]>
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" | |
87905682 | 22 | #include "xattr.h" |
9e4ded3f | 23 | #include "trace.h" |
51dd6249 | 24 | #include <trace/events/f2fs.h> |
e05df3b1 | 25 | |
68afcf2d | 26 | #define on_build_free_nids(nmi) mutex_is_locked(&(nm_i)->build_lock) |
f978f5a0 | 27 | |
e05df3b1 JK |
28 | static struct kmem_cache *nat_entry_slab; |
29 | static struct kmem_cache *free_nid_slab; | |
aec71382 | 30 | static struct kmem_cache *nat_entry_set_slab; |
e05df3b1 | 31 | |
6fb03f3a | 32 | bool available_free_memory(struct f2fs_sb_info *sbi, int type) |
cdfc41c1 | 33 | { |
6fb03f3a | 34 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
cdfc41c1 | 35 | struct sysinfo val; |
e5e7ea3c | 36 | unsigned long avail_ram; |
cdfc41c1 | 37 | unsigned long mem_size = 0; |
6fb03f3a | 38 | bool res = false; |
cdfc41c1 JK |
39 | |
40 | si_meminfo(&val); | |
e5e7ea3c JK |
41 | |
42 | /* only uses low memory */ | |
43 | avail_ram = val.totalram - val.totalhigh; | |
44 | ||
429511cd CY |
45 | /* |
46 | * give 25%, 25%, 50%, 50%, 50% memory for each components respectively | |
47 | */ | |
6fb03f3a | 48 | if (type == FREE_NIDS) { |
b8559dc2 CY |
49 | mem_size = (nm_i->nid_cnt[FREE_NID_LIST] * |
50 | sizeof(struct free_nid)) >> PAGE_SHIFT; | |
e5e7ea3c | 51 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); |
6fb03f3a | 52 | } else if (type == NAT_ENTRIES) { |
e5e7ea3c | 53 | mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >> |
09cbfeaf | 54 | PAGE_SHIFT; |
e5e7ea3c | 55 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); |
e589c2c4 JK |
56 | if (excess_cached_nats(sbi)) |
57 | res = false; | |
a1257023 JK |
58 | } else if (type == DIRTY_DENTS) { |
59 | if (sbi->sb->s_bdi->wb.dirty_exceeded) | |
60 | return false; | |
61 | mem_size = get_pages(sbi, F2FS_DIRTY_DENTS); | |
62 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); | |
e5e7ea3c JK |
63 | } else if (type == INO_ENTRIES) { |
64 | int i; | |
65 | ||
e5e7ea3c | 66 | for (i = 0; i <= UPDATE_INO; i++) |
8f73cbb7 KM |
67 | mem_size += sbi->im[i].ino_num * |
68 | sizeof(struct ino_entry); | |
69 | mem_size >>= PAGE_SHIFT; | |
e5e7ea3c | 70 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
429511cd | 71 | } else if (type == EXTENT_CACHE) { |
7441ccef JK |
72 | mem_size = (atomic_read(&sbi->total_ext_tree) * |
73 | sizeof(struct extent_tree) + | |
429511cd | 74 | atomic_read(&sbi->total_ext_node) * |
09cbfeaf | 75 | sizeof(struct extent_node)) >> PAGE_SHIFT; |
429511cd | 76 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
1e84371f | 77 | } else { |
1663cae4 JK |
78 | if (!sbi->sb->s_bdi->wb.dirty_exceeded) |
79 | return true; | |
6fb03f3a JK |
80 | } |
81 | return res; | |
cdfc41c1 JK |
82 | } |
83 | ||
e05df3b1 JK |
84 | static void clear_node_page_dirty(struct page *page) |
85 | { | |
86 | struct address_space *mapping = page->mapping; | |
e05df3b1 JK |
87 | unsigned int long flags; |
88 | ||
89 | if (PageDirty(page)) { | |
90 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
91 | radix_tree_tag_clear(&mapping->page_tree, | |
92 | page_index(page), | |
93 | PAGECACHE_TAG_DIRTY); | |
94 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
95 | ||
96 | clear_page_dirty_for_io(page); | |
4081363f | 97 | dec_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_NODES); |
e05df3b1 JK |
98 | } |
99 | ClearPageUptodate(page); | |
100 | } | |
101 | ||
102 | static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
103 | { | |
104 | pgoff_t index = current_nat_addr(sbi, nid); | |
105 | return get_meta_page(sbi, index); | |
106 | } | |
107 | ||
108 | static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
109 | { | |
110 | struct page *src_page; | |
111 | struct page *dst_page; | |
112 | pgoff_t src_off; | |
113 | pgoff_t dst_off; | |
114 | void *src_addr; | |
115 | void *dst_addr; | |
116 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
117 | ||
118 | src_off = current_nat_addr(sbi, nid); | |
119 | dst_off = next_nat_addr(sbi, src_off); | |
120 | ||
121 | /* get current nat block page with lock */ | |
122 | src_page = get_meta_page(sbi, src_off); | |
e05df3b1 | 123 | dst_page = grab_meta_page(sbi, dst_off); |
9850cf4a | 124 | f2fs_bug_on(sbi, PageDirty(src_page)); |
e05df3b1 JK |
125 | |
126 | src_addr = page_address(src_page); | |
127 | dst_addr = page_address(dst_page); | |
09cbfeaf | 128 | memcpy(dst_addr, src_addr, PAGE_SIZE); |
e05df3b1 JK |
129 | set_page_dirty(dst_page); |
130 | f2fs_put_page(src_page, 1); | |
131 | ||
132 | set_to_next_nat(nm_i, nid); | |
133 | ||
134 | return dst_page; | |
135 | } | |
136 | ||
e05df3b1 JK |
137 | static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n) |
138 | { | |
139 | return radix_tree_lookup(&nm_i->nat_root, n); | |
140 | } | |
141 | ||
142 | static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i, | |
143 | nid_t start, unsigned int nr, struct nat_entry **ep) | |
144 | { | |
145 | return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr); | |
146 | } | |
147 | ||
148 | static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e) | |
149 | { | |
150 | list_del(&e->list); | |
151 | radix_tree_delete(&nm_i->nat_root, nat_get_nid(e)); | |
152 | nm_i->nat_cnt--; | |
153 | kmem_cache_free(nat_entry_slab, e); | |
154 | } | |
155 | ||
309cc2b6 JK |
156 | static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i, |
157 | struct nat_entry *ne) | |
158 | { | |
159 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); | |
160 | struct nat_entry_set *head; | |
161 | ||
309cc2b6 JK |
162 | head = radix_tree_lookup(&nm_i->nat_set_root, set); |
163 | if (!head) { | |
80c54505 | 164 | head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS); |
309cc2b6 JK |
165 | |
166 | INIT_LIST_HEAD(&head->entry_list); | |
167 | INIT_LIST_HEAD(&head->set_list); | |
168 | head->set = set; | |
169 | head->entry_cnt = 0; | |
9be32d72 | 170 | f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head); |
309cc2b6 | 171 | } |
febeca6d CY |
172 | |
173 | if (get_nat_flag(ne, IS_DIRTY)) | |
174 | goto refresh_list; | |
175 | ||
309cc2b6 JK |
176 | nm_i->dirty_nat_cnt++; |
177 | head->entry_cnt++; | |
178 | set_nat_flag(ne, IS_DIRTY, true); | |
febeca6d CY |
179 | refresh_list: |
180 | if (nat_get_blkaddr(ne) == NEW_ADDR) | |
181 | list_del_init(&ne->list); | |
182 | else | |
183 | list_move_tail(&ne->list, &head->entry_list); | |
309cc2b6 JK |
184 | } |
185 | ||
186 | static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i, | |
0b28b71e | 187 | struct nat_entry_set *set, struct nat_entry *ne) |
309cc2b6 | 188 | { |
0b28b71e KM |
189 | list_move_tail(&ne->list, &nm_i->nat_entries); |
190 | set_nat_flag(ne, IS_DIRTY, false); | |
191 | set->entry_cnt--; | |
192 | nm_i->dirty_nat_cnt--; | |
309cc2b6 JK |
193 | } |
194 | ||
195 | static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i, | |
196 | nid_t start, unsigned int nr, struct nat_entry_set **ep) | |
197 | { | |
198 | return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep, | |
199 | start, nr); | |
200 | } | |
201 | ||
2dcf51ab | 202 | int need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 JK |
203 | { |
204 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
205 | struct nat_entry *e; | |
2dcf51ab | 206 | bool need = false; |
e05df3b1 | 207 | |
b873b798 | 208 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 | 209 | e = __lookup_nat_cache(nm_i, nid); |
2dcf51ab JK |
210 | if (e) { |
211 | if (!get_nat_flag(e, IS_CHECKPOINTED) && | |
212 | !get_nat_flag(e, HAS_FSYNCED_INODE)) | |
213 | need = true; | |
214 | } | |
b873b798 | 215 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 216 | return need; |
e05df3b1 JK |
217 | } |
218 | ||
2dcf51ab | 219 | bool is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid) |
479f40c4 JK |
220 | { |
221 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
222 | struct nat_entry *e; | |
2dcf51ab | 223 | bool is_cp = true; |
479f40c4 | 224 | |
b873b798 | 225 | down_read(&nm_i->nat_tree_lock); |
2dcf51ab JK |
226 | e = __lookup_nat_cache(nm_i, nid); |
227 | if (e && !get_nat_flag(e, IS_CHECKPOINTED)) | |
228 | is_cp = false; | |
b873b798 | 229 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 230 | return is_cp; |
479f40c4 JK |
231 | } |
232 | ||
88bd02c9 | 233 | bool need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino) |
b6fe5873 JK |
234 | { |
235 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
236 | struct nat_entry *e; | |
88bd02c9 | 237 | bool need_update = true; |
b6fe5873 | 238 | |
b873b798 | 239 | down_read(&nm_i->nat_tree_lock); |
88bd02c9 JK |
240 | e = __lookup_nat_cache(nm_i, ino); |
241 | if (e && get_nat_flag(e, HAS_LAST_FSYNC) && | |
242 | (get_nat_flag(e, IS_CHECKPOINTED) || | |
243 | get_nat_flag(e, HAS_FSYNCED_INODE))) | |
244 | need_update = false; | |
b873b798 | 245 | up_read(&nm_i->nat_tree_lock); |
88bd02c9 | 246 | return need_update; |
b6fe5873 JK |
247 | } |
248 | ||
5c9e4184 CY |
249 | static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid, |
250 | bool no_fail) | |
e05df3b1 JK |
251 | { |
252 | struct nat_entry *new; | |
253 | ||
5c9e4184 CY |
254 | if (no_fail) { |
255 | new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_NOFS); | |
256 | f2fs_radix_tree_insert(&nm_i->nat_root, nid, new); | |
257 | } else { | |
258 | new = kmem_cache_alloc(nat_entry_slab, GFP_NOFS); | |
259 | if (!new) | |
260 | return NULL; | |
261 | if (radix_tree_insert(&nm_i->nat_root, nid, new)) { | |
262 | kmem_cache_free(nat_entry_slab, new); | |
263 | return NULL; | |
264 | } | |
265 | } | |
266 | ||
e05df3b1 JK |
267 | memset(new, 0, sizeof(struct nat_entry)); |
268 | nat_set_nid(new, nid); | |
88bd02c9 | 269 | nat_reset_flag(new); |
e05df3b1 JK |
270 | list_add_tail(&new->list, &nm_i->nat_entries); |
271 | nm_i->nat_cnt++; | |
272 | return new; | |
273 | } | |
274 | ||
1515aef0 | 275 | static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid, |
e05df3b1 JK |
276 | struct f2fs_nat_entry *ne) |
277 | { | |
1515aef0 | 278 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 279 | struct nat_entry *e; |
9be32d72 | 280 | |
e05df3b1 JK |
281 | e = __lookup_nat_cache(nm_i, nid); |
282 | if (!e) { | |
5c9e4184 CY |
283 | e = grab_nat_entry(nm_i, nid, false); |
284 | if (e) | |
285 | node_info_from_raw_nat(&e->ni, ne); | |
1515aef0 | 286 | } else { |
0c0b471e EB |
287 | f2fs_bug_on(sbi, nat_get_ino(e) != le32_to_cpu(ne->ino) || |
288 | nat_get_blkaddr(e) != | |
289 | le32_to_cpu(ne->block_addr) || | |
1515aef0 | 290 | nat_get_version(e) != ne->version); |
e05df3b1 | 291 | } |
e05df3b1 JK |
292 | } |
293 | ||
294 | static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, | |
479f40c4 | 295 | block_t new_blkaddr, bool fsync_done) |
e05df3b1 JK |
296 | { |
297 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
298 | struct nat_entry *e; | |
9be32d72 | 299 | |
b873b798 | 300 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
301 | e = __lookup_nat_cache(nm_i, ni->nid); |
302 | if (!e) { | |
5c9e4184 | 303 | e = grab_nat_entry(nm_i, ni->nid, true); |
5c27f4ee | 304 | copy_node_info(&e->ni, ni); |
9850cf4a | 305 | f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR); |
e05df3b1 JK |
306 | } else if (new_blkaddr == NEW_ADDR) { |
307 | /* | |
308 | * when nid is reallocated, | |
309 | * previous nat entry can be remained in nat cache. | |
310 | * So, reinitialize it with new information. | |
311 | */ | |
5c27f4ee | 312 | copy_node_info(&e->ni, ni); |
9850cf4a | 313 | f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR); |
e05df3b1 JK |
314 | } |
315 | ||
e05df3b1 | 316 | /* sanity check */ |
9850cf4a JK |
317 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr); |
318 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR && | |
e05df3b1 | 319 | new_blkaddr == NULL_ADDR); |
9850cf4a | 320 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR && |
e05df3b1 | 321 | new_blkaddr == NEW_ADDR); |
9850cf4a | 322 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != NEW_ADDR && |
e05df3b1 JK |
323 | nat_get_blkaddr(e) != NULL_ADDR && |
324 | new_blkaddr == NEW_ADDR); | |
325 | ||
e1c42045 | 326 | /* increment version no as node is removed */ |
e05df3b1 JK |
327 | if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) { |
328 | unsigned char version = nat_get_version(e); | |
329 | nat_set_version(e, inc_node_version(version)); | |
26834466 JK |
330 | |
331 | /* in order to reuse the nid */ | |
332 | if (nm_i->next_scan_nid > ni->nid) | |
333 | nm_i->next_scan_nid = ni->nid; | |
e05df3b1 JK |
334 | } |
335 | ||
336 | /* change address */ | |
337 | nat_set_blkaddr(e, new_blkaddr); | |
88bd02c9 JK |
338 | if (new_blkaddr == NEW_ADDR || new_blkaddr == NULL_ADDR) |
339 | set_nat_flag(e, IS_CHECKPOINTED, false); | |
e05df3b1 | 340 | __set_nat_cache_dirty(nm_i, e); |
479f40c4 JK |
341 | |
342 | /* update fsync_mark if its inode nat entry is still alive */ | |
d5b692b7 CY |
343 | if (ni->nid != ni->ino) |
344 | e = __lookup_nat_cache(nm_i, ni->ino); | |
88bd02c9 JK |
345 | if (e) { |
346 | if (fsync_done && ni->nid == ni->ino) | |
347 | set_nat_flag(e, HAS_FSYNCED_INODE, true); | |
348 | set_nat_flag(e, HAS_LAST_FSYNC, fsync_done); | |
349 | } | |
b873b798 | 350 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
