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 <fs/nfs/nfsid.h>
44 #include <sys/smp.h>
45 #include <sys/sysctl.h>
46 #include <sys/taskqueue.h>
47 #include <rpc/rpc_com.h>
48 #include <vm/vm.h>
49 #include <vm/vm_object.h>
50 #include <vm/vm_page.h>
51 #include <vm/vm_param.h>
52 #include <vm/vm_map.h>
53 #include <vm/vm_kern.h>
54 #include <vm/vm_extern.h>
55 #include <vm/uma.h>
56
57 extern int nfscl_ticks;
58 extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
59 struct thread *);
60 extern int nfsrv_useacl;
61 int newnfs_numnfsd = 0;
62 struct nfsstatsv1 nfsstatsv1;
63 int nfs_numnfscbd = 0;
64 int nfscl_debuglevel = 0;
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 VNET_DEFINE(struct nfsstatsv1 *, nfsstatsv1_p);
75
76 VNET_DECLARE(struct nfssockreq, nfsrv_nfsuserdsock);
77 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_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
93 0, "Debug level for NFS client");
94 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
95 0, "Size of hash tables for uid/name mapping");
96 int nfs_pnfsiothreads = -1;
97 SYSCTL_INT(_vfs_nfs, OID_AUTO, pnfsiothreads, CTLFLAG_RW, &nfs_pnfsiothreads,
98 0, "Number of pNFS mirror I/O threads");
99
100 /*
101 * Defines for malloc
102 * (Here for FreeBSD, since they allocate storage.)
103 */
104 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
105 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
106 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
107 "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
108 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
109 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
110 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
111 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
112 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
113 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
114 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
115 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
116 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
117 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
118 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
119 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
120 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
121 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroff",
122 "NFS directory offset data");
123 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
124 "NFS local lock rollback");
125 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
126 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
127 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
128 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
129 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
130 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
131 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
132
133 /*
134 * Definition of mutex locks.
135 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
136 * and assorted other nfsd structures.
137 */
138 struct mtx newnfsd_mtx;
139 struct mtx nfs_sockl_mutex;
140 struct mtx nfs_state_mutex;
141 struct mtx nfs_nameid_mutex;
142 struct mtx nfs_req_mutex;
143 struct mtx nfs_slock_mutex;
144 struct mtx nfs_clstate_mutex;
145
146 /* local functions */
147 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
148
149 #ifdef __NO_STRICT_ALIGNMENT
150 /*
151 * These architectures don't need re-alignment, so just return.
152 */
153 int
newnfs_realign(struct mbuf ** pm,int how)154 newnfs_realign(struct mbuf **pm, int how)
155 {
156
157 return (0);
158 }
159 #else /* !__NO_STRICT_ALIGNMENT */
160 /*
161 * newnfs_realign:
162 *
163 * Check for badly aligned mbuf data and realign by copying the unaligned
164 * portion of the data into a new mbuf chain and freeing the portions
165 * of the old chain that were replaced.
166 *
167 * We cannot simply realign the data within the existing mbuf chain
168 * because the underlying buffers may contain other rpc commands and
169 * we cannot afford to overwrite them.
170 *
171 * We would prefer to avoid this situation entirely. The situation does
172 * not occur with NFS/UDP and is supposed to only occasionally occur
173 * with TCP. Use vfs.nfs.realign_count and realign_test to check this.
174 *
175 */
176 int
newnfs_realign(struct mbuf ** pm,int how)177 newnfs_realign(struct mbuf **pm, int how)
178 {
179 struct mbuf *m, *n;
180 int off, space;
181
182 ++nfs_realign_test;
183 while ((m = *pm) != NULL) {
184 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
185 /*
186 * NB: we can't depend on m_pkthdr.len to help us
187 * decide what to do here. May not be worth doing
188 * the m_length calculation as m_copyback will
189 * expand the mbuf chain below as needed.
190 */
191 space = m_length(m, NULL);
192 if (space >= MINCLSIZE) {
193 /* NB: m_copyback handles space > MCLBYTES */
194 n = m_getcl(how, MT_DATA, 0);
195 } else
196 n = m_get(how, MT_DATA);
197 if (n == NULL)
198 return (ENOMEM);
199 /*
200 * Align the remainder of the mbuf chain.
