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>
52 #include <vm/uma_int.h>
54 extern int nfscl_ticks;
55 extern int nfsrv_nfsuserd;
56 extern struct nfssockreq nfsrv_nfsuserdsock;
57 extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
59 extern int nfsrv_useacl;
60 struct mount nfsv4root_mnt;
61 int newnfs_numnfsd = 0;
62 struct nfsstats newnfsstats;
63 int nfs_numnfscbd = 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");
83 * (Here for FreeBSD, since they allocate storage.)
85 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
86 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
87 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
88 "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
89 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
90 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
91 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
92 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
93 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
94 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
95 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
96 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
97 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
98 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
99 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
100 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
101 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "New nfs vnode");
102 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "New nfs Direct IO buffer");
103 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
104 "New NFS directory offset data");
105 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
106 "New NFS local lock rollback");
109 * Definition of mutex locks.
110 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
111 * and assorted other nfsd structures.
112 * Giant is used to protect the nfsd list and count, which is just
113 * updated when nfsd's start/stop and is grabbed for nfsrvd_dorpc()
116 struct mtx newnfsd_mtx;
117 struct mtx nfs_sockl_mutex;
118 struct mtx nfs_state_mutex;
119 struct mtx nfs_nameid_mutex;
120 struct mtx nfs_req_mutex;
121 struct mtx nfs_slock_mutex;
123 /* local functions */
124 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
126 #ifdef __NO_STRICT_ALIGNMENT
128 * These architectures don't need re-alignment, so just return.
131 newnfs_realign(struct mbuf **pm)
136 #else /* !__NO_STRICT_ALIGNMENT */
140 * Check for badly aligned mbuf data and realign by copying the unaligned
141 * portion of the data into a new mbuf chain and freeing the portions
142 * of the old chain that were replaced.
144 * We cannot simply realign the data within the existing mbuf chain
145 * because the underlying buffers may contain other rpc commands and
146 * we cannot afford to overwrite them.
148 * We would prefer to avoid this situation entirely. The situation does
149 * not occur with NFS/UDP and is supposed to only occassionally occur
150 * with TCP. Use vfs.nfs.realign_count and realign_test to check this.
154 newnfs_realign(struct mbuf **pm)
160 while ((m = *pm) != NULL) {
161 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
163 * NB: we can't depend on m_pkthdr.len to help us
164 * decide what to do here. May not be worth doing
165 * the m_length calculation as m_copyback will
166 * expand the mbuf chain below as needed.
168 space = m_length(m, NULL);
169 if (space >= MINCLSIZE) {
170 /* NB: m_copyback handles space > MCLBYTES */
171 n = m_getcl(M_WAITOK, MT_DATA, 0);
173 n = m_get(M_WAITOK, MT_DATA);
177 * Align the remainder of the mbuf chain.
182 m_copyback(n, off, m->m_len, mtod(m, caddr_t));
194 #endif /* __NO_STRICT_ALIGNMENT */
198 nfsrv_object_create(struct vnode *vp, struct thread *td)
201 if (vp == NULL || vp->v_type != VREG)
203 (void) vfs_object_create(vp, td, td->td_ucred);
208 * Look up a file name. Basically just initialize stuff and call namei().
211 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
215 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, UIO_USERSPACE, fname,
219 NDFREE(ndp, NDF_ONLY_PNBUF);
225 * Copy NFS uid, gids to the cred structure.
228 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
231 KASSERT(nfscr->nfsc_ngroups >= 0,
232 ("newnfs_copycred: negative nfsc_ngroups"));
233 cr->cr_uid = nfscr->nfsc_uid;
234 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
238 * Map args from nfsmsleep() to msleep().
241 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
248 timeo = hz * ts->tv_sec;
249 nsecval = (u_int64_t)ts->tv_nsec;
250 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
252 timeo += (int)nsecval;
256 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
261 * Get the file system info for the server. For now, just assume FFS.
264 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
270 * There should be file system VFS OP(s) to get this information.