351 | } |
352 | ||
4660f9c0 | 353 | int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) |
e05df3b1 JK |
354 | { |
355 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1b38dc8e | 356 | int nr = nr_shrink; |
e05df3b1 | 357 | |
b873b798 JK |
358 | if (!down_write_trylock(&nm_i->nat_tree_lock)) |
359 | return 0; | |
e05df3b1 | 360 | |
e05df3b1 JK |
361 | while (nr_shrink && !list_empty(&nm_i->nat_entries)) { |
362 | struct nat_entry *ne; | |
363 | ne = list_first_entry(&nm_i->nat_entries, | |
364 | struct nat_entry, list); | |
365 | __del_from_nat_cache(nm_i, ne); | |
366 | nr_shrink--; | |
367 | } | |
b873b798 | 368 | up_write(&nm_i->nat_tree_lock); |
1b38dc8e | 369 | return nr - nr_shrink; |
e05df3b1 JK |
370 | } |
371 | ||
0a8165d7 | 372 | /* |
e1c42045 | 373 | * This function always returns success |
e05df3b1 JK |
374 | */ |
375 | void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni) | |
376 | { | |
377 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
378 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 379 | struct f2fs_journal *journal = curseg->journal; |
e05df3b1 JK |
380 | nid_t start_nid = START_NID(nid); |
381 | struct f2fs_nat_block *nat_blk; | |
382 | struct page *page = NULL; | |
383 | struct f2fs_nat_entry ne; | |
384 | struct nat_entry *e; | |
66a82d1f | 385 | pgoff_t index; |
e05df3b1 JK |
386 | int i; |
387 | ||
388 | ni->nid = nid; | |
389 | ||
390 | /* Check nat cache */ | |
b873b798 | 391 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
392 | e = __lookup_nat_cache(nm_i, nid); |
393 | if (e) { | |
394 | ni->ino = nat_get_ino(e); | |
395 | ni->blk_addr = nat_get_blkaddr(e); | |
396 | ni->version = nat_get_version(e); | |
b873b798 | 397 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 | 398 | return; |
1515aef0 | 399 | } |
e05df3b1 | 400 | |
3547ea96 JK |
401 | memset(&ne, 0, sizeof(struct f2fs_nat_entry)); |
402 | ||
e05df3b1 | 403 | /* Check current segment summary */ |
b7ad7512 | 404 | down_read(&curseg->journal_rwsem); |
dfc08a12 | 405 | i = lookup_journal_in_cursum(journal, NAT_JOURNAL, nid, 0); |
e05df3b1 | 406 | if (i >= 0) { |
dfc08a12 | 407 | ne = nat_in_journal(journal, i); |
e05df3b1 JK |
408 | node_info_from_raw_nat(ni, &ne); |
409 | } | |
b7ad7512 | 410 | up_read(&curseg->journal_rwsem); |
66a82d1f YH |
411 | if (i >= 0) { |
412 | up_read(&nm_i->nat_tree_lock); | |
e05df3b1 | 413 | goto cache; |
66a82d1f | 414 | } |
e05df3b1 JK |
415 | |
416 | /* Fill node_info from nat page */ | |
66a82d1f YH |
417 | index = current_nat_addr(sbi, nid); |
418 | up_read(&nm_i->nat_tree_lock); | |
419 | ||
420 | page = get_meta_page(sbi, index); | |
e05df3b1 JK |
421 | nat_blk = (struct f2fs_nat_block *)page_address(page); |
422 | ne = nat_blk->entries[nid - start_nid]; | |
423 | node_info_from_raw_nat(ni, &ne); | |
424 | f2fs_put_page(page, 1); | |
425 | cache: | |
426 | /* cache nat entry */ | |
b873b798 | 427 | down_write(&nm_i->nat_tree_lock); |
1515aef0 | 428 | cache_nat_entry(sbi, nid, &ne); |
b873b798 | 429 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
430 | } |
431 | ||
79344efb JK |
432 | /* |
433 | * readahead MAX_RA_NODE number of node pages. | |
434 | */ | |
435 | static void ra_node_pages(struct page *parent, int start, int n) | |
436 | { | |
437 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); | |
438 | struct blk_plug plug; | |
439 | int i, end; | |
440 | nid_t nid; | |
441 | ||
442 | blk_start_plug(&plug); | |
443 | ||
444 | /* Then, try readahead for siblings of the desired node */ | |
445 | end = start + n; | |
446 | end = min(end, NIDS_PER_BLOCK); | |
447 | for (i = start; i < end; i++) { | |
448 | nid = get_nid(parent, i, false); | |
449 | ra_node_page(sbi, nid); | |
450 | } | |
451 | ||
452 | blk_finish_plug(&plug); | |
453 | } | |
454 | ||
3cf45747 CY |
455 | pgoff_t get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs) |
456 | { | |
457 | const long direct_index = ADDRS_PER_INODE(dn->inode); | |
458 | const long direct_blks = ADDRS_PER_BLOCK; | |
459 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
460 | unsigned int skipped_unit = ADDRS_PER_BLOCK; | |
461 | int cur_level = dn->cur_level; | |
462 | int max_level = dn->max_level; | |
463 | pgoff_t base = 0; | |
464 | ||
465 | if (!dn->max_level) | |
466 | return pgofs + 1; | |
467 | ||
468 | while (max_level-- > cur_level) | |
469 | skipped_unit *= NIDS_PER_BLOCK; | |
470 | ||
471 | switch (dn->max_level) { | |
472 | case 3: | |
473 | base += 2 * indirect_blks; | |
474 | case 2: | |
475 | base += 2 * direct_blks; | |
476 | case 1: | |
477 | base += direct_index; | |
478 | break; | |
479 | default: | |
480 | f2fs_bug_on(F2FS_I_SB(dn->inode), 1); | |
481 | } | |
482 | ||
483 | return ((pgofs - base) / skipped_unit + 1) * skipped_unit + base; | |
484 | } | |
485 | ||
0a8165d7 | 486 | /* |
e05df3b1 JK |
487 | * The maximum depth is four. |
488 | * Offset[0] will have raw inode offset. | |
489 | */ | |
81ca7350 | 490 | static int get_node_path(struct inode *inode, long block, |
de93653f | 491 | int offset[4], unsigned int noffset[4]) |
e05df3b1 | 492 | { |
81ca7350 | 493 | const long direct_index = ADDRS_PER_INODE(inode); |
e05df3b1 JK |
494 | const long direct_blks = ADDRS_PER_BLOCK; |
495 | const long dptrs_per_blk = NIDS_PER_BLOCK; | |
496 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
497 | const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK; | |
498 | int n = 0; | |
499 | int level = 0; | |
500 | ||
501 | noffset[0] = 0; | |
502 | ||
503 | if (block < direct_index) { | |
25c0a6e5 | 504 | offset[n] = block; |
e05df3b1 JK |
505 | goto got; |
506 | } | |
507 | block -= direct_index; | |
508 | if (block < direct_blks) { | |
509 | offset[n++] = NODE_DIR1_BLOCK; | |
510 | noffset[n] = 1; | |
25c0a6e5 | 511 | offset[n] = block; |
e05df3b1 JK |
512 | level = 1; |
513 | goto got; | |
514 | } | |
515 | block -= direct_blks; | |
516 | if (block < direct_blks) { | |
517 | offset[n++] = NODE_DIR2_BLOCK; | |
518 | noffset[n] = 2; | |
25c0a6e5 | 519 | offset[n] = block; |
e05df3b1 JK |
520 | level = 1; |
521 | goto got; | |
522 | } | |
523 | block -= direct_blks; | |
524 | if (block < indirect_blks) { | |
525 | offset[n++] = NODE_IND1_BLOCK; | |
526 | noffset[n] = 3; | |
527 | offset[n++] = block / direct_blks; | |
528 | noffset[n] = 4 + offset[n - 1]; | |
25c0a6e5 | 529 | offset[n] = block % direct_blks; |
e05df3b1 JK |
530 | level = 2; |
531 | goto got; | |
532 | } | |
533 | block -= indirect_blks; | |
534 | if (block < indirect_blks) { | |
535 | offset[n++] = NODE_IND2_BLOCK; | |
536 | noffset[n] = 4 + dptrs_per_blk; | |
537 | offset[n++] = block / direct_blks; | |
538 | noffset[n] = 5 + dptrs_per_blk + offset[n - 1]; | |
25c0a6e5 | 539 | offset[n] = block % direct_blks; |
e05df3b1 JK |
540 | level = 2; |
541 | goto got; | |
542 | } | |
543 | block -= indirect_blks; | |
544 | if (block < dindirect_blks) { | |
545 | offset[n++] = NODE_DIND_BLOCK; | |
546 | noffset[n] = 5 + (dptrs_per_blk * 2); | |
547 | offset[n++] = block / indirect_blks; | |
548 | noffset[n] = 6 + (dptrs_per_blk * 2) + | |
549 | offset[n - 1] * (dptrs_per_blk + 1); | |
550 | offset[n++] = (block / direct_blks) % dptrs_per_blk; | |
551 | noffset[n] = 7 + (dptrs_per_blk * 2) + | |
552 | offset[n - 2] * (dptrs_per_blk + 1) + | |
553 | offset[n - 1]; | |
25c0a6e5 | 554 | offset[n] = block % direct_blks; |
e05df3b1 JK |
555 | level = 3; |
556 | goto got; | |
557 | } else { | |
adb6dc19 | 558 | return -E2BIG; |
e05df3b1 JK |
559 | } |
560 | got: | |
561 | return level; | |
562 | } | |
563 | ||
564 | /* | |
565 | * Caller should call f2fs_put_dnode(dn). | |
4f4124d0 CY |
566 | * Also, it should grab and release a rwsem by calling f2fs_lock_op() and |
567 | * f2fs_unlock_op() only if ro is not set RDONLY_NODE. | |
39936837 | 568 | * In the case of RDONLY_NODE, we don't need to care about mutex. |
e05df3b1 | 569 | */ |
266e97a8 | 570 | int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode) |
e05df3b1 | 571 | { |
4081363f | 572 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 | 573 | struct page *npage[4]; |
f1a3b98e | 574 | struct page *parent = NULL; |
e05df3b1 JK |
575 | int offset[4]; |
576 | unsigned int noffset[4]; | |
577 | nid_t nids[4]; | |
3cf45747 | 578 | int level, i = 0; |
e05df3b1 JK |
579 | int err = 0; |
580 | ||
81ca7350 | 581 | level = get_node_path(dn->inode, index, offset, noffset); |
adb6dc19 JK |
582 | if (level < 0) |
583 | return level; | |
e05df3b1 JK |
584 | |
585 | nids[0] = dn->inode->i_ino; | |
1646cfac | 586 | npage[0] = dn->inode_page; |
e05df3b1 | 587 | |
1646cfac JK |
588 | if (!npage[0]) { |
589 | npage[0] = get_node_page(sbi, nids[0]); | |
590 | if (IS_ERR(npage[0])) | |
591 | return PTR_ERR(npage[0]); | |
592 | } | |
f1a3b98e JK |
593 | |
594 | /* if inline_data is set, should not report any block indices */ | |
595 | if (f2fs_has_inline_data(dn->inode) && index) { | |
76629165 | 596 | err = -ENOENT; |
f1a3b98e JK |
597 | f2fs_put_page(npage[0], 1); |
598 | goto release_out; | |
599 | } | |
600 | ||
e05df3b1 | 601 | parent = npage[0]; |
52c2db3f CL |
602 | if (level != 0) |
603 | nids[1] = get_nid(parent, offset[0], true); | |
e05df3b1 JK |
604 | dn->inode_page = npage[0]; |
605 | dn->inode_page_locked = true; | |
606 | ||
607 | /* get indirect or direct nodes */ | |
608 | for (i = 1; i <= level; i++) { | |
609 | bool done = false; | |
610 | ||
266e97a8 | 611 | if (!nids[i] && mode == ALLOC_NODE) { |
e05df3b1 JK |
612 | /* alloc new node */ |
613 | if (!alloc_nid(sbi, &(nids[i]))) { | |
e05df3b1 JK |
614 | err = -ENOSPC; |
615 | goto release_pages; | |
616 | } | |
617 | ||
618 | dn->nid = nids[i]; | |
5f4ce6ab | 619 | npage[i] = new_node_page(dn, noffset[i]); |
e05df3b1 JK |
620 | if (IS_ERR(npage[i])) { |
621 | alloc_nid_failed(sbi, nids[i]); | |
e05df3b1 JK |
622 | err = PTR_ERR(npage[i]); |
623 | goto release_pages; | |
624 | } | |
625 | ||
626 | set_nid(parent, offset[i - 1], nids[i], i == 1); | |
627 | alloc_nid_done(sbi, nids[i]); | |
e05df3b1 | 628 | done = true; |
266e97a8 | 629 | } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) { |
e05df3b1 JK |
630 | npage[i] = get_node_page_ra(parent, offset[i - 1]); |
631 | if (IS_ERR(npage[i])) { | |
632 | err = PTR_ERR(npage[i]); | |
633 | goto release_pages; | |
634 | } | |
635 | done = true; | |
636 | } | |
637 | if (i == 1) { | |
638 | dn->inode_page_locked = false; | |
639 | unlock_page(parent); | |
640 | } else { | |
641 | f2fs_put_page(parent, 1); | |
642 | } | |
643 | ||
644 | if (!done) { | |
645 | npage[i] = get_node_page(sbi, nids[i]); | |
646 | if (IS_ERR(npage[i])) { | |
647 | err = PTR_ERR(npage[i]); | |
648 | f2fs_put_page(npage[0], 0); | |
649 | goto release_out; | |
650 | } | |
651 | } | |
652 | if (i < level) { | |
653 | parent = npage[i]; | |
654 | nids[i + 1] = get_nid(parent, offset[i], false); | |
655 | } | |
656 | } | |
657 | dn->nid = nids[level]; | |
658 | dn->ofs_in_node = offset[level]; | |
659 | dn->node_page = npage[level]; | |
7a2af766 CY |
660 | dn->data_blkaddr = datablock_addr(dn->inode, |
661 | dn->node_page, dn->ofs_in_node); | |
e05df3b1 JK |
662 | return 0; |
663 | ||
664 | release_pages: | |
665 | f2fs_put_page(parent, 1); | |
666 | if (i > 1) | |
667 | f2fs_put_page(npage[0], 0); | |
668 | release_out: | |
669 | dn->inode_page = NULL; | |
670 | dn->node_page = NULL; | |
3cf45747 CY |
671 | if (err == -ENOENT) { |
672 | dn->cur_level = i; | |
673 | dn->max_level = level; | |
0a2aa8fb | 674 | dn->ofs_in_node = offset[level]; |
3cf45747 | 675 | } |
e05df3b1 JK |
676 | return err; |
677 | } | |
678 | ||
679 | static void truncate_node(struct dnode_of_data *dn) | |
680 | { | |
4081363f | 681 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
682 | struct node_info ni; |
683 | ||
684 | get_node_info(sbi, dn->nid, &ni); | |
9850cf4a | 685 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
e05df3b1 | 686 | |
e05df3b1 | 687 | /* Deallocate node address */ |
71e9fec5 | 688 | invalidate_blocks(sbi, ni.blk_addr); |
000519f2 | 689 | dec_valid_node_count(sbi, dn->inode, dn->nid == dn->inode->i_ino); |
479f40c4 | 690 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
e05df3b1 JK |
691 | |
692 | if (dn->nid == dn->inode->i_ino) { | |
693 | remove_orphan_inode(sbi, dn->nid); | |
694 | dec_valid_inode_count(sbi); | |
0f18b462 | 695 | f2fs_inode_synced(dn->inode); |
e05df3b1 | 696 | } |
000519f2 | 697 | |
e05df3b1 | 698 | clear_node_page_dirty(dn->node_page); |
caf0047e | 699 | set_sbi_flag(sbi, SBI_IS_DIRTY); |
e05df3b1 JK |
700 | |
701 | f2fs_put_page(dn->node_page, 1); | |
bf39c00a JK |
702 | |
703 | invalidate_mapping_pages(NODE_MAPPING(sbi), | |
704 | dn->node_page->index, dn->node_page->index); | |
705 | ||
e05df3b1 | 706 | dn->node_page = NULL; |
51dd6249 | 707 | trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr); |
e05df3b1 JK |
708 | } |
709 | ||
710 | static int truncate_dnode(struct dnode_of_data *dn) | |
711 | { | |
e05df3b1 JK |
712 | struct page *page; |
713 | ||
714 | if (dn->nid == 0) | |
715 | return 1; | |
716 | ||
717 | /* get direct node */ | |
4081363f | 718 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
e05df3b1 JK |
719 | if (IS_ERR(page) && PTR_ERR(page) == -ENOENT) |
720 | return 1; | |
721 | else if (IS_ERR(page)) | |
