nfs_clsubs.c revision 258541 
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: head/sys/fs/nfsclient/nfs_clsubs.c 258541 2013-11-25 07:38:45Z attilio $");
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 nfsstats newnfsstats;
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
138int
139ncl_upgrade_vnlock(struct vnode *vp)
140{
141	int old_lock;
142
143	ASSERT_VOP_LOCKED(vp, "ncl_upgrade_vnlock");
144	old_lock = NFSVOPISLOCKED(vp);
145	if (old_lock != LK_EXCLUSIVE) {
146		KASSERT(old_lock == LK_SHARED,
147		    ("ncl_upgrade_vnlock: wrong old_lock %d", old_lock));
148		/* Upgrade to exclusive lock, this might block */
149		NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY);
150  	}
151	return (old_lock);
152}
153
154void
155ncl_downgrade_vnlock(struct vnode *vp, int old_lock)
156{
157	if (old_lock != LK_EXCLUSIVE) {
158		KASSERT(old_lock == LK_SHARED, ("wrong old_lock %d", old_lock));
159		/* Downgrade from exclusive lock. */
160		NFSVOPLOCK(vp, LK_DOWNGRADE | LK_RETRY);
161  	}
162}
163
164void
165ncl_printf(const char *fmt, ...)
166{
167	va_list ap;
168
169	mtx_lock(&Giant);
170	va_start(ap, fmt);
171	vprintf(fmt, ap);
172	va_end(ap);
173	mtx_unlock(&Giant);
174}
175
176#ifdef NFS_ACDEBUG
177#include <sys/sysctl.h>
178SYSCTL_DECL(_vfs_nfs);
179static int nfs_acdebug;
180SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
181#endif
182
183/*
184 * Check the time stamp
185 * If the cache is valid, copy contents to *vap and return 0
186 * otherwise return an error
187 */
188int
189ncl_getattrcache(struct vnode *vp, struct vattr *vaper)
190{
191	struct nfsnode *np;
192	struct vattr *vap;
193	struct nfsmount *nmp;
194	int timeo, mustflush;
195
196	np = VTONFS(vp);
197	vap = &np->n_vattr.na_vattr;
198	nmp = VFSTONFS(vp->v_mount);
199	mustflush = nfscl_mustflush(vp);	/* must be before mtx_lock() */
200#ifdef NFS_ACDEBUG
201	mtx_lock(&Giant);	/* ncl_printf() */
202#endif
203	mtx_lock(&np->n_mtx);
204	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
205	timeo = (time_second - np->n_mtime.tv_sec) / 10;
206
207#ifdef NFS_ACDEBUG
208	if (nfs_acdebug>1)
209		ncl_printf("nfs_getattrcache: initial timeo = %d\n", timeo);
210#endif
211
212	if (vap->va_type == VDIR) {
213		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
214			timeo = nmp->nm_acdirmin;
215		else if (timeo > nmp->nm_acdirmax)
216			timeo = nmp->nm_acdirmax;
217	} else {
218		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
219			timeo = nmp->nm_acregmin;
220		else if (timeo > nmp->nm_acregmax)
221			timeo = nmp->nm_acregmax;
222	}
223
224#ifdef NFS_ACDEBUG
225	if (nfs_acdebug > 2)
226		ncl_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
227			   nmp->nm_acregmin, nmp->nm_acregmax,
228			   nmp->nm_acdirmin, nmp->nm_acdirmax);
229
230	if (nfs_acdebug)
231		ncl_printf("nfs_getattrcache: age = %d; final timeo = %d\n",
232			   (time_second - np->n_attrstamp), timeo);
233#endif
234
235	if ((time_second - np->n_attrstamp) >= timeo &&
236	    (mustflush != 0 || np->n_attrstamp == 0)) {
237		newnfsstats.attrcache_misses++;
238		mtx_unlock(&np->n_mtx);
239#ifdef NFS_ACDEBUG
240		mtx_unlock(&Giant);	/* ncl_printf() */
241#endif
242		KDTRACE_NFS_ATTRCACHE_GET_MISS(vp);
243		return( ENOENT);
244	}
245	newnfsstats.attrcache_hits++;
246	if (vap->va_size != np->n_size) {
247		if (vap->va_type == VREG) {
248			if (np->n_flag & NMODIFIED) {
249				if (vap->va_size < np->n_size)
250					vap->va_size = np->n_size;
251				else
252					np->n_size = vap->va_size;
253			} else {
254				np->n_size = vap->va_size;
255			}
256			vnode_pager_setsize(vp, np->n_size);
257		} else {
258			np->n_size = vap->va_size;
259		}
260	}
261	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
262	if (np->n_flag & NCHG) {
263		if (np->n_flag & NACC)
264			vaper->va_atime = np->n_atim;
265		if (np->n_flag & NUPD)
266			vaper->va_mtime = np->n_mtim;
267	}
268	mtx_unlock(&np->n_mtx);
269#ifdef NFS_ACDEBUG
270	mtx_unlock(&Giant);	/* ncl_printf() */
271#endif
272	KDTRACE_NFS_ATTRCACHE_GET_HIT(vp, vap);
273	return (0);
274}
275
276static nfsuint64 nfs_nullcookie = { { 0, 0 } };
277/*
278 * This function finds the directory cookie that corresponds to the
279 * logical byte offset given.
