2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
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.
42 #include <fs/nfs/nfsport.h>
43 #include <sys/sysctl.h>
45 #include <vm/vm_object.h>
46 #include <vm/vm_page.h>
47 #include <vm/vm_param.h>
48 #include <vm/vm_map.h>
49 #include <vm/vm_kern.h>
50 #include <vm/vm_extern.h>
53 extern int nfscl_ticks;
54 extern int nfsrv_nfsuserd;
55 extern struct nfssockreq nfsrv_nfsuserdsock;
56 extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
58 extern int nfsrv_useacl;
59 struct mount nfsv4root_mnt;
60 int newnfs_numnfsd = 0;
61 struct nfsstats newnfsstats;
62 int nfs_numnfscbd = 0;
63 char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
64 struct callout newnfsd_callout;
65 void (*nfsd_call_servertimer)(void) = NULL;
66 void (*ncl_call_invalcaches)(struct vnode *) = NULL;
68 static int nfs_realign_test;
69 static int nfs_realign_count;
71 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "New NFS filesystem");
72 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
73 0, "Number of realign tests done");
74 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
75 0, "Number of mbuf realignments done");
76 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
77 nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
78 "NFSv4 callback addr for server to use");
82 * (Here for FreeBSD, since they allocate storage.)
84 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
85 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
86 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
87 "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
88 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
89 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
90 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
91 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
92 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
93 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
94 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
95 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
96 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
97 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
98 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
99 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
100 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "New nfs vnode");
101 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "New nfs Direct IO buffer");
102 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
103 "New NFS directory offset data");
104 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
105 "New NFS local lock rollback");
108 * Definition of mutex locks.
109 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
110 * and assorted other nfsd structures.
112 struct mtx newnfsd_mtx;
113 struct mtx nfs_sockl_mutex;
114 struct mtx nfs_state_mutex;
115 struct mtx nfs_nameid_mutex;
116 struct mtx nfs_req_mutex;
117 struct mtx nfs_slock_mutex;
119 /* local functions */
120 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
122 #ifdef __NO_STRICT_ALIGNMENT
124 * These architectures don't need re-alignment, so just return.
127 newnfs_realign(struct mbuf **pm)
132 #else /* !__NO_STRICT_ALIGNMENT */
136 * Check for badly aligned mbuf data and realign by copying the unaligned
137 * portion of the data into a new mbuf chain and freeing the portions
138 * of the old chain that were replaced.
140 * We cannot simply realign the data within the existing mbuf chain
141 * because the underlying buffers may contain other rpc commands and
142 * we cannot afford to overwrite them.
144 * We would prefer to avoid this situation entirely. The situation does
145 * not occur with NFS/UDP and is supposed to only occassionally occur
146 * with TCP. Use vfs.nfs.realign_count and realign_test to check this.
150 newnfs_realign(struct mbuf **pm)
156 while ((m = *pm) != NULL) {
157 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
159 * NB: we can't depend on m_pkthdr.len to help us
160 * decide what to do here. May not be worth doing
161 * the m_length calculation as m_copyback will
162 * expand the mbuf chain below as needed.
164 space = m_length(m, NULL);
165 if (space >= MINCLSIZE) {
166 /* NB: m_copyback handles space > MCLBYTES */
167 n = m_getcl(M_WAITOK, MT_DATA, 0);
169 n = m_get(M_WAITOK, MT_DATA);
173 * Align the remainder of the mbuf chain.
178 m_copyback(n, off, m->m_len, mtod(m, caddr_t));
190 #endif /* __NO_STRICT_ALIGNMENT */
194 nfsrv_object_create(struct vnode *vp, struct thread *td)
197 if (vp == NULL || vp->v_type != VREG)
199 (void) vfs_object_create(vp, td, td->td_ucred);
204 * Look up a file name. Basically just initialize stuff and call namei().
207 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
211 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, UIO_USERSPACE, fname,
215 NDFREE(ndp, NDF_ONLY_PNBUF);
221 * Copy NFS uid, gids to the cred structure.
224 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
227 KASSERT(nfscr->nfsc_ngroups >= 0,
228 ("newnfs_copycred: negative nfsc_ngroups"));
229 cr->cr_uid = nfscr->nfsc_uid;
230 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
234 * Map args from nfsmsleep() to msleep().
