1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1989, 1993
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Rick Macklem at The University of Guelph.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 *	from nfs_subs.c  8.8 (Berkeley) 5/22/95
35 */
36
37#include <sys/cdefs.h>
38__FBSDID("$FreeBSD$");
39
40/*
41 * These functions support the macros and help fiddle mbuf chains for
42 * the nfs op functions. They do things like create the rpc header and
43 * copy data between mbuf chains and uio lists.
44 */
45
46#include <sys/param.h>
47#include <sys/systm.h>
48#include <sys/kernel.h>
49#include <sys/bio.h>
50#include <sys/buf.h>
51#include <sys/proc.h>
52#include <sys/mount.h>
53#include <sys/vnode.h>
54#include <sys/namei.h>
55#include <sys/mbuf.h>
56#include <sys/socket.h>
57#include <sys/stat.h>
58#include <sys/malloc.h>
59#include <sys/sysent.h>
60#include <sys/syscall.h>
61#include <sys/sysproto.h>
62#include <sys/taskqueue.h>
63
64#include <vm/vm.h>
65#include <vm/vm_object.h>
66#include <vm/vm_extern.h>
67#include <vm/uma.h>
68
69#include <fs/nfs/nfsport.h>
70#include <fs/nfsclient/nfsnode.h>
71#include <fs/nfsclient/nfsmount.h>
72#include <fs/nfsclient/nfs.h>
73#include <fs/nfsclient/nfs_kdtrace.h>
74
75#include <netinet/in.h>
76
77/*
78 * Note that stdarg.h and the ANSI style va_start macro is used for both
79 * ANSI and traditional C compilers.
80 */
81#include <machine/stdarg.h>
82
83extern struct mtx ncl_iod_mutex;
84extern enum nfsiod_state ncl_iodwant[NFS_MAXASYNCDAEMON];
85extern struct nfsmount *ncl_iodmount[NFS_MAXASYNCDAEMON];
86extern int ncl_numasync;
87extern unsigned int ncl_iodmax;
88extern struct nfsstatsv1 nfsstatsv1;
89
90struct task	ncl_nfsiodnew_task;
91
92int
93ncl_uninit(struct vfsconf *vfsp)
94{
95	/*
96	 * XXX: Unloading of nfscl module is unsupported.
97	 */
98#if 0
99	int i;
100
101	/*
102	 * Tell all nfsiod processes to exit. Clear ncl_iodmax, and wakeup
103	 * any sleeping nfsiods so they check ncl_iodmax and exit.
104	 */
105	NFSLOCKIOD();
106	ncl_iodmax = 0;
107	for (i = 0; i < ncl_numasync; i++)
108		if (ncl_iodwant[i] == NFSIOD_AVAILABLE)
109			wakeup(&ncl_iodwant[i]);
110	/* The last nfsiod to exit will wake us up when ncl_numasync hits 0 */
111	while (ncl_numasync)
112		msleep(&ncl_numasync, &ncl_iod_mutex, PWAIT, "ioddie", 0);
113	NFSUNLOCKIOD();
114	ncl_nhuninit();
115	return (0);
116#else
117	return (EOPNOTSUPP);
118#endif
119}
120
121/* Returns with NFSLOCKNODE() held. */
122void
123ncl_dircookie_lock(struct nfsnode *np)
124{
125	NFSLOCKNODE(np);
126	while (np->n_flag & NDIRCOOKIELK)
127		(void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0);
128	np->n_flag |= NDIRCOOKIELK;
129}
130
131void
132ncl_dircookie_unlock(struct nfsnode *np)
133{
134	NFSLOCKNODE(np);
135	np->n_flag &= ~NDIRCOOKIELK;
136	wakeup(&np->n_flag);
137	NFSUNLOCKNODE(np);
138}
139
140bool
141ncl_excl_start(struct vnode *vp)
142{
143	struct nfsnode *np;
144	int vn_lk;
145
146	ASSERT_VOP_LOCKED(vp, "ncl_excl_start");
147	vn_lk = NFSVOPISLOCKED(vp);
148	if (vn_lk == LK_EXCLUSIVE)
149		return (false);
150	KASSERT(vn_lk == LK_SHARED,
151	    ("ncl_excl_start: wrong vnode lock %d", vn_lk));
152	/* Ensure exclusive access, this might block */
