Deleted Added
full compact
nfs_commonport.c (205572) nfs_commonport.c (205941)
1/*-
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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.
19 *
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
30 * SUCH DAMAGE.
31 *
32 */
33
34#include <sys/cdefs.h>
1/*-
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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.
19 *
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
30 * SUCH DAMAGE.
31 *
32 */
33
34#include <sys/cdefs.h>
35__FBSDID("$FreeBSD: head/sys/fs/nfs/nfs_commonport.c 205572 2010-03-24 02:02:02Z rmacklem $");
35__FBSDID("$FreeBSD: head/sys/fs/nfs/nfs_commonport.c 205941 2010-03-30 23:11:50Z rmacklem $");
36
37/*
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.
41 */
42#include <fs/nfs/nfsport.h>
43#include <sys/sysctl.h>
44#include <vm/vm.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>
51#include <vm/uma.h>
52#include <vm/uma_int.h>
53
54extern int nfscl_ticks;
55extern int nfsrv_nfsuserd;
56extern struct nfssockreq nfsrv_nfsuserdsock;
57extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
58 struct thread *);
59extern int nfsrv_useacl;
60struct mount nfsv4root_mnt;
61int newnfs_numnfsd = 0;
62struct nfsstats newnfsstats;
63int nfs_numnfscbd = 0;
64char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
65struct callout newnfsd_callout;
66void (*nfsd_call_servertimer)(void) = NULL;
67void (*ncl_call_invalcaches)(struct vnode *) = NULL;
68
69static int nfs_realign_test;
70static int nfs_realign_count;
71
72SYSCTL_NODE(_vfs, OID_AUTO, newnfs, CTLFLAG_RW, 0, "New NFS filesystem");
73SYSCTL_INT(_vfs_newnfs, OID_AUTO, newnfs_realign_test, CTLFLAG_RW, &nfs_realign_test, 0, "");
74SYSCTL_INT(_vfs_newnfs, OID_AUTO, newnfs_realign_count, CTLFLAG_RW, &nfs_realign_count, 0, "");
75SYSCTL_INT(_vfs_newnfs, OID_AUTO, nfs4acl_enable, CTLFLAG_RW, &nfsrv_useacl, 0, "");
76SYSCTL_STRING(_vfs_newnfs, OID_AUTO, callback_addr, CTLFLAG_RW,
77 nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr), "");
78
79/*
80 * Defines for malloc
81 * (Here for FreeBSD, since they allocate storage.)
82 */
83MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
84MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
36
37/*
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.
41 */
42#include <fs/nfs/nfsport.h>
43#include <sys/sysctl.h>
44#include <vm/vm.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>
51#include <vm/uma.h>
52#include <vm/uma_int.h>
53
54extern int nfscl_ticks;
55extern int nfsrv_nfsuserd;
56extern struct nfssockreq nfsrv_nfsuserdsock;
57extern void (*nfsd_call_recall)(struct vnode *, int, struct ucred *,
58 struct thread *);
59extern int nfsrv_useacl;
60struct mount nfsv4root_mnt;
61int newnfs_numnfsd = 0;
62struct nfsstats newnfsstats;
63int nfs_numnfscbd = 0;
64char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
65struct callout newnfsd_callout;
66void (*nfsd_call_servertimer)(void) = NULL;
67void (*ncl_call_invalcaches)(struct vnode *) = NULL;
68
69static int nfs_realign_test;
70static int nfs_realign_count;
71
72SYSCTL_NODE(_vfs, OID_AUTO, newnfs, CTLFLAG_RW, 0, "New NFS filesystem");
73SYSCTL_INT(_vfs_newnfs, OID_AUTO, newnfs_realign_test, CTLFLAG_RW, &nfs_realign_test, 0, "");
74SYSCTL_INT(_vfs_newnfs, OID_AUTO, newnfs_realign_count, CTLFLAG_RW, &nfs_realign_count, 0, "");
75SYSCTL_INT(_vfs_newnfs, OID_AUTO, nfs4acl_enable, CTLFLAG_RW, &nfsrv_useacl, 0, "");
76SYSCTL_STRING(_vfs_newnfs, OID_AUTO, callback_addr, CTLFLAG_RW,
77 nfsv4_callbackaddr, sizeof(nfsv4_callbackaddr), "");
78
79/*
80 * Defines for malloc
81 * (Here for FreeBSD, since they allocate storage.)