722 | return PTR_ERR(page); | |
723 | ||
724 | /* Make dnode_of_data for parameter */ | |
725 | dn->node_page = page; | |
726 | dn->ofs_in_node = 0; | |
727 | truncate_data_blocks(dn); | |
728 | truncate_node(dn); | |
729 | return 1; | |
730 | } | |
731 | ||
732 | static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs, | |
733 | int ofs, int depth) | |
734 | { | |
e05df3b1 JK |
735 | struct dnode_of_data rdn = *dn; |
736 | struct page *page; | |
737 | struct f2fs_node *rn; | |
738 | nid_t child_nid; | |
739 | unsigned int child_nofs; | |
740 | int freed = 0; | |
741 | int i, ret; | |
742 | ||
743 | if (dn->nid == 0) | |
744 | return NIDS_PER_BLOCK + 1; | |
745 | ||
51dd6249 NJ |
746 | trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr); |
747 | ||
4081363f | 748 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
51dd6249 NJ |
749 | if (IS_ERR(page)) { |
750 | trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page)); | |
e05df3b1 | 751 | return PTR_ERR(page); |
51dd6249 | 752 | } |
e05df3b1 | 753 | |
79344efb JK |
754 | ra_node_pages(page, ofs, NIDS_PER_BLOCK); |
755 | ||
45590710 | 756 | rn = F2FS_NODE(page); |
e05df3b1 JK |
757 | if (depth < 3) { |
758 | for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) { | |
759 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
760 | if (child_nid == 0) | |
761 | continue; | |
762 | rdn.nid = child_nid; | |
763 | ret = truncate_dnode(&rdn); | |
764 | if (ret < 0) | |
765 | goto out_err; | |
12719ae1 JK |
766 | if (set_nid(page, i, 0, false)) |
767 | dn->node_changed = true; | |
e05df3b1 JK |
768 | } |
769 | } else { | |
770 | child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1; | |
771 | for (i = ofs; i < NIDS_PER_BLOCK; i++) { | |
772 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
773 | if (child_nid == 0) { | |
774 | child_nofs += NIDS_PER_BLOCK + 1; | |
775 | continue; | |
776 | } | |
777 | rdn.nid = child_nid; | |
778 | ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1); | |
779 | if (ret == (NIDS_PER_BLOCK + 1)) { | |
12719ae1 JK |
780 | if (set_nid(page, i, 0, false)) |
781 | dn->node_changed = true; | |
e05df3b1 JK |
782 | child_nofs += ret; |
783 | } else if (ret < 0 && ret != -ENOENT) { | |
784 | goto out_err; | |
785 | } | |
786 | } | |
787 | freed = child_nofs; | |
788 | } | |
789 | ||
790 | if (!ofs) { | |
791 | /* remove current indirect node */ | |
792 | dn->node_page = page; | |
793 | truncate_node(dn); | |
794 | freed++; | |
795 | } else { | |
796 | f2fs_put_page(page, 1); | |
797 | } | |
51dd6249 | 798 | trace_f2fs_truncate_nodes_exit(dn->inode, freed); |
e05df3b1 JK |
799 | return freed; |
800 | ||
801 | out_err: | |
802 | f2fs_put_page(page, 1); | |
51dd6249 | 803 | trace_f2fs_truncate_nodes_exit(dn->inode, ret); |
e05df3b1 JK |
804 | return ret; |
805 | } | |
806 | ||
807 | static int truncate_partial_nodes(struct dnode_of_data *dn, | |
808 | struct f2fs_inode *ri, int *offset, int depth) | |
809 | { | |
e05df3b1 JK |
810 | struct page *pages[2]; |
811 | nid_t nid[3]; | |
812 | nid_t child_nid; | |
813 | int err = 0; | |
814 | int i; | |
815 | int idx = depth - 2; | |
816 | ||
817 | nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); | |
818 | if (!nid[0]) | |
819 | return 0; | |
820 | ||
821 | /* get indirect nodes in the path */ | |
a225dca3 | 822 | for (i = 0; i < idx + 1; i++) { |
e1c42045 | 823 | /* reference count'll be increased */ |
4081363f | 824 | pages[i] = get_node_page(F2FS_I_SB(dn->inode), nid[i]); |
e05df3b1 | 825 | if (IS_ERR(pages[i])) { |
e05df3b1 | 826 | err = PTR_ERR(pages[i]); |
a225dca3 | 827 | idx = i - 1; |
e05df3b1 JK |
828 | goto fail; |
829 | } | |
830 | nid[i + 1] = get_nid(pages[i], offset[i + 1], false); | |
831 | } | |
832 | ||
79344efb JK |
833 | ra_node_pages(pages[idx], offset[idx + 1], NIDS_PER_BLOCK); |
834 | ||
e05df3b1 | 835 | /* free direct nodes linked to a partial indirect node */ |
a225dca3 | 836 | for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) { |
e05df3b1 JK |
837 | child_nid = get_nid(pages[idx], i, false); |
838 | if (!child_nid) | |
839 | continue; | |
840 | dn->nid = child_nid; | |
841 | err = truncate_dnode(dn); | |
842 | if (err < 0) | |
843 | goto fail; | |
12719ae1 JK |
844 | if (set_nid(pages[idx], i, 0, false)) |
845 | dn->node_changed = true; | |
e05df3b1 JK |
846 | } |
847 | ||
a225dca3 | 848 | if (offset[idx + 1] == 0) { |
e05df3b1 JK |
849 | dn->node_page = pages[idx]; |
850 | dn->nid = nid[idx]; | |
851 | truncate_node(dn); | |
852 | } else { | |
853 | f2fs_put_page(pages[idx], 1); | |
854 | } | |
855 | offset[idx]++; | |
a225dca3 | 856 | offset[idx + 1] = 0; |
857 | idx--; | |
e05df3b1 | 858 | fail: |
a225dca3 | 859 | for (i = idx; i >= 0; i--) |
e05df3b1 | 860 | f2fs_put_page(pages[i], 1); |
51dd6249 NJ |
861 | |
862 | trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err); | |
863 | ||
e05df3b1 JK |
864 | return err; |
865 | } | |
866 | ||
0a8165d7 | 867 | /* |
e05df3b1 JK |
868 | * All the block addresses of data and nodes should be nullified. |
869 | */ | |
870 | int truncate_inode_blocks(struct inode *inode, pgoff_t from) | |
871 | { | |
4081363f | 872 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
e05df3b1 JK |
873 | int err = 0, cont = 1; |
874 | int level, offset[4], noffset[4]; | |
7dd690c8 | 875 | unsigned int nofs = 0; |
58bfaf44 | 876 | struct f2fs_inode *ri; |
e05df3b1 JK |
877 | struct dnode_of_data dn; |
878 | struct page *page; | |
879 | ||
51dd6249 NJ |
880 | trace_f2fs_truncate_inode_blocks_enter(inode, from); |
881 | ||
81ca7350 | 882 | level = get_node_path(inode, from, offset, noffset); |
adb6dc19 JK |
883 | if (level < 0) |
884 | return level; | |
ff373558 | 885 | |
e05df3b1 | 886 | page = get_node_page(sbi, inode->i_ino); |
51dd6249 NJ |
887 | if (IS_ERR(page)) { |
888 | trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page)); | |
e05df3b1 | 889 | return PTR_ERR(page); |
51dd6249 | 890 | } |
e05df3b1 JK |
891 | |
892 | set_new_dnode(&dn, inode, page, NULL, 0); | |
893 | unlock_page(page); | |
894 | ||
58bfaf44 | 895 | ri = F2FS_INODE(page); |
e05df3b1 JK |
896 | switch (level) { |
897 | case 0: | |
898 | case 1: | |
899 | nofs = noffset[1]; | |
900 | break; | |
901 | case 2: | |
902 | nofs = noffset[1]; | |
903 | if (!offset[level - 1]) | |
904 | goto skip_partial; | |
58bfaf44 | 905 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
906 | if (err < 0 && err != -ENOENT) |
907 | goto fail; | |
908 | nofs += 1 + NIDS_PER_BLOCK; | |
909 | break; | |
910 | case 3: | |
911 | nofs = 5 + 2 * NIDS_PER_BLOCK; | |
912 | if (!offset[level - 1]) | |
913 | goto skip_partial; | |
58bfaf44 | 914 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
915 | if (err < 0 && err != -ENOENT) |
916 | goto fail; | |
917 | break; | |
918 | default: | |
919 | BUG(); | |
920 | } | |
921 | ||
922 | skip_partial: | |
923 | while (cont) { | |
58bfaf44 | 924 | dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); |
e05df3b1 JK |
925 | switch (offset[0]) { |
926 | case NODE_DIR1_BLOCK: | |
927 | case NODE_DIR2_BLOCK: | |
928 | err = truncate_dnode(&dn); | |
929 | break; | |
930 | ||
931 | case NODE_IND1_BLOCK: | |
932 | case NODE_IND2_BLOCK: | |
933 | err = truncate_nodes(&dn, nofs, offset[1], 2); | |
934 | break; | |
935 | ||
936 | case NODE_DIND_BLOCK: | |
937 | err = truncate_nodes(&dn, nofs, offset[1], 3); | |
938 | cont = 0; | |
939 | break; | |
940 | ||
941 | default: | |
942 | BUG(); | |
943 | } | |
944 | if (err < 0 && err != -ENOENT) | |
945 | goto fail; | |
946 | if (offset[1] == 0 && | |
58bfaf44 | 947 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) { |
e05df3b1 | 948 | lock_page(page); |
ff373558 | 949 | BUG_ON(page->mapping != NODE_MAPPING(sbi)); |
fec1d657 | 950 | f2fs_wait_on_page_writeback(page, NODE, true); |
58bfaf44 | 951 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0; |
e05df3b1 JK |
952 | set_page_dirty(page); |
953 | unlock_page(page); | |
954 | } | |
955 | offset[1] = 0; | |
956 | offset[0]++; | |
957 | nofs += err; | |
958 | } | |
959 | fail: | |
960 | f2fs_put_page(page, 0); | |
51dd6249 | 961 | trace_f2fs_truncate_inode_blocks_exit(inode, err); |
e05df3b1 JK |
962 | return err > 0 ? 0 : err; |
963 | } | |
964 | ||
4f16fb0f JK |
965 | int truncate_xattr_node(struct inode *inode, struct page *page) |
966 | { | |
4081363f | 967 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
4f16fb0f JK |
968 | nid_t nid = F2FS_I(inode)->i_xattr_nid; |
969 | struct dnode_of_data dn; | |
970 | struct page *npage; | |
971 | ||
972 | if (!nid) | |
973 | return 0; | |
974 | ||
975 | npage = get_node_page(sbi, nid); | |
976 | if (IS_ERR(npage)) | |
977 | return PTR_ERR(npage); | |
978 | ||
205b9822 | 979 | f2fs_i_xnid_write(inode, 0); |
65985d93 | 980 | |
4f16fb0f JK |
981 | set_new_dnode(&dn, inode, page, npage, nid); |
982 | ||
983 | if (page) | |
01d2d1aa | 984 | dn.inode_page_locked = true; |
4f16fb0f JK |
985 | truncate_node(&dn); |
986 | return 0; | |
987 | } | |
988 | ||
39936837 | 989 | /* |
4f4124d0 CY |
990 | * Caller should grab and release a rwsem by calling f2fs_lock_op() and |
991 | * f2fs_unlock_op(). | |
39936837 | 992 | */ |
13ec7297 | 993 | int remove_inode_page(struct inode *inode) |
e05df3b1 | 994 | { |
e05df3b1 | 995 | struct dnode_of_data dn; |
13ec7297 | 996 | int err; |
e05df3b1 | 997 | |
c2e69583 | 998 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); |
13ec7297 CY |
999 | err = get_dnode_of_data(&dn, 0, LOOKUP_NODE); |
1000 | if (err) | |
1001 | return err; | |
e05df3b1 | 1002 | |
13ec7297 CY |
1003 | err = truncate_xattr_node(inode, dn.inode_page); |
1004 | if (err) { | |
c2e69583 | 1005 | f2fs_put_dnode(&dn); |
13ec7297 | 1006 | return err; |
e05df3b1 | 1007 | } |
c2e69583 JK |
1008 | |
1009 | /* remove potential inline_data blocks */ | |
1010 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
1011 | S_ISLNK(inode->i_mode)) | |
1012 | truncate_data_blocks_range(&dn, 1); | |
1013 | ||
e1c42045 | 1014 | /* 0 is possible, after f2fs_new_inode() has failed */ |
9850cf4a | 1015 | f2fs_bug_on(F2FS_I_SB(inode), |
0eb0adad | 1016 | inode->i_blocks != 0 && inode->i_blocks != 8); |
c2e69583 JK |
1017 | |
1018 | /* will put inode & node pages */ | |
71e9fec5 | 1019 | truncate_node(&dn); |
13ec7297 | 1020 | return 0; |
e05df3b1 JK |
1021 | } |
1022 | ||
a014e037 | 1023 | struct page *new_inode_page(struct inode *inode) |
e05df3b1 | 1024 | { |
e05df3b1 JK |
1025 | struct dnode_of_data dn; |
1026 | ||
1027 | /* allocate inode page for new inode */ | |
1028 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); | |
44a83ff6 JK |
1029 | |
1030 | /* caller should f2fs_put_page(page, 1); */ | |
5f4ce6ab | 1031 | return new_node_page(&dn, 0); |
e05df3b1 JK |
1032 | } |
1033 | ||
5f4ce6ab | 1034 | struct page *new_node_page(struct dnode_of_data *dn, unsigned int ofs) |
e05df3b1 | 1035 | { |
4081363f | 1036 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
25cc5d3b | 1037 | struct node_info new_ni; |
e05df3b1 JK |
1038 | struct page *page; |
1039 | int err; | |
1040 | ||
91942321 | 1041 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC))) |
e05df3b1 JK |
1042 | return ERR_PTR(-EPERM); |
1043 | ||
300e129c | 1044 | page = f2fs_grab_cache_page(NODE_MAPPING(sbi), dn->nid, false); |
e05df3b1 JK |
1045 | if (!page) |
1046 | return ERR_PTR(-ENOMEM); | |
1047 | ||
0abd675e | 1048 | if (unlikely((err = inc_valid_node_count(sbi, dn->inode, !ofs)))) |
9c02740c | 1049 | goto fail; |
0abd675e | 1050 | |
25cc5d3b JK |
1051 | #ifdef CONFIG_F2FS_CHECK_FS |
1052 | get_node_info(sbi, dn->nid, &new_ni); | |
1053 | f2fs_bug_on(sbi, new_ni.blk_addr != NULL_ADDR); | |
1054 | #endif | |
1055 | new_ni.nid = dn->nid; | |
e05df3b1 | 1056 | new_ni.ino = dn->inode->i_ino; |
25cc5d3b JK |
1057 | new_ni.blk_addr = NULL_ADDR; |
1058 | new_ni.flag = 0; | |
1059 | new_ni.version = 0; | |
479f40c4 | 1060 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
9c02740c | 1061 | |
fec1d657 | 1062 | f2fs_wait_on_page_writeback(page, NODE, true); |
9c02740c | 1063 | fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true); |
398b1ac5 | 1064 | set_cold_node(dn->inode, page); |
237c0790 JK |
1065 | if (!PageUptodate(page)) |
1066 | SetPageUptodate(page); | |
12719ae1 JK |
1067 | if (set_page_dirty(page)) |
1068 | dn->node_changed = true; | |
e05df3b1 | 1069 | |
4bc8e9bc | 1070 | if (f2fs_has_xattr_block(ofs)) |
205b9822 | 1071 | f2fs_i_xnid_write(dn->inode, dn->nid); |
479bd73a | 1072 | |
e05df3b1 JK |
1073 | if (ofs == 0) |
1074 | inc_valid_inode_count(sbi); | |
e05df3b1 JK |
1075 | return page; |
1076 | ||
1077 | fail: | |
71e9fec5 | 1078 | clear_node_page_dirty(page); |
e05df3b1 JK |
1079 | f2fs_put_page(page, 1); |
1080 | return ERR_PTR(err); | |
1081 | } | |
1082 | ||
56ae674c JK |
1083 | /* |
1084 | * Caller should do after getting the following values. | |
1085 | * 0: f2fs_put_page(page, 0) | |
86531d6b | 1086 | * LOCKED_PAGE or error: f2fs_put_page(page, 1) |
56ae674c | 1087 | */ |
04d328de | 1088 | static int read_node_page(struct page *page, int op_flags) |
e05df3b1 | 1089 | { |
4081363f | 1090 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1091 | struct node_info ni; |
cf04e8eb | 1092 | struct f2fs_io_info fio = { |
05ca3632 | 1093 | .sbi = sbi, |
