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
newnfs_realign(struct mbuf ** pm,int how)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
newnfs_realign(struct mbuf ** pm,int how)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
nfsrv_object_create(struct vnode * vp,struct thread * td)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
nfsrv_lookupfilename(struct nameidata * ndp,char * fname,NFSPROC_T * p __unused)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
newnfs_copycred(struct nfscred * nfscr,struct ucred * cr)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_and_egid(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups,
262 GID_NOGROUP);
263 }
264
265 /*
266 * Map args from nfsmsleep() to msleep().
267 */
268 int
nfsmsleep(void * chan,void * mutex,int prio,const char * wmesg,struct timespec * ts)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
nfsvno_getfs(struct nfsfsinfo * sip,int isdgram)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
nfsvno_pathconf(struct vnode * vp,int flag,long * retf,struct ucred * cred,struct thread * p)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
nfsrv_atroot(struct vnode * vp,uint64_t * retp)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 */
375 void
newnfs_setroot(struct ucred * cred)376 newnfs_setroot(struct ucred *cred)
377 {
378
379 cred->cr_uid = 0;
380 cred->cr_gid = 0;
381 cred->cr_ngroups = 0;
382 }
383
384 /*
385 * Get the client credential. Used for Renew and recovery.
386 */
387 struct ucred *
newnfs_getcred(void)388 newnfs_getcred(void)
389 {
390 struct ucred *cred;
391 struct thread *td = curthread;
392
393 cred = crdup(td->td_ucred);
394 newnfs_setroot(cred);
395 return (cred);
396 }
397
398 /*
399 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
400 * the sleep should be for 5 seconds.
401 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
402 * an event that never gets a wakeup. Only return EINTR or 0.
403 */
404 int
nfs_catnap(int prio,int errval,const char * wmesg)405 nfs_catnap(int prio, int errval, const char *wmesg)
406 {
407 static int non_event;
408 int ret;
409
410 if (errval == NFSERR_GRACE)
411 ret = tsleep(&non_event, prio, wmesg, 5 * hz);
412 else
413 ret = tsleep(&non_event, prio, wmesg, 1);
414 if (ret != EINTR)
415 ret = 0;
416 return (ret);
417 }
418
419 /*
420 * Get referral. For now, just fail.
421 */
422 struct nfsreferral *
nfsv4root_getreferral(struct vnode * vp,struct vnode * dvp,u_int32_t fileno)423 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
424 {
425
426 return (NULL);
427 }
428
429 static int
nfssvc_nfscommon(struct thread * td,struct nfssvc_args * uap)430 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
431 {
432 int error;
433
434 NFSD_CURVNET_SET(NFSD_TD_TO_VNET(td));
435 error = nfssvc_call(td, uap, td->td_ucred);
436 NFSD_CURVNET_RESTORE();
437 NFSEXITCODE(error);
438 return (error);
439 }
440
441 static int