271 * For now, assume ufs.
274 pref = NFS_MAXDGRAMDATA;
277 sip->fs_rtmax = NFS_MAXDATA;
278 sip->fs_rtpref = pref;
279 sip->fs_rtmult = NFS_FABLKSIZE;
280 sip->fs_wtmax = NFS_MAXDATA;
281 sip->fs_wtpref = pref;
282 sip->fs_wtmult = NFS_FABLKSIZE;
283 sip->fs_dtpref = pref;
284 sip->fs_maxfilesize = 0xffffffffffffffffull;
285 sip->fs_timedelta.tv_sec = 0;
286 sip->fs_timedelta.tv_nsec = 1;
287 sip->fs_properties = (NFSV3FSINFO_LINK |
288 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
289 NFSV3FSINFO_CANSETTIME);
293 * Do the pathconf vnode op.
296 nfsvno_pathconf(struct vnode *vp, int flag, register_t *retf,
297 struct ucred *cred, struct thread *p)
301 error = VOP_PATHCONF(vp, flag, retf);
302 if (error == EOPNOTSUPP || error == EINVAL) {
304 * Some file systems return EINVAL for name arguments not
305 * supported and some return EOPNOTSUPP for this case.
306 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
316 case _PC_CHOWN_RESTRICTED:
324 * Only happens if a _PC_xxx is added to the server,
325 * but this isn't updated.
328 printf("nfsrvd pathconf flag=%d not supp\n", flag);
336 /* Fake nfsrv_atroot. Just return 0 */
338 nfsrv_atroot(struct vnode *vp, long *retp)
345 * Set the credentials to refer to root.
346 * If only the various BSDen could agree on whether cr_gid is a separate
347 * field or cr_groups[0]...
350 newnfs_setroot(struct ucred *cred)
354 cred->cr_groups[0] = 0;
355 cred->cr_ngroups = 1;
359 * Get the client credential. Used for Renew and recovery.
365 struct thread *td = curthread;
367 cred = crdup(td->td_ucred);
368 newnfs_setroot(cred);
374 * Call the nfsd's timer function once/sec.
377 newnfs_timer(void *arg)
379 static time_t lasttime = 0;
381 * Call the server timer, if set up.
382 * The argument indicates if it is the next second and therefore
383 * leases should be checked.
385 if (lasttime != NFSD_MONOSEC) {
386 lasttime = NFSD_MONOSEC;
387 if (nfsd_call_servertimer != NULL)
388 (*nfsd_call_servertimer)();
390 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
395 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
396 * the sleep should be for 5 seconds.
397 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
398 * an event that never gets a wakeup. Only return EINTR or 0.
401 nfs_catnap(int prio, int errval, const char *wmesg)
403 static int non_event;
406 if (errval == NFSERR_GRACE)
407 ret = tsleep(&non_event, prio, wmesg, 5 * hz);
409 ret = tsleep(&non_event, prio, wmesg, 1);
416 * Get referral. For now, just fail.
