summaryrefslogtreecommitdiffstats
path: root/sys/gnu/fs/xfs/xfs_refcache.c
blob: 09918b5daa93d8a45978d3f0f13ae9401c440a52 (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
431
/*
 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
 * All Rights Reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_types.h"
#include "xfs_bit.h"
#include "xfs_log.h"
#include "xfs_inum.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_dir.h"
#include "xfs_dir2.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
#include "xfs_bmap_btree.h"
#include "xfs_alloc_btree.h"
#include "xfs_ialloc_btree.h"
#include "xfs_dir_sf.h"
#include "xfs_dir2_sf.h"
#include "xfs_attr_sf.h"
#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_inode_item.h"
#include "xfs_itable.h"
#include "xfs_btree.h"
#include "xfs_alloc.h"
#include "xfs_ialloc.h"
#include "xfs_bmap.h"
#include "xfs_attr.h"
#include "xfs_error.h"
#include "xfs_buf_item.h"
#include "xfs_refcache.h"

STATIC lock_t		xfs_refcache_lock;
STATIC xfs_inode_t	**xfs_refcache;
STATIC int		xfs_refcache_index;
STATIC int		xfs_refcache_busy;
STATIC int		xfs_refcache_count;

void
xfs_refcache_init(void)
{
	spinlock_init(&xfs_refcache_lock, "xfs_refcache");
}
/*
 * Insert the given inode into the reference cache.
 */
void
xfs_refcache_insert(
	xfs_inode_t	*ip)
{
	vnode_t		*vp;
	xfs_inode_t	*release_ip;
	xfs_inode_t	**refcache;

	ASSERT(ismrlocked(&(ip->i_iolock), MR_UPDATE));

	/*
	 * If an unmount is busy blowing entries out of the cache,
	 * then don't bother.
	 */
	if (xfs_refcache_busy) {
		return;
	}

	/*
	 * If we tuned the refcache down to zero, don't do anything.
	 */
	 if (!xfs_refcache_size) {
		return;
	}

	/*
	 * The inode is already in the refcache, so don't bother
	 * with it.
	 */
	if (ip->i_refcache != NULL) {
		return;
	}

	vp = XFS_ITOV(ip);
	/* ASSERT(vp->v_count > 0); */
	VN_HOLD(vp);

	/*
	 * We allocate the reference cache on use so that we don't
	 * waste the memory on systems not being used as NFS servers.
	 */
	if (xfs_refcache == NULL) {
		refcache = (xfs_inode_t **)kmem_zalloc(XFS_REFCACHE_SIZE_MAX *
						       sizeof(xfs_inode_t *),
						       KM_SLEEP);
	} else {
		refcache = NULL;
	}

	spin_lock(&xfs_refcache_lock);

	/*
	 * If we allocated memory for the refcache above and it still
	 * needs it, then use the memory we allocated.  Otherwise we'll
	 * free the memory below.
	 */
	if (refcache != NULL) {
		if (xfs_refcache == NULL) {
			xfs_refcache = refcache;
			refcache = NULL;
		}
	}

	/*
	 * If an unmount is busy clearing out the cache, don't add new
	 * entries to it.
	 */
	if (xfs_refcache_busy) {
		spin_unlock(&xfs_refcache_lock);
		VN_RELE(vp);
		/*
		 * If we allocated memory for the refcache above but someone
		 * else beat us to using it, then free the memory now.
		 */
		if (refcache != NULL) {
			kmem_free(refcache,
				  XFS_REFCACHE_SIZE_MAX * sizeof(xfs_inode_t *));
		}
		return;
	}
	release_ip = xfs_refcache[xfs_refcache_index];
	if (release_ip != NULL) {
		release_ip->i_refcache = NULL;
		xfs_refcache_count--;
		ASSERT(xfs_refcache_count >= 0);
	}
	xfs_refcache[xfs_refcache_index] = ip;
	ASSERT(ip->i_refcache == NULL);
	ip->i_refcache = &(xfs_refcache[xfs_refcache_index]);
	xfs_refcache_count++;
	ASSERT(xfs_refcache_count <= xfs_refcache_size);
	xfs_refcache_index++;
	if (xfs_refcache_index == xfs_refcache_size) {
		xfs_refcache_index = 0;
	}
	spin_unlock(&xfs_refcache_lock);

