]>
Commit | Line | Data |
---|---|---|
0a8165d7 | 1 | /* |
e05df3b1 JK |
2 | * fs/f2fs/node.c |
3 | * | |
4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
5 | * http://www.samsung.com/ | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
11 | #include <linux/fs.h> | |
12 | #include <linux/f2fs_fs.h> | |
13 | #include <linux/mpage.h> | |
14 | #include <linux/backing-dev.h> | |
15 | #include <linux/blkdev.h> | |
16 | #include <linux/pagevec.h> | |
17 | #include <linux/swap.h> | |
18 | ||
19 | #include "f2fs.h" | |
20 | #include "node.h" | |
21 | #include "segment.h" | |
9e4ded3f | 22 | #include "trace.h" |
51dd6249 | 23 | #include <trace/events/f2fs.h> |
e05df3b1 | 24 | |
f978f5a0 GZ |
25 | #define on_build_free_nids(nmi) mutex_is_locked(&nm_i->build_lock) |
26 | ||
e05df3b1 JK |
27 | static struct kmem_cache *nat_entry_slab; |
28 | static struct kmem_cache *free_nid_slab; | |
aec71382 | 29 | static struct kmem_cache *nat_entry_set_slab; |
e05df3b1 | 30 | |
6fb03f3a | 31 | bool available_free_memory(struct f2fs_sb_info *sbi, int type) |
cdfc41c1 | 32 | { |
6fb03f3a | 33 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
cdfc41c1 | 34 | struct sysinfo val; |
e5e7ea3c | 35 | unsigned long avail_ram; |
cdfc41c1 | 36 | unsigned long mem_size = 0; |
6fb03f3a | 37 | bool res = false; |
cdfc41c1 JK |
38 | |
39 | si_meminfo(&val); | |
e5e7ea3c JK |
40 | |
41 | /* only uses low memory */ | |
42 | avail_ram = val.totalram - val.totalhigh; | |
43 | ||
429511cd CY |
44 | /* |
45 | * give 25%, 25%, 50%, 50%, 50% memory for each components respectively | |
46 | */ | |
6fb03f3a | 47 | if (type == FREE_NIDS) { |
e5e7ea3c JK |
48 | mem_size = (nm_i->fcnt * sizeof(struct free_nid)) >> |
49 | PAGE_CACHE_SHIFT; | |
50 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); | |
6fb03f3a | 51 | } else if (type == NAT_ENTRIES) { |
e5e7ea3c JK |
52 | mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >> |
53 | PAGE_CACHE_SHIFT; | |
54 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); | |
a1257023 JK |
55 | } else if (type == DIRTY_DENTS) { |
56 | if (sbi->sb->s_bdi->wb.dirty_exceeded) | |
57 | return false; | |
58 | mem_size = get_pages(sbi, F2FS_DIRTY_DENTS); | |
59 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); | |
e5e7ea3c JK |
60 | } else if (type == INO_ENTRIES) { |
61 | int i; | |
62 | ||
e5e7ea3c | 63 | for (i = 0; i <= UPDATE_INO; i++) |
67298804 CY |
64 | mem_size += (sbi->im[i].ino_num * |
65 | sizeof(struct ino_entry)) >> PAGE_CACHE_SHIFT; | |
e5e7ea3c | 66 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
429511cd | 67 | } else if (type == EXTENT_CACHE) { |
7441ccef JK |
68 | mem_size = (atomic_read(&sbi->total_ext_tree) * |
69 | sizeof(struct extent_tree) + | |
429511cd CY |
70 | atomic_read(&sbi->total_ext_node) * |
71 | sizeof(struct extent_node)) >> PAGE_CACHE_SHIFT; | |
72 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); | |
1e84371f | 73 | } else { |
a88a341a | 74 | if (sbi->sb->s_bdi->wb.dirty_exceeded) |
1e84371f | 75 | return false; |
6fb03f3a JK |
76 | } |
77 | return res; | |
cdfc41c1 JK |
78 | } |
79 | ||
e05df3b1 JK |
80 | static void clear_node_page_dirty(struct page *page) |
81 | { | |
82 | struct address_space *mapping = page->mapping; | |
e05df3b1 JK |
83 | unsigned int long flags; |
84 | ||
85 | if (PageDirty(page)) { | |
86 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
87 | radix_tree_tag_clear(&mapping->page_tree, | |
88 | page_index(page), | |
89 | PAGECACHE_TAG_DIRTY); | |
90 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
91 | ||
92 | clear_page_dirty_for_io(page); | |
4081363f | 93 | dec_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_NODES); |
e05df3b1 JK |
94 | } |
95 | ClearPageUptodate(page); | |
96 | } | |
97 | ||
98 | static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
99 | { | |
100 | pgoff_t index = current_nat_addr(sbi, nid); | |
101 | return get_meta_page(sbi, index); | |
102 | } | |
103 | ||
104 | static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
105 | { | |
106 | struct page *src_page; | |
107 | struct page *dst_page; | |
108 | pgoff_t src_off; | |
109 | pgoff_t dst_off; | |
110 | void *src_addr; | |
111 | void *dst_addr; | |
112 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
113 | ||
114 | src_off = current_nat_addr(sbi, nid); | |
115 | dst_off = next_nat_addr(sbi, src_off); | |
116 | ||
117 | /* get current nat block page with lock */ | |
118 | src_page = get_meta_page(sbi, src_off); | |
e05df3b1 | 119 | dst_page = grab_meta_page(sbi, dst_off); |
9850cf4a | 120 | f2fs_bug_on(sbi, PageDirty(src_page)); |
e05df3b1 JK |
121 | |
122 | src_addr = page_address(src_page); | |
123 | dst_addr = page_address(dst_page); | |
124 | memcpy(dst_addr, src_addr, PAGE_CACHE_SIZE); | |
125 | set_page_dirty(dst_page); | |
126 | f2fs_put_page(src_page, 1); | |
127 | ||
128 | set_to_next_nat(nm_i, nid); | |
129 | ||
130 | return dst_page; | |
131 | } | |
132 | ||
e05df3b1 JK |
133 | static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n) |
134 | { | |
135 | return radix_tree_lookup(&nm_i->nat_root, n); | |
136 | } | |
137 | ||
138 | static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i, | |
139 | nid_t start, unsigned int nr, struct nat_entry **ep) | |
140 | { | |
141 | return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr); | |
142 | } | |
143 | ||
144 | static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e) | |
145 | { | |
146 | list_del(&e->list); | |
147 | radix_tree_delete(&nm_i->nat_root, nat_get_nid(e)); | |
148 | nm_i->nat_cnt--; | |
149 | kmem_cache_free(nat_entry_slab, e); | |
150 | } | |
151 | ||
309cc2b6 JK |
152 | static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i, |
153 | struct nat_entry *ne) | |
154 | { | |
155 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); | |
156 | struct nat_entry_set *head; | |
157 | ||
158 | if (get_nat_flag(ne, IS_DIRTY)) | |
159 | return; | |
9be32d72 | 160 | |
309cc2b6 JK |
161 | head = radix_tree_lookup(&nm_i->nat_set_root, set); |
162 | if (!head) { | |
80c54505 | 163 | head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS); |
309cc2b6 JK |
164 | |
165 | INIT_LIST_HEAD(&head->entry_list); | |
166 | INIT_LIST_HEAD(&head->set_list); | |
167 | head->set = set; | |
168 | head->entry_cnt = 0; | |
9be32d72 | 169 | f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head); |
309cc2b6 JK |
170 | } |
171 | list_move_tail(&ne->list, &head->entry_list); | |
172 | nm_i->dirty_nat_cnt++; | |
173 | head->entry_cnt++; | |
174 | set_nat_flag(ne, IS_DIRTY, true); | |
175 | } | |
176 | ||
177 | static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i, | |
178 | struct nat_entry *ne) | |
179 | { | |
20d047c8 | 180 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); |
309cc2b6 JK |
181 | struct nat_entry_set *head; |
182 | ||
183 | head = radix_tree_lookup(&nm_i->nat_set_root, set); | |
184 | if (head) { | |
185 | list_move_tail(&ne->list, &nm_i->nat_entries); | |
186 | set_nat_flag(ne, IS_DIRTY, false); | |
187 | head->entry_cnt--; | |
188 | nm_i->dirty_nat_cnt--; | |
189 | } | |
190 | } | |
191 | ||
192 | static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i, | |
193 | nid_t start, unsigned int nr, struct nat_entry_set **ep) | |
194 | { | |
195 | return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep, | |
196 | start, nr); | |
197 | } | |
198 | ||
2dcf51ab | 199 | int need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 JK |
200 | { |
201 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
202 | struct nat_entry *e; | |
2dcf51ab | 203 | bool need = false; |
e05df3b1 | 204 | |
8b26ef98 | 205 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 | 206 | e = __lookup_nat_cache(nm_i, nid); |
2dcf51ab JK |
207 | if (e) { |
208 | if (!get_nat_flag(e, IS_CHECKPOINTED) && | |
209 | !get_nat_flag(e, HAS_FSYNCED_INODE)) | |
210 | need = true; | |
211 | } | |
8b26ef98 | 212 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 213 | return need; |
e05df3b1 JK |
214 | } |
215 | ||
2dcf51ab | 216 | bool is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid) |
479f40c4 JK |
217 | { |
218 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
219 | struct nat_entry *e; | |
2dcf51ab | 220 | bool is_cp = true; |
479f40c4 | 221 | |
8b26ef98 | 222 | down_read(&nm_i->nat_tree_lock); |
2dcf51ab JK |
223 | e = __lookup_nat_cache(nm_i, nid); |
224 | if (e && !get_nat_flag(e, IS_CHECKPOINTED)) | |
225 | is_cp = false; | |
8b26ef98 | 226 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 227 | return is_cp; |
479f40c4 JK |
228 | } |
229 | ||
88bd02c9 | 230 | bool need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino) |
b6fe5873 JK |
231 | { |
232 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
233 | struct nat_entry *e; | |
88bd02c9 | 234 | bool need_update = true; |
b6fe5873 | 235 | |
8b26ef98 | 236 | down_read(&nm_i->nat_tree_lock); |
88bd02c9 JK |
237 | e = __lookup_nat_cache(nm_i, ino); |
238 | if (e && get_nat_flag(e, HAS_LAST_FSYNC) && | |
239 | (get_nat_flag(e, IS_CHECKPOINTED) || | |
240 | get_nat_flag(e, HAS_FSYNCED_INODE))) | |
241 | need_update = false; | |
8b26ef98 | 242 | up_read(&nm_i->nat_tree_lock); |
88bd02c9 | 243 | return need_update; |
b6fe5873 JK |
244 | } |
245 | ||
e05df3b1 JK |
246 | static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid) |
247 | { | |
248 | struct nat_entry *new; | |
249 | ||
80c54505 | 250 | new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_NOFS); |
9be32d72 | 251 | f2fs_radix_tree_insert(&nm_i->nat_root, nid, new); |
e05df3b1 JK |
252 | memset(new, 0, sizeof(struct nat_entry)); |
253 | nat_set_nid(new, nid); | |
88bd02c9 | 254 | nat_reset_flag(new); |
e05df3b1 JK |
255 | list_add_tail(&new->list, &nm_i->nat_entries); |
256 | nm_i->nat_cnt++; | |
257 | return new; | |
258 | } | |
259 | ||
260 | static void cache_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid, | |
261 | struct f2fs_nat_entry *ne) | |
262 | { | |
263 | struct nat_entry *e; | |
9be32d72 | 264 | |
8b26ef98 | 265 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
