xref: /freebsd-13.1/sys/fs/nfs/nfs_commonport.c (revision 3540f6b4)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Rick Macklem at The University of Guelph.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  */
35 
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 /*
40  * Functions that need to be different for different versions of BSD
41  * kernel should be kept here, along with any global storage specific
42  * to this BSD variant.
43  */
44 #include <fs/nfs/nfsport.h>
45 #include <sys/smp.h>
46 #include <sys/sysctl.h>
47 #include <sys/taskqueue.h>
48 #include <rpc/rpc_com.h>
49 #include <vm/vm.h>
50 #include <vm/vm_object.h>
51 #include <vm/vm_page.h>
52 #include <vm/vm_param.h>
53 #include <vm/vm_map.h>
54 #include <vm/vm_kern.h>
55 #include <vm/vm_extern.h>
56 #include <vm/uma.h>
57 
58 extern int nfscl_ticks;
59 extern nfsuserd_state nfsrv_nfsuserd;
60 extern struct nfssockreq nfsrv_nfsuserdsock;
61 extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
62     struct thread *);
63 extern int nfsrv_useacl;
64 struct mount nfsv4root_mnt;
65 int newnfs_numnfsd = 0;
66 struct nfsstatsv1 nfsstatsv1;
67 int nfs_numnfscbd = 0;
68 int nfscl_debuglevel = 0;
69 char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
70 int nfsrv_lughashsize = 100;
71 struct mtx nfsrv_dslock_mtx;
72 struct nfsdevicehead nfsrv_devidhead;
73 volatile int nfsrv_devidcnt = 0;
74 void (*ncl_call_invalcaches)(struct vnode *) = NULL;
75 vop_advlock_t *nfs_advlock_p = NULL;
76 vop_reclaim_t *nfs_reclaim_p = NULL;
77 uint32_t nfs_srvmaxio = NFS_SRVMAXIO;
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_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer");
125 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroff",
126     "NFS directory offset data");
127 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
128     "NFS local lock rollback");
129 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
130 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
131 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
132 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
133 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
134 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
135 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
136 
137 /*
138  * Definition of mutex locks.
139  * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
140  * and assorted other nfsd structures.
141  */
142 struct mtx newnfsd_mtx;
143 struct mtx nfs_sockl_mutex;
144 struct mtx nfs_state_mutex;
145 struct mtx nfs_nameid_mutex;
146 struct mtx nfs_req_mutex;
147 struct mtx nfs_slock_mutex;
148 struct mtx nfs_clstate_mutex;
149 
150 /* local functions */
151 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
152 
153 #ifdef __NO_STRICT_ALIGNMENT
154 /*
155  * These architectures don't need re-alignment, so just return.
156  */
157 int
newnfs_realign(struct mbuf ** pm,int how)158 newnfs_realign(struct mbuf **pm, int how)
159 {
160 
161 	return (0);
162 }
163 #else	/* !__NO_STRICT_ALIGNMENT */
164 /*
165  *	newnfs_realign:
166  *
167  *	Check for badly aligned mbuf data and realign by copying the unaligned
168  *	portion of the data into a new mbuf chain and freeing the portions
169  *	of the old chain that were replaced.
170  *
171  *	We cannot simply realign the data within the existing mbuf chain
172  *	because the underlying buffers may contain other rpc commands and
173  *	we cannot afford to overwrite them.
174  *
175  *	We would prefer to avoid this situation entirely.  The situation does
176  *	not occur with NFS/UDP and is supposed to only occasionally occur
177  *	with TCP.  Use vfs.nfs.realign_count and realign_test to check this.
178  *
179  */
180 int
newnfs_realign(struct mbuf ** pm,int how)181 newnfs_realign(struct mbuf **pm, int how)
182 {
183 	struct mbuf *m, *n;
184 	int off, space;
185 
186 	++nfs_realign_test;
187 	while ((m = *pm) != NULL) {
188 		if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
189 			/*
190 			 * NB: we can't depend on m_pkthdr.len to help us
191 			 * decide what to do here.  May not be worth doing
192 			 * the m_length calculation as m_copyback will
193 			 * expand the mbuf chain below as needed.
