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, "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 NFS client");
83 SYSCTL_INT(_vfs_nfs, OID_AUTO, userhashsize, CTLFLAG_RDTUN, &nfsrv_lughashsize,
84 0, "Size of hash tables for uid/name mapping");
88 * (Here for FreeBSD, since they allocate storage.)
90 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
91 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
92 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
93 "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
94 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
95 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
96 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
97 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
98 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
99 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
100 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
101 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
102 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
103 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
104 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
105 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
106 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "NFS vnode");
107 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "NFS Direct IO buffer");
108 MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
109 "NFS directory offset data");
110 MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
111 "NFS local lock rollback");
112 MALLOC_DEFINE(M_NEWNFSLAYOUT, "NFSCL layout", "NFSv4.1 Layout");
113 MALLOC_DEFINE(M_NEWNFSFLAYOUT, "NFSCL flayout", "NFSv4.1 File Layout");
114 MALLOC_DEFINE(M_NEWNFSDEVINFO, "NFSCL devinfo", "NFSv4.1 Device Info");
115 MALLOC_DEFINE(M_NEWNFSSOCKREQ, "NFSCL sockreq", "NFS Sock Req");
116 MALLOC_DEFINE(M_NEWNFSCLDS, "NFSCL session", "NFSv4.1 Session");
117 MALLOC_DEFINE(M_NEWNFSLAYRECALL, "NFSCL layrecall", "NFSv4.1 Layout Recall");
118 MALLOC_DEFINE(M_NEWNFSDSESSION, "NFSD session", "NFSD Session for a client");
121 * Definition of mutex locks.
122 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
123 * and assorted other nfsd structures.
125 struct mtx newnfsd_mtx;
126 struct mtx nfs_sockl_mutex;
127 struct mtx nfs_state_mutex;
128 struct mtx nfs_nameid_mutex;
129 struct mtx nfs_req_mutex;
130 struct mtx nfs_slock_mutex;
132 /* local functions */
133 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
135 #ifdef __NO_STRICT_ALIGNMENT
137 * These architectures don't need re-alignment, so just return.
140 newnfs_realign(struct mbuf **pm, int how)
145 #else /* !__NO_STRICT_ALIGNMENT */
149 * Check for badly aligned mbuf data and realign by copying the unaligned
150 * portion of the data into a new mbuf chain and freeing the portions
151 * of the old chain that were replaced.
153 * We cannot simply realign the data within the existing mbuf chain
154 * because the underlying buffers may contain other rpc commands and
155 * we cannot afford to overwrite them.
157 * We would prefer to avoid this situation entirely. The situation does
158 * not occur with NFS/UDP and is supposed to only occasionally occur
159 * with TCP. Use vfs.nfs.realign_count and realign_test to check this.
163 newnfs_realign(struct mbuf **pm, int how)
169 while ((m = *pm) != NULL) {
170 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
172 * NB: we can't depend on m_pkthdr.len to help us
173 * decide what to do here. May not be worth doing
174 * the m_length calculation as m_copyback will
175 * expand the mbuf chain below as needed.
177 space = m_length(m, NULL);
178 if (space >= MINCLSIZE) {
179 /* NB: m_copyback handles space > MCLBYTES */
180 n = m_getcl(how, MT_DATA, 0);
182 n = m_get(how, MT_DATA);
186 * Align the remainder of the mbuf chain.
191 m_copyback(n, off, m->m_len, mtod(m, caddr_t));
205 #endif /* __NO_STRICT_ALIGNMENT */
209 nfsrv_object_create(struct vnode *vp, struct thread *td)
212 if (vp == NULL || vp->v_type != VREG)
214 (void) vfs_object_create(vp, td, td->td_ucred);
219 * Look up a file name. Basically just initialize stuff and call namei().
222 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
226 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
230 NDFREE(ndp, NDF_ONLY_PNBUF);
236 * Copy NFS uid, gids to the cred structure.
239 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
242 KASSERT(nfscr->nfsc_ngroups >= 0,
243 ("newnfs_copycred: negative nfsc_ngroups"));
244 cr->cr_uid = nfscr->nfsc_uid;
245 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
249 * Map args from nfsmsleep() to msleep().
252 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
259 timeo = hz * ts->tv_sec;
260 nsecval = (u_int64_t)ts->tv_nsec;
261 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
263 timeo += (int)nsecval;
267 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
272 * Get the file system info for the server. For now, just assume FFS.
