]>
Commit | Line | Data |
---|---|---|
0a8165d7 | 1 | /* |
e05df3b1 JK |
2 | * fs/f2fs/node.c |
3 | * | |
4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
5 | * http://www.samsung.com/ | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
11 | #include <linux/fs.h> | |
12 | #include <linux/f2fs_fs.h> | |
13 | #include <linux/mpage.h> | |
14 | #include <linux/backing-dev.h> | |
15 | #include <linux/blkdev.h> | |
16 | #include <linux/pagevec.h> | |
17 | #include <linux/swap.h> | |
18 | ||
19 | #include "f2fs.h" | |
20 | #include "node.h" | |
21 | #include "segment.h" | |
9e4ded3f | 22 | #include "trace.h" |
51dd6249 | 23 | #include <trace/events/f2fs.h> |
e05df3b1 | 24 | |
f978f5a0 GZ |
25 | #define on_build_free_nids(nmi) mutex_is_locked(&nm_i->build_lock) |
26 | ||
e05df3b1 JK |
27 | static struct kmem_cache *nat_entry_slab; |
28 | static struct kmem_cache *free_nid_slab; | |
aec71382 | 29 | static struct kmem_cache *nat_entry_set_slab; |
e05df3b1 | 30 | |
6fb03f3a | 31 | bool available_free_memory(struct f2fs_sb_info *sbi, int type) |
cdfc41c1 | 32 | { |
6fb03f3a | 33 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
cdfc41c1 | 34 | struct sysinfo val; |
e5e7ea3c | 35 | unsigned long avail_ram; |
cdfc41c1 | 36 | unsigned long mem_size = 0; |
6fb03f3a | 37 | bool res = false; |
cdfc41c1 JK |
38 | |
39 | si_meminfo(&val); | |
e5e7ea3c JK |
40 | |
41 | /* only uses low memory */ | |
42 | avail_ram = val.totalram - val.totalhigh; | |
43 | ||
429511cd CY |
44 | /* |
45 | * give 25%, 25%, 50%, 50%, 50% memory for each components respectively | |
46 | */ | |
6fb03f3a | 47 | if (type == FREE_NIDS) { |
b8559dc2 CY |
48 | mem_size = (nm_i->nid_cnt[FREE_NID_LIST] * |
49 | sizeof(struct free_nid)) >> PAGE_SHIFT; | |
e5e7ea3c | 50 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); |
6fb03f3a | 51 | } else if (type == NAT_ENTRIES) { |
e5e7ea3c | 52 | mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >> |
09cbfeaf | 53 | PAGE_SHIFT; |
e5e7ea3c | 54 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); |
e589c2c4 JK |
55 | if (excess_cached_nats(sbi)) |
56 | res = false; | |
a1257023 JK |
57 | } else if (type == DIRTY_DENTS) { |
58 | if (sbi->sb->s_bdi->wb.dirty_exceeded) | |
59 | return false; | |
60 | mem_size = get_pages(sbi, F2FS_DIRTY_DENTS); | |
61 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); | |
e5e7ea3c JK |
62 | } else if (type == INO_ENTRIES) { |
63 | int i; | |
64 | ||
e5e7ea3c | 65 | for (i = 0; i <= UPDATE_INO; i++) |
67298804 | 66 | mem_size += (sbi->im[i].ino_num * |
09cbfeaf | 67 | sizeof(struct ino_entry)) >> PAGE_SHIFT; |
e5e7ea3c | 68 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
429511cd | 69 | } else if (type == EXTENT_CACHE) { |
7441ccef JK |
70 | mem_size = (atomic_read(&sbi->total_ext_tree) * |
71 | sizeof(struct extent_tree) + | |
429511cd | 72 | atomic_read(&sbi->total_ext_node) * |
09cbfeaf | 73 | sizeof(struct extent_node)) >> PAGE_SHIFT; |
429511cd | 74 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
1e84371f | 75 | } else { |
1663cae4 JK |
76 | if (!sbi->sb->s_bdi->wb.dirty_exceeded) |
77 | return true; | |
6fb03f3a JK |
78 | } |
79 | return res; | |
cdfc41c1 JK |
80 | } |
81 | ||
e05df3b1 JK |
82 | static void clear_node_page_dirty(struct page *page) |
83 | { | |
84 | struct address_space *mapping = page->mapping; | |
e05df3b1 JK |
85 | unsigned int long flags; |
86 | ||
87 | if (PageDirty(page)) { | |
88 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
89 | radix_tree_tag_clear(&mapping->page_tree, | |
90 | page_index(page), | |
91 | PAGECACHE_TAG_DIRTY); | |
92 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
93 | ||
94 | clear_page_dirty_for_io(page); | |
4081363f | 95 | dec_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_NODES); |
e05df3b1 JK |
96 | } |
97 | ClearPageUptodate(page); | |
98 | } | |
99 | ||
100 | static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
101 | { | |
102 | pgoff_t index = current_nat_addr(sbi, nid); | |
103 | return get_meta_page(sbi, index); | |
104 | } | |
105 | ||
106 | static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
107 | { | |
108 | struct page *src_page; | |
109 | struct page *dst_page; | |
110 | pgoff_t src_off; | |
111 | pgoff_t dst_off; | |
112 | void *src_addr; | |
113 | void *dst_addr; | |
114 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
115 | ||
116 | src_off = current_nat_addr(sbi, nid); | |
117 | dst_off = next_nat_addr(sbi, src_off); | |
118 | ||
119 | /* get current nat block page with lock */ | |
120 | src_page = get_meta_page(sbi, src_off); | |
e05df3b1 | 121 | dst_page = grab_meta_page(sbi, dst_off); |
9850cf4a | 122 | f2fs_bug_on(sbi, PageDirty(src_page)); |
e05df3b1 JK |
123 | |
124 | src_addr = page_address(src_page); | |
125 | dst_addr = page_address(dst_page); | |
09cbfeaf | 126 | memcpy(dst_addr, src_addr, PAGE_SIZE); |
e05df3b1 JK |
127 | set_page_dirty(dst_page); |
128 | f2fs_put_page(src_page, 1); | |
129 | ||
130 | set_to_next_nat(nm_i, nid); | |
131 | ||
132 | return dst_page; | |
133 | } | |
134 | ||
e05df3b1 JK |
135 | static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n) |
136 | { | |
137 | return radix_tree_lookup(&nm_i->nat_root, n); | |
138 | } | |
139 | ||
140 | static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i, | |
141 | nid_t start, unsigned int nr, struct nat_entry **ep) | |
142 | { | |
143 | return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr); | |
144 | } | |
145 | ||
146 | static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e) | |
147 | { | |
148 | list_del(&e->list); | |
149 | radix_tree_delete(&nm_i->nat_root, nat_get_nid(e)); | |
150 | nm_i->nat_cnt--; | |
151 | kmem_cache_free(nat_entry_slab, e); | |
152 | } | |
153 | ||
309cc2b6 JK |
154 | static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i, |
155 | struct nat_entry *ne) | |
156 | { | |
157 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); | |
158 | struct nat_entry_set *head; | |
159 | ||
160 | if (get_nat_flag(ne, IS_DIRTY)) | |
161 | return; | |
9be32d72 | 162 | |
309cc2b6 JK |
163 | head = radix_tree_lookup(&nm_i->nat_set_root, set); |
164 | if (!head) { | |
80c54505 | 165 | head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS); |
309cc2b6 JK |
166 | |
167 | INIT_LIST_HEAD(&head->entry_list); | |
168 | INIT_LIST_HEAD(&head->set_list); | |
169 | head->set = set; | |
170 | head->entry_cnt = 0; | |
9be32d72 | 171 | f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head); |
309cc2b6 JK |
172 | } |
173 | list_move_tail(&ne->list, &head->entry_list); | |
174 | nm_i->dirty_nat_cnt++; | |
175 | head->entry_cnt++; | |
176 | set_nat_flag(ne, IS_DIRTY, true); | |
177 | } | |
178 | ||
179 | static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i, | |
180 | struct nat_entry *ne) | |
181 | { | |
20d047c8 | 182 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); |
309cc2b6 JK |
183 | struct nat_entry_set *head; |
184 | ||
185 | head = radix_tree_lookup(&nm_i->nat_set_root, set); | |
186 | if (head) { | |
187 | list_move_tail(&ne->list, &nm_i->nat_entries); | |