cf04e8eb | 1094 | .type = NODE, |
04d328de MC |
1095 | .op = REQ_OP_READ, |
1096 | .op_flags = op_flags, | |
05ca3632 | 1097 | .page = page, |
4375a336 | 1098 | .encrypted_page = NULL, |
cf04e8eb | 1099 | }; |
e05df3b1 | 1100 | |
3bdad3c7 JK |
1101 | if (PageUptodate(page)) |
1102 | return LOCKED_PAGE; | |
1103 | ||
e05df3b1 JK |
1104 | get_node_info(sbi, page->index, &ni); |
1105 | ||
6bacf52f | 1106 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
2bca1e23 | 1107 | ClearPageUptodate(page); |
e05df3b1 | 1108 | return -ENOENT; |
393ff91f JK |
1109 | } |
1110 | ||
7a9d7548 | 1111 | fio.new_blkaddr = fio.old_blkaddr = ni.blk_addr; |
05ca3632 | 1112 | return f2fs_submit_page_bio(&fio); |
e05df3b1 JK |
1113 | } |
1114 | ||
0a8165d7 | 1115 | /* |
e05df3b1 JK |
1116 | * Readahead a node page |
1117 | */ | |
1118 | void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid) | |
1119 | { | |
e05df3b1 | 1120 | struct page *apage; |
56ae674c | 1121 | int err; |
e05df3b1 | 1122 | |
e8458725 CY |
1123 | if (!nid) |
1124 | return; | |
1125 | f2fs_bug_on(sbi, check_nid_range(sbi, nid)); | |
1126 | ||
999270de FL |
1127 | rcu_read_lock(); |
1128 | apage = radix_tree_lookup(&NODE_MAPPING(sbi)->page_tree, nid); | |
1129 | rcu_read_unlock(); | |
1130 | if (apage) | |
393ff91f | 1131 | return; |
e05df3b1 | 1132 | |
300e129c | 1133 | apage = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false); |
e05df3b1 JK |
1134 | if (!apage) |
1135 | return; | |
1136 | ||
70246286 | 1137 | err = read_node_page(apage, REQ_RAHEAD); |
86531d6b | 1138 | f2fs_put_page(apage, err ? 1 : 0); |
e05df3b1 JK |
1139 | } |
1140 | ||
17a0ee55 | 1141 | static struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid, |
0e022ea8 | 1142 | struct page *parent, int start) |
e05df3b1 | 1143 | { |
e05df3b1 | 1144 | struct page *page; |
0e022ea8 | 1145 | int err; |
e05df3b1 | 1146 | |
e05df3b1 JK |
1147 | if (!nid) |
1148 | return ERR_PTR(-ENOENT); | |
e8458725 | 1149 | f2fs_bug_on(sbi, check_nid_range(sbi, nid)); |
afcb7ca0 | 1150 | repeat: |
300e129c | 1151 | page = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false); |
e05df3b1 JK |
1152 | if (!page) |
1153 | return ERR_PTR(-ENOMEM); | |
1154 | ||
70fd7614 | 1155 | err = read_node_page(page, 0); |
86531d6b JK |
1156 | if (err < 0) { |
1157 | f2fs_put_page(page, 1); | |
e05df3b1 | 1158 | return ERR_PTR(err); |
86531d6b | 1159 | } else if (err == LOCKED_PAGE) { |
1f258ec1 | 1160 | err = 0; |
56ae674c | 1161 | goto page_hit; |
86531d6b | 1162 | } |
e05df3b1 | 1163 | |
0e022ea8 | 1164 | if (parent) |
79344efb | 1165 | ra_node_pages(parent, start + 1, MAX_RA_NODE); |
c718379b | 1166 | |
e05df3b1 | 1167 | lock_page(page); |
0e022ea8 | 1168 | |
4ef51a8f | 1169 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1170 | f2fs_put_page(page, 1); |
1171 | goto repeat; | |
1172 | } | |
1563ac75 | 1173 | |
1f258ec1 CY |
1174 | if (unlikely(!PageUptodate(page))) { |
1175 | err = -EIO; | |
1563ac75 | 1176 | goto out_err; |
1f258ec1 | 1177 | } |
704956ec CY |
1178 | |
1179 | if (!f2fs_inode_chksum_verify(sbi, page)) { | |
1180 | err = -EBADMSG; | |
1181 | goto out_err; | |
1182 | } | |
e0f56cb4 | 1183 | page_hit: |
0c9df7fb | 1184 | if(unlikely(nid != nid_of_node(page))) { |
1f258ec1 CY |
1185 | f2fs_msg(sbi->sb, KERN_WARNING, "inconsistent node block, " |
1186 | "nid:%lu, node_footer[nid:%u,ino:%u,ofs:%u,cpver:%llu,blkaddr:%u]", | |
1187 | nid, nid_of_node(page), ino_of_node(page), | |
1188 | ofs_of_node(page), cpver_of_node(page), | |
1189 | next_blkaddr_of_node(page)); | |
1f258ec1 | 1190 | err = -EINVAL; |
0c9df7fb | 1191 | out_err: |
ee605234 | 1192 | ClearPageUptodate(page); |
0c9df7fb | 1193 | f2fs_put_page(page, 1); |
1f258ec1 | 1194 | return ERR_PTR(err); |
0c9df7fb | 1195 | } |
e05df3b1 JK |
1196 | return page; |
1197 | } | |
1198 | ||
0e022ea8 CY |
1199 | struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) |
1200 | { | |
1201 | return __get_node_page(sbi, nid, NULL, 0); | |
1202 | } | |
1203 | ||
1204 | struct page *get_node_page_ra(struct page *parent, int start) | |
1205 | { | |
1206 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); | |
1207 | nid_t nid = get_nid(parent, start, false); | |
1208 | ||
1209 | return __get_node_page(sbi, nid, parent, start); | |
1210 | } | |
1211 | ||
2049d4fc JK |
1212 | static void flush_inline_data(struct f2fs_sb_info *sbi, nid_t ino) |
1213 | { | |
1214 | struct inode *inode; | |
1215 | struct page *page; | |
0f3311a8 | 1216 | int ret; |
2049d4fc JK |
1217 | |
1218 | /* should flush inline_data before evict_inode */ | |
1219 | inode = ilookup(sbi->sb, ino); | |
1220 | if (!inode) | |
1221 | return; | |
1222 | ||
4a6de50d | 1223 | page = pagecache_get_page(inode->i_mapping, 0, FGP_LOCK|FGP_NOWAIT, 0); |
2049d4fc JK |
1224 | if (!page) |
1225 | goto iput_out; | |
1226 | ||
1227 | if (!PageUptodate(page)) | |
1228 | goto page_out; | |
1229 | ||
1230 | if (!PageDirty(page)) | |
1231 | goto page_out; | |
1232 | ||
1233 | if (!clear_page_dirty_for_io(page)) | |
1234 | goto page_out; | |
1235 | ||
0f3311a8 CY |
1236 | ret = f2fs_write_inline_data(inode, page); |
1237 | inode_dec_dirty_pages(inode); | |
933439c8 | 1238 | remove_dirty_inode(inode); |
0f3311a8 | 1239 | if (ret) |
2049d4fc JK |
1240 | set_page_dirty(page); |
1241 | page_out: | |
4a6de50d | 1242 | f2fs_put_page(page, 1); |
2049d4fc JK |
1243 | iput_out: |
1244 | iput(inode); | |
1245 | } | |
1246 | ||
da011cc0 CY |
1247 | void move_node_page(struct page *node_page, int gc_type) |
1248 | { | |
1249 | if (gc_type == FG_GC) { | |
1250 | struct f2fs_sb_info *sbi = F2FS_P_SB(node_page); | |
1251 | struct writeback_control wbc = { | |
1252 | .sync_mode = WB_SYNC_ALL, | |
1253 | .nr_to_write = 1, | |
1254 | .for_reclaim = 0, | |
1255 | }; | |
1256 | ||
1257 | set_page_dirty(node_page); | |
1258 | f2fs_wait_on_page_writeback(node_page, NODE, true); | |
1259 | ||
1260 | f2fs_bug_on(sbi, PageWriteback(node_page)); | |
1261 | if (!clear_page_dirty_for_io(node_page)) | |
1262 | goto out_page; | |
1263 | ||
1264 | if (NODE_MAPPING(sbi)->a_ops->writepage(node_page, &wbc)) | |
1265 | unlock_page(node_page); | |
1266 | goto release_page; | |
1267 | } else { | |
1268 | /* set page dirty and write it */ | |
1269 | if (!PageWriteback(node_page)) | |
1270 | set_page_dirty(node_page); | |
1271 | } | |
1272 | out_page: | |
1273 | unlock_page(node_page); | |
1274 | release_page: | |
1275 | f2fs_put_page(node_page, 0); | |
1276 | } | |
1277 | ||
608514de JK |
1278 | static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino) |
1279 | { | |
1280 | pgoff_t index, end; | |
1281 | struct pagevec pvec; | |
1282 | struct page *last_page = NULL; | |
1283 | ||
1284 | pagevec_init(&pvec, 0); | |
1285 | index = 0; | |
1286 | end = ULONG_MAX; | |
1287 | ||
1288 | while (index <= end) { | |
1289 | int i, nr_pages; | |
1290 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1291 | PAGECACHE_TAG_DIRTY, | |
1292 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1293 | if (nr_pages == 0) | |
1294 | break; | |
1295 | ||
1296 | for (i = 0; i < nr_pages; i++) { | |
1297 | struct page *page = pvec.pages[i]; | |
1298 | ||
1299 | if (unlikely(f2fs_cp_error(sbi))) { | |
1300 | f2fs_put_page(last_page, 0); | |
1301 | pagevec_release(&pvec); | |
1302 | return ERR_PTR(-EIO); | |
1303 | } | |
1304 | ||
1305 | if (!IS_DNODE(page) || !is_cold_node(page)) | |
1306 | continue; | |
1307 | if (ino_of_node(page) != ino) | |
1308 | continue; | |
1309 | ||
1310 | lock_page(page); | |
1311 | ||
1312 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { | |
1313 | continue_unlock: | |
1314 | unlock_page(page); | |
1315 | continue; | |
1316 | } | |
1317 | if (ino_of_node(page) != ino) | |
1318 | goto continue_unlock; | |
1319 | ||
1320 | if (!PageDirty(page)) { | |
1321 | /* someone wrote it for us */ | |
1322 | goto continue_unlock; | |
1323 | } | |
1324 | ||
1325 | if (last_page) | |
1326 | f2fs_put_page(last_page, 0); | |
1327 | ||
1328 | get_page(page); | |
1329 | last_page = page; | |
1330 | unlock_page(page); | |
1331 | } | |
1332 | pagevec_release(&pvec); | |
1333 | cond_resched(); | |
1334 | } | |
1335 | return last_page; | |
1336 | } | |
1337 | ||
d68f735b | 1338 | static int __write_node_page(struct page *page, bool atomic, bool *submitted, |
b0af6d49 CY |
1339 | struct writeback_control *wbc, bool do_balance, |
1340 | enum iostat_type io_type) | |
faa24895 JK |
1341 | { |
1342 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); | |
1343 | nid_t nid; | |
1344 | struct node_info ni; | |
1345 | struct f2fs_io_info fio = { | |
1346 | .sbi = sbi, | |
1347 | .type = NODE, | |
1348 | .op = REQ_OP_WRITE, | |
1349 | .op_flags = wbc_to_write_flags(wbc), | |
1350 | .page = page, | |
1351 | .encrypted_page = NULL, | |
d68f735b | 1352 | .submitted = false, |
b0af6d49 | 1353 | .io_type = io_type, |
faa24895 JK |
1354 | }; |
1355 | ||
1356 | trace_f2fs_writepage(page, NODE); | |
1357 | ||
1358 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) | |
1359 | goto redirty_out; | |
1360 | if (unlikely(f2fs_cp_error(sbi))) | |
1361 | goto redirty_out; | |
1362 | ||
1363 | /* get old block addr of this node page */ | |
1364 | nid = nid_of_node(page); | |
1365 | f2fs_bug_on(sbi, page->index != nid); | |
1366 | ||
1367 | if (wbc->for_reclaim) { | |
1368 | if (!down_read_trylock(&sbi->node_write)) | |
1369 | goto redirty_out; | |
1370 | } else { | |
1371 | down_read(&sbi->node_write); | |
1372 | } | |
1373 | ||
1374 | get_node_info(sbi, nid, &ni); | |
1375 | ||
1376 | /* This page is already truncated */ | |
1377 | if (unlikely(ni.blk_addr == NULL_ADDR)) { | |
1378 | ClearPageUptodate(page); | |
1379 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1380 | up_read(&sbi->node_write); | |
1381 | unlock_page(page); | |
1382 | return 0; | |
1383 | } | |
1384 | ||
e7c75ab0 JK |
1385 | if (atomic && !test_opt(sbi, NOBARRIER)) |
1386 | fio.op_flags |= REQ_PREFLUSH | REQ_FUA; | |
1387 | ||
faa24895 JK |
1388 | set_page_writeback(page); |
1389 | fio.old_blkaddr = ni.blk_addr; | |
1390 | write_node_page(nid, &fio); | |
1391 | set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page)); | |
1392 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1393 | up_read(&sbi->node_write); | |
1394 | ||
d68f735b | 1395 | if (wbc->for_reclaim) { |
b9109b0e JK |
1396 | f2fs_submit_merged_write_cond(sbi, page->mapping->host, 0, |
1397 | page->index, NODE); | |
d68f735b JK |
1398 | submitted = NULL; |
1399 | } | |
faa24895 JK |
1400 | |
1401 | unlock_page(page); | |
1402 | ||
d68f735b | 1403 | if (unlikely(f2fs_cp_error(sbi))) { |
b9109b0e | 1404 | f2fs_submit_merged_write(sbi, NODE); |
d68f735b JK |
1405 | submitted = NULL; |
1406 | } | |
1407 | if (submitted) | |
1408 | *submitted = fio.submitted; | |
faa24895 | 1409 | |
401db79f YS |
1410 | if (do_balance) |
1411 | f2fs_balance_fs(sbi, false); | |
faa24895 JK |
1412 | return 0; |
1413 | ||
1414 | redirty_out: | |
1415 | redirty_page_for_writepage(wbc, page); | |
1416 | return AOP_WRITEPAGE_ACTIVATE; | |
1417 | } | |
1418 | ||
1419 | static int f2fs_write_node_page(struct page *page, | |
1420 | struct writeback_control *wbc) | |
1421 | { | |
b0af6d49 | 1422 | return __write_node_page(page, false, NULL, wbc, false, FS_NODE_IO); |
faa24895 JK |
1423 | } |
1424 | ||
26de9b11 | 1425 | int fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode, |
608514de | 1426 | struct writeback_control *wbc, bool atomic) |
e05df3b1 | 1427 | { |
e05df3b1 | 1428 | pgoff_t index, end; |
942fd319 | 1429 | pgoff_t last_idx = ULONG_MAX; |
e05df3b1 | 1430 | struct pagevec pvec; |
c267ec15 | 1431 | int ret = 0; |
608514de JK |
1432 | struct page *last_page = NULL; |
1433 | bool marked = false; | |
26de9b11 | 1434 | nid_t ino = inode->i_ino; |
52681375 | 1435 | |
608514de JK |
1436 | if (atomic) { |
1437 | last_page = last_fsync_dnode(sbi, ino); | |
1438 | if (IS_ERR_OR_NULL(last_page)) | |
1439 | return PTR_ERR_OR_ZERO(last_page); | |
1440 | } | |
1441 | retry: | |
52681375 JK |
1442 | pagevec_init(&pvec, 0); |
1443 | index = 0; | |
1444 | end = ULONG_MAX; | |
1445 | ||
1446 | while (index <= end) { | |
1447 | int i, nr_pages; | |
1448 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1449 | PAGECACHE_TAG_DIRTY, | |
1450 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1451 | if (nr_pages == 0) | |
1452 | break; | |
1453 | ||
1454 | for (i = 0; i < nr_pages; i++) { | |
1455 | struct page *page = pvec.pages[i]; | |
d68f735b | 1456 | bool submitted = false; |
52681375 JK |
1457 | |
1458 | if (unlikely(f2fs_cp_error(sbi))) { | |
608514de | 1459 | f2fs_put_page(last_page, 0); |
52681375 | 1460 | pagevec_release(&pvec); |
9de69279 CY |
1461 | ret = -EIO; |
1462 | goto out; | |
52681375 JK |
1463 | } |
1464 | ||
1465 | if (!IS_DNODE(page) || !is_cold_node(page)) | |
1466 | continue; | |
1467 | if (ino_of_node(page) != ino) | |
1468 | continue; | |
1469 | ||
1470 | lock_page(page); | |
1471 | ||
1472 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { | |
1473 | continue_unlock: | |
1474 | unlock_page(page); | |
1475 | continue; | |
1476 | } | |
1477 | if (ino_of_node(page) != ino) | |
1478 | goto continue_unlock; | |
1479 | ||
608514de | 1480 | if (!PageDirty(page) && page != last_page) { |
52681375 JK |
