xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision 830940567b49bb0c08dfaed40418999e76616909)
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("$FreeBSD$");
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 
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 *,
58     struct thread *);
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;
68 
69 static int nfs_realign_test;
70 static int nfs_realign_count;
71 
72 SYSCTL_NODE(_vfs, OID_AUTO, newnfs, CTLFLAG_RW, 0, "New NFS filesystem");
73 SYSCTL_INT(_vfs_newnfs, OID_AUTO, newnfs_realign_test, CTLFLAG_RW, &nfs_realign_test, 0, "");
74 SYSCTL_INT(_vfs_newnfs, OID_AUTO, newnfs_realign_count, CTLFLAG_RW, &nfs_realign_count, 0, "");
75 SYSCTL_INT(_vfs_newnfs, OID_AUTO, nfs4acl_enable, CTLFLAG_RW, &nfsrv_useacl, 0, "");
76 SYSCTL_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  */
83 MALLOC_DEFINE(M_NEWNFSRVCACHE, "NFSD srvcache", "NFSD Server Request Cache");
84 MALLOC_DEFINE(M_NEWNFSDCLIENT, "NFSD V4client", "NFSD V4 Client Id");
85 MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state", "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
86 MALLOC_DEFINE(M_NEWNFSDLOCK, "NFSD V4lock", "NFSD V4 byte range lock");
87 MALLOC_DEFINE(M_NEWNFSDLOCKFILE, "NFSD lckfile", "NFSD Open/Lock file");
88 MALLOC_DEFINE(M_NEWNFSSTRING, "NFSD string", "NFSD V4 long string");
89 MALLOC_DEFINE(M_NEWNFSUSERGROUP, "NFSD usrgroup", "NFSD V4 User/group map");
90 MALLOC_DEFINE(M_NEWNFSDREQ, "NFS req", "NFS request header");
91 MALLOC_DEFINE(M_NEWNFSFH, "NFS fh", "NFS file handle");
92 MALLOC_DEFINE(M_NEWNFSCLOWNER, "NFSCL owner", "NFSCL Open Owner");
93 MALLOC_DEFINE(M_NEWNFSCLOPEN, "NFSCL open", "NFSCL Open");
94 MALLOC_DEFINE(M_NEWNFSCLDELEG, "NFSCL deleg", "NFSCL Delegation");
95 MALLOC_DEFINE(M_NEWNFSCLCLIENT, "NFSCL client", "NFSCL Client");
96 MALLOC_DEFINE(M_NEWNFSCLLOCKOWNER, "NFSCL lckown", "NFSCL Lock Owner");
97 MALLOC_DEFINE(M_NEWNFSCLLOCK, "NFSCL lck", "NFSCL Lock");
98 MALLOC_DEFINE(M_NEWNFSV4NODE, "NEWNFSnode", "New nfs vnode");
99 MALLOC_DEFINE(M_NEWNFSDIRECTIO, "NEWdirectio", "New nfs Direct IO buffer");
100 MALLOC_DEFINE(M_NEWNFSDIROFF, "Newnfscl_diroff", "New NFS directory offset data");
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  */
110 struct mtx newnfsd_mtx;
111 struct mtx nfs_sockl_mutex;
112 struct mtx nfs_state_mutex;
113 struct mtx nfs_nameid_mutex;
114 struct mtx nfs_req_mutex;
115 struct mtx nfs_slock_mutex;
116 
117 /* local functions */
118 static int nfssvc_call(struct thread *, struct nfssvc_args *, struct ucred *);
119 
120 #if defined(__i386__)
121 /*
122  * These architectures don't need re-alignment, so just return.
123  */
124 void
125 newnfs_realign(struct mbuf **pm)
126 {
127 
128 	return;
129 }
130 #else
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  */
147 void
148 newnfs_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	/* !__i386__ */
189 
190 #ifdef notdef
191 static void
192 nfsrv_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  */
204 int
205 nfsrv_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  */
220 void
221 newnfs_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  */
231 int
232 nfsmsleep(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  */
254 void
255 nfsvno_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 */
284 int
285 nfsrv_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  */
296 void
297 newnfs_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  */
308 struct ucred *
309 newnfs_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  */
323 void
324 newnfs_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  */
346 int
347 nfs_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  */
361 struct nfsreferral *
362 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
363 {
364 
365 	return (NULL);
366 }
367 
368 static int
369 nfssvc_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 
377 static int
378 nfssvc_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  */
411 void
412 newnfs_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 #ifdef NFS4_ACL_EXTATTR_NAME
425 /*
426  * Determine if the file system supports NFSv4 ACLs.
427  * Return 1 if it does, 0 otherwise.
428  */
429 int
430 nfs_supportsnfsv4acls(struct mount *mp)
431 {
432 
433 	if (mp->mnt_stat.f_fstypename == NULL)
434 		return (0);
435 	if (strcmp(mp->mnt_stat.f_fstypename, "ufs") == 0) {
436 		/* Not yet */
437 		return (0);
438 	} else if (strcmp(mp->mnt_stat.f_fstypename, "zfs") == 0) {
439 		/* Always supports them */
440 		return (1);
441 	}
442 	return (0);
443 }
444 #endif	/* NFS4_ACL_EXTATTR_NAME */
445 
446 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
447 
448 /*
449  * Called once to initialize data structures...
450  */
451 static int
452 nfscommon_modevent(module_t mod, int type, void *data)
453 {
454 	int error = 0;
455 	static int loaded = 0;
456 
457 	switch (type) {
458 	case MOD_LOAD:
459 		if (loaded)
460 			return (0);
461 		newnfs_portinit();
462 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
463 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
464 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
465 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
466 		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
467 		    MTX_DEF);
468 		callout_init(&newnfsd_callout, CALLOUT_MPSAFE);
469 		newnfs_init();
470 		nfsd_call_nfscommon = nfssvc_nfscommon;
471 		loaded = 1;
472 		break;
473 
474 	case MOD_UNLOAD:
475 		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
476 		    nfs_numnfscbd != 0) {
477 			error = EBUSY;
478 			break;
479 		}
480 
481 		nfsd_call_nfscommon = NULL;
482 		callout_drain(&newnfsd_callout);
483 		/* and get rid of the mutexes */
484 		mtx_destroy(&nfs_nameid_mutex);
485 		mtx_destroy(&newnfsd_mtx);
486 		mtx_destroy(&nfs_state_mutex);
487 		mtx_destroy(&nfs_sockl_mutex);
488 		mtx_destroy(&nfs_slock_mutex);
489 		mtx_destroy(&nfs_req_mutex);
490 		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
491 		loaded = 0;
492 		break;
493 	default:
494 		error = EOPNOTSUPP;
495 		break;
496 	}
497 	return error;
498 }
499 static moduledata_t nfscommon_mod = {
500 	"nfscommon",
501 	nfscommon_modevent,
502 	NULL,
503 };
504 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
505 
506 /* So that loader and kldload(2) can find us, wherever we are.. */
507 MODULE_VERSION(nfscommon, 1);
508 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
509 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
510 
511