82 */
83MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
84MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
85MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state", "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
85MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
86 "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
86MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
87MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
88MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
89MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
90MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
91MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
92MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
93MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
94MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
95MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
96MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
97MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
98MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "New nfs vnode");
99MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "New nfs Direct IO buffer");
87MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
88MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
89MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
90MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
91MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
92MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
93MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
94MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
95MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
96MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
97MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
98MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
99MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "New nfs vnode");
100MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "New nfs Direct IO buffer");
100MALLOC_DEFINE(M_NEWNFSDIROFF, "Newnfscl_diroff", "New NFS directory offset data");
101MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
102 "New NFS directory offset data");
103MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
104 "New NFS local lock rollback");
101
102/*
103 * Definition of mutex locks.
104 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
105 * and assorted other nfsd structures.
106 * Giant is used to protect the nfsd list and count, which is just
107 * updated when nfsd's start/stop and is grabbed for nfsrvd_dorpc()
108 * for the VFS ops.
109 */
110struct mtx newnfsd_mtx;
111struct mtx nfs_sockl_mutex;
112struct mtx nfs_state_mutex;
113struct mtx nfs_nameid_mutex;
114struct mtx nfs_req_mutex;
115struct mtx nfs_slock_mutex;
116
117/* local functions */
118static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
119
120#ifdef __NO_STRICT_ALIGNMENT
121/*
122 * These architectures don't need re-alignment, so just return.
123 */
124void
125newnfs_realign(struct mbuf **pm)
126{
127
128 return;
129}
130#else /* !__NO_STRICT_ALIGNMENT */
131/*
132 * newnfs_realign:
133 *
134 * Check for badly aligned mbuf data and realign by copying the unaligned
135 * portion of the data into a new mbuf chain and freeing the portions
136 * of the old chain that were replaced.
137 *
138 * We cannot simply realign the data within the existing mbuf chain
139 * because the underlying buffers may contain other rpc commands and
140 * we cannot afford to overwrite them.
141 *
142 * We would prefer to avoid this situation entirely. The situation does
143 * not occur with NFS/UDP and is supposed to only occassionally occur
144 * with TCP. Use vfs.nfs.realign_count and realign_test to check this.
145 *
146 */
147void
148newnfs_realign(struct mbuf **pm)
149{
150 struct mbuf *m, *n;
151 int off, space;
152
153 ++nfs_realign_test;
154 while ((m = *pm) != NULL) {
155 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
156 /*
157 * NB: we can't depend on m_pkthdr.len to help us
158 * decide what to do here. May not be worth doing
159 * the m_length calculation as m_copyback will
160 * expand the mbuf chain below as needed.
161 */
162 space = m_length(m, NULL);
163 if (space >= MINCLSIZE) {
164 /* NB: m_copyback handles space > MCLBYTES */
165 n = m_getcl(M_WAITOK, MT_DATA, 0);
166 } else
167 n = m_get(M_WAITOK, MT_DATA);
168 if (n == NULL)
169 return;
170 /*
171 * Align the remainder of the mbuf chain.
172 */
173 n->m_len = 0;
174 off = 0;
175 while (m != NULL) {
176 m_copyback(n, off, m->m_len, mtod(m, caddr_t));
177 off += m->m_len;
178 m = m->m_next;
179 }
180 m_freem(*pm);
181 *pm = n;
182 ++nfs_realign_count;
183 break;
184 }
185 pm = &m->m_next;
186 }
187}
188#endif /* __NO_STRICT_ALIGNMENT */
189
190#ifdef notdef
191static void
192nfsrv_object_create(struct vnode *vp, struct thread *td)
193{
194
195 if (vp == NULL || vp->v_type != VREG)
196 return;
197 (void) vfs_object_create(vp, td, td->td_ucred);
198}
199#endif
200
201/*
202 * Look up a file name. Basically just initialize stuff and call namei().
203 */
204int
205nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
206{
207 int error;
208
209 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname, p);
210 error = namei(ndp);
211 if (!error) {
212 NDFREE(ndp, NDF_ONLY_PNBUF);
213 }
214 return (error);
215}
216
217/*
218 * Copy NFS uid, gids to the cred structure.
219 */
220void
221newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
222{
223
224 cr->cr_uid = nfscr->nfsc_uid;
225 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
226}
227
228/*
229 * Map args from nfsmsleep() to msleep().
