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_subs.c  8.8 (Berkeley) 5/22/95
33 */
34
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD: stable/11/sys/fs/nfsclient/nfs_clsubs.c 323142 2017-09-03 09:12:02Z kib $");
37
38/*
39 * These functions support the macros and help fiddle mbuf chains for
40 * the nfs op functions. They do things like create the rpc header and
41 * copy data between mbuf chains and uio lists.
42 */
43
44#include <sys/param.h>
45#include <sys/systm.h>
46#include <sys/kernel.h>
47#include <sys/bio.h>
48#include <sys/buf.h>
49#include <sys/proc.h>
50#include <sys/mount.h>
51#include <sys/vnode.h>
52#include <sys/namei.h>
53#include <sys/mbuf.h>
54#include <sys/socket.h>
55#include <sys/stat.h>
56#include <sys/malloc.h>
57#include <sys/sysent.h>
58#include <sys/syscall.h>
59#include <sys/sysproto.h>
60#include <sys/taskqueue.h>
61
62#include <vm/vm.h>
63#include <vm/vm_object.h>
64#include <vm/vm_extern.h>
65#include <vm/uma.h>
66
67#include <fs/nfs/nfsport.h>
68#include <fs/nfsclient/nfsnode.h>
69#include <fs/nfsclient/nfsmount.h>
70#include <fs/nfsclient/nfs.h>
71#include <fs/nfsclient/nfs_kdtrace.h>
72
73#include <netinet/in.h>
74
75/*
76 * Note that stdarg.h and the ANSI style va_start macro is used for both
77 * ANSI and traditional C compilers.
78 */
79#include <machine/stdarg.h>
80
81extern struct mtx ncl_iod_mutex;
82extern enum nfsiod_state ncl_iodwant[NFS_MAXASYNCDAEMON];
83extern struct nfsmount *ncl_iodmount[NFS_MAXASYNCDAEMON];
84extern int ncl_numasync;
85extern unsigned int ncl_iodmax;
86extern struct nfsstatsv1 nfsstatsv1;
87
88struct task	ncl_nfsiodnew_task;
89
90int
91ncl_uninit(struct vfsconf *vfsp)
92{
93	/*
94	 * XXX: Unloading of nfscl module is unsupported.
95	 */
96#if 0
97	int i;
98
99	/*
100	 * Tell all nfsiod processes to exit. Clear ncl_iodmax, and wakeup
101	 * any sleeping nfsiods so they check ncl_iodmax and exit.
102	 */
103	mtx_lock(&ncl_iod_mutex);
104	ncl_iodmax = 0;
105	for (i = 0; i < ncl_numasync; i++)
106		if (ncl_iodwant[i] == NFSIOD_AVAILABLE)
107			wakeup(&ncl_iodwant[i]);
108	/* The last nfsiod to exit will wake us up when ncl_numasync hits 0 */
109	while (ncl_numasync)
110		msleep(&ncl_numasync, &ncl_iod_mutex, PWAIT, "ioddie", 0);
111	mtx_unlock(&ncl_iod_mutex);
112	ncl_nhuninit();
113	return (0);
114#else
115	return (EOPNOTSUPP);
116#endif
117}
118
119void
120ncl_dircookie_lock(struct nfsnode *np)
121{
122	mtx_lock(&np->n_mtx);
123	while (np->n_flag & NDIRCOOKIELK)
124		(void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0);
125	np->n_flag |= NDIRCOOKIELK;
126	mtx_unlock(&np->n_mtx);
127}
128
129void
130ncl_dircookie_unlock(struct nfsnode *np)
131{
132	mtx_lock(&np->n_mtx);
133	np->n_flag &= ~NDIRCOOKIELK;
134	wakeup(&np->n_flag);
135	mtx_unlock(&np->n_mtx);
136}
137
138bool
139ncl_excl_start(struct vnode *vp)
140{
141	struct nfsnode *np;
142	int vn_lk;
143
144	ASSERT_VOP_LOCKED(vp, "ncl_excl_start");
145	vn_lk = NFSVOPISLOCKED(vp);
146	if (vn_lk == LK_EXCLUSIVE)
147		return (false);
148	KASSERT(vn_lk == LK_SHARED,
149	    ("ncl_excl_start: wrong vnode lock %d", vn_lk));