201 */
202 n->m_len = 0;
203 off = 0;
204 while (m != NULL) {
205 m_copyback(n, off, m->m_len, mtod(m, caddr_t));
206 off += m->m_len;
207 m = m->m_next;
208 }
209 m_freem(*pm);
210 *pm = n;
211 ++nfs_realign_count;
212 break;
213 }
214 pm = &m->m_next;
215 }
216
217 return (0);
218 }
219 #endif /* __NO_STRICT_ALIGNMENT */
220
221 #ifdef notdef
222 static void
nfsrv_object_create(struct vnode * vp,struct thread * td)223 nfsrv_object_create(struct vnode *vp, struct thread *td)
224 {
225
226 if (vp == NULL || vp->v_type != VREG)
227 return;
228 (void) vfs_object_create(vp, td, td->td_ucred);
229 }
230 #endif
231
232 /*
233 * Look up a file name. Basically just initialize stuff and call namei().
234 */
235 int
nfsrv_lookupfilename(struct nameidata * ndp,char * fname,NFSPROC_T * p __unused)236 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p __unused)
237 {
238 int error;
239
240 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname);
241 error = namei(ndp);
242 if (!error) {
243 NDFREE_PNBUF(ndp);
244 }
245 return (error);
246 }
247
248 /*
249 * Copy NFS uid, gids to the cred structure.
250 */
251 void
newnfs_copycred(struct nfscred * nfscr,struct ucred * cr)252 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
253 {
254
255 KASSERT(nfscr->nfsc_ngroups >= 0,
256 ("newnfs_copycred: negative nfsc_ngroups"));
257 cr->cr_uid = nfscr->nfsc_uid;
258 crsetgroups_and_egid(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups,
259 GID_NOGROUP);
260 }
261
262 /*
263 * Map args from nfsmsleep() to msleep().
264 */
265 int
nfsmsleep(void * chan,void * mutex,int prio,const char * wmesg,struct timespec * ts)266 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
267 struct timespec *ts)
268 {
269 u_int64_t nsecval;
270 int error, timeo;
271
272 if (ts) {
273 timeo = hz * ts->tv_sec;
274 nsecval = (u_int64_t)ts->tv_nsec;
275 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
276 1000000000;
277 timeo += (int)nsecval;
278 } else {
279 timeo = 0;
280 }
281 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
282 return (error);
283 }
284
285 /*
286 * Get the file system info for the server. For now, just assume FFS.
287 */
288 void
nfsvno_getfs(struct nfsfsinfo * sip,int isdgram)289 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
290 {
291 int pref;
292
293 /*
294 * XXX
295 * There should be file system VFS OP(s) to get this information.
296 * For now, assume ufs.
297 */
298 if (isdgram)
299 pref = NFS_MAXDGRAMDATA;
300 else
301 pref = nfs_srvmaxio;
302 sip->fs_rtmax = nfs_srvmaxio;
303 sip->fs_rtpref = pref;
304 sip->fs_rtmult = NFS_FABLKSIZE;
305 sip->fs_wtmax = nfs_srvmaxio;
306 sip->fs_wtpref = pref;
307 sip->fs_wtmult = NFS_FABLKSIZE;
308 sip->fs_dtpref = pref;
309 sip->fs_maxfilesize = 0xffffffffffffffffull;
310 sip->fs_timedelta.tv_sec = 0;
311 sip->fs_timedelta.tv_nsec = 1;
312 sip->fs_properties = (NFSV3FSINFO_LINK |
313 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
314 NFSV3FSINFO_CANSETTIME);
315 }
316
317 /*
318 * Do the pathconf vnode op.
319 */
320 int
nfsvno_pathconf(struct vnode * vp,int flag,long * retf,struct ucred * cred,struct thread * p)321 nfsvno_pathconf(struct vnode *vp, int flag, long *retf,
322 struct ucred *cred, struct thread *p)
323 {
324 int error;
325
326 error = VOP_PATHCONF(vp, flag, retf);
327 if (error == EOPNOTSUPP || error == EINVAL) {
328 /*
329 * Some file systems return EINVAL for name arguments not
330 * supported and some return EOPNOTSUPP for this case.
331 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
332 * just fake them.
333 */
334 switch (flag) {
335 case _PC_LINK_MAX:
336 *retf = NFS_LINK_MAX;
337 break;
338 case _PC_NAME_MAX:
339 *retf = NAME_MAX;
340 break;
341 case _PC_CHOWN_RESTRICTED:
342 *retf = 1;
343 break;
344 case _PC_NO_TRUNC:
345 *retf = 1;
346 break;
347 default:
348 /*
349 * Only happens if a _PC_xxx is added to the server,
350 * but this isn't updated.