nfssvc_call(struct thread * p,struct nfssvc_args * uap,struct ucred * cred)442 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
443 {
444 int error = EINVAL, i, j;
445 struct nfsd_idargs nid;
446 struct nfsd_oidargs onid;
447 struct {
448 int vers; /* Just the first field of nfsstats. */
449 } nfsstatver;
450
451 if (uap->flag & NFSSVC_IDNAME) {
452 if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
453 error = copyin(uap->argp, &nid, sizeof(nid));
454 else {
455 error = copyin(uap->argp, &onid, sizeof(onid));
456 if (error == 0) {
457 nid.nid_flag = onid.nid_flag;
458 nid.nid_uid = onid.nid_uid;
459 nid.nid_gid = onid.nid_gid;
460 nid.nid_usermax = onid.nid_usermax;
461 nid.nid_usertimeout = onid.nid_usertimeout;
462 nid.nid_name = onid.nid_name;
463 nid.nid_namelen = onid.nid_namelen;
464 nid.nid_ngroup = 0;
465 nid.nid_grps = NULL;
466 }
467 }
468 if (error)
469 goto out;
470 error = nfssvc_idname(&nid);
471 goto out;
472 } else if (uap->flag & NFSSVC_GETSTATS) {
473 if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
474 /* Copy fields to the old ext_nfsstat structure. */
475 oldnfsstats.attrcache_hits =
476 NFSD_VNET(nfsstatsv1_p)->attrcache_hits;
477 oldnfsstats.attrcache_misses =
478 NFSD_VNET(nfsstatsv1_p)->attrcache_misses;
479 oldnfsstats.lookupcache_hits =
480 NFSD_VNET(nfsstatsv1_p)->lookupcache_hits;
481 oldnfsstats.lookupcache_misses =
482 NFSD_VNET(nfsstatsv1_p)->lookupcache_misses;
483 oldnfsstats.direofcache_hits =
484 NFSD_VNET(nfsstatsv1_p)->direofcache_hits;
485 oldnfsstats.direofcache_misses =
486 NFSD_VNET(nfsstatsv1_p)->direofcache_misses;
487 oldnfsstats.accesscache_hits =
488 NFSD_VNET(nfsstatsv1_p)->accesscache_hits;
489 oldnfsstats.accesscache_misses =
490 NFSD_VNET(nfsstatsv1_p)->accesscache_misses;
491 oldnfsstats.biocache_reads =
492 NFSD_VNET(nfsstatsv1_p)->biocache_reads;
493 oldnfsstats.read_bios =
494 NFSD_VNET(nfsstatsv1_p)->read_bios;
495 oldnfsstats.read_physios =
496 NFSD_VNET(nfsstatsv1_p)->read_physios;
497 oldnfsstats.biocache_writes =
498 NFSD_VNET(nfsstatsv1_p)->biocache_writes;
499 oldnfsstats.write_bios =
500 NFSD_VNET(nfsstatsv1_p)->write_bios;
501 oldnfsstats.write_physios =
502 NFSD_VNET(nfsstatsv1_p)->write_physios;
503 oldnfsstats.biocache_readlinks =
504 NFSD_VNET(nfsstatsv1_p)->biocache_readlinks;
505 oldnfsstats.readlink_bios =
506 NFSD_VNET(nfsstatsv1_p)->readlink_bios;
507 oldnfsstats.biocache_readdirs =
508 NFSD_VNET(nfsstatsv1_p)->biocache_readdirs;
509 oldnfsstats.readdir_bios =
510 NFSD_VNET(nfsstatsv1_p)->readdir_bios;
511 for (i = 0; i < NFSV4_NPROCS; i++)
512 oldnfsstats.rpccnt[i] =
513 NFSD_VNET(nfsstatsv1_p)->rpccnt[i];
514 oldnfsstats.rpcretries =
515 NFSD_VNET(nfsstatsv1_p)->rpcretries;
516 for (i = 0; i < NFSV4OP_NOPS; i++)
517 oldnfsstats.srvrpccnt[i] =
518 NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
519 for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
520 i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
521 oldnfsstats.srvrpccnt[j] =
522 NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