194 			 */
195 			space = m_length(m, NULL);
196 			if (space >= MINCLSIZE) {
197 				/* NB: m_copyback handles space > MCLBYTES */
198 				n = m_getcl(how, MT_DATA, 0);
199 			} else
200 				n = m_get(how, MT_DATA);
201 			if (n == NULL)
202 				return (ENOMEM);
203 			/*
204 			 * Align the remainder of the mbuf chain.
205 			 */
206 			n->m_len = 0;
207 			off = 0;
208 			while (m != NULL) {
209 				m_copyback(n, off, m->m_len, mtod(m, caddr_t));
210 				off += m->m_len;
211 				m = m->m_next;
212 			}
213 			m_freem(*pm);
214 			*pm = n;
215 			++nfs_realign_count;
216 			break;
217 		}
218 		pm = &m->m_next;
219 	}
220 
221 	return (0);
222 }
223 #endif	/* __NO_STRICT_ALIGNMENT */
224 
225 #ifdef notdef
226 static void
nfsrv_object_create(struct vnode * vp,struct thread * td)227 nfsrv_object_create(struct vnode *vp, struct thread *td)
228 {
229 
230 	if (vp == NULL || vp->v_type != VREG)
231 		return;
232 	(void) vfs_object_create(vp, td, td->td_ucred);
233 }
234 #endif
235 
236 /*
237  * Look up a file name. Basically just initialize stuff and call namei().
238  */
239 int
nfsrv_lookupfilename(struct nameidata * ndp,char * fname,NFSPROC_T * p)240 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
241 {
242 	int error;
243 
244 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
245 	    p);
246 	error = namei(ndp);
247 	if (!error) {
248 		NDFREE(ndp, NDF_ONLY_PNBUF);
249 	}
250 	return (error);
251 }
252 
253 /*
254  * Copy NFS uid, gids to the cred structure.
255  */
256 void
newnfs_copycred(struct nfscred * nfscr,struct ucred * cr)257 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
258 {
259 
260 	KASSERT(nfscr->nfsc_ngroups >= 0,
261 	    ("newnfs_copycred: negative nfsc_ngroups"));
262 	cr->cr_uid = nfscr->nfsc_uid;
263 	crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
264 }
265 
266 /*
267  * Map args from nfsmsleep() to msleep().
268  */
269 int
nfsmsleep(void * chan,void * mutex,int prio,const char * wmesg,struct timespec * ts)270 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
271     struct timespec *ts)
272 {
273 	u_int64_t nsecval;
274 	int error, timeo;
275 
276 	if (ts) {
277 		timeo = hz * ts->tv_sec;
278 		nsecval = (u_int64_t)ts->tv_nsec;
279 		nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
280 		    1000000000;
281 		timeo += (int)nsecval;
282 	} else {
283 		timeo = 0;
284 	}
285 	error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
286 	return (error);
287 }
288 
289 /*
290  * Get the file system info for the server. For now, just assume FFS.
291  */
292 void
nfsvno_getfs(struct nfsfsinfo * sip,int isdgram)293 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
294 {
295 	int pref;
296 
297 	/*
298 	 * XXX
299 	 * There should be file system VFS OP(s) to get this information.
300 	 * For now, assume ufs.
301 	 */
302 	if (isdgram)
303 		pref = NFS_MAXDGRAMDATA;
304 	else
305 		pref = nfs_srvmaxio;
306 	sip->fs_rtmax = nfs_srvmaxio;
307 	sip->fs_rtpref = pref;
308 	sip->fs_rtmult = NFS_FABLKSIZE;
309 	sip->fs_wtmax = nfs_srvmaxio;
310 	sip->fs_wtpref = pref;
311 	sip->fs_wtmult = NFS_FABLKSIZE;
312 	sip->fs_dtpref = pref;
313 	sip->fs_maxfilesize = 0xffffffffffffffffull;
314 	sip->fs_timedelta.tv_sec = 0;
315 	sip->fs_timedelta.tv_nsec = 1;
316 	sip->fs_properties = (NFSV3FSINFO_LINK |
317 	    NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
318 	    NFSV3FSINFO_CANSETTIME);
319 }
320 
321 /*
322  * Do the pathconf vnode op.
