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("$FreeBSD$");
37
38#include "opt_kdtrace.h"
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 nfsstats newnfsstats;
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	mtx_lock(&ncl_iod_mutex);
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	mtx_unlock(&ncl_iod_mutex);
114	ncl_nhuninit();
115	return (0);
116#else
117	return (EOPNOTSUPP);
118#endif
119}
120
121void
122ncl_dircookie_lock(struct nfsnode *np)
123{
124	mtx_lock(&np->n_mtx);
125	while (np->n_flag & NDIRCOOKIELK)
126		(void) msleep(&np->n_flag, &np->n_mtx, PZERO, "nfsdirlk", 0);
127	np->n_flag |= NDIRCOOKIELK;
128	mtx_unlock(&np->n_mtx);
129}
130
131void
132ncl_dircookie_unlock(struct nfsnode *np)
133{
134	mtx_lock(&np->n_mtx);
135	np->n_flag &= ~NDIRCOOKIELK;
136	wakeup(&np->n_flag);
137	mtx_unlock(&np->n_mtx);
138}
139
140int
141ncl_upgrade_vnlock(struct vnode *vp)
142{
143	int old_lock;
144
145	ASSERT_VOP_LOCKED(vp, "ncl_upgrade_vnlock");
146	old_lock = NFSVOPISLOCKED(vp);
147	if (old_lock != LK_EXCLUSIVE) {
148		KASSERT(old_lock == LK_SHARED,
149		    ("ncl_upgrade_vnlock: wrong old_lock %d", old_lock));
150		/* Upgrade to exclusive lock, this might block */
151		NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY);
152  	}
153	return (old_lock);
154}
155
156void
157ncl_downgrade_vnlock(struct vnode *vp, int old_lock)
158{
159	if (old_lock != LK_EXCLUSIVE) {
160		KASSERT(old_lock == LK_SHARED, ("wrong old_lock %d", old_lock));
161		/* Downgrade from exclusive lock. */
162		NFSVOPLOCK(vp, LK_DOWNGRADE | LK_RETRY);
163  	}
164}
165
166void
167ncl_printf(const char *fmt, ...)
168{
169	va_list ap;
170
171	mtx_lock(&Giant);
172	va_start(ap, fmt);
173	vprintf(fmt, ap);
174	va_end(ap);
175	mtx_unlock(&Giant);
176}
177
178#ifdef NFS_ACDEBUG
179#include <sys/sysctl.h>
180SYSCTL_DECL(_vfs_nfs);
181static int nfs_acdebug;
182SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
183#endif
184
185/*
186 * Check the time stamp
187 * If the cache is valid, copy contents to *vap and return 0
188 * otherwise return an error
189 */
190int
191ncl_getattrcache(struct vnode *vp, struct vattr *vaper)
192{
193	struct nfsnode *np;
194	struct vattr *vap;
195	struct nfsmount *nmp;
196	int timeo, mustflush;
197
198	np = VTONFS(vp);
199	vap = &np->n_vattr.na_vattr;
200	nmp = VFSTONFS(vp->v_mount);
201	mustflush = nfscl_mustflush(vp);	/* must be before mtx_lock() */
202#ifdef NFS_ACDEBUG
203	mtx_lock(&Giant);	/* ncl_printf() */
204#endif
205	mtx_lock(&np->n_mtx);
206	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
207	timeo = (time_second - np->n_mtime.tv_sec) / 10;
208
209#ifdef NFS_ACDEBUG
210	if (nfs_acdebug>1)
211		ncl_printf("nfs_getattrcache: initial timeo = %d\n", timeo);
212#endif
213
214	if (vap->va_type == VDIR) {
215		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
216			timeo = nmp->nm_acdirmin;
217		else if (timeo > nmp->nm_acdirmax)
218			timeo = nmp->nm_acdirmax;
219	} else {
220		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
221			timeo = nmp->nm_acregmin;
222		else if (timeo > nmp->nm_acregmax)
223			timeo = nmp->nm_acregmax;
224	}
225
226#ifdef NFS_ACDEBUG
227	if (nfs_acdebug > 2)
228		ncl_printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
229			   nmp->nm_acregmin, nmp->nm_acregmax,
230			   nmp->nm_acdirmin, nmp->nm_acdirmax);
231
232	if (nfs_acdebug)
233		ncl_printf("nfs_getattrcache: age = %d; final timeo = %d\n",
234			   (time_second - np->n_attrstamp), timeo);
235#endif
236
237	if ((time_second - np->n_attrstamp) >= timeo &&
238	    (mustflush != 0 || np->n_attrstamp == 0)) {
239		newnfsstats.attrcache_misses++;
240		mtx_unlock(&np->n_mtx);
241#ifdef NFS_ACDEBUG
242		mtx_unlock(&Giant);	/* ncl_printf() */
243#endif
244		KDTRACE_NFS_ATTRCACHE_GET_MISS(vp);
245		return( ENOENT);
246	}
247	newnfsstats.attrcache_hits++;
248	if (vap->va_size != np->n_size) {
249		if (vap->va_type == VREG) {
250			if (np->n_flag & NMODIFIED) {
251				if (vap->va_size < np->n_size)
252					vap->va_size = np->n_size;
253				else
254					np->n_size = vap->va_size;
255			} else {
256				np->n_size = vap->va_size;
257			}
258			vnode_pager_setsize(vp, np->n_size);
259		} else {
260			np->n_size = vap->va_size;
261		}
262	}
263	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
264	if (np->n_flag & NCHG) {
265		if (np->n_flag & NACC)
266			vaper->va_atime = np->n_atim;
267		if (np->n_flag & NUPD)
268			vaper->va_mtime = np->n_mtim;
269	}
270	mtx_unlock(&np->n_mtx);
271#ifdef NFS_ACDEBUG
272	mtx_unlock(&Giant);	/* ncl_printf() */
273#endif
274	KDTRACE_NFS_ATTRCACHE_GET_HIT(vp, vap);
275	return (0);
276}
277
278static nfsuint64 nfs_nullcookie = { { 0, 0 } };
279/*
280 * This function finds the directory cookie that corresponds to the
281 * logical byte offset given.