523 oldnfsstats.reserved_0 = 0;
524 oldnfsstats.reserved_1 = 0;
525 oldnfsstats.rpcrequests =
526 NFSD_VNET(nfsstatsv1_p)->rpcrequests;
527 oldnfsstats.rpctimeouts =
528 NFSD_VNET(nfsstatsv1_p)->rpctimeouts;
529 oldnfsstats.rpcunexpected =
530 NFSD_VNET(nfsstatsv1_p)->rpcunexpected;
531 oldnfsstats.rpcinvalid =
532 NFSD_VNET(nfsstatsv1_p)->rpcinvalid;
533 oldnfsstats.srvcache_inproghits =
534 NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits;
535 oldnfsstats.reserved_2 = 0;
536 oldnfsstats.srvcache_nonidemdonehits =
537 NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
538 oldnfsstats.srvcache_misses =
539 NFSD_VNET(nfsstatsv1_p)->srvcache_misses;
540 oldnfsstats.srvcache_tcppeak =
541 NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak;
542 oldnfsstats.srvcache_size =
543 NFSD_VNET(nfsstatsv1_p)->srvcache_size;
544 oldnfsstats.srvclients =
545 NFSD_VNET(nfsstatsv1_p)->srvclients;
546 oldnfsstats.srvopenowners =
547 NFSD_VNET(nfsstatsv1_p)->srvopenowners;
548 oldnfsstats.srvopens =
549 NFSD_VNET(nfsstatsv1_p)->srvopens;
550 oldnfsstats.srvlockowners =
551 NFSD_VNET(nfsstatsv1_p)->srvlockowners;
552 oldnfsstats.srvlocks =
553 NFSD_VNET(nfsstatsv1_p)->srvlocks;
554 oldnfsstats.srvdelegates =
555 NFSD_VNET(nfsstatsv1_p)->srvdelegates;
556 for (i = 0; i < NFSV4OP_CBNOPS; i++)
557 oldnfsstats.cbrpccnt[i] =
558 NFSD_VNET(nfsstatsv1_p)->cbrpccnt[i];
559 oldnfsstats.clopenowners =
560 NFSD_VNET(nfsstatsv1_p)->clopenowners;
561 oldnfsstats.clopens = NFSD_VNET(nfsstatsv1_p)->clopens;
562 oldnfsstats.cllockowners =
563 NFSD_VNET(nfsstatsv1_p)->cllockowners;
564 oldnfsstats.cllocks = NFSD_VNET(nfsstatsv1_p)->cllocks;
565 oldnfsstats.cldelegates =
566 NFSD_VNET(nfsstatsv1_p)->cldelegates;
567 oldnfsstats.cllocalopenowners =
568 NFSD_VNET(nfsstatsv1_p)->cllocalopenowners;
569 oldnfsstats.cllocalopens =
570 NFSD_VNET(nfsstatsv1_p)->cllocalopens;
571 oldnfsstats.cllocallockowners =
572 NFSD_VNET(nfsstatsv1_p)->cllocallockowners;
573 oldnfsstats.cllocallocks =
574 NFSD_VNET(nfsstatsv1_p)->cllocallocks;
575 error = copyout(&oldnfsstats, uap->argp,
576 sizeof (oldnfsstats));
577 } else {
578 error = copyin(uap->argp, &nfsstatver,
579 sizeof(nfsstatver));
580 if (error == 0) {
581 if (nfsstatver.vers == NFSSTATS_OV1) {
582 /* Copy nfsstatsv1 to nfsstatsov1. */
583 nfsstatsov1.attrcache_hits =
584 NFSD_VNET(nfsstatsv1_p)->attrcache_hits;
585 nfsstatsov1.attrcache_misses =
586 NFSD_VNET(nfsstatsv1_p)->attrcache_misses;
587 nfsstatsov1.lookupcache_hits =
588 NFSD_VNET(nfsstatsv1_p)->lookupcache_hits;
589 nfsstatsov1.lookupcache_misses =
590 NFSD_VNET(nfsstatsv1_p)->lookupcache_misses;
591 nfsstatsov1.direofcache_hits =
592 NFSD_VNET(nfsstatsv1_p)->direofcache_hits;
593 nfsstatsov1.direofcache_misses =
594 NFSD_VNET(nfsstatsv1_p)->direofcache_misses;
595 nfsstatsov1.accesscache_hits =
596 NFSD_VNET(nfsstatsv1_p)->accesscache_hits;
597 nfsstatsov1.accesscache_misses =
598 NFSD_VNET(nfsstatsv1_p)->accesscache_misses;