237 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
244 timeo = hz * ts->tv_sec;
245 nsecval = (u_int64_t)ts->tv_nsec;
246 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
248 timeo += (int)nsecval;
252 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
257 * Get the file system info for the server. For now, just assume FFS.
260 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
266 * There should be file system VFS OP(s) to get this information.
267 * For now, assume ufs.
270 pref = NFS_MAXDGRAMDATA;
273 sip->fs_rtmax = NFS_MAXDATA;
274 sip->fs_rtpref = pref;
275 sip->fs_rtmult = NFS_FABLKSIZE;
276 sip->fs_wtmax = NFS_MAXDATA;
277 sip->fs_wtpref = pref;
278 sip->fs_wtmult = NFS_FABLKSIZE;
279 sip->fs_dtpref = pref;
280 sip->fs_maxfilesize = 0xffffffffffffffffull;
281 sip->fs_timedelta.tv_sec = 0;
282 sip->fs_timedelta.tv_nsec = 1;
283 sip->fs_properties = (NFSV3FSINFO_LINK |
284 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
285 NFSV3FSINFO_CANSETTIME);
289 * Do the pathconf vnode op.
292 nfsvno_pathconf(struct vnode *vp, int flag, register_t *retf,
293 struct ucred *cred, struct thread *p)
297 error = VOP_PATHCONF(vp, flag, retf);
298 if (error == EOPNOTSUPP || error == EINVAL) {
300 * Some file systems return EINVAL for name arguments not
301 * supported and some return EOPNOTSUPP for this case.
302 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
312 case _PC_CHOWN_RESTRICTED:
320 * Only happens if a _PC_xxx is added to the server,
321 * but this isn't updated.
324 printf("nfsrvd pathconf flag=%d not supp\n", flag);
332 /* Fake nfsrv_atroot. Just return 0 */
334 nfsrv_atroot(struct vnode *vp, long *retp)
341 * Set the credentials to refer to root.
342 * If only the various BSDen could agree on whether cr_gid is a separate
343 * field or cr_groups[0]...
346 newnfs_setroot(struct ucred *cred)
350 cred->cr_groups[0] = 0;
351 cred->cr_ngroups = 1;
355 * Get the client credential. Used for Renew and recovery.
361 struct thread *td = curthread;
363 cred = crdup(td->td_ucred);
364 newnfs_setroot(cred);
370 * Call the nfsd's timer function once/sec.
373 newnfs_timer(void *arg)
375 static time_t lasttime = 0;
377 * Call the server timer, if set up.
378 * The argument indicates if it is the next second and therefore
379 * leases should be checked.
381 if (lasttime != NFSD_MONOSEC) {
382 lasttime = NFSD_MONOSEC;
383 if (nfsd_call_servertimer != NULL)
384 (*nfsd_call_servertimer)();
386 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
391 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
392 * the sleep should be for 5 seconds.
393 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
394 * an event that never gets a wakeup. Only return EINTR or 0.
397 nfs_catnap(int prio, int errval, const char *wmesg)
399 static int non_event;
402 if (errval == NFSERR_GRACE)
403 ret = tsleep(&non_event, prio, wmesg, 5 * hz);
405 ret = tsleep(&non_event, prio, wmesg, 1);
412 * Get referral. For now, just fail.