188 | set_nat_flag(ne, IS_DIRTY, false); | |
189 | head->entry_cnt--; | |
190 | nm_i->dirty_nat_cnt--; | |
191 | } | |
192 | } | |
193 | ||
194 | static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i, | |
195 | nid_t start, unsigned int nr, struct nat_entry_set **ep) | |
196 | { | |
197 | return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep, | |
198 | start, nr); | |
199 | } | |
200 | ||
2dcf51ab | 201 | int need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 JK |
202 | { |
203 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
204 | struct nat_entry *e; | |
2dcf51ab | 205 | bool need = false; |
e05df3b1 | 206 | |
b873b798 | 207 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 | 208 | e = __lookup_nat_cache(nm_i, nid); |
2dcf51ab JK |
209 | if (e) { |
210 | if (!get_nat_flag(e, IS_CHECKPOINTED) && | |
211 | !get_nat_flag(e, HAS_FSYNCED_INODE)) | |
212 | need = true; | |
213 | } | |
b873b798 | 214 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 215 | return need; |
e05df3b1 JK |
216 | } |
217 | ||
2dcf51ab | 218 | bool is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid) |
479f40c4 JK |
219 | { |
220 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
221 | struct nat_entry *e; | |
2dcf51ab | 222 | bool is_cp = true; |
479f40c4 | 223 | |
b873b798 | 224 | down_read(&nm_i->nat_tree_lock); |
2dcf51ab JK |
225 | e = __lookup_nat_cache(nm_i, nid); |
226 | if (e && !get_nat_flag(e, IS_CHECKPOINTED)) | |
227 | is_cp = false; | |
b873b798 | 228 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 229 | return is_cp; |
479f40c4 JK |
230 | } |
231 | ||
88bd02c9 | 232 | bool need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino) |
b6fe5873 JK |
233 | { |
234 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
235 | struct nat_entry *e; | |
88bd02c9 | 236 | bool need_update = true; |
b6fe5873 | 237 | |
b873b798 | 238 | down_read(&nm_i->nat_tree_lock); |
88bd02c9 JK |
239 | e = __lookup_nat_cache(nm_i, ino); |
240 | if (e && get_nat_flag(e, HAS_LAST_FSYNC) && | |
241 | (get_nat_flag(e, IS_CHECKPOINTED) || | |
242 | get_nat_flag(e, HAS_FSYNCED_INODE))) | |
243 | need_update = false; | |
b873b798 | 244 | up_read(&nm_i->nat_tree_lock); |
88bd02c9 | 245 | return need_update; |
b6fe5873 JK |
246 | } |
247 | ||
5c9e4184 CY |
248 | static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid, |
249 | bool no_fail) | |
e05df3b1 JK |
250 | { |
251 | struct nat_entry *new; | |
252 | ||
5c9e4184 CY |
253 | if (no_fail) { |
254 | new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_NOFS); | |
255 | f2fs_radix_tree_insert(&nm_i->nat_root, nid, new); | |
256 | } else { | |
257 | new = kmem_cache_alloc(nat_entry_slab, GFP_NOFS); | |
258 | if (!new) | |
259 | return NULL; | |
260 | if (radix_tree_insert(&nm_i->nat_root, nid, new)) { | |
261 | kmem_cache_free(nat_entry_slab, new); | |
262 | return NULL; | |
263 | } | |
264 | } | |
265 | ||
e05df3b1 JK |
266 | memset(new, 0, sizeof(struct nat_entry)); |
267 | nat_set_nid(new, nid); | |
88bd02c9 | 268 | nat_reset_flag(new); |
e05df3b1 JK |
269 | list_add_tail(&new->list, &nm_i->nat_entries); |
270 | nm_i->nat_cnt++; | |
271 | return new; | |
272 | } | |
273 | ||
1515aef0 | 274 | static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid, |
e05df3b1 JK |
275 | struct f2fs_nat_entry *ne) |
276 | { | |
1515aef0 | 277 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 278 | struct nat_entry *e; |
9be32d72 | 279 | |
e05df3b1 JK |
280 | e = __lookup_nat_cache(nm_i, nid); |
281 | if (!e) { | |
5c9e4184 CY |
282 | e = grab_nat_entry(nm_i, nid, false); |
283 | if (e) | |
284 | node_info_from_raw_nat(&e->ni, ne); | |
1515aef0 | 285 | } else { |
0c0b471e EB |
286 | f2fs_bug_on(sbi, nat_get_ino(e) != le32_to_cpu(ne->ino) || |
287 | nat_get_blkaddr(e) != | |
288 | le32_to_cpu(ne->block_addr) || | |
1515aef0 | 289 | nat_get_version(e) != ne->version); |
e05df3b1 | 290 | } |
e05df3b1 JK |
291 | } |
292 | ||
293 | static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, | |
479f40c4 | 294 | block_t new_blkaddr, bool fsync_done) |
e05df3b1 JK |
295 | { |
296 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
297 | struct nat_entry *e; | |
9be32d72 | 298 | |
b873b798 | 299 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
300 | e = __lookup_nat_cache(nm_i, ni->nid); |
301 | if (!e) { | |
5c9e4184 | 302 | e = grab_nat_entry(nm_i, ni->nid, true); |
5c27f4ee | 303 | copy_node_info(&e->ni, ni); |
9850cf4a | 304 | f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR); |
e05df3b1 JK |
305 | } else if (new_blkaddr == NEW_ADDR) { |
306 | /* | |
307 | * when nid is reallocated, | |
308 | * previous nat entry can be remained in nat cache. | |
309 | * So, reinitialize it with new information. | |
310 | */ | |
5c27f4ee | 311 | copy_node_info(&e->ni, ni); |
9850cf4a | 312 | f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR); |
e05df3b1 JK |
313 | } |
314 | ||
e05df3b1 | 315 | /* sanity check */ |
9850cf4a JK |
316 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr); |
317 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR && | |
e05df3b1 | 318 | new_blkaddr == NULL_ADDR); |
9850cf4a | 319 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR && |
e05df3b1 | 320 | new_blkaddr == NEW_ADDR); |
9850cf4a | 321 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != NEW_ADDR && |
e05df3b1 JK |
322 | nat_get_blkaddr(e) != NULL_ADDR && |
323 | new_blkaddr == NEW_ADDR); | |
324 | ||
e1c42045 | 325 | /* increment version no as node is removed */ |
e05df3b1 JK |
326 | if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) { |
327 | unsigned char version = nat_get_version(e); | |
328 | nat_set_version(e, inc_node_version(version)); | |
26834466 JK |
329 | |
330 | /* in order to reuse the nid */ | |
331 | if (nm_i->next_scan_nid > ni->nid) | |
332 | nm_i->next_scan_nid = ni->nid; | |
e05df3b1 JK |
333 | } |
334 | ||
335 | /* change address */ | |
336 | nat_set_blkaddr(e, new_blkaddr); | |
88bd02c9 JK |
337 | if (new_blkaddr == NEW_ADDR || new_blkaddr == NULL_ADDR) |
338 | set_nat_flag(e, IS_CHECKPOINTED, false); | |
e05df3b1 | 339 | __set_nat_cache_dirty(nm_i, e); |
479f40c4 JK |
340 | |
341 | /* update fsync_mark if its inode nat entry is still alive */ | |
d5b692b7 CY |
342 | if (ni->nid != ni->ino) |
343 | e = __lookup_nat_cache(nm_i, ni->ino); | |
88bd02c9 JK |
344 | if (e) { |
345 | if (fsync_done && ni->nid == ni->ino) | |
346 | set_nat_flag(e, HAS_FSYNCED_INODE, true); | |
347 | set_nat_flag(e, HAS_LAST_FSYNC, fsync_done); | |
348 | } | |
b873b798 | 349 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
350 | } |
351 | ||
4660f9c0 | 352 | int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) |
e05df3b1 JK |
353 | { |
354 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1b38dc8e | 355 | int nr = nr_shrink; |
e05df3b1 | 356 | |
b873b798 JK |
357 | if (!down_write_trylock(&nm_i->nat_tree_lock)) |
358 | return 0; | |
e05df3b1 | 359 | |
e05df3b1 JK |
360 | while (nr_shrink && !list_empty(&nm_i->nat_entries)) { |
361 | struct nat_entry *ne; | |
362 | ne = list_first_entry(&nm_i->nat_entries, | |