280 */
281nfsuint64 *
282ncl_getcookie(struct nfsnode *np, off_t off, int add)
283{
284	struct nfsdmap *dp, *dp2;
285	int pos;
286	nfsuint64 *retval = NULL;
287
288	pos = (uoff_t)off / NFS_DIRBLKSIZ;
289	if (pos == 0 || off < 0) {
290		KASSERT(!add, ("nfs getcookie add at <= 0"));
291		return (&nfs_nullcookie);
292	}
293	pos--;
294	dp = LIST_FIRST(&np->n_cookies);
295	if (!dp) {
296		if (add) {
297			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
298				M_NFSDIROFF, M_WAITOK);
299			dp->ndm_eocookie = 0;
300			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
301		} else
302			goto out;
303	}
304	while (pos >= NFSNUMCOOKIES) {
305		pos -= NFSNUMCOOKIES;
306		if (LIST_NEXT(dp, ndm_list)) {
307			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
308			    pos >= dp->ndm_eocookie)
309				goto out;
310			dp = LIST_NEXT(dp, ndm_list);
311		} else if (add) {
312			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
313				M_NFSDIROFF, M_WAITOK);
314			dp2->ndm_eocookie = 0;
315			LIST_INSERT_AFTER(dp, dp2, ndm_list);
316			dp = dp2;
317		} else
318			goto out;
319	}
320	if (pos >= dp->ndm_eocookie) {
321		if (add)
322			dp->ndm_eocookie = pos + 1;
323		else
324			goto out;
325	}
326	retval = &dp->ndm_cookies[pos];
327out:
328	return (retval);
329}
330
331/*
332 * Invalidate cached directory information, except for the actual directory
333 * blocks (which are invalidated separately).
334 * Done mainly to avoid the use of stale offset cookies.
335 */
336void
337ncl_invaldir(struct vnode *vp)
338{
339	struct nfsnode *np = VTONFS(vp);
340
341	KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir"));
342	ncl_dircookie_lock(np);
343	np->n_direofoffset = 0;
344	np->n_cookieverf.nfsuquad[0] = 0;
345	np->n_cookieverf.nfsuquad[1] = 0;
346	if (LIST_FIRST(&np->n_cookies))
347		LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
348	ncl_dircookie_unlock(np);
349}
350
351/*
352 * The write verifier has changed (probably due to a server reboot), so all
353 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
354 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
355 * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
356 * mount point.
357 *
358 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
359 * writes are not clusterable.
360 */
361void
362ncl_clearcommit(struct mount *mp)
363{
364	struct vnode *vp, *nvp;
365	struct buf *bp, *nbp;
366	struct bufobj *bo;
367
368	MNT_VNODE_FOREACH_ALL(vp, mp, nvp) {
369		bo = &vp->v_bufobj;
370		vholdl(vp);
371		VI_UNLOCK(vp);
372		BO_LOCK(bo);
373		TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
374			if (!BUF_ISLOCKED(bp) &&
375			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
376				== (B_DELWRI | B_NEEDCOMMIT))
377				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
378		}
379		BO_UNLOCK(bo);
380		vdrop(vp);
381	}
382}
383
384/*
385 * Called once to initialize data structures...
386 */
387int
388ncl_init(struct vfsconf *vfsp)
389{
390	int i;
391
392	/* Ensure async daemons disabled */
393	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
394		ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE;
395		ncl_iodmount[i] = NULL;
396	}
397	TASK_INIT(&ncl_nfsiodnew_task, 0, ncl_nfsiodnew_tq, NULL);
398	ncl_nhinit();			/* Init the nfsnode table */
399
400	return (0);
401}
402
403