230 */
231int
232nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
233 struct timespec *ts)
234{
235 u_int64_t nsecval;
236 int error, timeo;
237
238 if (ts) {
239 timeo = hz * ts->tv_sec;
240 nsecval = (u_int64_t)ts->tv_nsec;
241 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
242 1000000000;
243 timeo += (int)nsecval;
244 } else {
245 timeo = 0;
246 }
247 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
248 return (error);
249}
250
251/*
252 * Get the file system info for the server. For now, just assume FFS.
253 */
254void
255nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
256{
257 int pref;
258
259 /*
260 * XXX
261 * There should be file system VFS OP(s) to get this information.
262 * For now, assume ufs.
263 */
264 if (isdgram)
265 pref = NFS_MAXDGRAMDATA;
266 else
267 pref = NFS_MAXDATA;
268 sip->fs_rtmax = NFS_MAXDATA;
269 sip->fs_rtpref = pref;
270 sip->fs_rtmult = NFS_FABLKSIZE;
271 sip->fs_wtmax = NFS_MAXDATA;
272 sip->fs_wtpref = pref;
273 sip->fs_wtmult = NFS_FABLKSIZE;
274 sip->fs_dtpref = pref;
275 sip->fs_maxfilesize = 0xffffffffffffffffull;
276 sip->fs_timedelta.tv_sec = 0;
277 sip->fs_timedelta.tv_nsec = 1;
278 sip->fs_properties = (NFSV3FSINFO_LINK |
279 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
280 NFSV3FSINFO_CANSETTIME);
281}
282
283/* Fake nfsrv_atroot. Just return 0 */
284int
285nfsrv_atroot(struct vnode *vp, long *retp)
286{
287
288 return (0);
289}
290
291/*
292 * Set the credentials to refer to root.
293 * If only the various BSDen could agree on whether cr_gid is a separate
294 * field or cr_groups[0]...
295 */
296void
297newnfs_setroot(struct ucred *cred)
298{
299
300 cred->cr_uid = 0;
301 cred->cr_groups[0] = 0;
302 cred->cr_ngroups = 1;
303}
304
305/*
306 * Get the client credential. Used for Renew and recovery.
307 */
308struct ucred *
309newnfs_getcred(void)
310{
311 struct ucred *cred;
312 struct thread *td = curthread;
313
314 cred = crdup(td->td_ucred);
315 newnfs_setroot(cred);
316 return (cred);
317}
318
319/*
320 * Nfs timer routine
321 * Call the nfsd's timer function once/sec.
322 */
323void
324newnfs_timer(void *arg)
325{
326 static time_t lasttime = 0;
327 /*
328 * Call the server timer, if set up.
329 * The argument indicates if it is the next second and therefore
330 * leases should be checked.
331 */
332 if (lasttime != NFSD_MONOSEC) {
333 lasttime = NFSD_MONOSEC;
334 if (nfsd_call_servertimer != NULL)
335 (*nfsd_call_servertimer)();
336 }
337 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
338}
339
340
341/*
342 * sleep for a short period of time.
343 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
344 * an event that never gets a wakeup. Only return EINTR or 0.
345 */
346int
347nfs_catnap(int prio, const char *wmesg)
348{
349 static int non_event;
350 int ret;
351
352 ret = tsleep(&non_event, prio, wmesg, 1);
353 if (ret != EINTR)
354 ret = 0;
355 return (ret);
356}
357
358/*
359 * Get referral. For now, just fail.
360 */
361struct nfsreferral *
362nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
363{
364
365 return (NULL);
366}
367
368static int
369nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
370{
371 int error;
372
373 error = nfssvc_call(td, uap, td->td_ucred);
374 return (error);
375}
376
377static int
378nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
379{
380 int error = EINVAL;
381 struct nfsd_idargs nid;
382
383 if (uap->flag & NFSSVC_IDNAME) {
384 error = copyin(uap->argp, (caddr_t)&nid, sizeof (nid));
385 if (error)
386 return (error);
387 error = nfssvc_idname(&nid);
388 return (error);
389 } else if (uap->flag & NFSSVC_GETSTATS) {
390 error = copyout(&newnfsstats,
391 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
392 return (error);
393 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
394 u_short sockport;
395
396 error = copyin(uap->argp, (caddr_t)&sockport,
397 sizeof (u_short));
398 if (!error)
399 error = nfsrv_nfsuserdport(sockport, p);
400 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
401 nfsrv_nfsuserddelport();
402 error = 0;
403 }
404 return (error);
405}
406
407/*
408 * called by all three modevent routines, so that it gets things
409 * initialized soon enough.
