nfs_clsubs.c revision 221973 
1235368Sgnn/*-
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: head/sys/fs/nfsclient/nfs_clsubs.c 221973 2011-05-15 20:52:43Z rmacklem $");
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
72#include <netinet/in.h>
73
74/*
75 * Note that stdarg.h and the ANSI style va_start macro is used for both
76 * ANSI and traditional C compilers.
77 */
78#include <machine/stdarg.h>
79
80extern struct mtx ncl_iod_mutex;
81extern enum nfsiod_state ncl_iodwant[NFS_MAXASYNCDAEMON];
82extern struct nfsmount *ncl_iodmount[NFS_MAXASYNCDAEMON];
83extern int ncl_numasync;
84extern unsigned int ncl_iodmax;
85extern struct nfsstats newnfsstats;
86
87struct task	ncl_nfsiodnew_task;
88
89int
90ncl_uninit(struct vfsconf *vfsp)
91{
92	/*
93	 * XXX: Unloading of nfscl module is unsupported.
94	 */
95#if 0
96	int i;
97
98	/*
99	 * Tell all nfsiod processes to exit. Clear ncl_iodmax, and wakeup
100	 * any sleeping nfsiods so they check ncl_iodmax and exit.
101	 */
102	mtx_lock(&ncl_iod_mutex);
103	ncl_iodmax = 0;
104	for (i = 0; i < ncl_numasync; i++)
105		if (ncl_iodwant[i] == NFSIOD_AVAILABLE)
106			wakeup(&ncl_iodwant[i]);
107	/* The last nfsiod to exit will wake us up when ncl_numasync hits 0 */
108	while (ncl_numasync)
109		msleep(&ncl_numasync, &ncl_iod_mutex, PWAIT, "ioddie", 0);
110	mtx_unlock(&ncl_iod_mutex);
111	ncl_nhuninit();
112	return (0);
113#else
114	return (EOPNOTSUPP);
115#endif
116}
117
118void
119ncl_dircookie_lock(struct nfsnode *np)
120{
121	mtx_lock(&np->n_mtx);
122	while (np->n_flag & NDIRCOOKIELK)
123		(void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0);
124	np->n_flag |= NDIRCOOKIELK;
125	mtx_unlock(&np->n_mtx);
126}
127
128void
129ncl_dircookie_unlock(struct nfsnode *np)
130{
131	mtx_lock(&np->n_mtx);
132	np->n_flag &= ~NDIRCOOKIELK;
133	wakeup(&np->n_flag);
134	mtx_unlock(&np->n_mtx);
135}
136
137int
138ncl_upgrade_vnlock(struct vnode *vp)
139{
140	int old_lock;
141
142	ASSERT_VOP_LOCKED(vp, "ncl_upgrade_vnlock");
143	old_lock = VOP_ISLOCKED(vp);
144	if (old_lock != LK_EXCLUSIVE) {
145		KASSERT(old_lock == LK_SHARED,
146		    ("ncl_upgrade_vnlock: wrong old_lock %d", old_lock));
147		/* Upgrade to exclusive lock, this might block */
148		vn_lock(vp, LK_UPGRADE | LK_RETRY);
149  	}
150	return (old_lock);
151}
152
153void
154ncl_downgrade_vnlock(struct vnode *vp, int old_lock)
155{
156	if (old_lock != LK_EXCLUSIVE) {
157		KASSERT(old_lock == LK_SHARED, ("wrong old_lock %d", old_lock));
158		/* Downgrade from exclusive lock. */
159		vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
160  	}
161}
162
163void
164ncl_printf(const char *fmt, ...)
165{
166	va_list ap;
167
168	mtx_lock(&Giant);
169	va_start(ap, fmt);
170	vprintf(fmt, ap);
171	va_end(ap);
172	mtx_unlock(&Giant);
173}
174
175#ifdef NFS_ACDEBUG
176#include <sys/sysctl.h>
177SYSCTL_DECL(_vfs_nfs);
178static int nfs_acdebug;
179SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
180#endif
181
182/*
183 * Check the time stamp
184 * If the cache is valid, copy contents to *vap and return 0
185 * otherwise return an error
186 */
187int
188ncl_getattrcache(struct vnode *vp, struct vattr *vaper)
189{
190	struct nfsnode *np;
191	struct vattr *vap;
192	struct nfsmount *nmp;
193	int timeo, mustflush;
194
195	np = VTONFS(vp);
196	vap = &np->n_vattr.na_vattr;
197	nmp = VFSTONFS(vp->v_mount);
198	mustflush = nfscl_mustflush(vp);	/* must be before mtx_lock() */
199#ifdef NFS_ACDEBUG
200	mtx_lock(&Giant);	/* ncl_printf() */
201#endif
202	mtx_lock(&np->n_mtx);
203	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
204	timeo = (time_second - np->n_mtime.tv_sec) / 10;
205
206#ifdef NFS_ACDEBUG
207	if (nfs_acdebug>1)
208		ncl_printf("nfs_getattrcache: initial timeo = %d\n", timeo);
209#endif
210
211	if (vap->va_type == VDIR) {
212		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
213			timeo = nmp->nm_acdirmin;
214		else if (timeo > nmp->nm_acdirmax)
215			timeo = nmp->nm_acdirmax;
216	} else {
217		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
218			timeo = nmp->nm_acregmin;
219		else if (timeo > nmp->nm_acregmax)
220			timeo = nmp->nm_acregmax;
221	}
222
223#ifdef NFS_ACDEBUG
224	if (nfs_acdebug > 2)
225		ncl_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
226			   nmp->nm_acregmin, nmp->nm_acregmax,
227			   nmp->nm_acdirmin, nmp->nm_acdirmax);
228
229	if (nfs_acdebug)
230		ncl_printf("nfs_getattrcache: age = %d; final timeo = %d\n",
231			   (time_second - np->n_attrstamp), timeo);