275 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
281 * There should be file system VFS OP(s) to get this information.
282 * For now, assume ufs.
285 pref = NFS_MAXDGRAMDATA;
288 sip->fs_rtmax = NFS_SRVMAXIO;
289 sip->fs_rtpref = pref;
290 sip->fs_rtmult = NFS_FABLKSIZE;
291 sip->fs_wtmax = NFS_SRVMAXIO;
292 sip->fs_wtpref = pref;
293 sip->fs_wtmult = NFS_FABLKSIZE;
294 sip->fs_dtpref = pref;
295 sip->fs_maxfilesize = 0xffffffffffffffffull;
296 sip->fs_timedelta.tv_sec = 0;
297 sip->fs_timedelta.tv_nsec = 1;
298 sip->fs_properties = (NFSV3FSINFO_LINK |
299 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
300 NFSV3FSINFO_CANSETTIME);
304 * Do the pathconf vnode op.
307 nfsvno_pathconf(struct vnode *vp, int flag, register_t *retf,
308 struct ucred *cred, struct thread *p)
312 error = VOP_PATHCONF(vp, flag, retf);
313 if (error == EOPNOTSUPP || error == EINVAL) {
315 * Some file systems return EINVAL for name arguments not
316 * supported and some return EOPNOTSUPP for this case.
317 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
327 case _PC_CHOWN_RESTRICTED:
335 * Only happens if a _PC_xxx is added to the server,
336 * but this isn't updated.
339 printf("nfsrvd pathconf flag=%d not supp\n", flag);
347 /* Fake nfsrv_atroot. Just return 0 */
349 nfsrv_atroot(struct vnode *vp, long *retp)
356 * Set the credentials to refer to root.
357 * If only the various BSDen could agree on whether cr_gid is a separate
358 * field or cr_groups[0]...
361 newnfs_setroot(struct ucred *cred)
365 cred->cr_groups[0] = 0;
366 cred->cr_ngroups = 1;
370 * Get the client credential. Used for Renew and recovery.
376 struct thread *td = curthread;
378 cred = crdup(td->td_ucred);
379 newnfs_setroot(cred);
385 * Call the nfsd's timer function once/sec.
388 newnfs_timer(void *arg)
390 static time_t lasttime = 0;
392 * Call the server timer, if set up.
393 * The argument indicates if it is the next second and therefore
394 * leases should be checked.
396 if (lasttime != NFSD_MONOSEC) {
397 lasttime = NFSD_MONOSEC;
398 if (nfsd_call_servertimer != NULL)
399 (*nfsd_call_servertimer)();
401 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
406 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
407 * the sleep should be for 5 seconds.
408 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
409 * an event that never gets a wakeup. Only return EINTR or 0.
412 nfs_catnap(int prio, int errval, const char *wmesg)
414 static int non_event;
417 if (errval == NFSERR_GRACE)
418 ret = tsleep(&non_event, prio, wmesg, 5 * hz);
420 ret = tsleep(&non_event, prio, wmesg, 1);
427 * Get referral. For now, just fail.
