2 * f2fs extent cache support
4 * Copyright (c) 2015 Motorola Mobility
5 * Copyright (c) 2015 Samsung Electronics
6 * Authors: Jaegeuk Kim <jaegeuk@kernel.org>
7 * Chao Yu <chao2.yu@samsung.com>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License version 2 as
11 * published by the Free Software Foundation.
15 #include <linux/f2fs_fs.h>
19 #include <trace/events/f2fs.h>
21 static struct kmem_cache *extent_tree_slab;
22 static struct kmem_cache *extent_node_slab;
24 static struct extent_node *__attach_extent_node(struct f2fs_sb_info *sbi,
25 struct extent_tree *et, struct extent_info *ei,
26 struct rb_node *parent, struct rb_node **p)
28 struct extent_node *en;
30 en = kmem_cache_alloc(extent_node_slab, GFP_ATOMIC);
35 INIT_LIST_HEAD(&en->list);
37 rb_link_node(&en->rb_node, parent, p);
38 rb_insert_color(&en->rb_node, &et->root);
40 atomic_inc(&sbi->total_ext_node);
44 static void __detach_extent_node(struct f2fs_sb_info *sbi,
45 struct extent_tree *et, struct extent_node *en)
47 rb_erase(&en->rb_node, &et->root);
49 atomic_dec(&sbi->total_ext_node);
51 if (et->cached_en == en)
55 static struct extent_tree *__grab_extent_tree(struct inode *inode)
57 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
58 struct extent_tree *et;
59 nid_t ino = inode->i_ino;
61 down_write(&sbi->extent_tree_lock);
62 et = radix_tree_lookup(&sbi->extent_tree_root, ino);
64 et = f2fs_kmem_cache_alloc(extent_tree_slab, GFP_NOFS);
65 f2fs_radix_tree_insert(&sbi->extent_tree_root, ino, et);
66 memset(et, 0, sizeof(struct extent_tree));
70 rwlock_init(&et->lock);
71 atomic_set(&et->refcount, 0);
73 sbi->total_ext_tree++;
75 atomic_inc(&et->refcount);
76 up_write(&sbi->extent_tree_lock);
78 /* never died until evict_inode */
79 F2FS_I(inode)->extent_tree = et;
84 static struct extent_node *__lookup_extent_tree(struct f2fs_sb_info *sbi,
85 struct extent_tree *et, unsigned int fofs)
87 struct rb_node *node = et->root.rb_node;
88 struct extent_node *en = et->cached_en;
91 struct extent_info *cei = &en->ei;
93 if (cei->fofs <= fofs && cei->fofs + cei->len > fofs) {
94 stat_inc_cached_node_hit(sbi);
100 en = rb_entry(node, struct extent_node, rb_node);
102 if (fofs < en->ei.fofs) {
103 node = node->rb_left;
104 } else if (fofs >= en->ei.fofs + en->ei.len) {
105 node = node->rb_right;
107 stat_inc_rbtree_node_hit(sbi);
114 static struct extent_node *__init_extent_tree(struct f2fs_sb_info *sbi,
115 struct extent_tree *et, struct extent_info *ei)
117 struct rb_node **p = &et->root.rb_node;
118 struct extent_node *en;
120 en = __attach_extent_node(sbi, et, ei, NULL, p);
124 et->largest = en->ei;
129 static unsigned int __free_extent_tree(struct f2fs_sb_info *sbi,
130 struct extent_tree *et, bool free_all)
132 struct rb_node *node, *next;
133 struct extent_node *en;
134 unsigned int count = et->count;
136 node = rb_first(&et->root);
138 next = rb_next(node);
139 en = rb_entry(node, struct extent_node, rb_node);
142 spin_lock(&sbi->extent_lock);
143 if (!list_empty(&en->list))
144 list_del_init(&en->list);
145 spin_unlock(&sbi->extent_lock);
148 if (free_all || list_empty(&en->list)) {
149 __detach_extent_node(sbi, et, en);
150 kmem_cache_free(extent_node_slab, en);
155 return count - et->count;