410 */
411void
412newnfs_portinit(void)
413{
414 static int inited = 0;
415
416 if (inited)
417 return;
418 inited = 1;
419 /* Initialize SMP locks used by both client and server. */
420 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
421 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
422}
423
424/*
425 * Determine if the file system supports NFSv4 ACLs.
426 * Return 1 if it does, 0 otherwise.
427 */
428int
429nfs_supportsnfsv4acls(struct mount *mp)
430{
431
432 if (mp->mnt_stat.f_fstypename == NULL)
433 return (0);
434 if (strcmp(mp->mnt_stat.f_fstypename, "ufs") == 0) {
435 /* Not yet */
436 return (0);
437 } else if (strcmp(mp->mnt_stat.f_fstypename, "zfs") == 0) {
438 /* Always supports them */
439 return (1);
440 }
441 return (0);
442}
443
444extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
445
446/*
447 * Called once to initialize data structures...
448 */
449static int
450nfscommon_modevent(module_t mod, int type, void *data)
451{
452 int error = 0;
453 static int loaded = 0;
454
455 switch (type) {
456 case MOD_LOAD:
457 if (loaded)
458 return (0);
459 newnfs_portinit();
460 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
461 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
462 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
463 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
464 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
465 MTX_DEF);
466 callout_init(&newnfsd_callout, CALLOUT_MPSAFE);
467 newnfs_init();
468 nfsd_call_nfscommon = nfssvc_nfscommon;
469 loaded = 1;
470 break;
471
472 case MOD_UNLOAD:
473 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
474 nfs_numnfscbd != 0) {
475 error = EBUSY;
476 break;
477 }
478
479 nfsd_call_nfscommon = NULL;
480 callout_drain(&newnfsd_callout);
481 /* and get rid of the mutexes */
482 mtx_destroy(&nfs_nameid_mutex);
483 mtx_destroy(&newnfsd_mtx);
484 mtx_destroy(&nfs_state_mutex);
485 mtx_destroy(&nfs_sockl_mutex);
486 mtx_destroy(&nfs_slock_mutex);
487 mtx_destroy(&nfs_req_mutex);
488 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
489 loaded = 0;
490 break;
491 default:
492 error = EOPNOTSUPP;
493 break;
494 }
495 return error;
496}
497static moduledata_t nfscommon_mod = {
498 "nfscommon",
499 nfscommon_modevent,
500 NULL,
501};
502DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
503
504/* So that loader and kldload(2) can find us, wherever we are.. */
505MODULE_VERSION(nfscommon, 1);
506MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
507MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
508
105
106/*
107 * Definition of mutex locks.
108 * newnfsd_mtx is used in nfsrvd_nfsd() to protect the nfs socket list
109 * and assorted other nfsd structures.
110 * Giant is used to protect the nfsd list and count, which is just
111 * updated when nfsd's start/stop and is grabbed for nfsrvd_dorpc()
112 * for the VFS ops.
113 */
114struct mtx newnfsd_mtx;
115struct mtx nfs_sockl_mutex;
116struct mtx nfs_state_mutex;
117struct mtx nfs_nameid_mutex;
118struct mtx nfs_req_mutex;
119struct mtx nfs_slock_mutex;
120
121/* local functions */
122static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
123
124#ifdef __NO_STRICT_ALIGNMENT
125/*
126 * These architectures don't need re-alignment, so just return.
127 */
128void
129newnfs_realign(struct mbuf **pm)
130{
131
132 return;
133}
134#else /* !__NO_STRICT_ALIGNMENT */
135/*
136 * newnfs_realign:
137 *
138 * Check for badly aligned mbuf data and realign by copying the unaligned
139 * portion of the data into a new mbuf chain and freeing the portions
140 * of the old chain that were replaced.
141 *
142 * We cannot simply realign the data within the existing mbuf chain
143 * because the underlying buffers may contain other rpc commands and
144 * we cannot afford to overwrite them.
145 *
146 * We would prefer to avoid this situation entirely. The situation does
147 * not occur with NFS/UDP and is supposed to only occassionally occur
148 * with TCP. Use vfs.nfs.realign_count and realign_test to check this.
149 *
150 */
151void
152newnfs_realign(struct mbuf **pm)
153{
154 struct mbuf *m, *n;
155 int off, space;
156
157 ++nfs_realign_test;
158 while ((m = *pm) != NULL) {
159 if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
160 /*
161 * NB: we can't depend on m_pkthdr.len to help us
162 * decide what to do here. May not be worth doing
163 * the m_length calculation as m_copyback will
164 * expand the mbuf chain below as needed.
