xref: /netbsd-src/sys/fs/nfs/client/nfs_clsubs.c (revision 946379e7b37692fc43f68eb0d1c10daa0a7f3b6c)
1 /*	$NetBSD: nfs_clsubs.c,v 1.1.1.1 2013/09/30 07:18:58 dholland Exp $	*/
2 /*-
3  * Copyright (c) 1989, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Rick Macklem at The University of Guelph.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	from nfs_subs.c  8.8 (Berkeley) 5/22/95
34  */
35 
36 #include <sys/cdefs.h>
37 /* __FBSDID("FreeBSD: head/sys/fs/nfsclient/nfs_clsubs.c 234386 2012-04-17 16:28:22Z mckusick "); */
38 __RCSID("$NetBSD: nfs_clsubs.c,v 1.1.1.1 2013/09/30 07:18:58 dholland Exp $");
39 
40 #include "opt_kdtrace.h"
41 
42 /*
43  * These functions support the macros and help fiddle mbuf chains for
44  * the nfs op functions. They do things like create the rpc header and
45  * copy data between mbuf chains and uio lists.
46  */
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/kernel.h>
51 #include <sys/bio.h>
52 #include <sys/buf.h>
53 #include <sys/proc.h>
54 #include <sys/mount.h>
55 #include <sys/vnode.h>
56 #include <sys/namei.h>
57 #include <sys/mbuf.h>
58 #include <sys/socket.h>
59 #include <sys/stat.h>
60 #include <sys/malloc.h>
61 #include <sys/sysent.h>
62 #include <sys/syscall.h>
63 #include <sys/sysproto.h>
64 #include <sys/taskqueue.h>
65 
66 #include <vm/vm.h>
67 #include <vm/vm_object.h>
68 #include <vm/vm_extern.h>
69 #include <vm/uma.h>
70 
71 #include <fs/nfs/nfsport.h>
72 #include <fs/nfsclient/nfsnode.h>
73 #include <fs/nfsclient/nfsmount.h>
74 #include <fs/nfsclient/nfs.h>
75 #include <fs/nfsclient/nfs_kdtrace.h>
76 
77 #include <netinet/in.h>
78 
79 /*
80  * Note that stdarg.h and the ANSI style va_start macro is used for both
81  * ANSI and traditional C compilers.
82  */
83 #include <machine/stdarg.h>
84 
85 extern struct mtx ncl_iod_mutex;
86 extern enum nfsiod_state ncl_iodwant[NFS_MAXASYNCDAEMON];
87 extern struct nfsmount *ncl_iodmount[NFS_MAXASYNCDAEMON];
88 extern int ncl_numasync;
89 extern unsigned int ncl_iodmax;
90 extern struct nfsstats newnfsstats;
91 
92 struct task	ncl_nfsiodnew_task;
93 
94 int
95 ncl_uninit(struct vfsconf *vfsp)
96 {
97 	/*
98 	 * XXX: Unloading of nfscl module is unsupported.
99 	 */
100 #if 0
101 	int i;
102 
103 	/*
104 	 * Tell all nfsiod processes to exit. Clear ncl_iodmax, and wakeup
105 	 * any sleeping nfsiods so they check ncl_iodmax and exit.
106 	 */
107 	mtx_lock(&ncl_iod_mutex);
108 	ncl_iodmax = 0;
109 	for (i = 0; i < ncl_numasync; i++)
110 		if (ncl_iodwant[i] == NFSIOD_AVAILABLE)
111 			wakeup(&ncl_iodwant[i]);
112 	/* The last nfsiod to exit will wake us up when ncl_numasync hits 0 */
113 	while (ncl_numasync)
114 		msleep(&ncl_numasync, &ncl_iod_mutex, PWAIT, "ioddie", 0);
115 	mtx_unlock(&ncl_iod_mutex);
116 	ncl_nhuninit();
117 	return (0);
118 #else
119 	return (EOPNOTSUPP);
120 #endif
121 }
122 
123 void
124 ncl_dircookie_lock(struct nfsnode *np)
125 {
126 	mtx_lock(&np->n_mtx);
127 	while (np->n_flag & NDIRCOOKIELK)
128 		(void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0);
129 	np->n_flag |= NDIRCOOKIELK;
130 	mtx_unlock(&np->n_mtx);
131 }
132 
133 void
134 ncl_dircookie_unlock(struct nfsnode *np)
135 {
136 	mtx_lock(&np->n_mtx);
137 	np->n_flag &= ~NDIRCOOKIELK;
138 	wakeup(&np->n_flag);
139 	mtx_unlock(&np->n_mtx);
140 }
141 
142 int
143 ncl_upgrade_vnlock(struct vnode *vp)
144 {
145 	int old_lock;
146 
147 	ASSERT_VOP_LOCKED(vp, "ncl_upgrade_vnlock");
148 	old_lock = NFSVOPISLOCKED(vp);
149 	if (old_lock != LK_EXCLUSIVE) {
150 		KASSERT(old_lock == LK_SHARED,
151 		    ("ncl_upgrade_vnlock: wrong old_lock %d", old_lock));
152 		/* Upgrade to exclusive lock, this might block */
153 		NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY);
154   	}
155 	return (old_lock);
156 }
157 
158 void
159 ncl_downgrade_vnlock(struct vnode *vp, int old_lock)
160 {
161 	if (old_lock != LK_EXCLUSIVE) {
162 		KASSERT(old_lock == LK_SHARED, ("wrong old_lock %d", old_lock));
163 		/* Downgrade from exclusive lock. */
164 		NFSVOPLOCK(vp, LK_DOWNGRADE | LK_RETRY);
165   	}
166 }
167 
168 void
169 ncl_printf(const char *fmt, ...)