1481 | /* someone wrote it for us */ |
1482 | goto continue_unlock; | |
1483 | } | |
1484 | ||
1485 | f2fs_wait_on_page_writeback(page, NODE, true); | |
1486 | BUG_ON(PageWriteback(page)); | |
52681375 | 1487 | |
d29fd172 JK |
1488 | set_fsync_mark(page, 0); |
1489 | set_dentry_mark(page, 0); | |
1490 | ||
608514de JK |
1491 | if (!atomic || page == last_page) { |
1492 | set_fsync_mark(page, 1); | |
26de9b11 JK |
1493 | if (IS_INODE(page)) { |
1494 | if (is_inode_flag_set(inode, | |
1495 | FI_DIRTY_INODE)) | |
1496 | update_inode(inode, page); | |
608514de | 1497 | set_dentry_mark(page, |
52681375 | 1498 | need_dentry_mark(sbi, ino)); |
26de9b11 | 1499 | } |
608514de JK |
1500 | /* may be written by other thread */ |
1501 | if (!PageDirty(page)) | |
1502 | set_page_dirty(page); | |
1503 | } | |
1504 | ||
1505 | if (!clear_page_dirty_for_io(page)) | |
1506 | goto continue_unlock; | |
52681375 | 1507 | |
e7c75ab0 | 1508 | ret = __write_node_page(page, atomic && |
d68f735b | 1509 | page == last_page, |
b0af6d49 CY |
1510 | &submitted, wbc, true, |
1511 | FS_NODE_IO); | |
c267ec15 | 1512 | if (ret) { |
52681375 | 1513 | unlock_page(page); |
608514de JK |
1514 | f2fs_put_page(last_page, 0); |
1515 | break; | |
d68f735b | 1516 | } else if (submitted) { |
942fd319 | 1517 | last_idx = page->index; |
608514de | 1518 | } |
3f5f4959 | 1519 | |
608514de JK |
1520 | if (page == last_page) { |
1521 | f2fs_put_page(page, 0); | |
1522 | marked = true; | |
52681375 | 1523 | break; |
c267ec15 | 1524 | } |
52681375 JK |
1525 | } |
1526 | pagevec_release(&pvec); | |
1527 | cond_resched(); | |
1528 | ||
608514de | 1529 | if (ret || marked) |
52681375 JK |
1530 | break; |
1531 | } | |
608514de JK |
1532 | if (!ret && atomic && !marked) { |
1533 | f2fs_msg(sbi->sb, KERN_DEBUG, | |
1534 | "Retry to write fsync mark: ino=%u, idx=%lx", | |
1535 | ino, last_page->index); | |
1536 | lock_page(last_page); | |
d40a43af | 1537 | f2fs_wait_on_page_writeback(last_page, NODE, true); |
608514de JK |
1538 | set_page_dirty(last_page); |
1539 | unlock_page(last_page); | |
1540 | goto retry; | |
1541 | } | |
9de69279 | 1542 | out: |
942fd319 | 1543 | if (last_idx != ULONG_MAX) |
b9109b0e | 1544 | f2fs_submit_merged_write_cond(sbi, NULL, ino, last_idx, NODE); |
c267ec15 | 1545 | return ret ? -EIO: 0; |
52681375 JK |
1546 | } |
1547 | ||
401db79f | 1548 | int sync_node_pages(struct f2fs_sb_info *sbi, struct writeback_control *wbc, |
b0af6d49 | 1549 | bool do_balance, enum iostat_type io_type) |
52681375 JK |
1550 | { |
1551 | pgoff_t index, end; | |
1552 | struct pagevec pvec; | |
1553 | int step = 0; | |
12bb0a8f | 1554 | int nwritten = 0; |
3f5f4959 | 1555 | int ret = 0; |
e05df3b1 JK |
1556 | |
1557 | pagevec_init(&pvec, 0); | |
1558 | ||
1559 | next_step: | |
1560 | index = 0; | |
80dd9c0e | 1561 | end = ULONG_MAX; |
e05df3b1 JK |
1562 | |
1563 | while (index <= end) { | |
1564 | int i, nr_pages; | |
4ef51a8f | 1565 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, |
e05df3b1 JK |
1566 | PAGECACHE_TAG_DIRTY, |
1567 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1568 | if (nr_pages == 0) | |
1569 | break; | |
1570 | ||
1571 | for (i = 0; i < nr_pages; i++) { | |
1572 | struct page *page = pvec.pages[i]; | |
d68f735b | 1573 | bool submitted = false; |
e05df3b1 | 1574 | |
6d5a1495 CY |
1575 | if (unlikely(f2fs_cp_error(sbi))) { |
1576 | pagevec_release(&pvec); | |
3f5f4959 CY |
1577 | ret = -EIO; |
1578 | goto out; | |
6d5a1495 CY |
1579 | } |
1580 | ||
e05df3b1 JK |
1581 | /* |
1582 | * flushing sequence with step: | |
1583 | * 0. indirect nodes | |
1584 | * 1. dentry dnodes | |
1585 | * 2. file dnodes | |
1586 | */ | |
1587 | if (step == 0 && IS_DNODE(page)) | |
1588 | continue; | |
1589 | if (step == 1 && (!IS_DNODE(page) || | |
1590 | is_cold_node(page))) | |
1591 | continue; | |
1592 | if (step == 2 && (!IS_DNODE(page) || | |
1593 | !is_cold_node(page))) | |
1594 | continue; | |
9a4cbc9e | 1595 | lock_node: |
52681375 | 1596 | if (!trylock_page(page)) |
e05df3b1 JK |
1597 | continue; |
1598 | ||
4ef51a8f | 1599 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
e05df3b1 JK |
1600 | continue_unlock: |
1601 | unlock_page(page); | |
1602 | continue; | |
1603 | } | |
e05df3b1 JK |
1604 | |
1605 | if (!PageDirty(page)) { | |
1606 | /* someone wrote it for us */ | |
1607 | goto continue_unlock; | |
1608 | } | |
1609 | ||
2049d4fc | 1610 | /* flush inline_data */ |
52681375 | 1611 | if (is_inline_node(page)) { |
2049d4fc JK |
1612 | clear_inline_node(page); |
1613 | unlock_page(page); | |
1614 | flush_inline_data(sbi, ino_of_node(page)); | |
9a4cbc9e | 1615 | goto lock_node; |
2049d4fc JK |
1616 | } |
1617 | ||
fa3d2bdf JK |
1618 | f2fs_wait_on_page_writeback(page, NODE, true); |
1619 | ||
1620 | BUG_ON(PageWriteback(page)); | |
e05df3b1 JK |
1621 | if (!clear_page_dirty_for_io(page)) |
1622 | goto continue_unlock; | |
1623 | ||
52681375 JK |
1624 | set_fsync_mark(page, 0); |
1625 | set_dentry_mark(page, 0); | |
52746519 | 1626 | |
401db79f | 1627 | ret = __write_node_page(page, false, &submitted, |
b0af6d49 | 1628 | wbc, do_balance, io_type); |
d68f735b | 1629 | if (ret) |
52746519 | 1630 | unlock_page(page); |
d68f735b | 1631 | else if (submitted) |
3f5f4959 | 1632 | nwritten++; |
e05df3b1 JK |
1633 | |
1634 | if (--wbc->nr_to_write == 0) | |
1635 | break; | |
1636 | } | |
1637 | pagevec_release(&pvec); | |
1638 | cond_resched(); | |
1639 | ||
1640 | if (wbc->nr_to_write == 0) { | |
1641 | step = 2; | |
1642 | break; | |
1643 | } | |
1644 | } | |
1645 | ||
1646 | if (step < 2) { | |
1647 | step++; | |
1648 | goto next_step; | |
1649 | } | |
3f5f4959 CY |
1650 | out: |
1651 | if (nwritten) | |
b9109b0e | 1652 | f2fs_submit_merged_write(sbi, NODE); |
3f5f4959 | 1653 | return ret; |
e05df3b1 JK |
1654 | } |
1655 | ||
cfe58f9d JK |
1656 | int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino) |
1657 | { | |
80dd9c0e | 1658 | pgoff_t index = 0, end = ULONG_MAX; |
cfe58f9d | 1659 | struct pagevec pvec; |
280db3c8 | 1660 | int ret2, ret = 0; |
cfe58f9d JK |
1661 | |
1662 | pagevec_init(&pvec, 0); | |
4ef51a8f JK |
1663 | |
1664 | while (index <= end) { | |
1665 | int i, nr_pages; | |
1666 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1667 | PAGECACHE_TAG_WRITEBACK, | |
1668 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1669 | if (nr_pages == 0) | |
1670 | break; | |
cfe58f9d JK |
1671 | |
1672 | for (i = 0; i < nr_pages; i++) { | |
1673 | struct page *page = pvec.pages[i]; | |
1674 | ||
1675 | /* until radix tree lookup accepts end_index */ | |
cfb271d4 | 1676 | if (unlikely(page->index > end)) |
cfe58f9d JK |
1677 | continue; |
1678 | ||
4bf08ff6 | 1679 | if (ino && ino_of_node(page) == ino) { |
fec1d657 | 1680 | f2fs_wait_on_page_writeback(page, NODE, true); |
4bf08ff6 CY |
1681 | if (TestClearPageError(page)) |
1682 | ret = -EIO; | |
1683 | } | |
cfe58f9d JK |
1684 | } |
1685 | pagevec_release(&pvec); | |
1686 | cond_resched(); | |
1687 | } | |
1688 | ||
280db3c8 | 1689 | ret2 = filemap_check_errors(NODE_MAPPING(sbi)); |
cfe58f9d JK |
1690 | if (!ret) |
1691 | ret = ret2; | |
1692 | return ret; | |
1693 | } | |
1694 | ||
e05df3b1 JK |
1695 | static int f2fs_write_node_pages(struct address_space *mapping, |
1696 | struct writeback_control *wbc) | |
1697 | { | |
4081363f | 1698 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
9dfa1baf | 1699 | struct blk_plug plug; |
50c8cdb3 | 1700 | long diff; |
e05df3b1 | 1701 | |
0771fcc7 CY |
1702 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
1703 | goto skip_write; | |
1704 | ||
4660f9c0 JK |
1705 | /* balancing f2fs's metadata in background */ |
1706 | f2fs_balance_fs_bg(sbi); | |
e05df3b1 | 1707 | |
a7fdffbd | 1708 | /* collect a number of dirty node pages and write together */ |
87d6f890 | 1709 | if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE)) |
d3baf95d | 1710 | goto skip_write; |
a7fdffbd | 1711 | |
d31c7c3f YH |
1712 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
1713 | ||
50c8cdb3 | 1714 | diff = nr_pages_to_write(sbi, NODE, wbc); |
fb5566da | 1715 | wbc->sync_mode = WB_SYNC_NONE; |
9dfa1baf | 1716 | blk_start_plug(&plug); |
b0af6d49 | 1717 | sync_node_pages(sbi, wbc, true, FS_NODE_IO); |
9dfa1baf | 1718 | blk_finish_plug(&plug); |
50c8cdb3 | 1719 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
e05df3b1 | 1720 | return 0; |
d3baf95d JK |
1721 | |
1722 | skip_write: | |
1723 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES); | |
d31c7c3f | 1724 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
d3baf95d | 1725 | return 0; |
e05df3b1 JK |
1726 | } |
1727 | ||
1728 | static int f2fs_set_node_page_dirty(struct page *page) | |
1729 | { | |
26c6b887 JK |
1730 | trace_f2fs_set_page_dirty(page, NODE); |
1731 | ||
237c0790 JK |
1732 | if (!PageUptodate(page)) |
1733 | SetPageUptodate(page); | |
e05df3b1 | 1734 | if (!PageDirty(page)) { |
fe76b796 | 1735 | f2fs_set_page_dirty_nobuffers(page); |
4081363f | 1736 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 | 1737 | SetPagePrivate(page); |
9e4ded3f | 1738 | f2fs_trace_pid(page); |
e05df3b1 JK |
1739 | return 1; |
1740 | } | |
1741 | return 0; | |
1742 | } | |
1743 | ||
0a8165d7 | 1744 | /* |
e05df3b1 JK |
1745 | * Structure of the f2fs node operations |
1746 | */ | |
1747 | const struct address_space_operations f2fs_node_aops = { | |
1748 | .writepage = f2fs_write_node_page, | |
1749 | .writepages = f2fs_write_node_pages, | |
1750 | .set_page_dirty = f2fs_set_node_page_dirty, | |
487261f3 CY |
1751 | .invalidatepage = f2fs_invalidate_page, |
1752 | .releasepage = f2fs_release_page, | |
5b7a487c WG |
1753 | #ifdef CONFIG_MIGRATION |
1754 | .migratepage = f2fs_migrate_page, | |
1755 | #endif | |
e05df3b1 JK |
1756 | }; |
1757 | ||
8a7ed66a JK |
1758 | static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i, |
1759 | nid_t n) | |
e05df3b1 | 1760 | { |
8a7ed66a | 1761 | return radix_tree_lookup(&nm_i->free_nid_root, n); |
e05df3b1 JK |
1762 | } |
1763 | ||
eb0aa4b8 JK |
1764 | static int __insert_nid_to_list(struct f2fs_sb_info *sbi, |
1765 | struct free_nid *i, enum nid_list list, bool new) | |
e05df3b1 | 1766 | { |
b8559dc2 CY |
1767 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1768 | ||
eb0aa4b8 JK |
1769 | if (new) { |
1770 | int err = radix_tree_insert(&nm_i->free_nid_root, i->nid, i); | |
1771 | if (err) | |
1772 | return err; | |
1773 | } | |
1774 | ||
b8559dc2 CY |
1775 | f2fs_bug_on(sbi, list == FREE_NID_LIST ? i->state != NID_NEW : |
1776 | i->state != NID_ALLOC); | |
1777 | nm_i->nid_cnt[list]++; | |
1778 | list_add_tail(&i->list, &nm_i->nid_list[list]); | |
eb0aa4b8 | 1779 | return 0; |
b8559dc2 CY |
1780 | } |
1781 | ||
1782 | static void __remove_nid_from_list(struct f2fs_sb_info *sbi, | |
eb0aa4b8 | 1783 | struct free_nid *i, enum nid_list list, bool reuse) |
b8559dc2 CY |
1784 | { |
1785 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1786 | ||
1787 | f2fs_bug_on(sbi, list == FREE_NID_LIST ? i->state != NID_NEW : | |
1788 | i->state != NID_ALLOC); | |
1789 | nm_i->nid_cnt[list]--; | |
e05df3b1 | 1790 | list_del(&i->list); |
eb0aa4b8 JK |
1791 | if (!reuse) |
1792 | radix_tree_delete(&nm_i->free_nid_root, i->nid); | |
e05df3b1 JK |
1793 | } |
1794 | ||
4ac91242 CY |
1795 | /* return if the nid is recognized as free */ |
1796 | static bool add_free_nid(struct f2fs_sb_info *sbi, nid_t nid, bool build) | |
e05df3b1 | 1797 | { |
6fb03f3a | 1798 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
30a61ddf | 1799 | struct free_nid *i, *e; |
59bbd474 | 1800 | struct nat_entry *ne; |
30a61ddf CY |
1801 | int err = -EINVAL; |
1802 | bool ret = false; | |
9198aceb JK |
1803 | |
1804 | /* 0 nid should not be used */ | |
cfb271d4 | 1805 | if (unlikely(nid == 0)) |
4ac91242 | 1806 | return false; |
59bbd474 | 1807 | |
7bd59381 | 1808 | i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); |
e05df3b1 JK |
1809 | i->nid = nid; |
1810 | i->state = NID_NEW; | |
1811 | ||
30a61ddf CY |
1812 | if (radix_tree_preload(GFP_NOFS)) |
1813 | goto err; | |
769ec6e5 | 1814 | |
b8559dc2 | 1815 | spin_lock(&nm_i->nid_list_lock); |
30a61ddf CY |
1816 | |
1817 | if (build) { | |
1818 | /* | |
1819 | * Thread A Thread B | |
1820 | * - f2fs_create | |
1821 | * - f2fs_new_inode | |
1822 | * - alloc_nid | |
1823 | * - __insert_nid_to_list(ALLOC_NID_LIST) | |
1824 | * - f2fs_balance_fs_bg | |
1825 | * - build_free_nids | |
1826 | * - __build_free_nids | |
1827 | * - scan_nat_page | |
1828 | * - add_free_nid | |
1829 | * - __lookup_nat_cache | |
1830 | * - f2fs_add_link | |
1831 | * - init_inode_metadata | |
1832 | * - new_inode_page | |
1833 | * - new_node_page | |
1834 | * - set_node_addr | |
1835 | * - alloc_nid_done | |
1836 | * - __remove_nid_from_list(ALLOC_NID_LIST) | |
1837 | * - __insert_nid_to_list(FREE_NID_LIST) | |
1838 | */ | |
1839 | ne = __lookup_nat_cache(nm_i, nid); | |
1840 | if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) || | |
1841 | nat_get_blkaddr(ne) != NULL_ADDR)) | |
1842 | goto err_out; | |
1843 | ||
1844 | e = __lookup_free_nid_list(nm_i, nid); | |
1845 | if (e) { | |
1846 | if (e->state == NID_NEW) | |
1847 | ret = true; | |
1848 | goto err_out; | |
1849 | } | |
1850 | } | |
1851 | ret = true; | |
