1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
6 * Tracking the up-to-date-ness of a local buffer_head with respect to
9 * Copyright (C) 2002, 2004, 2005 Oracle. All rights reserved.
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public
13 * License as published by the Free Software Foundation; either
14 * version 2 of the License, or (at your option) any later version.
16 * This program is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * General Public License for more details.
21 * You should have received a copy of the GNU General Public
22 * License along with this program; if not, write to the
23 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
24 * Boston, MA 021110-1307, USA.
26 * Standard buffer head caching flags (uptodate, etc) are insufficient
27 * in a clustered environment - a buffer may be marked up to date on
28 * our local node but could have been modified by another cluster
29 * member. As a result an additional (and performant) caching scheme
30 * is required. A further requirement is that we consume as little
31 * memory as possible - we never pin buffer_head structures in order
34 * We track the existence of up to date buffers on the inodes which
35 * are associated with them. Because we don't want to pin
36 * buffer_heads, this is only a (strong) hint and several other checks
37 * are made in the I/O path to ensure that we don't use a stale or
38 * invalid buffer without going to disk:
39 * - buffer_jbd is used liberally - if a bh is in the journal on
40 * this node then it *must* be up to date.
41 * - the standard buffer_uptodate() macro is used to detect buffers
42 * which may be invalid (even if we have an up to date tracking
45 * For a full understanding of how this code works together, one
46 * should read the callers in dlmglue.c, the I/O functions in
47 * buffer_head_io.c and ocfs2_journal_access in journal.c
51 #include <linux/types.h>
52 #include <linux/slab.h>
53 #include <linux/highmem.h>
54 #include <linux/buffer_head.h>
55 #include <linux/rbtree.h>
56 #include <linux/jbd.h>
58 #define MLOG_MASK_PREFIX ML_UPTODATE
60 #include <cluster/masklog.h>
67 struct ocfs2_meta_cache_item {
68 struct rb_node c_node;
72 static kmem_cache_t *ocfs2_uptodate_cachep = NULL;
74 void ocfs2_metadata_cache_init(struct inode *inode)
76 struct ocfs2_inode_info *oi = OCFS2_I(inode);
77 struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
79 oi->ip_flags |= OCFS2_INODE_CACHE_INLINE;
80 ci->ci_num_cached = 0;
83 /* No lock taken here as 'root' is not expected to be visible to other
85 static unsigned int ocfs2_purge_copied_metadata_tree(struct rb_root *root)
87 unsigned int purged = 0;
89 struct ocfs2_meta_cache_item *item;
91 while ((node = rb_last(root)) != NULL) {
92 item = rb_entry(node, struct ocfs2_meta_cache_item, c_node);
94 mlog(0, "Purge item %llu\n",
95 (unsigned long long) item->c_block);
97 rb_erase(&item->c_node, root);
98 kmem_cache_free(ocfs2_uptodate_cachep, item);
105 /* Called from locking and called from ocfs2_clear_inode. Dump the
106 * cache for a given inode.
108 * This function is a few more lines longer than necessary due to some
109 * accounting done here, but I think it's worth tracking down those
110 * bugs sooner -- Mark */
111 void ocfs2_metadata_cache_purge(struct inode *inode)
113 struct ocfs2_inode_info *oi = OCFS2_I(inode);
114 unsigned int tree, to_purge, purged;
115 struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
116 struct rb_root root = RB_ROOT;
118 spin_lock(&oi->ip_lock);
119 tree = !(oi->ip_flags & OCFS2_INODE_CACHE_INLINE);
120 to_purge = ci->ci_num_cached;
122 mlog(0, "Purge %u %s items from Inode %llu\n", to_purge,
123 tree ? "array" : "tree", (unsigned long long)oi->ip_blkno);
125 /* If we're a tree, save off the root so that we can safely
126 * initialize the cache. We do the work to free tree members
127 * without the spinlock. */
129 root = ci->ci_cache.ci_tree;
131 ocfs2_metadata_cache_init(inode);
132 spin_unlock(&oi->ip_lock);
134 purged = ocfs2_purge_copied_metadata_tree(&root);
135 /* If possible, track the number wiped so that we can more
136 * easily detect counting errors. Unfortunately, this is only
137 * meaningful for trees. */
138 if (tree && purged != to_purge)
139 mlog(ML_ERROR, "Inode %llu, count = %u, purged = %u\n",
140 (unsigned long long)oi->ip_blkno, to_purge, purged);
143 /* Returns the index in the cache array, -1 if not found.
