 0) {
861			taskqueue_free(pnfsioq);
862			pnfsioq = NULL;
863			return (ret);
864		}
865	}
866	pio->inprog = 1;
867	TASK_INIT(&pio->tsk, 0, func, context);
868	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
869	if (ret != 0)
870		pio->inprog = 0;
871	return (ret);
872}
873
874extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
875
876/*
877 * Called once to initialize data structures...
878 */
879static int
880nfscommon_modevent(module_t mod, int type, void *data)
881{
882	int error = 0;
883	static int loaded = 0;
884
885	switch (type) {
886	case MOD_LOAD:
887		if (loaded)
888			goto out;
889		newnfs_portinit();
890		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
891		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
892		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
893		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
894		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
895		    MTX_DEF);
896		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
897		TAILQ_INIT(&nfsrv_devidhead);
898		callout_init(&newnfsd_callout, 1);
899		newnfs_init();
900		nfsd_call_nfscommon = nfssvc_nfscommon;
901		loaded = 1;
902		break;
903
904	case MOD_UNLOAD:
905		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != NOTRUNNING ||
906		    nfs_numnfscbd != 0) {
907			error = EBUSY;
908			break;
909		}
910
911		nfsd_call_nfscommon = NULL;
912		callout_drain(&newnfsd_callout);
913		/* Clean out the name<-->id cache. */
914		nfsrv_cleanusergroup();
915		/* and get rid of the mutexes */
916		mtx_destroy(&nfs_nameid_mutex);
917		mtx_destroy(&newnfsd_mtx);
918		mtx_destroy(&nfs_state_mutex);
919		mtx_destroy(&nfs_clstate_mutex);
920		mtx_destroy(&nfs_sockl_mutex);
921		mtx_destroy(&nfs_slock_mutex);
922		mtx_destroy(&nfs_req_mutex);
923		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
924		mtx_destroy(&nfsrv_dslock_mtx);
925		loaded = 0;
926		break;
927	default:
928		error = EOPNOTSUPP;
929		break;
930	}
931
932out:
933	NFSEXITCODE(error);
934	return error;
935}
936static moduledata_t nfscommon_mod = {
937	"nfscommon",
938	nfscommon_modevent,
939	NULL,
940};
941DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
942
943/* So that loader and kldload(2) can find us, wherever we are.. */
944MODULE_VERSION(nfscommon, 1);
945MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
946MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
947