351 */
352 *retf = 0;
353 printf("nfsrvd pathconf flag=%d not supp\n", flag);
354 }
355 error = 0;
356 }
357 NFSEXITCODE(error);
358 return (error);
359 }
360
361 /* Fake nfsrv_atroot. Just return 0 */
362 int
nfsrv_atroot(struct vnode * vp,uint64_t * retp)363 nfsrv_atroot(struct vnode *vp, uint64_t *retp)
364 {
365
366 return (0);
367 }
368
369 /*
370 * Set the credentials to refer to root.
371 */
372 void
newnfs_setroot(struct ucred * cred)373 newnfs_setroot(struct ucred *cred)
374 {
375
376 cred->cr_uid = 0;
377 cred->cr_gid = 0;
378 cred->cr_ngroups = 0;
379 }
380
381 /*
382 * Get the client credential. Used for Renew and recovery.
383 */
384 struct ucred *
newnfs_getcred(void)385 newnfs_getcred(void)
386 {
387 struct ucred *cred;
388 struct thread *td = curthread;
389
390 cred = crdup(td->td_ucred);
391 newnfs_setroot(cred);
392 return (cred);
393 }
394
395 /*
396 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
397 * the sleep should be for 5 seconds.
398 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
399 * an event that never gets a wakeup. Only return EINTR or 0.
400 */
401 int
nfs_catnap(int prio,int errval,const char * wmesg)402 nfs_catnap(int prio, int errval, const char *wmesg)
403 {
404 static int non_event;
405 int ret;
406
407 if (errval == NFSERR_GRACE)
408 ret = tsleep(&non_event, prio, wmesg, 5 * hz);
409 else
410 ret = tsleep(&non_event, prio, wmesg, 1);
411 if (ret != EINTR)
412 ret = 0;
413 return (ret);
414 }
415
416 /*
417 * Get referral. For now, just fail.
418 */
419 struct nfsreferral *
nfsv4root_getreferral(struct vnode * vp,struct vnode * dvp,u_int32_t fileno)420 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
421 {
422
423 return (NULL);
424 }
425
426 static int
nfssvc_nfscommon(struct thread * td,struct nfssvc_args * uap)427 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
428 {
429 int error;
430
431 CURVNET_SET(TD_TO_VNET(td));
432 error = nfssvc_call(td, uap, td->td_ucred);
433 CURVNET_RESTORE();
434 NFSEXITCODE(error);
435 return (error);
436 }
437
438 static int
nfssvc_call(struct thread * p,struct nfssvc_args * uap,struct ucred * cred)439 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
440 {
441 int error = EINVAL, i, j;
442 struct nfsd_idargs nid;
443 struct nfsd_oidargs onid;
444 struct {
445 int vers; /* Just the first field of nfsstats. */
446 } nfsstatver;
447
448 if (uap->flag & NFSSVC_IDNAME) {
449 if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
450 error = copyin(uap->argp, &nid, sizeof(nid));
451 else {
452 error = copyin(uap->argp, &onid, sizeof(onid));
453 if (error == 0) {
454 nid.nid_flag = onid.nid_flag;
455 nid.nid_uid = onid.nid_uid;
456 nid.nid_gid = onid.nid_gid;
457 nid.nid_usermax = onid.nid_usermax;
458 nid.nid_usertimeout = onid.nid_usertimeout;
459 nid.nid_name = onid.nid_name;
460 nid.nid_namelen = onid.nid_namelen;
461 nid.nid_ngroup = 0;
462 nid.nid_grps = NULL;
463 }
464 }
465 if (error)
466 goto out;
467 error = nfssvc_idname(&nid);
468 goto out;
469 } else if (uap->flag & NFSSVC_GETSTATS) {
470 if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
471 /* Copy fields to the old ext_nfsstat structure. */