430 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
437 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
441 error = nfssvc_call(td, uap, td->td_ucred);
447 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
450 struct nfsd_idargs nid;
451 struct nfsd_oidargs onid;
453 if (uap->flag & NFSSVC_IDNAME) {
454 if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
455 error = copyin(uap->argp, &nid, sizeof(nid));
457 error = copyin(uap->argp, &onid, sizeof(onid));
459 nid.nid_flag = onid.nid_flag;
460 nid.nid_uid = onid.nid_uid;
461 nid.nid_gid = onid.nid_gid;
462 nid.nid_usermax = onid.nid_usermax;
463 nid.nid_usertimeout = onid.nid_usertimeout;
464 nid.nid_name = onid.nid_name;
465 nid.nid_namelen = onid.nid_namelen;
472 error = nfssvc_idname(&nid);
474 } else if (uap->flag & NFSSVC_GETSTATS) {
475 error = copyout(&newnfsstats,
476 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
478 if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
479 newnfsstats.attrcache_hits = 0;
480 newnfsstats.attrcache_misses = 0;
481 newnfsstats.lookupcache_hits = 0;
482 newnfsstats.lookupcache_misses = 0;
483 newnfsstats.direofcache_hits = 0;
484 newnfsstats.direofcache_misses = 0;
485 newnfsstats.accesscache_hits = 0;
486 newnfsstats.accesscache_misses = 0;
487 newnfsstats.biocache_reads = 0;
488 newnfsstats.read_bios = 0;
489 newnfsstats.read_physios = 0;
490 newnfsstats.biocache_writes = 0;
491 newnfsstats.write_bios = 0;
492 newnfsstats.write_physios = 0;
493 newnfsstats.biocache_readlinks = 0;
494 newnfsstats.readlink_bios = 0;
495 newnfsstats.biocache_readdirs = 0;
496 newnfsstats.readdir_bios = 0;
497 newnfsstats.rpcretries = 0;
498 newnfsstats.rpcrequests = 0;
499 newnfsstats.rpctimeouts = 0;
500 newnfsstats.rpcunexpected = 0;
501 newnfsstats.rpcinvalid = 0;
502 bzero(newnfsstats.rpccnt,
503 sizeof(newnfsstats.rpccnt));
505 if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
506 newnfsstats.srvrpc_errs = 0;
507 newnfsstats.srv_errs = 0;
508 newnfsstats.srvcache_inproghits = 0;
509 newnfsstats.srvcache_idemdonehits = 0;
510 newnfsstats.srvcache_nonidemdonehits = 0;
511 newnfsstats.srvcache_misses = 0;
512 newnfsstats.srvcache_tcppeak = 0;
513 newnfsstats.srvclients = 0;
514 newnfsstats.srvopenowners = 0;
515 newnfsstats.srvopens = 0;
516 newnfsstats.srvlockowners = 0;
517 newnfsstats.srvlocks = 0;
518 newnfsstats.srvdelegates = 0;
519 newnfsstats.clopenowners = 0;
520 newnfsstats.clopens = 0;
521 newnfsstats.cllockowners = 0;
522 newnfsstats.cllocks = 0;
523 newnfsstats.cldelegates = 0;
524 newnfsstats.cllocalopenowners = 0;
525 newnfsstats.cllocalopens = 0;
526 newnfsstats.cllocallockowners = 0;
527 newnfsstats.cllocallocks = 0;
528 bzero(newnfsstats.srvrpccnt,
529 sizeof(newnfsstats.srvrpccnt));
530 bzero(newnfsstats.cbrpccnt,
531 sizeof(newnfsstats.cbrpccnt));
535 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
538 error = copyin(uap->argp, (caddr_t)&sockport,
541 error = nfsrv_nfsuserdport(sockport, p);
542 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
543 nfsrv_nfsuserddelport();
553 * called by all three modevent routines, so that it gets things
554 * initialized soon enough.
557 newnfs_portinit(void)
559 static int inited = 0;
564 /* Initialize SMP locks used by both client and server. */
565 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
566 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
570 * Determine if the file system supports NFSv4 ACLs.
571 * Return 1 if it does, 0 otherwise.
574 nfs_supportsnfsv4acls(struct vnode *vp)
579 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
581 if (nfsrv_useacl == 0)
583 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
584 if (error == 0 && retval != 0)
589 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
592 * Called once to initialize data structures...
595 nfscommon_modevent(module_t mod, int type, void *data)
598 static int loaded = 0;
605 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
606 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
607 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
608 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
609 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
611 callout_init(&newnfsd_callout, 1);
613 nfsd_call_nfscommon = nfssvc_nfscommon;
618 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
619 nfs_numnfscbd != 0) {
624 nfsd_call_nfscommon = NULL;
625 callout_drain(&newnfsd_callout);
626 /* Clean out the name<-->id cache. */
627 nfsrv_cleanusergroup();
628 /* and get rid of the mutexes */
629 mtx_destroy(&nfs_nameid_mutex);
630 mtx_destroy(&newnfsd_mtx);
631 mtx_destroy(&nfs_state_mutex);
632 mtx_destroy(&nfs_sockl_mutex);
633 mtx_destroy(&nfs_slock_mutex);
634 mtx_destroy(&nfs_req_mutex);
635 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
647 static moduledata_t nfscommon_mod = {
652 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
654 /* So that loader and kldload(2) can find us, wherever we are.. */
655 MODULE_VERSION(nfscommon, 1);
656 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
657 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);