323  */
324 int
nfsvno_pathconf(struct vnode * vp,int flag,long * retf,struct ucred * cred,struct thread * p)325 nfsvno_pathconf(struct vnode *vp, int flag, long *retf,
326     struct ucred *cred, struct thread *p)
327 {
328 	int error;
329 
330 	error = VOP_PATHCONF(vp, flag, retf);
331 	if (error == EOPNOTSUPP || error == EINVAL) {
332 		/*
333 		 * Some file systems return EINVAL for name arguments not
334 		 * supported and some return EOPNOTSUPP for this case.
335 		 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
336 		 * just fake them.
337 		 */
338 		switch (flag) {
339 		case _PC_LINK_MAX:
340 			*retf = NFS_LINK_MAX;
341 			break;
342 		case _PC_NAME_MAX:
343 			*retf = NAME_MAX;
344 			break;
345 		case _PC_CHOWN_RESTRICTED:
346 			*retf = 1;
347 			break;
348 		case _PC_NO_TRUNC:
349 			*retf = 1;
350 			break;
351 		default:
352 			/*
353 			 * Only happens if a _PC_xxx is added to the server,
354 			 * but this isn't updated.
355 			 */
356 			*retf = 0;
357 			printf("nfsrvd pathconf flag=%d not supp\n", flag);
358 		}
359 		error = 0;
360 	}
361 	NFSEXITCODE(error);
362 	return (error);
363 }
364 
365 /* Fake nfsrv_atroot. Just return 0 */
366 int
nfsrv_atroot(struct vnode * vp,uint64_t * retp)367 nfsrv_atroot(struct vnode *vp, uint64_t *retp)
368 {
369 
370 	return (0);
371 }
372 
373 /*
374  * Set the credentials to refer to root.
375  * If only the various BSDen could agree on whether cr_gid is a separate
376  * field or cr_groups[0]...
377  */
378 void
newnfs_setroot(struct ucred * cred)379 newnfs_setroot(struct ucred *cred)
380 {
381 
382 	cred->cr_uid = 0;
383 	cred->cr_groups[0] = 0;
384 	cred->cr_ngroups = 1;
385 }
386 
387 /*
388  * Get the client credential. Used for Renew and recovery.
389  */
390 struct ucred *
newnfs_getcred(void)391 newnfs_getcred(void)
392 {
393 	struct ucred *cred;
394 	struct thread *td = curthread;
395 
396 	cred = crdup(td->td_ucred);
397 	newnfs_setroot(cred);
398 	return (cred);
399 }
400 
401 /*
402  * Sleep for a short period of time unless errval == NFSERR_GRACE, where
403  * the sleep should be for 5 seconds.
404  * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
405  * an event that never gets a wakeup. Only return EINTR or 0.
406  */
407 int
nfs_catnap(int prio,int errval,const char * wmesg)408 nfs_catnap(int prio, int errval, const char *wmesg)
409 {
410 	static int non_event;
411 	int ret;
412 
413 	if (errval == NFSERR_GRACE)
414 		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
415 	else
416 		ret = tsleep(&non_event, prio, wmesg, 1);
417 	if (ret != EINTR)
418 		ret = 0;
419 	return (ret);
420 }
421 
422 /*
423  * Get referral. For now, just fail.
