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 int nfsrv_lughashsize = 100;
67 void (*nfsd_call_servertimer)(void) = NULL;
68 void (*ncl_call_invalcaches)(struct vnode *) = NULL;
70 static int nfs_realign_test;
71 static int nfs_realign_count;
73 SYSCTL_NODE(_vfs, OID_AUTO, nfs, CTLFLAG_RW, 0, "New NFS filesystem");
74 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_test, CTLFLAG_RW, &nfs_realign_test,
75 0, "Number of realign tests done");
76 SYSCTL_INT(_vfs_nfs, OID_AUTO, realign_count, CTLFLAG_RW, &nfs_realign_count,
77 0, "Number of mbuf realignments done");
78 SYSCTL_STRING(_vfs_nfs, OID_AUTO, callback_addr, CTLFLAG_RW,
79 nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr),
80 "NFSv4 callback addr for server to use");
81 SYSCTL_INT(_vfs_nfs, OID_AUTO, debuglevel, CTLFLAG_RW, &nfscl_debuglevel,
82 0, "Debug level for new nfs client");
83 TUNABLE_INT("vfs.nfs.userhashsize", &nfsrv_lughashsize);
84 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
85 0, "Size of hash tables for uid/name mapping");
89 * (Here for FreeBSD, since they allocate storage.)
91 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
92 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
93 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
94 "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
95 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
96 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
97 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
98 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
99 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
100 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
101 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
102 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
103 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
104 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
105 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
106 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
107 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "New nfs vnode");
108 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "New nfs Direct IO buffer");
109 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
110 "New NFS directory offset data");
111 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
112 "New NFS local lock rollback");
113 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
114 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
115 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
116 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
117 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
118 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
119 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
122 * Definition of mutex locks.
123 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
124 * and assorted other nfsd structures.
126 struct mtx newnfsd_mtx;
127 struct mtx nfs_sockl_mutex;
128 struct mtx nfs_state_mutex;
129 struct mtx nfs_nameid_mutex;
130 struct mtx nfs_req_mutex;
131 struct mtx nfs_slock_mutex;
132 struct mtx nfs_clstate_mutex;
134 /* local functions */
135 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
137 #ifdef __NO_STRICT_ALIGNMENT
139 * These architectures don't need re-alignment, so just return.
142 newnfs_realign(struct mbuf **pm, int how)
147 #else /* !__NO_STRICT_ALIGNMENT */
151 * Check for badly aligned mbuf data and realign by copying the unaligned
152 * portion of the data into a new mbuf chain and freeing the portions
153 * of the old chain that were replaced.
155 * We cannot simply realign the data within the existing mbuf chain
156 * because the underlying buffers may contain other rpc commands and
157 * we cannot afford to overwrite them.
159 * We would prefer to avoid this situation entirely. The situation does
160 * not occur with NFS/UDP and is supposed to only occassionally occur
161 * with TCP. Use vfs.nfs.realign_count and realign_test to check this.
165 newnfs_realign(struct mbuf **pm, int how)
171 while ((m = *pm) != NULL) {
172 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
174 * NB: we can't depend on m_pkthdr.len to help us
175 * decide what to do here. May not be worth doing
176 * the m_length calculation as m_copyback will
177 * expand the mbuf chain below as needed.
179 space = m_length(m, NULL);
180 if (space >= MINCLSIZE) {
181 /* NB: m_copyback handles space > MCLBYTES */
182 n = m_getcl(how, MT_DATA, 0);
184 n = m_get(how, MT_DATA);
188 * Align the remainder of the mbuf chain.
193 m_copyback(n, off, m->m_len, mtod(m, caddr_t));
207 #endif /* __NO_STRICT_ALIGNMENT */
211 nfsrv_object_create(struct vnode *vp, struct thread *td)
214 if (vp == NULL || vp->v_type != VREG)
216 (void) vfs_object_create(vp, td, td->td_ucred);
221 * Look up a file name. Basically just initialize stuff and call namei().
224 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
228 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
232 NDFREE(ndp, NDF_ONLY_PNBUF);
238 * Copy NFS uid, gids to the cred structure.
241 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
244 KASSERT(nfscr->nfsc_ngroups >= 0,
245 ("newnfs_copycred: negative nfsc_ngroups"));
246 cr->cr_uid = nfscr->nfsc_uid;
247 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
251 * Map args from nfsmsleep() to msleep().
254 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
261 timeo = hz * ts->tv_sec;
262 nsecval = (u_int64_t)ts->tv_nsec;
263 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
265 timeo += (int)nsecval;
269 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
274 * Get the file system info for the server. For now, just assume FFS.
