xref: /freebsd/sys/fs/nfs/nfs_commonport.c (revision f4f8f02054f3abb6ceb84aefcdecc78d5c8b462f)
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, realign_test, CTLFLAG_RW, &nfs_realign_test, 0, "");
74 SYSCTL_INT(_vfs_newnfs, OID_AUTO, 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  *	nfs_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 void
147 newnfs_realign(struct mbuf **pm)
148 {
149 	struct mbuf *m;
150 	struct mbuf *n = NULL;
151 	int off = 0;
152 
153 	++nfs_realign_test;
154 	while ((m = *pm) != NULL) {
155 		if ((m->m_len & 0x3) || (mtod(m, intptr_t) & 0x3)) {
156 			MGET(n, M_WAIT, MT_DATA);
157 			if (m->m_len >= MINCLSIZE) {
158 				MCLGET(n, M_WAIT);
159 			}
160 			n->m_len = 0;
161 			break;
162 		}
163 		pm = &m->m_next;
164 	}
165 
166 	/*
167 	 * If n is non-NULL, loop on m copying data, then replace the
168 	 * portion of the chain that had to be realigned.
169 	 */
170 	if (n != NULL) {
171 		++nfs_realign_count;
172 		while (m) {
173 			m_copyback(n, off, m->m_len, mtod(m, caddr_t));
174 			off += m->m_len;
175 			m = m->m_next;
176 		}
177 		m_freem(*pm);
178 		*pm = n;
179 	}
180 }
181 #endif	/* newnfs_realign */
182 
183 #ifdef notdef
184 static void
185 nfsrv_object_create(struct vnode *vp, struct thread *td)
186 {
187 
188 	if (vp == NULL || vp->v_type != VREG)
189 		return;
190 	(void) vfs_object_create(vp, td, td->td_ucred);
191 }
192 #endif
193 
194 /*
195  * Look up a file name. Basically just initialize stuff and call namei().
196  */
197 int
198 nfsrv_lookupfilename(struct nameidata *ndp, char *fname, NFSPROC_T *p)
199 {
200 	int error;
201 
202 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, fname, p);
203 	error = namei(ndp);
204 	if (!error) {
205 		NDFREE(ndp, NDF_ONLY_PNBUF);
206 	}
207 	return (error);
208 }
209 
210 /*
211  * Copy NFS uid, gids to the cred structure.
212  */
213 void
214 newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
215 {
216 	int ngroups, i;
217 
218 	cr->cr_uid = nfscr->nfsc_uid;
219 	ngroups = (nfscr->nfsc_ngroups < NGROUPS) ?
220 	    nfscr->nfsc_ngroups : NGROUPS;
221 	for (i = 0; i < ngroups; i++)
222 		cr->cr_groups[i] = nfscr->nfsc_groups[i];
223 	cr->cr_ngroups = ngroups;
224 }
225 
226 /*
227  * Map args from nfsmsleep() to msleep().
228  */
229 int
230 nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
231     struct timespec *ts)
232 {
233 	u_int64_t nsecval;
234 	int error, timeo;
235 
236 	if (ts) {
237 		timeo = hz * ts->tv_sec;
238 		nsecval = (u_int64_t)ts->tv_nsec;
239 		nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
240 		    1000000000;
241 		timeo += (int)nsecval;
242 	} else {
243 		timeo = 0;
244 	}
245 	error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
246 	return (error);
247 }
248 
249 /*
250  * Get the file system info for the server. For now, just assume FFS.
251  */
252 void
253 nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
254 {
255 	int pref;
256 
257 	/*
258 	 * XXX
259 	 * There should be file system VFS OP(s) to get this information.
260 	 * For now, assume ufs.
261 	 */
262 	if (isdgram)
263 		pref = NFS_MAXDGRAMDATA;
264 	else
265 		pref = NFS_MAXDATA;
266 	sip->fs_rtmax = NFS_MAXDATA;
267 	sip->fs_rtpref = pref;
268 	sip->fs_rtmult = NFS_FABLKSIZE;
269 	sip->fs_wtmax = NFS_MAXDATA;
270 	sip->fs_wtpref = pref;
271 	sip->fs_wtmult = NFS_FABLKSIZE;
272 	sip->fs_dtpref = pref;
273 	sip->fs_maxfilesize = 0xffffffffffffffffull;
274 	sip->fs_timedelta.tv_sec = 0;
275 	sip->fs_timedelta.tv_nsec = 1;
276 	sip->fs_properties = (NFSV3FSINFO_LINK |
277 	    NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
278 	    NFSV3FSINFO_CANSETTIME);
279 }
280 
281 /* Fake nfsrv_atroot. Just return 0 */
282 int
283 nfsrv_atroot(struct vnode *vp, long *retp)
284 {
285 
286 	return (0);
287 }
288 
289 /*
290  * Set the credentials to refer to root.
291  * If only the various BSDen could agree on whether cr_gid is a separate
292  * field or cr_groups[0]...