424  */
425 struct nfsreferral *
nfsv4root_getreferral(struct vnode * vp,struct vnode * dvp,u_int32_t fileno)426 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
427 {
428 
429 	return (NULL);
430 }
431 
432 static int
nfssvc_nfscommon(struct thread * td,struct nfssvc_args * uap)433 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
434 {
435 	int error;
436 
437 	error = nfssvc_call(td, uap, td->td_ucred);
438 	NFSEXITCODE(error);
439 	return (error);
440 }
441 
442 static int
nfssvc_call(struct thread * p,struct nfssvc_args * uap,struct ucred * cred)443 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
444 {
445 	int error = EINVAL, i, j;
446 	struct nfsd_idargs nid;
447 	struct nfsd_oidargs onid;
448 	struct {
449 		int vers;	/* Just the first field of nfsstats. */
450 	} nfsstatver;
451 
452 	if (uap->flag & NFSSVC_IDNAME) {
453 		if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
454 			error = copyin(uap->argp, &nid, sizeof(nid));
455 		else {
456 			error = copyin(uap->argp, &onid, sizeof(onid));
457 			if (error == 0) {
458 				nid.nid_flag = onid.nid_flag;
459 				nid.nid_uid = onid.nid_uid;
460 				nid.nid_gid = onid.nid_gid;
461 				nid.nid_usermax = onid.nid_usermax;
462 				nid.nid_usertimeout = onid.nid_usertimeout;
463 				nid.nid_name = onid.nid_name;
464 				nid.nid_namelen = onid.nid_namelen;
465 				nid.nid_ngroup = 0;
466 				nid.nid_grps = NULL;
467 			}
468 		}
469 		if (error)
470 			goto out;
471 		error = nfssvc_idname(&nid);
472 		goto out;
473 	} else if (uap->flag & NFSSVC_GETSTATS) {
474 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
475 			/* Copy fields to the old ext_nfsstat structure. */
476 			oldnfsstats.attrcache_hits =
477 			    nfsstatsv1.attrcache_hits;
478 			oldnfsstats.attrcache_misses =
479 			    nfsstatsv1.attrcache_misses;
480 			oldnfsstats.lookupcache_hits =
481 			    nfsstatsv1.lookupcache_hits;
482 			oldnfsstats.lookupcache_misses =
483 			    nfsstatsv1.lookupcache_misses;
484 			oldnfsstats.direofcache_hits =
485 			    nfsstatsv1.direofcache_hits;
486 			oldnfsstats.direofcache_misses =
487 			    nfsstatsv1.direofcache_misses;
488 			oldnfsstats.accesscache_hits =
489 			    nfsstatsv1.accesscache_hits;
490 			oldnfsstats.accesscache_misses =
491 			    nfsstatsv1.accesscache_misses;
492 			oldnfsstats.biocache_reads =
493 			    nfsstatsv1.biocache_reads;
494 			oldnfsstats.read_bios =
495 			    nfsstatsv1.read_bios;
496 			oldnfsstats.read_physios =
497 			    nfsstatsv1.read_physios;
498 			oldnfsstats.biocache_writes =
499 			    nfsstatsv1.biocache_writes;
500 			oldnfsstats.write_bios =
501 			    nfsstatsv1.write_bios;
502 			oldnfsstats.write_physios =
503 			    nfsstatsv1.write_physios;
504 			oldnfsstats.biocache_readlinks =
505 			    nfsstatsv1.biocache_readlinks;
506 			oldnfsstats.readlink_bios =
507 			    nfsstatsv1.readlink_bios;
508 			oldnfsstats.biocache_readdirs =
509 			    nfsstatsv1.biocache_readdirs;
510 			oldnfsstats.readdir_bios =
511 			    nfsstatsv1.readdir_bios;
512 			for (i = 0; i < NFSV4_NPROCS; i++)
513 				oldnfsstats.rpccnt[i] = nfsstatsv1.rpccnt[i];
514 			oldnfsstats.rpcretries = nfsstatsv1.rpcretries;
515 			for (i = 0; i < NFSV4OP_NOPS; i++)
516 				oldnfsstats.srvrpccnt[i] =
517 				    nfsstatsv1.srvrpccnt[i];
518 			for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