415 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
422 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
426 error = nfssvc_call(td, uap, td->td_ucred);
432 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
435 struct nfsd_idargs nid;
437 if (uap->flag & NFSSVC_IDNAME) {
438 error = copyin(uap->argp, (caddr_t)&nid, sizeof (nid));
441 error = nfssvc_idname(&nid);
443 } else if (uap->flag & NFSSVC_GETSTATS) {
444 error = copyout(&newnfsstats,
445 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
447 if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
448 newnfsstats.attrcache_hits = 0;
449 newnfsstats.attrcache_misses = 0;
450 newnfsstats.lookupcache_hits = 0;
451 newnfsstats.lookupcache_misses = 0;
452 newnfsstats.direofcache_hits = 0;
453 newnfsstats.direofcache_misses = 0;
454 newnfsstats.accesscache_hits = 0;
455 newnfsstats.accesscache_misses = 0;
456 newnfsstats.biocache_reads = 0;
457 newnfsstats.read_bios = 0;
458 newnfsstats.read_physios = 0;
459 newnfsstats.biocache_writes = 0;
460 newnfsstats.write_bios = 0;
461 newnfsstats.write_physios = 0;
462 newnfsstats.biocache_readlinks = 0;
463 newnfsstats.readlink_bios = 0;
464 newnfsstats.biocache_readdirs = 0;
465 newnfsstats.readdir_bios = 0;
466 newnfsstats.rpcretries = 0;
467 newnfsstats.rpcrequests = 0;
468 newnfsstats.rpctimeouts = 0;
469 newnfsstats.rpcunexpected = 0;
470 newnfsstats.rpcinvalid = 0;
471 bzero(newnfsstats.rpccnt,
472 sizeof(newnfsstats.rpccnt));
474 if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
475 newnfsstats.srvrpc_errs = 0;
476 newnfsstats.srv_errs = 0;
477 newnfsstats.srvcache_inproghits = 0;
478 newnfsstats.srvcache_idemdonehits = 0;
479 newnfsstats.srvcache_nonidemdonehits = 0;
480 newnfsstats.srvcache_misses = 0;
481 newnfsstats.srvcache_tcppeak = 0;
482 newnfsstats.srvclients = 0;
483 newnfsstats.srvopenowners = 0;
484 newnfsstats.srvopens = 0;
485 newnfsstats.srvlockowners = 0;
486 newnfsstats.srvlocks = 0;
487 newnfsstats.srvdelegates = 0;
488 newnfsstats.clopenowners = 0;
489 newnfsstats.clopens = 0;
490 newnfsstats.cllockowners = 0;
491 newnfsstats.cllocks = 0;
492 newnfsstats.cldelegates = 0;
493 newnfsstats.cllocalopenowners = 0;
494 newnfsstats.cllocalopens = 0;
495 newnfsstats.cllocallockowners = 0;
496 newnfsstats.cllocallocks = 0;
497 bzero(newnfsstats.srvrpccnt,
498 sizeof(newnfsstats.srvrpccnt));
499 bzero(newnfsstats.cbrpccnt,
500 sizeof(newnfsstats.cbrpccnt));
504 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
507 error = copyin(uap->argp, (caddr_t)&sockport,
510 error = nfsrv_nfsuserdport(sockport, p);
511 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
512 nfsrv_nfsuserddelport();
522 * called by all three modevent routines, so that it gets things
523 * initialized soon enough.
526 newnfs_portinit(void)
528 static int inited = 0;
533 /* Initialize SMP locks used by both client and server. */
534 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
535 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
539 * Determine if the file system supports NFSv4 ACLs.
540 * Return 1 if it does, 0 otherwise.
543 nfs_supportsnfsv4acls(struct vnode *vp)
548 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
550 if (nfsrv_useacl == 0)
552 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
553 if (error == 0 && retval != 0)
558 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
561 * Called once to initialize data structures...
564 nfscommon_modevent(module_t mod, int type, void *data)
567 static int loaded = 0;
574 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
575 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
576 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
577 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
578 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
580 callout_init(&newnfsd_callout, CALLOUT_MPSAFE);
582 nfsd_call_nfscommon = nfssvc_nfscommon;
587 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
588 nfs_numnfscbd != 0) {
593 nfsd_call_nfscommon = NULL;
594 callout_drain(&newnfsd_callout);
595 /* and get rid of the mutexes */
596 mtx_destroy(&nfs_nameid_mutex);
597 mtx_destroy(&newnfsd_mtx);
598 mtx_destroy(&nfs_state_mutex);
599 mtx_destroy(&nfs_sockl_mutex);
600 mtx_destroy(&nfs_slock_mutex);
601 mtx_destroy(&nfs_req_mutex);
602 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
614 static moduledata_t nfscommon_mod = {
619 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
621 /* So that loader and kldload(2) can find us, wherever we are.. */
622 MODULE_VERSION(nfscommon, 1);
623 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
624 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);