150	/* Ensure exclusive access, this might block */
151	np = VTONFS(vp);
152	lockmgr(&np->n_excl, LK_EXCLUSIVE, NULL);
153	return (true);
154}
155
156void
157ncl_excl_finish(struct vnode *vp, bool old_lock)
158{
159	struct nfsnode *np;
160
161	if (!old_lock)
162		return;
163	np = VTONFS(vp);
164	lockmgr(&np->n_excl, LK_RELEASE, NULL);
165}
166
167#ifdef NFS_ACDEBUG
168#include <sys/sysctl.h>
169SYSCTL_DECL(_vfs_nfs);
170static int nfs_acdebug;
171SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
172#endif
173
174/*
175 * Check the time stamp
176 * If the cache is valid, copy contents to *vap and return 0
177 * otherwise return an error
178 */
179int
180ncl_getattrcache(struct vnode *vp, struct vattr *vaper)
181{
182	struct nfsnode *np;
183	struct vattr *vap;
184	struct nfsmount *nmp;
185	int timeo, mustflush;
186
187	np = VTONFS(vp);
188	vap = &np->n_vattr.na_vattr;
189	nmp = VFSTONFS(vp->v_mount);
190	mustflush = nfscl_mustflush(vp);	/* must be before mtx_lock() */
191	mtx_lock(&np->n_mtx);
192	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
193	timeo = (time_second - np->n_mtime.tv_sec) / 10;
194
195#ifdef NFS_ACDEBUG
196	if (nfs_acdebug>1)
197		printf("ncl_getattrcache: initial timeo = %d\n", timeo);
198#endif
199
200	if (vap->va_type == VDIR) {
201		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
202			timeo = nmp->nm_acdirmin;
203		else if (timeo > nmp->nm_acdirmax)
204			timeo = nmp->nm_acdirmax;
205	} else {
206		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
207			timeo = nmp->nm_acregmin;
208		else if (timeo > nmp->nm_acregmax)
209			timeo = nmp->nm_acregmax;
210	}
211
212#ifdef NFS_ACDEBUG
213	if (nfs_acdebug > 2)
214		printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
215		    nmp->nm_acregmin, nmp->nm_acregmax,
216		    nmp->nm_acdirmin, nmp->nm_acdirmax);
217
218	if (nfs_acdebug)
219		printf("ncl_getattrcache: age = %d; final timeo = %d\n",
220		    (time_second - np->n_attrstamp), timeo);
221#endif
222
223	if ((time_second - np->n_attrstamp) >= timeo &&
224	    (mustflush != 0 || np->n_attrstamp == 0)) {
225		nfsstatsv1.attrcache_misses++;
226		mtx_unlock(&np->n_mtx);
227		KDTRACE_NFS_ATTRCACHE_GET_MISS(vp);
228		return( ENOENT);
229	}
230	nfsstatsv1.attrcache_hits++;
231	if (vap->va_size != np->n_size) {
232		if (vap->va_type == VREG) {
233			if (np->n_flag & NMODIFIED) {
234				if (vap->va_size < np->n_size)
235					vap->va_size = np->n_size;
236				else
237					np->n_size = vap->va_size;
238			} else {
239				np->n_size = vap->va_size;
240			}
241			vnode_pager_setsize(vp, np->n_size);
242		} else {
243			np->n_size = vap->va_size;
244		}
245	}
246	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
247	if (np->n_flag & NCHG) {
248		if (np->n_flag & NACC)
249			vaper->va_atime = np->n_atim;
250		if (np->n_flag & NUPD)
251			vaper->va_mtime = np->n_mtim;
252	}
253	mtx_unlock(&np->n_mtx);
254	KDTRACE_NFS_ATTRCACHE_GET_HIT(vp, vap);
255	return (0);
256}
257
258static nfsuint64 nfs_nullcookie = { { 0, 0 } };
259/*
260 * This function finds the directory cookie that corresponds to the
261 * logical byte offset given.