266 | e = __lookup_nat_cache(nm_i, nid); |
267 | if (!e) { | |
268 | e = grab_nat_entry(nm_i, nid); | |
94dac22e | 269 | node_info_from_raw_nat(&e->ni, ne); |
e05df3b1 | 270 | } |
8b26ef98 | 271 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
272 | } |
273 | ||
274 | static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, | |
479f40c4 | 275 | block_t new_blkaddr, bool fsync_done) |
e05df3b1 JK |
276 | { |
277 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
278 | struct nat_entry *e; | |
9be32d72 | 279 | |
8b26ef98 | 280 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
281 | e = __lookup_nat_cache(nm_i, ni->nid); |
282 | if (!e) { | |
283 | e = grab_nat_entry(nm_i, ni->nid); | |
5c27f4ee | 284 | copy_node_info(&e->ni, ni); |
9850cf4a | 285 | f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR); |
e05df3b1 JK |
286 | } else if (new_blkaddr == NEW_ADDR) { |
287 | /* | |
288 | * when nid is reallocated, | |
289 | * previous nat entry can be remained in nat cache. | |
290 | * So, reinitialize it with new information. | |
291 | */ | |
5c27f4ee | 292 | copy_node_info(&e->ni, ni); |
9850cf4a | 293 | f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR); |
e05df3b1 JK |
294 | } |
295 | ||
e05df3b1 | 296 | /* sanity check */ |
9850cf4a JK |
297 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr); |
298 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR && | |
e05df3b1 | 299 | new_blkaddr == NULL_ADDR); |
9850cf4a | 300 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR && |
e05df3b1 | 301 | new_blkaddr == NEW_ADDR); |
9850cf4a | 302 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != NEW_ADDR && |
e05df3b1 JK |
303 | nat_get_blkaddr(e) != NULL_ADDR && |
304 | new_blkaddr == NEW_ADDR); | |
305 | ||
e1c42045 | 306 | /* increment version no as node is removed */ |
e05df3b1 JK |
307 | if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) { |
308 | unsigned char version = nat_get_version(e); | |
309 | nat_set_version(e, inc_node_version(version)); | |
26834466 JK |
310 | |
311 | /* in order to reuse the nid */ | |
312 | if (nm_i->next_scan_nid > ni->nid) | |
313 | nm_i->next_scan_nid = ni->nid; | |
e05df3b1 JK |
314 | } |
315 | ||
316 | /* change address */ | |
317 | nat_set_blkaddr(e, new_blkaddr); | |
88bd02c9 JK |
318 | if (new_blkaddr == NEW_ADDR || new_blkaddr == NULL_ADDR) |
319 | set_nat_flag(e, IS_CHECKPOINTED, false); | |
e05df3b1 | 320 | __set_nat_cache_dirty(nm_i, e); |
479f40c4 JK |
321 | |
322 | /* update fsync_mark if its inode nat entry is still alive */ | |
d5b692b7 CY |
323 | if (ni->nid != ni->ino) |
324 | e = __lookup_nat_cache(nm_i, ni->ino); | |
88bd02c9 JK |
325 | if (e) { |
326 | if (fsync_done && ni->nid == ni->ino) | |
327 | set_nat_flag(e, HAS_FSYNCED_INODE, true); | |
328 | set_nat_flag(e, HAS_LAST_FSYNC, fsync_done); | |
329 | } | |
8b26ef98 | 330 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
331 | } |
332 | ||
4660f9c0 | 333 | int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) |
e05df3b1 JK |
334 | { |
335 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1b38dc8e | 336 | int nr = nr_shrink; |
e05df3b1 | 337 | |
1b38dc8e | 338 | if (!down_write_trylock(&nm_i->nat_tree_lock)) |
e05df3b1 JK |
339 | return 0; |
340 | ||
e05df3b1 JK |
341 | while (nr_shrink && !list_empty(&nm_i->nat_entries)) { |
342 | struct nat_entry *ne; | |
343 | ne = list_first_entry(&nm_i->nat_entries, | |
344 | struct nat_entry, list); | |
345 | __del_from_nat_cache(nm_i, ne); | |
346 | nr_shrink--; | |
347 | } | |
8b26ef98 | 348 | up_write(&nm_i->nat_tree_lock); |
1b38dc8e | 349 | return nr - nr_shrink; |
e05df3b1 JK |
350 | } |
351 | ||
0a8165d7 | 352 | /* |
e1c42045 | 353 | * This function always returns success |
e05df3b1 JK |
354 | */ |
355 | void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni) | |
356 | { | |
357 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
358 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
359 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
360 | nid_t start_nid = START_NID(nid); | |
361 | struct f2fs_nat_block *nat_blk; | |
362 | struct page *page = NULL; | |
363 | struct f2fs_nat_entry ne; | |
364 | struct nat_entry *e; | |
365 | int i; | |
366 | ||
367 | ni->nid = nid; | |
368 | ||
369 | /* Check nat cache */ | |
8b26ef98 | 370 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
371 | e = __lookup_nat_cache(nm_i, nid); |
372 | if (e) { | |
373 | ni->ino = nat_get_ino(e); | |
374 | ni->blk_addr = nat_get_blkaddr(e); | |
375 | ni->version = nat_get_version(e); | |
376 | } | |
8b26ef98 | 377 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
378 | if (e) |
379 | return; | |
380 | ||
3547ea96 JK |
381 | memset(&ne, 0, sizeof(struct f2fs_nat_entry)); |
382 | ||
e05df3b1 JK |
383 | /* Check current segment summary */ |
384 | mutex_lock(&curseg->curseg_mutex); | |
385 | i = lookup_journal_in_cursum(sum, NAT_JOURNAL, nid, 0); | |
386 | if (i >= 0) { | |
387 | ne = nat_in_journal(sum, i); | |
388 | node_info_from_raw_nat(ni, &ne); | |
389 | } | |
390 | mutex_unlock(&curseg->curseg_mutex); | |
391 | if (i >= 0) | |
392 | goto cache; | |
393 | ||
394 | /* Fill node_info from nat page */ | |
395 | page = get_current_nat_page(sbi, start_nid); | |
396 | nat_blk = (struct f2fs_nat_block *)page_address(page); | |
397 | ne = nat_blk->entries[nid - start_nid]; | |
398 | node_info_from_raw_nat(ni, &ne); | |
399 | f2fs_put_page(page, 1); | |
400 | cache: | |
401 | /* cache nat entry */ | |
402 | cache_nat_entry(NM_I(sbi), nid, &ne); | |
403 | } | |
404 | ||
0a8165d7 | 405 | /* |
e05df3b1 JK |
406 | * The maximum depth is four. |
407 | * Offset[0] will have raw inode offset. | |
408 | */ | |
de93653f JK |
409 | static int get_node_path(struct f2fs_inode_info *fi, long block, |
410 | int offset[4], unsigned int noffset[4]) | |
e05df3b1 | 411 | { |
de93653f | 412 | const long direct_index = ADDRS_PER_INODE(fi); |
e05df3b1 JK |
413 | const long direct_blks = ADDRS_PER_BLOCK; |
414 | const long dptrs_per_blk = NIDS_PER_BLOCK; | |
415 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
416 | const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK; | |
417 | int n = 0; | |
418 | int level = 0; | |
419 | ||
420 | noffset[0] = 0; | |
421 | ||
422 | if (block < direct_index) { | |
25c0a6e5 | 423 | offset[n] = block; |
e05df3b1 JK |
424 | goto got; |
425 | } | |
426 | block -= direct_index; | |
427 | if (block < direct_blks) { | |
428 | offset[n++] = NODE_DIR1_BLOCK; | |
429 | noffset[n] = 1; | |
25c0a6e5 | 430 | offset[n] = block; |
e05df3b1 JK |
431 | level = 1; |
432 | goto got; | |
433 | } | |
434 | block -= direct_blks; | |
435 | if (block < direct_blks) { | |
436 | offset[n++] = NODE_DIR2_BLOCK; | |
437 | noffset[n] = 2; | |
25c0a6e5 | 438 | offset[n] = block; |
e05df3b1 JK |
439 | level = 1; |
440 | goto got; | |
441 | } | |
442 | block -= direct_blks; | |
443 | if (block < indirect_blks) { | |
444 | offset[n++] = NODE_IND1_BLOCK; | |
445 | noffset[n] = 3; | |
446 | offset[n++] = block / direct_blks; | |
447 | noffset[n] = 4 + offset[n - 1]; | |
25c0a6e5 | 448 | offset[n] = block % direct_blks; |
e05df3b1 JK |
449 | level = 2; |
450 | goto got; | |
451 | } | |
452 | block -= indirect_blks; | |
453 | if (block < indirect_blks) { | |
454 | offset[n++] = NODE_IND2_BLOCK; | |
455 | noffset[n] = 4 + dptrs_per_blk; | |
456 | offset[n++] = block / direct_blks; | |
457 | noffset[n] = 5 + dptrs_per_blk + offset[n - 1]; | |
25c0a6e5 | 458 | offset[n] = block % direct_blks; |
e05df3b1 JK |
459 | level = 2; |
460 | goto got; | |
461 | } | |
462 | block -= indirect_blks; | |
463 | if (block < dindirect_blks) { | |
464 | offset[n++] = NODE_DIND_BLOCK; | |
465 | noffset[n] = 5 + (dptrs_per_blk * 2); | |
466 | offset[n++] = block / indirect_blks; | |
467 | noffset[n] = 6 + (dptrs_per_blk * 2) + | |
468 | offset[n - 1] * (dptrs_per_blk + 1); | |
469 | offset[n++] = (block / direct_blks) % dptrs_per_blk; | |
470 | noffset[n] = 7 + (dptrs_per_blk * 2) + | |
471 | offset[n - 2] * (dptrs_per_blk + 1) + | |
472 | offset[n - 1]; | |
25c0a6e5 | 473 | offset[n] = block % direct_blks; |
e05df3b1 JK |
474 | level = 3; |
475 | goto got; | |
476 | } else { | |
477 | BUG(); | |
478 | } | |
479 | got: | |
480 | return level; | |
481 | } | |
482 | ||
483 | /* | |
484 | * Caller should call f2fs_put_dnode(dn). | |
4f4124d0 CY |
485 | * Also, it should grab and release a rwsem by calling f2fs_lock_op() and |
486 | * f2fs_unlock_op() only if ro is not set RDONLY_NODE. | |
39936837 | 487 | * In the case of RDONLY_NODE, we don't need to care about mutex. |
e05df3b1 | 488 | */ |
266e97a8 | 489 | int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode) |
e05df3b1 | 490 | { |
4081363f | 491 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 | 492 | struct page *npage[4]; |
f1a3b98e | 493 | struct page *parent = NULL; |
e05df3b1 JK |
494 | int offset[4]; |
495 | unsigned int noffset[4]; | |
496 | nid_t nids[4]; | |
497 | int level, i; | |
498 | int err = 0; | |
499 | ||
de93653f | 500 | level = get_node_path(F2FS_I(dn->inode), index, offset, noffset); |
e05df3b1 JK |
501 | |
502 | nids[0] = dn->inode->i_ino; | |
1646cfac | 503 | npage[0] = dn->inode_page; |
e05df3b1 | 504 | |
1646cfac JK |
505 | if (!npage[0]) { |
506 | npage[0] = get_node_page(sbi, nids[0]); | |
507 | if (IS_ERR(npage[0])) | |
508 | return PTR_ERR(npage[0]); | |
509 | } | |
f1a3b98e JK |
510 | |
511 | /* if inline_data is set, should not report any block indices */ | |
512 | if (f2fs_has_inline_data(dn->inode) && index) { | |
76629165 | 513 | err = -ENOENT; |
f1a3b98e JK |
514 | f2fs_put_page(npage[0], 1); |
515 | goto release_out; | |
516 | } | |
517 | ||
e05df3b1 | 518 | parent = npage[0]; |
52c2db3f CL |
519 | if (level != 0) |
520 | nids[1] = get_nid(parent, offset[0], true); | |
e05df3b1 JK |
521 | dn->inode_page = npage[0]; |
522 | dn->inode_page_locked = true; | |
523 | ||
524 | /* get indirect or direct nodes */ | |
525 | for (i = 1; i <= level; i++) { | |
526 | bool done = false; | |
527 | ||
266e97a8 | 528 | if (!nids[i] && mode == ALLOC_NODE) { |