519 			    i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
520 				oldnfsstats.srvrpccnt[j] =
521 				    nfsstatsv1.srvrpccnt[i];
522 			oldnfsstats.reserved_0 = 0;
523 			oldnfsstats.reserved_1 = 0;
524 			oldnfsstats.rpcrequests = nfsstatsv1.rpcrequests;
525 			oldnfsstats.rpctimeouts = nfsstatsv1.rpctimeouts;
526 			oldnfsstats.rpcunexpected = nfsstatsv1.rpcunexpected;
527 			oldnfsstats.rpcinvalid = nfsstatsv1.rpcinvalid;
528 			oldnfsstats.srvcache_inproghits =
529 			    nfsstatsv1.srvcache_inproghits;
530 			oldnfsstats.reserved_2 = 0;
531 			oldnfsstats.srvcache_nonidemdonehits =
532 			    nfsstatsv1.srvcache_nonidemdonehits;
533 			oldnfsstats.srvcache_misses =
534 			    nfsstatsv1.srvcache_misses;
535 			oldnfsstats.srvcache_tcppeak =
536 			    nfsstatsv1.srvcache_tcppeak;
537 			oldnfsstats.srvcache_size = nfsstatsv1.srvcache_size;
538 			oldnfsstats.srvclients = nfsstatsv1.srvclients;
539 			oldnfsstats.srvopenowners = nfsstatsv1.srvopenowners;
540 			oldnfsstats.srvopens = nfsstatsv1.srvopens;
541 			oldnfsstats.srvlockowners = nfsstatsv1.srvlockowners;
542 			oldnfsstats.srvlocks = nfsstatsv1.srvlocks;
543 			oldnfsstats.srvdelegates = nfsstatsv1.srvdelegates;
544 			for (i = 0; i < NFSV4OP_CBNOPS; i++)
545 				oldnfsstats.cbrpccnt[i] =
546 				    nfsstatsv1.cbrpccnt[i];
547 			oldnfsstats.clopenowners = nfsstatsv1.clopenowners;
548 			oldnfsstats.clopens = nfsstatsv1.clopens;
549 			oldnfsstats.cllockowners = nfsstatsv1.cllockowners;
550 			oldnfsstats.cllocks = nfsstatsv1.cllocks;
551 			oldnfsstats.cldelegates = nfsstatsv1.cldelegates;
552 			oldnfsstats.cllocalopenowners =
553 			    nfsstatsv1.cllocalopenowners;
554 			oldnfsstats.cllocalopens = nfsstatsv1.cllocalopens;
555 			oldnfsstats.cllocallockowners =
556 			    nfsstatsv1.cllocallockowners;
557 			oldnfsstats.cllocallocks = nfsstatsv1.cllocallocks;
558 			error = copyout(&oldnfsstats, uap->argp,
559 			    sizeof (oldnfsstats));
560 		} else {
561 			error = copyin(uap->argp, &nfsstatver,
562 			    sizeof(nfsstatver));
563 			if (error == 0) {
564 				if (nfsstatver.vers == NFSSTATS_OV1) {
565 					/* Copy nfsstatsv1 to nfsstatsov1. */
566 					nfsstatsov1.attrcache_hits =
567 					    nfsstatsv1.attrcache_hits;
568 					nfsstatsov1.attrcache_misses =
569 					    nfsstatsv1.attrcache_misses;
570 					nfsstatsov1.lookupcache_hits =
571 					    nfsstatsv1.lookupcache_hits;
572 					nfsstatsov1.lookupcache_misses =
573 					    nfsstatsv1.lookupcache_misses;
574 					nfsstatsov1.direofcache_hits =
575 					    nfsstatsv1.direofcache_hits;
576 					nfsstatsov1.direofcache_misses =
577 					    nfsstatsv1.direofcache_misses;
578 					nfsstatsov1.accesscache_hits =
579 					    nfsstatsv1.accesscache_hits;
580 					nfsstatsov1.accesscache_misses =
581 					    nfsstatsv1.accesscache_misses;
582 					nfsstatsov1.biocache_reads =
583 					    nfsstatsv1.biocache_reads;
584 					nfsstatsov1.read_bios =
585 					    nfsstatsv1.read_bios;
586 					nfsstatsov1.read_physios =
587 					    nfsstatsv1.read_physios;
588 					nfsstatsov1.biocache_writes =
589 					    nfsstatsv1.biocache_writes;
590 					nfsstatsov1.write_bios =
591 					    nfsstatsv1.write_bios;
592 					nfsstatsov1.write_physios =
593 					    nfsstatsv1.write_physios;
594 					nfsstatsov1.biocache_readlinks =
