nfs_commonport.c revision 220648
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 220648 2011-04-14 23:46:15Z 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");
85MALLOC_DEFINE(M_NEWNFSDSTATE, "NFSD V4state",
86    "NFSD V4 State (Openowner, Open, Lockowner, Delegation");
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");
101MALLOC_DEFINE(M_NEWNFSDIROFF, "NFSCL diroffdiroff",
102    "New NFS directory offset data");
103MALLOC_DEFINE(M_NEWNFSDROLLBACK, "NFSD rollback",
104    "New NFS local lock rollback");
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 | MPSAFE, UIO_USERSPACE, fname,
214	    p);
215	error = namei(ndp);
216	if (!error) {
217		NDFREE(ndp, NDF_ONLY_PNBUF);
218	}
219	return (error);
220}
221
222/*
223 * Copy NFS uid, gids to the cred structure.
224 */
225void
226newnfs_copycred(struct nfscred *nfscr, struct ucred *cr)
227{
228
229	KASSERT(nfscr->nfsc_ngroups >= 0,
230	    ("newnfs_copycred: negative nfsc_ngroups"));
231	cr->cr_uid = nfscr->nfsc_uid;
232	crsetgroups(cr, nfscr->nfsc_ngroups, nfscr->nfsc_groups);
233}
234
235/*
236 * Map args from nfsmsleep() to msleep().
237 */
238int
239nfsmsleep(void *chan, void *mutex, int prio, const char *wmesg,
240    struct timespec *ts)
241{
242	u_int64_t nsecval;
243	int error, timeo;
244
245	if (ts) {
246		timeo = hz * ts->tv_sec;
247		nsecval = (u_int64_t)ts->tv_nsec;
248		nsecval = ((nsecval * ((u_int64_t)hz)) + 500000000) /
249		    1000000000;
250		timeo += (int)nsecval;
251	} else {
252		timeo = 0;
253	}
254	error = msleep(chan, (struct mtx *)mutex, prio, wmesg, timeo);
255	return (error);
256}
257
258/*
259 * Get the file system info for the server. For now, just assume FFS.
260 */
261void
262nfsvno_getfs(struct nfsfsinfo *sip, int isdgram)
263{
264	int pref;
265
266	/*
267	 * XXX
268	 * There should be file system VFS OP(s) to get this information.
269	 * For now, assume ufs.
270	 */
271	if (isdgram)
272		pref = NFS_MAXDGRAMDATA;
273	else
274		pref = NFS_MAXDATA;
275	sip->fs_rtmax = NFS_MAXDATA;
276	sip->fs_rtpref = pref;
277	sip->fs_rtmult = NFS_FABLKSIZE;
278	sip->fs_wtmax = NFS_MAXDATA;
279	sip->fs_wtpref = pref;
280	sip->fs_wtmult = NFS_FABLKSIZE;
281	sip->fs_dtpref = pref;
282	sip->fs_maxfilesize = 0xffffffffffffffffull;
283	sip->fs_timedelta.tv_sec = 0;
284	sip->fs_timedelta.tv_nsec = 1;
285	sip->fs_properties = (NFSV3FSINFO_LINK |
286	    NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
287	    NFSV3FSINFO_CANSETTIME);
288}
289
290/* Fake nfsrv_atroot. Just return 0 */
291int
292nfsrv_atroot(struct vnode *vp, long *retp)
293{
294
295	return (0);
296}
297
298/*
299 * Set the credentials to refer to root.
300 * If only the various BSDen could agree on whether cr_gid is a separate
301 * field or cr_groups[0]...
302 */
303void
304newnfs_setroot(struct ucred *cred)
305{
306
307	cred->cr_uid = 0;
308	cred->cr_groups[0] = 0;
309	cred->cr_ngroups = 1;
310}
311
312/*
313 * Get the client credential. Used for Renew and recovery.
314 */
315struct ucred *
316newnfs_getcred(void)
317{
318	struct ucred *cred;
319	struct thread *td = curthread;
320
321	cred = crdup(td->td_ucred);
322	newnfs_setroot(cred);
323	return (cred);
324}
325
326/*
327 * Nfs timer routine
328 * Call the nfsd's timer function once/sec.
329 */
330void
331newnfs_timer(void *arg)
332{
333	static time_t lasttime = 0;
334	/*
335	 * Call the server timer, if set up.
336	 * The argument indicates if it is the next second and therefore
337	 * leases should be checked.
338	 */
339	if (lasttime != NFSD_MONOSEC) {
340		lasttime = NFSD_MONOSEC;
341		if (nfsd_call_servertimer != NULL)
342			(*nfsd_call_servertimer)();
343	}
344	callout_reset(&newnfsd_callout, nfscl_ticks, newnfs_timer, NULL);
345}
346
347
348/*
349 * Sleep for a short period of time unless errval == NFSERR_GRACE, where
350 * the sleep should be for 5 seconds.
351 * Since lbolt doesn't exist in FreeBSD-CURRENT, just use a timeout on
352 * an event that never gets a wakeup. Only return EINTR or 0.
353 */
354int
355nfs_catnap(int prio, int errval, const char *wmesg)
356{
357	static int non_event;
358	int ret;
359
360	if (errval == NFSERR_GRACE)
361		ret = tsleep(&non_event, prio, wmesg, 5 * hz);
362	else
363		ret = tsleep(&non_event, prio, wmesg, 1);
364	if (ret != EINTR)
365		ret = 0;
366	return (ret);
367}
368
369/*
370 * Get referral. For now, just fail.