158 static void __drop_largest_extent(struct inode *inode,
159 pgoff_t fofs, unsigned int len)
161 struct extent_info *largest = &F2FS_I(inode)->extent_tree->largest;
163 if (fofs < largest->fofs + largest->len && fofs + len > largest->fofs)
167 void f2fs_drop_largest_extent(struct inode *inode, pgoff_t fofs)
169 if (!f2fs_may_extent_tree(inode))
172 __drop_largest_extent(inode, fofs, 1);
175 void f2fs_init_extent_tree(struct inode *inode, struct f2fs_extent *i_ext)
177 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
178 struct extent_tree *et;
179 struct extent_node *en;
180 struct extent_info ei;
182 if (!f2fs_may_extent_tree(inode))
185 et = __grab_extent_tree(inode);
187 if (!i_ext || le32_to_cpu(i_ext->len) < F2FS_MIN_EXTENT_LEN)
190 set_extent_info(&ei, le32_to_cpu(i_ext->fofs),
191 le32_to_cpu(i_ext->blk), le32_to_cpu(i_ext->len));
193 write_lock(&et->lock);
197 en = __init_extent_tree(sbi, et, &ei);
199 spin_lock(&sbi->extent_lock);
200 list_add_tail(&en->list, &sbi->extent_list);
201 spin_unlock(&sbi->extent_lock);
204 write_unlock(&et->lock);
207 static bool f2fs_lookup_extent_tree(struct inode *inode, pgoff_t pgofs,
208 struct extent_info *ei)
210 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
211 struct extent_tree *et = F2FS_I(inode)->extent_tree;
212 struct extent_node *en;
215 f2fs_bug_on(sbi, !et);
217 trace_f2fs_lookup_extent_tree_start(inode, pgofs);
219 read_lock(&et->lock);
221 if (et->largest.fofs <= pgofs &&
222 et->largest.fofs + et->largest.len > pgofs) {
225 stat_inc_largest_node_hit(sbi);
229 en = __lookup_extent_tree(sbi, et, pgofs);
232 spin_lock(&sbi->extent_lock);
233 if (!list_empty(&en->list))
234 list_move_tail(&en->list, &sbi->extent_list);
236 spin_unlock(&sbi->extent_lock);
240 stat_inc_total_hit(sbi);
241 read_unlock(&et->lock);
243 trace_f2fs_lookup_extent_tree_end(inode, pgofs, ei);
249 * lookup extent at @fofs, if hit, return the extent
250 * if not, return NULL and
251 * @prev_ex: extent before fofs
252 * @next_ex: extent after fofs
253 * @insert_p: insert point for new extent at fofs
254 * in order to simpfy the insertion after.
255 * tree must stay unchanged between lookup and insertion.
257 static struct extent_node *__lookup_extent_tree_ret(struct extent_tree *et,
259 struct extent_node **prev_ex,
260 struct extent_node **next_ex,
261 struct rb_node ***insert_p,
262 struct rb_node **insert_parent)
264 struct rb_node **pnode = &et->root.rb_node;
265 struct rb_node *parent = NULL, *tmp_node;
266 struct extent_node *en = et->cached_en;
269 *insert_parent = NULL;
273 if (RB_EMPTY_ROOT(&et->root))
277 struct extent_info *cei = &en->ei;
279 if (cei->fofs <= fofs && cei->fofs + cei->len > fofs)
280 goto lookup_neighbors;
285 en = rb_entry(*pnode, struct extent_node, rb_node);
287 if (fofs < en->ei.fofs)
288 pnode = &(*pnode)->rb_left;
289 else if (fofs >= en->ei.fofs + en->ei.len)
290 pnode = &(*pnode)->rb_right;
292 goto lookup_neighbors;
296 *insert_parent = parent;
298 en = rb_entry(parent, struct extent_node, rb_node);
300 if (parent && fofs > en->ei.fofs)
301 tmp_node = rb_next(parent);
302 *next_ex = tmp_node ?
303 rb_entry(tmp_node, struct extent_node, rb_node) : NULL;
306 if (parent && fofs < en->ei.fofs)
307 tmp_node = rb_prev(parent);
308 *prev_ex = tmp_node ?
