1191783Srmacklem/*-
2191783Srmacklem * Copyright (c) 1989, 1993
3191783Srmacklem *	The Regents of the University of California.  All rights reserved.
4191783Srmacklem *
5191783Srmacklem * This code is derived from software contributed to Berkeley by
6191783Srmacklem * Rick Macklem at The University of Guelph.
7191783Srmacklem *
8191783Srmacklem * Redistribution and use in source and binary forms, with or without
9191783Srmacklem * modification, are permitted provided that the following conditions
10191783Srmacklem * are met:
11191783Srmacklem * 1. Redistributions of source code must retain the above copyright
12191783Srmacklem *    notice, this list of conditions and the following disclaimer.
13191783Srmacklem * 2. Redistributions in binary form must reproduce the above copyright
14191783Srmacklem *    notice, this list of conditions and the following disclaimer in the
15191783Srmacklem *    documentation and/or other materials provided with the distribution.
16191783Srmacklem * 4. Neither the name of the University nor the names of its contributors
17191783Srmacklem *    may be used to endorse or promote products derived from this software
18191783Srmacklem *    without specific prior written permission.
19191783Srmacklem *
20191783Srmacklem * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21191783Srmacklem * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22191783Srmacklem * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23191783Srmacklem * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24191783Srmacklem * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25191783Srmacklem * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26191783Srmacklem * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27191783Srmacklem * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28191783Srmacklem * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29191783Srmacklem * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30191783Srmacklem * SUCH DAMAGE.
31191783Srmacklem *
32191783Srmacklem *	from nfs_node.c	8.6 (Berkeley) 5/22/95
33191783Srmacklem */
34191783Srmacklem
35191783Srmacklem#include <sys/cdefs.h>
36191783Srmacklem__FBSDID("$FreeBSD: releng/10.2/sys/fs/nfsclient/nfs_clnode.c 248084 2013-03-09 02:32:23Z attilio $");
37191783Srmacklem
38223280Srmacklem#include "opt_kdtrace.h"
39223280Srmacklem
40191783Srmacklem#include <sys/param.h>
41191783Srmacklem#include <sys/systm.h>
42214048Srmacklem#include <sys/fcntl.h>
43191783Srmacklem#include <sys/lock.h>
44191783Srmacklem#include <sys/malloc.h>
45191783Srmacklem#include <sys/mount.h>
46191783Srmacklem#include <sys/namei.h>
47191783Srmacklem#include <sys/proc.h>
48191783Srmacklem#include <sys/socket.h>
49191783Srmacklem#include <sys/sysctl.h>
50224606Srmacklem#include <sys/taskqueue.h>
51191783Srmacklem#include <sys/vnode.h>
52191783Srmacklem
53191783Srmacklem#include <vm/uma.h>
54191783Srmacklem
55191783Srmacklem#include <fs/nfs/nfsport.h>
56191783Srmacklem#include <fs/nfsclient/nfsnode.h>
57191783Srmacklem#include <fs/nfsclient/nfsmount.h>
58191783Srmacklem#include <fs/nfsclient/nfs.h>
59223280Srmacklem#include <fs/nfsclient/nfs_kdtrace.h>
60191783Srmacklem
61214048Srmacklem#include <nfs/nfs_lock.h>
62214048Srmacklem
63191783Srmacklemextern struct vop_vector newnfs_vnodeops;
64191783Srmacklemextern struct buf_ops buf_ops_newnfs;
65191783SrmacklemMALLOC_DECLARE(M_NEWNFSREQ);
66191783Srmacklem
67191783Srmacklemuma_zone_t newnfsnode_zone;
68191783Srmacklem
69224606Srmacklemstatic void	nfs_freesillyrename(void *arg, __unused int pending);
70224606Srmacklem
71191783Srmacklemvoid
72191783Srmacklemncl_nhinit(void)
73191783Srmacklem{
74191783Srmacklem
75191783Srmacklem	newnfsnode_zone = uma_zcreate("NCLNODE", sizeof(struct nfsnode), NULL,
76191783Srmacklem	    NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
77191783Srmacklem}
78191783Srmacklem
79191783Srmacklemvoid
80191783Srmacklemncl_nhuninit(void)
81191783Srmacklem{
82191783Srmacklem	uma_zdestroy(newnfsnode_zone);
83191783Srmacklem}
84191783Srmacklem
85191783Srmacklem/*
86191783Srmacklem * ONLY USED FOR THE ROOT DIRECTORY. nfscl_nget() does the rest. If this
87191783Srmacklem * function is going to be used to get Regular Files, code must be added
88191783Srmacklem * to fill in the "struct nfsv4node".
89191783Srmacklem * Look up a vnode/nfsnode by file handle.