165 */
166 space = m_length(m, NULL);
167 if (space >= MINCLSIZE) {
168 /* NB: m_copyback handles space > MCLBYTES */
169 n = m_getcl(M_WAITOK, MT_DATA, 0);
170 } else
171 n = m_get(M_WAITOK, MT_DATA);
172 if (n == NULL)
173 return;
174 /*
175 * Align the remainder of the mbuf chain.
176 */
177 n->m_len = 0;
178 off = 0;
179 while (m != NULL) {
180 m_copyback(n, off, m->m_len, mtod(m, caddr_t));
181 off += m->m_len;
182 m = m->m_next;
183 }
184 m_freem(*pm);
185 *pm = n;
186 ++nfs_realign_count;
187 break;
188 }
189 pm = &m->m_next;
190 }
191}
192#endif /* __NO_STRICT_ALIGNMENT */
193
194#ifdef notdef
195static void
196nfsrv_object_create(struct vnode *vp, struct thread *td)
197{
198
199 if (vp == NULL || vp->v_type != VREG)
200 return;
201 (void) vfs_object_create(vp, td, td->td_ucred);
202}
203#endif
204
205/*
206 * Look up a file name. Basically just initialize stuff and call namei().
207 */
208int
209nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
210{
211 int error;
212
213 NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname, p);
214 error = namei(ndp);
215 if (!error) {
216 NDFREE(ndp, NDF_ONLY_PNBUF);
217 }
218 return (error);
219}
220
221/*
222 * Copy NFS uid, gids to the cred structure.
223 */
224void
225newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
226{
227
228 cr->cr_uid = nfscr->nfsc_uid;
229 crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
230}
231
232/*
233 * Map args from nfsmsleep() to msleep().
234 */
235int
236nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
237 struct timespec *ts)
238{
239 u_int64_t nsecval;
240 int error, timeo;
241
242 if (ts) {
243 timeo = hz * ts->tv_sec;
244 nsecval = (u_int64_t)ts->tv_nsec;
245 nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
246 1000000000;
247 timeo += (int)nsecval;
248 } else {
249 timeo = 0;
250 }
251 error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
252 return (error);
253}
254
255/*
256 * Get the file system info for the server. For now, just assume FFS.
257 */
258void
259nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
260{
261 int pref;
262
263 /*
264 * XXX
265 * There should be file system VFS OP(s) to get this information.
266 * For now, assume ufs.
267 */
268 if (isdgram)
269 pref = NFS_MAXDGRAMDATA;
270 else
271 pref = NFS_MAXDATA;
272 sip->fs_rtmax = NFS_MAXDATA;
273 sip->fs_rtpref = pref;
274 sip->fs_rtmult = NFS_FABLKSIZE;
275 sip->fs_wtmax = NFS_MAXDATA;
276 sip->fs_wtpref = pref;
277 sip->fs_wtmult = NFS_FABLKSIZE;
278 sip->fs_dtpref = pref;
279 sip->fs_maxfilesize = 0xffffffffffffffffull;
280 sip->fs_timedelta.tv_sec = 0;
281 sip->fs_timedelta.tv_nsec = 1;
282 sip->fs_properties = (NFSV3FSINFO_LINK |
283 NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
284 NFSV3FSINFO_CANSETTIME);
285}
286
287/* Fake nfsrv_atroot. Just return 0 */
288int
289nfsrv_atroot(struct vnode *vp, long *retp)
290{
291
292 return (0);
293}
294
295/*
296 * Set the credentials to refer to root.
297 * If only the various BSDen could agree on whether cr_gid is a separate
298 * field or cr_groups[0]...
299 */
300void
301newnfs_setroot(struct ucred *cred)
302{
303
304 cred->cr_uid = 0;
305 cred->cr_groups[0] = 0;
306 cred->cr_ngroups = 1;
307}
308
309/*
310 * Get the client credential. Used for Renew and recovery.
311 */
312struct ucred *
313newnfs_getcred(void)
314{
315 struct ucred *cred;
316 struct thread *td = curthread;
317
318 cred = crdup(td->td_ucred);
319 newnfs_setroot(cred);
320 return (cred);
321}
322
323/*
324 * Nfs timer routine
325 * Call the nfsd's timer function once/sec.
326 */
327void
328newnfs_timer(void *arg)
329{
330 static time_t lasttime = 0;
331 /*
332 * Call the server timer, if set up.
333 * The argument indicates if it is the next second and therefore
334 * leases should be checked.