e05df3b1 JK |
529 | /* alloc new node */ |
530 | if (!alloc_nid(sbi, &(nids[i]))) { | |
e05df3b1 JK |
531 | err = -ENOSPC; |
532 | goto release_pages; | |
533 | } | |
534 | ||
535 | dn->nid = nids[i]; | |
8ae8f162 | 536 | npage[i] = new_node_page(dn, noffset[i], NULL); |
e05df3b1 JK |
537 | if (IS_ERR(npage[i])) { |
538 | alloc_nid_failed(sbi, nids[i]); | |
e05df3b1 JK |
539 | err = PTR_ERR(npage[i]); |
540 | goto release_pages; | |
541 | } | |
542 | ||
543 | set_nid(parent, offset[i - 1], nids[i], i == 1); | |
544 | alloc_nid_done(sbi, nids[i]); | |
93bae099 | 545 | dn->node_changed = true; |
e05df3b1 | 546 | done = true; |
266e97a8 | 547 | } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) { |
e05df3b1 JK |
548 | npage[i] = get_node_page_ra(parent, offset[i - 1]); |
549 | if (IS_ERR(npage[i])) { | |
550 | err = PTR_ERR(npage[i]); | |
551 | goto release_pages; | |
552 | } | |
553 | done = true; | |
554 | } | |
555 | if (i == 1) { | |
556 | dn->inode_page_locked = false; | |
557 | unlock_page(parent); | |
558 | } else { | |
559 | f2fs_put_page(parent, 1); | |
560 | } | |
561 | ||
562 | if (!done) { | |
563 | npage[i] = get_node_page(sbi, nids[i]); | |
564 | if (IS_ERR(npage[i])) { | |
565 | err = PTR_ERR(npage[i]); | |
566 | f2fs_put_page(npage[0], 0); | |
567 | goto release_out; | |
568 | } | |
569 | } | |
570 | if (i < level) { | |
571 | parent = npage[i]; | |
572 | nids[i + 1] = get_nid(parent, offset[i], false); | |
573 | } | |
574 | } | |
575 | dn->nid = nids[level]; | |
576 | dn->ofs_in_node = offset[level]; | |
577 | dn->node_page = npage[level]; | |
578 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
579 | return 0; | |
580 | ||
581 | release_pages: | |
582 | f2fs_put_page(parent, 1); | |
583 | if (i > 1) | |
584 | f2fs_put_page(npage[0], 0); | |
585 | release_out: | |
586 | dn->inode_page = NULL; | |
587 | dn->node_page = NULL; | |
588 | return err; | |
589 | } | |
590 | ||
591 | static void truncate_node(struct dnode_of_data *dn) | |
592 | { | |
4081363f | 593 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
594 | struct node_info ni; |
595 | ||
596 | get_node_info(sbi, dn->nid, &ni); | |
71e9fec5 | 597 | if (dn->inode->i_blocks == 0) { |
9850cf4a | 598 | f2fs_bug_on(sbi, ni.blk_addr != NULL_ADDR); |
71e9fec5 JK |
599 | goto invalidate; |
600 | } | |
9850cf4a | 601 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
e05df3b1 | 602 | |
e05df3b1 | 603 | /* Deallocate node address */ |
71e9fec5 | 604 | invalidate_blocks(sbi, ni.blk_addr); |
ef86d709 | 605 | dec_valid_node_count(sbi, dn->inode); |
479f40c4 | 606 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
e05df3b1 JK |
607 | |
608 | if (dn->nid == dn->inode->i_ino) { | |
609 | remove_orphan_inode(sbi, dn->nid); | |
610 | dec_valid_inode_count(sbi); | |
611 | } else { | |
612 | sync_inode_page(dn); | |
613 | } | |
71e9fec5 | 614 | invalidate: |
e05df3b1 | 615 | clear_node_page_dirty(dn->node_page); |
caf0047e | 616 | set_sbi_flag(sbi, SBI_IS_DIRTY); |
e05df3b1 JK |
617 | |
618 | f2fs_put_page(dn->node_page, 1); | |
bf39c00a JK |
619 | |
620 | invalidate_mapping_pages(NODE_MAPPING(sbi), | |
621 | dn->node_page->index, dn->node_page->index); | |
622 | ||
e05df3b1 | 623 | dn->node_page = NULL; |
51dd6249 | 624 | trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr); |
e05df3b1 JK |
625 | } |
626 | ||
627 | static int truncate_dnode(struct dnode_of_data *dn) | |
628 | { | |
e05df3b1 JK |
629 | struct page *page; |
630 | ||
631 | if (dn->nid == 0) | |
632 | return 1; | |
633 | ||
634 | /* get direct node */ | |
4081363f | 635 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
e05df3b1 JK |
636 | if (IS_ERR(page) && PTR_ERR(page) == -ENOENT) |
637 | return 1; | |
638 | else if (IS_ERR(page)) | |
639 | return PTR_ERR(page); | |
640 | ||
641 | /* Make dnode_of_data for parameter */ | |
642 | dn->node_page = page; | |
643 | dn->ofs_in_node = 0; | |
644 | truncate_data_blocks(dn); | |
645 | truncate_node(dn); | |
646 | return 1; | |
647 | } | |
648 | ||
649 | static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs, | |
650 | int ofs, int depth) | |
651 | { | |
e05df3b1 JK |
652 | struct dnode_of_data rdn = *dn; |
653 | struct page *page; | |
654 | struct f2fs_node *rn; | |
655 | nid_t child_nid; | |
656 | unsigned int child_nofs; | |
657 | int freed = 0; | |
658 | int i, ret; | |
659 | ||
660 | if (dn->nid == 0) | |
661 | return NIDS_PER_BLOCK + 1; | |
662 | ||
51dd6249 NJ |
663 | trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr); |
664 | ||
4081363f | 665 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
51dd6249 NJ |
666 | if (IS_ERR(page)) { |
667 | trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page)); | |
e05df3b1 | 668 | return PTR_ERR(page); |
51dd6249 | 669 | } |
e05df3b1 | 670 | |
45590710 | 671 | rn = F2FS_NODE(page); |
e05df3b1 JK |
672 | if (depth < 3) { |
673 | for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) { | |
674 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
675 | if (child_nid == 0) | |
676 | continue; | |
677 | rdn.nid = child_nid; | |
678 | ret = truncate_dnode(&rdn); | |
679 | if (ret < 0) | |
680 | goto out_err; | |
681 | set_nid(page, i, 0, false); | |
93bae099 | 682 | dn->node_changed = true; |
e05df3b1 JK |
683 | } |
684 | } else { | |
685 | child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1; | |
686 | for (i = ofs; i < NIDS_PER_BLOCK; i++) { | |
687 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
688 | if (child_nid == 0) { | |
689 | child_nofs += NIDS_PER_BLOCK + 1; | |
690 | continue; | |
691 | } | |
692 | rdn.nid = child_nid; | |
693 | ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1); | |
694 | if (ret == (NIDS_PER_BLOCK + 1)) { | |
695 | set_nid(page, i, 0, false); | |
93bae099 | 696 | dn->node_changed = true; |
e05df3b1 JK |
697 | child_nofs += ret; |
698 | } else if (ret < 0 && ret != -ENOENT) { | |
699 | goto out_err; | |
700 | } | |
701 | } | |
702 | freed = child_nofs; | |
703 | } | |
704 | ||
705 | if (!ofs) { | |
706 | /* remove current indirect node */ | |
707 | dn->node_page = page; | |
708 | truncate_node(dn); | |
709 | freed++; | |
710 | } else { | |
711 | f2fs_put_page(page, 1); | |
712 | } | |
51dd6249 | 713 | trace_f2fs_truncate_nodes_exit(dn->inode, freed); |
e05df3b1 JK |
714 | return freed; |
715 | ||
716 | out_err: | |
717 | f2fs_put_page(page, 1); | |
51dd6249 | 718 | trace_f2fs_truncate_nodes_exit(dn->inode, ret); |
e05df3b1 JK |
719 | return ret; |
720 | } | |
721 | ||
722 | static int truncate_partial_nodes(struct dnode_of_data *dn, | |
723 | struct f2fs_inode *ri, int *offset, int depth) | |
724 | { | |
e05df3b1 JK |
725 | struct page *pages[2]; |
726 | nid_t nid[3]; | |
727 | nid_t child_nid; | |
728 | int err = 0; | |
729 | int i; | |
730 | int idx = depth - 2; | |
731 | ||
732 | nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); | |
733 | if (!nid[0]) | |
734 | return 0; | |
735 | ||
736 | /* get indirect nodes in the path */ | |
a225dca3 | 737 | for (i = 0; i < idx + 1; i++) { |
e1c42045 | 738 | /* reference count'll be increased */ |
4081363f | 739 | pages[i] = get_node_page(F2FS_I_SB(dn->inode), nid[i]); |
e05df3b1 | 740 | if (IS_ERR(pages[i])) { |
e05df3b1 | 741 | err = PTR_ERR(pages[i]); |
a225dca3 | 742 | idx = i - 1; |
e05df3b1 JK |
743 | goto fail; |
744 | } | |
745 | nid[i + 1] = get_nid(pages[i], offset[i + 1], false); | |
746 | } | |
747 | ||
748 | /* free direct nodes linked to a partial indirect node */ | |
a225dca3 | 749 | for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) { |
e05df3b1 JK |
750 | child_nid = get_nid(pages[idx], i, false); |
751 | if (!child_nid) | |
752 | continue; | |
753 | dn->nid = child_nid; | |
754 | err = truncate_dnode(dn); | |
755 | if (err < 0) | |
756 | goto fail; | |
757 | set_nid(pages[idx], i, 0, false); | |
93bae099 | 758 | dn->node_changed = true; |
e05df3b1 JK |
759 | } |
760 | ||
a225dca3 | 761 | if (offset[idx + 1] == 0) { |
e05df3b1 JK |
762 | dn->node_page = pages[idx]; |
763 | dn->nid = nid[idx]; | |
764 | truncate_node(dn); | |
765 | } else { | |
766 | f2fs_put_page(pages[idx], 1); | |
767 | } | |
768 | offset[idx]++; | |
a225dca3 | 769 | offset[idx + 1] = 0; |
770 | idx--; | |
e05df3b1 | 771 | fail: |
a225dca3 | 772 | for (i = idx; i >= 0; i--) |
e05df3b1 | 773 | f2fs_put_page(pages[i], 1); |
51dd6249 NJ |
774 | |
775 | trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err); | |
776 | ||
e05df3b1 JK |
777 | return err; |
778 | } | |
779 | ||
0a8165d7 | 780 | /* |
e05df3b1 JK |
781 | * All the block addresses of data and nodes should be nullified. |
782 | */ | |
783 | int truncate_inode_blocks(struct inode *inode, pgoff_t from) | |
784 | { | |
4081363f | 785 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
e05df3b1 JK |
786 | int err = 0, cont = 1; |
787 | int level, offset[4], noffset[4]; | |
7dd690c8 | 788 | unsigned int nofs = 0; |
58bfaf44 | 789 | struct f2fs_inode *ri; |
e05df3b1 JK |
790 | struct dnode_of_data dn; |
791 | struct page *page; | |
792 | ||
51dd6249 NJ |
793 | trace_f2fs_truncate_inode_blocks_enter(inode, from); |
794 | ||
de93653f | 795 | level = get_node_path(F2FS_I(inode), from, offset, noffset); |
afcb7ca0 | 796 | restart: |
e05df3b1 | 797 | page = get_node_page(sbi, inode->i_ino); |
51dd6249 NJ |
798 | if (IS_ERR(page)) { |
799 | trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page)); | |
e05df3b1 | 800 | return PTR_ERR(page); |
51dd6249 | 801 | } |
e05df3b1 JK |
802 | |
803 | set_new_dnode(&dn, inode, page, NULL, 0); | |
804 | unlock_page(page); | |
805 | ||
58bfaf44 | 806 | ri = F2FS_INODE(page); |
e05df3b1 JK |
807 | switch (level) { |
808 | case 0: | |
809 | case 1: | |
810 | nofs = noffset[1]; | |
811 | break; | |
812 | case 2: | |
813 | nofs = noffset[1]; | |
814 | if (!offset[level - 1]) | |
815 | goto skip_partial; | |
58bfaf44 | 816 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
817 | if (err < 0 && err != -ENOENT) |
818 | goto fail; | |
819 | nofs += 1 + NIDS_PER_BLOCK; | |
820 | break; | |
821 | case 3: | |
822 | nofs = 5 + 2 * NIDS_PER_BLOCK; | |