	/*
	 * Save the pointer to the inode to be released so that we can
	 * VN_RELE it once we've dropped our inode locks in xfs_rwunlock().
	 * The pointer may be NULL, but that's OK.
	 */
	ip->i_release = release_ip;

	/*
	 * If we allocated memory for the refcache above but someone
	 * else beat us to using it, then free the memory now.
	 */
	if (refcache != NULL) {
		kmem_free(refcache,
			  XFS_REFCACHE_SIZE_MAX * sizeof(xfs_inode_t *));
	}
}


/*
 * If the given inode is in the reference cache, purge its entry and
 * release the reference on the vnode.
 */
void
xfs_refcache_purge_ip(
	xfs_inode_t	*ip)
{
	vnode_t	*vp;
	int	error;

	/*
	 * If we're not pointing to our entry in the cache, then
	 * we must not be in the cache.
	 */
	if (ip->i_refcache == NULL) {
		return;
	}

	spin_lock(&xfs_refcache_lock);
	if (ip->i_refcache == NULL) {
		spin_unlock(&xfs_refcache_lock);
		return;
	}

	/*
	 * Clear both our pointer to the cache entry and its pointer
	 * back to us.
	 */
	ASSERT(*(ip->i_refcache) == ip);
	*(ip->i_refcache) = NULL;
	ip->i_refcache = NULL;
	xfs_refcache_count--;
	ASSERT(xfs_refcache_count >= 0);
	spin_unlock(&xfs_refcache_lock);

	vp = XFS_ITOV(ip);
	/* ASSERT(vp->v_count > 1); */
	VOP_RELEASE(vp, error);
	VN_RELE(vp);
}


/*
 * This is called from the XFS unmount code to purge all entries for the
 * given mount from the cache.  It uses the refcache busy counter to
 * make sure that new entries are not added to the cache as we purge them.
 */
void
xfs_refcache_purge_mp(
	xfs_mount_t	*mp)
{
	vnode_t		*vp;
	int		error, i;
	xfs_inode_t	*ip;

	if (xfs_refcache == NULL) {
		return;
	}

	spin_lock(&xfs_refcache_lock);
	/*
	 * Bumping the busy counter keeps new entries from being added
	 * to the cache.  We use a counter since multiple unmounts could
	 * be in here simultaneously.
	 */
	xfs_refcache_busy++;

	for (i = 0; i < xfs_refcache_size; i++) {
		ip = xfs_refcache[i];
		if ((ip != NULL) && (ip->i_mount == mp)) {
			xfs_refcache[i] = NULL;
			ip->i_refcache = NULL;
			xfs_refcache_count--;
			ASSERT(xfs_refcache_count >= 0);
			spin_unlock(&xfs_refcache_lock);
			vp = XFS_ITOV(ip);
			VOP_RELEASE(vp, error);
			VN_RELE(vp);
			spin_lock(&xfs_refcache_lock);
		}
	}

	xfs_refcache_busy--;
	ASSERT(xfs_refcache_busy >= 0);
	spin_unlock(&xfs_refcache_lock);
}


/*
 * This is called from the XFS sync code to ensure that the refcache
 * is emptied out over time.  We purge a small number of entries with
 * each call.
 */
void
xfs_refcache_purge_some(xfs_mount_t *mp)
{
	int		error, i;
	xfs_inode_t	*ip;
	int		iplist_index;
	xfs_inode_t	**iplist;

	if ((xfs_refcache == NULL) || (xfs_refcache_count == 0)) {
		return;
	}

	iplist_index = 0;
	iplist = (xfs_inode_t **)kmem_zalloc(xfs_refcache_purge_count *
					  sizeof(xfs_inode_t *), KM_SLEEP);

	spin_lock(&xfs_refcache_lock);

