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 int nfscl_debuglevel = 0;
64 char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
65 struct callout newnfsd_callout;
66 void (*nfsd_call_servertimer)(void) = NULL;
67 void (*ncl_call_invalcaches)(struct vnode *) = NULL;
69 static int nfs_realign_test;
70 static int nfs_realign_count;
72 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "New NFS filesystem");
73 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
74 0, "Number of realign tests done");
75 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
76 0, "Number of mbuf realignments done");
77 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
78 nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
79 "NFSv4 callback addr for server to use");
80 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
81 0, "Debug level for new nfs client");
85 * (Here for FreeBSD, since they allocate storage.)
87 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
88 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
89 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
90 "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
91 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
92 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
93 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
94 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
95 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
96 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
97 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
98 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
99 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
100 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
101 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
102 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
103 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "New nfs vnode");
104 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "New nfs Direct IO buffer");
105 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
106 "New NFS directory offset data");
107 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
108 "New NFS local lock rollback");
109 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
110 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
111 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
112 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
113 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
114 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
115 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
118 * Definition of mutex locks.
119 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
120 * and assorted other nfsd structures.
122 struct mtx newnfsd_mtx;
123 struct mtx nfs_sockl_mutex;
124 struct mtx nfs_state_mutex;
125 struct mtx nfs_nameid_mutex;
126 struct mtx nfs_req_mutex;
127 struct mtx nfs_slock_mutex;
129 /* local functions */
130 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
132 #ifdef __NO_STRICT_ALIGNMENT
134 * These architectures don't need re-alignment, so just return.
137 newnfs_realign(struct mbuf **pm, int how)
142 #else /* !__NO_STRICT_ALIGNMENT */
146 * Check for badly aligned mbuf data and realign by copying the unaligned
147 * portion of the data into a new mbuf chain and freeing the portions
148 * of the old chain that were replaced.
150 * We cannot simply realign the data within the existing mbuf chain
151 * because the underlying buffers may contain other rpc commands and
152 * we cannot afford to overwrite them.
154 * We would prefer to avoid this situation entirely. The situation does
155 * not occur with NFS/UDP and is supposed to only occassionally occur
156 * with TCP. Use vfs.nfs.realign_count and realign_test to check this.
160 newnfs_realign(struct mbuf **pm, int how)
166 while ((m = *pm) != NULL) {
167 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
169 * NB: we can't depend on m_pkthdr.len to help us
170 * decide what to do here. May not be worth doing
171 * the m_length calculation as m_copyback will
172 * expand the mbuf chain below as needed.
174 space = m_length(m, NULL);
175 if (space >= MINCLSIZE) {
176 /* NB: m_copyback handles space > MCLBYTES */
177 n = m_getcl(how, MT_DATA, 0);
179 n = m_get(how, MT_DATA);
183 * Align the remainder of the mbuf chain.
188 m_copyback(n, off, m->m_len, mtod(m, caddr_t));
202 #endif /* __NO_STRICT_ALIGNMENT */
206 nfsrv_object_create(struct vnode *vp, struct thread *td)
209 if (vp == NULL || vp->v_type != VREG)
211 (void) vfs_object_create(vp, td, td->td_ucred);
216 * Look up a file name. Basically just initialize stuff and call namei().
219 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
223 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
227 NDFREE(ndp, NDF_ONLY_PNBUF);
233 * Copy NFS uid, gids to the cred structure.
236 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
239 KASSERT(nfscr->nfsc_ngroups >= 0,
240 ("newnfs_copycred: negative nfsc_ngroups"));
241 cr->cr_uid = nfscr->nfsc_uid;
242 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
246 * Map args from nfsmsleep() to msleep().
249 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
256 timeo = hz * ts->tv_sec;
257 nsecval = (u_int64_t)ts->tv_nsec;
258 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
260 timeo += (int)nsecval;
264 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
269 * Get the file system info for the server. For now, just assume FFS.
272 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
278 * There should be file system VFS OP(s) to get this information.