823 | if (!offset[level - 1]) | |
824 | goto skip_partial; | |
58bfaf44 | 825 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
826 | if (err < 0 && err != -ENOENT) |
827 | goto fail; | |
828 | break; | |
829 | default: | |
830 | BUG(); | |
831 | } | |
832 | ||
833 | skip_partial: | |
834 | while (cont) { | |
58bfaf44 | 835 | dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); |
e05df3b1 JK |
836 | switch (offset[0]) { |
837 | case NODE_DIR1_BLOCK: | |
838 | case NODE_DIR2_BLOCK: | |
839 | err = truncate_dnode(&dn); | |
840 | break; | |
841 | ||
842 | case NODE_IND1_BLOCK: | |
843 | case NODE_IND2_BLOCK: | |
844 | err = truncate_nodes(&dn, nofs, offset[1], 2); | |
845 | break; | |
846 | ||
847 | case NODE_DIND_BLOCK: | |
848 | err = truncate_nodes(&dn, nofs, offset[1], 3); | |
849 | cont = 0; | |
850 | break; | |
851 | ||
852 | default: | |
853 | BUG(); | |
854 | } | |
855 | if (err < 0 && err != -ENOENT) | |
856 | goto fail; | |
857 | if (offset[1] == 0 && | |
58bfaf44 | 858 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) { |
e05df3b1 | 859 | lock_page(page); |
4ef51a8f | 860 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
861 | f2fs_put_page(page, 1); |
862 | goto restart; | |
863 | } | |
3cb5ad15 | 864 | f2fs_wait_on_page_writeback(page, NODE); |
58bfaf44 | 865 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0; |
e05df3b1 JK |
866 | set_page_dirty(page); |
867 | unlock_page(page); | |
868 | } | |
869 | offset[1] = 0; | |
870 | offset[0]++; | |
871 | nofs += err; | |
872 | } | |
873 | fail: | |
874 | f2fs_put_page(page, 0); | |
51dd6249 | 875 | trace_f2fs_truncate_inode_blocks_exit(inode, err); |
e05df3b1 JK |
876 | return err > 0 ? 0 : err; |
877 | } | |
878 | ||
4f16fb0f JK |
879 | int truncate_xattr_node(struct inode *inode, struct page *page) |
880 | { | |
4081363f | 881 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
4f16fb0f JK |
882 | nid_t nid = F2FS_I(inode)->i_xattr_nid; |
883 | struct dnode_of_data dn; | |
884 | struct page *npage; | |
885 | ||
886 | if (!nid) | |
887 | return 0; | |
888 | ||
889 | npage = get_node_page(sbi, nid); | |
890 | if (IS_ERR(npage)) | |
891 | return PTR_ERR(npage); | |
892 | ||
893 | F2FS_I(inode)->i_xattr_nid = 0; | |
65985d93 JK |
894 | |
895 | /* need to do checkpoint during fsync */ | |
896 | F2FS_I(inode)->xattr_ver = cur_cp_version(F2FS_CKPT(sbi)); | |
897 | ||
4f16fb0f JK |
898 | set_new_dnode(&dn, inode, page, npage, nid); |
899 | ||
900 | if (page) | |
01d2d1aa | 901 | dn.inode_page_locked = true; |
4f16fb0f JK |
902 | truncate_node(&dn); |
903 | return 0; | |
904 | } | |
905 | ||
39936837 | 906 | /* |
4f4124d0 CY |
907 | * Caller should grab and release a rwsem by calling f2fs_lock_op() and |
908 | * f2fs_unlock_op(). | |
39936837 | 909 | */ |
13ec7297 | 910 | int remove_inode_page(struct inode *inode) |
e05df3b1 | 911 | { |
e05df3b1 | 912 | struct dnode_of_data dn; |
13ec7297 | 913 | int err; |
e05df3b1 | 914 | |
c2e69583 | 915 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); |
13ec7297 CY |
916 | err = get_dnode_of_data(&dn, 0, LOOKUP_NODE); |
917 | if (err) | |
918 | return err; | |
e05df3b1 | 919 | |
13ec7297 CY |
920 | err = truncate_xattr_node(inode, dn.inode_page); |
921 | if (err) { | |
c2e69583 | 922 | f2fs_put_dnode(&dn); |
13ec7297 | 923 | return err; |
e05df3b1 | 924 | } |
c2e69583 JK |
925 | |
926 | /* remove potential inline_data blocks */ | |
927 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
928 | S_ISLNK(inode->i_mode)) | |
929 | truncate_data_blocks_range(&dn, 1); | |
930 | ||
e1c42045 | 931 | /* 0 is possible, after f2fs_new_inode() has failed */ |
9850cf4a JK |
932 | f2fs_bug_on(F2FS_I_SB(inode), |
933 | inode->i_blocks != 0 && inode->i_blocks != 1); | |
c2e69583 JK |
934 | |
935 | /* will put inode & node pages */ | |
71e9fec5 | 936 | truncate_node(&dn); |
13ec7297 | 937 | return 0; |
e05df3b1 JK |
938 | } |
939 | ||
a014e037 | 940 | struct page *new_inode_page(struct inode *inode) |
e05df3b1 | 941 | { |
e05df3b1 JK |
942 | struct dnode_of_data dn; |
943 | ||
944 | /* allocate inode page for new inode */ | |
945 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); | |
44a83ff6 JK |
946 | |
947 | /* caller should f2fs_put_page(page, 1); */ | |
8ae8f162 | 948 | return new_node_page(&dn, 0, NULL); |
e05df3b1 JK |
949 | } |
950 | ||
8ae8f162 JK |
951 | struct page *new_node_page(struct dnode_of_data *dn, |
952 | unsigned int ofs, struct page *ipage) | |
e05df3b1 | 953 | { |
4081363f | 954 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
955 | struct node_info old_ni, new_ni; |
956 | struct page *page; | |
957 | int err; | |
958 | ||
6bacf52f | 959 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
e05df3b1 JK |
960 | return ERR_PTR(-EPERM); |
961 | ||
54b591df | 962 | page = grab_cache_page(NODE_MAPPING(sbi), dn->nid); |
e05df3b1 JK |
963 | if (!page) |
964 | return ERR_PTR(-ENOMEM); | |
965 | ||
6bacf52f | 966 | if (unlikely(!inc_valid_node_count(sbi, dn->inode))) { |
9c02740c JK |
967 | err = -ENOSPC; |
968 | goto fail; | |
969 | } | |
e05df3b1 | 970 | |
9c02740c | 971 | get_node_info(sbi, dn->nid, &old_ni); |
e05df3b1 JK |
972 | |
973 | /* Reinitialize old_ni with new node page */ | |
9850cf4a | 974 | f2fs_bug_on(sbi, old_ni.blk_addr != NULL_ADDR); |
e05df3b1 JK |
975 | new_ni = old_ni; |
976 | new_ni.ino = dn->inode->i_ino; | |
479f40c4 | 977 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
9c02740c | 978 | |
54b591df | 979 | f2fs_wait_on_page_writeback(page, NODE); |
9c02740c | 980 | fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true); |
398b1ac5 | 981 | set_cold_node(dn->inode, page); |
9c02740c JK |
982 | SetPageUptodate(page); |
983 | set_page_dirty(page); | |
e05df3b1 | 984 | |
4bc8e9bc | 985 | if (f2fs_has_xattr_block(ofs)) |
479bd73a JK |
986 | F2FS_I(dn->inode)->i_xattr_nid = dn->nid; |
987 | ||
e05df3b1 | 988 | dn->node_page = page; |
8ae8f162 JK |
989 | if (ipage) |
990 | update_inode(dn->inode, ipage); | |
991 | else | |
992 | sync_inode_page(dn); | |
e05df3b1 JK |
993 | if (ofs == 0) |
994 | inc_valid_inode_count(sbi); | |
995 | ||
996 | return page; | |
997 | ||
998 | fail: | |
71e9fec5 | 999 | clear_node_page_dirty(page); |
e05df3b1 JK |
1000 | f2fs_put_page(page, 1); |
1001 | return ERR_PTR(err); | |
1002 | } | |
1003 | ||
56ae674c JK |
1004 | /* |
1005 | * Caller should do after getting the following values. | |
1006 | * 0: f2fs_put_page(page, 0) | |
86531d6b | 1007 | * LOCKED_PAGE or error: f2fs_put_page(page, 1) |
56ae674c | 1008 | */ |
93dfe2ac | 1009 | static int read_node_page(struct page *page, int rw) |
e05df3b1 | 1010 | { |
4081363f | 1011 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1012 | struct node_info ni; |
cf04e8eb | 1013 | struct f2fs_io_info fio = { |
05ca3632 | 1014 | .sbi = sbi, |
cf04e8eb JK |
1015 | .type = NODE, |
1016 | .rw = rw, | |
05ca3632 | 1017 | .page = page, |
4375a336 | 1018 | .encrypted_page = NULL, |
cf04e8eb | 1019 | }; |
e05df3b1 JK |
1020 | |
1021 | get_node_info(sbi, page->index, &ni); | |
1022 | ||
6bacf52f | 1023 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
2bca1e23 | 1024 | ClearPageUptodate(page); |
e05df3b1 | 1025 | return -ENOENT; |
393ff91f JK |
1026 | } |
1027 | ||
56ae674c JK |
1028 | if (PageUptodate(page)) |
1029 | return LOCKED_PAGE; | |
393ff91f | 1030 | |
cf04e8eb | 1031 | fio.blk_addr = ni.blk_addr; |
05ca3632 | 1032 | return f2fs_submit_page_bio(&fio); |
e05df3b1 JK |
1033 | } |
1034 | ||
0a8165d7 | 1035 | /* |
e05df3b1 JK |
1036 | * Readahead a node page |
1037 | */ | |
1038 | void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid) | |
1039 | { | |
e05df3b1 | 1040 | struct page *apage; |
56ae674c | 1041 | int err; |
e05df3b1 | 1042 | |
4ef51a8f | 1043 | apage = find_get_page(NODE_MAPPING(sbi), nid); |
393ff91f JK |
1044 | if (apage && PageUptodate(apage)) { |
1045 | f2fs_put_page(apage, 0); | |
1046 | return; | |
1047 | } | |
e05df3b1 JK |
1048 | f2fs_put_page(apage, 0); |
1049 | ||
4ef51a8f | 1050 | apage = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1051 | if (!apage) |
1052 | return; | |
1053 | ||
56ae674c | 1054 | err = read_node_page(apage, READA); |
86531d6b | 1055 | f2fs_put_page(apage, err ? 1 : 0); |
e05df3b1 JK |
1056 | } |
1057 | ||
1058 | struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) | |
1059 | { | |
56ae674c JK |
1060 | struct page *page; |
1061 | int err; | |
afcb7ca0 | 1062 | repeat: |
54b591df | 1063 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1064 | if (!page) |
1065 | return ERR_PTR(-ENOMEM); | |
1066 | ||
1067 | err = read_node_page(page, READ_SYNC); | |
86531d6b JK |
1068 | if (err < 0) { |
1069 | f2fs_put_page(page, 1); | |
e05df3b1 | 1070 | return ERR_PTR(err); |
e1c51b9f CY |
1071 | } else if (err == LOCKED_PAGE) { |
1072 | goto page_hit; | |
86531d6b | 1073 | } |
e05df3b1 | 1074 | |
e1c51b9f CY |
1075 | lock_page(page); |
1076 | ||
1077 | if (unlikely(!PageUptodate(page))) { | |
393ff91f JK |
1078 | f2fs_put_page(page, 1); |
1079 | return ERR_PTR(-EIO); | |
1080 | } | |
4ef51a8f | 1081 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1082 | f2fs_put_page(page, 1); |
1083 | goto repeat; | |
1084 | } | |
e1c51b9f CY |
1085 | page_hit: |
1086 | f2fs_bug_on(sbi, nid != nid_of_node(page)); | |
e05df3b1 JK |
1087 | return page; |
1088 | } | |
1089 | ||
0a8165d7 | 1090 | /* |
e05df3b1 JK |
1091 | * Return a locked page for the desired node page. |
1092 | * And, readahead MAX_RA_NODE number of node pages. | |
1093 | */ | |
1094 | struct page *get_node_page_ra(struct page *parent, int start) | |
1095 | { | |
4081363f | 1096 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); |
c718379b | 1097 | struct blk_plug plug; |
e05df3b1 | 1098 | struct page *page; |
56ae674c JK |
1099 | int err, i, end; |
1100 | nid_t nid; | |
e05df3b1 JK |
1101 | |
1102 | /* First, try getting the desired direct node. */ | |
1103 | nid = get_nid(parent, start, false); | |
1104 | if (!nid) | |
1105 | return ERR_PTR(-ENOENT); | |
afcb7ca0 | 1106 | repeat: |
4ef51a8f | 1107 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1108 | if (!page) |
