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;
133 /* local functions */
134 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
136 #ifdef __NO_STRICT_ALIGNMENT
138 * These architectures don't need re-alignment, so just return.
141 newnfs_realign(struct mbuf **pm, int how)
146 #else /* !__NO_STRICT_ALIGNMENT */
150 * Check for badly aligned mbuf data and realign by copying the unaligned
151 * portion of the data into a new mbuf chain and freeing the portions
152 * of the old chain that were replaced.
154 * We cannot simply realign the data within the existing mbuf chain
155 * because the underlying buffers may contain other rpc commands and
156 * we cannot afford to overwrite them.
158 * We would prefer to avoid this situation entirely. The situation does
159 * not occur with NFS/UDP and is supposed to only occassionally occur
160 * with TCP. Use vfs.nfs.realign_count and realign_test to check this.
164 newnfs_realign(struct mbuf **pm, int how)
170 while ((m = *pm) != NULL) {
171 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
173 * NB: we can't depend on m_pkthdr.len to help us
174 * decide what to do here. May not be worth doing
175 * the m_length calculation as m_copyback will
176 * expand the mbuf chain below as needed.
178 space = m_length(m, NULL);
179 if (space >= MINCLSIZE) {
180 /* NB: m_copyback handles space > MCLBYTES */
181 n = m_getcl(how, MT_DATA, 0);
183 n = m_get(how, MT_DATA);
187 * Align the remainder of the mbuf chain.
192 m_copyback(n, off, m->m_len, mtod(m, caddr_t));
206 #endif /* __NO_STRICT_ALIGNMENT */
210 nfsrv_object_create(struct vnode *vp, struct thread *td)
213 if (vp == NULL || vp->v_type != VREG)
215 (void) vfs_object_create(vp, td, td->td_ucred);
220 * Look up a file name. Basically just initialize stuff and call namei().
223 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
227 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname,
231 NDFREE(ndp, NDF_ONLY_PNBUF);
237 * Copy NFS uid, gids to the cred structure.
240 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
243 KASSERT(nfscr->nfsc_ngroups >= 0,
244 ("newnfs_copycred: negative nfsc_ngroups"));
245 cr->cr_uid = nfscr->nfsc_uid;
246 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
250 * Map args from nfsmsleep() to msleep().
253 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
260 timeo = hz * ts->tv_sec;
261 nsecval = (u_int64_t)ts->tv_nsec;
262 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
264 timeo += (int)nsecval;
268 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
273 * Get the file system info for the server. For now, just assume FFS.
276 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
282 * There should be file system VFS OP(s) to get this information.
283 * For now, assume ufs.
286 pref = NFS_MAXDGRAMDATA;
289 sip->fs_rtmax = NFS_SRVMAXIO;
290 sip->fs_rtpref = pref;
291 sip->fs_rtmult = NFS_FABLKSIZE;
292 sip->fs_wtmax = NFS_SRVMAXIO;
293 sip->fs_wtpref = pref;
294 sip->fs_wtmult = NFS_FABLKSIZE;
295 sip->fs_dtpref = pref;
296 sip->fs_maxfilesize = 0xffffffffffffffffull;
297 sip->fs_timedelta.tv_sec = 0;
298 sip->fs_timedelta.tv_nsec = 1;
299 sip->fs_properties = (NFSV3FSINFO_LINK |
300 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
301 NFSV3FSINFO_CANSETTIME);
305 * Do the pathconf vnode op.
308 nfsvno_pathconf(struct vnode *vp, int flag, register_t *retf,
309 struct ucred *cred, struct thread *p)
313 error = VOP_PATHCONF(vp, flag, retf);
314 if (error == EOPNOTSUPP || error == EINVAL) {
316 * Some file systems return EINVAL for name arguments not
317 * supported and some return EOPNOTSUPP for this case.
318 * So the NFSv3 Pathconf RPC doesn't fail for these cases,
328 case _PC_CHOWN_RESTRICTED:
336 * Only happens if a _PC_xxx is added to the server,
337 * but this isn't updated.
340 printf("nfsrvd pathconf flag=%d not supp\n", flag);
348 /* Fake nfsrv_atroot. Just return 0 */
350 nfsrv_atroot(struct vnode *vp, long *retp)
357 * Set the credentials to refer to root.
358 * If only the various BSDen could agree on whether cr_gid is a separate
359 * field or cr_groups[0]...
362 newnfs_setroot(struct ucred *cred)
366 cred->cr_groups[0] = 0;
367 cred->cr_ngroups = 1;
371 * Get the client credential. Used for Renew and recovery.
377 struct thread *td = curthread;
379 cred = crdup(td->td_ucred);
380 newnfs_setroot(cred);
386 * Call the nfsd's timer function once/sec.
389 newnfs_timer(void *arg)
391 static time_t lasttime = 0;
393 * Call the server timer, if set up.
394 * The argument indicates if it is the next second and therefore
395 * leases should be checked.
397 if (lasttime != NFSD_MONOSEC) {
398 lasttime = NFSD_MONOSEC;
399 if (nfsd_call_servertimer != NULL)
400 (*nfsd_call_servertimer)();
402 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
407 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
408 * the sleep should be for 5 seconds.
409 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
410 * an event that never gets a wakeup. Only return EINTR or 0.
413 nfs_catnap(int prio, int errval, const char *wmesg)
415 static int non_event;
418 if (errval == NFSERR_GRACE)
419 ret = tsleep(&non_event, prio, wmesg, 5 * hz);
421 ret = tsleep(&non_event, prio, wmesg, 1);
428 * Get referral. For now, just fail.
