xref: /openbmc/linux/fs/ocfs2/uptodate.c (revision e868d61272caa648214046a096e5a6bfc068dc8c)
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * uptodate.c
5  *
6  * Tracking the up-to-date-ness of a local buffer_head with respect to
7  * the cluster.
8  *
9  * Copyright (C) 2002, 2004, 2005 Oracle.  All rights reserved.
10  *
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.
15  *
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.
20  *
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.
25  *
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
32  * to cache them.
33  *
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
43  * 	  item for them)
44  *
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
48  */
49 
50 #include <linux/fs.h>
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>
57 
58 #define MLOG_MASK_PREFIX ML_UPTODATE
59 
60 #include <cluster/masklog.h>
61 
62 #include "ocfs2.h"
63 
64 #include "inode.h"
65 #include "uptodate.h"
66 
67 struct ocfs2_meta_cache_item {
68 	struct rb_node	c_node;
69 	sector_t	c_block;
70 };
71 
72 static struct kmem_cache *ocfs2_uptodate_cachep = NULL;
73 
74 void ocfs2_metadata_cache_init(struct inode *inode)
75 {
76 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
77 	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
78 
79 	oi->ip_flags |= OCFS2_INODE_CACHE_INLINE;
80 	ci->ci_num_cached = 0;
81 }
82 
83 /* No lock taken here as 'root' is not expected to be visible to other
84  * processes. */
85 static unsigned int ocfs2_purge_copied_metadata_tree(struct rb_root *root)
86 {
87 	unsigned int purged = 0;
88 	struct rb_node *node;
89 	struct ocfs2_meta_cache_item *item;
90 
91 	while ((node = rb_last(root)) != NULL) {
92 		item = rb_entry(node, struct ocfs2_meta_cache_item, c_node);
93 
94 		mlog(0, "Purge item %llu\n",
95 		     (unsigned long long) item->c_block);
96 
97 		rb_erase(&item->c_node, root);
98 		kmem_cache_free(ocfs2_uptodate_cachep, item);
99 
100 		purged++;
101 	}
102 	return purged;
103 }
104 
105 /* Called from locking and called from ocfs2_clear_inode. Dump the
106  * cache for a given inode.
107  *
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)
112 {
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;
117 
118 	spin_lock(&oi->ip_lock);
119 	tree = !(oi->ip_flags & OCFS2_INODE_CACHE_INLINE);
120 	to_purge = ci->ci_num_cached;
121 
122 	mlog(0, "Purge %u %s items from Inode %llu\n", to_purge,
123 	     tree ? "array" : "tree", (unsigned long long)oi->ip_blkno);
124 
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. */
128 	if (tree)
129 		root = ci->ci_cache.ci_tree;
130 
131 	ocfs2_metadata_cache_init(inode);
132 	spin_unlock(&oi->ip_lock);
133 
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);
141 }
142 
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,
146 				    sector_t item)
147 {
148 	int i;
149 
150 	for (i = 0; i < ci->ci_num_cached; i++) {
151 		if (item == ci->ci_cache.ci_array[i])
152 			return i;
153 	}
154 
155 	return -1;
156 }
157 
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,
162 			sector_t block)
163 {
164 	struct rb_node * n = ci->ci_cache.ci_tree.rb_node;
165 	struct ocfs2_meta_cache_item *item = NULL;
166 
167 	while (n) {
168 		item = rb_entry(n, struct ocfs2_meta_cache_item, c_node);
169 
170 		if (block < item->c_block)
171 			n = n->rb_left;
172 		else if (block > item->c_block)
173 			n = n->rb_right;
174 		else
175 			return item;
176 	}
177 
178 	return NULL;
179 }
180 
181 static int ocfs2_buffer_cached(struct ocfs2_inode_info *oi,
182 			       struct buffer_head *bh)
183 {
184 	int index = -1;
185 	struct ocfs2_meta_cache_item *item = NULL;
186 
187 	spin_lock(&oi->ip_lock);
188 
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));
193 
194 	if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE)
195 		index = ocfs2_search_cache_array(&oi->ip_metadata_cache,
196 						 bh->b_blocknr);
197 	else
198 		item = ocfs2_search_cache_tree(&oi->ip_metadata_cache,
199 					       bh->b_blocknr);
200 
201 	spin_unlock(&oi->ip_lock);
202 
203 	mlog(0, "index = %d, item = %p\n", index, item);
204 
205 	return (index != -1) || (item != NULL);
206 }
207 
208 /* Warning: even if it returns true, this does *not* guarantee that
209  * the block is stored in our inode metadata cache.