363 | struct nat_entry, list); | |
364 | __del_from_nat_cache(nm_i, ne); | |
365 | nr_shrink--; | |
366 | } | |
b873b798 | 367 | up_write(&nm_i->nat_tree_lock); |
1b38dc8e | 368 | return nr - nr_shrink; |
e05df3b1 JK |
369 | } |
370 | ||
0a8165d7 | 371 | /* |
e1c42045 | 372 | * This function always returns success |
e05df3b1 JK |
373 | */ |
374 | void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni) | |
375 | { | |
376 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
377 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 378 | struct f2fs_journal *journal = curseg->journal; |
e05df3b1 JK |
379 | nid_t start_nid = START_NID(nid); |
380 | struct f2fs_nat_block *nat_blk; | |
381 | struct page *page = NULL; | |
382 | struct f2fs_nat_entry ne; | |
383 | struct nat_entry *e; | |
384 | int i; | |
385 | ||
386 | ni->nid = nid; | |
387 | ||
388 | /* Check nat cache */ | |
b873b798 | 389 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
390 | e = __lookup_nat_cache(nm_i, nid); |
391 | if (e) { | |
392 | ni->ino = nat_get_ino(e); | |
393 | ni->blk_addr = nat_get_blkaddr(e); | |
394 | ni->version = nat_get_version(e); | |
b873b798 | 395 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 | 396 | return; |
1515aef0 | 397 | } |
e05df3b1 | 398 | |
3547ea96 JK |
399 | memset(&ne, 0, sizeof(struct f2fs_nat_entry)); |
400 | ||
e05df3b1 | 401 | /* Check current segment summary */ |
b7ad7512 | 402 | down_read(&curseg->journal_rwsem); |
dfc08a12 | 403 | i = lookup_journal_in_cursum(journal, NAT_JOURNAL, nid, 0); |
e05df3b1 | 404 | if (i >= 0) { |
dfc08a12 | 405 | ne = nat_in_journal(journal, i); |
e05df3b1 JK |
406 | node_info_from_raw_nat(ni, &ne); |
407 | } | |
b7ad7512 | 408 | up_read(&curseg->journal_rwsem); |
e05df3b1 JK |
409 | if (i >= 0) |
410 | goto cache; | |
411 | ||
412 | /* Fill node_info from nat page */ | |
413 | page = get_current_nat_page(sbi, start_nid); | |
414 | nat_blk = (struct f2fs_nat_block *)page_address(page); | |
415 | ne = nat_blk->entries[nid - start_nid]; | |
416 | node_info_from_raw_nat(ni, &ne); | |
417 | f2fs_put_page(page, 1); | |
418 | cache: | |
b873b798 | 419 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 | 420 | /* cache nat entry */ |
b873b798 | 421 | down_write(&nm_i->nat_tree_lock); |
1515aef0 | 422 | cache_nat_entry(sbi, nid, &ne); |
b873b798 | 423 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
424 | } |
425 | ||
79344efb JK |
426 | /* |
427 | * readahead MAX_RA_NODE number of node pages. | |
428 | */ | |
429 | static void ra_node_pages(struct page *parent, int start, int n) | |
430 | { | |
431 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); | |
432 | struct blk_plug plug; | |
433 | int i, end; | |
434 | nid_t nid; | |
435 | ||
436 | blk_start_plug(&plug); | |
437 | ||
438 | /* Then, try readahead for siblings of the desired node */ | |
439 | end = start + n; | |
440 | end = min(end, NIDS_PER_BLOCK); | |
441 | for (i = start; i < end; i++) { | |
442 | nid = get_nid(parent, i, false); | |
443 | ra_node_page(sbi, nid); | |
444 | } | |
445 | ||
446 | blk_finish_plug(&plug); | |
447 | } | |
448 | ||
3cf45747 CY |
449 | pgoff_t get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs) |
450 | { | |
451 | const long direct_index = ADDRS_PER_INODE(dn->inode); | |
452 | const long direct_blks = ADDRS_PER_BLOCK; | |
453 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
454 | unsigned int skipped_unit = ADDRS_PER_BLOCK; | |
455 | int cur_level = dn->cur_level; | |
456 | int max_level = dn->max_level; | |
457 | pgoff_t base = 0; | |
458 | ||
459 | if (!dn->max_level) | |
460 | return pgofs + 1; | |
461 | ||
462 | while (max_level-- > cur_level) | |
463 | skipped_unit *= NIDS_PER_BLOCK; | |
464 | ||
465 | switch (dn->max_level) { | |
466 | case 3: | |
467 | base += 2 * indirect_blks; | |
468 | case 2: | |
469 | base += 2 * direct_blks; | |
470 | case 1: | |
471 | base += direct_index; | |
472 | break; | |
473 | default: | |
474 | f2fs_bug_on(F2FS_I_SB(dn->inode), 1); | |
475 | } | |
476 | ||
477 | return ((pgofs - base) / skipped_unit + 1) * skipped_unit + base; | |
478 | } | |
479 | ||
0a8165d7 | 480 | /* |
e05df3b1 JK |
481 | * The maximum depth is four. |
482 | * Offset[0] will have raw inode offset. | |
483 | */ | |
81ca7350 | 484 | static int get_node_path(struct inode *inode, long block, |
de93653f | 485 | int offset[4], unsigned int noffset[4]) |
e05df3b1 | 486 | { |
81ca7350 | 487 | const long direct_index = ADDRS_PER_INODE(inode); |
e05df3b1 JK |
488 | const long direct_blks = ADDRS_PER_BLOCK; |
489 | const long dptrs_per_blk = NIDS_PER_BLOCK; | |
490 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
491 | const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK; | |
492 | int n = 0; | |
493 | int level = 0; | |
494 | ||
495 | noffset[0] = 0; | |
496 | ||
497 | if (block < direct_index) { | |
25c0a6e5 | 498 | offset[n] = block; |
e05df3b1 JK |
499 | goto got; |
500 | } | |
501 | block -= direct_index; | |
502 | if (block < direct_blks) { | |
503 | offset[n++] = NODE_DIR1_BLOCK; | |
504 | noffset[n] = 1; | |
25c0a6e5 | 505 | offset[n] = block; |
e05df3b1 JK |
506 | level = 1; |
507 | goto got; | |
508 | } | |
509 | block -= direct_blks; | |
510 | if (block < direct_blks) { | |
511 | offset[n++] = NODE_DIR2_BLOCK; | |
512 | noffset[n] = 2; | |
25c0a6e5 | 513 | offset[n] = block; |
e05df3b1 JK |
514 | level = 1; |
515 | goto got; | |
516 | } | |
517 | block -= direct_blks; | |
518 | if (block < indirect_blks) { | |
519 | offset[n++] = NODE_IND1_BLOCK; | |
520 | noffset[n] = 3; | |
521 | offset[n++] = block / direct_blks; | |
522 | noffset[n] = 4 + offset[n - 1]; | |
25c0a6e5 | 523 | offset[n] = block % direct_blks; |
e05df3b1 JK |
524 | level = 2; |
525 | goto got; | |
526 | } | |
527 | block -= indirect_blks; | |
528 | if (block < indirect_blks) { | |
529 | offset[n++] = NODE_IND2_BLOCK; | |
530 | noffset[n] = 4 + dptrs_per_blk; | |
531 | offset[n++] = block / direct_blks; | |
532 | noffset[n] = 5 + dptrs_per_blk + offset[n - 1]; | |
25c0a6e5 | 533 | offset[n] = block % direct_blks; |
e05df3b1 JK |
534 | level = 2; |
535 | goto got; | |
536 | } | |
537 | block -= indirect_blks; | |
538 | if (block < dindirect_blks) { | |
539 | offset[n++] = NODE_DIND_BLOCK; | |
540 | noffset[n] = 5 + (dptrs_per_blk * 2); | |
541 | offset[n++] = block / indirect_blks; | |
542 | noffset[n] = 6 + (dptrs_per_blk * 2) + | |
543 | offset[n - 1] * (dptrs_per_blk + 1); | |
544 | offset[n++] = (block / direct_blks) % dptrs_per_blk; | |
545 | noffset[n] = 7 + (dptrs_per_blk * 2) + | |
546 | offset[n - 2] * (dptrs_per_blk + 1) + | |
547 | offset[n - 1]; | |
25c0a6e5 | 548 | offset[n] = block % direct_blks; |
e05df3b1 JK |
549 | level = 3; |
550 | goto got; | |
551 | } else { | |
552 | BUG(); | |
553 | } | |
554 | got: | |
555 | return level; | |
556 | } | |
557 | ||
558 | /* | |
559 | * Caller should call f2fs_put_dnode(dn). | |
4f4124d0 CY |
560 | * Also, it should grab and release a rwsem by calling f2fs_lock_op() and |
561 | * f2fs_unlock_op() only if ro is not set RDONLY_NODE. | |
