summaryrefslogtreecommitdiff
path: root/sys/nfs/nfs_kq.c
blob: b6b227c52a75708e57a431258034101ff193ca20 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
/*	$OpenBSD: nfs_kq.c,v 1.22 2014/11/15 00:03:12 tedu Exp $ */
/*	$NetBSD: nfs_kq.c,v 1.7 2003/10/30 01:43:10 simonb Exp $	*/

/*-
 * Copyright (c) 2002 The NetBSD Foundation, Inc.
 * All rights reserved.
 *
 * This code is derived from software contributed to The NetBSD Foundation
 * by Jaromir Dolecek.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
 * POSSIBILITY OF SUCH DAMAGE.
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/proc.h>
#include <sys/mount.h>
#include <sys/malloc.h>
#include <sys/vnode.h>
#include <sys/unistd.h>
#include <sys/file.h>
#include <sys/kthread.h>
#include <sys/rwlock.h>
#include <sys/queue.h>

#include <nfs/rpcv2.h>
#include <nfs/nfsproto.h>
#include <nfs/nfs.h>
#include <nfs/nfsnode.h>
#include <nfs/nfs_var.h>

void	nfs_kqpoll(void *);

void	filt_nfsdetach(struct knote *);
int	filt_nfsread(struct knote *, long);
int	filt_nfsvnode(struct knote *, long);

struct kevq {
	SLIST_ENTRY(kevq)	kev_link;
	struct vnode		*vp;
	u_int			usecount;
	u_int			flags;
#define KEVQ_BUSY	0x01	/* currently being processed */
#define KEVQ_WANT	0x02	/* want to change this entry */
	struct timespec		omtime;	/* old modification time */
	struct timespec		octime;	/* old change time */
	nlink_t			onlink;	/* old number of references to file */
};
SLIST_HEAD(kevqlist, kevq);

struct rwlock nfskevq_lock = RWLOCK_INITIALIZER("nfskqlk");
struct proc *pnfskq;
struct kevqlist kevlist = SLIST_HEAD_INITIALIZER(kevlist);

/*
 * This quite simplistic routine periodically checks for server changes
 * of any of the watched files every NFS_MINATTRTIMO/2 seconds.
 * Only changes in size, modification time, change time and nlinks
 * are being checked, everything else is ignored.
 * The routine only calls VOP_GETATTR() when it's likely it would get
 * some new data, i.e. when the vnode expires from attrcache. This
 * should give same result as periodically running stat(2) from userland,
 * while keeping CPU/network usage low, and still provide proper kevent
 * semantics.
 * The poller thread is created when first vnode is added to watch list,
 * and exits when the watch list is empty. The overhead of thread creation
 * isn't really important, neither speed of attach and detach of knote.
 */
/* ARGSUSED */
void
nfs_kqpoll(void *arg)
{
	struct kevq *ke;
	struct vattr attr;
	struct proc *p = pnfskq;
	u_quad_t osize;
	int error;

	for(;;) {
		rw_enter_write(&nfskevq_lock);
		SLIST_FOREACH(ke, &kevlist, kev_link) {
			struct nfsnode *np = VTONFS(ke->vp);

#ifdef DEBUG
			printf("nfs_kqpoll on: ");
			VOP_PRINT(ke->vp);
#endif
			/* skip if still in attrcache */
			if (nfs_getattrcache(ke->vp, &attr) != ENOENT)
				continue;

			/*
			 * Mark entry busy, release lock and check
			 * for changes.
			 */
			ke->flags |= KEVQ_BUSY;
			rw_exit_write(&nfskevq_lock);

			/* save v_size, nfs_getattr() updates it */
			osize = np->n_size;

			error = VOP_GETATTR(ke->vp, &attr, p->p_ucred, p);
			if (error == ESTALE) {
				NFS_INVALIDATE_ATTRCACHE(np);
				VN_KNOTE(ke->vp, NOTE_DELETE);
				goto next;
			}

			/* following is a bit fragile, but about best
			 * we can get */
			if (attr.va_size != osize) {
				int flags = NOTE_WRITE;

				if (attr.va_size > osize)
					flags |= NOTE_EXTEND;
				else
					flags |= NOTE_TRUNCATE;

				VN_KNOTE(ke->vp, flags);
				ke->omtime = attr.va_mtime;
			} else if (attr.va_mtime.tv_sec != ke->omtime.tv_sec
			    || attr.va_mtime.tv_nsec != ke->omtime.tv_nsec) {
				VN_KNOTE(ke->vp, NOTE_WRITE);
				ke->omtime = attr.va_mtime;
			}

			if (attr.va_ctime.tv_sec != ke->octime.tv_sec
			    || attr.va_ctime.tv_nsec != ke->octime.tv_nsec) {
				VN_KNOTE(ke->vp, NOTE_ATTRIB);
				ke->octime = attr.va_ctime;
			}

			if (attr.va_nlink != ke->onlink) {
				VN_KNOTE(ke->vp, NOTE_LINK);
				ke->onlink = attr.va_nlink;
			}

next:
			rw_enter_write(&nfskevq_lock);
			ke->flags &= ~KEVQ_BUSY;
			if (ke->flags & KEVQ_WANT) {
				ke->flags &= ~KEVQ_WANT;
				wakeup(ke);
			}
		}

		if (SLIST_EMPTY(&kevlist)) {
			/* Nothing more to watch, exit */
			pnfskq = NULL;
			rw_exit_write(&nfskevq_lock);
			kthread_exit(0);
		}
		rw_exit_write(&nfskevq_lock);