210  *
211  * This can be called under lock_buffer()
212  */
213 int ocfs2_buffer_uptodate(struct inode *inode,
214 			  struct buffer_head *bh)
215 {
216 	/* Doesn't matter if the bh is in our cache or not -- if it's
217 	 * not marked uptodate then we know it can't have correct
218 	 * data. */
219 	if (!buffer_uptodate(bh))
220 		return 0;
221 
222 	/* OCFS2 does not allow multiple nodes to be changing the same
223 	 * block at the same time. */
224 	if (buffer_jbd(bh))
225 		return 1;
226 
227 	/* Ok, locally the buffer is marked as up to date, now search
228 	 * our cache to see if we can trust that. */
229 	return ocfs2_buffer_cached(OCFS2_I(inode), bh);
230 }
231 
232 /*
233  * Determine whether a buffer is currently out on a read-ahead request.
234  * ip_io_sem should be held to serialize submitters with the logic here.
235  */
236 int ocfs2_buffer_read_ahead(struct inode *inode,
237 			    struct buffer_head *bh)
238 {
239 	return buffer_locked(bh) && ocfs2_buffer_cached(OCFS2_I(inode), bh);
240 }
241 
242 /* Requires ip_lock */
243 static void ocfs2_append_cache_array(struct ocfs2_caching_info *ci,
244 				     sector_t block)
245 {
246 	BUG_ON(ci->ci_num_cached >= OCFS2_INODE_MAX_CACHE_ARRAY);
247 
248 	mlog(0, "block %llu takes position %u\n", (unsigned long long) block,
249 	     ci->ci_num_cached);
250 
251 	ci->ci_cache.ci_array[ci->ci_num_cached] = block;
252 	ci->ci_num_cached++;
253 }
254 
255 /* By now the caller should have checked that the item does *not*
256  * exist in the tree.
257  * Requires ip_lock. */
258 static void __ocfs2_insert_cache_tree(struct ocfs2_caching_info *ci,
259 				      struct ocfs2_meta_cache_item *new)
260 {
261 	sector_t block = new->c_block;
262 	struct rb_node *parent = NULL;
263 	struct rb_node **p = &ci->ci_cache.ci_tree.rb_node;
264 	struct ocfs2_meta_cache_item *tmp;
265 
266 	mlog(0, "Insert block %llu num = %u\n", (unsigned long long) block,
267 	     ci->ci_num_cached);
268 
269 	while(*p) {
270 		parent = *p;
271 
272 		tmp = rb_entry(parent, struct ocfs2_meta_cache_item, c_node);
273 
274 		if (block < tmp->c_block)
275 			p = &(*p)->rb_left;
276 		else if (block > tmp->c_block)
277 			p = &(*p)->rb_right;
278 		else {
279 			/* This should never happen! */
280 			mlog(ML_ERROR, "Duplicate block %llu cached!\n",
281 			     (unsigned long long) block);
282 			BUG();
283 		}
284 	}
285 
286 	rb_link_node(&new->c_node, parent, p);
287 	rb_insert_color(&new->c_node, &ci->ci_cache.ci_tree);
288 	ci->ci_num_cached++;
289 }
290 
291 static inline int ocfs2_insert_can_use_array(struct ocfs2_inode_info *oi,
292 					     struct ocfs2_caching_info *ci)
293 {
294 	assert_spin_locked(&oi->ip_lock);
295 
296 	return (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) &&
297 		(ci->ci_num_cached < OCFS2_INODE_MAX_CACHE_ARRAY);
298 }
299 
300 /* tree should be exactly OCFS2_INODE_MAX_CACHE_ARRAY wide. NULL the
301  * pointers in tree after we use them - this allows caller to detect
302  * when to free in case of error. */
303 static void ocfs2_expand_cache(struct ocfs2_inode_info *oi,
304 			       struct ocfs2_meta_cache_item **tree)
305 {
306 	int i;
307 	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
308 
309 	mlog_bug_on_msg(ci->ci_num_cached != OCFS2_INODE_MAX_CACHE_ARRAY,
310 			"Inode %llu, num cached = %u, should be %u\n",
311 			(unsigned long long)oi->ip_blkno, ci->ci_num_cached,
312 			OCFS2_INODE_MAX_CACHE_ARRAY);
313 	mlog_bug_on_msg(!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE),
314 			"Inode %llu not marked as inline anymore!\n",
315 			(unsigned long long)oi->ip_blkno);
316 	assert_spin_locked(&oi->ip_lock);
317 
318 	/* Be careful to initialize the tree members *first* because