232#endif
233
234	if ((time_second - np->n_attrstamp) >= timeo &&
235	    (mustflush != 0 || np->n_attrstamp == 0)) {
236		newnfsstats.attrcache_misses++;
237		mtx_unlock(&np->n_mtx);
238#ifdef NFS_ACDEBUG
239		mtx_unlock(&Giant);	/* ncl_printf() */
240#endif
241		return( ENOENT);
242	}
243	newnfsstats.attrcache_hits++;
244	if (vap->va_size != np->n_size) {
245		if (vap->va_type == VREG) {
246			if (np->n_flag & NMODIFIED) {
247				if (vap->va_size < np->n_size)
248					vap->va_size = np->n_size;
249				else
250					np->n_size = vap->va_size;
251			} else {
252				np->n_size = vap->va_size;
253			}
254			vnode_pager_setsize(vp, np->n_size);
255		} else {
256			np->n_size = vap->va_size;
257		}
258	}
259	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
260	if (np->n_flag & NCHG) {
261		if (np->n_flag & NACC)
262			vaper->va_atime = np->n_atim;
263		if (np->n_flag & NUPD)
264			vaper->va_mtime = np->n_mtim;
265	}
266	mtx_unlock(&np->n_mtx);
267#ifdef NFS_ACDEBUG
268	mtx_unlock(&Giant);	/* ncl_printf() */
269#endif
270	return (0);
271}
272
273static nfsuint64 nfs_nullcookie = { { 0, 0 } };
274/*
275 * This function finds the directory cookie that corresponds to the
276 * logical byte offset given.
277 */
278nfsuint64 *
279ncl_getcookie(struct nfsnode *np, off_t off, int add)
280{
281	struct nfsdmap *dp, *dp2;
282	int pos;
283	nfsuint64 *retval = NULL;
284
285	pos = (uoff_t)off / NFS_DIRBLKSIZ;
286	if (pos == 0 || off < 0) {
287		KASSERT(!add, ("nfs getcookie add at <= 0"));
288		return (&nfs_nullcookie);
289	}
290	pos--;
291	dp = LIST_FIRST(&np->n_cookies);
292	if (!dp) {
293		if (add) {
294			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
295				M_NFSDIROFF, M_WAITOK);
296			dp->ndm_eocookie = 0;
297			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
298		} else
299			goto out;
300	}
301	while (pos >= NFSNUMCOOKIES) {
302		pos -= NFSNUMCOOKIES;
303		if (LIST_NEXT(dp, ndm_list)) {
304			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
305			    pos >= dp->ndm_eocookie)
306				goto out;
307			dp = LIST_NEXT(dp, ndm_list);
308		} else if (add) {
309			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
310				M_NFSDIROFF, M_WAITOK);
311			dp2->ndm_eocookie = 0;
312			LIST_INSERT_AFTER(dp, dp2, ndm_list);
313			dp = dp2;
314		} else
315			goto out;
316	}
317	if (pos >= dp->ndm_eocookie) {
318		if (add)
319			dp->ndm_eocookie = pos + 1;
320		else
321			goto out;
322	}
323	retval = &dp->ndm_cookies[pos];
324out:
325	return (retval);
326}
327
328/*
329 * Invalidate cached directory information, except for the actual directory
330 * blocks (which are invalidated separately).
331 * Done mainly to avoid the use of stale offset cookies.
332 */
333void
334ncl_invaldir(struct vnode *vp)
335{
336	struct nfsnode *np = VTONFS(vp);
337
338	KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir"));
339	ncl_dircookie_lock(np);
340	np->n_direofoffset = 0;
341	np->n_cookieverf.nfsuquad[0] = 0;
342	np->n_cookieverf.nfsuquad[1] = 0;
343	if (LIST_FIRST(&np->n_cookies))
344		LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
345	ncl_dircookie_unlock(np);
346}
347
348/*
349 * The write verifier has changed (probably due to a server reboot), so all
350 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
351 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
352 * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
353 * mount point.
354 *
355 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
356 * writes are not clusterable.
357 */
358void
359ncl_clearcommit(struct mount *mp)
360{
361	struct vnode *vp, *nvp;
362	struct buf *bp, *nbp;
363	struct bufobj *bo;
364
365	MNT_ILOCK(mp);
366	MNT_VNODE_FOREACH(vp, mp, nvp) {
367		bo = &vp->v_bufobj;
368		VI_LOCK(vp);
369		if (vp->v_iflag & VI_DOOMED) {
370			VI_UNLOCK(vp);
371			continue;
372		}
373		vholdl(vp);
374		VI_UNLOCK(vp);
375		MNT_IUNLOCK(mp);
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		MNT_ILOCK(mp);
386	}
387	MNT_IUNLOCK(mp);
388}
389
390/*
391 * Called once to initialize data structures...
392 */
393int
394ncl_init(struct vfsconf *vfsp)
395{
396	int i;
397
398	/* Ensure async daemons disabled */
399	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
400		ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE;
401		ncl_iodmount[i] = NULL;
402	}
403	TASK_INIT(&ncl_nfsiodnew_task, 0, ncl_nfsiodnew_tq, NULL);
404	ncl_nhinit();			/* Init the nfsnode table */
405
406	return (0);
407}
408
409