eb0aa4b8 | 1852 | err = __insert_nid_to_list(sbi, i, FREE_NID_LIST, true); |
30a61ddf | 1853 | err_out: |
eb0aa4b8 JK |
1854 | spin_unlock(&nm_i->nid_list_lock); |
1855 | radix_tree_preload_end(); | |
30a61ddf CY |
1856 | err: |
1857 | if (err) | |
e05df3b1 | 1858 | kmem_cache_free(free_nid_slab, i); |
30a61ddf | 1859 | return ret; |
e05df3b1 JK |
1860 | } |
1861 | ||
b8559dc2 | 1862 | static void remove_free_nid(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 | 1863 | { |
b8559dc2 | 1864 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1865 | struct free_nid *i; |
cf0ee0f0 CY |
1866 | bool need_free = false; |
1867 | ||
b8559dc2 | 1868 | spin_lock(&nm_i->nid_list_lock); |
8a7ed66a | 1869 | i = __lookup_free_nid_list(nm_i, nid); |
e05df3b1 | 1870 | if (i && i->state == NID_NEW) { |
eb0aa4b8 | 1871 | __remove_nid_from_list(sbi, i, FREE_NID_LIST, false); |
cf0ee0f0 | 1872 | need_free = true; |
e05df3b1 | 1873 | } |
b8559dc2 | 1874 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 CY |
1875 | |
1876 | if (need_free) | |
1877 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1878 | } |
1879 | ||
9f7e4a2c | 1880 | static void update_free_nid_bitmap(struct f2fs_sb_info *sbi, nid_t nid, |
346fe752 | 1881 | bool set, bool build) |
4ac91242 CY |
1882 | { |
1883 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1884 | unsigned int nat_ofs = NAT_BLOCK_OFFSET(nid); | |
1885 | unsigned int nid_ofs = nid - START_NID(nid); | |
1886 | ||
1887 | if (!test_bit_le(nat_ofs, nm_i->nat_block_bitmap)) | |
1888 | return; | |
1889 | ||
1890 | if (set) | |
23380b85 | 1891 | __set_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]); |
4ac91242 | 1892 | else |
23380b85 | 1893 | __clear_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]); |
586d1492 | 1894 | |
586d1492 CY |
1895 | if (set) |
1896 | nm_i->free_nid_count[nat_ofs]++; | |
1897 | else if (!build) | |
1898 | nm_i->free_nid_count[nat_ofs]--; | |
4ac91242 CY |
1899 | } |
1900 | ||
6fb03f3a | 1901 | static void scan_nat_page(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
1902 | struct page *nat_page, nid_t start_nid) |
1903 | { | |
6fb03f3a | 1904 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 JK |
1905 | struct f2fs_nat_block *nat_blk = page_address(nat_page); |
1906 | block_t blk_addr; | |
4ac91242 | 1907 | unsigned int nat_ofs = NAT_BLOCK_OFFSET(start_nid); |
e05df3b1 JK |
1908 | int i; |
1909 | ||
586d1492 CY |
1910 | if (test_bit_le(nat_ofs, nm_i->nat_block_bitmap)) |
1911 | return; | |
1912 | ||
23380b85 | 1913 | __set_bit_le(nat_ofs, nm_i->nat_block_bitmap); |
4ac91242 | 1914 | |
e05df3b1 JK |
1915 | i = start_nid % NAT_ENTRY_PER_BLOCK; |
1916 | ||
1917 | for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { | |
4ac91242 | 1918 | bool freed = false; |
23d38844 | 1919 | |
cfb271d4 | 1920 | if (unlikely(start_nid >= nm_i->max_nid)) |
04431c44 | 1921 | break; |
23d38844 HL |
1922 | |
1923 | blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); | |
9850cf4a | 1924 | f2fs_bug_on(sbi, blk_addr == NEW_ADDR); |
3a2ad567 | 1925 | if (blk_addr == NULL_ADDR) |
4ac91242 | 1926 | freed = add_free_nid(sbi, start_nid, true); |
346fe752 CY |
1927 | spin_lock(&NM_I(sbi)->nid_list_lock); |
1928 | update_free_nid_bitmap(sbi, start_nid, freed, true); | |
1929 | spin_unlock(&NM_I(sbi)->nid_list_lock); | |
4ac91242 CY |
1930 | } |
1931 | } | |
1932 | ||
1933 | static void scan_free_nid_bits(struct f2fs_sb_info *sbi) | |
1934 | { | |
1935 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1936 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1937 | struct f2fs_journal *journal = curseg->journal; | |
1938 | unsigned int i, idx; | |
4ac91242 CY |
1939 | |
1940 | down_read(&nm_i->nat_tree_lock); | |
1941 | ||
1942 | for (i = 0; i < nm_i->nat_blocks; i++) { | |
1943 | if (!test_bit_le(i, nm_i->nat_block_bitmap)) | |
1944 | continue; | |
586d1492 CY |
1945 | if (!nm_i->free_nid_count[i]) |
1946 | continue; | |
4ac91242 CY |
1947 | for (idx = 0; idx < NAT_ENTRY_PER_BLOCK; idx++) { |
1948 | nid_t nid; | |
1949 | ||
1950 | if (!test_bit_le(idx, nm_i->free_nid_bitmap[i])) | |
1951 | continue; | |
1952 | ||
1953 | nid = i * NAT_ENTRY_PER_BLOCK + idx; | |
1954 | add_free_nid(sbi, nid, true); | |
1955 | ||
f0cdbfe6 | 1956 | if (nm_i->nid_cnt[FREE_NID_LIST] >= MAX_FREE_NIDS) |
4ac91242 CY |
1957 | goto out; |
1958 | } | |
1959 | } | |
1960 | out: | |
1961 | down_read(&curseg->journal_rwsem); | |
1962 | for (i = 0; i < nats_in_cursum(journal); i++) { | |
1963 | block_t addr; | |
1964 | nid_t nid; | |
1965 | ||
1966 | addr = le32_to_cpu(nat_in_journal(journal, i).block_addr); | |
1967 | nid = le32_to_cpu(nid_in_journal(journal, i)); | |
1968 | if (addr == NULL_ADDR) | |
1969 | add_free_nid(sbi, nid, true); | |
1970 | else | |
1971 | remove_free_nid(sbi, nid); | |
e05df3b1 | 1972 | } |
4ac91242 CY |
1973 | up_read(&curseg->journal_rwsem); |
1974 | up_read(&nm_i->nat_tree_lock); | |
e05df3b1 JK |
1975 | } |
1976 | ||
22ad0b6a | 1977 | static void __build_free_nids(struct f2fs_sb_info *sbi, bool sync, bool mount) |
e05df3b1 | 1978 | { |
e05df3b1 JK |
1979 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1980 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 1981 | struct f2fs_journal *journal = curseg->journal; |
8760952d | 1982 | int i = 0; |
55008d84 | 1983 | nid_t nid = nm_i->next_scan_nid; |
e05df3b1 | 1984 | |
e9cdd307 YH |
1985 | if (unlikely(nid >= nm_i->max_nid)) |
1986 | nid = 0; | |
1987 | ||
55008d84 | 1988 | /* Enough entries */ |
b8559dc2 | 1989 | if (nm_i->nid_cnt[FREE_NID_LIST] >= NAT_ENTRY_PER_BLOCK) |
55008d84 | 1990 | return; |
e05df3b1 | 1991 | |
3a2ad567 | 1992 | if (!sync && !available_free_memory(sbi, FREE_NIDS)) |
55008d84 | 1993 | return; |
e05df3b1 | 1994 | |
4ac91242 CY |
1995 | if (!mount) { |
1996 | /* try to find free nids in free_nid_bitmap */ | |
1997 | scan_free_nid_bits(sbi); | |
1998 | ||
1999 | if (nm_i->nid_cnt[FREE_NID_LIST]) | |
2000 | return; | |
22ad0b6a JK |
2001 | } |
2002 | ||
55008d84 | 2003 | /* readahead nat pages to be scanned */ |
26879fb1 CY |
2004 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES, |
2005 | META_NAT, true); | |
e05df3b1 | 2006 | |
b873b798 | 2007 | down_read(&nm_i->nat_tree_lock); |
a5131193 | 2008 | |
e05df3b1 JK |
2009 | while (1) { |
2010 | struct page *page = get_current_nat_page(sbi, nid); | |
2011 | ||
6fb03f3a | 2012 | scan_nat_page(sbi, page, nid); |
e05df3b1 JK |
2013 | f2fs_put_page(page, 1); |
2014 | ||
2015 | nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); | |
cfb271d4 | 2016 | if (unlikely(nid >= nm_i->max_nid)) |
e05df3b1 | 2017 | nid = 0; |
55008d84 | 2018 | |
a6d494b6 | 2019 | if (++i >= FREE_NID_PAGES) |
e05df3b1 JK |
2020 | break; |
2021 | } | |
2022 | ||
55008d84 JK |
2023 | /* go to the next free nat pages to find free nids abundantly */ |
2024 | nm_i->next_scan_nid = nid; | |
e05df3b1 JK |
2025 | |
2026 | /* find free nids from current sum_pages */ | |
b7ad7512 | 2027 | down_read(&curseg->journal_rwsem); |
dfc08a12 CY |
2028 | for (i = 0; i < nats_in_cursum(journal); i++) { |
2029 | block_t addr; | |
2030 | ||
2031 | addr = le32_to_cpu(nat_in_journal(journal, i).block_addr); | |
2032 | nid = le32_to_cpu(nid_in_journal(journal, i)); | |
e05df3b1 | 2033 | if (addr == NULL_ADDR) |
6fb03f3a | 2034 | add_free_nid(sbi, nid, true); |
e05df3b1 | 2035 | else |
b8559dc2 | 2036 | remove_free_nid(sbi, nid); |
e05df3b1 | 2037 | } |
b7ad7512 | 2038 | up_read(&curseg->journal_rwsem); |
b873b798 | 2039 | up_read(&nm_i->nat_tree_lock); |
2db2388f CY |
2040 | |
2041 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid), | |
ea1a29a0 | 2042 | nm_i->ra_nid_pages, META_NAT, false); |
e05df3b1 JK |
2043 | } |
2044 | ||
22ad0b6a | 2045 | void build_free_nids(struct f2fs_sb_info *sbi, bool sync, bool mount) |
2411cf5b CY |
2046 | { |
2047 | mutex_lock(&NM_I(sbi)->build_lock); | |
22ad0b6a | 2048 | __build_free_nids(sbi, sync, mount); |
2411cf5b CY |
2049 | mutex_unlock(&NM_I(sbi)->build_lock); |
2050 | } | |
2051 | ||
e05df3b1 JK |
2052 | /* |
2053 | * If this function returns success, caller can obtain a new nid | |
2054 | * from second parameter of this function. | |
2055 | * The returned nid could be used ino as well as nid when inode is created. | |
2056 | */ | |
2057 | bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) | |
2058 | { | |
2059 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2060 | struct free_nid *i = NULL; | |
e05df3b1 | 2061 | retry: |
cb78942b | 2062 | #ifdef CONFIG_F2FS_FAULT_INJECTION |
55523519 CY |
2063 | if (time_to_inject(sbi, FAULT_ALLOC_NID)) { |
2064 | f2fs_show_injection_info(FAULT_ALLOC_NID); | |
cb78942b | 2065 | return false; |
55523519 | 2066 | } |
cb78942b | 2067 | #endif |
b8559dc2 | 2068 | spin_lock(&nm_i->nid_list_lock); |
e05df3b1 | 2069 | |
04d47e67 CY |
2070 | if (unlikely(nm_i->available_nids == 0)) { |
2071 | spin_unlock(&nm_i->nid_list_lock); | |
2072 | return false; | |
2073 | } | |
e05df3b1 | 2074 | |
55008d84 | 2075 | /* We should not use stale free nids created by build_free_nids */ |
b8559dc2 CY |
2076 | if (nm_i->nid_cnt[FREE_NID_LIST] && !on_build_free_nids(nm_i)) { |
2077 | f2fs_bug_on(sbi, list_empty(&nm_i->nid_list[FREE_NID_LIST])); | |
2078 | i = list_first_entry(&nm_i->nid_list[FREE_NID_LIST], | |
2079 | struct free_nid, list); | |
55008d84 | 2080 | *nid = i->nid; |
b8559dc2 | 2081 | |
eb0aa4b8 | 2082 | __remove_nid_from_list(sbi, i, FREE_NID_LIST, true); |
55008d84 | 2083 | i->state = NID_ALLOC; |
eb0aa4b8 | 2084 | __insert_nid_to_list(sbi, i, ALLOC_NID_LIST, false); |
04d47e67 | 2085 | nm_i->available_nids--; |
4ac91242 | 2086 | |
346fe752 | 2087 | update_free_nid_bitmap(sbi, *nid, false, false); |
4ac91242 | 2088 | |
b8559dc2 | 2089 | spin_unlock(&nm_i->nid_list_lock); |
55008d84 JK |
2090 | return true; |
2091 | } | |
b8559dc2 | 2092 | spin_unlock(&nm_i->nid_list_lock); |
55008d84 JK |
2093 | |
2094 | /* Let's scan nat pages and its caches to get free nids */ | |
22ad0b6a | 2095 | build_free_nids(sbi, true, false); |
55008d84 | 2096 | goto retry; |
e05df3b1 JK |
2097 | } |
2098 | ||
0a8165d7 | 2099 | /* |
e05df3b1 JK |
2100 | * alloc_nid() should be called prior to this function. |
2101 | */ | |
2102 | void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) | |
2103 | { | |
2104 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2105 | struct free_nid *i; | |
2106 | ||
b8559dc2 | 2107 | spin_lock(&nm_i->nid_list_lock); |
8a7ed66a | 2108 | i = __lookup_free_nid_list(nm_i, nid); |
b8559dc2 | 2109 | f2fs_bug_on(sbi, !i); |
eb0aa4b8 | 2110 | __remove_nid_from_list(sbi, i, ALLOC_NID_LIST, false); |
b8559dc2 | 2111 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 CY |
2112 | |
2113 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
2114 | } |
2115 | ||
0a8165d7 | 2116 | /* |
e05df3b1 JK |
2117 | * alloc_nid() should be called prior to this function. |
2118 | */ | |
2119 | void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) | |
2120 | { | |
49952fa1 JK |
2121 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
2122 | struct free_nid *i; | |
cf0ee0f0 | 2123 | bool need_free = false; |
49952fa1 | 2124 | |
65985d93 JK |
2125 | if (!nid) |
2126 | return; | |
2127 | ||
b8559dc2 | 2128 | spin_lock(&nm_i->nid_list_lock); |
8a7ed66a | 2129 | i = __lookup_free_nid_list(nm_i, nid); |
b8559dc2 CY |
2130 | f2fs_bug_on(sbi, !i); |
2131 | ||
6fb03f3a | 2132 | if (!available_free_memory(sbi, FREE_NIDS)) { |
eb0aa4b8 | 2133 | __remove_nid_from_list(sbi, i, ALLOC_NID_LIST, false); |
cf0ee0f0 | 2134 | need_free = true; |
95630cba | 2135 | } else { |
eb0aa4b8 | 2136 | __remove_nid_from_list(sbi, i, ALLOC_NID_LIST, true); |
95630cba | 2137 | i->state = NID_NEW; |
eb0aa4b8 | 2138 | __insert_nid_to_list(sbi, i, FREE_NID_LIST, false); |
95630cba | 2139 | } |
04d47e67 CY |
2140 | |
2141 | nm_i->available_nids++; | |
2142 | ||
346fe752 | 2143 | update_free_nid_bitmap(sbi, nid, true, false); |
4ac91242 | 2144 | |
b8559dc2 | 2145 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 CY |
2146 | |
2147 | if (need_free) | |
2148 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
2149 | } |
2150 | ||
31696580 CY |
2151 | int try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink) |
2152 | { | |
2153 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2154 | struct free_nid *i, *next; | |
2155 | int nr = nr_shrink; | |
2156 | ||
b8559dc2 | 2157 | if (nm_i->nid_cnt[FREE_NID_LIST] <= MAX_FREE_NIDS) |
ad4edb83 JK |
2158 | return 0; |
2159 | ||
31696580 CY |
2160 | if (!mutex_trylock(&nm_i->build_lock)) |
2161 | return 0; | |
2162 | ||
b8559dc2 CY |
2163 | spin_lock(&nm_i->nid_list_lock); |
2164 | list_for_each_entry_safe(i, next, &nm_i->nid_list[FREE_NID_LIST], | |
2165 | list) { | |
2166 | if (nr_shrink <= 0 || | |
2167 | nm_i->nid_cnt[FREE_NID_LIST] <= MAX_FREE_NIDS) | |
31696580 | 2168 | break; |
b8559dc2 | 2169 | |
eb0aa4b8 | 2170 | __remove_nid_from_list(sbi, i, FREE_NID_LIST, false); |
31696580 CY |
2171 | kmem_cache_free(free_nid_slab, i); |
2172 | nr_shrink--; | |
31696580 | 2173 | } |
b8559dc2 | 2174 | spin_unlock(&nm_i->nid_list_lock); |
31696580 CY |
2175 | mutex_unlock(&nm_i->build_lock); |
2176 | ||
2177 | return nr - nr_shrink; | |
2178 | } | |
2179 | ||
