nfs_clnode.c revision 302210
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 *	from nfs_node.c	8.6 (Berkeley) 5/22/95
33 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: head/sys/fs/nfsclient/nfs_clnode.c 302210 2016-06-26 14:18:28Z kib $");
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/fcntl.h>
41#include <sys/lock.h>
42#include <sys/malloc.h>
43#include <sys/mount.h>
44#include <sys/namei.h>
45#include <sys/proc.h>
46#include <sys/socket.h>
47#include <sys/sysctl.h>
48#include <sys/taskqueue.h>
49#include <sys/vnode.h>
50
51#include <vm/uma.h>
52
53#include <fs/nfs/nfsport.h>
54#include <fs/nfsclient/nfsnode.h>
55#include <fs/nfsclient/nfsmount.h>
56#include <fs/nfsclient/nfs.h>
57#include <fs/nfsclient/nfs_kdtrace.h>
58
59#include <nfs/nfs_lock.h>
60
61extern struct vop_vector newnfs_vnodeops;
62extern struct buf_ops buf_ops_newnfs;
63MALLOC_DECLARE(M_NEWNFSREQ);
64
65uma_zone_t newnfsnode_zone;
66
67const char nfs_vnode_tag[] = "nfs";
68
69static void	nfs_freesillyrename(void *arg, __unused int pending);
70
71void
72ncl_nhinit(void)
73{
74
75	newnfsnode_zone = uma_zcreate("NCLNODE", sizeof(struct nfsnode), NULL,
76	    NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
77}
78
79void
80ncl_nhuninit(void)
81{
82	uma_zdestroy(newnfsnode_zone);
83}
84
85/*
86 * ONLY USED FOR THE ROOT DIRECTORY. nfscl_nget() does the rest. If this
87 * function is going to be used to get Regular Files, code must be added
88 * to fill in the "struct nfsv4node".
89 * Look up a vnode/nfsnode by file handle.
90 * Callers must check for mount points!!
91 * In all cases, a pointer to a
92 * nfsnode structure is returned.
93 */
94int
95ncl_nget(struct mount *mntp, u_int8_t *fhp, int fhsize, struct nfsnode **npp,
96    int lkflags)
97{
98	struct thread *td = curthread;	/* XXX */
99	struct nfsnode *np;
100	struct vnode *vp;
101	struct vnode *nvp;
102	int error;
103	u_int hash;
104	struct nfsmount *nmp;
105	struct nfsfh *nfhp;
106
107	nmp = VFSTONFS(mntp);
108	*npp = NULL;
109
110	hash = fnv_32_buf(fhp, fhsize, FNV1_32_INIT);
111
112	MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) + fhsize,
113	    M_NFSFH, M_WAITOK);
114	bcopy(fhp, &nfhp->nfh_fh[0], fhsize);
115	nfhp->nfh_len = fhsize;
116	error = vfs_hash_get(mntp, hash, lkflags,
117	    td, &nvp, newnfs_vncmpf, nfhp);
118	FREE(nfhp, M_NFSFH);
119	if (error)
120		return (error);
121	if (nvp != NULL) {
122		*npp = VTONFS(nvp);
123		return (0);
124	}
125	np = uma_zalloc(newnfsnode_zone, M_WAITOK | M_ZERO);
126
127	error = getnewvnode(nfs_vnode_tag, mntp, &newnfs_vnodeops, &nvp);
128	if (error) {
129		uma_zfree(newnfsnode_zone, np);
130		return (error);
131	}
132	vp = nvp;
133	KASSERT(vp->v_bufobj.bo_bsize != 0, ("ncl_nget: bo_bsize == 0"));
134	vp->v_bufobj.bo_ops = &buf_ops_newnfs;
135	vp->v_data = np;
136	np->n_vnode = vp;
137	/*
138	 * Initialize the mutex even if the vnode is going to be a loser.
139	 * This simplifies the logic in reclaim, which can then unconditionally
140	 * destroy the mutex (in the case of the loser, or if hash_insert
141	 * happened to return an error no special casing is needed).