599 nfsstatsov1.biocache_reads =
600 NFSD_VNET(nfsstatsv1_p)->biocache_reads;
601 nfsstatsov1.read_bios =
602 NFSD_VNET(nfsstatsv1_p)->read_bios;
603 nfsstatsov1.read_physios =
604 NFSD_VNET(nfsstatsv1_p)->read_physios;
605 nfsstatsov1.biocache_writes =
606 NFSD_VNET(nfsstatsv1_p)->biocache_writes;
607 nfsstatsov1.write_bios =
608 NFSD_VNET(nfsstatsv1_p)->write_bios;
609 nfsstatsov1.write_physios =
610 NFSD_VNET(nfsstatsv1_p)->write_physios;
611 nfsstatsov1.biocache_readlinks =
612 NFSD_VNET(nfsstatsv1_p)->biocache_readlinks;
613 nfsstatsov1.readlink_bios =
614 NFSD_VNET(nfsstatsv1_p)->readlink_bios;
615 nfsstatsov1.biocache_readdirs =
616 NFSD_VNET(nfsstatsv1_p)->biocache_readdirs;
617 nfsstatsov1.readdir_bios =
618 NFSD_VNET(nfsstatsv1_p)->readdir_bios;
619 for (i = 0; i < NFSV42_OLDNPROCS; i++)
620 nfsstatsov1.rpccnt[i] =
621 NFSD_VNET(nfsstatsv1_p)->rpccnt[i];
622 nfsstatsov1.rpcretries =
623 NFSD_VNET(nfsstatsv1_p)->rpcretries;
624 for (i = 0; i < NFSV42_PURENOPS; i++)
625 nfsstatsov1.srvrpccnt[i] =
626 NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
627 for (i = NFSV42_NOPS,
628 j = NFSV42_PURENOPS;
629 i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
630 i++, j++)
631 nfsstatsov1.srvrpccnt[j] =
632 NFSD_VNET(nfsstatsv1_p)->srvrpccnt[i];
633 nfsstatsov1.reserved_0 = 0;
634 nfsstatsov1.reserved_1 = 0;
635 nfsstatsov1.rpcrequests =
636 NFSD_VNET(nfsstatsv1_p)->rpcrequests;
637 nfsstatsov1.rpctimeouts =
638 NFSD_VNET(nfsstatsv1_p)->rpctimeouts;
639 nfsstatsov1.rpcunexpected =
640 NFSD_VNET(nfsstatsv1_p)->rpcunexpected;
641 nfsstatsov1.rpcinvalid =
642 NFSD_VNET(nfsstatsv1_p)->rpcinvalid;
643 nfsstatsov1.srvcache_inproghits =
644 NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits;
645 nfsstatsov1.reserved_2 = 0;
646 nfsstatsov1.srvcache_nonidemdonehits =
647 NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
648 nfsstatsov1.srvcache_misses =
649 NFSD_VNET(nfsstatsv1_p)->srvcache_misses;
650 nfsstatsov1.srvcache_tcppeak =
651 NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak;
652 nfsstatsov1.srvcache_size =
653 NFSD_VNET(nfsstatsv1_p)->srvcache_size;
654 nfsstatsov1.srvclients =
655 NFSD_VNET(nfsstatsv1_p)->srvclients;
656 nfsstatsov1.srvopenowners =
657 NFSD_VNET(nfsstatsv1_p)->srvopenowners;
658 nfsstatsov1.srvopens =
659 NFSD_VNET(nfsstatsv1_p)->srvopens;
660 nfsstatsov1.srvlockowners =
661 NFSD_VNET(nfsstatsv1_p)->srvlockowners;
662 nfsstatsov1.srvlocks =
663 NFSD_VNET(nfsstatsv1_p)->srvlocks;
664 nfsstatsov1.srvdelegates =
665 NFSD_VNET(nfsstatsv1_p)->srvdelegates;
666 for (i = 0; i < NFSV42_CBNOPS; i++)
667 nfsstatsov1.cbrpccnt[i] =
668 NFSD_VNET(nfsstatsv1_p)->cbrpccnt[i];
669 nfsstatsov1.clopenowners =
670 NFSD_VNET(nfsstatsv1_p)->clopenowners;
671 nfsstatsov1.clopens =
672 NFSD_VNET(nfsstatsv1_p)->clopens;
673 nfsstatsov1.cllockowners =
674 NFSD_VNET(nfsstatsv1_p)->cllockowners;
675 nfsstatsov1.cllocks =
676 NFSD_VNET(nfsstatsv1_p)->cllocks;
677 nfsstatsov1.cldelegates =