70cfed88 | 2180 | void recover_inline_xattr(struct inode *inode, struct page *page) |
28cdce04 | 2181 | { |
28cdce04 CY |
2182 | void *src_addr, *dst_addr; |
2183 | size_t inline_size; | |
2184 | struct page *ipage; | |
2185 | struct f2fs_inode *ri; | |
2186 | ||
4081363f | 2187 | ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino); |
9850cf4a | 2188 | f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage)); |
28cdce04 | 2189 | |
e3b4d43f JK |
2190 | ri = F2FS_INODE(page); |
2191 | if (!(ri->i_inline & F2FS_INLINE_XATTR)) { | |
91942321 | 2192 | clear_inode_flag(inode, FI_INLINE_XATTR); |
e3b4d43f JK |
2193 | goto update_inode; |
2194 | } | |
2195 | ||
28cdce04 CY |
2196 | dst_addr = inline_xattr_addr(ipage); |
2197 | src_addr = inline_xattr_addr(page); | |
2198 | inline_size = inline_xattr_size(inode); | |
2199 | ||
fec1d657 | 2200 | f2fs_wait_on_page_writeback(ipage, NODE, true); |
28cdce04 | 2201 | memcpy(dst_addr, src_addr, inline_size); |
e3b4d43f | 2202 | update_inode: |
28cdce04 CY |
2203 | update_inode(inode, ipage); |
2204 | f2fs_put_page(ipage, 1); | |
2205 | } | |
2206 | ||
d260081c | 2207 | int recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr) |
abb2366c | 2208 | { |
4081363f | 2209 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
abb2366c | 2210 | nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid; |
87905682 YH |
2211 | nid_t new_xnid; |
2212 | struct dnode_of_data dn; | |
abb2366c | 2213 | struct node_info ni; |
d260081c | 2214 | struct page *xpage; |
abb2366c | 2215 | |
abb2366c JK |
2216 | if (!prev_xnid) |
2217 | goto recover_xnid; | |
2218 | ||
d260081c | 2219 | /* 1: invalidate the previous xattr nid */ |
abb2366c | 2220 | get_node_info(sbi, prev_xnid, &ni); |
9850cf4a | 2221 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
abb2366c | 2222 | invalidate_blocks(sbi, ni.blk_addr); |
000519f2 | 2223 | dec_valid_node_count(sbi, inode, false); |
479f40c4 | 2224 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
abb2366c JK |
2225 | |
2226 | recover_xnid: | |
d260081c | 2227 | /* 2: update xattr nid in inode */ |
87905682 YH |
2228 | if (!alloc_nid(sbi, &new_xnid)) |
2229 | return -ENOSPC; | |
2230 | ||
2231 | set_new_dnode(&dn, inode, NULL, NULL, new_xnid); | |
2232 | xpage = new_node_page(&dn, XATTR_NODE_OFFSET); | |
2233 | if (IS_ERR(xpage)) { | |
2234 | alloc_nid_failed(sbi, new_xnid); | |
2235 | return PTR_ERR(xpage); | |
2236 | } | |
2237 | ||
2238 | alloc_nid_done(sbi, new_xnid); | |
d260081c CY |
2239 | update_inode_page(inode); |
2240 | ||
2241 | /* 3: update and set xattr node page dirty */ | |
87905682 | 2242 | memcpy(F2FS_NODE(xpage), F2FS_NODE(page), VALID_XATTR_BLOCK_SIZE); |
abb2366c | 2243 | |
d260081c CY |
2244 | set_page_dirty(xpage); |
2245 | f2fs_put_page(xpage, 1); | |
abb2366c | 2246 | |
d260081c | 2247 | return 0; |
abb2366c JK |
2248 | } |
2249 | ||
e05df3b1 JK |
2250 | int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) |
2251 | { | |
58bfaf44 | 2252 | struct f2fs_inode *src, *dst; |
e05df3b1 JK |
2253 | nid_t ino = ino_of_node(page); |
2254 | struct node_info old_ni, new_ni; | |
2255 | struct page *ipage; | |
2256 | ||
e8271fa3 JK |
2257 | get_node_info(sbi, ino, &old_ni); |
2258 | ||
2259 | if (unlikely(old_ni.blk_addr != NULL_ADDR)) | |
2260 | return -EINVAL; | |
e8ea9b3d | 2261 | retry: |
300e129c | 2262 | ipage = f2fs_grab_cache_page(NODE_MAPPING(sbi), ino, false); |
e8ea9b3d JK |
2263 | if (!ipage) { |
2264 | congestion_wait(BLK_RW_ASYNC, HZ/50); | |
2265 | goto retry; | |
2266 | } | |
e05df3b1 | 2267 | |
e1c42045 | 2268 | /* Should not use this inode from free nid list */ |
b8559dc2 | 2269 | remove_free_nid(sbi, ino); |
e05df3b1 | 2270 | |
237c0790 JK |
2271 | if (!PageUptodate(ipage)) |
2272 | SetPageUptodate(ipage); | |
e05df3b1 JK |
2273 | fill_node_footer(ipage, ino, ino, 0, true); |
2274 | ||
58bfaf44 JK |
2275 | src = F2FS_INODE(page); |
2276 | dst = F2FS_INODE(ipage); | |
e05df3b1 | 2277 | |
58bfaf44 JK |
2278 | memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); |
2279 | dst->i_size = 0; | |
2280 | dst->i_blocks = cpu_to_le64(1); | |
2281 | dst->i_links = cpu_to_le32(1); | |
2282 | dst->i_xattr_nid = 0; | |
7a2af766 | 2283 | dst->i_inline = src->i_inline & (F2FS_INLINE_XATTR | F2FS_EXTRA_ATTR); |
5c57132e | 2284 | if (dst->i_inline & F2FS_EXTRA_ATTR) { |
7a2af766 | 2285 | dst->i_extra_isize = src->i_extra_isize; |
5c57132e CY |
2286 | if (f2fs_sb_has_project_quota(sbi->sb) && |
2287 | F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize), | |
2288 | i_projid)) | |
2289 | dst->i_projid = src->i_projid; | |
2290 | } | |
e05df3b1 JK |
2291 | |
2292 | new_ni = old_ni; | |
2293 | new_ni.ino = ino; | |
2294 | ||
0abd675e | 2295 | if (unlikely(inc_valid_node_count(sbi, NULL, true))) |
65e5cd0a | 2296 | WARN_ON(1); |
479f40c4 | 2297 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
e05df3b1 | 2298 | inc_valid_inode_count(sbi); |
617deb8c | 2299 | set_page_dirty(ipage); |
e05df3b1 JK |
2300 | f2fs_put_page(ipage, 1); |
2301 | return 0; | |
2302 | } | |
2303 | ||
2304 | int restore_node_summary(struct f2fs_sb_info *sbi, | |
2305 | unsigned int segno, struct f2fs_summary_block *sum) | |
2306 | { | |
2307 | struct f2fs_node *rn; | |
2308 | struct f2fs_summary *sum_entry; | |
e05df3b1 | 2309 | block_t addr; |
9ecf4b80 | 2310 | int i, idx, last_offset, nrpages; |
e05df3b1 JK |
2311 | |
2312 | /* scan the node segment */ | |
2313 | last_offset = sbi->blocks_per_seg; | |
2314 | addr = START_BLOCK(sbi, segno); | |
2315 | sum_entry = &sum->entries[0]; | |
2316 | ||
9ecf4b80 | 2317 | for (i = 0; i < last_offset; i += nrpages, addr += nrpages) { |
664ba972 | 2318 | nrpages = min(last_offset - i, BIO_MAX_PAGES); |
393ff91f | 2319 | |
e1c42045 | 2320 | /* readahead node pages */ |
26879fb1 | 2321 | ra_meta_pages(sbi, addr, nrpages, META_POR, true); |
e05df3b1 | 2322 | |
9ecf4b80 | 2323 | for (idx = addr; idx < addr + nrpages; idx++) { |
2b947003 | 2324 | struct page *page = get_tmp_page(sbi, idx); |
9af0ff1c | 2325 | |
9ecf4b80 CY |
2326 | rn = F2FS_NODE(page); |
2327 | sum_entry->nid = rn->footer.nid; | |
2328 | sum_entry->version = 0; | |
2329 | sum_entry->ofs_in_node = 0; | |
2330 | sum_entry++; | |
2331 | f2fs_put_page(page, 1); | |
9af0ff1c | 2332 | } |
bac4eef6 | 2333 | |
9ecf4b80 | 2334 | invalidate_mapping_pages(META_MAPPING(sbi), addr, |
bac4eef6 | 2335 | addr + nrpages); |
e05df3b1 | 2336 | } |
9ecf4b80 | 2337 | return 0; |
e05df3b1 JK |
2338 | } |
2339 | ||
aec71382 | 2340 | static void remove_nats_in_journal(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
2341 | { |
2342 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2343 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2344 | struct f2fs_journal *journal = curseg->journal; |
e05df3b1 JK |
2345 | int i; |
2346 | ||
b7ad7512 | 2347 | down_write(&curseg->journal_rwsem); |
dfc08a12 | 2348 | for (i = 0; i < nats_in_cursum(journal); i++) { |
e05df3b1 JK |
2349 | struct nat_entry *ne; |
2350 | struct f2fs_nat_entry raw_ne; | |
dfc08a12 | 2351 | nid_t nid = le32_to_cpu(nid_in_journal(journal, i)); |
e05df3b1 | 2352 | |
dfc08a12 | 2353 | raw_ne = nat_in_journal(journal, i); |
9be32d72 | 2354 | |
e05df3b1 | 2355 | ne = __lookup_nat_cache(nm_i, nid); |
e05df3b1 | 2356 | if (!ne) { |
5c9e4184 | 2357 | ne = grab_nat_entry(nm_i, nid, true); |
9be32d72 | 2358 | node_info_from_raw_nat(&ne->ni, &raw_ne); |
e05df3b1 | 2359 | } |
04d47e67 CY |
2360 | |
2361 | /* | |
2362 | * if a free nat in journal has not been used after last | |
2363 | * checkpoint, we should remove it from available nids, | |
2364 | * since later we will add it again. | |
2365 | */ | |
2366 | if (!get_nat_flag(ne, IS_DIRTY) && | |
2367 | le32_to_cpu(raw_ne.block_addr) == NULL_ADDR) { | |
2368 | spin_lock(&nm_i->nid_list_lock); | |
2369 | nm_i->available_nids--; | |
2370 | spin_unlock(&nm_i->nid_list_lock); | |
2371 | } | |
2372 | ||
e05df3b1 | 2373 | __set_nat_cache_dirty(nm_i, ne); |
e05df3b1 | 2374 | } |
dfc08a12 | 2375 | update_nats_in_cursum(journal, -i); |
b7ad7512 | 2376 | up_write(&curseg->journal_rwsem); |
e05df3b1 JK |
2377 | } |
2378 | ||
309cc2b6 JK |
2379 | static void __adjust_nat_entry_set(struct nat_entry_set *nes, |
2380 | struct list_head *head, int max) | |
e05df3b1 | 2381 | { |
309cc2b6 | 2382 | struct nat_entry_set *cur; |
e05df3b1 | 2383 | |
309cc2b6 JK |
2384 | if (nes->entry_cnt >= max) |
2385 | goto add_out; | |
e05df3b1 | 2386 | |
309cc2b6 JK |
2387 | list_for_each_entry(cur, head, set_list) { |
2388 | if (cur->entry_cnt >= nes->entry_cnt) { | |
2389 | list_add(&nes->set_list, cur->set_list.prev); | |
2390 | return; | |
2391 | } | |
aec71382 | 2392 | } |
309cc2b6 JK |
2393 | add_out: |
2394 | list_add_tail(&nes->set_list, head); | |
2395 | } | |
e05df3b1 | 2396 | |
9f7e4a2c | 2397 | static void __update_nat_bits(struct f2fs_sb_info *sbi, nid_t start_nid, |
22ad0b6a JK |
2398 | struct page *page) |
2399 | { | |
2400 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2401 | unsigned int nat_index = start_nid / NAT_ENTRY_PER_BLOCK; | |
2402 | struct f2fs_nat_block *nat_blk = page_address(page); | |
2403 | int valid = 0; | |
2404 | int i; | |
2405 | ||
2406 | if (!enabled_nat_bits(sbi, NULL)) | |
2407 | return; | |
2408 | ||
2409 | for (i = 0; i < NAT_ENTRY_PER_BLOCK; i++) { | |
2410 | if (start_nid == 0 && i == 0) | |
2411 | valid++; | |
2412 | if (nat_blk->entries[i].block_addr) | |
2413 | valid++; | |
2414 | } | |
2415 | if (valid == 0) { | |
23380b85 JK |
2416 | __set_bit_le(nat_index, nm_i->empty_nat_bits); |
2417 | __clear_bit_le(nat_index, nm_i->full_nat_bits); | |
22ad0b6a JK |
2418 | return; |
2419 | } | |
2420 | ||
23380b85 | 2421 | __clear_bit_le(nat_index, nm_i->empty_nat_bits); |
22ad0b6a | 2422 | if (valid == NAT_ENTRY_PER_BLOCK) |
23380b85 | 2423 | __set_bit_le(nat_index, nm_i->full_nat_bits); |
22ad0b6a | 2424 | else |
23380b85 | 2425 | __clear_bit_le(nat_index, nm_i->full_nat_bits); |
22ad0b6a JK |
2426 | } |
2427 | ||
309cc2b6 | 2428 | static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, |
22ad0b6a | 2429 | struct nat_entry_set *set, struct cp_control *cpc) |
309cc2b6 JK |
2430 | { |
2431 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2432 | struct f2fs_journal *journal = curseg->journal; |
309cc2b6 JK |
2433 | nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK; |
2434 | bool to_journal = true; | |
2435 | struct f2fs_nat_block *nat_blk; | |
2436 | struct nat_entry *ne, *cur; | |
2437 | struct page *page = NULL; | |
e05df3b1 | 2438 | |
aec71382 CY |
2439 | /* |
2440 | * there are two steps to flush nat entries: | |
2441 | * #1, flush nat entries to journal in current hot data summary block. | |
2442 | * #2, flush nat entries to nat page. | |
2443 | */ | |
22ad0b6a JK |
2444 | if (enabled_nat_bits(sbi, cpc) || |
2445 | !__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL)) | |
309cc2b6 JK |
2446 | to_journal = false; |
2447 | ||
2448 | if (to_journal) { | |
b7ad7512 | 2449 | down_write(&curseg->journal_rwsem); |
309cc2b6 JK |
2450 | } else { |
2451 | page = get_next_nat_page(sbi, start_nid); | |
2452 | nat_blk = page_address(page); | |
2453 | f2fs_bug_on(sbi, !nat_blk); | |
2454 | } | |
aec71382 | 2455 | |
309cc2b6 JK |
2456 | /* flush dirty nats in nat entry set */ |
2457 | list_for_each_entry_safe(ne, cur, &set->entry_list, list) { | |
2458 | struct f2fs_nat_entry *raw_ne; | |
2459 | nid_t nid = nat_get_nid(ne); | |
2460 | int offset; | |
2461 | ||
febeca6d | 2462 | f2fs_bug_on(sbi, nat_get_blkaddr(ne) == NEW_ADDR); |
aec71382 CY |
2463 | |
2464 | if (to_journal) { | |
dfc08a12 | 2465 | offset = lookup_journal_in_cursum(journal, |
309cc2b6 JK |
2466 | NAT_JOURNAL, nid, 1); |
2467 | f2fs_bug_on(sbi, offset < 0); | |
dfc08a12 CY |
2468 | raw_ne = &nat_in_journal(journal, offset); |
2469 | nid_in_journal(journal, offset) = cpu_to_le32(nid); | |
aec71382 | 2470 | } else { |
309cc2b6 | 2471 | raw_ne = &nat_blk->entries[nid - start_nid]; |
e05df3b1 | 2472 | } |
309cc2b6 | 2473 | raw_nat_from_node_info(raw_ne, &ne->ni); |
309cc2b6 | 2474 | nat_reset_flag(ne); |
0b28b71e | 2475 | __clear_nat_cache_dirty(NM_I(sbi), set, ne); |
04d47e67 | 2476 | if (nat_get_blkaddr(ne) == NULL_ADDR) { |
309cc2b6 | 2477 | add_free_nid(sbi, nid, false); |
04d47e67 CY |
2478 | spin_lock(&NM_I(sbi)->nid_list_lock); |
2479 | NM_I(sbi)->available_nids++; | |
346fe752 | 2480 | update_free_nid_bitmap(sbi, nid, true, false); |
4ac91242 CY |
2481 | spin_unlock(&NM_I(sbi)->nid_list_lock); |
2482 | } else { | |
2483 | spin_lock(&NM_I(sbi)->nid_list_lock); | |
346fe752 | 2484 | update_free_nid_bitmap(sbi, nid, false, false); |
04d47e67 CY |
2485 | spin_unlock(&NM_I(sbi)->nid_list_lock); |
2486 | } | |
309cc2b6 | 2487 | } |
e05df3b1 | 2488 | |
22ad0b6a | 2489 | if (to_journal) { |
b7ad7512 | 2490 | up_write(&curseg->journal_rwsem); |
22ad0b6a JK |
2491 | } else { |
2492 | __update_nat_bits(sbi, start_nid, page); | |
309cc2b6 | 2493 | f2fs_put_page(page, 1); |
22ad0b6a | 2494 | } |
aec71382 | 2495 | |
59c9081b YH |
2496 | /* Allow dirty nats by node block allocation in write_begin */ |
2497 | if (!set->entry_cnt) { | |
2498 | radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); | |
2499 | kmem_cache_free(nat_entry_set_slab, set); | |
2500 | } | |
309cc2b6 | 2501 | } |