90191783Srmacklem * Callers must check for mount points!!
91191783Srmacklem * In all cases, a pointer to a
92191783Srmacklem * nfsnode structure is returned.
93191783Srmacklem */
94191783Srmacklemint
95220732Srmacklemncl_nget(struct mount *mntp, u_int8_t *fhp, int fhsize, struct nfsnode **npp,
96220732Srmacklem    int lkflags)
97191783Srmacklem{
98191783Srmacklem	struct thread *td = curthread;	/* XXX */
99191783Srmacklem	struct nfsnode *np;
100191783Srmacklem	struct vnode *vp;
101191783Srmacklem	struct vnode *nvp;
102191783Srmacklem	int error;
103191783Srmacklem	u_int hash;
104191783Srmacklem	struct nfsmount *nmp;
105191783Srmacklem	struct nfsfh *nfhp;
106191783Srmacklem
107191783Srmacklem	nmp = VFSTONFS(mntp);
108191783Srmacklem	*npp = NULL;
109191783Srmacklem
110191783Srmacklem	hash = fnv_32_buf(fhp, fhsize, FNV1_32_INIT);
111191783Srmacklem
112191783Srmacklem	MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) + fhsize,
113191783Srmacklem	    M_NFSFH, M_WAITOK);
114191783Srmacklem	bcopy(fhp, &nfhp->nfh_fh[0], fhsize);
115191783Srmacklem	nfhp->nfh_len = fhsize;
116220732Srmacklem	error = vfs_hash_get(mntp, hash, lkflags,
117191783Srmacklem	    td, &nvp, newnfs_vncmpf, nfhp);
118191783Srmacklem	FREE(nfhp, M_NFSFH);
119191783Srmacklem	if (error)
120191783Srmacklem		return (error);
121191783Srmacklem	if (nvp != NULL) {
122191783Srmacklem		*npp = VTONFS(nvp);
123191783Srmacklem		return (0);
124191783Srmacklem	}
125191783Srmacklem	np = uma_zalloc(newnfsnode_zone, M_WAITOK | M_ZERO);
126191783Srmacklem
127191783Srmacklem	error = getnewvnode("newnfs", mntp, &newnfs_vnodeops, &nvp);
128191783Srmacklem	if (error) {
129191783Srmacklem		uma_zfree(newnfsnode_zone, np);
130191783Srmacklem		return (error);
131191783Srmacklem	}
132191783Srmacklem	vp = nvp;
133230605Srmacklem	KASSERT(vp->v_bufobj.bo_bsize != 0, ("ncl_nget: bo_bsize == 0"));
134191783Srmacklem	vp->v_bufobj.bo_ops = &buf_ops_newnfs;
135191783Srmacklem	vp->v_data = np;
136191783Srmacklem	np->n_vnode = vp;
137191783Srmacklem	/*
138191783Srmacklem	 * Initialize the mutex even if the vnode is going to be a loser.
139191783Srmacklem	 * This simplifies the logic in reclaim, which can then unconditionally
140191783Srmacklem	 * destroy the mutex (in the case of the loser, or if hash_insert
141191783Srmacklem	 * happened to return an error no special casing is needed).
142191783Srmacklem	 */
143191783Srmacklem	mtx_init(&np->n_mtx, "NEWNFSnode lock", NULL, MTX_DEF | MTX_DUPOK);
144191783Srmacklem	/*
145191783Srmacklem	 * NFS supports recursive and shared locking.
