summaryrefslogtreecommitdiff
path: root/sys/kern/vfs_sync.c
blob: 549962182f6c50c665e5bd64b9d45b7cc6aac3b9 (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
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
/*       $OpenBSD: vfs_sync.c,v 1.30 2005/05/25 23:17:47 niklas Exp $  */

/*
 *  Portions of this code are:
 *
 * Copyright (c) 1989, 1993
 *	The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * 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.
 * 3. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
 */

/*
 * Syncer daemon
 */

#include <sys/queue.h>
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/proc.h>
#include <sys/mount.h>
#include <sys/vnode.h>
#include <sys/buf.h>
#include <sys/malloc.h>

#include <sys/kernel.h>
#include <sys/sched.h>

#ifdef FFS_SOFTUPDATES
int   softdep_process_worklist(struct mount *);
#endif

/*
 * The workitem queue.
 */
#define SYNCER_MAXDELAY	32		/* maximum sync delay time */
#define SYNCER_DEFAULT 30		/* default sync delay time */
int syncer_maxdelay = SYNCER_MAXDELAY;	/* maximum delay time */
time_t syncdelay = SYNCER_DEFAULT;	/* time to delay syncing vnodes */

int rushjob = 0;			/* number of slots to run ASAP */
int stat_rush_requests = 0;		/* number of rush requests */

static int syncer_delayno = 0;
static long syncer_mask;
LIST_HEAD(synclist, vnode);
static struct synclist *syncer_workitem_pending;

struct proc *syncerproc;

/*
 * The workitem queue.
 *
 * It is useful to delay writes of file data and filesystem metadata
 * for tens of seconds so that quickly created and deleted files need
 * not waste disk bandwidth being created and removed. To realize this,
 * we append vnodes to a "workitem" queue. When running with a soft
 * updates implementation, most pending metadata dependencies should
 * not wait for more than a few seconds. Thus, mounted on block devices
 * are delayed only about a half the time that file data is delayed.
 * Similarly, directory updates are more critical, so are only delayed
 * about a third the time that file data is delayed. Thus, there are
 * SYNCER_MAXDELAY queues that are processed round-robin at a rate of
 * one each second (driven off the filesystem syncer process). The
 * syncer_delayno variable indicates the next queue that is to be processed.
 * Items that need to be processed soon are placed in this queue:
 *
 *	syncer_workitem_pending[syncer_delayno]
 *
 * A delay of fifteen seconds is done by placing the request fifteen
 * entries later in the queue:
 *
 *	syncer_workitem_pending[(syncer_delayno + 15) & syncer_mask]
 *
 */

void
vn_initialize_syncerd()

{
	syncer_workitem_pending = hashinit(syncer_maxdelay, M_VNODE, M_WAITOK,
	    &syncer_mask);
	syncer_maxdelay = syncer_mask + 1;
}

/*
 * Add an item to the syncer work queue.
 */
void
vn_syncer_add_to_worklist(vp, delay)
	struct vnode *vp;
	int delay;
{
	int s, slot;

	if (delay > syncer_maxdelay - 2)
		delay = syncer_maxdelay - 2;
	slot = (syncer_delayno + delay) & syncer_mask;

	s = splbio();
	if (vp->v_bioflag & VBIOONSYNCLIST)
		LIST_REMOVE(vp, v_synclist);

	vp->v_bioflag |= VBIOONSYNCLIST;
	LIST_INSERT_HEAD(&syncer_workitem_pending[slot], vp, v_synclist);
	splx(s);
}

/*
 * System filesystem synchronizer daemon.
 */

void
sched_sync(p)
	struct proc *p;
{
	struct synclist *slp;
	struct vnode *vp;
	long starttime;
	int s;

	syncerproc = curproc;

	for (;;) {
		starttime = time_second;