144 * Requires ip_lock. */
145 static int ocfs2_search_cache_array(struct ocfs2_caching_info *ci,
150 for (i = 0; i < ci->ci_num_cached; i++) {
151 if (item == ci->ci_cache.ci_array[i])
158 /* Returns the cache item if found, otherwise NULL.
159 * Requires ip_lock. */
160 static struct ocfs2_meta_cache_item *
161 ocfs2_search_cache_tree(struct ocfs2_caching_info *ci,
164 struct rb_node * n = ci->ci_cache.ci_tree.rb_node;
165 struct ocfs2_meta_cache_item *item = NULL;
168 item = rb_entry(n, struct ocfs2_meta_cache_item, c_node);
170 if (block < item->c_block)
172 else if (block > item->c_block)
181 static int ocfs2_buffer_cached(struct ocfs2_inode_info *oi,
182 struct buffer_head *bh)
185 struct ocfs2_meta_cache_item *item = NULL;
187 spin_lock(&oi->ip_lock);
189 mlog(0, "Inode %llu, query block %llu (inline = %u)\n",
190 (unsigned long long)oi->ip_blkno,
191 (unsigned long long) bh->b_blocknr,
192 !!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE));
194 if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE)
195 index = ocfs2_search_cache_array(&oi->ip_metadata_cache,
198 item = ocfs2_search_cache_tree(&oi->ip_metadata_cache,
201 spin_unlock(&oi->ip_lock);
203 mlog(0, "index = %d, item = %p\n", index, item);
205 return (index != -1) || (item != NULL);
208 /* Warning: even if it returns true, this does *not* guarantee that
209 * the block is stored in our inode metadata cache. */
210 int ocfs2_buffer_uptodate(struct inode *inode,
211 struct buffer_head *bh)
213 /* Doesn't matter if the bh is in our cache or not -- if it's
214 * not marked uptodate then we know it can't have correct
216 if (!buffer_uptodate(bh))
219 /* OCFS2 does not allow multiple nodes to be changing the same
220 * block at the same time. */
224 /* Ok, locally the buffer is marked as up to date, now search
225 * our cache to see if we can trust that. */
226 return ocfs2_buffer_cached(OCFS2_I(inode), bh);
229 /* Requires ip_lock */
230 static void ocfs2_append_cache_array(struct ocfs2_caching_info *ci,
233 BUG_ON(ci->ci_num_cached >= OCFS2_INODE_MAX_CACHE_ARRAY);
235 mlog(0, "block %llu takes position %u\n", (unsigned long long) block,
238 ci->ci_cache.ci_array[ci->ci_num_cached] = block;
242 /* By now the caller should have checked that the item does *not*
244 * Requires ip_lock. */
245 static void __ocfs2_insert_cache_tree(struct ocfs2_caching_info *ci,
246 struct ocfs2_meta_cache_item *new)
248 sector_t block = new->c_block;
249 struct rb_node *parent = NULL;
250 struct rb_node **p = &ci->ci_cache.ci_tree.rb_node;
251 struct ocfs2_meta_cache_item *tmp;
253 mlog(0, "Insert block %llu num = %u\n", (unsigned long long) block,
259 tmp = rb_entry(parent, struct ocfs2_meta_cache_item, c_node);
261 if (block < tmp->c_block)
263 else if (block > tmp->c_block)
266 /* This should never happen! */
267 mlog(ML_ERROR, "Duplicate block %llu cached!\n",
268 (unsigned long long) block);
273 rb_link_node(&new->c_node, parent, p);
274 rb_insert_color(&new->c_node, &ci->ci_cache.ci_tree);
278 static inline int ocfs2_insert_can_use_array(struct ocfs2_inode_info *oi,
279 struct ocfs2_caching_info *ci)
281 assert_spin_locked(&oi->ip_lock);
283 return (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) &&
284 (ci->ci_num_cached < OCFS2_INODE_MAX_CACHE_ARRAY);