472 oldnfsstats.attrcache_hits =
473 VNET(nfsstatsv1_p)->attrcache_hits;
474 oldnfsstats.attrcache_misses =
475 VNET(nfsstatsv1_p)->attrcache_misses;
476 oldnfsstats.lookupcache_hits =
477 VNET(nfsstatsv1_p)->lookupcache_hits;
478 oldnfsstats.lookupcache_misses =
479 VNET(nfsstatsv1_p)->lookupcache_misses;
480 oldnfsstats.direofcache_hits =
481 VNET(nfsstatsv1_p)->direofcache_hits;
482 oldnfsstats.direofcache_misses =
483 VNET(nfsstatsv1_p)->direofcache_misses;
484 oldnfsstats.accesscache_hits =
485 VNET(nfsstatsv1_p)->accesscache_hits;
486 oldnfsstats.accesscache_misses =
487 VNET(nfsstatsv1_p)->accesscache_misses;
488 oldnfsstats.biocache_reads =
489 VNET(nfsstatsv1_p)->biocache_reads;
490 oldnfsstats.read_bios =
491 VNET(nfsstatsv1_p)->read_bios;
492 oldnfsstats.read_physios =
493 VNET(nfsstatsv1_p)->read_physios;
494 oldnfsstats.biocache_writes =
495 VNET(nfsstatsv1_p)->biocache_writes;
496 oldnfsstats.write_bios =
497 VNET(nfsstatsv1_p)->write_bios;
498 oldnfsstats.write_physios =
499 VNET(nfsstatsv1_p)->write_physios;
500 oldnfsstats.biocache_readlinks =
501 VNET(nfsstatsv1_p)->biocache_readlinks;
502 oldnfsstats.readlink_bios =
503 VNET(nfsstatsv1_p)->readlink_bios;
504 oldnfsstats.biocache_readdirs =
505 VNET(nfsstatsv1_p)->biocache_readdirs;
506 oldnfsstats.readdir_bios =
507 VNET(nfsstatsv1_p)->readdir_bios;
508 for (i = 0; i < NFSV4_NPROCS; i++)
509 oldnfsstats.rpccnt[i] =
510 VNET(nfsstatsv1_p)->rpccnt[i];
511 oldnfsstats.rpcretries =
512 VNET(nfsstatsv1_p)->rpcretries;
513 for (i = 0; i < NFSV4OP_NOPS; i++)
514 oldnfsstats.srvrpccnt[i] =
515 VNET(nfsstatsv1_p)->srvrpccnt[i];
516 for (i = NFSV42_NOPS, j = NFSV4OP_NOPS;
517 i < NFSV42_NOPS + NFSV4OP_FAKENOPS; i++, j++)
518 oldnfsstats.srvrpccnt[j] =
519 VNET(nfsstatsv1_p)->srvrpccnt[i];
520 oldnfsstats.reserved_0 = 0;
521 oldnfsstats.reserved_1 = 0;
522 oldnfsstats.rpcrequests =
523 VNET(nfsstatsv1_p)->rpcrequests;
524 oldnfsstats.rpctimeouts =
525 VNET(nfsstatsv1_p)->rpctimeouts;
526 oldnfsstats.rpcunexpected =
527 VNET(nfsstatsv1_p)->rpcunexpected;
528 oldnfsstats.rpcinvalid =
529 VNET(nfsstatsv1_p)->rpcinvalid;
530 oldnfsstats.srvcache_inproghits =
531 VNET(nfsstatsv1_p)->srvcache_inproghits;
532 oldnfsstats.reserved_2 = 0;
533 oldnfsstats.srvcache_nonidemdonehits =
534 VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
535 oldnfsstats.srvcache_misses =
536 VNET(nfsstatsv1_p)->srvcache_misses;
537 oldnfsstats.srvcache_tcppeak =
538 VNET(nfsstatsv1_p)->srvcache_tcppeak;
539 oldnfsstats.srvcache_size =
540 VNET(nfsstatsv1_p)->srvcache_size;
541 oldnfsstats.srvclients =
542 VNET(nfsstatsv1_p)->srvclients;
543 oldnfsstats.srvopenowners =
544 VNET(nfsstatsv1_p)->srvopenowners;
545 oldnfsstats.srvopens =
546 VNET(nfsstatsv1_p)->srvopens;
547 oldnfsstats.srvlockowners =
548 VNET(nfsstatsv1_p)->srvlockowners;
549 oldnfsstats.srvlocks =
550 VNET(nfsstatsv1_p)->srvlocks;
551 oldnfsstats.srvdelegates =
552 VNET(nfsstatsv1_p)->srvdelegates;
553 for (i = 0; i < NFSV4OP_CBNOPS; i++)
554 oldnfsstats.cbrpccnt[i] =
555 VNET(nfsstatsv1_p)->cbrpccnt[i];
556 oldnfsstats.clopenowners =
557 VNET(nfsstatsv1_p)->clopenowners;