319 	 * once the ci_tree is used, the array is junk... */
320 	for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++)
321 		tree[i]->c_block = ci->ci_cache.ci_array[i];
322 
323 	oi->ip_flags &= ~OCFS2_INODE_CACHE_INLINE;
324 	ci->ci_cache.ci_tree = RB_ROOT;
325 	/* this will be set again by __ocfs2_insert_cache_tree */
326 	ci->ci_num_cached = 0;
327 
328 	for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) {
329 		__ocfs2_insert_cache_tree(ci, tree[i]);
330 		tree[i] = NULL;
331 	}
332 
333 	mlog(0, "Expanded %llu to a tree cache: flags 0x%x, num = %u\n",
334 	     (unsigned long long)oi->ip_blkno, oi->ip_flags, ci->ci_num_cached);
335 }
336 
337 /* Slow path function - memory allocation is necessary. See the
338  * comment above ocfs2_set_buffer_uptodate for more information. */
339 static void __ocfs2_set_buffer_uptodate(struct ocfs2_inode_info *oi,
340 					sector_t block,
341 					int expand_tree)
342 {
343 	int i;
344 	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
345 	struct ocfs2_meta_cache_item *new = NULL;
346 	struct ocfs2_meta_cache_item *tree[OCFS2_INODE_MAX_CACHE_ARRAY] =
347 		{ NULL, };
348 
349 	mlog(0, "Inode %llu, block %llu, expand = %d\n",
350 	     (unsigned long long)oi->ip_blkno,
351 	     (unsigned long long)block, expand_tree);
352 
353 	new = kmem_cache_alloc(ocfs2_uptodate_cachep, GFP_NOFS);
354 	if (!new) {
355 		mlog_errno(-ENOMEM);
356 		return;
357 	}
358 	new->c_block = block;
359 
360 	if (expand_tree) {
361 		/* Do *not* allocate an array here - the removal code
362 		 * has no way of tracking that. */
363 		for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) {
364 			tree[i] = kmem_cache_alloc(ocfs2_uptodate_cachep,
365 						   GFP_NOFS);
366 			if (!tree[i]) {
367 				mlog_errno(-ENOMEM);
368 				goto out_free;
369 			}
370 
371 			/* These are initialized in ocfs2_expand_cache! */
372 		}
373 	}
374 
375 	spin_lock(&oi->ip_lock);
376 	if (ocfs2_insert_can_use_array(oi, ci)) {
377 		mlog(0, "Someone cleared the tree underneath us\n");
378 		/* Ok, items were removed from the cache in between
379 		 * locks. Detect this and revert back to the fast path */
380 		ocfs2_append_cache_array(ci, block);
381 		spin_unlock(&oi->ip_lock);
382 		goto out_free;
383 	}
384 
385 	if (expand_tree)
386 		ocfs2_expand_cache(oi, tree);
387 
388 	__ocfs2_insert_cache_tree(ci, new);
389 	spin_unlock(&oi->ip_lock);
390 
391 	new = NULL;
392 out_free:
393 	if (new)
394 		kmem_cache_free(ocfs2_uptodate_cachep, new);
395 
396 	/* If these were used, then ocfs2_expand_cache re-set them to
397 	 * NULL for us. */
398 	if (tree[0]) {
399 		for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++)
400 			if (tree[i])
401 				kmem_cache_free(ocfs2_uptodate_cachep,
402 						tree[i]);
403 	}
404 }
405 
406 /* Item insertion is guarded by ip_io_mutex, so the insertion path takes
407  * advantage of this by not rechecking for a duplicate insert during
408  * the slow case. Additionally, if the cache needs to be bumped up to
409  * a tree, the code will not recheck after acquiring the lock --
410  * multiple paths cannot be expanding to a tree at the same time.
411  *
412  * The slow path takes into account that items can be removed
413  * (including the whole tree wiped and reset) when this process it out
414  * allocating memory. In those cases, it reverts back to the fast
415  * path.
416  *
417  * Note that this function may actually fail to insert the block if
418  * memory cannot be allocated. This is not fatal however (but may
419  * result in a performance penalty)
420  *
421  * Readahead buffers can be passed in here before the I/O request is
422  * completed.