287 /* tree should be exactly OCFS2_INODE_MAX_CACHE_ARRAY wide. NULL the
288 * pointers in tree after we use them - this allows caller to detect
289 * when to free in case of error. */
290 static void ocfs2_expand_cache(struct ocfs2_inode_info *oi,
291 struct ocfs2_meta_cache_item **tree)
294 struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
296 mlog_bug_on_msg(ci->ci_num_cached != OCFS2_INODE_MAX_CACHE_ARRAY,
297 "Inode %llu, num cached = %u, should be %u\n",
298 (unsigned long long)oi->ip_blkno, ci->ci_num_cached,
299 OCFS2_INODE_MAX_CACHE_ARRAY);
300 mlog_bug_on_msg(!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE),
301 "Inode %llu not marked as inline anymore!\n",
302 (unsigned long long)oi->ip_blkno);
303 assert_spin_locked(&oi->ip_lock);
305 /* Be careful to initialize the tree members *first* because
306 * once the ci_tree is used, the array is junk... */
307 for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++)
308 tree[i]->c_block = ci->ci_cache.ci_array[i];
310 oi->ip_flags &= ~OCFS2_INODE_CACHE_INLINE;
311 ci->ci_cache.ci_tree = RB_ROOT;
312 /* this will be set again by __ocfs2_insert_cache_tree */
313 ci->ci_num_cached = 0;
315 for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) {
316 __ocfs2_insert_cache_tree(ci, tree[i]);
320 mlog(0, "Expanded %llu to a tree cache: flags 0x%x, num = %u\n",
321 (unsigned long long)oi->ip_blkno, oi->ip_flags, ci->ci_num_cached);
324 /* Slow path function - memory allocation is necessary. See the
325 * comment above ocfs2_set_buffer_uptodate for more information. */
326 static void __ocfs2_set_buffer_uptodate(struct ocfs2_inode_info *oi,
331 struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
332 struct ocfs2_meta_cache_item *new = NULL;
333 struct ocfs2_meta_cache_item *tree[OCFS2_INODE_MAX_CACHE_ARRAY] =
336 mlog(0, "Inode %llu, block %llu, expand = %d\n",
337 (unsigned long long)oi->ip_blkno,
338 (unsigned long long)block, expand_tree);
340 new = kmem_cache_alloc(ocfs2_uptodate_cachep, GFP_NOFS);
345 new->c_block = block;
348 /* Do *not* allocate an array here - the removal code
349 * has no way of tracking that. */
350 for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) {
351 tree[i] = kmem_cache_alloc(ocfs2_uptodate_cachep,
358 /* These are initialized in ocfs2_expand_cache! */
362 spin_lock(&oi->ip_lock);
363 if (ocfs2_insert_can_use_array(oi, ci)) {
364 mlog(0, "Someone cleared the tree underneath us\n");
365 /* Ok, items were removed from the cache in between
366 * locks. Detect this and revert back to the fast path */
367 ocfs2_append_cache_array(ci, block);
368 spin_unlock(&oi->ip_lock);
373 ocfs2_expand_cache(oi, tree);
375 __ocfs2_insert_cache_tree(ci, new);
376 spin_unlock(&oi->ip_lock);
381 kmem_cache_free(ocfs2_uptodate_cachep, new);
383 /* If these were used, then ocfs2_expand_cache re-set them to
386 for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++)
388 kmem_cache_free(ocfs2_uptodate_cachep,
393 /* Item insertion is guarded by ip_io_mutex, so the insertion path takes
394 * advantage of this by not rechecking for a duplicate insert during
395 * the slow case. Additionally, if the cache needs to be bumped up to
396 * a tree, the code will not recheck after acquiring the lock --
397 * multiple paths cannot be expanding to a tree at the same time.