309 rb_entry(tmp_node, struct extent_node, rb_node) : NULL;
313 if (fofs == en->ei.fofs) {
314 /* lookup prev node for merging backward later */
315 tmp_node = rb_prev(&en->rb_node);
316 *prev_ex = tmp_node ?
317 rb_entry(tmp_node, struct extent_node, rb_node) : NULL;
319 if (fofs == en->ei.fofs + en->ei.len - 1) {
320 /* lookup next node for merging frontward later */
321 tmp_node = rb_next(&en->rb_node);
322 *next_ex = tmp_node ?
323 rb_entry(tmp_node, struct extent_node, rb_node) : NULL;
328 static struct extent_node *__try_merge_extent_node(struct f2fs_sb_info *sbi,
329 struct extent_tree *et, struct extent_info *ei,
330 struct extent_node **den,
331 struct extent_node *prev_ex,
332 struct extent_node *next_ex)
334 struct extent_node *en = NULL;
336 if (prev_ex && __is_back_mergeable(ei, &prev_ex->ei)) {
337 prev_ex->ei.len += ei->len;
342 if (next_ex && __is_front_mergeable(ei, &next_ex->ei)) {
344 __detach_extent_node(sbi, et, prev_ex);
347 next_ex->ei.fofs = ei->fofs;
348 next_ex->ei.blk = ei->blk;
349 next_ex->ei.len += ei->len;
354 __try_update_largest_extent(et, en);
360 static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi,
361 struct extent_tree *et, struct extent_info *ei,
362 struct rb_node **insert_p,
363 struct rb_node *insert_parent)
365 struct rb_node **p = &et->root.rb_node;
366 struct rb_node *parent = NULL;
367 struct extent_node *en = NULL;
369 if (insert_p && insert_parent) {
370 parent = insert_parent;
377 en = rb_entry(parent, struct extent_node, rb_node);
379 if (ei->fofs < en->ei.fofs)
381 else if (ei->fofs >= en->ei.fofs + en->ei.len)
387 en = __attach_extent_node(sbi, et, ei, parent, p);
391 __try_update_largest_extent(et, en);
396 static unsigned int f2fs_update_extent_tree_range(struct inode *inode,
397 pgoff_t fofs, block_t blkaddr, unsigned int len)
399 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
400 struct extent_tree *et = F2FS_I(inode)->extent_tree;
401 struct extent_node *en = NULL, *en1 = NULL;
402 struct extent_node *prev_en = NULL, *next_en = NULL;
403 struct extent_info ei, dei, prev;
404 struct rb_node **insert_p = NULL, *insert_parent = NULL;
405 unsigned int end = fofs + len;
406 unsigned int pos = (unsigned int)fofs;
411 trace_f2fs_update_extent_tree_range(inode, fofs, blkaddr, len);
413 write_lock(&et->lock);
415 if (is_inode_flag_set(F2FS_I(inode), FI_NO_EXTENT)) {
416 write_unlock(&et->lock);
424 * drop largest extent before lookup, in case it's already
425 * been shrunk from extent tree
427 __drop_largest_extent(inode, fofs, len);
429 /* 1. lookup first extent node in range [fofs, fofs + len - 1] */
430 en = __lookup_extent_tree_ret(et, fofs, &prev_en, &next_en,
431 &insert_p, &insert_parent);
435 /* 2. invlidate all extent nodes in range [fofs, fofs + len - 1] */
436 while (en && en->ei.fofs < end) {
437 unsigned int org_end;
438 int parts = 0; /* # of parts current extent split into */
440 next_en = en1 = NULL;
443 org_end = dei.fofs + dei.len;
444 f2fs_bug_on(sbi, pos >= org_end);
446 if (pos > dei.fofs && pos - dei.fofs >= F2FS_MIN_EXTENT_LEN) {
447 en->ei.len = pos - en->ei.fofs;
452 if (end < org_end && org_end - end >= F2FS_MIN_EXTENT_LEN) {
454 set_extent_info(&ei, end,
455 end - dei.fofs + dei.blk,
457 en1 = __insert_extent_tree(sbi, et, &ei,
462 en->ei.blk += end - dei.fofs;
463 en->ei.len -= end - dei.fofs;
470 struct rb_node *node = rb_next(&en->rb_node);
473 rb_entry(node, struct extent_node, rb_node)
478 __try_update_largest_extent(et, en);
480 __detach_extent_node(sbi, et, en);
483 * if original extent is split into zero or two parts, extent
484 * tree has been altered by deletion or insertion, therefore
485 * invalidate pointers regard to tree.