		/* wait a while before checking for changes again */
		tsleep(pnfskq, PSOCK, "nfskqpw", NFS_MINATTRTIMO * hz / 2);

	}
}

void
filt_nfsdetach(struct knote *kn)
{
	struct vnode *vp = (struct vnode *)kn->kn_hook;
	struct kevq *ke;

	SLIST_REMOVE(&vp->v_selectinfo.si_note, kn, knote, kn_selnext);

	/* Remove the vnode from watch list */
	rw_enter_write(&nfskevq_lock);
	SLIST_FOREACH(ke, &kevlist, kev_link) {
		if (ke->vp == vp) {
			while (ke->flags & KEVQ_BUSY) {
				ke->flags |= KEVQ_WANT;
				rw_exit_write(&nfskevq_lock);
				(void) tsleep(ke, PSOCK, "nfskqdet", 0);
				rw_enter_write(&nfskevq_lock);
			}

			if (ke->usecount > 1) {
				/* keep, other kevents need this */
				ke->usecount--;
			} else {
				/* last user, g/c */
				SLIST_REMOVE(&kevlist, ke, kevq, kev_link);
				free(ke, M_KEVENT, sizeof(*ke));
			}
			break;
		}
	}
	rw_exit_write(&nfskevq_lock);
}

int
filt_nfsread(struct knote *kn, long hint)
{
	struct vnode *vp = (struct vnode *)kn->kn_hook;
	struct nfsnode *np = VTONFS(vp);

	/*
	 * filesystem is gone, so set the EOF flag and schedule
	 * the knote for deletion.
	 */
	if (hint == NOTE_REVOKE) {
		kn->kn_flags |= (EV_EOF | EV_ONESHOT);
		return (1);
	}

	kn->kn_data = np->n_size - kn->kn_fp->f_offset;
#ifdef DEBUG
	printf("nfsread event. %lld\n", kn->kn_data);
#endif
	if (kn->kn_data == 0 && kn->kn_sfflags & NOTE_EOF) {
		kn->kn_fflags |= NOTE_EOF;
		return (1);
	}
        return (kn->kn_data != 0);
}

int
filt_nfsvnode(struct knote *kn, long hint)
{
	if (kn->kn_sfflags & hint)
		kn->kn_fflags |= hint;
	if (hint == NOTE_REVOKE) {
		kn->kn_flags |= EV_EOF;
		return (1);
	}
	return (kn->kn_fflags != 0);
}

static const struct filterops nfsread_filtops = 
	{ 1, NULL, filt_nfsdetach, filt_nfsread };
static const struct filterops nfsvnode_filtops = 
	{ 1, NULL, filt_nfsdetach, filt_nfsvnode };

int
nfs_kqfilter(void *v)
{
	struct vop_kqfilter_args *ap = v;
	struct vnode *vp;
	struct knote *kn;
	struct kevq *ke;
	int error = 0;
	struct vattr attr;
	struct proc *p = curproc;	/* XXX */

	vp = ap->a_vp;
	kn = ap->a_kn;

#ifdef DEBUG
	printf("nfs_kqfilter(%d) on: ", kn->kn_filter);
	VOP_PRINT(vp);
#endif

	switch (kn->kn_filter) {
	case EVFILT_READ:
		kn->kn_fop = &nfsread_filtops;
		break;
	case EVFILT_VNODE:
		kn->kn_fop = &nfsvnode_filtops;
		break;
	default:
		return (EINVAL);
	}

	kn->kn_hook = vp;

	/*
	 * Put the vnode to watched list.
	 */
	
	/*
	 * Fetch current attributes. It's only needed when the vnode
	 * is not watched yet, but we need to do this without lock
	 * held. This is likely cheap due to attrcache, so do it now.
	 */ 
	memset(&attr, 0, sizeof(attr));
	(void) VOP_GETATTR(vp, &attr, p->p_ucred, p);

	rw_enter_write(&nfskevq_lock);

	/* ensure the poller is running */
	if (!pnfskq) {
		error = kthread_create(nfs_kqpoll, NULL, &pnfskq,
				"nfskqpoll");
		if (error)
			goto out;
	}

	SLIST_FOREACH(ke, &kevlist, kev_link)
		if (ke->vp == vp)
			break;

	if (ke) {
		/* already watched, so just bump usecount */
		ke->usecount++;
	} else {
		/* need a new one */
		ke = malloc(sizeof(*ke), M_KEVENT, M_WAITOK);
		ke->vp = vp;
		ke->usecount = 1;
		ke->flags = 0;
		ke->omtime = attr.va_mtime;
		ke->octime = attr.va_ctime;
		ke->onlink = attr.va_nlink;
		SLIST_INSERT_HEAD(&kevlist, ke, kev_link);
	}

	/* kick the poller */
	wakeup(pnfskq);

	SLIST_INSERT_HEAD(&vp->v_selectinfo.si_note, kn, kn_selnext);

out:
	rw_exit_write(&nfskevq_lock);
	return (error);
}