170 {
171 	va_list ap;
172 
173 	mtx_lock(&Giant);
174 	va_start(ap, fmt);
175 	vprintf(fmt, ap);
176 	va_end(ap);
177 	mtx_unlock(&Giant);
178 }
179 
180 #ifdef NFS_ACDEBUG
181 #include <sys/sysctl.h>
182 SYSCTL_DECL(_vfs_nfs);
183 static int nfs_acdebug;
184 SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
185 #endif
186 
187 /*
188  * Check the time stamp
189  * If the cache is valid, copy contents to *vap and return 0
190  * otherwise return an error
191  */
192 int
193 ncl_getattrcache(struct vnode *vp, struct vattr *vaper)
194 {
195 	struct nfsnode *np;
196 	struct vattr *vap;
197 	struct nfsmount *nmp;
198 	int timeo, mustflush;
199 
200 	np = VTONFS(vp);
201 	vap = &np->n_vattr.na_vattr;
202 	nmp = VFSTONFS(vp->v_mount);
203 	mustflush = nfscl_mustflush(vp);	/* must be before mtx_lock() */
204 #ifdef NFS_ACDEBUG
205 	mtx_lock(&Giant);	/* ncl_printf() */
206 #endif
207 	mtx_lock(&np->n_mtx);
208 	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
209 	timeo = (time_second - np->n_mtime.tv_sec) / 10;
210 
211 #ifdef NFS_ACDEBUG
212 	if (nfs_acdebug>1)
213 		ncl_printf("nfs_getattrcache: initial timeo = %d\n", timeo);
214 #endif
215 
216 	if (vap->va_type == VDIR) {
217 		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
218 			timeo = nmp->nm_acdirmin;
219 		else if (timeo > nmp->nm_acdirmax)
220 			timeo = nmp->nm_acdirmax;
221 	} else {
222 		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
223 			timeo = nmp->nm_acregmin;
224 		else if (timeo > nmp->nm_acregmax)
225 			timeo = nmp->nm_acregmax;
226 	}
227 
228 #ifdef NFS_ACDEBUG
229 	if (nfs_acdebug > 2)
230 		ncl_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
231 			   nmp->nm_acregmin, nmp->nm_acregmax,
232 			   nmp->nm_acdirmin, nmp->nm_acdirmax);
233 
234 	if (nfs_acdebug)
235 		ncl_printf("nfs_getattrcache: age = %d; final timeo = %d\n",
236 			   (time_second - np->n_attrstamp), timeo);
237 #endif
238 
239 	if ((time_second - np->n_attrstamp) >= timeo &&
240 	    (mustflush != 0 || np->n_attrstamp == 0)) {
241 		newnfsstats.attrcache_misses++;
242 		mtx_unlock(&np->n_mtx);
243 #ifdef NFS_ACDEBUG
244 		mtx_unlock(&Giant);	/* ncl_printf() */
245 #endif
246 		KDTRACE_NFS_ATTRCACHE_GET_MISS(vp);
247 		return( ENOENT);
248 	}
249 	newnfsstats.attrcache_hits++;
250 	if (vap->va_size != np->n_size) {
251 		if (vap->va_type == VREG) {
252 			if (np->n_flag & NMODIFIED) {
253 				if (vap->va_size < np->n_size)
254 					vap->va_size = np->n_size;
255 				else
256 					np->n_size = vap->va_size;
257 			} else {
258 				np->n_size = vap->va_size;
259 			}
260 			vnode_pager_setsize(vp, np->n_size);
261 		} else {
262 			np->n_size = vap->va_size;
263 		}
264 	}
265 	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
266 	if (np->n_flag & NCHG) {
267 		if (np->n_flag & NACC)
268 			vaper->va_atime = np->n_atim;
269 		if (np->n_flag & NUPD)
270 			vaper->va_mtime = np->n_mtim;
271 	}
272 	mtx_unlock(&np->n_mtx);
273 #ifdef NFS_ACDEBUG
274 	mtx_unlock(&Giant);	/* ncl_printf() */
275 #endif
276 	KDTRACE_NFS_ATTRCACHE_GET_HIT(vp, vap);
277 	return (0);
278 }
279 
280 static nfsuint64 nfs_nullcookie = { { 0, 0 } };
281 /*
282  * This function finds the directory cookie that corresponds to the
283  * logical byte offset given.