558 oldnfsstats.clopens = VNET(nfsstatsv1_p)->clopens;
559 oldnfsstats.cllockowners =
560 VNET(nfsstatsv1_p)->cllockowners;
561 oldnfsstats.cllocks = VNET(nfsstatsv1_p)->cllocks;
562 oldnfsstats.cldelegates =
563 VNET(nfsstatsv1_p)->cldelegates;
564 oldnfsstats.cllocalopenowners =
565 VNET(nfsstatsv1_p)->cllocalopenowners;
566 oldnfsstats.cllocalopens =
567 VNET(nfsstatsv1_p)->cllocalopens;
568 oldnfsstats.cllocallockowners =
569 VNET(nfsstatsv1_p)->cllocallockowners;
570 oldnfsstats.cllocallocks =
571 VNET(nfsstatsv1_p)->cllocallocks;
572 error = copyout(&oldnfsstats, uap->argp,
573 sizeof (oldnfsstats));
574 } else {
575 error = copyin(uap->argp, &nfsstatver,
576 sizeof(nfsstatver));
577 if (error == 0) {
578 if (nfsstatver.vers == NFSSTATS_OV1) {
579 /* Copy nfsstatsv1 to nfsstatsov1. */
580 nfsstatsov1.attrcache_hits =
581 VNET(nfsstatsv1_p)->attrcache_hits;
582 nfsstatsov1.attrcache_misses =
583 VNET(nfsstatsv1_p)->attrcache_misses;
584 nfsstatsov1.lookupcache_hits =
585 VNET(nfsstatsv1_p)->lookupcache_hits;
586 nfsstatsov1.lookupcache_misses =
587 VNET(nfsstatsv1_p)->lookupcache_misses;
588 nfsstatsov1.direofcache_hits =
589 VNET(nfsstatsv1_p)->direofcache_hits;
590 nfsstatsov1.direofcache_misses =
591 VNET(nfsstatsv1_p)->direofcache_misses;
592 nfsstatsov1.accesscache_hits =
593 VNET(nfsstatsv1_p)->accesscache_hits;
594 nfsstatsov1.accesscache_misses =
595 VNET(nfsstatsv1_p)->accesscache_misses;
596 nfsstatsov1.biocache_reads =
597 VNET(nfsstatsv1_p)->biocache_reads;
598 nfsstatsov1.read_bios =
599 VNET(nfsstatsv1_p)->read_bios;
600 nfsstatsov1.read_physios =
601 VNET(nfsstatsv1_p)->read_physios;
602 nfsstatsov1.biocache_writes =
603 VNET(nfsstatsv1_p)->biocache_writes;
604 nfsstatsov1.write_bios =
605 VNET(nfsstatsv1_p)->write_bios;
606 nfsstatsov1.write_physios =
607 VNET(nfsstatsv1_p)->write_physios;
608 nfsstatsov1.biocache_readlinks =
609 VNET(nfsstatsv1_p)->biocache_readlinks;
610 nfsstatsov1.readlink_bios =
611 VNET(nfsstatsv1_p)->readlink_bios;
612 nfsstatsov1.biocache_readdirs =
613 VNET(nfsstatsv1_p)->biocache_readdirs;
614 nfsstatsov1.readdir_bios =
615 VNET(nfsstatsv1_p)->readdir_bios;
616 for (i = 0; i < NFSV42_OLDNPROCS; i++)
617 nfsstatsov1.rpccnt[i] =
618 VNET(nfsstatsv1_p)->rpccnt[i];
619 nfsstatsov1.rpcretries =
620 VNET(nfsstatsv1_p)->rpcretries;
621 for (i = 0; i < NFSV42_PURENOPS; i++)
622 nfsstatsov1.srvrpccnt[i] =
623 VNET(nfsstatsv1_p)->srvrpccnt[i];
624 for (i = NFSV42_NOPS,
625 j = NFSV42_PURENOPS;
626 i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
627 i++, j++)
628 nfsstatsov1.srvrpccnt[j] =
629 VNET(nfsstatsv1_p)->srvrpccnt[i];
630 nfsstatsov1.reserved_0 = 0;
631 nfsstatsov1.reserved_1 = 0;
632 nfsstatsov1.rpcrequests =
633 VNET(nfsstatsv1_p)->rpcrequests;
634 nfsstatsov1.rpctimeouts =
635 VNET(nfsstatsv1_p)->rpctimeouts;
636 nfsstatsov1.rpcunexpected =
637 VNET(nfsstatsv1_p)->rpcunexpected;
638 nfsstatsov1.rpcinvalid =
639 VNET(nfsstatsv1_p)->rpcinvalid;
640 nfsstatsov1.srvcache_inproghits =
641 VNET(nfsstatsv1_p)->srvcache_inproghits;
642 nfsstatsov1.reserved_2 = 0;
643 nfsstatsov1.srvcache_nonidemdonehits =