aec71382 | 2502 | |
309cc2b6 JK |
2503 | /* |
2504 | * This function is called during the checkpointing process. | |
2505 | */ | |
22ad0b6a | 2506 | void flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc) |
309cc2b6 JK |
2507 | { |
2508 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2509 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2510 | struct f2fs_journal *journal = curseg->journal; |
7aed0d45 | 2511 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
309cc2b6 JK |
2512 | struct nat_entry_set *set, *tmp; |
2513 | unsigned int found; | |
2514 | nid_t set_idx = 0; | |
2515 | LIST_HEAD(sets); | |
2516 | ||
20d047c8 CL |
2517 | if (!nm_i->dirty_nat_cnt) |
2518 | return; | |
a5131193 | 2519 | |
b873b798 | 2520 | down_write(&nm_i->nat_tree_lock); |
a5131193 | 2521 | |
309cc2b6 JK |
2522 | /* |
2523 | * if there are no enough space in journal to store dirty nat | |
2524 | * entries, remove all entries from journal and merge them | |
2525 | * into nat entry set. | |
2526 | */ | |
900f7362 | 2527 | if (enabled_nat_bits(sbi, cpc) || |
22ad0b6a | 2528 | !__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL)) |
309cc2b6 JK |
2529 | remove_nats_in_journal(sbi); |
2530 | ||
309cc2b6 | 2531 | while ((found = __gang_lookup_nat_set(nm_i, |
7aed0d45 | 2532 | set_idx, SETVEC_SIZE, setvec))) { |
309cc2b6 JK |
2533 | unsigned idx; |
2534 | set_idx = setvec[found - 1]->set + 1; | |
2535 | for (idx = 0; idx < found; idx++) | |
2536 | __adjust_nat_entry_set(setvec[idx], &sets, | |
dfc08a12 | 2537 | MAX_NAT_JENTRIES(journal)); |
e05df3b1 | 2538 | } |
aec71382 | 2539 | |
309cc2b6 JK |
2540 | /* flush dirty nats in nat entry set */ |
2541 | list_for_each_entry_safe(set, tmp, &sets, set_list) | |
22ad0b6a | 2542 | __flush_nat_entry_set(sbi, set, cpc); |
309cc2b6 | 2543 | |
b873b798 | 2544 | up_write(&nm_i->nat_tree_lock); |
59c9081b | 2545 | /* Allow dirty nats by node block allocation in write_begin */ |
e05df3b1 JK |
2546 | } |
2547 | ||
22ad0b6a JK |
2548 | static int __get_nat_bitmaps(struct f2fs_sb_info *sbi) |
2549 | { | |
2550 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
2551 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2552 | unsigned int nat_bits_bytes = nm_i->nat_blocks / BITS_PER_BYTE; | |
2553 | unsigned int i; | |
2554 | __u64 cp_ver = cur_cp_version(ckpt); | |
22ad0b6a JK |
2555 | block_t nat_bits_addr; |
2556 | ||
2557 | if (!enabled_nat_bits(sbi, NULL)) | |
2558 | return 0; | |
2559 | ||
2560 | nm_i->nat_bits_blocks = F2FS_BYTES_TO_BLK((nat_bits_bytes << 1) + 8 + | |
2561 | F2FS_BLKSIZE - 1); | |
2562 | nm_i->nat_bits = kzalloc(nm_i->nat_bits_blocks << F2FS_BLKSIZE_BITS, | |
2563 | GFP_KERNEL); | |
2564 | if (!nm_i->nat_bits) | |
2565 | return -ENOMEM; | |
2566 | ||
2567 | nat_bits_addr = __start_cp_addr(sbi) + sbi->blocks_per_seg - | |
2568 | nm_i->nat_bits_blocks; | |
2569 | for (i = 0; i < nm_i->nat_bits_blocks; i++) { | |
2570 | struct page *page = get_meta_page(sbi, nat_bits_addr++); | |
2571 | ||
2572 | memcpy(nm_i->nat_bits + (i << F2FS_BLKSIZE_BITS), | |
2573 | page_address(page), F2FS_BLKSIZE); | |
2574 | f2fs_put_page(page, 1); | |
2575 | } | |
2576 | ||
ced2c7ea | 2577 | cp_ver |= (cur_cp_crc(ckpt) << 32); |
22ad0b6a JK |
2578 | if (cpu_to_le64(cp_ver) != *(__le64 *)nm_i->nat_bits) { |
2579 | disable_nat_bits(sbi, true); | |
2580 | return 0; | |
2581 | } | |
2582 | ||
2583 | nm_i->full_nat_bits = nm_i->nat_bits + 8; | |
2584 | nm_i->empty_nat_bits = nm_i->full_nat_bits + nat_bits_bytes; | |
2585 | ||
2586 | f2fs_msg(sbi->sb, KERN_NOTICE, "Found nat_bits in checkpoint"); | |
2587 | return 0; | |
2588 | } | |
2589 | ||
bd80a4b9 | 2590 | static inline void load_free_nid_bitmap(struct f2fs_sb_info *sbi) |
7041d5d2 CY |
2591 | { |
2592 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2593 | unsigned int i = 0; | |
2594 | nid_t nid, last_nid; | |
2595 | ||
2596 | if (!enabled_nat_bits(sbi, NULL)) | |
2597 | return; | |
2598 | ||
2599 | for (i = 0; i < nm_i->nat_blocks; i++) { | |
2600 | i = find_next_bit_le(nm_i->empty_nat_bits, nm_i->nat_blocks, i); | |
2601 | if (i >= nm_i->nat_blocks) | |
2602 | break; | |
2603 | ||
2604 | __set_bit_le(i, nm_i->nat_block_bitmap); | |
2605 | ||
2606 | nid = i * NAT_ENTRY_PER_BLOCK; | |
2607 | last_nid = (i + 1) * NAT_ENTRY_PER_BLOCK; | |
2608 | ||
346fe752 | 2609 | spin_lock(&NM_I(sbi)->nid_list_lock); |
7041d5d2 | 2610 | for (; nid < last_nid; nid++) |
346fe752 CY |
2611 | update_free_nid_bitmap(sbi, nid, true, true); |
2612 | spin_unlock(&NM_I(sbi)->nid_list_lock); | |
7041d5d2 CY |
2613 | } |
2614 | ||
2615 | for (i = 0; i < nm_i->nat_blocks; i++) { | |
2616 | i = find_next_bit_le(nm_i->full_nat_bits, nm_i->nat_blocks, i); | |
2617 | if (i >= nm_i->nat_blocks) | |
2618 | break; | |
2619 | ||
2620 | __set_bit_le(i, nm_i->nat_block_bitmap); | |
2621 | } | |
2622 | } | |
2623 | ||
e05df3b1 JK |
2624 | static int init_node_manager(struct f2fs_sb_info *sbi) |
2625 | { | |
2626 | struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); | |
2627 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2628 | unsigned char *version_bitmap; | |
22ad0b6a JK |
2629 | unsigned int nat_segs; |
2630 | int err; | |
e05df3b1 JK |
2631 | |
2632 | nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); | |
2633 | ||
2634 | /* segment_count_nat includes pair segment so divide to 2. */ | |
2635 | nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; | |
22ad0b6a JK |
2636 | nm_i->nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); |
2637 | nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nm_i->nat_blocks; | |
7ee0eeab | 2638 | |
b63da15e | 2639 | /* not used nids: 0, node, meta, (and root counted as valid node) */ |
04d47e67 CY |
2640 | nm_i->available_nids = nm_i->max_nid - sbi->total_valid_node_count - |
2641 | F2FS_RESERVED_NODE_NUM; | |
b8559dc2 CY |
2642 | nm_i->nid_cnt[FREE_NID_LIST] = 0; |
2643 | nm_i->nid_cnt[ALLOC_NID_LIST] = 0; | |
e05df3b1 | 2644 | nm_i->nat_cnt = 0; |
cdfc41c1 | 2645 | nm_i->ram_thresh = DEF_RAM_THRESHOLD; |
ea1a29a0 | 2646 | nm_i->ra_nid_pages = DEF_RA_NID_PAGES; |
2304cb0c | 2647 | nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD; |
e05df3b1 | 2648 | |
8a7ed66a | 2649 | INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC); |
b8559dc2 CY |
2650 | INIT_LIST_HEAD(&nm_i->nid_list[FREE_NID_LIST]); |
2651 | INIT_LIST_HEAD(&nm_i->nid_list[ALLOC_NID_LIST]); | |
769ec6e5 JK |
2652 | INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO); |
2653 | INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO); | |
e05df3b1 | 2654 | INIT_LIST_HEAD(&nm_i->nat_entries); |
e05df3b1 JK |
2655 | |
2656 | mutex_init(&nm_i->build_lock); | |
b8559dc2 | 2657 | spin_lock_init(&nm_i->nid_list_lock); |
b873b798 | 2658 | init_rwsem(&nm_i->nat_tree_lock); |
e05df3b1 | 2659 | |
e05df3b1 | 2660 | nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); |
79b5793b | 2661 | nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); |
e05df3b1 JK |
2662 | version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); |
2663 | if (!version_bitmap) | |
2664 | return -EFAULT; | |
2665 | ||
79b5793b AG |
2666 | nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, |
2667 | GFP_KERNEL); | |
2668 | if (!nm_i->nat_bitmap) | |
2669 | return -ENOMEM; | |
599a09b2 | 2670 | |
22ad0b6a JK |
2671 | err = __get_nat_bitmaps(sbi); |
2672 | if (err) | |
2673 | return err; | |
2674 | ||
599a09b2 CY |
2675 | #ifdef CONFIG_F2FS_CHECK_FS |
2676 | nm_i->nat_bitmap_mir = kmemdup(version_bitmap, nm_i->bitmap_size, | |
2677 | GFP_KERNEL); | |
2678 | if (!nm_i->nat_bitmap_mir) | |
2679 | return -ENOMEM; | |
2680 | #endif | |
2681 | ||
e05df3b1 JK |
2682 | return 0; |
2683 | } | |
2684 | ||
9f7e4a2c | 2685 | static int init_free_nid_cache(struct f2fs_sb_info *sbi) |
4ac91242 CY |
2686 | { |
2687 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2688 | ||
a7c3e901 | 2689 | nm_i->free_nid_bitmap = kvzalloc(nm_i->nat_blocks * |
4ac91242 CY |
2690 | NAT_ENTRY_BITMAP_SIZE, GFP_KERNEL); |
2691 | if (!nm_i->free_nid_bitmap) | |
2692 | return -ENOMEM; | |
2693 | ||
a7c3e901 | 2694 | nm_i->nat_block_bitmap = kvzalloc(nm_i->nat_blocks / 8, |
4ac91242 CY |
2695 | GFP_KERNEL); |
2696 | if (!nm_i->nat_block_bitmap) | |
2697 | return -ENOMEM; | |
586d1492 | 2698 | |
a7c3e901 | 2699 | nm_i->free_nid_count = kvzalloc(nm_i->nat_blocks * |
586d1492 CY |
2700 | sizeof(unsigned short), GFP_KERNEL); |
2701 | if (!nm_i->free_nid_count) | |
2702 | return -ENOMEM; | |
4ac91242 CY |
2703 | return 0; |
2704 | } | |
2705 | ||
e05df3b1 JK |
2706 | int build_node_manager(struct f2fs_sb_info *sbi) |
2707 | { | |
2708 | int err; | |
2709 | ||
2710 | sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL); | |
2711 | if (!sbi->nm_info) | |
2712 | return -ENOMEM; | |
2713 | ||
2714 | err = init_node_manager(sbi); | |
2715 | if (err) | |
2716 | return err; | |
2717 | ||
4ac91242 CY |
2718 | err = init_free_nid_cache(sbi); |
2719 | if (err) | |
2720 | return err; | |
2721 | ||
7041d5d2 CY |
2722 | /* load free nid status from nat_bits table */ |
2723 | load_free_nid_bitmap(sbi); | |
2724 | ||
22ad0b6a | 2725 | build_free_nids(sbi, true, true); |
e05df3b1 JK |
2726 | return 0; |
2727 | } | |
2728 | ||
2729 | void destroy_node_manager(struct f2fs_sb_info *sbi) | |
2730 | { | |
2731 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2732 | struct free_nid *i, *next_i; | |
2733 | struct nat_entry *natvec[NATVEC_SIZE]; | |
7aed0d45 | 2734 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
e05df3b1 JK |
2735 | nid_t nid = 0; |
2736 | unsigned int found; | |
2737 | ||
2738 | if (!nm_i) | |
2739 | return; | |
2740 | ||
2741 | /* destroy free nid list */ | |
b8559dc2 CY |
2742 | spin_lock(&nm_i->nid_list_lock); |
2743 | list_for_each_entry_safe(i, next_i, &nm_i->nid_list[FREE_NID_LIST], | |
2744 | list) { | |
eb0aa4b8 | 2745 | __remove_nid_from_list(sbi, i, FREE_NID_LIST, false); |
b8559dc2 | 2746 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 | 2747 | kmem_cache_free(free_nid_slab, i); |
b8559dc2 | 2748 | spin_lock(&nm_i->nid_list_lock); |
e05df3b1 | 2749 | } |
b8559dc2 CY |
2750 | f2fs_bug_on(sbi, nm_i->nid_cnt[FREE_NID_LIST]); |
2751 | f2fs_bug_on(sbi, nm_i->nid_cnt[ALLOC_NID_LIST]); | |
2752 | f2fs_bug_on(sbi, !list_empty(&nm_i->nid_list[ALLOC_NID_LIST])); | |
2753 | spin_unlock(&nm_i->nid_list_lock); | |
e05df3b1 JK |
2754 | |
2755 | /* destroy nat cache */ | |
b873b798 | 2756 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2757 | while ((found = __gang_lookup_nat_cache(nm_i, |
2758 | nid, NATVEC_SIZE, natvec))) { | |
2759 | unsigned idx; | |
7aed0d45 | 2760 | |
b6ce391e GZ |
2761 | nid = nat_get_nid(natvec[found - 1]) + 1; |
2762 | for (idx = 0; idx < found; idx++) | |
2763 | __del_from_nat_cache(nm_i, natvec[idx]); | |
e05df3b1 | 2764 | } |
9850cf4a | 2765 | f2fs_bug_on(sbi, nm_i->nat_cnt); |
7aed0d45 JK |
2766 | |
2767 | /* destroy nat set cache */ | |
2768 | nid = 0; | |
2769 | while ((found = __gang_lookup_nat_set(nm_i, | |
2770 | nid, SETVEC_SIZE, setvec))) { | |
2771 | unsigned idx; | |
2772 | ||
2773 | nid = setvec[found - 1]->set + 1; | |
2774 | for (idx = 0; idx < found; idx++) { | |
2775 | /* entry_cnt is not zero, when cp_error was occurred */ | |
2776 | f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list)); | |
2777 | radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set); | |
2778 | kmem_cache_free(nat_entry_set_slab, setvec[idx]); | |
2779 | } | |
2780 | } | |
b873b798 | 2781 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 | 2782 | |
4ac91242 CY |
2783 | kvfree(nm_i->nat_block_bitmap); |
2784 | kvfree(nm_i->free_nid_bitmap); | |
586d1492 | 2785 | kvfree(nm_i->free_nid_count); |
4ac91242 | 2786 | |
e05df3b1 | 2787 | kfree(nm_i->nat_bitmap); |
22ad0b6a | 2788 | kfree(nm_i->nat_bits); |
599a09b2 CY |
2789 | #ifdef CONFIG_F2FS_CHECK_FS |
2790 | kfree(nm_i->nat_bitmap_mir); | |
2791 | #endif | |
e05df3b1 JK |
2792 | sbi->nm_info = NULL; |
2793 | kfree(nm_i); | |
2794 | } | |
2795 | ||
6e6093a8 | 2796 | int __init create_node_manager_caches(void) |
e05df3b1 JK |
2797 | { |
2798 | nat_entry_slab = f2fs_kmem_cache_create("nat_entry", | |
e8512d2e | 2799 | sizeof(struct nat_entry)); |
e05df3b1 | 2800 | if (!nat_entry_slab) |
aec71382 | 2801 | goto fail; |
e05df3b1 JK |
2802 | |
2803 | free_nid_slab = f2fs_kmem_cache_create("free_nid", | |
e8512d2e | 2804 | sizeof(struct free_nid)); |
aec71382 | 2805 | if (!free_nid_slab) |
ce3e6d25 | 2806 | goto destroy_nat_entry; |
aec71382 CY |
2807 | |
2808 | nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set", | |
2809 | sizeof(struct nat_entry_set)); | |
2810 | if (!nat_entry_set_slab) | |
ce3e6d25 | 2811 | goto destroy_free_nid; |
e05df3b1 | 2812 | return 0; |
aec71382 | 2813 | |
ce3e6d25 | 2814 | destroy_free_nid: |
aec71382 | 2815 | kmem_cache_destroy(free_nid_slab); |
ce3e6d25 | 2816 | destroy_nat_entry: |
aec71382 CY |
2817 | kmem_cache_destroy(nat_entry_slab); |
2818 | fail: | |
2819 | return -ENOMEM; | |
e05df3b1 JK |
2820 | } |
2821 | ||
2822 | void destroy_node_manager_caches(void) | |
2823 | { | |
aec71382 | 2824 | kmem_cache_destroy(nat_entry_set_slab); |
e05df3b1 JK |
2825 | kmem_cache_destroy(free_nid_slab); |
2826 | kmem_cache_destroy(nat_entry_slab); | |
2827 | } |