277 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
283 * There should be file system VFS OP(s) to get this information.
284 * For now, assume ufs.
287 pref = NFS_MAXDGRAMDATA;
290 sip->fs_rtmax = NFS_SRVMAXIO;
291 sip->fs_rtpref = pref;
292 sip->fs_rtmult = NFS_FABLKSIZE;
293 sip->fs_wtmax = NFS_SRVMAXIO;
294 sip->fs_wtpref = pref;
295 sip->fs_wtmult = NFS_FABLKSIZE;
296 sip->fs_dtpref = pref;
297 sip->fs_maxfilesize = 0xffffffffffffffffull;
298 sip->fs_timedelta.tv_sec = 0;
299 sip->fs_timedelta.tv_nsec = 1;
300 sip->fs_properties = (NFSV3FSINFO_LINK |
301 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
302 NFSV3FSINFO_CANSETTIME);
306 * Do the pathconf vnode op.
309 nfsvno_pathconf(struct vnode *vp, int flag, register_t *retf,
310 struct ucred *cred, struct thread *p)
314 error = VOP_PATHCONF(vp, flag, retf);
315 if (error == EOPNOTSUPP || error == EINVAL) {
317 * Some file systems return EINVAL for name arguments not
318 * supported and some return EOPNOTSUPP for this case.
319 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
329 case _PC_CHOWN_RESTRICTED:
337 * Only happens if a _PC_xxx is added to the server,
338 * but this isn't updated.
341 printf("nfsrvd pathconf flag=%d not supp\n", flag);
349 /* Fake nfsrv_atroot. Just return 0 */
351 nfsrv_atroot(struct vnode *vp, long *retp)
358 * Set the credentials to refer to root.
359 * If only the various BSDen could agree on whether cr_gid is a separate
360 * field or cr_groups[0]...
363 newnfs_setroot(struct ucred *cred)
367 cred->cr_groups[0] = 0;
368 cred->cr_ngroups = 1;
372 * Get the client credential. Used for Renew and recovery.
378 struct thread *td = curthread;
380 cred = crdup(td->td_ucred);
381 newnfs_setroot(cred);
387 * Call the nfsd's timer function once/sec.
390 newnfs_timer(void *arg)
392 static time_t lasttime = 0;
394 * Call the server timer, if set up.
395 * The argument indicates if it is the next second and therefore
396 * leases should be checked.
398 if (lasttime != NFSD_MONOSEC) {
399 lasttime = NFSD_MONOSEC;
400 if (nfsd_call_servertimer != NULL)
401 (*nfsd_call_servertimer)();
403 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
408 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
409 * the sleep should be for 5 seconds.
410 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
411 * an event that never gets a wakeup. Only return EINTR or 0.
414 nfs_catnap(int prio, int errval, const char *wmesg)
416 static int non_event;
419 if (errval == NFSERR_GRACE)
420 ret = tsleep(&non_event, prio, wmesg, 5 * hz);
422 ret = tsleep(&non_event, prio, wmesg, 1);
429 * Get referral. For now, just fail.
432 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
439 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
443 error = nfssvc_call(td, uap, td->td_ucred);
449 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
452 struct nfsd_idargs nid;
453 struct nfsd_oidargs onid;
455 if (uap->flag & NFSSVC_IDNAME) {
456 if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
457 error = copyin(uap->argp, &nid, sizeof(nid));
459 error = copyin(uap->argp, &onid, sizeof(onid));
461 nid.nid_flag = onid.nid_flag;
462 nid.nid_uid = onid.nid_uid;
463 nid.nid_gid = onid.nid_gid;
464 nid.nid_usermax = onid.nid_usermax;
465 nid.nid_usertimeout = onid.nid_usertimeout;
466 nid.nid_name = onid.nid_name;
467 nid.nid_namelen = onid.nid_namelen;
474 error = nfssvc_idname(&nid);
476 } else if (uap->flag & NFSSVC_GETSTATS) {
477 error = copyout(&newnfsstats,
478 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
480 if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
481 newnfsstats.attrcache_hits = 0;
482 newnfsstats.attrcache_misses = 0;
483 newnfsstats.lookupcache_hits = 0;
484 newnfsstats.lookupcache_misses = 0;
485 newnfsstats.direofcache_hits = 0;
486 newnfsstats.direofcache_misses = 0;
487 newnfsstats.accesscache_hits = 0;
488 newnfsstats.accesscache_misses = 0;
489 newnfsstats.biocache_reads = 0;
490 newnfsstats.read_bios = 0;
491 newnfsstats.read_physios = 0;
492 newnfsstats.biocache_writes = 0;
493 newnfsstats.write_bios = 0;