153	np = VTONFS(vp);
154	lockmgr(&np->n_excl, LK_EXCLUSIVE, NULL);
155	return (true);
156}
157
158void
159ncl_excl_finish(struct vnode *vp, bool old_lock)
160{
161	struct nfsnode *np;
162
163	if (!old_lock)
164		return;
165	np = VTONFS(vp);
166	lockmgr(&np->n_excl, LK_RELEASE, NULL);
167}
168
169#ifdef NFS_ACDEBUG
170#include <sys/sysctl.h>
171SYSCTL_DECL(_vfs_nfs);
172static int nfs_acdebug;
173SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
174#endif
175
176/*
177 * Check the time stamp
178 * If the cache is valid, copy contents to *vap and return 0
179 * otherwise return an error
180 */
181int
182ncl_getattrcache(struct vnode *vp, struct vattr *vaper)
183{
184	struct nfsnode *np;
185	struct vattr *vap;
186	struct nfsmount *nmp;
187	int timeo, mustflush;
188	u_quad_t nsize;
189	bool setnsize;
190
191	np = VTONFS(vp);
192	vap = &np->n_vattr.na_vattr;
193	nmp = VFSTONFS(vp->v_mount);
194	mustflush = nfscl_mustflush(vp);	/* must be before mtx_lock() */
195	NFSLOCKNODE(np);
196	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
197	timeo = (time_second - np->n_mtime.tv_sec) / 10;
198
199#ifdef NFS_ACDEBUG
200	if (nfs_acdebug>1)
201		printf("ncl_getattrcache: initial timeo = %d\n", timeo);
202#endif
203
204	if (vap->va_type == VDIR) {
205		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
206			timeo = nmp->nm_acdirmin;
207		else if (timeo > nmp->nm_acdirmax)
208			timeo = nmp->nm_acdirmax;
209	} else {
210		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
211			timeo = nmp->nm_acregmin;
212		else if (timeo > nmp->nm_acregmax)
213			timeo = nmp->nm_acregmax;
214	}
215
216#ifdef NFS_ACDEBUG
217	if (nfs_acdebug > 2)
218		printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
219		    nmp->nm_acregmin, nmp->nm_acregmax,
220		    nmp->nm_acdirmin, nmp->nm_acdirmax);
221
222	if (nfs_acdebug)
223		printf("ncl_getattrcache: age = %d; final timeo = %d\n",
224		    (time_second - np->n_attrstamp), timeo);
225#endif
226
227	if ((time_second - np->n_attrstamp) >= timeo &&
228	    (mustflush != 0 || np->n_attrstamp == 0)) {
229		nfsstatsv1.attrcache_misses++;
230		NFSUNLOCKNODE(np);
231		KDTRACE_NFS_ATTRCACHE_GET_MISS(vp);
232		return( ENOENT);
233	}
234	nfsstatsv1.attrcache_hits++;
235	setnsize = false;
236	if (vap->va_size != np->n_size) {
237		if (vap->va_type == VREG) {
238			if (np->n_flag & NMODIFIED) {
239				if (vap->va_size < np->n_size)
240					vap->va_size = np->n_size;
241				else
242					np->n_size = vap->va_size;
243			} else {
244				np->n_size = vap->va_size;
245			}
246			setnsize = ncl_pager_setsize(vp, &nsize);
247		} else {
248			np->n_size = vap->va_size;
249		}
250	}
251	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
252	if (np->n_flag & NCHG) {
253		if (np->n_flag & NACC)
254			vaper->va_atime = np->n_atim;
255		if (np->n_flag & NUPD)
256			vaper->va_mtime = np->n_mtim;
257	}
258	NFSUNLOCKNODE(np);
259	if (setnsize)
260		vnode_pager_setsize(vp, nsize);
261	KDTRACE_NFS_ATTRCACHE_GET_HIT(vp, vap);
262	return (0);
263}
264
265static nfsuint64 nfs_nullcookie = { { 0, 0 } };
266/*
267 * This function finds the directory cookie that corresponds to the
268 * logical byte offset given.