399 * The slow path takes into account that items can be removed
400 * (including the whole tree wiped and reset) when this process it out
401 * allocating memory. In those cases, it reverts back to the fast
404 * Note that this function may actually fail to insert the block if
405 * memory cannot be allocated. This is not fatal however (but may
406 * result in a performance penalty) */
407 void ocfs2_set_buffer_uptodate(struct inode *inode,
408 struct buffer_head *bh)
411 struct ocfs2_inode_info *oi = OCFS2_I(inode);
412 struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
414 /* The block may very well exist in our cache already, so avoid
415 * doing any more work in that case. */
416 if (ocfs2_buffer_cached(oi, bh))
419 mlog(0, "Inode %llu, inserting block %llu\n",
420 (unsigned long long)oi->ip_blkno,
421 (unsigned long long)bh->b_blocknr);
423 /* No need to recheck under spinlock - insertion is guarded by
425 spin_lock(&oi->ip_lock);
426 if (ocfs2_insert_can_use_array(oi, ci)) {
427 /* Fast case - it's an array and there's a free
429 ocfs2_append_cache_array(ci, bh->b_blocknr);
430 spin_unlock(&oi->ip_lock);
435 if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) {
436 /* We need to bump things up to a tree. */
439 spin_unlock(&oi->ip_lock);
441 __ocfs2_set_buffer_uptodate(oi, bh->b_blocknr, expand);
444 /* Called against a newly allocated buffer. Most likely nobody should
445 * be able to read this sort of metadata while it's still being
446 * allocated, but this is careful to take ip_io_mutex anyway. */
447 void ocfs2_set_new_buffer_uptodate(struct inode *inode,
448 struct buffer_head *bh)
450 struct ocfs2_inode_info *oi = OCFS2_I(inode);
452 /* This should definitely *not* exist in our cache */
453 BUG_ON(ocfs2_buffer_cached(oi, bh));
455 set_buffer_uptodate(bh);
457 mutex_lock(&oi->ip_io_mutex);
458 ocfs2_set_buffer_uptodate(inode, bh);
459 mutex_unlock(&oi->ip_io_mutex);
462 /* Requires ip_lock. */
463 static void ocfs2_remove_metadata_array(struct ocfs2_caching_info *ci,
466 sector_t *array = ci->ci_cache.ci_array;
469 BUG_ON(index < 0 || index >= OCFS2_INODE_MAX_CACHE_ARRAY);
470 BUG_ON(index >= ci->ci_num_cached);
471 BUG_ON(!ci->ci_num_cached);
473 mlog(0, "remove index %d (num_cached = %u\n", index,
478 /* don't need to copy if the array is now empty, or if we
479 * removed at the tail */
480 if (ci->ci_num_cached && index < ci->ci_num_cached) {
481 bytes = sizeof(sector_t) * (ci->ci_num_cached - index);
482 memmove(&array[index], &array[index + 1], bytes);
486 /* Requires ip_lock. */
487 static void ocfs2_remove_metadata_tree(struct ocfs2_caching_info *ci,
488 struct ocfs2_meta_cache_item *item)
490 mlog(0, "remove block %llu from tree\n",
491 (unsigned long long) item->c_block);
493 rb_erase(&item->c_node, &ci->ci_cache.ci_tree);
497 /* Called when we remove a chunk of metadata from an inode. We don't
498 * bother reverting things to an inlined array in the case of a remove
499 * which moves us back under the limit. */
500 void ocfs2_remove_from_cache(struct inode *inode,
501 struct buffer_head *bh)
504 sector_t block = bh->b_blocknr;
505 struct ocfs2_meta_cache_item *item = NULL;
506 struct ocfs2_inode_info *oi = OCFS2_I(inode);
507 struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
509 spin_lock(&oi->ip_lock);
510 mlog(0, "Inode %llu, remove %llu, items = %u, array = %u\n",
511 (unsigned long long)oi->ip_blkno,
512 (unsigned long long) block, ci->ci_num_cached,
513 oi->ip_flags & OCFS2_INODE_CACHE_INLINE);
515 if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) {
516 index = ocfs2_search_cache_array(ci, block);
518 ocfs2_remove_metadata_array(ci, index);
520 item = ocfs2_search_cache_tree(ci, block);
522 ocfs2_remove_metadata_tree(ci, item);
524 spin_unlock(&oi->ip_lock);
527 kmem_cache_free(ocfs2_uptodate_cachep, item);
530 int __init init_ocfs2_uptodate_cache(void)
532 ocfs2_uptodate_cachep = kmem_cache_create("ocfs2_uptodate",
533 sizeof(struct ocfs2_meta_cache_item),
534 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
535 if (!ocfs2_uptodate_cachep)
538 mlog(0, "%u inlined cache items per inode.\n",
539 OCFS2_INODE_MAX_CACHE_ARRAY);
544 void exit_ocfs2_uptodate_cache(void)
546 if (ocfs2_uptodate_cachep)
547 kmem_cache_destroy(ocfs2_uptodate_cachep);