284  */
285 nfsuint64 *
286 ncl_getcookie(struct nfsnode *np, off_t off, int add)
287 {
288 	struct nfsdmap *dp, *dp2;
289 	int pos;
290 	nfsuint64 *retval = NULL;
291 
292 	pos = (uoff_t)off / NFS_DIRBLKSIZ;
293 	if (pos == 0 || off < 0) {
294 		KASSERT(!add, ("nfs getcookie add at <= 0"));
295 		return (&nfs_nullcookie);
296 	}
297 	pos--;
298 	dp = LIST_FIRST(&np->n_cookies);
299 	if (!dp) {
300 		if (add) {
301 			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
302 				M_NFSDIROFF, M_WAITOK);
303 			dp->ndm_eocookie = 0;
304 			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
305 		} else
306 			goto out;
307 	}
308 	while (pos >= NFSNUMCOOKIES) {
309 		pos -= NFSNUMCOOKIES;
310 		if (LIST_NEXT(dp, ndm_list)) {
311 			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
312 			    pos >= dp->ndm_eocookie)
313 				goto out;
314 			dp = LIST_NEXT(dp, ndm_list);
315 		} else if (add) {
316 			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
317 				M_NFSDIROFF, M_WAITOK);
318 			dp2->ndm_eocookie = 0;
319 			LIST_INSERT_AFTER(dp, dp2, ndm_list);
320 			dp = dp2;
321 		} else
322 			goto out;
323 	}
324 	if (pos >= dp->ndm_eocookie) {
325 		if (add)
326 			dp->ndm_eocookie = pos + 1;
327 		else
328 			goto out;
329 	}
330 	retval = &dp->ndm_cookies[pos];
331 out:
332 	return (retval);
333 }
334 
335 /*
336  * Invalidate cached directory information, except for the actual directory
337  * blocks (which are invalidated separately).
338  * Done mainly to avoid the use of stale offset cookies.
339  */
340 void
341 ncl_invaldir(struct vnode *vp)
342 {
343 	struct nfsnode *np = VTONFS(vp);
344 
345 	KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir"));
346 	ncl_dircookie_lock(np);
347 	np->n_direofoffset = 0;
348 	np->n_cookieverf.nfsuquad[0] = 0;
349 	np->n_cookieverf.nfsuquad[1] = 0;
350 	if (LIST_FIRST(&np->n_cookies))
351 		LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
352 	ncl_dircookie_unlock(np);
353 }
354 
355 /*
356  * The write verifier has changed (probably due to a server reboot), so all
357  * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
358  * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
359  * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
360  * mount point.
361  *
362  * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
363  * writes are not clusterable.
364  */
365 void
366 ncl_clearcommit(struct mount *mp)
367 {
368 	struct vnode *vp, *nvp;
369 	struct buf *bp, *nbp;
370 	struct bufobj *bo;
371 
372 	MNT_VNODE_FOREACH_ALL(vp, mp, nvp) {
373 		bo = &vp->v_bufobj;
374 		vholdl(vp);
375 		VI_UNLOCK(vp);
376 		BO_LOCK(bo);
377 		TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
378 			if (!BUF_ISLOCKED(bp) &&
379 			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
380 				== (B_DELWRI | B_NEEDCOMMIT))
381 				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
382 		}
383 		BO_UNLOCK(bo);
384 		vdrop(vp);
385 	}
386 }
387 
388 /*
389  * Called once to initialize data structures...
390  */
391 int
392 ncl_init(struct vfsconf *vfsp)
393 {
394 	int i;
395 
396 	/* Ensure async daemons disabled */
397 	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
398 		ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE;
399 		ncl_iodmount[i] = NULL;
400 	}
401 	TASK_INIT(&ncl_nfsiodnew_task, 0, ncl_nfsiodnew_tq, NULL);
402 	ncl_nhinit();			/* Init the nfsnode table */
403 
404 	return (0);
405 }
406 
407