1109 | return ERR_PTR(-ENOMEM); | |
1110 | ||
66d36a29 | 1111 | err = read_node_page(page, READ_SYNC); |
86531d6b JK |
1112 | if (err < 0) { |
1113 | f2fs_put_page(page, 1); | |
e05df3b1 | 1114 | return ERR_PTR(err); |
86531d6b | 1115 | } else if (err == LOCKED_PAGE) { |
56ae674c | 1116 | goto page_hit; |
86531d6b | 1117 | } |
e05df3b1 | 1118 | |
c718379b JK |
1119 | blk_start_plug(&plug); |
1120 | ||
e05df3b1 JK |
1121 | /* Then, try readahead for siblings of the desired node */ |
1122 | end = start + MAX_RA_NODE; | |
1123 | end = min(end, NIDS_PER_BLOCK); | |
1124 | for (i = start + 1; i < end; i++) { | |
e1c51b9f CY |
1125 | nid_t tnid = get_nid(parent, i, false); |
1126 | if (!tnid) | |
e05df3b1 | 1127 | continue; |
e1c51b9f | 1128 | ra_node_page(sbi, tnid); |
e05df3b1 JK |
1129 | } |
1130 | ||
c718379b JK |
1131 | blk_finish_plug(&plug); |
1132 | ||
e05df3b1 | 1133 | lock_page(page); |
e1c51b9f CY |
1134 | if (unlikely(!PageUptodate(page))) { |
1135 | f2fs_put_page(page, 1); | |
1136 | return ERR_PTR(-EIO); | |
1137 | } | |
4ef51a8f | 1138 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1139 | f2fs_put_page(page, 1); |
1140 | goto repeat; | |
1141 | } | |
e0f56cb4 | 1142 | page_hit: |
e1c51b9f | 1143 | f2fs_bug_on(sbi, nid != nid_of_node(page)); |
e05df3b1 JK |
1144 | return page; |
1145 | } | |
1146 | ||
1147 | void sync_inode_page(struct dnode_of_data *dn) | |
1148 | { | |
1149 | if (IS_INODE(dn->node_page) || dn->inode_page == dn->node_page) { | |
1150 | update_inode(dn->inode, dn->node_page); | |
1151 | } else if (dn->inode_page) { | |
1152 | if (!dn->inode_page_locked) | |
1153 | lock_page(dn->inode_page); | |
1154 | update_inode(dn->inode, dn->inode_page); | |
1155 | if (!dn->inode_page_locked) | |
1156 | unlock_page(dn->inode_page); | |
1157 | } else { | |
39936837 | 1158 | update_inode_page(dn->inode); |
e05df3b1 | 1159 | } |
93bae099 | 1160 | dn->node_changed = true; |
e05df3b1 JK |
1161 | } |
1162 | ||
1163 | int sync_node_pages(struct f2fs_sb_info *sbi, nid_t ino, | |
1164 | struct writeback_control *wbc) | |
1165 | { | |
e05df3b1 JK |
1166 | pgoff_t index, end; |
1167 | struct pagevec pvec; | |
1168 | int step = ino ? 2 : 0; | |
1169 | int nwritten = 0, wrote = 0; | |
1170 | ||
1171 | pagevec_init(&pvec, 0); | |
1172 | ||
1173 | next_step: | |
1174 | index = 0; | |
1175 | end = LONG_MAX; | |
1176 | ||
1177 | while (index <= end) { | |
1178 | int i, nr_pages; | |
4ef51a8f | 1179 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, |
e05df3b1 JK |
1180 | PAGECACHE_TAG_DIRTY, |
1181 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1182 | if (nr_pages == 0) | |
1183 | break; | |
1184 | ||
1185 | for (i = 0; i < nr_pages; i++) { | |
1186 | struct page *page = pvec.pages[i]; | |
1187 | ||
1188 | /* | |
1189 | * flushing sequence with step: | |
1190 | * 0. indirect nodes | |
1191 | * 1. dentry dnodes | |
1192 | * 2. file dnodes | |
1193 | */ | |
1194 | if (step == 0 && IS_DNODE(page)) | |
1195 | continue; | |
1196 | if (step == 1 && (!IS_DNODE(page) || | |
1197 | is_cold_node(page))) | |
1198 | continue; | |
1199 | if (step == 2 && (!IS_DNODE(page) || | |
1200 | !is_cold_node(page))) | |
1201 | continue; | |
1202 | ||
1203 | /* | |
1204 | * If an fsync mode, | |
1205 | * we should not skip writing node pages. | |
1206 | */ | |
1207 | if (ino && ino_of_node(page) == ino) | |
1208 | lock_page(page); | |
1209 | else if (!trylock_page(page)) | |
1210 | continue; | |
1211 | ||
4ef51a8f | 1212 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
e05df3b1 JK |
1213 | continue_unlock: |
1214 | unlock_page(page); | |
1215 | continue; | |
1216 | } | |
1217 | if (ino && ino_of_node(page) != ino) | |
1218 | goto continue_unlock; | |
1219 | ||
1220 | if (!PageDirty(page)) { | |
1221 | /* someone wrote it for us */ | |
1222 | goto continue_unlock; | |
1223 | } | |
1224 | ||
1225 | if (!clear_page_dirty_for_io(page)) | |
1226 | goto continue_unlock; | |
1227 | ||
1228 | /* called by fsync() */ | |
1229 | if (ino && IS_DNODE(page)) { | |
e05df3b1 | 1230 | set_fsync_mark(page, 1); |
2dcf51ab JK |
1231 | if (IS_INODE(page)) |
1232 | set_dentry_mark(page, | |
1233 | need_dentry_mark(sbi, ino)); | |
e05df3b1 JK |
1234 | nwritten++; |
1235 | } else { | |
1236 | set_fsync_mark(page, 0); | |
1237 | set_dentry_mark(page, 0); | |
1238 | } | |
52746519 JK |
1239 | |
1240 | if (NODE_MAPPING(sbi)->a_ops->writepage(page, wbc)) | |
1241 | unlock_page(page); | |
1242 | else | |
1243 | wrote++; | |
e05df3b1 JK |
1244 | |
1245 | if (--wbc->nr_to_write == 0) | |
1246 | break; | |
1247 | } | |
1248 | pagevec_release(&pvec); | |
1249 | cond_resched(); | |
1250 | ||
1251 | if (wbc->nr_to_write == 0) { | |
1252 | step = 2; | |
1253 | break; | |
1254 | } | |
1255 | } | |
1256 | ||
1257 | if (step < 2) { | |
1258 | step++; | |
1259 | goto next_step; | |
1260 | } | |
1261 | ||
1262 | if (wrote) | |
458e6197 | 1263 | f2fs_submit_merged_bio(sbi, NODE, WRITE); |
e05df3b1 JK |
1264 | return nwritten; |
1265 | } | |
1266 | ||
cfe58f9d JK |
1267 | int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino) |
1268 | { | |
cfe58f9d JK |
1269 | pgoff_t index = 0, end = LONG_MAX; |
1270 | struct pagevec pvec; | |
cfe58f9d JK |
1271 | int ret2 = 0, ret = 0; |
1272 | ||
1273 | pagevec_init(&pvec, 0); | |
4ef51a8f JK |
1274 | |
1275 | while (index <= end) { | |
1276 | int i, nr_pages; | |
1277 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1278 | PAGECACHE_TAG_WRITEBACK, | |
1279 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1280 | if (nr_pages == 0) | |
1281 | break; | |
cfe58f9d JK |
1282 | |
1283 | for (i = 0; i < nr_pages; i++) { | |
1284 | struct page *page = pvec.pages[i]; | |
1285 | ||
1286 | /* until radix tree lookup accepts end_index */ | |
cfb271d4 | 1287 | if (unlikely(page->index > end)) |
cfe58f9d JK |
1288 | continue; |
1289 | ||
4bf08ff6 | 1290 | if (ino && ino_of_node(page) == ino) { |
3cb5ad15 | 1291 | f2fs_wait_on_page_writeback(page, NODE); |
4bf08ff6 CY |
1292 | if (TestClearPageError(page)) |
1293 | ret = -EIO; | |
1294 | } | |
cfe58f9d JK |
1295 | } |
1296 | pagevec_release(&pvec); | |
1297 | cond_resched(); | |
1298 | } | |
1299 | ||
4ef51a8f | 1300 | if (unlikely(test_and_clear_bit(AS_ENOSPC, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d | 1301 | ret2 = -ENOSPC; |
4ef51a8f | 1302 | if (unlikely(test_and_clear_bit(AS_EIO, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d JK |
1303 | ret2 = -EIO; |
1304 | if (!ret) | |
1305 | ret = ret2; | |
1306 | return ret; | |
1307 | } | |
1308 | ||
e05df3b1 JK |
1309 | static int f2fs_write_node_page(struct page *page, |
1310 | struct writeback_control *wbc) | |
1311 | { | |
4081363f | 1312 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1313 | nid_t nid; |
e05df3b1 | 1314 | struct node_info ni; |
fb5566da | 1315 | struct f2fs_io_info fio = { |
05ca3632 | 1316 | .sbi = sbi, |
fb5566da | 1317 | .type = NODE, |
6c311ec6 | 1318 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
05ca3632 | 1319 | .page = page, |
4375a336 | 1320 | .encrypted_page = NULL, |
fb5566da | 1321 | }; |
e05df3b1 | 1322 | |
ecda0de3 CY |
1323 | trace_f2fs_writepage(page, NODE); |
1324 | ||
caf0047e | 1325 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
87a9bd26 | 1326 | goto redirty_out; |
cf779cab JK |
1327 | if (unlikely(f2fs_cp_error(sbi))) |
1328 | goto redirty_out; | |
87a9bd26 | 1329 | |
3cb5ad15 | 1330 | f2fs_wait_on_page_writeback(page, NODE); |
e05df3b1 | 1331 | |
e05df3b1 JK |
1332 | /* get old block addr of this node page */ |
1333 | nid = nid_of_node(page); | |
9850cf4a | 1334 | f2fs_bug_on(sbi, page->index != nid); |
e05df3b1 | 1335 | |
25b93346 JK |
1336 | if (wbc->for_reclaim) { |
1337 | if (!down_read_trylock(&sbi->node_write)) | |
1338 | goto redirty_out; | |
1339 | } else { | |
1340 | down_read(&sbi->node_write); | |
1341 | } | |
1342 | ||
e05df3b1 JK |
1343 | get_node_info(sbi, nid, &ni); |
1344 | ||
1345 | /* This page is already truncated */ | |
6bacf52f | 1346 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
2bca1e23 | 1347 | ClearPageUptodate(page); |
39936837 | 1348 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
25b93346 | 1349 | up_read(&sbi->node_write); |
39936837 JK |
1350 | unlock_page(page); |
1351 | return 0; | |
1352 | } | |
e05df3b1 JK |
1353 | |
1354 | set_page_writeback(page); | |
cf04e8eb | 1355 | fio.blk_addr = ni.blk_addr; |
05ca3632 | 1356 | write_node_page(nid, &fio); |
cf04e8eb | 1357 | set_node_addr(sbi, &ni, fio.blk_addr, is_fsync_dnode(page)); |
e05df3b1 | 1358 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
b3582c68 | 1359 | up_read(&sbi->node_write); |
e05df3b1 | 1360 | unlock_page(page); |
27c6bd60 JK |
1361 | |
1362 | if (wbc->for_reclaim) | |
1363 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1364 | ||
e05df3b1 | 1365 | return 0; |
87a9bd26 JK |
1366 | |
1367 | redirty_out: | |
76f60268 | 1368 | redirty_page_for_writepage(wbc, page); |
87a9bd26 | 1369 | return AOP_WRITEPAGE_ACTIVATE; |
e05df3b1 JK |
1370 | } |
1371 | ||
1372 | static int f2fs_write_node_pages(struct address_space *mapping, | |
1373 | struct writeback_control *wbc) | |
1374 | { | |
4081363f | 1375 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 1376 | long diff; |
e05df3b1 | 1377 | |
e5748434 CY |
1378 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
1379 | ||
4660f9c0 JK |
1380 | /* balancing f2fs's metadata in background */ |
1381 | f2fs_balance_fs_bg(sbi); | |
e05df3b1 | 1382 | |
a7fdffbd | 1383 | /* collect a number of dirty node pages and write together */ |
87d6f890 | 1384 | if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE)) |
d3baf95d | 1385 | goto skip_write; |
a7fdffbd | 1386 | |
50c8cdb3 | 1387 | diff = nr_pages_to_write(sbi, NODE, wbc); |
fb5566da | 1388 | wbc->sync_mode = WB_SYNC_NONE; |
e05df3b1 | 1389 | sync_node_pages(sbi, 0, wbc); |
50c8cdb3 | 1390 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
e05df3b1 | 1391 | return 0; |
d3baf95d JK |
1392 | |
1393 | skip_write: | |
1394 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES); | |