678 NFSD_VNET(nfsstatsv1_p)->cldelegates;
679 nfsstatsov1.cllocalopenowners =
680 NFSD_VNET(nfsstatsv1_p)->cllocalopenowners;
681 nfsstatsov1.cllocalopens =
682 NFSD_VNET(nfsstatsv1_p)->cllocalopens;
683 nfsstatsov1.cllocallockowners =
684 NFSD_VNET(nfsstatsv1_p)->cllocallockowners;
685 nfsstatsov1.cllocallocks =
686 NFSD_VNET(nfsstatsv1_p)->cllocallocks;
687 nfsstatsov1.srvstartcnt =
688 NFSD_VNET(nfsstatsv1_p)->srvstartcnt;
689 nfsstatsov1.srvdonecnt =
690 NFSD_VNET(nfsstatsv1_p)->srvdonecnt;
691 for (i = NFSV42_NOPS,
692 j = NFSV42_PURENOPS;
693 i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
694 i++, j++) {
695 nfsstatsov1.srvbytes[j] =
696 NFSD_VNET(nfsstatsv1_p)->srvbytes[i];
697 nfsstatsov1.srvops[j] =
698 NFSD_VNET(nfsstatsv1_p)->srvops[i];
699 nfsstatsov1.srvduration[j] =
700 NFSD_VNET(nfsstatsv1_p)->srvduration[i];
701 }
702 nfsstatsov1.busyfrom =
703 NFSD_VNET(nfsstatsv1_p)->busyfrom;
704 nfsstatsov1.busyfrom =
705 NFSD_VNET(nfsstatsv1_p)->busyfrom;
706 error = copyout(&nfsstatsov1, uap->argp,
707 sizeof(nfsstatsov1));
708 } else if (nfsstatver.vers != NFSSTATS_V1)
709 error = EPERM;
710 else
711 error = copyout(NFSD_VNET(nfsstatsv1_p),
712 uap->argp, sizeof(nfsstatsv1));
713 }
714 }
715 if (error == 0) {
716 if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
717 NFSD_VNET(nfsstatsv1_p)->attrcache_hits = 0;
718 NFSD_VNET(nfsstatsv1_p)->attrcache_misses = 0;
719 NFSD_VNET(nfsstatsv1_p)->lookupcache_hits = 0;
720 NFSD_VNET(nfsstatsv1_p)->lookupcache_misses = 0;
721 NFSD_VNET(nfsstatsv1_p)->direofcache_hits = 0;
722 NFSD_VNET(nfsstatsv1_p)->direofcache_misses = 0;
723 NFSD_VNET(nfsstatsv1_p)->accesscache_hits = 0;
724 NFSD_VNET(nfsstatsv1_p)->accesscache_misses = 0;
725 NFSD_VNET(nfsstatsv1_p)->biocache_reads = 0;
726 NFSD_VNET(nfsstatsv1_p)->read_bios = 0;
727 NFSD_VNET(nfsstatsv1_p)->read_physios = 0;
728 NFSD_VNET(nfsstatsv1_p)->biocache_writes = 0;
729 NFSD_VNET(nfsstatsv1_p)->write_bios = 0;
730 NFSD_VNET(nfsstatsv1_p)->write_physios = 0;
731 NFSD_VNET(nfsstatsv1_p)->biocache_readlinks = 0;
732 NFSD_VNET(nfsstatsv1_p)->readlink_bios = 0;
733 NFSD_VNET(nfsstatsv1_p)->biocache_readdirs = 0;
734 NFSD_VNET(nfsstatsv1_p)->readdir_bios = 0;
735 NFSD_VNET(nfsstatsv1_p)->rpcretries = 0;
736 NFSD_VNET(nfsstatsv1_p)->rpcrequests = 0;
737 NFSD_VNET(nfsstatsv1_p)->rpctimeouts = 0;
738 NFSD_VNET(nfsstatsv1_p)->rpcunexpected = 0;
739 NFSD_VNET(nfsstatsv1_p)->rpcinvalid = 0;
740 bzero(NFSD_VNET(nfsstatsv1_p)->rpccnt,
741 sizeof(NFSD_VNET(nfsstatsv1_p)->rpccnt));
742 }
743 if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
744 NFSD_VNET(nfsstatsv1_p)->srvcache_inproghits = 0;
745 NFSD_VNET(nfsstatsv1_p)->srvcache_nonidemdonehits = 0;
746 NFSD_VNET(nfsstatsv1_p)->srvcache_misses = 0;
747 NFSD_VNET(nfsstatsv1_p)->srvcache_tcppeak = 0;
748 bzero(NFSD_VNET(nfsstatsv1_p)->srvrpccnt,
749 sizeof(NFSD_VNET(nfsstatsv1_p)->srvrpccnt));
750 bzero(NFSD_VNET(nfsstatsv1_p)->cbrpccnt,
751 sizeof(NFSD_VNET(nfsstatsv1_p)->cbrpccnt));
752 }