293  */
294 void
295 newnfs_setroot(struct ucred *cred)
296 {
297 
298 	cred->cr_uid = 0;
299 	cred->cr_groups[0] = 0;
300 	cred->cr_ngroups = 1;
301 }
302 
303 /*
304  * Get the client credential. Used for Renew and recovery.
305  */
306 struct ucred *
307 newnfs_getcred(void)
308 {
309 	struct ucred *cred;
310 	struct thread *td = curthread;
311 
312 	cred = crdup(td->td_ucred);
313 	newnfs_setroot(cred);
314 	return (cred);
315 }
316 
317 /*
318  * Nfs timer routine
319  * Call the nfsd's timer function once/sec.
320  */
321 void
322 newnfs_timer(void *arg)
323 {
324 	static time_t lasttime = 0;
325 	/*
326 	 * Call the server timer, if set up.
327 	 * The argument indicates if it is the next second and therefore
328 	 * leases should be checked.
329 	 */
330 	if (lasttime != NFSD_MONOSEC) {
331 		lasttime = NFSD_MONOSEC;
332 		if (nfsd_call_servertimer != NULL)
333 			(*nfsd_call_servertimer)();
334 	}
335 	callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
336 }
337 
338 
339 /*
340  * sleep for a short period of time.
341  * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
342  * an event that never gets a wakeup. Only return EINTR or 0.
343  */
344 int
345 nfs_catnap(int prio, const char *wmesg)
346 {
347 	static int non_event;
348 	int ret;
349 
350 	ret = tsleep(&non_event, prio, wmesg, 1);
351 	if (ret != EINTR)
352 		ret = 0;
353 	return (ret);
354 }
355 
356 /*
357  * Get referral. For now, just fail.
358  */
359 struct nfsreferral *
360 nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
361 {
362 
363 	return (NULL);
364 }
365 
366 static int
367 nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
368 {
369 	int error;
370 
371 	error = nfssvc_call(td, uap, td->td_ucred);
372 	return (error);
373 }
374 
375 static int
376 nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
377 {
378 	int error = EINVAL;
379 	struct nfsd_idargs nid;
380 
381 	if (uap->flag & NFSSVC_IDNAME) {
382 		error = copyin(uap->argp, (caddr_t)&nid, sizeof (nid));
383 		if (error)
384 			return (error);
385 		error = nfssvc_idname(&nid);
386 		return (error);
387 	} else if (uap->flag & NFSSVC_GETSTATS) {
388 		error = copyout(&newnfsstats,
389 		    CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
390 		return (error);
391 	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
392 		u_short sockport;
393 
394 		error = copyin(uap->argp, (caddr_t)&sockport,
395 		    sizeof (u_short));
396 		if (!error)
397 			error = nfsrv_nfsuserdport(sockport, p);
398 	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
399 		nfsrv_nfsuserddelport();
400 		error = 0;
401 	}
402 	return (error);
403 }
404 
405 /*
406  * called by all three modevent routines, so that it gets things
407  * initialized soon enough.
408  */
409 void
410 newnfs_portinit(void)
411 {
412 	static int inited = 0;
413 
414 	if (inited)
415 		return;
416 	inited = 1;
417 	/* Initialize SMP locks used by both client and server. */
418 	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
419 	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
420 }
421 
422 extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
423 
424 /*
425  * Called once to initialize data structures...
426  */
427 static int
428 nfscommon_modevent(module_t mod, int type, void *data)
429 {
430 	int error = 0;
431 	static int loaded = 0;
432 
433 	switch (type) {
434 	case MOD_LOAD:
435 		if (loaded)
436 			return (0);
437 		newnfs_portinit();
438 		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
439 		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
440 		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
441 		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
442 		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
443 		    MTX_DEF);
444 		callout_init(&newnfsd_callout, CALLOUT_MPSAFE);
445 		newnfs_init();
446 		nfsd_call_nfscommon = nfssvc_nfscommon;
447 		loaded = 1;
448 		break;
449 
450 	case MOD_UNLOAD:
451 		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
452 		    nfs_numnfscbd != 0) {
453 			error = EBUSY;
454 			break;
455 		}
456 
457 		nfsd_call_nfscommon = NULL;
458 		callout_drain(&newnfsd_callout);
459 		/* and get rid of the mutexes */
460 		mtx_destroy(&nfs_nameid_mutex);
461 		mtx_destroy(&newnfsd_mtx);
462 		mtx_destroy(&nfs_state_mutex);
463 		mtx_destroy(&nfs_sockl_mutex);
464 		mtx_destroy(&nfs_slock_mutex);
465 		mtx_destroy(&nfs_req_mutex);
466 		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
467 		loaded = 0;
468 		break;
469 	default:
470 		error = EOPNOTSUPP;
471 		break;
472 	}
473 	return error;
474 }
475 static moduledata_t nfscommon_mod = {
476 	"nfscommon",
477 	nfscommon_modevent,
478 	NULL,
479 };
480 DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
481 
482 /* So that loader and kldload(2) can find us, wherever we are.. */
483 MODULE_VERSION(nfscommon, 1);
484 MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
485 MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
486 
487