1395 | return 0; | |
e05df3b1 JK |
1396 | } |
1397 | ||
1398 | static int f2fs_set_node_page_dirty(struct page *page) | |
1399 | { | |
26c6b887 JK |
1400 | trace_f2fs_set_page_dirty(page, NODE); |
1401 | ||
e05df3b1 JK |
1402 | SetPageUptodate(page); |
1403 | if (!PageDirty(page)) { | |
1404 | __set_page_dirty_nobuffers(page); | |
4081363f | 1405 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 | 1406 | SetPagePrivate(page); |
9e4ded3f | 1407 | f2fs_trace_pid(page); |
e05df3b1 JK |
1408 | return 1; |
1409 | } | |
1410 | return 0; | |
1411 | } | |
1412 | ||
0a8165d7 | 1413 | /* |
e05df3b1 JK |
1414 | * Structure of the f2fs node operations |
1415 | */ | |
1416 | const struct address_space_operations f2fs_node_aops = { | |
1417 | .writepage = f2fs_write_node_page, | |
1418 | .writepages = f2fs_write_node_pages, | |
1419 | .set_page_dirty = f2fs_set_node_page_dirty, | |
487261f3 CY |
1420 | .invalidatepage = f2fs_invalidate_page, |
1421 | .releasepage = f2fs_release_page, | |
e05df3b1 JK |
1422 | }; |
1423 | ||
8a7ed66a JK |
1424 | static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i, |
1425 | nid_t n) | |
e05df3b1 | 1426 | { |
8a7ed66a | 1427 | return radix_tree_lookup(&nm_i->free_nid_root, n); |
e05df3b1 JK |
1428 | } |
1429 | ||
8a7ed66a JK |
1430 | static void __del_from_free_nid_list(struct f2fs_nm_info *nm_i, |
1431 | struct free_nid *i) | |
e05df3b1 JK |
1432 | { |
1433 | list_del(&i->list); | |
8a7ed66a | 1434 | radix_tree_delete(&nm_i->free_nid_root, i->nid); |
e05df3b1 JK |
1435 | } |
1436 | ||
6fb03f3a | 1437 | static int add_free_nid(struct f2fs_sb_info *sbi, nid_t nid, bool build) |
e05df3b1 | 1438 | { |
6fb03f3a | 1439 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1440 | struct free_nid *i; |
59bbd474 JK |
1441 | struct nat_entry *ne; |
1442 | bool allocated = false; | |
e05df3b1 | 1443 | |
6fb03f3a | 1444 | if (!available_free_memory(sbi, FREE_NIDS)) |
23d38844 | 1445 | return -1; |
9198aceb JK |
1446 | |
1447 | /* 0 nid should not be used */ | |
cfb271d4 | 1448 | if (unlikely(nid == 0)) |
9198aceb | 1449 | return 0; |
59bbd474 | 1450 | |
7bd59381 GZ |
1451 | if (build) { |
1452 | /* do not add allocated nids */ | |
8b26ef98 | 1453 | down_read(&nm_i->nat_tree_lock); |
7bd59381 | 1454 | ne = __lookup_nat_cache(nm_i, nid); |
8a7ed66a | 1455 | if (ne && |
7ef35e3b JK |
1456 | (!get_nat_flag(ne, IS_CHECKPOINTED) || |
1457 | nat_get_blkaddr(ne) != NULL_ADDR)) | |
7bd59381 | 1458 | allocated = true; |
8b26ef98 | 1459 | up_read(&nm_i->nat_tree_lock); |
7bd59381 GZ |
1460 | if (allocated) |
1461 | return 0; | |
e05df3b1 | 1462 | } |
7bd59381 GZ |
1463 | |
1464 | i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); | |
e05df3b1 JK |
1465 | i->nid = nid; |
1466 | i->state = NID_NEW; | |
1467 | ||
769ec6e5 JK |
1468 | if (radix_tree_preload(GFP_NOFS)) { |
1469 | kmem_cache_free(free_nid_slab, i); | |
1470 | return 0; | |
1471 | } | |
1472 | ||
e05df3b1 | 1473 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1474 | if (radix_tree_insert(&nm_i->free_nid_root, i->nid, i)) { |
e05df3b1 | 1475 | spin_unlock(&nm_i->free_nid_list_lock); |
769ec6e5 | 1476 | radix_tree_preload_end(); |
e05df3b1 JK |
1477 | kmem_cache_free(free_nid_slab, i); |
1478 | return 0; | |
1479 | } | |
1480 | list_add_tail(&i->list, &nm_i->free_nid_list); | |
1481 | nm_i->fcnt++; | |
1482 | spin_unlock(&nm_i->free_nid_list_lock); | |
769ec6e5 | 1483 | radix_tree_preload_end(); |
e05df3b1 JK |
1484 | return 1; |
1485 | } | |
1486 | ||
1487 | static void remove_free_nid(struct f2fs_nm_info *nm_i, nid_t nid) | |
1488 | { | |
1489 | struct free_nid *i; | |
cf0ee0f0 CY |
1490 | bool need_free = false; |
1491 | ||
e05df3b1 | 1492 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1493 | i = __lookup_free_nid_list(nm_i, nid); |
e05df3b1 | 1494 | if (i && i->state == NID_NEW) { |
8a7ed66a | 1495 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1496 | nm_i->fcnt--; |
cf0ee0f0 | 1497 | need_free = true; |
e05df3b1 JK |
1498 | } |
1499 | spin_unlock(&nm_i->free_nid_list_lock); | |
cf0ee0f0 CY |
1500 | |
1501 | if (need_free) | |
1502 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1503 | } |
1504 | ||
6fb03f3a | 1505 | static void scan_nat_page(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
1506 | struct page *nat_page, nid_t start_nid) |
1507 | { | |
6fb03f3a | 1508 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 JK |
1509 | struct f2fs_nat_block *nat_blk = page_address(nat_page); |
1510 | block_t blk_addr; | |
e05df3b1 JK |
1511 | int i; |
1512 | ||
e05df3b1 JK |
1513 | i = start_nid % NAT_ENTRY_PER_BLOCK; |
1514 | ||
1515 | for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { | |
23d38844 | 1516 | |
cfb271d4 | 1517 | if (unlikely(start_nid >= nm_i->max_nid)) |
04431c44 | 1518 | break; |
23d38844 HL |
1519 | |
1520 | blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); | |
9850cf4a | 1521 | f2fs_bug_on(sbi, blk_addr == NEW_ADDR); |
23d38844 | 1522 | if (blk_addr == NULL_ADDR) { |
6fb03f3a | 1523 | if (add_free_nid(sbi, start_nid, true) < 0) |
23d38844 HL |
1524 | break; |
1525 | } | |
e05df3b1 | 1526 | } |
e05df3b1 JK |
1527 | } |
1528 | ||
1529 | static void build_free_nids(struct f2fs_sb_info *sbi) | |
1530 | { | |
e05df3b1 JK |
1531 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1532 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1533 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
8760952d | 1534 | int i = 0; |
55008d84 | 1535 | nid_t nid = nm_i->next_scan_nid; |
e05df3b1 | 1536 | |
55008d84 JK |
1537 | /* Enough entries */ |
1538 | if (nm_i->fcnt > NAT_ENTRY_PER_BLOCK) | |
1539 | return; | |
e05df3b1 | 1540 | |
55008d84 | 1541 | /* readahead nat pages to be scanned */ |
26879fb1 CY |
1542 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES, |
1543 | META_NAT, true); | |
e05df3b1 JK |
1544 | |
1545 | while (1) { | |
1546 | struct page *page = get_current_nat_page(sbi, nid); | |
1547 | ||
6fb03f3a | 1548 | scan_nat_page(sbi, page, nid); |
e05df3b1 JK |
1549 | f2fs_put_page(page, 1); |
1550 | ||
1551 | nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); | |
cfb271d4 | 1552 | if (unlikely(nid >= nm_i->max_nid)) |
e05df3b1 | 1553 | nid = 0; |
55008d84 | 1554 | |
a6d494b6 | 1555 | if (++i >= FREE_NID_PAGES) |
e05df3b1 JK |
1556 | break; |
1557 | } | |
1558 | ||
55008d84 JK |
1559 | /* go to the next free nat pages to find free nids abundantly */ |
1560 | nm_i->next_scan_nid = nid; | |
e05df3b1 JK |
1561 | |
1562 | /* find free nids from current sum_pages */ | |
1563 | mutex_lock(&curseg->curseg_mutex); | |
1564 | for (i = 0; i < nats_in_cursum(sum); i++) { | |
1565 | block_t addr = le32_to_cpu(nat_in_journal(sum, i).block_addr); | |
1566 | nid = le32_to_cpu(nid_in_journal(sum, i)); | |
1567 | if (addr == NULL_ADDR) | |
6fb03f3a | 1568 | add_free_nid(sbi, nid, true); |
e05df3b1 JK |
1569 | else |
1570 | remove_free_nid(nm_i, nid); | |
1571 | } | |
1572 | mutex_unlock(&curseg->curseg_mutex); | |
2db2388f CY |
1573 | |
1574 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid), | |
ea1a29a0 | 1575 | nm_i->ra_nid_pages, META_NAT, false); |
e05df3b1 JK |
1576 | } |
1577 | ||
1578 | /* | |
1579 | * If this function returns success, caller can obtain a new nid | |
1580 | * from second parameter of this function. | |
1581 | * The returned nid could be used ino as well as nid when inode is created. | |
1582 | */ | |
1583 | bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) | |
1584 | { | |
1585 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1586 | struct free_nid *i = NULL; | |
e05df3b1 | 1587 | retry: |
7ee0eeab | 1588 | if (unlikely(sbi->total_valid_node_count + 1 > nm_i->available_nids)) |
55008d84 | 1589 | return false; |
e05df3b1 | 1590 | |
e05df3b1 | 1591 | spin_lock(&nm_i->free_nid_list_lock); |
e05df3b1 | 1592 | |
55008d84 | 1593 | /* We should not use stale free nids created by build_free_nids */ |
f978f5a0 | 1594 | if (nm_i->fcnt && !on_build_free_nids(nm_i)) { |
24928634 JK |
1595 | struct node_info ni; |
1596 | ||
9850cf4a | 1597 | f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list)); |
2d7b822a | 1598 | list_for_each_entry(i, &nm_i->free_nid_list, list) |
55008d84 JK |
1599 | if (i->state == NID_NEW) |
1600 | break; | |
e05df3b1 | 1601 | |
9850cf4a | 1602 | f2fs_bug_on(sbi, i->state != NID_NEW); |
55008d84 JK |
1603 | *nid = i->nid; |
1604 | i->state = NID_ALLOC; | |
1605 | nm_i->fcnt--; | |
1606 | spin_unlock(&nm_i->free_nid_list_lock); | |
24928634 JK |
1607 | |
1608 | /* check nid is allocated already */ | |
1609 | get_node_info(sbi, *nid, &ni); | |
1610 | if (ni.blk_addr != NULL_ADDR) { | |
1611 | alloc_nid_done(sbi, *nid); | |
1612 | goto retry; | |
1613 | } | |
55008d84 JK |
1614 | return true; |
1615 | } | |
e05df3b1 | 1616 | spin_unlock(&nm_i->free_nid_list_lock); |
55008d84 JK |
1617 | |
1618 | /* Let's scan nat pages and its caches to get free nids */ | |
1619 | mutex_lock(&nm_i->build_lock); | |
55008d84 | 1620 | build_free_nids(sbi); |
55008d84 JK |
1621 | mutex_unlock(&nm_i->build_lock); |
1622 | goto retry; | |
e05df3b1 JK |
1623 | } |
1624 | ||
0a8165d7 | 1625 | /* |
e05df3b1 JK |
1626 | * alloc_nid() should be called prior to this function. |
1627 | */ | |
1628 | void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) | |
1629 | { | |
1630 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1631 | struct free_nid *i; | |
1632 | ||
1633 | spin_lock(&nm_i->free_nid_list_lock); | |
8a7ed66a | 1634 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1635 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
8a7ed66a | 1636 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1637 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1638 | |
1639 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1640 | } |
1641 | ||
0a8165d7 | 1642 | /* |
e05df3b1 JK |
1643 | * alloc_nid() should be called prior to this function. |
1644 | */ | |
1645 | void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) | |
1646 | { | |
49952fa1 JK |
1647 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1648 | struct free_nid *i; | |