	/*
	 * Store any inodes we find in the next several entries
	 * into the iplist array to be released after dropping
	 * the spinlock.  We always start looking from the currently
	 * oldest place in the cache.  We move the refcache index
	 * forward as we go so that we are sure to eventually clear
	 * out the entire cache when the system goes idle.
	 */
	for (i = 0; i < xfs_refcache_purge_count; i++) {
		ip = xfs_refcache[xfs_refcache_index];
		if (ip != NULL) {
			xfs_refcache[xfs_refcache_index] = NULL;
			ip->i_refcache = NULL;
			xfs_refcache_count--;
			ASSERT(xfs_refcache_count >= 0);
			iplist[iplist_index] = ip;
			iplist_index++;
		}
		xfs_refcache_index++;
		if (xfs_refcache_index == xfs_refcache_size) {
			xfs_refcache_index = 0;
		}
	}

	spin_unlock(&xfs_refcache_lock);

	/*
	 * Now drop the inodes we collected.
	 */
	for (i = 0; i < iplist_index; i++) {
		VOP_RELEASE(XFS_ITOV(iplist[i]), error);
		VN_RELE(XFS_ITOV(iplist[i]));
	}

	kmem_free(iplist, xfs_refcache_purge_count *
			  sizeof(xfs_inode_t *));
}

/*
 * This is called when the refcache is dynamically resized
 * via a sysctl.
 *
 * If the new size is smaller than the old size, purge all
 * entries in slots greater than the new size, and move
 * the index if necessary.
 *
 * If the refcache hasn't even been allocated yet, or the
 * new size is larger than the old size, just set the value
 * of xfs_refcache_size.
 */

void
xfs_refcache_resize(int xfs_refcache_new_size)
{
	int		i;
	xfs_inode_t	*ip;
	int		iplist_index = 0;
	xfs_inode_t	**iplist;
	int		error;

	/*
	 * If the new size is smaller than the current size,
	 * purge entries to create smaller cache, and
	 * reposition index if necessary.
	 * Don't bother if no refcache yet.
	 */
	if (xfs_refcache && (xfs_refcache_new_size < xfs_refcache_size)) {

		iplist = (xfs_inode_t **)kmem_zalloc(XFS_REFCACHE_SIZE_MAX *
				sizeof(xfs_inode_t *), KM_SLEEP);

		spin_lock(&xfs_refcache_lock);

		for (i = xfs_refcache_new_size; i < xfs_refcache_size; i++) {
			ip = xfs_refcache[i];
			if (ip != NULL) {
				xfs_refcache[i] = NULL;
				ip->i_refcache = NULL;
				xfs_refcache_count--;
				ASSERT(xfs_refcache_count >= 0);
				iplist[iplist_index] = ip;
				iplist_index++;
			}
		}

		xfs_refcache_size = xfs_refcache_new_size;

		/*
		 * Move index to beginning of cache if it's now past the end
		 */
		if (xfs_refcache_index >= xfs_refcache_new_size)
			xfs_refcache_index = 0;

		spin_unlock(&xfs_refcache_lock);

		/*
		 * Now drop the inodes we collected.
		 */
		for (i = 0; i < iplist_index; i++) {
			VOP_RELEASE(XFS_ITOV(iplist[i]), error);
			VN_RELE(XFS_ITOV(iplist[i]));
		}

		kmem_free(iplist, XFS_REFCACHE_SIZE_MAX *
				  sizeof(xfs_inode_t *));
	} else {
		spin_lock(&xfs_refcache_lock);
		xfs_refcache_size = xfs_refcache_new_size;
		spin_unlock(&xfs_refcache_lock);
	}
}

void
xfs_refcache_iunlock(
	xfs_inode_t	*ip,
	uint		lock_flags)
{
	xfs_inode_t	*release_ip;
	int		error;

	release_ip = ip->i_release;
	ip->i_release = NULL;

	xfs_iunlock(ip, lock_flags);

	if (release_ip != NULL) {
		VOP_RELEASE(XFS_ITOV(release_ip), error);
		VN_RELE(XFS_ITOV(release_ip));
	}
}

void
xfs_refcache_destroy(void)
{
	if (xfs_refcache) {
		kmem_free(xfs_refcache,
			XFS_REFCACHE_SIZE_MAX * sizeof(xfs_inode_t *));
		xfs_refcache = NULL;
	}
	spinlock_destroy(&xfs_refcache_lock);
}
OpenPOWER on IntegriCloud