644 VNET(nfsstatsv1_p)->srvcache_nonidemdonehits;
645 nfsstatsov1.srvcache_misses =
646 VNET(nfsstatsv1_p)->srvcache_misses;
647 nfsstatsov1.srvcache_tcppeak =
648 VNET(nfsstatsv1_p)->srvcache_tcppeak;
649 nfsstatsov1.srvcache_size =
650 VNET(nfsstatsv1_p)->srvcache_size;
651 nfsstatsov1.srvclients =
652 VNET(nfsstatsv1_p)->srvclients;
653 nfsstatsov1.srvopenowners =
654 VNET(nfsstatsv1_p)->srvopenowners;
655 nfsstatsov1.srvopens =
656 VNET(nfsstatsv1_p)->srvopens;
657 nfsstatsov1.srvlockowners =
658 VNET(nfsstatsv1_p)->srvlockowners;
659 nfsstatsov1.srvlocks =
660 VNET(nfsstatsv1_p)->srvlocks;
661 nfsstatsov1.srvdelegates =
662 VNET(nfsstatsv1_p)->srvdelegates;
663 for (i = 0; i < NFSV42_CBNOPS; i++)
664 nfsstatsov1.cbrpccnt[i] =
665 VNET(nfsstatsv1_p)->cbrpccnt[i];
666 nfsstatsov1.clopenowners =
667 VNET(nfsstatsv1_p)->clopenowners;
668 nfsstatsov1.clopens =
669 VNET(nfsstatsv1_p)->clopens;
670 nfsstatsov1.cllockowners =
671 VNET(nfsstatsv1_p)->cllockowners;
672 nfsstatsov1.cllocks =
673 VNET(nfsstatsv1_p)->cllocks;
674 nfsstatsov1.cldelegates =
675 VNET(nfsstatsv1_p)->cldelegates;
676 nfsstatsov1.cllocalopenowners =
677 VNET(nfsstatsv1_p)->cllocalopenowners;
678 nfsstatsov1.cllocalopens =
679 VNET(nfsstatsv1_p)->cllocalopens;
680 nfsstatsov1.cllocallockowners =
681 VNET(nfsstatsv1_p)->cllocallockowners;
682 nfsstatsov1.cllocallocks =
683 VNET(nfsstatsv1_p)->cllocallocks;
684 nfsstatsov1.srvstartcnt =
685 VNET(nfsstatsv1_p)->srvstartcnt;
686 nfsstatsov1.srvdonecnt =
687 VNET(nfsstatsv1_p)->srvdonecnt;
688 for (i = NFSV42_NOPS,
689 j = NFSV42_PURENOPS;
690 i < NFSV42_NOPS + NFSV4OP_FAKENOPS;
691 i++, j++) {
692 nfsstatsov1.srvbytes[j] =
693 VNET(nfsstatsv1_p)->srvbytes[i];
694 nfsstatsov1.srvops[j] =
695 VNET(nfsstatsv1_p)->srvops[i];
696 nfsstatsov1.srvduration[j] =
697 VNET(nfsstatsv1_p)->srvduration[i];
698 }
699 nfsstatsov1.busyfrom =
700 VNET(nfsstatsv1_p)->busyfrom;
701 nfsstatsov1.busyfrom =
702 VNET(nfsstatsv1_p)->busyfrom;
703 error = copyout(&nfsstatsov1, uap->argp,
704 sizeof(nfsstatsov1));
705 } else if (nfsstatver.vers != NFSSTATS_V1)
706 error = EPERM;
707 else
708 error = copyout(VNET(nfsstatsv1_p),
709 uap->argp, sizeof(nfsstatsv1));
710 }
711 }
712 if (error == 0) {
713 if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
714 VNET(nfsstatsv1_p)->attrcache_hits = 0;
715 VNET(nfsstatsv1_p)->attrcache_misses = 0;
716 VNET(nfsstatsv1_p)->lookupcache_hits = 0;
717 VNET(nfsstatsv1_p)->lookupcache_misses = 0;
718 VNET(nfsstatsv1_p)->direofcache_hits = 0;
719 VNET(nfsstatsv1_p)->direofcache_misses = 0;
720 VNET(nfsstatsv1_p)->accesscache_hits = 0;
721 VNET(nfsstatsv1_p)->accesscache_misses = 0;
722 VNET(nfsstatsv1_p)->biocache_reads = 0;
723 VNET(nfsstatsv1_p)->read_bios = 0;
724 VNET(nfsstatsv1_p)->read_physios = 0;
725 VNET(nfsstatsv1_p)->biocache_writes = 0;
726 VNET(nfsstatsv1_p)->write_bios = 0;
727 VNET(nfsstatsv1_p)->write_physios = 0;
728 VNET(nfsstatsv1_p)->biocache_readlinks = 0;
729 VNET(nfsstatsv1_p)->readlink_bios = 0;
730 VNET(nfsstatsv1_p)->biocache_readdirs = 0;
731 VNET(nfsstatsv1_p)->readdir_bios = 0;
732 VNET(nfsstatsv1_p)->rpcretries = 0;
733 VNET(nfsstatsv1_p)->rpcrequests = 0;