39936837 | 562 | * In the case of RDONLY_NODE, we don't need to care about mutex. |
e05df3b1 | 563 | */ |
266e97a8 | 564 | int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode) |
e05df3b1 | 565 | { |
4081363f | 566 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 | 567 | struct page *npage[4]; |
f1a3b98e | 568 | struct page *parent = NULL; |
e05df3b1 JK |
569 | int offset[4]; |
570 | unsigned int noffset[4]; | |
571 | nid_t nids[4]; | |
3cf45747 | 572 | int level, i = 0; |
e05df3b1 JK |
573 | int err = 0; |
574 | ||
81ca7350 | 575 | level = get_node_path(dn->inode, index, offset, noffset); |
e05df3b1 JK |
576 | |
577 | nids[0] = dn->inode->i_ino; | |
1646cfac | 578 | npage[0] = dn->inode_page; |
e05df3b1 | 579 | |
1646cfac JK |
580 | if (!npage[0]) { |
581 | npage[0] = get_node_page(sbi, nids[0]); | |
582 | if (IS_ERR(npage[0])) | |
583 | return PTR_ERR(npage[0]); | |
584 | } | |
f1a3b98e JK |
585 | |
586 | /* if inline_data is set, should not report any block indices */ | |
587 | if (f2fs_has_inline_data(dn->inode) && index) { | |
76629165 | 588 | err = -ENOENT; |
f1a3b98e JK |
589 | f2fs_put_page(npage[0], 1); |
590 | goto release_out; | |
591 | } | |
592 | ||
e05df3b1 | 593 | parent = npage[0]; |
52c2db3f CL |
594 | if (level != 0) |
595 | nids[1] = get_nid(parent, offset[0], true); | |
e05df3b1 JK |
596 | dn->inode_page = npage[0]; |
597 | dn->inode_page_locked = true; | |
598 | ||
599 | /* get indirect or direct nodes */ | |
600 | for (i = 1; i <= level; i++) { | |
601 | bool done = false; | |
602 | ||
266e97a8 | 603 | if (!nids[i] && mode == ALLOC_NODE) { |
e05df3b1 JK |
604 | /* alloc new node */ |
605 | if (!alloc_nid(sbi, &(nids[i]))) { | |
e05df3b1 JK |
606 | err = -ENOSPC; |
607 | goto release_pages; | |
608 | } | |
609 | ||
610 | dn->nid = nids[i]; | |
8ae8f162 | 611 | npage[i] = new_node_page(dn, noffset[i], NULL); |
e05df3b1 JK |
612 | if (IS_ERR(npage[i])) { |
613 | alloc_nid_failed(sbi, nids[i]); | |
e05df3b1 JK |
614 | err = PTR_ERR(npage[i]); |
615 | goto release_pages; | |
616 | } | |
617 | ||
618 | set_nid(parent, offset[i - 1], nids[i], i == 1); | |
619 | alloc_nid_done(sbi, nids[i]); | |
e05df3b1 | 620 | done = true; |
266e97a8 | 621 | } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) { |
e05df3b1 JK |
622 | npage[i] = get_node_page_ra(parent, offset[i - 1]); |
623 | if (IS_ERR(npage[i])) { | |
624 | err = PTR_ERR(npage[i]); | |
625 | goto release_pages; | |
626 | } | |
627 | done = true; | |
628 | } | |
629 | if (i == 1) { | |
630 | dn->inode_page_locked = false; | |
631 | unlock_page(parent); | |
632 | } else { | |
633 | f2fs_put_page(parent, 1); | |
634 | } | |
635 | ||
636 | if (!done) { | |
637 | npage[i] = get_node_page(sbi, nids[i]); | |
638 | if (IS_ERR(npage[i])) { | |
639 | err = PTR_ERR(npage[i]); | |
640 | f2fs_put_page(npage[0], 0); | |
641 | goto release_out; | |
642 | } | |
643 | } | |
644 | if (i < level) { | |
645 | parent = npage[i]; | |
646 | nids[i + 1] = get_nid(parent, offset[i], false); | |
647 | } | |
648 | } | |
649 | dn->nid = nids[level]; | |
650 | dn->ofs_in_node = offset[level]; | |
651 | dn->node_page = npage[level]; | |
652 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
653 | return 0; | |
654 | ||
655 | release_pages: | |
656 | f2fs_put_page(parent, 1); | |
657 | if (i > 1) | |
658 | f2fs_put_page(npage[0], 0); | |
659 | release_out: | |
660 | dn->inode_page = NULL; | |
661 | dn->node_page = NULL; | |
3cf45747 CY |
662 | if (err == -ENOENT) { |
663 | dn->cur_level = i; | |
664 | dn->max_level = level; | |
0a2aa8fb | 665 | dn->ofs_in_node = offset[level]; |
3cf45747 | 666 | } |
e05df3b1 JK |
667 | return err; |
668 | } | |
669 | ||
670 | static void truncate_node(struct dnode_of_data *dn) | |
671 | { | |
4081363f | 672 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
673 | struct node_info ni; |
674 | ||
675 | get_node_info(sbi, dn->nid, &ni); | |
71e9fec5 | 676 | if (dn->inode->i_blocks == 0) { |
9850cf4a | 677 | f2fs_bug_on(sbi, ni.blk_addr != NULL_ADDR); |
71e9fec5 JK |
678 | goto invalidate; |
679 | } | |
9850cf4a | 680 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
e05df3b1 | 681 | |
e05df3b1 | 682 | /* Deallocate node address */ |
71e9fec5 | 683 | invalidate_blocks(sbi, ni.blk_addr); |
ef86d709 | 684 | dec_valid_node_count(sbi, dn->inode); |
479f40c4 | 685 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
e05df3b1 JK |
686 | |
687 | if (dn->nid == dn->inode->i_ino) { | |
688 | remove_orphan_inode(sbi, dn->nid); | |
689 | dec_valid_inode_count(sbi); | |
0f18b462 | 690 | f2fs_inode_synced(dn->inode); |
e05df3b1 | 691 | } |
71e9fec5 | 692 | invalidate: |
e05df3b1 | 693 | clear_node_page_dirty(dn->node_page); |
caf0047e | 694 | set_sbi_flag(sbi, SBI_IS_DIRTY); |
e05df3b1 JK |
695 | |
696 | f2fs_put_page(dn->node_page, 1); | |
bf39c00a JK |
697 | |
698 | invalidate_mapping_pages(NODE_MAPPING(sbi), | |
699 | dn->node_page->index, dn->node_page->index); | |
700 | ||
e05df3b1 | 701 | dn->node_page = NULL; |
51dd6249 | 702 | trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr); |
e05df3b1 JK |
703 | } |
704 | ||
705 | static int truncate_dnode(struct dnode_of_data *dn) | |
706 | { | |
e05df3b1 JK |
707 | struct page *page; |
708 | ||
709 | if (dn->nid == 0) | |
710 | return 1; | |
711 | ||
712 | /* get direct node */ | |
4081363f | 713 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
e05df3b1 JK |
714 | if (IS_ERR(page) && PTR_ERR(page) == -ENOENT) |
715 | return 1; | |
716 | else if (IS_ERR(page)) | |
717 | return PTR_ERR(page); | |
718 | ||
719 | /* Make dnode_of_data for parameter */ | |
720 | dn->node_page = page; | |
721 | dn->ofs_in_node = 0; | |
722 | truncate_data_blocks(dn); | |
723 | truncate_node(dn); | |
724 | return 1; | |
725 | } | |
726 | ||
727 | static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs, | |
728 | int ofs, int depth) | |
729 | { | |
e05df3b1 JK |
730 | struct dnode_of_data rdn = *dn; |
731 | struct page *page; | |
732 | struct f2fs_node *rn; | |
733 | nid_t child_nid; | |
734 | unsigned int child_nofs; | |
735 | int freed = 0; | |
736 | int i, ret; | |
737 | ||
738 | if (dn->nid == 0) | |
739 | return NIDS_PER_BLOCK + 1; | |
740 | ||
51dd6249 NJ |
741 | trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr); |
742 | ||
4081363f | 743 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
51dd6249 NJ |
744 | if (IS_ERR(page)) { |
745 | trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page)); | |
e05df3b1 | 746 | return PTR_ERR(page); |
51dd6249 | 747 | } |
e05df3b1 | 748 | |
79344efb JK |
749 | ra_node_pages(page, ofs, NIDS_PER_BLOCK); |
750 | ||
45590710 | 751 | rn = F2FS_NODE(page); |
e05df3b1 JK |
752 | if (depth < 3) { |
753 | for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) { | |
754 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
755 | if (child_nid == 0) | |
756 | continue; | |
757 | rdn.nid = child_nid; | |
758 | ret = truncate_dnode(&rdn); | |
759 | if (ret < 0) | |
760 | goto out_err; | |
12719ae1 JK |
761 | if (set_nid(page, i, 0, false)) |
762 | dn->node_changed = true; | |
e05df3b1 JK |