		/*
		 * Push files whose dirty time has expired.
		 */
		s = splbio();
		slp = &syncer_workitem_pending[syncer_delayno];

		syncer_delayno += 1;
		if (syncer_delayno == syncer_maxdelay)
			syncer_delayno = 0;

		while ((vp = LIST_FIRST(slp)) != NULL) {
			if (vn_lock(vp, LK_EXCLUSIVE | LK_NOWAIT, p) != 0) {
				/*
				 * If we fail to get the lock, we move this
				 * vnode one second ahead in time.
				 * XXX - no good, but the best we can do.
				 */
				vn_syncer_add_to_worklist(vp, 1);
				continue;
			}
			splx(s);
			(void) VOP_FSYNC(vp, p->p_ucred, MNT_LAZY, p);
			VOP_UNLOCK(vp, 0, p);
			s = splbio();
			if (LIST_FIRST(slp) == vp) {
				/*
				 * Note: disk vps can remain on the
				 * worklist too with no dirty blocks, but
				 * since sync_fsync() moves it to a different
				 * slot we are safe.
				 */
				if (LIST_FIRST(&vp->v_dirtyblkhd) == NULL &&
				    vp->v_type != VBLK) {
					vprint("fsync failed", vp);
					if (vp->v_mount != NULL)
						printf("mounted on: %s\n",
						    vp->v_mount->mnt_stat.f_mntonname);
					panic("sched_sync: fsync failed");
				}
				/*
				 * Put us back on the worklist.  The worklist
				 * routine will remove us from our current
				 * position and then add us back in at a later
				 * position.
				 */
				vn_syncer_add_to_worklist(vp, syncdelay);
			}
		}

		splx(s);

#ifdef FFS_SOFTUPDATES
		/*
		 * Do soft update processing.
		 */
		softdep_process_worklist(NULL);
#endif

		/*
		 * The variable rushjob allows the kernel to speed up the
		 * processing of the filesystem syncer process. A rushjob
		 * value of N tells the filesystem syncer to process the next
		 * N seconds worth of work on its queue ASAP. Currently rushjob
		 * is used by the soft update code to speed up the filesystem
		 * syncer process when the incore state is getting so far
		 * ahead of the disk that the kernel memory pool is being
		 * threatened with exhaustion.
		 */
		if (rushjob > 0) {
			rushjob -= 1;
			continue;
		}
		/*
		 * If it has taken us less than a second to process the
		 * current work, then wait. Otherwise start right over
		 * again. We can still lose time if any single round
		 * takes more than two seconds, but it does not really
		 * matter as we are just trying to generally pace the
		 * filesystem activity.
		 */
		if (time_second == starttime)
			tsleep(&lbolt, PPAUSE, "syncer", 0);
	}
}

/*
 * Request the syncer daemon to speed up its work.
 * We never push it to speed up more than half of its
 * normal turn time, otherwise it could take over the cpu.
 */
int
speedup_syncer()
{
	int s;

	SCHED_LOCK(s);
	if (syncerproc && syncerproc->p_wchan == &lbolt)
		setrunnable(syncerproc);
	SCHED_UNLOCK(s);
	if (rushjob < syncdelay / 2) {
		rushjob += 1;
		stat_rush_requests += 1;
		return 1;
	}
	return 0;
}

/*
 * Routine to create and manage a filesystem syncer vnode.
 */
#define sync_close nullop
int   sync_fsync(void *);
int   sync_inactive(void *);
#define sync_reclaim nullop
#define sync_lock vop_generic_lock
#define sync_unlock vop_generic_unlock
int   sync_print(void *);
#define sync_islocked vop_generic_islocked

int (**sync_vnodeop_p)(void *);
struct vnodeopv_entry_desc sync_vnodeop_entries[] = {
      { &vop_default_desc, vn_default_error },
      { &vop_close_desc, sync_close },                /* close */
      { &vop_fsync_desc, sync_fsync },                /* fsync */
      { &vop_inactive_desc, sync_inactive },          /* inactive */
      { &vop_reclaim_desc, sync_reclaim },            /* reclaim */
      { &vop_lock_desc, sync_lock },                  /* lock */
      { &vop_unlock_desc, sync_unlock },              /* unlock */
      { &vop_print_desc, sync_print },                /* print */
      { &vop_islocked_desc, sync_islocked },          /* islocked */
      { (struct vnodeop_desc*)NULL, (int(*)(void *))NULL }
};
struct vnodeopv_desc sync_vnodeop_opv_desc = {
	&sync_vnodeop_p, sync_vnodeop_entries
};