419 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
426 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
430 error = nfssvc_call(td, uap, td->td_ucred);
436 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
439 struct nfsd_idargs nid;
441 if (uap->flag & NFSSVC_IDNAME) {
442 error = copyin(uap->argp, (caddr_t)&nid, sizeof (nid));
445 error = nfssvc_idname(&nid);
447 } else if (uap->flag & NFSSVC_GETSTATS) {
448 error = copyout(&newnfsstats,
449 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
451 if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
452 newnfsstats.attrcache_hits = 0;
453 newnfsstats.attrcache_misses = 0;
454 newnfsstats.lookupcache_hits = 0;
455 newnfsstats.lookupcache_misses = 0;
456 newnfsstats.direofcache_hits = 0;
457 newnfsstats.direofcache_misses = 0;
458 newnfsstats.accesscache_hits = 0;
459 newnfsstats.accesscache_misses = 0;
460 newnfsstats.biocache_reads = 0;
461 newnfsstats.read_bios = 0;
462 newnfsstats.read_physios = 0;
463 newnfsstats.biocache_writes = 0;
464 newnfsstats.write_bios = 0;
465 newnfsstats.write_physios = 0;
466 newnfsstats.biocache_readlinks = 0;
467 newnfsstats.readlink_bios = 0;
468 newnfsstats.biocache_readdirs = 0;
469 newnfsstats.readdir_bios = 0;
470 newnfsstats.rpcretries = 0;
471 newnfsstats.rpcrequests = 0;
472 newnfsstats.rpctimeouts = 0;
473 newnfsstats.rpcunexpected = 0;
474 newnfsstats.rpcinvalid = 0;
475 bzero(newnfsstats.rpccnt,
476 sizeof(newnfsstats.rpccnt));
478 if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
479 newnfsstats.srvrpc_errs = 0;
480 newnfsstats.srv_errs = 0;
481 newnfsstats.srvcache_inproghits = 0;
482 newnfsstats.srvcache_idemdonehits = 0;
483 newnfsstats.srvcache_nonidemdonehits = 0;
484 newnfsstats.srvcache_misses = 0;
485 newnfsstats.srvcache_tcppeak = 0;
486 newnfsstats.srvcache_size = 0;
487 newnfsstats.srvclients = 0;
488 newnfsstats.srvopenowners = 0;
489 newnfsstats.srvopens = 0;
490 newnfsstats.srvlockowners = 0;
491 newnfsstats.srvlocks = 0;
492 newnfsstats.srvdelegates = 0;
493 newnfsstats.clopenowners = 0;
494 newnfsstats.clopens = 0;
495 newnfsstats.cllockowners = 0;
496 newnfsstats.cllocks = 0;
497 newnfsstats.cldelegates = 0;
498 newnfsstats.cllocalopenowners = 0;
499 newnfsstats.cllocalopens = 0;
500 newnfsstats.cllocallockowners = 0;
501 newnfsstats.cllocallocks = 0;
502 bzero(newnfsstats.srvrpccnt,
503 sizeof(newnfsstats.srvrpccnt));
504 bzero(newnfsstats.cbrpccnt,
505 sizeof(newnfsstats.cbrpccnt));
509 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
512 error = copyin(uap->argp, (caddr_t)&sockport,
515 error = nfsrv_nfsuserdport(sockport, p);
516 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
517 nfsrv_nfsuserddelport();
527 * called by all three modevent routines, so that it gets things
528 * initialized soon enough.
531 newnfs_portinit(void)
533 static int inited = 0;
538 /* Initialize SMP locks used by both client and server. */
539 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
540 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
544 * Determine if the file system supports NFSv4 ACLs.
545 * Return 1 if it does, 0 otherwise.
548 nfs_supportsnfsv4acls(struct vnode *vp)
553 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
555 if (nfsrv_useacl == 0)
557 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
558 if (error == 0 && retval != 0)
563 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
566 * Called once to initialize data structures...
569 nfscommon_modevent(module_t mod, int type, void *data)
572 static int loaded = 0;
579 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
580 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
581 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
582 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
583 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
585 callout_init(&newnfsd_callout, CALLOUT_MPSAFE);
587 nfsd_call_nfscommon = nfssvc_nfscommon;
592 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
593 nfs_numnfscbd != 0) {
598 nfsd_call_nfscommon = NULL;
599 callout_drain(&newnfsd_callout);
600 /* and get rid of the mutexes */
601 mtx_destroy(&nfs_nameid_mutex);
602 mtx_destroy(&newnfsd_mtx);
603 mtx_destroy(&nfs_state_mutex);
604 mtx_destroy(&nfs_sockl_mutex);
605 mtx_destroy(&nfs_slock_mutex);
606 mtx_destroy(&nfs_req_mutex);
607 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
619 static moduledata_t nfscommon_mod = {
624 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
626 /* So that loader and kldload(2) can find us, wherever we are.. */
627 MODULE_VERSION(nfscommon, 1);
628 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
629 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);