763 | } |
764 | } else { | |
765 | child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1; | |
766 | for (i = ofs; i < NIDS_PER_BLOCK; i++) { | |
767 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
768 | if (child_nid == 0) { | |
769 | child_nofs += NIDS_PER_BLOCK + 1; | |
770 | continue; | |
771 | } | |
772 | rdn.nid = child_nid; | |
773 | ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1); | |
774 | if (ret == (NIDS_PER_BLOCK + 1)) { | |
12719ae1 JK |
775 | if (set_nid(page, i, 0, false)) |
776 | dn->node_changed = true; | |
e05df3b1 JK |
777 | child_nofs += ret; |
778 | } else if (ret < 0 && ret != -ENOENT) { | |
779 | goto out_err; | |
780 | } | |
781 | } | |
782 | freed = child_nofs; | |
783 | } | |
784 | ||
785 | if (!ofs) { | |
786 | /* remove current indirect node */ | |
787 | dn->node_page = page; | |
788 | truncate_node(dn); | |
789 | freed++; | |
790 | } else { | |
791 | f2fs_put_page(page, 1); | |
792 | } | |
51dd6249 | 793 | trace_f2fs_truncate_nodes_exit(dn->inode, freed); |
e05df3b1 JK |
794 | return freed; |
795 | ||
796 | out_err: | |
797 | f2fs_put_page(page, 1); | |
51dd6249 | 798 | trace_f2fs_truncate_nodes_exit(dn->inode, ret); |
e05df3b1 JK |
799 | return ret; |
800 | } | |
801 | ||
802 | static int truncate_partial_nodes(struct dnode_of_data *dn, | |
803 | struct f2fs_inode *ri, int *offset, int depth) | |
804 | { | |
e05df3b1 JK |
805 | struct page *pages[2]; |
806 | nid_t nid[3]; | |
807 | nid_t child_nid; | |
808 | int err = 0; | |
809 | int i; | |
810 | int idx = depth - 2; | |
811 | ||
812 | nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); | |
813 | if (!nid[0]) | |
814 | return 0; | |
815 | ||
816 | /* get indirect nodes in the path */ | |
a225dca3 | 817 | for (i = 0; i < idx + 1; i++) { |
e1c42045 | 818 | /* reference count'll be increased */ |
4081363f | 819 | pages[i] = get_node_page(F2FS_I_SB(dn->inode), nid[i]); |
e05df3b1 | 820 | if (IS_ERR(pages[i])) { |
e05df3b1 | 821 | err = PTR_ERR(pages[i]); |
a225dca3 | 822 | idx = i - 1; |
e05df3b1 JK |
823 | goto fail; |
824 | } | |
825 | nid[i + 1] = get_nid(pages[i], offset[i + 1], false); | |
826 | } | |
827 | ||
79344efb JK |
828 | ra_node_pages(pages[idx], offset[idx + 1], NIDS_PER_BLOCK); |
829 | ||
e05df3b1 | 830 | /* free direct nodes linked to a partial indirect node */ |
a225dca3 | 831 | for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) { |
e05df3b1 JK |
832 | child_nid = get_nid(pages[idx], i, false); |
833 | if (!child_nid) | |
834 | continue; | |
835 | dn->nid = child_nid; | |
836 | err = truncate_dnode(dn); | |
837 | if (err < 0) | |
838 | goto fail; | |
12719ae1 JK |
839 | if (set_nid(pages[idx], i, 0, false)) |
840 | dn->node_changed = true; | |
e05df3b1 JK |
841 | } |
842 | ||
a225dca3 | 843 | if (offset[idx + 1] == 0) { |
e05df3b1 JK |
844 | dn->node_page = pages[idx]; |
845 | dn->nid = nid[idx]; | |
846 | truncate_node(dn); | |
847 | } else { | |
848 | f2fs_put_page(pages[idx], 1); | |
849 | } | |
850 | offset[idx]++; | |
a225dca3 | 851 | offset[idx + 1] = 0; |
852 | idx--; | |
e05df3b1 | 853 | fail: |
a225dca3 | 854 | for (i = idx; i >= 0; i--) |
e05df3b1 | 855 | f2fs_put_page(pages[i], 1); |
51dd6249 NJ |
856 | |
857 | trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err); | |
858 | ||
e05df3b1 JK |
859 | return err; |
860 | } | |
861 | ||
0a8165d7 | 862 | /* |
e05df3b1 JK |
863 | * All the block addresses of data and nodes should be nullified. |
864 | */ | |
865 | int truncate_inode_blocks(struct inode *inode, pgoff_t from) | |
866 | { | |
4081363f | 867 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
e05df3b1 JK |
868 | int err = 0, cont = 1; |
869 | int level, offset[4], noffset[4]; | |
7dd690c8 | 870 | unsigned int nofs = 0; |
58bfaf44 | 871 | struct f2fs_inode *ri; |
e05df3b1 JK |
872 | struct dnode_of_data dn; |
873 | struct page *page; | |
874 | ||
51dd6249 NJ |
875 | trace_f2fs_truncate_inode_blocks_enter(inode, from); |
876 | ||
81ca7350 | 877 | level = get_node_path(inode, from, offset, noffset); |
ff373558 | 878 | |
e05df3b1 | 879 | page = get_node_page(sbi, inode->i_ino); |
51dd6249 NJ |
880 | if (IS_ERR(page)) { |
881 | trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page)); | |
e05df3b1 | 882 | return PTR_ERR(page); |
51dd6249 | 883 | } |
e05df3b1 JK |
884 | |
885 | set_new_dnode(&dn, inode, page, NULL, 0); | |
886 | unlock_page(page); | |
887 | ||
58bfaf44 | 888 | ri = F2FS_INODE(page); |
e05df3b1 JK |
889 | switch (level) { |
890 | case 0: | |
891 | case 1: | |
892 | nofs = noffset[1]; | |
893 | break; | |
894 | case 2: | |
895 | nofs = noffset[1]; | |
896 | if (!offset[level - 1]) | |
897 | goto skip_partial; | |
58bfaf44 | 898 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
899 | if (err < 0 && err != -ENOENT) |
900 | goto fail; | |
901 | nofs += 1 + NIDS_PER_BLOCK; | |
902 | break; | |
903 | case 3: | |
904 | nofs = 5 + 2 * NIDS_PER_BLOCK; | |
905 | if (!offset[level - 1]) | |
906 | goto skip_partial; | |
58bfaf44 | 907 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
908 | if (err < 0 && err != -ENOENT) |
909 | goto fail; | |
910 | break; | |
911 | default: | |
912 | BUG(); | |
913 | } | |
914 | ||
915 | skip_partial: | |
916 | while (cont) { | |
58bfaf44 | 917 | dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); |
e05df3b1 JK |
918 | switch (offset[0]) { |
919 | case NODE_DIR1_BLOCK: | |
920 | case NODE_DIR2_BLOCK: | |
921 | err = truncate_dnode(&dn); | |
922 | break; | |
923 | ||
924 | case NODE_IND1_BLOCK: | |
925 | case NODE_IND2_BLOCK: | |
926 | err = truncate_nodes(&dn, nofs, offset[1], 2); | |
927 | break; | |
928 | ||
929 | case NODE_DIND_BLOCK: | |
930 | err = truncate_nodes(&dn, nofs, offset[1], 3); | |
931 | cont = 0; | |
932 | break; | |
933 | ||
934 | default: | |
935 | BUG(); | |
936 | } | |
937 | if (err < 0 && err != -ENOENT) | |
938 | goto fail; | |
939 | if (offset[1] == 0 && | |
58bfaf44 | 940 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) { |
e05df3b1 | 941 | lock_page(page); |
ff373558 | 942 | BUG_ON(page->mapping != NODE_MAPPING(sbi)); |
fec1d657 | 943 | f2fs_wait_on_page_writeback(page, NODE, true); |
58bfaf44 | 944 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0; |
e05df3b1 JK |
945 | set_page_dirty(page); |
946 | unlock_page(page); | |
947 | } | |
948 | offset[1] = 0; | |
949 | offset[0]++; | |
950 | nofs += err; | |
951 | } | |
952 | fail: | |
953 | f2fs_put_page(page, 0); | |
51dd6249 | 954 | trace_f2fs_truncate_inode_blocks_exit(inode, err); |
e05df3b1 JK |
955 | return err > 0 ? 0 : err; |
956 | } | |
957 | ||
4f16fb0f JK |
958 | int truncate_xattr_node(struct inode *inode, struct page *page) |
959 | { | |
4081363f | 960 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
4f16fb0f JK |
961 | nid_t nid = F2FS_I(inode)->i_xattr_nid; |
962 | struct dnode_of_data dn; | |
963 | struct page *npage; | |
964 | ||
965 | if (!nid) | |
966 | return 0; | |
967 | ||
968 | npage = get_node_page(sbi, nid); | |
969 | if (IS_ERR(npage)) | |
970 | return PTR_ERR(npage); | |
971 | ||
205b9822 | 972 | f2fs_i_xnid_write(inode, 0); |
65985d93 JK |