/*
 * Create a new filesystem syncer vnode for the specified mount point.
 */
int
vfs_allocate_syncvnode(mp)
	struct mount *mp;
{
	struct vnode *vp;
	static long start, incr, next;
	int error;

	/* Allocate a new vnode */
	if ((error = getnewvnode(VT_VFS, mp, sync_vnodeop_p, &vp)) != 0) {
		mp->mnt_syncer = NULL;
		return (error);
	}
	vp->v_writecount = 1;
	vp->v_type = VNON;
	/*
	 * Place the vnode onto the syncer worklist. We attempt to
	 * scatter them about on the list so that they will go off
	 * at evenly distributed times even if all the filesystems
	 * are mounted at once.
	 */
	next += incr;
	if (next == 0 || next > syncer_maxdelay) {
		start /= 2;
		incr /= 2;
		if (start == 0) {
			start = syncer_maxdelay / 2;
			incr = syncer_maxdelay;
		}
		next = start;
	}
	vn_syncer_add_to_worklist(vp, next);
	mp->mnt_syncer = vp;
	return (0);
}

/*
 * Do a lazy sync of the filesystem.
 */
int
sync_fsync(v)
	void *v;
{
	struct vop_fsync_args /* {
		struct vnodeop_desc *a_desc;
		struct vnode *a_vp;
		struct ucred *a_cred;
		int a_waitfor;
		struct proc *a_p;
	} */ *ap = v;
	struct vnode *syncvp = ap->a_vp;
	struct mount *mp = syncvp->v_mount;
	int asyncflag;

	/*
	 * We only need to do something if this is a lazy evaluation.
	 */
	if (ap->a_waitfor != MNT_LAZY)
		return (0);

	/*
	 * Move ourselves to the back of the sync list.
	 */
	vn_syncer_add_to_worklist(syncvp, syncdelay);

	/*
	 * Walk the list of vnodes pushing all that are dirty and
	 * not already on the sync list.
	 */
	simple_lock(&mountlist_slock);
	if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, ap->a_p) == 0) {
		asyncflag = mp->mnt_flag & MNT_ASYNC;
		mp->mnt_flag &= ~MNT_ASYNC;
		VFS_SYNC(mp, MNT_LAZY, ap->a_cred, ap->a_p);
		if (asyncflag)
			mp->mnt_flag |= MNT_ASYNC;
		vfs_unbusy(mp, ap->a_p);
	} else
		simple_unlock(&mountlist_slock);

	return (0);
}

/*
 * The syncer vnode is no longer needed and is being decommissioned.
 */
int
sync_inactive(v)
	void *v;
{
	struct vop_inactive_args /* {
		struct vnodeop_desc *a_desc;
		struct vnode *a_vp;
		struct proc *a_p;
	} */ *ap = v;

	struct vnode *vp = ap->a_vp;
	int s;

	if (vp->v_usecount == 0) {
		VOP_UNLOCK(vp, 0, ap->a_p);
		return (0);
	}

	vp->v_mount->mnt_syncer = NULL;

	s = splbio();

	LIST_REMOVE(vp, v_synclist);
	vp->v_bioflag &= ~VBIOONSYNCLIST;

	splx(s);

	vp->v_writecount = 0;
	vput(vp);

	return (0);
}

/*
 * Print out a syncer vnode.
 */
int
sync_print(v)
	void *v;

{
	struct vop_print_args /* {
		struct vnodeop_desc *a_desc;
		struct vnode *a_vp;
	} */ *ap = v;
	struct vnode *vp = ap->a_vp;

	printf("syncer vnode");
	if (vp->v_vnlock != NULL)
		lockmgr_printinfo(vp->v_vnlock);
	printf("\n");
	return (0);
}