423  */
424 void ocfs2_set_buffer_uptodate(struct inode *inode,
425 			       struct buffer_head *bh)
426 {
427 	int expand;
428 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
429 	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
430 
431 	/* The block may very well exist in our cache already, so avoid
432 	 * doing any more work in that case. */
433 	if (ocfs2_buffer_cached(oi, bh))
434 		return;
435 
436 	mlog(0, "Inode %llu, inserting block %llu\n",
437 	     (unsigned long long)oi->ip_blkno,
438 	     (unsigned long long)bh->b_blocknr);
439 
440 	/* No need to recheck under spinlock - insertion is guarded by
441 	 * ip_io_mutex */
442 	spin_lock(&oi->ip_lock);
443 	if (ocfs2_insert_can_use_array(oi, ci)) {
444 		/* Fast case - it's an array and there's a free
445 		 * spot. */
446 		ocfs2_append_cache_array(ci, bh->b_blocknr);
447 		spin_unlock(&oi->ip_lock);
448 		return;
449 	}
450 
451 	expand = 0;
452 	if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) {
453 		/* We need to bump things up to a tree. */
454 		expand = 1;
455 	}
456 	spin_unlock(&oi->ip_lock);
457 
458 	__ocfs2_set_buffer_uptodate(oi, bh->b_blocknr, expand);
459 }
460 
461 /* Called against a newly allocated buffer. Most likely nobody should
462  * be able to read this sort of metadata while it's still being
463  * allocated, but this is careful to take ip_io_mutex anyway. */
464 void ocfs2_set_new_buffer_uptodate(struct inode *inode,
465 				   struct buffer_head *bh)
466 {
467 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
468 
469 	/* This should definitely *not* exist in our cache */
470 	BUG_ON(ocfs2_buffer_cached(oi, bh));
471 
472 	set_buffer_uptodate(bh);
473 
474 	mutex_lock(&oi->ip_io_mutex);
475 	ocfs2_set_buffer_uptodate(inode, bh);
476 	mutex_unlock(&oi->ip_io_mutex);
477 }
478 
479 /* Requires ip_lock. */
480 static void ocfs2_remove_metadata_array(struct ocfs2_caching_info *ci,
481 					int index)
482 {
483 	sector_t *array = ci->ci_cache.ci_array;
484 	int bytes;
485 
486 	BUG_ON(index < 0 || index >= OCFS2_INODE_MAX_CACHE_ARRAY);
487 	BUG_ON(index >= ci->ci_num_cached);
488 	BUG_ON(!ci->ci_num_cached);
489 
490 	mlog(0, "remove index %d (num_cached = %u\n", index,
491 	     ci->ci_num_cached);
492 
493 	ci->ci_num_cached--;
494 
495 	/* don't need to copy if the array is now empty, or if we
496 	 * removed at the tail */
497 	if (ci->ci_num_cached && index < ci->ci_num_cached) {
498 		bytes = sizeof(sector_t) * (ci->ci_num_cached - index);
499 		memmove(&array[index], &array[index + 1], bytes);
500 	}
501 }
502 
503 /* Requires ip_lock. */
504 static void ocfs2_remove_metadata_tree(struct ocfs2_caching_info *ci,
505 				       struct ocfs2_meta_cache_item *item)
506 {
507 	mlog(0, "remove block %llu from tree\n",
508 	     (unsigned long long) item->c_block);
509 
510 	rb_erase(&item->c_node, &ci->ci_cache.ci_tree);
511 	ci->ci_num_cached--;
512 }
513 
514 /* Called when we remove a chunk of metadata from an inode. We don't
515  * bother reverting things to an inlined array in the case of a remove
516  * which moves us back under the limit. */
517 void ocfs2_remove_from_cache(struct inode *inode,
518 			     struct buffer_head *bh)
519 {
520 	int index;
521 	sector_t block = bh->b_blocknr;
522 	struct ocfs2_meta_cache_item *item = NULL;
523 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
524 	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
525 
526 	spin_lock(&oi->ip_lock);
527 	mlog(0, "Inode %llu, remove %llu, items = %u, array = %u\n",
528 	     (unsigned long long)oi->ip_blkno,
529 	     (unsigned long long) block, ci->ci_num_cached,
530 	     oi->ip_flags & OCFS2_INODE_CACHE_INLINE);
531 
532 	if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) {
533 		index = ocfs2_search_cache_array(ci, block);
534 		if (index != -1)
535 			ocfs2_remove_metadata_array(ci, index);
536 	} else {
537 		item = ocfs2_search_cache_tree(ci, block);
538 		if (item)
539 			ocfs2_remove_metadata_tree(ci, item);
540 	}
541 	spin_unlock(&oi->ip_lock);
542 
543 	if (item)
544 		kmem_cache_free(ocfs2_uptodate_cachep, item);
545 }
546 
547 int __init init_ocfs2_uptodate_cache(void)
548 {
549 	ocfs2_uptodate_cachep = kmem_cache_create("ocfs2_uptodate",
550 				  sizeof(struct ocfs2_meta_cache_item),
551 				  0, SLAB_HWCACHE_ALIGN, NULL, NULL);
552 	if (!ocfs2_uptodate_cachep)
553 		return -ENOMEM;
554 
555 	mlog(0, "%u inlined cache items per inode.\n",
556 	     OCFS2_INODE_MAX_CACHE_ARRAY);
557 
558 	return 0;
559 }
560 
561 void exit_ocfs2_uptodate_cache(void)
562 {
563 	if (ocfs2_uptodate_cachep)
564 		kmem_cache_destroy(ocfs2_uptodate_cachep);
565 }
566