753 }
754 goto out;
755 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
756 u_short sockport;
757 struct nfsuserd_args nargs;
758
759 if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
760 error = copyin(uap->argp, (caddr_t)&sockport,
761 sizeof (u_short));
762 if (error == 0) {
763 nargs.nuserd_family = AF_INET;
764 nargs.nuserd_port = sockport;
765 }
766 } else {
767 /*
768 * New nfsuserd_args structure, which indicates
769 * which IP version to use along with the port#.
770 */
771 error = copyin(uap->argp, &nargs, sizeof(nargs));
772 }
773 if (!error)
774 error = nfsrv_nfsuserdport(&nargs, p);
775 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
776 nfsrv_nfsuserddelport();
777 error = 0;
778 }
779
780 out:
781 NFSEXITCODE(error);
782 return (error);
783 }
784
785 /*
786 * called by all three modevent routines, so that it gets things
787 * initialized soon enough.
788 */
789 void
newnfs_portinit(void)790 newnfs_portinit(void)
791 {
792 static int inited = 0;
793
794 if (inited)
795 return;
796 inited = 1;
797 /* Initialize SMP locks used by both client and server. */
798 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
799 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
800 mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
801 }
802
803 /*
804 * Determine if the file system supports NFSv4 ACLs.
805 * Return 1 if it does, 0 otherwise.
806 */
807 int
nfs_supportsnfsv4acls(struct vnode * vp)808 nfs_supportsnfsv4acls(struct vnode *vp)
809 {
810 int error;
811 long retval;
812
813 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
814
815 if (nfsrv_useacl == 0)
816 return (0);
817 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
818 if (error == 0 && retval != 0)
819 return (1);
820 return (0);
821 }
822
823 /*
824 * These are the first fields of all the context structures passed into
825 * nfs_pnfsio().
826 */
827 struct pnfsio {
828 int done;
829 int inprog;
830 struct task tsk;
831 };
832
833 /*
834 * Do a mirror I/O on a pNFS thread.
835 */
836 int
nfs_pnfsio(task_fn_t * func,void * context)837 nfs_pnfsio(task_fn_t *func, void *context)
838 {
839 struct pnfsio *pio;
840 int ret;
841 static struct taskqueue *pnfsioq = NULL;
842
843 pio = (struct pnfsio *)context;
844 if (pnfsioq == NULL) {
845 if (nfs_pnfsiothreads == 0)
846 return (EPERM);
847 if (nfs_pnfsiothreads < 0)
848 nfs_pnfsiothreads = mp_ncpus * 4;
849 pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
850 taskqueue_thread_enqueue, &pnfsioq);
851 if (pnfsioq == NULL)
852 return (ENOMEM);
853 ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
854 0, "pnfsiot");
855 if (ret != 0) {
856 taskqueue_free(pnfsioq);
857 pnfsioq = NULL;
858 return (ret);
859 }
860 }
861 pio->inprog = 1;
862 TASK_INIT(&pio->tsk, 0, func, context);
863 ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
864 if (ret != 0)
865 pio->inprog = 0;
866 return (ret);
867 }
868
869 /*
870 * Initialize everything that needs to be initialized for a vnet.