489 insert_parent = NULL;
492 /* update in global extent list */
493 spin_lock(&sbi->extent_lock);
494 if (!parts && !list_empty(&en->list))
497 list_add_tail(&en1->list, &sbi->extent_list);
498 spin_unlock(&sbi->extent_lock);
500 /* release extent node */
502 kmem_cache_free(extent_node_slab, en);
507 /* 3. update extent in extent cache */
509 struct extent_node *den = NULL;
511 set_extent_info(&ei, fofs, blkaddr, len);
512 en1 = __try_merge_extent_node(sbi, et, &ei, &den,
515 en1 = __insert_extent_tree(sbi, et, &ei,
516 insert_p, insert_parent);
518 /* give up extent_cache, if split and small updates happen */
520 prev.len < F2FS_MIN_EXTENT_LEN &&
521 et->largest.len < F2FS_MIN_EXTENT_LEN) {
523 set_inode_flag(F2FS_I(inode), FI_NO_EXTENT);
526 spin_lock(&sbi->extent_lock);
528 if (list_empty(&en1->list))
529 list_add_tail(&en1->list, &sbi->extent_list);
531 list_move_tail(&en1->list, &sbi->extent_list);
533 if (den && !list_empty(&den->list))
534 list_del(&den->list);
535 spin_unlock(&sbi->extent_lock);
538 kmem_cache_free(extent_node_slab, den);
541 if (is_inode_flag_set(F2FS_I(inode), FI_NO_EXTENT))
542 __free_extent_tree(sbi, et, true);
544 write_unlock(&et->lock);
546 return !__is_extent_same(&prev, &et->largest);
549 unsigned int f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink)
551 struct extent_tree *treevec[EXT_TREE_VEC_SIZE];
552 struct extent_node *en, *tmp;
553 unsigned long ino = F2FS_ROOT_INO(sbi);
554 struct radix_tree_root *root = &sbi->extent_tree_root;
556 unsigned int node_cnt = 0, tree_cnt = 0;
559 if (!test_opt(sbi, EXTENT_CACHE))
562 if (!down_write_trylock(&sbi->extent_tree_lock))
565 /* 1. remove unreferenced extent tree */
566 while ((found = radix_tree_gang_lookup(root,
567 (void **)treevec, ino, EXT_TREE_VEC_SIZE))) {
570 ino = treevec[found - 1]->ino + 1;
571 for (i = 0; i < found; i++) {
572 struct extent_tree *et = treevec[i];
574 if (!atomic_read(&et->refcount)) {
575 write_lock(&et->lock);
576 node_cnt += __free_extent_tree(sbi, et, true);
577 write_unlock(&et->lock);
579 radix_tree_delete(root, et->ino);
580 kmem_cache_free(extent_tree_slab, et);
581 sbi->total_ext_tree--;
584 if (node_cnt + tree_cnt >= nr_shrink)
589 up_write(&sbi->extent_tree_lock);
591 /* 2. remove LRU extent entries */
592 if (!down_write_trylock(&sbi->extent_tree_lock))
595 remained = nr_shrink - (node_cnt + tree_cnt);
597 spin_lock(&sbi->extent_lock);
598 list_for_each_entry_safe(en, tmp, &sbi->extent_list, list) {
601 list_del_init(&en->list);
603 spin_unlock(&sbi->extent_lock);
606 * reset ino for searching victims from beginning of global extent tree.