734 VNET(nfsstatsv1_p)->rpctimeouts = 0;
735 VNET(nfsstatsv1_p)->rpcunexpected = 0;
736 VNET(nfsstatsv1_p)->rpcinvalid = 0;
737 bzero(VNET(nfsstatsv1_p)->rpccnt,
738 sizeof(VNET(nfsstatsv1_p)->rpccnt));
739 }
740 if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
741 VNET(nfsstatsv1_p)->srvcache_inproghits = 0;
742 VNET(nfsstatsv1_p)->srvcache_nonidemdonehits = 0;
743 VNET(nfsstatsv1_p)->srvcache_misses = 0;
744 VNET(nfsstatsv1_p)->srvcache_tcppeak = 0;
745 bzero(VNET(nfsstatsv1_p)->srvrpccnt,
746 sizeof(VNET(nfsstatsv1_p)->srvrpccnt));
747 bzero(VNET(nfsstatsv1_p)->cbrpccnt,
748 sizeof(VNET(nfsstatsv1_p)->cbrpccnt));
749 }
750 }
751 goto out;
752 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
753 u_short sockport;
754 struct nfsuserd_args nargs;
755
756 if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
757 error = copyin(uap->argp, (caddr_t)&sockport,
758 sizeof (u_short));
759 if (error == 0) {
760 nargs.nuserd_family = AF_INET;
761 nargs.nuserd_port = sockport;
762 }
763 } else {
764 /*
765 * New nfsuserd_args structure, which indicates
766 * which IP version to use along with the port#.
767 */
768 error = copyin(uap->argp, &nargs, sizeof(nargs));
769 }
770 if (!error)
771 error = nfsrv_nfsuserdport(&nargs, p);
772 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
773 nfsrv_nfsuserddelport();
774 error = 0;
775 }
776
777 out:
778 NFSEXITCODE(error);
779 return (error);
780 }
781
782 /*
783 * called by all three modevent routines, so that it gets things
784 * initialized soon enough.
785 */
786 void
newnfs_portinit(void)787 newnfs_portinit(void)
788 {
789 static int inited = 0;
790
791 if (inited)
792 return;
793 inited = 1;
794 /* Initialize SMP locks used by both client and server. */
795 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
796 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
797 mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
798 }
799
800 /*
801 * Determine if the file system supports NFSv4 ACLs.
802 * Return 1 if it does, 0 otherwise.
803 */
804 int
nfs_supportsnfsv4acls(struct vnode * vp)805 nfs_supportsnfsv4acls(struct vnode *vp)
806 {
807 int error;
808 long retval;
809
810 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
811
812 if (nfsrv_useacl == 0)
813 return (0);
814 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
815 if (error == 0 && retval != 0)
816 return (1);
817 return (0);
818 }
819
820 /*
821 * Determine if the file system supports POSIX draft ACLs.
822 * Return 1 if it does, 0 otherwise.
823 */
824 int
nfs_supportsposixacls(struct vnode * vp)825 nfs_supportsposixacls(struct vnode *vp)
826 {
827 int error;
828 long retval;
829
830 ASSERT_VOP_LOCKED(vp, "nfs supports posixacls");
831
832 if (nfsrv_useacl == 0)
833 return (0);
834 error = VOP_PATHCONF(vp, _PC_ACL_EXTENDED, &retval);
835 if (error == 0 && retval != 0)
836 return (1);
837 return (0);
838 }
839
840 /*
841 * These are the first fields of all the context structures passed into
842 * nfs_pnfsio().
843 */
844 struct pnfsio {
845 int done;
846 int inprog;
847 struct task tsk;
848 };
849
850 /*
851 * Do a mirror I/O on a pNFS thread.