335 */
336 if (lasttime != NFSD_MONOSEC) {
337 lasttime = NFSD_MONOSEC;
338 if (nfsd_call_servertimer != NULL)
339 (*nfsd_call_servertimer)();
340 }
341 callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
342}
343
344
345/*
346 * sleep for a short period of time.
347 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
348 * an event that never gets a wakeup. Only return EINTR or 0.
349 */
350int
351nfs_catnap(int prio, const char *wmesg)
352{
353 static int non_event;
354 int ret;
355
356 ret = tsleep(&non_event, prio, wmesg, 1);
357 if (ret != EINTR)
358 ret = 0;
359 return (ret);
360}
361
362/*
363 * Get referral. For now, just fail.
364 */
365struct nfsreferral *
366nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
367{
368
369 return (NULL);
370}
371
372static int
373nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
374{
375 int error;
376
377 error = nfssvc_call(td, uap, td->td_ucred);
378 return (error);
379}
380
381static int
382nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
383{
384 int error = EINVAL;
385 struct nfsd_idargs nid;
386
387 if (uap->flag & NFSSVC_IDNAME) {
388 error = copyin(uap->argp, (caddr_t)&nid, sizeof (nid));
389 if (error)
390 return (error);
391 error = nfssvc_idname(&nid);
392 return (error);
393 } else if (uap->flag & NFSSVC_GETSTATS) {
394 error = copyout(&newnfsstats,
395 CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
396 return (error);
397 } else if (uap->flag & NFSSVC_NFSUSERDPORT) {
398 u_short sockport;
399
400 error = copyin(uap->argp, (caddr_t)&sockport,
401 sizeof (u_short));
402 if (!error)
403 error = nfsrv_nfsuserdport(sockport, p);
404 } else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
405 nfsrv_nfsuserddelport();
406 error = 0;
407 }
408 return (error);
409}
410
411/*
412 * called by all three modevent routines, so that it gets things
413 * initialized soon enough.
414 */
415void
416newnfs_portinit(void)
417{
418 static int inited = 0;
419
420 if (inited)
421 return;
422 inited = 1;
423 /* Initialize SMP locks used by both client and server. */
424 mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
425 mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
426}
427
428/*
429 * Determine if the file system supports NFSv4 ACLs.
430 * Return 1 if it does, 0 otherwise.
431 */
432int
433nfs_supportsnfsv4acls(struct mount *mp)
434{
435
436 if (mp->mnt_stat.f_fstypename == NULL)
437 return (0);
438 if (strcmp(mp->mnt_stat.f_fstypename, "ufs") == 0) {
439 /* Not yet */
440 return (0);
441 } else if (strcmp(mp->mnt_stat.f_fstypename, "zfs") == 0) {
442 /* Always supports them */
443 return (1);
444 }
445 return (0);
446}
447
448extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
449
450/*
451 * Called once to initialize data structures...
452 */
453static int
454nfscommon_modevent(module_t mod, int type, void *data)
455{
456 int error = 0;
457 static int loaded = 0;
458
459 switch (type) {
460 case MOD_LOAD:
461 if (loaded)
462 return (0);
463 newnfs_portinit();
464 mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
465 mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
466 mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
467 mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
468 mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
469 MTX_DEF);
470 callout_init(&newnfsd_callout, CALLOUT_MPSAFE);
471 newnfs_init();
472 nfsd_call_nfscommon = nfssvc_nfscommon;
473 loaded = 1;
474 break;
475
476 case MOD_UNLOAD:
477 if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
478 nfs_numnfscbd != 0) {
479 error = EBUSY;
480 break;
481 }
482
483 nfsd_call_nfscommon = NULL;
484 callout_drain(&newnfsd_callout);
485 /* and get rid of the mutexes */
486 mtx_destroy(&nfs_nameid_mutex);
487 mtx_destroy(&newnfsd_mtx);
488 mtx_destroy(&nfs_state_mutex);
489 mtx_destroy(&nfs_sockl_mutex);
490 mtx_destroy(&nfs_slock_mutex);
491 mtx_destroy(&nfs_req_mutex);
492 mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
493 loaded = 0;
494 break;
495 default:
496 error = EOPNOTSUPP;
497 break;
498 }
499 return error;
500}
501static moduledata_t nfscommon_mod = {
502 "nfscommon",
503 nfscommon_modevent,
504 NULL,
505};
506DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
507
508/* So that loader and kldload(2) can find us, wherever we are.. */
509MODULE_VERSION(nfscommon, 1);
510MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
511MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
512