973 | |
974 | /* need to do checkpoint during fsync */ | |
975 | F2FS_I(inode)->xattr_ver = cur_cp_version(F2FS_CKPT(sbi)); | |
976 | ||
4f16fb0f JK |
977 | set_new_dnode(&dn, inode, page, npage, nid); |
978 | ||
979 | if (page) | |
01d2d1aa | 980 | dn.inode_page_locked = true; |
4f16fb0f JK |
981 | truncate_node(&dn); |
982 | return 0; | |
983 | } | |
984 | ||
39936837 | 985 | /* |
4f4124d0 CY |
986 | * Caller should grab and release a rwsem by calling f2fs_lock_op() and |
987 | * f2fs_unlock_op(). | |
39936837 | 988 | */ |
13ec7297 | 989 | int remove_inode_page(struct inode *inode) |
e05df3b1 | 990 | { |
e05df3b1 | 991 | struct dnode_of_data dn; |
13ec7297 | 992 | int err; |
e05df3b1 | 993 | |
c2e69583 | 994 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); |
13ec7297 CY |
995 | err = get_dnode_of_data(&dn, 0, LOOKUP_NODE); |
996 | if (err) | |
997 | return err; | |
e05df3b1 | 998 | |
13ec7297 CY |
999 | err = truncate_xattr_node(inode, dn.inode_page); |
1000 | if (err) { | |
c2e69583 | 1001 | f2fs_put_dnode(&dn); |
13ec7297 | 1002 | return err; |
e05df3b1 | 1003 | } |
c2e69583 JK |
1004 | |
1005 | /* remove potential inline_data blocks */ | |
1006 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
1007 | S_ISLNK(inode->i_mode)) | |
1008 | truncate_data_blocks_range(&dn, 1); | |
1009 | ||
e1c42045 | 1010 | /* 0 is possible, after f2fs_new_inode() has failed */ |
9850cf4a JK |
1011 | f2fs_bug_on(F2FS_I_SB(inode), |
1012 | inode->i_blocks != 0 && inode->i_blocks != 1); | |
c2e69583 JK |
1013 | |
1014 | /* will put inode & node pages */ | |
71e9fec5 | 1015 | truncate_node(&dn); |
13ec7297 | 1016 | return 0; |
e05df3b1 JK |
1017 | } |
1018 | ||
a014e037 | 1019 | struct page *new_inode_page(struct inode *inode) |
e05df3b1 | 1020 | { |
e05df3b1 JK |
1021 | struct dnode_of_data dn; |
1022 | ||
1023 | /* allocate inode page for new inode */ | |
1024 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); | |
44a83ff6 JK |
1025 | |
1026 | /* caller should f2fs_put_page(page, 1); */ | |
8ae8f162 | 1027 | return new_node_page(&dn, 0, NULL); |
e05df3b1 JK |
1028 | } |
1029 | ||
8ae8f162 JK |
1030 | struct page *new_node_page(struct dnode_of_data *dn, |
1031 | unsigned int ofs, struct page *ipage) | |
e05df3b1 | 1032 | { |
4081363f | 1033 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
1034 | struct node_info old_ni, new_ni; |
1035 | struct page *page; | |
1036 | int err; | |
1037 | ||
91942321 | 1038 | if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC))) |
e05df3b1 JK |
1039 | return ERR_PTR(-EPERM); |
1040 | ||
300e129c | 1041 | page = f2fs_grab_cache_page(NODE_MAPPING(sbi), dn->nid, false); |
e05df3b1 JK |
1042 | if (!page) |
1043 | return ERR_PTR(-ENOMEM); | |
1044 | ||
6bacf52f | 1045 | if (unlikely(!inc_valid_node_count(sbi, dn->inode))) { |
9c02740c JK |
1046 | err = -ENOSPC; |
1047 | goto fail; | |
1048 | } | |
e05df3b1 | 1049 | |
9c02740c | 1050 | get_node_info(sbi, dn->nid, &old_ni); |
e05df3b1 JK |
1051 | |
1052 | /* Reinitialize old_ni with new node page */ | |
9850cf4a | 1053 | f2fs_bug_on(sbi, old_ni.blk_addr != NULL_ADDR); |
e05df3b1 JK |
1054 | new_ni = old_ni; |
1055 | new_ni.ino = dn->inode->i_ino; | |
479f40c4 | 1056 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
9c02740c | 1057 | |
fec1d657 | 1058 | f2fs_wait_on_page_writeback(page, NODE, true); |
9c02740c | 1059 | fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true); |
398b1ac5 | 1060 | set_cold_node(dn->inode, page); |
237c0790 JK |
1061 | if (!PageUptodate(page)) |
1062 | SetPageUptodate(page); | |
12719ae1 JK |
1063 | if (set_page_dirty(page)) |
1064 | dn->node_changed = true; | |
e05df3b1 | 1065 | |
4bc8e9bc | 1066 | if (f2fs_has_xattr_block(ofs)) |
205b9822 | 1067 | f2fs_i_xnid_write(dn->inode, dn->nid); |
479bd73a | 1068 | |
e05df3b1 JK |
1069 | if (ofs == 0) |
1070 | inc_valid_inode_count(sbi); | |
e05df3b1 JK |
1071 | return page; |
1072 | ||
1073 | fail: | |
71e9fec5 | 1074 | clear_node_page_dirty(page); |
e05df3b1 JK |
1075 | f2fs_put_page(page, 1); |
1076 | return ERR_PTR(err); | |
1077 | } | |
1078 | ||
56ae674c JK |
1079 | /* |
1080 | * Caller should do after getting the following values. | |
1081 | * 0: f2fs_put_page(page, 0) | |
86531d6b | 1082 | * LOCKED_PAGE or error: f2fs_put_page(page, 1) |
56ae674c | 1083 | */ |
04d328de | 1084 | static int read_node_page(struct page *page, int op_flags) |
e05df3b1 | 1085 | { |
4081363f | 1086 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1087 | struct node_info ni; |
cf04e8eb | 1088 | struct f2fs_io_info fio = { |
05ca3632 | 1089 | .sbi = sbi, |
cf04e8eb | 1090 | .type = NODE, |
04d328de MC |
1091 | .op = REQ_OP_READ, |
1092 | .op_flags = op_flags, | |
05ca3632 | 1093 | .page = page, |
4375a336 | 1094 | .encrypted_page = NULL, |
cf04e8eb | 1095 | }; |
e05df3b1 | 1096 | |
3bdad3c7 JK |
1097 | if (PageUptodate(page)) |
1098 | return LOCKED_PAGE; | |
1099 | ||
e05df3b1 JK |
1100 | get_node_info(sbi, page->index, &ni); |
1101 | ||
6bacf52f | 1102 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
2bca1e23 | 1103 | ClearPageUptodate(page); |
e05df3b1 | 1104 | return -ENOENT; |
393ff91f JK |
1105 | } |
1106 | ||
7a9d7548 | 1107 | fio.new_blkaddr = fio.old_blkaddr = ni.blk_addr; |
05ca3632 | 1108 | return f2fs_submit_page_bio(&fio); |
e05df3b1 JK |
1109 | } |
1110 | ||
0a8165d7 | 1111 | /* |
e05df3b1 JK |
1112 | * Readahead a node page |
1113 | */ | |
1114 | void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid) | |
1115 | { | |
e05df3b1 | 1116 | struct page *apage; |
56ae674c | 1117 | int err; |
e05df3b1 | 1118 | |
e8458725 CY |
1119 | if (!nid) |
1120 | return; | |
1121 | f2fs_bug_on(sbi, check_nid_range(sbi, nid)); | |
1122 | ||
999270de FL |
1123 | rcu_read_lock(); |
1124 | apage = radix_tree_lookup(&NODE_MAPPING(sbi)->page_tree, nid); | |
1125 | rcu_read_unlock(); | |
1126 | if (apage) | |
393ff91f | 1127 | return; |
e05df3b1 | 1128 | |
300e129c | 1129 | apage = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false); |
e05df3b1 JK |
1130 | if (!apage) |
1131 | return; | |
1132 | ||
70246286 | 1133 | err = read_node_page(apage, REQ_RAHEAD); |
86531d6b | 1134 | f2fs_put_page(apage, err ? 1 : 0); |
e05df3b1 JK |
1135 | } |
1136 | ||
17a0ee55 | 1137 | static struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid, |
0e022ea8 | 1138 | struct page *parent, int start) |
e05df3b1 | 1139 | { |
e05df3b1 | 1140 | struct page *page; |
0e022ea8 | 1141 | int err; |
e05df3b1 | 1142 | |
e05df3b1 JK |
1143 | if (!nid) |
1144 | return ERR_PTR(-ENOENT); | |
e8458725 | 1145 | f2fs_bug_on(sbi, check_nid_range(sbi, nid)); |
afcb7ca0 | 1146 | repeat: |
300e129c | 1147 | page = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false); |
e05df3b1 JK |
1148 | if (!page) |
1149 | return ERR_PTR(-ENOMEM); | |
1150 | ||
70fd7614 | 1151 | err = read_node_page(page, 0); |
86531d6b JK |
1152 | if (err < 0) { |
1153 | f2fs_put_page(page, 1); | |
e05df3b1 | 1154 | return ERR_PTR(err); |
86531d6b | 1155 | } else if (err == LOCKED_PAGE) { |
56ae674c | 1156 | goto page_hit; |
86531d6b | 1157 | } |
e05df3b1 | 1158 | |