595 					    nfsstatsv1.biocache_readlinks;
596 					nfsstatsov1.readlink_bios =
597 					    nfsstatsv1.readlink_bios;
598 					nfsstatsov1.biocache_readdirs =
599 					    nfsstatsv1.biocache_readdirs;
600 					nfsstatsov1.readdir_bios =
601 					    nfsstatsv1.readdir_bios;
602 					for (i = 0; i < NFSV42_NPROCS; i++)
603 						nfsstatsov1.rpccnt[i] =
604 						    nfsstatsv1.rpccnt[i];
605 					nfsstatsov1.rpcretries =
606 					    nfsstatsv1.rpcretries;
607 					for (i = 0; i < NFSV42_PURENOPS; i++)
608 						nfsstatsov1.srvrpccnt[i] =
609 						    nfsstatsv1.srvrpccnt[i];
610 					for (i = NFSV42_NOPS,
611 					     j = NFSV42_PURENOPS;
612 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
613 					     i++, j++)
614 						nfsstatsov1.srvrpccnt[j] =
615 						    nfsstatsv1.srvrpccnt[i];
616 					nfsstatsov1.reserved_0 = 0;
617 					nfsstatsov1.reserved_1 = 0;
618 					nfsstatsov1.rpcrequests =
619 					    nfsstatsv1.rpcrequests;
620 					nfsstatsov1.rpctimeouts =
621 					    nfsstatsv1.rpctimeouts;
622 					nfsstatsov1.rpcunexpected =
623 					    nfsstatsv1.rpcunexpected;
624 					nfsstatsov1.rpcinvalid =
625 					    nfsstatsv1.rpcinvalid;
626 					nfsstatsov1.srvcache_inproghits =
627 					    nfsstatsv1.srvcache_inproghits;
628 					nfsstatsov1.reserved_2 = 0;
629 					nfsstatsov1.srvcache_nonidemdonehits =
630 					    nfsstatsv1.srvcache_nonidemdonehits;
631 					nfsstatsov1.srvcache_misses =
632 					    nfsstatsv1.srvcache_misses;
633 					nfsstatsov1.srvcache_tcppeak =
634 					    nfsstatsv1.srvcache_tcppeak;
635 					nfsstatsov1.srvcache_size =
636 					    nfsstatsv1.srvcache_size;
637 					nfsstatsov1.srvclients =
638 					    nfsstatsv1.srvclients;
639 					nfsstatsov1.srvopenowners =
640 					    nfsstatsv1.srvopenowners;
641 					nfsstatsov1.srvopens =
642 					    nfsstatsv1.srvopens;
643 					nfsstatsov1.srvlockowners =
644 					    nfsstatsv1.srvlockowners;
645 					nfsstatsov1.srvlocks =
646 					    nfsstatsv1.srvlocks;
647 					nfsstatsov1.srvdelegates =
648 					    nfsstatsv1.srvdelegates;
649 					for (i = 0; i < NFSV42_CBNOPS; i++)
650 						nfsstatsov1.cbrpccnt[i] =
651 						    nfsstatsv1.cbrpccnt[i];
652 					nfsstatsov1.clopenowners =
653 					    nfsstatsv1.clopenowners;
654 					nfsstatsov1.clopens =
655 					    nfsstatsv1.clopens;
656 					nfsstatsov1.cllockowners =
657 					    nfsstatsv1.cllockowners;
658 					nfsstatsov1.cllocks =
659 					    nfsstatsv1.cllocks;
660 					nfsstatsov1.cldelegates =
661 					    nfsstatsv1.cldelegates;
662 					nfsstatsov1.cllocalopenowners =
663 					    nfsstatsv1.cllocalopenowners;
664 					nfsstatsov1.cllocalopens =
665 					    nfsstatsv1.cllocalopens;
666 					nfsstatsov1.cllocallockowners =
667 					    nfsstatsv1.cllocallockowners;
668 					nfsstatsov1.cllocallocks =
669 					    nfsstatsv1.cllocallocks;
670 					nfsstatsov1.srvstartcnt =
671 					    nfsstatsv1.srvstartcnt;
672 					nfsstatsov1.srvdonecnt =
673 					    nfsstatsv1.srvdonecnt;
674 					for (i = NFSV42_NOPS,
675 					     j = NFSV42_PURENOPS;
676 					     i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
677 					     i++, j++) {
678 						nfsstatsov1.srvbytes[j] =
679 						    nfsstatsv1.srvbytes[i];
680 						nfsstatsov1.srvops[j] =
681 						    nfsstatsv1.srvops[i];
682 						nfsstatsov1.srvduration[j] =