279 * For now, assume ufs.
282 pref = NFS_MAXDGRAMDATA;
285 sip->fs_rtmax = NFS_MAXDATA;
286 sip->fs_rtpref = pref;
287 sip->fs_rtmult = NFS_FABLKSIZE;
288 sip->fs_wtmax = NFS_MAXDATA;
289 sip->fs_wtpref = pref;
290 sip->fs_wtmult = NFS_FABLKSIZE;
291 sip->fs_dtpref = pref;
292 sip->fs_maxfilesize = 0xffffffffffffffffull;
293 sip->fs_timedelta.tv_sec = 0;
294 sip->fs_timedelta.tv_nsec = 1;
295 sip->fs_properties = (NFSV3FSINFO_LINK |
296 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
297 NFSV3FSINFO_CANSETTIME);
301 * Do the pathconf vnode op.
304 nfsvno_pathconf(struct vnode *vp, int flag, register_t *retf,
305 struct ucred *cred, struct thread *p)
309 error = VOP_PATHCONF(vp, flag, retf);
310 if (error == EOPNOTSUPP || error == EINVAL) {
312 * Some file systems return EINVAL for name arguments not
313 * supported and some return EOPNOTSUPP for this case.
314 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
324 case _PC_CHOWN_RESTRICTED:
332 * Only happens if a _PC_xxx is added to the server,
333 * but this isn't updated.
336 printf("nfsrvd pathconf flag=%d not supp\n", flag);
344 /* Fake nfsrv_atroot. Just return 0 */
346 nfsrv_atroot(struct vnode *vp, long *retp)
353 * Set the credentials to refer to root.
354 * If only the various BSDen could agree on whether cr_gid is a separate
355 * field or cr_groups[0]...
358 newnfs_setroot(struct ucred *cred)
362 cred->cr_groups[0] = 0;
363 cred->cr_ngroups = 1;
367 * Get the client credential. Used for Renew and recovery.
373 struct thread *td = curthread;
375 cred = crdup(td->td_ucred);
376 newnfs_setroot(cred);
382 * Call the nfsd's timer function once/sec.
385 newnfs_timer(void *arg)
387 static time_t lasttime = 0;
389 * Call the server timer, if set up.
390 * The argument indicates if it is the next second and therefore
391 * leases should be checked.
393 if (lasttime != NFSD_MONOSEC) {
394 lasttime = NFSD_MONOSEC;
395 if (nfsd_call_servertimer != NULL)
396 (*nfsd_call_servertimer)();
398 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
403 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
404 * the sleep should be for 5 seconds.
405 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
406 * an event that never gets a wakeup. Only return EINTR or 0.
409 nfs_catnap(int prio, int errval, const char *wmesg)
411 static int non_event;
414 if (errval == NFSERR_GRACE)
415 ret = tsleep(&non_event, prio, wmesg, 5 * hz);
417 ret = tsleep(&non_event, prio, wmesg, 1);
424 * Get referral. For now, just fail.