cf0ee0f0 | 1649 | bool need_free = false; |
49952fa1 | 1650 | |
65985d93 JK |
1651 | if (!nid) |
1652 | return; | |
1653 | ||
49952fa1 | 1654 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1655 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1656 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
6fb03f3a | 1657 | if (!available_free_memory(sbi, FREE_NIDS)) { |
8a7ed66a | 1658 | __del_from_free_nid_list(nm_i, i); |
cf0ee0f0 | 1659 | need_free = true; |
95630cba HL |
1660 | } else { |
1661 | i->state = NID_NEW; | |
1662 | nm_i->fcnt++; | |
1663 | } | |
49952fa1 | 1664 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1665 | |
1666 | if (need_free) | |
1667 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1668 | } |
1669 | ||
31696580 CY |
1670 | int try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink) |
1671 | { | |
1672 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1673 | struct free_nid *i, *next; | |
1674 | int nr = nr_shrink; | |
1675 | ||
1676 | if (!mutex_trylock(&nm_i->build_lock)) | |
1677 | return 0; | |
1678 | ||
1679 | spin_lock(&nm_i->free_nid_list_lock); | |
1680 | list_for_each_entry_safe(i, next, &nm_i->free_nid_list, list) { | |
1681 | if (nr_shrink <= 0 || nm_i->fcnt <= NAT_ENTRY_PER_BLOCK) | |
1682 | break; | |
1683 | if (i->state == NID_ALLOC) | |
1684 | continue; | |
1685 | __del_from_free_nid_list(nm_i, i); | |
31696580 | 1686 | kmem_cache_free(free_nid_slab, i); |
f7409d0f | 1687 | nm_i->fcnt--; |
31696580 | 1688 | nr_shrink--; |
31696580 CY |
1689 | } |
1690 | spin_unlock(&nm_i->free_nid_list_lock); | |
1691 | mutex_unlock(&nm_i->build_lock); | |
1692 | ||
1693 | return nr - nr_shrink; | |
1694 | } | |
1695 | ||
70cfed88 | 1696 | void recover_inline_xattr(struct inode *inode, struct page *page) |
28cdce04 | 1697 | { |
28cdce04 CY |
1698 | void *src_addr, *dst_addr; |
1699 | size_t inline_size; | |
1700 | struct page *ipage; | |
1701 | struct f2fs_inode *ri; | |
1702 | ||
4081363f | 1703 | ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino); |
9850cf4a | 1704 | f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage)); |
28cdce04 | 1705 | |
e3b4d43f JK |
1706 | ri = F2FS_INODE(page); |
1707 | if (!(ri->i_inline & F2FS_INLINE_XATTR)) { | |
1708 | clear_inode_flag(F2FS_I(inode), FI_INLINE_XATTR); | |
1709 | goto update_inode; | |
1710 | } | |
1711 | ||
28cdce04 CY |
1712 | dst_addr = inline_xattr_addr(ipage); |
1713 | src_addr = inline_xattr_addr(page); | |
1714 | inline_size = inline_xattr_size(inode); | |
1715 | ||
54b591df | 1716 | f2fs_wait_on_page_writeback(ipage, NODE); |
28cdce04 | 1717 | memcpy(dst_addr, src_addr, inline_size); |
e3b4d43f | 1718 | update_inode: |
28cdce04 CY |
1719 | update_inode(inode, ipage); |
1720 | f2fs_put_page(ipage, 1); | |
1721 | } | |
1722 | ||
1c35a90e | 1723 | void recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr) |
abb2366c | 1724 | { |
4081363f | 1725 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
abb2366c JK |
1726 | nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid; |
1727 | nid_t new_xnid = nid_of_node(page); | |
1728 | struct node_info ni; | |
1729 | ||
abb2366c JK |
1730 | /* 1: invalidate the previous xattr nid */ |
1731 | if (!prev_xnid) | |
1732 | goto recover_xnid; | |
1733 | ||
1734 | /* Deallocate node address */ | |
1735 | get_node_info(sbi, prev_xnid, &ni); | |
9850cf4a | 1736 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
abb2366c JK |
1737 | invalidate_blocks(sbi, ni.blk_addr); |
1738 | dec_valid_node_count(sbi, inode); | |
479f40c4 | 1739 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
abb2366c JK |
1740 | |
1741 | recover_xnid: | |
1742 | /* 2: allocate new xattr nid */ | |
1743 | if (unlikely(!inc_valid_node_count(sbi, inode))) | |
9850cf4a | 1744 | f2fs_bug_on(sbi, 1); |
abb2366c JK |
1745 | |
1746 | remove_free_nid(NM_I(sbi), new_xnid); | |
1747 | get_node_info(sbi, new_xnid, &ni); | |
1748 | ni.ino = inode->i_ino; | |
479f40c4 | 1749 | set_node_addr(sbi, &ni, NEW_ADDR, false); |
abb2366c JK |
1750 | F2FS_I(inode)->i_xattr_nid = new_xnid; |
1751 | ||
1752 | /* 3: update xattr blkaddr */ | |
1753 | refresh_sit_entry(sbi, NEW_ADDR, blkaddr); | |
479f40c4 | 1754 | set_node_addr(sbi, &ni, blkaddr, false); |
abb2366c JK |
1755 | |
1756 | update_inode_page(inode); | |
abb2366c JK |
1757 | } |
1758 | ||
e05df3b1 JK |
1759 | int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) |
1760 | { | |
58bfaf44 | 1761 | struct f2fs_inode *src, *dst; |
e05df3b1 JK |
1762 | nid_t ino = ino_of_node(page); |
1763 | struct node_info old_ni, new_ni; | |
1764 | struct page *ipage; | |
1765 | ||
e8271fa3 JK |
1766 | get_node_info(sbi, ino, &old_ni); |
1767 | ||
1768 | if (unlikely(old_ni.blk_addr != NULL_ADDR)) | |
1769 | return -EINVAL; | |
1770 | ||
4ef51a8f | 1771 | ipage = grab_cache_page(NODE_MAPPING(sbi), ino); |
e05df3b1 JK |
1772 | if (!ipage) |
1773 | return -ENOMEM; | |
1774 | ||
e1c42045 | 1775 | /* Should not use this inode from free nid list */ |
e05df3b1 JK |
1776 | remove_free_nid(NM_I(sbi), ino); |
1777 | ||
e05df3b1 JK |
1778 | SetPageUptodate(ipage); |
1779 | fill_node_footer(ipage, ino, ino, 0, true); | |
1780 | ||
58bfaf44 JK |
1781 | src = F2FS_INODE(page); |
1782 | dst = F2FS_INODE(ipage); | |
e05df3b1 | 1783 | |
58bfaf44 JK |
1784 | memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); |
1785 | dst->i_size = 0; | |
1786 | dst->i_blocks = cpu_to_le64(1); | |
1787 | dst->i_links = cpu_to_le32(1); | |
1788 | dst->i_xattr_nid = 0; | |
617deb8c | 1789 | dst->i_inline = src->i_inline & F2FS_INLINE_XATTR; |
e05df3b1 JK |
1790 | |
1791 | new_ni = old_ni; | |
1792 | new_ni.ino = ino; | |
1793 | ||
cfb271d4 | 1794 | if (unlikely(!inc_valid_node_count(sbi, NULL))) |
65e5cd0a | 1795 | WARN_ON(1); |
479f40c4 | 1796 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
e05df3b1 | 1797 | inc_valid_inode_count(sbi); |
617deb8c | 1798 | set_page_dirty(ipage); |
e05df3b1 JK |
1799 | f2fs_put_page(ipage, 1); |
1800 | return 0; | |
1801 | } | |
1802 | ||
1803 | int restore_node_summary(struct f2fs_sb_info *sbi, | |
1804 | unsigned int segno, struct f2fs_summary_block *sum) | |
1805 | { | |
1806 | struct f2fs_node *rn; | |
1807 | struct f2fs_summary *sum_entry; | |
e05df3b1 | 1808 | block_t addr; |
90a893c7 | 1809 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
9ecf4b80 | 1810 | int i, idx, last_offset, nrpages; |
e05df3b1 JK |
1811 | |
1812 | /* scan the node segment */ | |
1813 | last_offset = sbi->blocks_per_seg; | |
1814 | addr = START_BLOCK(sbi, segno); | |
1815 | sum_entry = &sum->entries[0]; | |
1816 | ||
9ecf4b80 | 1817 | for (i = 0; i < last_offset; i += nrpages, addr += nrpages) { |
9af0ff1c | 1818 | nrpages = min(last_offset - i, bio_blocks); |
393ff91f | 1819 | |
e1c42045 | 1820 | /* readahead node pages */ |
26879fb1 | 1821 | ra_meta_pages(sbi, addr, nrpages, META_POR, true); |
e05df3b1 | 1822 | |
9ecf4b80 | 1823 | for (idx = addr; idx < addr + nrpages; idx++) { |
2b947003 | 1824 | struct page *page = get_tmp_page(sbi, idx); |
9af0ff1c | 1825 | |
9ecf4b80 CY |
1826 | rn = F2FS_NODE(page); |
1827 | sum_entry->nid = rn->footer.nid; | |
1828 | sum_entry->version = 0; | |
1829 | sum_entry->ofs_in_node = 0; | |
1830 | sum_entry++; | |
1831 | f2fs_put_page(page, 1); | |
9af0ff1c | 1832 | } |
bac4eef6 | 1833 | |
9ecf4b80 | 1834 | invalidate_mapping_pages(META_MAPPING(sbi), addr, |
bac4eef6 | 1835 | addr + nrpages); |
e05df3b1 | 1836 | } |
9ecf4b80 | 1837 | return 0; |
e05df3b1 JK |
1838 | } |
1839 | ||
aec71382 | 1840 | static void remove_nats_in_journal(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
1841 | { |
1842 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1843 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1844 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1845 | int i; | |
1846 | ||
1847 | mutex_lock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1848 | for (i = 0; i < nats_in_cursum(sum); i++) { |
1849 | struct nat_entry *ne; | |
1850 | struct f2fs_nat_entry raw_ne; | |
1851 | nid_t nid = le32_to_cpu(nid_in_journal(sum, i)); | |
1852 | ||
1853 | raw_ne = nat_in_journal(sum, i); | |
9be32d72 | 1854 | |
8b26ef98 | 1855 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 | 1856 | ne = __lookup_nat_cache(nm_i, nid); |
e05df3b1 | 1857 | if (!ne) { |
9be32d72 JK |
1858 | ne = grab_nat_entry(nm_i, nid); |
1859 | node_info_from_raw_nat(&ne->ni, &raw_ne); | |
e05df3b1 | 1860 | } |
e05df3b1 | 1861 | __set_nat_cache_dirty(nm_i, ne); |
8b26ef98 | 1862 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
1863 | } |
1864 | update_nats_in_cursum(sum, -i); | |
1865 | mutex_unlock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1866 | } |
1867 | ||
309cc2b6 JK |
1868 | static void __adjust_nat_entry_set(struct nat_entry_set *nes, |
1869 | struct list_head *head, int max) | |
e05df3b1 | 1870 | { |
309cc2b6 | 1871 | struct nat_entry_set *cur; |
e05df3b1 | 1872 | |
309cc2b6 JK |
1873 | if (nes->entry_cnt >= max) |
1874 | goto add_out; | |
e05df3b1 | 1875 | |
309cc2b6 JK |
1876 | list_for_each_entry(cur, head, set_list) { |
1877 | if (cur->entry_cnt >= nes->entry_cnt) { | |
1878 | list_add(&nes->set_list, cur->set_list.prev); | |
1879 | return; | |
1880 | } | |
aec71382 | 1881 | } |
309cc2b6 JK |
1882 | add_out: |
1883 | list_add_tail(&nes->set_list, head); | |
1884 | } | |
e05df3b1 | 1885 | |
309cc2b6 JK |
1886 | static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, |
1887 | struct nat_entry_set *set) | |
1888 | { | |
1889 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1890 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1891 | nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK; | |
1892 | bool to_journal = true; | |
1893 | struct f2fs_nat_block *nat_blk; | |
1894 | struct nat_entry *ne, *cur; | |
1895 | struct page *page = NULL; | |
57ed1e95 | 1896 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1897 | |
aec71382 CY |
1898 | /* |
1899 | * there are two steps to flush nat entries: | |
1900 | * #1, flush nat entries to journal in current hot data summary block. | |