683 						    nfsstatsv1.srvduration[i];
684 					}
685 					nfsstatsov1.busyfrom =
686 					    nfsstatsv1.busyfrom;
687 					nfsstatsov1.busyfrom =
688 					    nfsstatsv1.busyfrom;
689 					error = copyout(&nfsstatsov1, uap->argp,
690 					    sizeof(nfsstatsov1));
691 				} else if (nfsstatver.vers != NFSSTATS_V1)
692 					error = EPERM;
693 				else
694 					error = copyout(&nfsstatsv1, uap->argp,
695 					    sizeof(nfsstatsv1));
696 			}
697 		}
698 		if (error == 0) {
699 			if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
700 				nfsstatsv1.attrcache_hits = 0;
701 				nfsstatsv1.attrcache_misses = 0;
702 				nfsstatsv1.lookupcache_hits = 0;
703 				nfsstatsv1.lookupcache_misses = 0;
704 				nfsstatsv1.direofcache_hits = 0;
705 				nfsstatsv1.direofcache_misses = 0;
706 				nfsstatsv1.accesscache_hits = 0;
707 				nfsstatsv1.accesscache_misses = 0;
708 				nfsstatsv1.biocache_reads = 0;
709 				nfsstatsv1.read_bios = 0;
710 				nfsstatsv1.read_physios = 0;
711 				nfsstatsv1.biocache_writes = 0;
712 				nfsstatsv1.write_bios = 0;
713 				nfsstatsv1.write_physios = 0;
714 				nfsstatsv1.biocache_readlinks = 0;
715 				nfsstatsv1.readlink_bios = 0;
716 				nfsstatsv1.biocache_readdirs = 0;
717 				nfsstatsv1.readdir_bios = 0;
718 				nfsstatsv1.rpcretries = 0;
719 				nfsstatsv1.rpcrequests = 0;
720 				nfsstatsv1.rpctimeouts = 0;
721 				nfsstatsv1.rpcunexpected = 0;
722 				nfsstatsv1.rpcinvalid = 0;
723 				bzero(nfsstatsv1.rpccnt,
724 				    sizeof(nfsstatsv1.rpccnt));
725 			}
726 			if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
727 				nfsstatsv1.srvcache_inproghits = 0;
728 				nfsstatsv1.srvcache_nonidemdonehits = 0;
729 				nfsstatsv1.srvcache_misses = 0;
730 				nfsstatsv1.srvcache_tcppeak = 0;
731 				bzero(nfsstatsv1.srvrpccnt,
732 				    sizeof(nfsstatsv1.srvrpccnt));
733 				bzero(nfsstatsv1.cbrpccnt,
734 				    sizeof(nfsstatsv1.cbrpccnt));
735 			}
736 		}
737 		goto out;
738 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
739 		u_short sockport;
740 		struct nfsuserd_args nargs;
741 
742 		if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
743 			error = copyin(uap->argp, (caddr_t)&sockport,
744 			    sizeof (u_short));
745 			if (error == 0) {
746 				nargs.nuserd_family = AF_INET;
747 				nargs.nuserd_port = sockport;
748 			}
749 		} else {
750 			/*
751 			 * New nfsuserd_args structure, which indicates
752 			 * which IP version to use along with the port#.
753 			 */
754 			error = copyin(uap->argp, &nargs, sizeof(nargs));
755 		}
756 		if (!error)
757 			error = nfsrv_nfsuserdport(&nargs, p);
758 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
759 		nfsrv_nfsuserddelport();
760 		error = 0;
761 	}
762 
763 out:
764 	NFSEXITCODE(error);
765 	return (error);
766 }
767 
768 /*
769  * called by all three modevent routines, so that it gets things
770  * initialized soon enough.
771  */
772 void
newnfs_portinit(void)773 newnfs_portinit(void)
774 {
775 	static int inited = 0;
776 
777 	if (inited)
778 		return;
779 	inited = 1;
780 	/* Initialize SMP locks used by both client and server. */
781 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
782 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
783 	mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
784 }
785 
786 /*
787  * Determine if the file system supports NFSv4 ACLs.