269 */
270nfsuint64 *
271ncl_getcookie(struct nfsnode *np, off_t off, int add)
272{
273	struct nfsdmap *dp, *dp2;
274	int pos;
275	nfsuint64 *retval = NULL;
276
277	pos = (uoff_t)off / NFS_DIRBLKSIZ;
278	if (pos == 0 || off < 0) {
279		KASSERT(!add, ("nfs getcookie add at <= 0"));
280		return (&nfs_nullcookie);
281	}
282	pos--;
283	dp = LIST_FIRST(&np->n_cookies);
284	if (!dp) {
285		if (add) {
286			dp = malloc(sizeof (struct nfsdmap),
287				M_NFSDIROFF, M_WAITOK);
288			dp->ndm_eocookie = 0;
289			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
290		} else
291			goto out;
292	}
293	while (pos >= NFSNUMCOOKIES) {
294		pos -= NFSNUMCOOKIES;
295		if (LIST_NEXT(dp, ndm_list)) {
296			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
297			    pos >= dp->ndm_eocookie)
298				goto out;
299			dp = LIST_NEXT(dp, ndm_list);
300		} else if (add) {
301			dp2 = malloc(sizeof (struct nfsdmap),
302				M_NFSDIROFF, M_WAITOK);
303			dp2->ndm_eocookie = 0;
304			LIST_INSERT_AFTER(dp, dp2, ndm_list);
305			dp = dp2;
306		} else
307			goto out;
308	}
309	if (pos >= dp->ndm_eocookie) {
310		if (add)
311			dp->ndm_eocookie = pos + 1;
312		else
313			goto out;
314	}
315	retval = &dp->ndm_cookies[pos];
316out:
317	return (retval);
318}
319
320/*
321 * Invalidate cached directory information, except for the actual directory
322 * blocks (which are invalidated separately).
323 * Done mainly to avoid the use of stale offset cookies.
324 */
325void
326ncl_invaldir(struct vnode *vp)
327{
328	struct nfsnode *np = VTONFS(vp);
329
330	KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir"));
331	ncl_dircookie_lock(np);
332	np->n_direofoffset = 0;
333	NFSUNLOCKNODE(np);
334	np->n_cookieverf.nfsuquad[0] = 0;
335	np->n_cookieverf.nfsuquad[1] = 0;
336	if (LIST_FIRST(&np->n_cookies))
337		LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
338	ncl_dircookie_unlock(np);
339}
340
341/*
342 * The write verifier has changed (probably due to a server reboot), so all
343 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
344 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
345 * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
346 * mount point.
347 *
348 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
349 * writes are not clusterable.
350 */
351void
352ncl_clearcommit(struct mount *mp)
353{
354	struct vnode *vp, *nvp;
355	struct buf *bp, *nbp;
356	struct bufobj *bo;
357
358	MNT_VNODE_FOREACH_ALL(vp, mp, nvp) {
359		bo = &vp->v_bufobj;
360		vholdl(vp);
361		VI_UNLOCK(vp);
362		BO_LOCK(bo);
363		TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
364			if (!BUF_ISLOCKED(bp) &&
365			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
366				== (B_DELWRI | B_NEEDCOMMIT))
367				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
368		}
369		BO_UNLOCK(bo);
370		vdrop(vp);
371	}
372}
373
374/*
375 * Called once to initialize data structures...
376 */
377int
378ncl_init(struct vfsconf *vfsp)
379{
380	int i;
381
382	/* Ensure async daemons disabled */
383	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
384		ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE;
385		ncl_iodmount[i] = NULL;
386	}
387	TASK_INIT(&ncl_nfsiodnew_task, 0, ncl_nfsiodnew_tq, NULL);
388	ncl_nhinit();			/* Init the nfsnode table */
389
390	return (0);
391}
392