871 */
872 static void
nfs_vnetinit(const void * unused __unused)873 nfs_vnetinit(const void *unused __unused)
874 {
875
876 if (IS_DEFAULT_VNET(curvnet))
877 NFSD_VNET(nfsstatsv1_p) = &nfsstatsv1;
878 else
879 NFSD_VNET(nfsstatsv1_p) = malloc(sizeof(struct nfsstatsv1),
880 M_TEMP, M_WAITOK | M_ZERO);
881 mtx_init(&NFSD_VNET(nfsrv_nfsuserdsock).nr_mtx, "nfsuserd",
882 NULL, MTX_DEF);
883 }
884 VNET_SYSINIT(nfs_vnetinit, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
885 nfs_vnetinit, NULL);
886
887 static void
nfs_cleanup(void * unused __unused)888 nfs_cleanup(void *unused __unused)
889 {
890
891 mtx_destroy(&NFSD_VNET(nfsrv_nfsuserdsock).nr_mtx);
892 if (!IS_DEFAULT_VNET(curvnet)) {
893 free(NFSD_VNET(nfsstatsv1_p), M_TEMP);
894 NFSD_VNET(nfsstatsv1_p) = NULL;
895 }
896 /* Clean out the name<-->id cache. */
897 nfsrv_cleanusergroup();
898 }
899 VNET_SYSUNINIT(nfs_cleanup, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
900 nfs_cleanup, NULL);
901
902 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
903
904 /*
905 * Called once to initialize data structures...
906 */
907 static int
nfscommon_modevent(module_t mod,int type,void * data)908 nfscommon_modevent(module_t mod, int type, void *data)
909 {
910 int error = 0;
911 static int loaded = 0;
912
913 switch (type) {
914 case MOD_LOAD:
915 if (loaded)
916 goto out;
917 newnfs_portinit();
918 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
919 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
920 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
921 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
922 mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
923 TAILQ_INIT(&nfsrv_devidhead);
924 newnfs_init();
925 nfsd_call_nfscommon = nfssvc_nfscommon;
926 loaded = 1;
927 break;
928
929 case MOD_UNLOAD:
930 if (newnfs_numnfsd != 0 ||
931 NFSD_VNET(nfsrv_nfsuserd) != NOTRUNNING ||
932 nfs_numnfscbd != 0) {
933 error = EBUSY;
934 break;
935 }
936
937 nfsd_call_nfscommon = NULL;
938 /* and get rid of the mutexes */
939 mtx_destroy(&nfs_nameid_mutex);
940 mtx_destroy(&newnfsd_mtx);
941 mtx_destroy(&nfs_state_mutex);
942 mtx_destroy(&nfs_clstate_mutex);
943 mtx_destroy(&nfs_sockl_mutex);
944 mtx_destroy(&nfs_slock_mutex);
945 mtx_destroy(&nfs_req_mutex);
946 mtx_destroy(&nfsrv_dslock_mtx);
947 loaded = 0;
948 break;
949 default:
950 error = EOPNOTSUPP;
951 break;
952 }
953
954 out:
955 NFSEXITCODE(error);
956 return error;
957 }
958 static moduledata_t nfscommon_mod = {
959 "nfscommon",
960 nfscommon_modevent,
961 NULL,
962 };
963 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
964
965 /* So that loader and kldload(2) can find us, wherever we are.. */
966 MODULE_VERSION(nfscommon, 1);
967 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
968 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
969