788  * Return 1 if it does, 0 otherwise.
789  */
790 int
nfs_supportsnfsv4acls(struct vnode * vp)791 nfs_supportsnfsv4acls(struct vnode *vp)
792 {
793 	int error;
794 	long retval;
795 
796 	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
797 
798 	if (nfsrv_useacl == 0)
799 		return (0);
800 	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
801 	if (error == 0 && retval != 0)
802 		return (1);
803 	return (0);
804 }
805 
806 /*
807  * These are the first fields of all the context structures passed into
808  * nfs_pnfsio().
809  */
810 struct pnfsio {
811 	int		done;
812 	int		inprog;
813 	struct task	tsk;
814 };
815 
816 /*
817  * Do a mirror I/O on a pNFS thread.
818  */
819 int
nfs_pnfsio(task_fn_t * func,void * context)820 nfs_pnfsio(task_fn_t *func, void *context)
821 {
822 	struct pnfsio *pio;
823 	int ret;
824 	static struct taskqueue *pnfsioq = NULL;
825 
826 	pio = (struct pnfsio *)context;
827 	if (pnfsioq == NULL) {
828 		if (nfs_pnfsiothreads == 0)
829 			return (EPERM);
830 		if (nfs_pnfsiothreads < 0)
831 			nfs_pnfsiothreads = mp_ncpus * 4;
832 		pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
833 		    taskqueue_thread_enqueue, &pnfsioq);
834 		if (pnfsioq == NULL)
835 			return (ENOMEM);
836 		ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
837 		    0, "pnfsiot");
838 		if (ret != 0) {
839 			taskqueue_free(pnfsioq);
840 			pnfsioq = NULL;
841 			return (ret);
842 		}
843 	}
844 	pio->inprog = 1;
845 	TASK_INIT(&pio->tsk, 0, func, context);
846 	ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
847 	if (ret != 0)
848 		pio->inprog = 0;
849 	return (ret);
850 }
851 
852 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
853 
854 /*
855  * Called once to initialize data structures...
856  */
857 static int
nfscommon_modevent(module_t mod,int type,void * data)858 nfscommon_modevent(module_t mod, int type, void *data)
859 {
860 	int error = 0;
861 	static int loaded = 0;
862 
863 	switch (type) {
864 	case MOD_LOAD:
865 		if (loaded)
866 			goto out;
867 		newnfs_portinit();
868 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
869 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
870 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
871 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
872 		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
873 		    MTX_DEF);
874 		mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
875 		TAILQ_INIT(&nfsrv_devidhead);
876 		newnfs_init();
877 		nfsd_call_nfscommon = nfssvc_nfscommon;
878 		loaded = 1;
879 		break;
880 
881 	case MOD_UNLOAD:
882 		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != NOTRUNNING ||
883 		    nfs_numnfscbd != 0) {
884 			error = EBUSY;
885 			break;
886 		}
887 
888 		nfsd_call_nfscommon = NULL;
889 		/* Clean out the name<-->id cache. */
890 		nfsrv_cleanusergroup();
891 		/* and get rid of the mutexes */
892 		mtx_destroy(&nfs_nameid_mutex);
893 		mtx_destroy(&newnfsd_mtx);
894 		mtx_destroy(&nfs_state_mutex);
895 		mtx_destroy(&nfs_clstate_mutex);
896 		mtx_destroy(&nfs_sockl_mutex);
897 		mtx_destroy(&nfs_slock_mutex);
898 		mtx_destroy(&nfs_req_mutex);
899 		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
900 		mtx_destroy(&nfsrv_dslock_mtx);
901 		loaded = 0;
902 		break;
903 	default:
904 		error = EOPNOTSUPP;
905 		break;
906 	}
907 
908 out:
909 	NFSEXITCODE(error);
910 	return error;
911 }
912 static moduledata_t nfscommon_mod = {
913 	"nfscommon",
914 	nfscommon_modevent,
915 	NULL,
916 };
917 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
918 
919 /* So that loader and kldload(2) can find us, wherever we are.. */
920 MODULE_VERSION(nfscommon, 1);
921 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
922 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
923