146191783Srmacklem	 */
147211531Sjhb	lockmgr(vp->v_vnlock, LK_EXCLUSIVE | LK_NOWITNESS, NULL);
148191783Srmacklem	VN_LOCK_AREC(vp);
149191783Srmacklem	VN_LOCK_ASHARE(vp);
150191783Srmacklem	/*
151191783Srmacklem	 * Are we getting the root? If so, make sure the vnode flags
152191783Srmacklem	 * are correct
153191783Srmacklem	 */
154191783Srmacklem	if ((fhsize == nmp->nm_fhsize) &&
155191783Srmacklem	    !bcmp(fhp, nmp->nm_fh, fhsize)) {
156191783Srmacklem		if (vp->v_type == VNON)
157191783Srmacklem			vp->v_type = VDIR;
158191783Srmacklem		vp->v_vflag |= VV_ROOT;
159191783Srmacklem	}
160191783Srmacklem
161191783Srmacklem	MALLOC(np->n_fhp, struct nfsfh *, sizeof (struct nfsfh) + fhsize,
162191783Srmacklem	    M_NFSFH, M_WAITOK);
163191783Srmacklem	bcopy(fhp, np->n_fhp->nfh_fh, fhsize);
164191783Srmacklem	np->n_fhp->nfh_len = fhsize;
165191783Srmacklem	error = insmntque(vp, mntp);
166191783Srmacklem	if (error != 0) {
167191783Srmacklem		*npp = NULL;
168191783Srmacklem		FREE((caddr_t)np->n_fhp, M_NFSFH);
169191783Srmacklem		mtx_destroy(&np->n_mtx);
170191783Srmacklem		uma_zfree(newnfsnode_zone, np);
171191783Srmacklem		return (error);
172191783Srmacklem	}
173220732Srmacklem	error = vfs_hash_insert(vp, hash, lkflags,
174191783Srmacklem	    td, &nvp, newnfs_vncmpf, np->n_fhp);
175191783Srmacklem	if (error)
176191783Srmacklem		return (error);
177191783Srmacklem	if (nvp != NULL) {
178191783Srmacklem		*npp = VTONFS(nvp);
179191783Srmacklem		/* vfs_hash_insert() vput()'s the losing vnode */
180191783Srmacklem		return (0);
181191783Srmacklem	}
182191783Srmacklem	*npp = np;
183191783Srmacklem
184191783Srmacklem	return (0);
185191783Srmacklem}
186191783Srmacklem
187224606Srmacklem/*
188224606Srmacklem * Do the vrele(sp->s_dvp) as a separate task in order to avoid a
189224606Srmacklem * deadlock because of a LOR when vrele() locks the directory vnode.
190224606Srmacklem */
191224606Srmacklemstatic void
192224606Srmacklemnfs_freesillyrename(void *arg, __unused int pending)
193224606Srmacklem{
194224606Srmacklem	struct sillyrename *sp;
195224606Srmacklem
196224606Srmacklem	sp = arg;
197224606Srmacklem	vrele(sp->s_dvp);
198224606Srmacklem	free(sp, M_NEWNFSREQ);
199224606Srmacklem}
200224606Srmacklem
201191783Srmacklemint
202191783Srmacklemncl_inactive(struct vop_inactive_args *ap)
203191783Srmacklem{
204191783Srmacklem	struct nfsnode *np;
205191783Srmacklem	struct sillyrename *sp;
206193125Srmacklem	struct vnode *vp = ap->a_vp;
207237244Srmacklem	boolean_t retv;
208191783Srmacklem
209193125Srmacklem	np = VTONFS(vp);
210192337Srmacklem
211193125Srmacklem	if (NFS_ISV4(vp) && vp->v_type == VREG) {
212192928Srmacklem		/*
213192928Srmacklem		 * Since mmap()'d files do I/O after VOP_CLOSE(), the NFSv4
214237244Srmacklem		 * Close operations are delayed until now. Any dirty
215237244Srmacklem		 * buffers/pages must be flushed before the close, so that the
216237244Srmacklem		 * stateid is available for the writes.
217192928Srmacklem		 */
218237244Srmacklem		if (vp->v_object != NULL) {
219248084Sattilio			VM_OBJECT_WLOCK(vp->v_object);
220237244Srmacklem			retv = vm_object_page_clean(vp->v_object, 0, 0,
221237244Srmacklem			    OBJPC_SYNC);
222248084Sattilio			VM_OBJECT_WUNLOCK(vp->v_object);
223237244Srmacklem		} else
224237244Srmacklem			retv = TRUE;
225237244Srmacklem		if (retv == TRUE) {
226237244Srmacklem			(void)ncl_flush(vp, MNT_WAIT, NULL, ap->a_td, 1, 0);
227237244Srmacklem			(void)nfsrpc_close(vp, 1, ap->a_td);
228237244Srmacklem		}
229192928Srmacklem	}
230192337Srmacklem
231220731Srmacklem	mtx_lock(&np->n_mtx);
232193125Srmacklem	if (vp->v_type != VDIR) {
233191783Srmacklem		sp = np->n_sillyrename;
234191783Srmacklem		np->n_sillyrename = NULL;
235191783Srmacklem	} else
236191783Srmacklem		sp = NULL;
237191783Srmacklem	if (sp) {
238220731Srmacklem		mtx_unlock(&np->n_mtx);
239193125Srmacklem		(void) ncl_vinvalbuf(vp, 0, ap->a_td, 1);
240191783Srmacklem		/*
241191783Srmacklem		 * Remove the silly file that was rename'd earlier
242191783Srmacklem		 */
243193125Srmacklem		ncl_removeit(sp, vp);
244191783Srmacklem		crfree(sp->s_cred);
245224606Srmacklem		TASK_INIT(&sp->s_task, 0, nfs_freesillyrename, sp);
246224606Srmacklem		taskqueue_enqueue(taskqueue_thread, &sp->s_task);
247220731Srmacklem		mtx_lock(&np->n_mtx);
248191783Srmacklem	}
249191783Srmacklem	np->n_flag &= NMODIFIED;
250220731Srmacklem	mtx_unlock(&np->n_mtx);
251191783Srmacklem	return (0);
252191783Srmacklem}
253191783Srmacklem
254191783Srmacklem/*
255191783Srmacklem * Reclaim an nfsnode so that it can be used for other purposes.