431 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
438 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
442 error = nfssvc_call(td, uap, td->td_ucred);
448 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
451 struct nfsd_idargs nid;
452 struct nfsd_oidargs onid;
454 if (uap->flag & NFSSVC_IDNAME) {
455 if ((uap->flag & NFSSVC_NEWSTRUCT) != 0)
456 error = copyin(uap->argp, &nid, sizeof(nid));
458 error = copyin(uap->argp, &onid, sizeof(onid));
460 nid.nid_flag = onid.nid_flag;
461 nid.nid_uid = onid.nid_uid;
462 nid.nid_gid = onid.nid_gid;
463 nid.nid_usermax = onid.nid_usermax;
464 nid.nid_usertimeout = onid.nid_usertimeout;
465 nid.nid_name = onid.nid_name;
466 nid.nid_namelen = onid.nid_namelen;
473 error = nfssvc_idname(&nid);
475 } else if (uap->flag & NFSSVC_GETSTATS) {
476 error = copyout(&newnfsstats,
477 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
479 if ((uap->flag & NFSSVC_ZEROCLTSTATS) != 0) {
480 newnfsstats.attrcache_hits = 0;
481 newnfsstats.attrcache_misses = 0;
482 newnfsstats.lookupcache_hits = 0;
483 newnfsstats.lookupcache_misses = 0;
484 newnfsstats.direofcache_hits = 0;
485 newnfsstats.direofcache_misses = 0;
486 newnfsstats.accesscache_hits = 0;
487 newnfsstats.accesscache_misses = 0;
488 newnfsstats.biocache_reads = 0;
489 newnfsstats.read_bios = 0;
490 newnfsstats.read_physios = 0;
491 newnfsstats.biocache_writes = 0;
492 newnfsstats.write_bios = 0;
493 newnfsstats.write_physios = 0;
494 newnfsstats.biocache_readlinks = 0;
495 newnfsstats.readlink_bios = 0;
496 newnfsstats.biocache_readdirs = 0;
497 newnfsstats.readdir_bios = 0;
498 newnfsstats.rpcretries = 0;
499 newnfsstats.rpcrequests = 0;
500 newnfsstats.rpctimeouts = 0;
501 newnfsstats.rpcunexpected = 0;
502 newnfsstats.rpcinvalid = 0;
503 bzero(newnfsstats.rpccnt,
504 sizeof(newnfsstats.rpccnt));
506 if ((uap->flag & NFSSVC_ZEROSRVSTATS) != 0) {
507 newnfsstats.srvrpc_errs = 0;
508 newnfsstats.srv_errs = 0;
509 newnfsstats.srvcache_inproghits = 0;
510 newnfsstats.srvcache_idemdonehits = 0;
511 newnfsstats.srvcache_nonidemdonehits = 0;
512 newnfsstats.srvcache_misses = 0;
513 newnfsstats.srvcache_tcppeak = 0;
514 newnfsstats.srvclients = 0;
515 newnfsstats.srvopenowners = 0;
516 newnfsstats.srvopens = 0;
517 newnfsstats.srvlockowners = 0;
518 newnfsstats.srvlocks = 0;
519 newnfsstats.srvdelegates = 0;
520 newnfsstats.clopenowners = 0;
521 newnfsstats.clopens = 0;
522 newnfsstats.cllockowners = 0;
523 newnfsstats.cllocks = 0;
524 newnfsstats.cldelegates = 0;
525 newnfsstats.cllocalopenowners = 0;
526 newnfsstats.cllocalopens = 0;
527 newnfsstats.cllocallockowners = 0;
528 newnfsstats.cllocallocks = 0;
529 bzero(newnfsstats.srvrpccnt,
530 sizeof(newnfsstats.srvrpccnt));
531 bzero(newnfsstats.cbrpccnt,
532 sizeof(newnfsstats.cbrpccnt));
536 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
539 error = copyin(uap->argp, (caddr_t)&sockport,
542 error = nfsrv_nfsuserdport(sockport, p);
543 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
544 nfsrv_nfsuserddelport();
554 * called by all three modevent routines, so that it gets things
555 * initialized soon enough.
558 newnfs_portinit(void)
560 static int inited = 0;
565 /* Initialize SMP locks used by both client and server. */
566 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
567 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
571 * Determine if the file system supports NFSv4 ACLs.
572 * Return 1 if it does, 0 otherwise.
575 nfs_supportsnfsv4acls(struct vnode *vp)
580 ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
582 if (nfsrv_useacl == 0)
584 error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
585 if (error == 0 && retval != 0)
590 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
593 * Called once to initialize data structures...
596 nfscommon_modevent(module_t mod, int type, void *data)
599 static int loaded = 0;
606 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
607 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
608 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
609 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
610 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
612 callout_init(&newnfsd_callout, CALLOUT_MPSAFE);
614 nfsd_call_nfscommon = nfssvc_nfscommon;
619 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
620 nfs_numnfscbd != 0) {
625 nfsd_call_nfscommon = NULL;
626 callout_drain(&newnfsd_callout);
627 /* and get rid of the mutexes */
628 mtx_destroy(&nfs_nameid_mutex);
629 mtx_destroy(&newnfsd_mtx);
630 mtx_destroy(&nfs_state_mutex);
631 mtx_destroy(&nfs_sockl_mutex);
632 mtx_destroy(&nfs_slock_mutex);
633 mtx_destroy(&nfs_req_mutex);
634 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
646 static moduledata_t nfscommon_mod = {
651 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
653 /* So that loader and kldload(2) can find us, wherever we are.. */
654 MODULE_VERSION(nfscommon, 1);
655 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
656 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);