371 */
372struct nfsreferral *
373nfsv4root_getreferral(struct vnode *vp, struct vnode *dvp, u_int32_t fileno)
374{
375
376	return (NULL);
377}
378
379static int
380nfssvc_nfscommon(struct thread *td, struct nfssvc_args *uap)
381{
382	int error;
383
384	error = nfssvc_call(td, uap, td->td_ucred);
385	return (error);
386}
387
388static int
389nfssvc_call(struct thread *p, struct nfssvc_args *uap, struct ucred *cred)
390{
391	int error = EINVAL;
392	struct nfsd_idargs nid;
393
394	if (uap->flag & NFSSVC_IDNAME) {
395		error = copyin(uap->argp, (caddr_t)&nid, sizeof (nid));
396		if (error)
397			return (error);
398		error = nfssvc_idname(&nid);
399		return (error);
400	} else if (uap->flag & NFSSVC_GETSTATS) {
401		error = copyout(&newnfsstats,
402		    CAST_USER_ADDR_T(uap->argp), sizeof (newnfsstats));
403		return (error);
404	} else if (uap->flag & NFSSVC_NFSUSERDPORT) {
405		u_short sockport;
406
407		error = copyin(uap->argp, (caddr_t)&sockport,
408		    sizeof (u_short));
409		if (!error)
410			error = nfsrv_nfsuserdport(sockport, p);
411	} else if (uap->flag & NFSSVC_NFSUSERDDELPORT) {
412		nfsrv_nfsuserddelport();
413		error = 0;
414	}
415	return (error);
416}
417
418/*
419 * called by all three modevent routines, so that it gets things
420 * initialized soon enough.
421 */
422void
423newnfs_portinit(void)
424{
425	static int inited = 0;
426
427	if (inited)
428		return;
429	inited = 1;
430	/* Initialize SMP locks used by both client and server. */
431	mtx_init(&newnfsd_mtx, "newnfsd_mtx", NULL, MTX_DEF);
432	mtx_init(&nfs_state_mutex, "nfs_state_mutex", NULL, MTX_DEF);
433}
434
435/*
436 * Determine if the file system supports NFSv4 ACLs.
437 * Return 1 if it does, 0 otherwise.
438 */
439int
440nfs_supportsnfsv4acls(struct vnode *vp)
441{
442	int error;
443	register_t retval;
444
445	ASSERT_VOP_LOCKED(vp, "nfs supports nfsv4acls");
446
447	if (nfsrv_useacl == 0)
448		return (0);
449	error = VOP_PATHCONF(vp, _PC_ACL_NFS4, &retval);
450	if (error == 0 && retval != 0)
451		return (1);
452	return (0);
453}
454
455extern int (*nfsd_call_nfscommon)(struct thread *, struct nfssvc_args *);
456
457/*
458 * Called once to initialize data structures...
459 */
460static int
461nfscommon_modevent(module_t mod, int type, void *data)
462{
463	int error = 0;
464	static int loaded = 0;
465
466	switch (type) {
467	case MOD_LOAD:
468		if (loaded)
469			return (0);
470		newnfs_portinit();
471		mtx_init(&nfs_nameid_mutex, "nfs_nameid_mutex", NULL, MTX_DEF);
472		mtx_init(&nfs_sockl_mutex, "nfs_sockl_mutex", NULL, MTX_DEF);
473		mtx_init(&nfs_slock_mutex, "nfs_slock_mutex", NULL, MTX_DEF);
474		mtx_init(&nfs_req_mutex, "nfs_req_mutex", NULL, MTX_DEF);
475		mtx_init(&nfsrv_nfsuserdsock.nr_mtx, "nfsuserd", NULL,
476		    MTX_DEF);
477		callout_init(&newnfsd_callout, CALLOUT_MPSAFE);
478		newnfs_init();
479		nfsd_call_nfscommon = nfssvc_nfscommon;
480		loaded = 1;
481		break;
482
483	case MOD_UNLOAD:
484		if (newnfs_numnfsd != 0 || nfsrv_nfsuserd != 0 ||
485		    nfs_numnfscbd != 0) {
486			error = EBUSY;
487			break;
488		}
489
490		nfsd_call_nfscommon = NULL;
491		callout_drain(&newnfsd_callout);
492		/* and get rid of the mutexes */
493		mtx_destroy(&nfs_nameid_mutex);
494		mtx_destroy(&newnfsd_mtx);
495		mtx_destroy(&nfs_state_mutex);
496		mtx_destroy(&nfs_sockl_mutex);
497		mtx_destroy(&nfs_slock_mutex);
498		mtx_destroy(&nfs_req_mutex);
499		mtx_destroy(&nfsrv_nfsuserdsock.nr_mtx);
500		loaded = 0;
501		break;
502	default:
503		error = EOPNOTSUPP;
504		break;
505	}
506	return error;
507}
508static moduledata_t nfscommon_mod = {
509	"nfscommon",
510	nfscommon_modevent,
511	NULL,
512};
513DECLARE_MODULE(nfscommon, nfscommon_mod, SI_SUB_VFS, SI_ORDER_ANY);
514
515/* So that loader and kldload(2) can find us, wherever we are.. */
516MODULE_VERSION(nfscommon, 1);
517MODULE_DEPEND(nfscommon, nfssvc, 1, 1, 1);
518MODULE_DEPEND(nfscommon, krpc, 1, 1, 1);
519
520