608 ino = F2FS_ROOT_INO(sbi);
610 while ((found = radix_tree_gang_lookup(root,
611 (void **)treevec, ino, EXT_TREE_VEC_SIZE))) {
614 ino = treevec[found - 1]->ino + 1;
615 for (i = 0; i < found; i++) {
616 struct extent_tree *et = treevec[i];
618 write_lock(&et->lock);
619 node_cnt += __free_extent_tree(sbi, et, false);
620 write_unlock(&et->lock);
622 if (node_cnt + tree_cnt >= nr_shrink)
627 up_write(&sbi->extent_tree_lock);
629 trace_f2fs_shrink_extent_tree(sbi, node_cnt, tree_cnt);
631 return node_cnt + tree_cnt;
634 unsigned int f2fs_destroy_extent_node(struct inode *inode)
636 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
637 struct extent_tree *et = F2FS_I(inode)->extent_tree;
638 unsigned int node_cnt = 0;
643 write_lock(&et->lock);
644 node_cnt = __free_extent_tree(sbi, et, true);
645 write_unlock(&et->lock);
650 void f2fs_destroy_extent_tree(struct inode *inode)
652 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
653 struct extent_tree *et = F2FS_I(inode)->extent_tree;
654 unsigned int node_cnt = 0;
659 if (inode->i_nlink && !is_bad_inode(inode) && et->count) {
660 atomic_dec(&et->refcount);
664 /* free all extent info belong to this extent tree */
665 node_cnt = f2fs_destroy_extent_node(inode);
667 /* delete extent tree entry in radix tree */
668 down_write(&sbi->extent_tree_lock);
669 atomic_dec(&et->refcount);
670 f2fs_bug_on(sbi, atomic_read(&et->refcount) || et->count);
671 radix_tree_delete(&sbi->extent_tree_root, inode->i_ino);
672 kmem_cache_free(extent_tree_slab, et);
673 sbi->total_ext_tree--;
674 up_write(&sbi->extent_tree_lock);
676 F2FS_I(inode)->extent_tree = NULL;
678 trace_f2fs_destroy_extent_tree(inode, node_cnt);
681 bool f2fs_lookup_extent_cache(struct inode *inode, pgoff_t pgofs,
682 struct extent_info *ei)
684 if (!f2fs_may_extent_tree(inode))
687 return f2fs_lookup_extent_tree(inode, pgofs, ei);
690 void f2fs_update_extent_cache(struct dnode_of_data *dn)
692 struct f2fs_inode_info *fi = F2FS_I(dn->inode);
695 if (!f2fs_may_extent_tree(dn->inode))
698 f2fs_bug_on(F2FS_I_SB(dn->inode), dn->data_blkaddr == NEW_ADDR);
701 fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
704 if (f2fs_update_extent_tree_range(dn->inode, fofs, dn->data_blkaddr, 1))
708 void f2fs_update_extent_cache_range(struct dnode_of_data *dn,
709 pgoff_t fofs, block_t blkaddr, unsigned int len)
712 if (!f2fs_may_extent_tree(dn->inode))
715 if (f2fs_update_extent_tree_range(dn->inode, fofs, blkaddr, len))
719 void init_extent_cache_info(struct f2fs_sb_info *sbi)
721 INIT_RADIX_TREE(&sbi->extent_tree_root, GFP_NOIO);
722 init_rwsem(&sbi->extent_tree_lock);
723 INIT_LIST_HEAD(&sbi->extent_list);
724 spin_lock_init(&sbi->extent_lock);
725 sbi->total_ext_tree = 0;
726 atomic_set(&sbi->total_ext_node, 0);
729 int __init create_extent_cache(void)
731 extent_tree_slab = f2fs_kmem_cache_create("f2fs_extent_tree",
732 sizeof(struct extent_tree));
733 if (!extent_tree_slab)
735 extent_node_slab = f2fs_kmem_cache_create("f2fs_extent_node",
736 sizeof(struct extent_node));
737 if (!extent_node_slab) {
738 kmem_cache_destroy(extent_tree_slab);
744 void destroy_extent_cache(void)
746 kmem_cache_destroy(extent_node_slab);
747 kmem_cache_destroy(extent_tree_slab);