142	 */
143	mtx_init(&np->n_mtx, "NEWNFSnode lock", NULL, MTX_DEF | MTX_DUPOK);
144	/*
145	 * NFS supports recursive and shared locking.
146	 */
147	lockmgr(vp->v_vnlock, LK_EXCLUSIVE | LK_NOWITNESS, NULL);
148	VN_LOCK_AREC(vp);
149	VN_LOCK_ASHARE(vp);
150	/*
151	 * Are we getting the root? If so, make sure the vnode flags
152	 * are correct
153	 */
154	if ((fhsize == nmp->nm_fhsize) &&
155	    !bcmp(fhp, nmp->nm_fh, fhsize)) {
156		if (vp->v_type == VNON)
157			vp->v_type = VDIR;
158		vp->v_vflag |= VV_ROOT;
159	}
160
161	MALLOC(np->n_fhp, struct nfsfh *, sizeof (struct nfsfh) + fhsize,
162	    M_NFSFH, M_WAITOK);
163	bcopy(fhp, np->n_fhp->nfh_fh, fhsize);
164	np->n_fhp->nfh_len = fhsize;
165	error = insmntque(vp, mntp);
166	if (error != 0) {
167		*npp = NULL;
168		FREE((caddr_t)np->n_fhp, M_NFSFH);
169		mtx_destroy(&np->n_mtx);
170		uma_zfree(newnfsnode_zone, np);
171		return (error);
172	}
173	error = vfs_hash_insert(vp, hash, lkflags,
174	    td, &nvp, newnfs_vncmpf, np->n_fhp);
175	if (error)
176		return (error);
177	if (nvp != NULL) {
178		*npp = VTONFS(nvp);
179		/* vfs_hash_insert() vput()'s the losing vnode */
180		return (0);
181	}
182	*npp = np;
183
184	return (0);
185}
186
187/*
188 * Do the vrele(sp->s_dvp) as a separate task in order to avoid a
189 * deadlock because of a LOR when vrele() locks the directory vnode.
190 */
191static void
192nfs_freesillyrename(void *arg, __unused int pending)
193{
194	struct sillyrename *sp;
195
196	sp = arg;
197	vrele(sp->s_dvp);
198	free(sp, M_NEWNFSREQ);
199}
200
201static void
202ncl_releasesillyrename(struct vnode *vp, struct thread *td)
203{
204	struct nfsnode *np;
205	struct sillyrename *sp;
206
207	ASSERT_VOP_ELOCKED(vp, "releasesillyrename");
208	np = VTONFS(vp);
209	mtx_assert(&np->n_mtx, MA_OWNED);
210	if (vp->v_type != VDIR) {
211		sp = np->n_sillyrename;
212		np->n_sillyrename = NULL;
213	} else
214		sp = NULL;
215	if (sp != NULL) {
216		mtx_unlock(&np->n_mtx);
217		(void) ncl_vinvalbuf(vp, 0, td, 1);
218		/*
219		 * Remove the silly file that was rename'd earlier
220		 */
221		ncl_removeit(sp, vp);
222		crfree(sp->s_cred);
223		TASK_INIT(&sp->s_task, 0, nfs_freesillyrename, sp);
224		taskqueue_enqueue(taskqueue_thread, &sp->s_task);
225		mtx_lock(&np->n_mtx);
226	}
227}
228
229int
230ncl_inactive(struct vop_inactive_args *ap)
231{
232	struct vnode *vp = ap->a_vp;
233	struct nfsnode *np;
234	boolean_t retv;
235
236	if (NFS_ISV4(vp) && vp->v_type == VREG) {
237		/*
238		 * Since mmap()'d files do I/O after VOP_CLOSE(), the NFSv4
239		 * Close operations are delayed until now. Any dirty
240		 * buffers/pages must be flushed before the close, so that the
241		 * stateid is available for the writes.