282 */
283nfsuint64 *
284ncl_getcookie(struct nfsnode *np, off_t off, int add)
285{
286	struct nfsdmap *dp, *dp2;
287	int pos;
288	nfsuint64 *retval = NULL;
289
290	pos = (uoff_t)off / NFS_DIRBLKSIZ;
291	if (pos == 0 || off < 0) {
292		KASSERT(!add, ("nfs getcookie add at <= 0"));
293		return (&nfs_nullcookie);
294	}
295	pos--;
296	dp = LIST_FIRST(&np->n_cookies);
297	if (!dp) {
298		if (add) {
299			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
300				M_NFSDIROFF, M_WAITOK);
301			dp->ndm_eocookie = 0;
302			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
303		} else
304			goto out;
305	}
306	while (pos >= NFSNUMCOOKIES) {
307		pos -= NFSNUMCOOKIES;
308		if (LIST_NEXT(dp, ndm_list)) {
309			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
310			    pos >= dp->ndm_eocookie)
311				goto out;
312			dp = LIST_NEXT(dp, ndm_list);
313		} else if (add) {
314			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
315				M_NFSDIROFF, M_WAITOK);
316			dp2->ndm_eocookie = 0;
317			LIST_INSERT_AFTER(dp, dp2, ndm_list);
318			dp = dp2;
319		} else
320			goto out;
321	}
322	if (pos >= dp->ndm_eocookie) {
323		if (add)
324			dp->ndm_eocookie = pos + 1;
325		else
326			goto out;
327	}
328	retval = &dp->ndm_cookies[pos];
329out:
330	return (retval);
331}
332
333/*
334 * Invalidate cached directory information, except for the actual directory
335 * blocks (which are invalidated separately).
336 * Done mainly to avoid the use of stale offset cookies.
337 */
338void
339ncl_invaldir(struct vnode *vp)
340{
341	struct nfsnode *np = VTONFS(vp);
342
343	KASSERT(vp->v_type == VDIR, ("nfs: invaldir not dir"));
344	ncl_dircookie_lock(np);
345	np->n_direofoffset = 0;
346	np->n_cookieverf.nfsuquad[0] = 0;
347	np->n_cookieverf.nfsuquad[1] = 0;
348	if (LIST_FIRST(&np->n_cookies))
349		LIST_FIRST(&np->n_cookies)->ndm_eocookie = 0;
350	ncl_dircookie_unlock(np);
351}
352
353/*
354 * The write verifier has changed (probably due to a server reboot), so all
355 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
356 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
357 * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
358 * mount point.
359 *
360 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
361 * writes are not clusterable.
362 */
363void
364ncl_clearcommit(struct mount *mp)
365{
366	struct vnode *vp, *nvp;
367	struct buf *bp, *nbp;
368	struct bufobj *bo;
369
370	MNT_VNODE_FOREACH_ALL(vp, mp, nvp) {
371		bo = &vp->v_bufobj;
372		vholdl(vp);
373		VI_UNLOCK(vp);
374		BO_LOCK(bo);
375		TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) {
376			if (!BUF_ISLOCKED(bp) &&
377			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
378				== (B_DELWRI | B_NEEDCOMMIT))
379				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
380		}
381		BO_UNLOCK(bo);
382		vdrop(vp);
383	}
384}
385
386/*
387 * Called once to initialize data structures...
388 */
389int
390ncl_init(struct vfsconf *vfsp)
391{
392	int i;
393
394	/* Ensure async daemons disabled */
395	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
396		ncl_iodwant[i] = NFSIOD_NOT_AVAILABLE;
397		ncl_iodmount[i] = NULL;
398	}
399	TASK_INIT(&ncl_nfsiodnew_task, 0, ncl_nfsiodnew_tq, NULL);
400	ncl_nhinit();			/* Init the nfsnode table */
401
402	return (0);
403}
404
405