427 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
434 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
438 error = nfssvc_call(td, uap, td->td_ucred);
444 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
447 struct nfsd_idargs nid;
449 if (uap->flag & NFSSVC_IDNAME) {
450 error = copyin(uap->argp, (caddr_t)&nid, sizeof (nid));
453 error = nfssvc_idname(&nid);
455 } else if (uap->flag & NFSSVC_GETSTATS) {
456 error = copyout(&newnfsstats,
457 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
459 if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
460 newnfsstats.attrcache_hits = 0;
461 newnfsstats.attrcache_misses = 0;
462 newnfsstats.lookupcache_hits = 0;
463 newnfsstats.lookupcache_misses = 0;
464 newnfsstats.direofcache_hits = 0;
465 newnfsstats.direofcache_misses = 0;
466 newnfsstats.accesscache_hits = 0;
467 newnfsstats.accesscache_misses = 0;
468 newnfsstats.biocache_reads = 0;
469 newnfsstats.read_bios = 0;
470 newnfsstats.read_physios = 0;
471 newnfsstats.biocache_writes = 0;
472 newnfsstats.write_bios = 0;
473 newnfsstats.write_physios = 0;
474 newnfsstats.biocache_readlinks = 0;
475 newnfsstats.readlink_bios = 0;
476 newnfsstats.biocache_readdirs = 0;
477 newnfsstats.readdir_bios = 0;
478 newnfsstats.rpcretries = 0;
479 newnfsstats.rpcrequests = 0;
480 newnfsstats.rpctimeouts = 0;
481 newnfsstats.rpcunexpected = 0;
482 newnfsstats.rpcinvalid = 0;
483 bzero(newnfsstats.rpccnt,
484 sizeof(newnfsstats.rpccnt));
486 if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
487 newnfsstats.srvrpc_errs = 0;
488 newnfsstats.srv_errs = 0;
489 newnfsstats.srvcache_inproghits = 0;
490 newnfsstats.srvcache_idemdonehits = 0;
491 newnfsstats.srvcache_nonidemdonehits = 0;
492 newnfsstats.srvcache_misses = 0;
493 newnfsstats.srvcache_tcppeak = 0;
494 newnfsstats.srvclients = 0;
495 newnfsstats.srvopenowners = 0;
496 newnfsstats.srvopens = 0;
497 newnfsstats.srvlockowners = 0;
498 newnfsstats.srvlocks = 0;
499 newnfsstats.srvdelegates = 0;
500 newnfsstats.clopenowners = 0;
501 newnfsstats.clopens = 0;
502 newnfsstats.cllockowners = 0;
503 newnfsstats.cllocks = 0;
504 newnfsstats.cldelegates = 0;
505 newnfsstats.cllocalopenowners = 0;
506 newnfsstats.cllocalopens = 0;
507 newnfsstats.cllocallockowners = 0;
508 newnfsstats.cllocallocks = 0;
509 bzero(newnfsstats.srvrpccnt,
510 sizeof(newnfsstats.srvrpccnt));
511 bzero(newnfsstats.cbrpccnt,
512 sizeof(newnfsstats.cbrpccnt));
516 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
519 error = copyin(uap->argp, (caddr_t)&sockport,
522 error = nfsrv_nfsuserdport(sockport, p);
523 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
524 nfsrv_nfsuserddelport();
534 * called by all three modevent routines, so that it gets things
535 * initialized soon enough.
538 newnfs_portinit(void)
540 static int inited = 0;
545 /* Initialize SMP locks used by both client and server. */
546 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
547 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
551 * Determine if the file system supports NFSv4 ACLs.
552 * Return 1 if it does, 0 otherwise.
555 nfs_supportsnfsv4acls(struct vnode *vp)
560 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
562 if (nfsrv_useacl == 0)
564 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
565 if (error == 0 && retval != 0)
570 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
573 * Called once to initialize data structures...
576 nfscommon_modevent(module_t mod, int type, void *data)
579 static int loaded = 0;
586 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
587 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
588 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
589 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
590 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
592 callout_init(&newnfsd_callout, CALLOUT_MPSAFE);
594 nfsd_call_nfscommon = nfssvc_nfscommon;
599 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
600 nfs_numnfscbd != 0) {
605 nfsd_call_nfscommon = NULL;
606 callout_drain(&newnfsd_callout);
607 /* and get rid of the mutexes */
608 mtx_destroy(&nfs_nameid_mutex);
609 mtx_destroy(&newnfsd_mtx);
610 mtx_destroy(&nfs_state_mutex);
611 mtx_destroy(&nfs_sockl_mutex);
612 mtx_destroy(&nfs_slock_mutex);
613 mtx_destroy(&nfs_req_mutex);
614 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
626 static moduledata_t nfscommon_mod = {
631 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
633 /* So that loader and kldload(2) can find us, wherever we are.. */
634 MODULE_VERSION(nfscommon, 1);
635 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
636 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);