242		 */
243		if (vp->v_object != NULL) {
244			VM_OBJECT_WLOCK(vp->v_object);
245			retv = vm_object_page_clean(vp->v_object, 0, 0,
246			    OBJPC_SYNC);
247			VM_OBJECT_WUNLOCK(vp->v_object);
248		} else
249			retv = TRUE;
250		if (retv == TRUE) {
251			(void)ncl_flush(vp, MNT_WAIT, NULL, ap->a_td, 1, 0);
252			(void)nfsrpc_close(vp, 1, ap->a_td);
253		}
254	}
255
256	np = VTONFS(vp);
257	mtx_lock(&np->n_mtx);
258	ncl_releasesillyrename(vp, ap->a_td);
259
260	/*
261	 * NMODIFIED means that there might be dirty/stale buffers
262	 * associated with the NFS vnode.  None of the other flags are
263	 * meaningful after the vnode is unused.
264	 */
265	np->n_flag &= NMODIFIED;
266	mtx_unlock(&np->n_mtx);
267	return (0);
268}
269
270/*
271 * Reclaim an nfsnode so that it can be used for other purposes.
272 */
273int
274ncl_reclaim(struct vop_reclaim_args *ap)
275{
276	struct vnode *vp = ap->a_vp;
277	struct nfsnode *np = VTONFS(vp);
278	struct nfsdmap *dp, *dp2;
279
280	/*
281	 * If the NLM is running, give it a chance to abort pending
282	 * locks.
283	 */
284	if (nfs_reclaim_p != NULL)
285		nfs_reclaim_p(ap);
286
287	mtx_lock(&np->n_mtx);
288	ncl_releasesillyrename(vp, ap->a_td);
289	mtx_unlock(&np->n_mtx);
290
291	/*
292	 * Destroy the vm object and flush associated pages.
293	 */
294	vnode_destroy_vobject(vp);
295
296	if (NFS_ISV4(vp) && vp->v_type == VREG)
297		/*
298		 * We can now safely close any remaining NFSv4 Opens for
299		 * this file. Most opens will have already been closed by
300		 * ncl_inactive(), but there are cases where it is not
301		 * called, so we need to do it again here.
302		 */
303		(void) nfsrpc_close(vp, 1, ap->a_td);
304
305	vfs_hash_remove(vp);
306
307	/*
308	 * Call nfscl_reclaimnode() to save attributes in the delegation,
309	 * as required.
310	 */
311	if (vp->v_type == VREG)
312		nfscl_reclaimnode(vp);
313
314	/*
315	 * Free up any directory cookie structures and
316	 * large file handle structures that might be associated with
317	 * this nfs node.
318	 */
319	if (vp->v_type == VDIR) {
320		dp = LIST_FIRST(&np->n_cookies);
321		while (dp) {
322			dp2 = dp;
323			dp = LIST_NEXT(dp, ndm_list);
324			FREE((caddr_t)dp2, M_NFSDIROFF);
325		}
326	}
327	if (np->n_writecred != NULL)
328		crfree(np->n_writecred);
329	FREE((caddr_t)np->n_fhp, M_NFSFH);
330	if (np->n_v4 != NULL)
331		FREE((caddr_t)np->n_v4, M_NFSV4NODE);
332	mtx_destroy(&np->n_mtx);
333	uma_zfree(newnfsnode_zone, vp->v_data);
334	vp->v_data = NULL;
335	return (0);
336}
337
338/*
339 * Invalidate both the access and attribute caches for this vnode.
340 */
341void
342ncl_invalcaches(struct vnode *vp)
343{
344	struct nfsnode *np = VTONFS(vp);
345	int i;
346
347	mtx_lock(&np->n_mtx);
348	for (i = 0; i < NFS_ACCESSCACHESIZE; i++)
349		np->n_accesscache[i].stamp = 0;
350	KDTRACE_NFS_ACCESSCACHE_FLUSH_DONE(vp);
351	np->n_attrstamp = 0;
352	KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
353	mtx_unlock(&np->n_mtx);
354}
355