852 */
853 int
nfs_pnfsio(task_fn_t * func,void * context)854 nfs_pnfsio(task_fn_t *func, void *context)
855 {
856 struct pnfsio *pio;
857 int ret;
858 static struct taskqueue *pnfsioq = NULL;
859
860 pio = (struct pnfsio *)context;
861 if (pnfsioq == NULL) {
862 if (nfs_pnfsiothreads == 0)
863 return (EPERM);
864 if (nfs_pnfsiothreads < 0)
865 nfs_pnfsiothreads = mp_ncpus * 4;
866 pnfsioq = taskqueue_create("pnfsioq", M_WAITOK,
867 taskqueue_thread_enqueue, &pnfsioq);
868 if (pnfsioq == NULL)
869 return (ENOMEM);
870 ret = taskqueue_start_threads(&pnfsioq, nfs_pnfsiothreads,
871 0, "pnfsiot");
872 if (ret != 0) {
873 taskqueue_free(pnfsioq);
874 pnfsioq = NULL;
875 return (ret);
876 }
877 }
878 pio->inprog = 1;
879 TASK_INIT(&pio->tsk, 0, func, context);
880 ret = taskqueue_enqueue(pnfsioq, &pio->tsk);
881 if (ret != 0)
882 pio->inprog = 0;
883 return (ret);
884 }
885
886 /*
887 * Initialize everything that needs to be initialized for a vnet.
888 */
889 static void
nfs_vnetinit(const void * unused __unused)890 nfs_vnetinit(const void *unused __unused)
891 {
892
893 if (IS_DEFAULT_VNET(curvnet))
894 VNET(nfsstatsv1_p) = &nfsstatsv1;
895 else
896 VNET(nfsstatsv1_p) = malloc(sizeof(struct nfsstatsv1),
897 M_TEMP, M_WAITOK | M_ZERO);
898 mtx_init(&VNET(nfsrv_nfsuserdsock).nr_mtx, "nfsuserd",
899 NULL, MTX_DEF);
900 }
901 VNET_SYSINIT(nfs_vnetinit, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
902 nfs_vnetinit, NULL);
903
904 static void
nfs_cleanup(void * unused __unused)905 nfs_cleanup(void *unused __unused)
906 {
907
908 mtx_destroy(&VNET(nfsrv_nfsuserdsock).nr_mtx);
909 if (!IS_DEFAULT_VNET(curvnet)) {
910 free(VNET(nfsstatsv1_p), M_TEMP);
911 VNET(nfsstatsv1_p) = NULL;
912 }
913 /* Clean out the name<-->id cache. */
914 nfsrv_cleanusergroup();
915 }
916 VNET_SYSUNINIT(nfs_cleanup, SI_SUB_VNET_DONE, SI_ORDER_FIRST,
917 nfs_cleanup, NULL);
918
919 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
920
921 /*
922 * Called once to initialize data structures...
923 */
924 static int
nfscommon_modevent(module_t mod,int type,void * data)925 nfscommon_modevent(module_t mod, int type, void *data)
926 {
927 int error = 0;
928 static int loaded = 0;
929
930 switch (type) {
931 case MOD_LOAD:
932 if (loaded)
933 goto out;
934 newnfs_portinit();
935 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
936 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
937 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
938 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
939 mtx_init(&nfsrv_dslock_mtx, "nfs4ds", NULL, MTX_DEF);
940 TAILQ_INIT(&nfsrv_devidhead);
941 newnfs_init();
942 nfsd_call_nfscommon = nfssvc_nfscommon;
943 loaded = 1;
944 break;
945
946 case MOD_UNLOAD:
947 if (newnfs_numnfsd != 0 ||
948 VNET(nfsrv_nfsuserd) != NOTRUNNING ||
949 nfs_numnfscbd != 0) {
950 error = EBUSY;
951 break;
952 }
953
954 nfsd_call_nfscommon = NULL;
955 /* and get rid of the mutexes */
956 mtx_destroy(&nfs_nameid_mutex);
957 mtx_destroy(&newnfsd_mtx);
958 mtx_destroy(&nfs_state_mutex);
959 mtx_destroy(&nfs_clstate_mutex);
960 mtx_destroy(&nfs_sockl_mutex);
961 mtx_destroy(&nfs_slock_mutex);
962 mtx_destroy(&nfs_req_mutex);
963 mtx_destroy(&nfsrv_dslock_mtx);
964 loaded = 0;
965 break;
966 default:
967 error = EOPNOTSUPP;
968 break;
969 }
970
971 out:
972 NFSEXITCODE(error);
973 return error;
974 }
975 static moduledata_t nfscommon_mod = {
976 "nfscommon",
977 nfscommon_modevent,
978 NULL,
979 };
980 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
981
982 /* So that loader and kldload(2) can find us, wherever we are.. */
983 MODULE_VERSION(nfscommon, 1);
984 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
985 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
986