1901 | * #2, flush nat entries to nat page. | |
1902 | */ | |
309cc2b6 JK |
1903 | if (!__has_cursum_space(sum, set->entry_cnt, NAT_JOURNAL)) |
1904 | to_journal = false; | |
1905 | ||
1906 | if (to_journal) { | |
1907 | mutex_lock(&curseg->curseg_mutex); | |
1908 | } else { | |
1909 | page = get_next_nat_page(sbi, start_nid); | |
1910 | nat_blk = page_address(page); | |
1911 | f2fs_bug_on(sbi, !nat_blk); | |
1912 | } | |
aec71382 | 1913 | |
309cc2b6 JK |
1914 | /* flush dirty nats in nat entry set */ |
1915 | list_for_each_entry_safe(ne, cur, &set->entry_list, list) { | |
1916 | struct f2fs_nat_entry *raw_ne; | |
1917 | nid_t nid = nat_get_nid(ne); | |
1918 | int offset; | |
1919 | ||
1920 | if (nat_get_blkaddr(ne) == NEW_ADDR) | |
1921 | continue; | |
aec71382 CY |
1922 | |
1923 | if (to_journal) { | |
309cc2b6 JK |
1924 | offset = lookup_journal_in_cursum(sum, |
1925 | NAT_JOURNAL, nid, 1); | |
1926 | f2fs_bug_on(sbi, offset < 0); | |
1927 | raw_ne = &nat_in_journal(sum, offset); | |
1928 | nid_in_journal(sum, offset) = cpu_to_le32(nid); | |
aec71382 | 1929 | } else { |
309cc2b6 | 1930 | raw_ne = &nat_blk->entries[nid - start_nid]; |
e05df3b1 | 1931 | } |
309cc2b6 | 1932 | raw_nat_from_node_info(raw_ne, &ne->ni); |
e05df3b1 | 1933 | |
8b26ef98 | 1934 | down_write(&NM_I(sbi)->nat_tree_lock); |
309cc2b6 JK |
1935 | nat_reset_flag(ne); |
1936 | __clear_nat_cache_dirty(NM_I(sbi), ne); | |
8b26ef98 | 1937 | up_write(&NM_I(sbi)->nat_tree_lock); |
aec71382 | 1938 | |
309cc2b6 JK |
1939 | if (nat_get_blkaddr(ne) == NULL_ADDR) |
1940 | add_free_nid(sbi, nid, false); | |
1941 | } | |
e05df3b1 | 1942 | |
309cc2b6 JK |
1943 | if (to_journal) |
1944 | mutex_unlock(&curseg->curseg_mutex); | |
1945 | else | |
1946 | f2fs_put_page(page, 1); | |
aec71382 | 1947 | |
80ec2e91 CL |
1948 | f2fs_bug_on(sbi, set->entry_cnt); |
1949 | ||
57ed1e95 | 1950 | down_write(&nm_i->nat_tree_lock); |
80ec2e91 | 1951 | radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); |
57ed1e95 | 1952 | up_write(&nm_i->nat_tree_lock); |
80ec2e91 | 1953 | kmem_cache_free(nat_entry_set_slab, set); |
309cc2b6 | 1954 | } |
aec71382 | 1955 | |
309cc2b6 JK |
1956 | /* |
1957 | * This function is called during the checkpointing process. | |
1958 | */ | |
1959 | void flush_nat_entries(struct f2fs_sb_info *sbi) | |
1960 | { | |
1961 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1962 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1963 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
7aed0d45 | 1964 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
309cc2b6 JK |
1965 | struct nat_entry_set *set, *tmp; |
1966 | unsigned int found; | |
1967 | nid_t set_idx = 0; | |
1968 | LIST_HEAD(sets); | |
1969 | ||
20d047c8 CL |
1970 | if (!nm_i->dirty_nat_cnt) |
1971 | return; | |
309cc2b6 JK |
1972 | /* |
1973 | * if there are no enough space in journal to store dirty nat | |
1974 | * entries, remove all entries from journal and merge them | |
1975 | * into nat entry set. | |
1976 | */ | |
1977 | if (!__has_cursum_space(sum, nm_i->dirty_nat_cnt, NAT_JOURNAL)) | |
1978 | remove_nats_in_journal(sbi); | |
1979 | ||
57ed1e95 | 1980 | down_write(&nm_i->nat_tree_lock); |
309cc2b6 | 1981 | while ((found = __gang_lookup_nat_set(nm_i, |
7aed0d45 | 1982 | set_idx, SETVEC_SIZE, setvec))) { |
309cc2b6 JK |
1983 | unsigned idx; |
1984 | set_idx = setvec[found - 1]->set + 1; | |
1985 | for (idx = 0; idx < found; idx++) | |
1986 | __adjust_nat_entry_set(setvec[idx], &sets, | |
1987 | MAX_NAT_JENTRIES(sum)); | |
e05df3b1 | 1988 | } |
57ed1e95 | 1989 | up_write(&nm_i->nat_tree_lock); |
aec71382 | 1990 | |
309cc2b6 JK |
1991 | /* flush dirty nats in nat entry set */ |
1992 | list_for_each_entry_safe(set, tmp, &sets, set_list) | |
1993 | __flush_nat_entry_set(sbi, set); | |
1994 | ||
9850cf4a | 1995 | f2fs_bug_on(sbi, nm_i->dirty_nat_cnt); |
e05df3b1 JK |
1996 | } |
1997 | ||
1998 | static int init_node_manager(struct f2fs_sb_info *sbi) | |
1999 | { | |
2000 | struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); | |
2001 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2002 | unsigned char *version_bitmap; | |
2003 | unsigned int nat_segs, nat_blocks; | |
2004 | ||
2005 | nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); | |
2006 | ||
2007 | /* segment_count_nat includes pair segment so divide to 2. */ | |
2008 | nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; | |
2009 | nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); | |
b63da15e | 2010 | |
7ee0eeab JK |
2011 | nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks; |
2012 | ||
b63da15e | 2013 | /* not used nids: 0, node, meta, (and root counted as valid node) */ |
c200b1aa | 2014 | nm_i->available_nids = nm_i->max_nid - F2FS_RESERVED_NODE_NUM; |
e05df3b1 JK |
2015 | nm_i->fcnt = 0; |
2016 | nm_i->nat_cnt = 0; | |
cdfc41c1 | 2017 | nm_i->ram_thresh = DEF_RAM_THRESHOLD; |
ea1a29a0 | 2018 | nm_i->ra_nid_pages = DEF_RA_NID_PAGES; |
e05df3b1 | 2019 | |
8a7ed66a | 2020 | INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC); |
e05df3b1 | 2021 | INIT_LIST_HEAD(&nm_i->free_nid_list); |
769ec6e5 JK |
2022 | INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO); |
2023 | INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO); | |
e05df3b1 | 2024 | INIT_LIST_HEAD(&nm_i->nat_entries); |
e05df3b1 JK |
2025 | |
2026 | mutex_init(&nm_i->build_lock); | |
2027 | spin_lock_init(&nm_i->free_nid_list_lock); | |
8b26ef98 | 2028 | init_rwsem(&nm_i->nat_tree_lock); |
e05df3b1 | 2029 | |
e05df3b1 | 2030 | nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); |
79b5793b | 2031 | nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); |
e05df3b1 JK |
2032 | version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); |
2033 | if (!version_bitmap) | |
2034 | return -EFAULT; | |
2035 | ||
79b5793b AG |
2036 | nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, |
2037 | GFP_KERNEL); | |
2038 | if (!nm_i->nat_bitmap) | |
2039 | return -ENOMEM; | |
e05df3b1 JK |
2040 | return 0; |
2041 | } | |
2042 | ||
2043 | int build_node_manager(struct f2fs_sb_info *sbi) | |
2044 | { | |
2045 | int err; | |
2046 | ||
2047 | sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL); | |
2048 | if (!sbi->nm_info) | |
2049 | return -ENOMEM; | |
2050 | ||
2051 | err = init_node_manager(sbi); | |
2052 | if (err) | |
2053 | return err; | |
2054 | ||
2055 | build_free_nids(sbi); | |
2056 | return 0; | |
2057 | } | |
2058 | ||
2059 | void destroy_node_manager(struct f2fs_sb_info *sbi) | |
2060 | { | |
2061 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2062 | struct free_nid *i, *next_i; | |
2063 | struct nat_entry *natvec[NATVEC_SIZE]; | |
7aed0d45 | 2064 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
e05df3b1 JK |
2065 | nid_t nid = 0; |
2066 | unsigned int found; | |
2067 | ||
2068 | if (!nm_i) | |
2069 | return; | |
2070 | ||
2071 | /* destroy free nid list */ | |
2072 | spin_lock(&nm_i->free_nid_list_lock); | |
2073 | list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) { | |
9850cf4a | 2074 | f2fs_bug_on(sbi, i->state == NID_ALLOC); |
8a7ed66a | 2075 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 2076 | nm_i->fcnt--; |
cf0ee0f0 CY |
2077 | spin_unlock(&nm_i->free_nid_list_lock); |
2078 | kmem_cache_free(free_nid_slab, i); | |
2079 | spin_lock(&nm_i->free_nid_list_lock); | |
e05df3b1 | 2080 | } |
9850cf4a | 2081 | f2fs_bug_on(sbi, nm_i->fcnt); |
e05df3b1 JK |
2082 | spin_unlock(&nm_i->free_nid_list_lock); |
2083 | ||
2084 | /* destroy nat cache */ | |
8b26ef98 | 2085 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2086 | while ((found = __gang_lookup_nat_cache(nm_i, |
2087 | nid, NATVEC_SIZE, natvec))) { | |
2088 | unsigned idx; | |
7aed0d45 | 2089 | |
b6ce391e GZ |
2090 | nid = nat_get_nid(natvec[found - 1]) + 1; |
2091 | for (idx = 0; idx < found; idx++) | |
2092 | __del_from_nat_cache(nm_i, natvec[idx]); | |
e05df3b1 | 2093 | } |
9850cf4a | 2094 | f2fs_bug_on(sbi, nm_i->nat_cnt); |
7aed0d45 JK |
2095 | |
2096 | /* destroy nat set cache */ | |
2097 | nid = 0; | |
2098 | while ((found = __gang_lookup_nat_set(nm_i, | |
2099 | nid, SETVEC_SIZE, setvec))) { | |
2100 | unsigned idx; | |
2101 | ||
2102 | nid = setvec[found - 1]->set + 1; | |
2103 | for (idx = 0; idx < found; idx++) { | |
2104 | /* entry_cnt is not zero, when cp_error was occurred */ | |
2105 | f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list)); | |
2106 | radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set); | |
2107 | kmem_cache_free(nat_entry_set_slab, setvec[idx]); | |
2108 | } | |
2109 | } | |
8b26ef98 | 2110 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2111 | |
2112 | kfree(nm_i->nat_bitmap); | |
2113 | sbi->nm_info = NULL; | |
2114 | kfree(nm_i); | |
2115 | } | |
2116 | ||
6e6093a8 | 2117 | int __init create_node_manager_caches(void) |
e05df3b1 JK |
2118 | { |
2119 | nat_entry_slab = f2fs_kmem_cache_create("nat_entry", | |
e8512d2e | 2120 | sizeof(struct nat_entry)); |
e05df3b1 | 2121 | if (!nat_entry_slab) |
aec71382 | 2122 | goto fail; |
e05df3b1 JK |
2123 | |
2124 | free_nid_slab = f2fs_kmem_cache_create("free_nid", | |
e8512d2e | 2125 | sizeof(struct free_nid)); |
aec71382 | 2126 | if (!free_nid_slab) |
ce3e6d25 | 2127 | goto destroy_nat_entry; |
aec71382 CY |
2128 | |
2129 | nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set", | |
2130 | sizeof(struct nat_entry_set)); | |
2131 | if (!nat_entry_set_slab) | |
ce3e6d25 | 2132 | goto destroy_free_nid; |
e05df3b1 | 2133 | return 0; |
aec71382 | 2134 | |
ce3e6d25 | 2135 | destroy_free_nid: |
aec71382 | 2136 | kmem_cache_destroy(free_nid_slab); |
ce3e6d25 | 2137 | destroy_nat_entry: |
aec71382 CY |
2138 | kmem_cache_destroy(nat_entry_slab); |
2139 | fail: | |
2140 | return -ENOMEM; | |
e05df3b1 JK |
2141 | } |
2142 | ||
2143 | void destroy_node_manager_caches(void) | |
2144 | { | |
aec71382 | 2145 | kmem_cache_destroy(nat_entry_set_slab); |
e05df3b1 JK |
2146 | kmem_cache_destroy(free_nid_slab); |
2147 | kmem_cache_destroy(nat_entry_slab); | |
2148 | } |