0e022ea8 | 1159 | if (parent) |
79344efb | 1160 | ra_node_pages(parent, start + 1, MAX_RA_NODE); |
c718379b | 1161 | |
e05df3b1 | 1162 | lock_page(page); |
0e022ea8 | 1163 | |
4ef51a8f | 1164 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1165 | f2fs_put_page(page, 1); |
1166 | goto repeat; | |
1167 | } | |
1563ac75 CY |
1168 | |
1169 | if (unlikely(!PageUptodate(page))) | |
1170 | goto out_err; | |
e0f56cb4 | 1171 | page_hit: |
0c9df7fb YS |
1172 | if(unlikely(nid != nid_of_node(page))) { |
1173 | f2fs_bug_on(sbi, 1); | |
1174 | ClearPageUptodate(page); | |
1175 | out_err: | |
1176 | f2fs_put_page(page, 1); | |
1177 | return ERR_PTR(-EIO); | |
1178 | } | |
e05df3b1 JK |
1179 | return page; |
1180 | } | |
1181 | ||
0e022ea8 CY |
1182 | struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) |
1183 | { | |
1184 | return __get_node_page(sbi, nid, NULL, 0); | |
1185 | } | |
1186 | ||
1187 | struct page *get_node_page_ra(struct page *parent, int start) | |
1188 | { | |
1189 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); | |
1190 | nid_t nid = get_nid(parent, start, false); | |
1191 | ||
1192 | return __get_node_page(sbi, nid, parent, start); | |
1193 | } | |
1194 | ||
2049d4fc JK |
1195 | static void flush_inline_data(struct f2fs_sb_info *sbi, nid_t ino) |
1196 | { | |
1197 | struct inode *inode; | |
1198 | struct page *page; | |
0f3311a8 | 1199 | int ret; |
2049d4fc JK |
1200 | |
1201 | /* should flush inline_data before evict_inode */ | |
1202 | inode = ilookup(sbi->sb, ino); | |
1203 | if (!inode) | |
1204 | return; | |
1205 | ||
4a6de50d | 1206 | page = pagecache_get_page(inode->i_mapping, 0, FGP_LOCK|FGP_NOWAIT, 0); |
2049d4fc JK |
1207 | if (!page) |
1208 | goto iput_out; | |
1209 | ||
1210 | if (!PageUptodate(page)) | |
1211 | goto page_out; | |
1212 | ||
1213 | if (!PageDirty(page)) | |
1214 | goto page_out; | |
1215 | ||
1216 | if (!clear_page_dirty_for_io(page)) | |
1217 | goto page_out; | |
1218 | ||
0f3311a8 CY |
1219 | ret = f2fs_write_inline_data(inode, page); |
1220 | inode_dec_dirty_pages(inode); | |
933439c8 | 1221 | remove_dirty_inode(inode); |
0f3311a8 | 1222 | if (ret) |
2049d4fc JK |
1223 | set_page_dirty(page); |
1224 | page_out: | |
4a6de50d | 1225 | f2fs_put_page(page, 1); |
2049d4fc JK |
1226 | iput_out: |
1227 | iput(inode); | |
1228 | } | |
1229 | ||
da011cc0 CY |
1230 | void move_node_page(struct page *node_page, int gc_type) |
1231 | { | |
1232 | if (gc_type == FG_GC) { | |
1233 | struct f2fs_sb_info *sbi = F2FS_P_SB(node_page); | |
1234 | struct writeback_control wbc = { | |
1235 | .sync_mode = WB_SYNC_ALL, | |
1236 | .nr_to_write = 1, | |
1237 | .for_reclaim = 0, | |
1238 | }; | |
1239 | ||
1240 | set_page_dirty(node_page); | |
1241 | f2fs_wait_on_page_writeback(node_page, NODE, true); | |
1242 | ||
1243 | f2fs_bug_on(sbi, PageWriteback(node_page)); | |
1244 | if (!clear_page_dirty_for_io(node_page)) | |
1245 | goto out_page; | |
1246 | ||
1247 | if (NODE_MAPPING(sbi)->a_ops->writepage(node_page, &wbc)) | |
1248 | unlock_page(node_page); | |
1249 | goto release_page; | |
1250 | } else { | |
1251 | /* set page dirty and write it */ | |
1252 | if (!PageWriteback(node_page)) | |
1253 | set_page_dirty(node_page); | |
1254 | } | |
1255 | out_page: | |
1256 | unlock_page(node_page); | |
1257 | release_page: | |
1258 | f2fs_put_page(node_page, 0); | |
1259 | } | |
1260 | ||
608514de JK |
1261 | static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino) |
1262 | { | |
1263 | pgoff_t index, end; | |
1264 | struct pagevec pvec; | |
1265 | struct page *last_page = NULL; | |
1266 | ||
1267 | pagevec_init(&pvec, 0); | |
1268 | index = 0; | |
1269 | end = ULONG_MAX; | |
1270 | ||
1271 | while (index <= end) { | |
1272 | int i, nr_pages; | |
1273 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1274 | PAGECACHE_TAG_DIRTY, | |
1275 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1276 | if (nr_pages == 0) | |
1277 | break; | |
1278 | ||
1279 | for (i = 0; i < nr_pages; i++) { | |
1280 | struct page *page = pvec.pages[i]; | |
1281 | ||
1282 | if (unlikely(f2fs_cp_error(sbi))) { | |
1283 | f2fs_put_page(last_page, 0); | |
1284 | pagevec_release(&pvec); | |
1285 | return ERR_PTR(-EIO); | |
1286 | } | |
1287 | ||
1288 | if (!IS_DNODE(page) || !is_cold_node(page)) | |
1289 | continue; | |
1290 | if (ino_of_node(page) != ino) | |
1291 | continue; | |
1292 | ||
1293 | lock_page(page); | |
1294 | ||
1295 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { | |
1296 | continue_unlock: | |
1297 | unlock_page(page); | |
1298 | continue; | |
1299 | } | |
1300 | if (ino_of_node(page) != ino) | |
1301 | goto continue_unlock; | |
1302 | ||
1303 | if (!PageDirty(page)) { | |
1304 | /* someone wrote it for us */ | |
1305 | goto continue_unlock; | |
1306 | } | |
1307 | ||
1308 | if (last_page) | |
1309 | f2fs_put_page(last_page, 0); | |
1310 | ||
1311 | get_page(page); | |
1312 | last_page = page; | |
1313 | unlock_page(page); | |
1314 | } | |
1315 | pagevec_release(&pvec); | |
1316 | cond_resched(); | |
1317 | } | |
1318 | return last_page; | |
1319 | } | |
1320 | ||
d68f735b | 1321 | static int __write_node_page(struct page *page, bool atomic, bool *submitted, |
faa24895 JK |
1322 | struct writeback_control *wbc) |
1323 | { | |
1324 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); | |
1325 | nid_t nid; | |
1326 | struct node_info ni; | |
1327 | struct f2fs_io_info fio = { | |
1328 | .sbi = sbi, | |
1329 | .type = NODE, | |
1330 | .op = REQ_OP_WRITE, | |
1331 | .op_flags = wbc_to_write_flags(wbc), | |
1332 | .page = page, | |
1333 | .encrypted_page = NULL, | |
d68f735b | 1334 | .submitted = false, |
faa24895 JK |
1335 | }; |
1336 | ||
1337 | trace_f2fs_writepage(page, NODE); | |
1338 | ||
1339 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) | |
1340 | goto redirty_out; | |
1341 | if (unlikely(f2fs_cp_error(sbi))) | |
1342 | goto redirty_out; | |
1343 | ||
1344 | /* get old block addr of this node page */ | |
1345 | nid = nid_of_node(page); | |
1346 | f2fs_bug_on(sbi, page->index != nid); | |
1347 | ||
1348 | if (wbc->for_reclaim) { | |
1349 | if (!down_read_trylock(&sbi->node_write)) | |
1350 | goto redirty_out; | |
1351 | } else { | |
1352 | down_read(&sbi->node_write); | |
1353 | } | |
1354 | ||
1355 | get_node_info(sbi, nid, &ni); | |
1356 | ||
1357 | /* This page is already truncated */ | |
1358 | if (unlikely(ni.blk_addr == NULL_ADDR)) { | |
1359 | ClearPageUptodate(page); | |
1360 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1361 | up_read(&sbi->node_write); | |
1362 | unlock_page(page); | |
1363 | return 0; | |
1364 | } | |
1365 | ||
e7c75ab0 JK |
1366 | if (atomic && !test_opt(sbi, NOBARRIER)) |
1367 | fio.op_flags |= REQ_PREFLUSH | REQ_FUA; | |
1368 | ||
faa24895 JK |
1369 | set_page_writeback(page); |
1370 | fio.old_blkaddr = ni.blk_addr; | |
1371 | write_node_page(nid, &fio); | |
1372 | set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page)); | |
1373 | dec_page_count(sbi, F2FS_DIRTY_NODES); | |
1374 | up_read(&sbi->node_write); | |
1375 | ||
d68f735b | 1376 | if (wbc->for_reclaim) { |