262 */
263nfsuint64 *
264ncl_getcookie(struct nfsnode *np, off_t off, int add)
265{
266	struct nfsdmap *dp, *dp2;
267	int pos;
268	nfsuint64 *retval = NULL;
269
270	pos = (uoff_t)off / NFS_DIRBLKSIZ;
271	if (pos == 0 || off < 0) {
272		KASSERT(!add, ("nfs getcookie add at <= 0"));
273		return (&nfs_nullcookie);
274	}
275	pos--;
276	dp = LIST_FIRST(&np->n_cookies);
277	if (!dp) {
278		if (add) {
279			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
280				M_NFSDIROFF, M_WAITOK);
281			dp->ndm_eocookie = 0;
282			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
283		} else
284			goto out;
285	}
286	while (pos >= NFSNUMCOOKIES) {
287		pos -= NFSNUMCOOKIES;
288		if (LIST_NEXT(dp, ndm_list)) {
289			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
290			    pos >= dp->ndm_eocookie)
291				goto out;
292			dp = LIST_NEXT(dp, ndm_list);
293		} else if (add) {
294			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
295				M_NFSDIROFF, M_WAITOK);
296			dp2->ndm_eocookie = 0;
297			LIST_INSERT_AFTER(dp, dp2, ndm_list);
298			dp = dp2;
299		} else
300			goto out;
301	}
302	if (pos >= dp->ndm_eocookie) {
303		if (add)
304			dp->ndm_eocookie = pos + 1;
305		else
306			goto out;
307	}
308	retval = &dp->ndm_cookies[pos];
309out:
310	return (retval);
311}
312
313/*
314 * Invalidate cached directory information, except for the actual directory
315 * blocks (which are invalidated separately).
316 * Done mainly to avoid the use of stale offset cookies.
317 */
318void
319ncl_invaldir(struct vnode *vp)
320{
321	struct nfsnode *np = VTONFS(vp);
322
323	KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir"));
324	ncl_dircookie_lock(np);
325	np->n_direofoffset = 0;
326	np->n_cookieverf.nfsuquad[0] = 0;
327	np->n_cookieverf.nfsuquad[1] = 0;
328	if (LIST_FIRST(&np->n_cookies))
329		LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
330	ncl_dircookie_unlock(np);
331}
332
333/*
334 * The write verifier has changed (probably due to a server reboot), so all
335 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
336 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
337 * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
338 * mount point.
339 *
340 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
341 * writes are not clusterable.
342 */
343void
344ncl_clearcommit(struct mount *mp)
345{
346	struct vnode *vp, *nvp;
347	struct buf *bp, *nbp;
348	struct bufobj *bo;
349
350	MNT_VNODE_FOREACH_ALL(vp, mp, nvp) {
351		bo = &vp->v_bufobj;
352		vholdl(vp);
353		VI_UNLOCK(vp);
354		BO_LOCK(bo);
355		TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
356			if (!BUF_ISLOCKED(bp) &&
357			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
358				== (B_DELWRI | B_NEEDCOMMIT))
359				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
360		}
361		BO_UNLOCK(bo);
362		vdrop(vp);
363	}
364}
365
366/*
367 * Called once to initialize data structures...
368 */
369int
370ncl_init(struct vfsconf *vfsp)
371{
372	int i;
373
374	/* Ensure async daemons disabled */
375	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
376		ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE;
377		ncl_iodmount[i] = NULL;
378	}
379	TASK_INIT(&ncl_nfsiodnew_task, 0, ncl_nfsiodnew_tq, NULL);
380	ncl_nhinit();			/* Init the nfsnode table */
381
382	return (0);
383}
384
385