494 newnfsstats.write_physios = 0;
495 newnfsstats.biocache_readlinks = 0;
496 newnfsstats.readlink_bios = 0;
497 newnfsstats.biocache_readdirs = 0;
498 newnfsstats.readdir_bios = 0;
499 newnfsstats.rpcretries = 0;
500 newnfsstats.rpcrequests = 0;
501 newnfsstats.rpctimeouts = 0;
502 newnfsstats.rpcunexpected = 0;
503 newnfsstats.rpcinvalid = 0;
504 bzero(newnfsstats.rpccnt,
505 sizeof(newnfsstats.rpccnt));
507 if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
508 newnfsstats.srvrpc_errs = 0;
509 newnfsstats.srv_errs = 0;
510 newnfsstats.srvcache_inproghits = 0;
511 newnfsstats.srvcache_idemdonehits = 0;
512 newnfsstats.srvcache_nonidemdonehits = 0;
513 newnfsstats.srvcache_misses = 0;
514 newnfsstats.srvcache_tcppeak = 0;
515 newnfsstats.srvclients = 0;
516 newnfsstats.srvopenowners = 0;
517 newnfsstats.srvopens = 0;
518 newnfsstats.srvlockowners = 0;
519 newnfsstats.srvlocks = 0;
520 newnfsstats.srvdelegates = 0;
521 newnfsstats.clopenowners = 0;
522 newnfsstats.clopens = 0;
523 newnfsstats.cllockowners = 0;
524 newnfsstats.cllocks = 0;
525 newnfsstats.cldelegates = 0;
526 newnfsstats.cllocalopenowners = 0;
527 newnfsstats.cllocalopens = 0;
528 newnfsstats.cllocallockowners = 0;
529 newnfsstats.cllocallocks = 0;
530 bzero(newnfsstats.srvrpccnt,
531 sizeof(newnfsstats.srvrpccnt));
532 bzero(newnfsstats.cbrpccnt,
533 sizeof(newnfsstats.cbrpccnt));
537 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
539 struct nfsuserd_args nargs;
541 if ((uap->flag & NFSSVC_NEWSTRUCT) == 0) {
542 error = copyin(uap->argp, (caddr_t)&sockport,
545 nargs.nuserd_family = AF_INET;
546 nargs.nuserd_port = sockport;
550 * New nfsuserd_args structure, which indicates
551 * which IP version to use along with the port#.
553 error = copyin(uap->argp, &nargs, sizeof(nargs));
556 error = nfsrv_nfsuserdport(&nargs, p);
557 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
558 nfsrv_nfsuserddelport();
568 * called by all three modevent routines, so that it gets things
569 * initialized soon enough.
572 newnfs_portinit(void)
574 static int inited = 0;
579 /* Initialize SMP locks used by both client and server. */
580 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
581 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
582 mtx_init(&nfs_clstate_mutex, "nfs_clstate_mutex", NULL, MTX_DEF);
586 * Determine if the file system supports NFSv4 ACLs.
587 * Return 1 if it does, 0 otherwise.
590 nfs_supportsnfsv4acls(struct vnode *vp)
595 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
597 if (nfsrv_useacl == 0)
599 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
600 if (error == 0 && retval != 0)
605 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
608 * Called once to initialize data structures...
611 nfscommon_modevent(module_t mod, int type, void *data)
614 static int loaded = 0;
621 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
622 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
623 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
624 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
625 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
627 callout_init(&newnfsd_callout, 1);
629 nfsd_call_nfscommon = nfssvc_nfscommon;
634 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
635 nfs_numnfscbd != 0) {
640 nfsd_call_nfscommon = NULL;
641 callout_drain(&newnfsd_callout);
642 /* Clean out the name<-->id cache. */
643 nfsrv_cleanusergroup();
644 /* and get rid of the mutexes */
645 mtx_destroy(&nfs_nameid_mutex);
646 mtx_destroy(&newnfsd_mtx);
647 mtx_destroy(&nfs_state_mutex);
648 mtx_destroy(&nfs_clstate_mutex);
649 mtx_destroy(&nfs_sockl_mutex);
650 mtx_destroy(&nfs_slock_mutex);
651 mtx_destroy(&nfs_req_mutex);
652 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
664 static moduledata_t nfscommon_mod = {
669 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
671 /* So that loader and kldload(2) can find us, wherever we are.. */
672 MODULE_VERSION(nfscommon, 1);
673 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
674 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);