256191783Srmacklem */
257191783Srmacklemint
258191783Srmacklemncl_reclaim(struct vop_reclaim_args *ap)
259191783Srmacklem{
260191783Srmacklem	struct vnode *vp = ap->a_vp;
261191783Srmacklem	struct nfsnode *np = VTONFS(vp);
262191783Srmacklem	struct nfsdmap *dp, *dp2;
263191783Srmacklem
264191783Srmacklem	/*
265191783Srmacklem	 * If the NLM is running, give it a chance to abort pending
266191783Srmacklem	 * locks.
267191783Srmacklem	 */
268214048Srmacklem	if (nfs_reclaim_p != NULL)
269214048Srmacklem		nfs_reclaim_p(ap);
270191783Srmacklem
271191783Srmacklem	/*
272191783Srmacklem	 * Destroy the vm object and flush associated pages.
273191783Srmacklem	 */
274191783Srmacklem	vnode_destroy_vobject(vp);
275191783Srmacklem
276237200Srmacklem	if (NFS_ISV4(vp) && vp->v_type == VREG)
277237200Srmacklem		/*
278237200Srmacklem		 * We can now safely close any remaining NFSv4 Opens for
279237200Srmacklem		 * this file. Most opens will have already been closed by
280237200Srmacklem		 * ncl_inactive(), but there are cases where it is not
281237200Srmacklem		 * called, so we need to do it again here.
282237200Srmacklem		 */
283237200Srmacklem		(void) nfsrpc_close(vp, 1, ap->a_td);
284237200Srmacklem
285191783Srmacklem	vfs_hash_remove(vp);
286191783Srmacklem
287191783Srmacklem	/*
288191783Srmacklem	 * Call nfscl_reclaimnode() to save attributes in the delegation,
289191783Srmacklem	 * as required.
290191783Srmacklem	 */
291191783Srmacklem	if (vp->v_type == VREG)
292191783Srmacklem		nfscl_reclaimnode(vp);
293191783Srmacklem
294191783Srmacklem	/*
295191783Srmacklem	 * Free up any directory cookie structures and
296191783Srmacklem	 * large file handle structures that might be associated with
297191783Srmacklem	 * this nfs node.
298191783Srmacklem	 */
299191783Srmacklem	if (vp->v_type == VDIR) {
300191783Srmacklem		dp = LIST_FIRST(&np->n_cookies);
301191783Srmacklem		while (dp) {
302191783Srmacklem			dp2 = dp;
303191783Srmacklem			dp = LIST_NEXT(dp, ndm_list);
304191783Srmacklem			FREE((caddr_t)dp2, M_NFSDIROFF);
305191783Srmacklem		}
306191783Srmacklem	}
307235332Srmacklem	if (np->n_writecred != NULL)
308235332Srmacklem		crfree(np->n_writecred);
309191783Srmacklem	FREE((caddr_t)np->n_fhp, M_NFSFH);
310191783Srmacklem	if (np->n_v4 != NULL)
311191783Srmacklem		FREE((caddr_t)np->n_v4, M_NFSV4NODE);
312191783Srmacklem	mtx_destroy(&np->n_mtx);
313191783Srmacklem	uma_zfree(newnfsnode_zone, vp->v_data);
314191783Srmacklem	vp->v_data = NULL;
315191783Srmacklem	return (0);
316191783Srmacklem}
317191783Srmacklem
318191783Srmacklem/*
319191783Srmacklem * Invalidate both the access and attribute caches for this vnode.
320191783Srmacklem */
321191783Srmacklemvoid
322191783Srmacklemncl_invalcaches(struct vnode *vp)
323191783Srmacklem{
324191783Srmacklem	struct nfsnode *np = VTONFS(vp);
325191783Srmacklem	int i;
326191783Srmacklem
327191783Srmacklem	mtx_lock(&np->n_mtx);
328191783Srmacklem	for (i = 0; i < NFS_ACCESSCACHESIZE; i++)
329191783Srmacklem		np->n_accesscache[i].stamp = 0;
330223280Srmacklem	KDTRACE_NFS_ACCESSCACHE_FLUSH_DONE(vp);
331191783Srmacklem	np->n_attrstamp = 0;
332223280Srmacklem	KDTRACE_NFS_ATTRCACHE_FLUSH_DONE(vp);
333191783Srmacklem	mtx_unlock(&np->n_mtx);
334191783Srmacklem}
335191783Srmacklem
336