942fd319 JK |
1377 | f2fs_submit_merged_bio_cond(sbi, page->mapping->host, 0, |
1378 | page->index, NODE, WRITE); | |
d68f735b JK |
1379 | submitted = NULL; |
1380 | } | |
faa24895 JK |
1381 | |
1382 | unlock_page(page); | |
1383 | ||
d68f735b | 1384 | if (unlikely(f2fs_cp_error(sbi))) { |
faa24895 | 1385 | f2fs_submit_merged_bio(sbi, NODE, WRITE); |
d68f735b JK |
1386 | submitted = NULL; |
1387 | } | |
1388 | if (submitted) | |
1389 | *submitted = fio.submitted; | |
faa24895 JK |
1390 | |
1391 | return 0; | |
1392 | ||
1393 | redirty_out: | |
1394 | redirty_page_for_writepage(wbc, page); | |
1395 | return AOP_WRITEPAGE_ACTIVATE; | |
1396 | } | |
1397 | ||
1398 | static int f2fs_write_node_page(struct page *page, | |
1399 | struct writeback_control *wbc) | |
1400 | { | |
d68f735b | 1401 | return __write_node_page(page, false, NULL, wbc); |
faa24895 JK |
1402 | } |
1403 | ||
26de9b11 | 1404 | int fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode, |
608514de | 1405 | struct writeback_control *wbc, bool atomic) |
e05df3b1 | 1406 | { |
e05df3b1 | 1407 | pgoff_t index, end; |
942fd319 | 1408 | pgoff_t last_idx = ULONG_MAX; |
e05df3b1 | 1409 | struct pagevec pvec; |
c267ec15 | 1410 | int ret = 0; |
608514de JK |
1411 | struct page *last_page = NULL; |
1412 | bool marked = false; | |
26de9b11 | 1413 | nid_t ino = inode->i_ino; |
52681375 | 1414 | |
608514de JK |
1415 | if (atomic) { |
1416 | last_page = last_fsync_dnode(sbi, ino); | |
1417 | if (IS_ERR_OR_NULL(last_page)) | |
1418 | return PTR_ERR_OR_ZERO(last_page); | |
1419 | } | |
1420 | retry: | |
52681375 JK |
1421 | pagevec_init(&pvec, 0); |
1422 | index = 0; | |
1423 | end = ULONG_MAX; | |
1424 | ||
1425 | while (index <= end) { | |
1426 | int i, nr_pages; | |
1427 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1428 | PAGECACHE_TAG_DIRTY, | |
1429 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1430 | if (nr_pages == 0) | |
1431 | break; | |
1432 | ||
1433 | for (i = 0; i < nr_pages; i++) { | |
1434 | struct page *page = pvec.pages[i]; | |
d68f735b | 1435 | bool submitted = false; |
52681375 JK |
1436 | |
1437 | if (unlikely(f2fs_cp_error(sbi))) { | |
608514de | 1438 | f2fs_put_page(last_page, 0); |
52681375 | 1439 | pagevec_release(&pvec); |
9de69279 CY |
1440 | ret = -EIO; |
1441 | goto out; | |
52681375 JK |
1442 | } |
1443 | ||
1444 | if (!IS_DNODE(page) || !is_cold_node(page)) | |
1445 | continue; | |
1446 | if (ino_of_node(page) != ino) | |
1447 | continue; | |
1448 | ||
1449 | lock_page(page); | |
1450 | ||
1451 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { | |
1452 | continue_unlock: | |
1453 | unlock_page(page); | |
1454 | continue; | |
1455 | } | |
1456 | if (ino_of_node(page) != ino) | |
1457 | goto continue_unlock; | |
1458 | ||
608514de | 1459 | if (!PageDirty(page) && page != last_page) { |
52681375 JK |
1460 | /* someone wrote it for us */ |
1461 | goto continue_unlock; | |
1462 | } | |
1463 | ||
1464 | f2fs_wait_on_page_writeback(page, NODE, true); | |
1465 | BUG_ON(PageWriteback(page)); | |
52681375 | 1466 | |
608514de JK |
1467 | if (!atomic || page == last_page) { |
1468 | set_fsync_mark(page, 1); | |
26de9b11 JK |
1469 | if (IS_INODE(page)) { |
1470 | if (is_inode_flag_set(inode, | |
1471 | FI_DIRTY_INODE)) | |
1472 | update_inode(inode, page); | |
608514de | 1473 | set_dentry_mark(page, |
52681375 | 1474 | need_dentry_mark(sbi, ino)); |
26de9b11 | 1475 | } |
608514de JK |
1476 | /* may be written by other thread */ |
1477 | if (!PageDirty(page)) | |
1478 | set_page_dirty(page); | |
1479 | } | |
1480 | ||
1481 | if (!clear_page_dirty_for_io(page)) | |
1482 | goto continue_unlock; | |
52681375 | 1483 | |
e7c75ab0 | 1484 | ret = __write_node_page(page, atomic && |
d68f735b JK |
1485 | page == last_page, |
1486 | &submitted, wbc); | |
c267ec15 | 1487 | if (ret) { |
52681375 | 1488 | unlock_page(page); |
608514de JK |
1489 | f2fs_put_page(last_page, 0); |
1490 | break; | |
d68f735b | 1491 | } else if (submitted) { |
942fd319 | 1492 | last_idx = page->index; |
608514de | 1493 | } |
3f5f4959 | 1494 | |
608514de JK |
1495 | if (page == last_page) { |
1496 | f2fs_put_page(page, 0); | |
1497 | marked = true; | |
52681375 | 1498 | break; |
c267ec15 | 1499 | } |
52681375 JK |
1500 | } |
1501 | pagevec_release(&pvec); | |
1502 | cond_resched(); | |
1503 | ||
608514de | 1504 | if (ret || marked) |
52681375 JK |
1505 | break; |
1506 | } | |
608514de JK |
1507 | if (!ret && atomic && !marked) { |
1508 | f2fs_msg(sbi->sb, KERN_DEBUG, | |
1509 | "Retry to write fsync mark: ino=%u, idx=%lx", | |
1510 | ino, last_page->index); | |
1511 | lock_page(last_page); | |
d40a43af | 1512 | f2fs_wait_on_page_writeback(last_page, NODE, true); |
608514de JK |
1513 | set_page_dirty(last_page); |
1514 | unlock_page(last_page); | |
1515 | goto retry; | |
1516 | } | |
9de69279 | 1517 | out: |
942fd319 JK |
1518 | if (last_idx != ULONG_MAX) |
1519 | f2fs_submit_merged_bio_cond(sbi, NULL, ino, last_idx, | |
1520 | NODE, WRITE); | |
c267ec15 | 1521 | return ret ? -EIO: 0; |
52681375 JK |
1522 | } |
1523 | ||
1524 | int sync_node_pages(struct f2fs_sb_info *sbi, struct writeback_control *wbc) | |
1525 | { | |
1526 | pgoff_t index, end; | |
1527 | struct pagevec pvec; | |
1528 | int step = 0; | |
12bb0a8f | 1529 | int nwritten = 0; |
3f5f4959 | 1530 | int ret = 0; |
e05df3b1 JK |
1531 | |
1532 | pagevec_init(&pvec, 0); | |
1533 | ||
1534 | next_step: | |
1535 | index = 0; | |
80dd9c0e | 1536 | end = ULONG_MAX; |
e05df3b1 JK |
1537 | |
1538 | while (index <= end) { | |
1539 | int i, nr_pages; | |
4ef51a8f | 1540 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, |
e05df3b1 JK |
1541 | PAGECACHE_TAG_DIRTY, |
1542 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1543 | if (nr_pages == 0) | |
1544 | break; | |
1545 | ||
1546 | for (i = 0; i < nr_pages; i++) { | |
1547 | struct page *page = pvec.pages[i]; | |
d68f735b | 1548 | bool submitted = false; |
e05df3b1 | 1549 | |
6d5a1495 CY |
1550 | if (unlikely(f2fs_cp_error(sbi))) { |
1551 | pagevec_release(&pvec); | |
3f5f4959 CY |
1552 | ret = -EIO; |
1553 | goto out; | |
6d5a1495 CY |
1554 | } |
1555 | ||
e05df3b1 JK |
1556 | /* |
1557 | * flushing sequence with step: | |
1558 | * 0. indirect nodes | |
1559 | * 1. dentry dnodes | |
1560 | * 2. file dnodes | |
1561 | */ | |
1562 | if (step == 0 && IS_DNODE(page)) | |
1563 | continue; | |
1564 | if (step == 1 && (!IS_DNODE(page) || | |
1565 | is_cold_node(page))) | |
1566 | continue; | |
1567 | if (step == 2 && (!IS_DNODE(page) || | |
1568 | !is_cold_node(page))) | |
1569 | continue; | |
9a4cbc9e | 1570 | lock_node: |
52681375 | 1571 | if (!trylock_page(page)) |
e05df3b1 JK |
1572 | continue; |
1573 | ||
4ef51a8f | 1574 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
e05df3b1 JK |
1575 | continue_unlock: |
1576 | unlock_page(page); | |
1577 | continue; | |
1578 | } | |
e05df3b1 JK |
1579 | |
1580 | if (!PageDirty(page)) { | |
1581 | /* someone wrote it for us */ | |
1582 | goto continue_unlock; | |
1583 | } | |
1584 | ||
2049d4fc | 1585 | /* flush inline_data */ |
52681375 | 1586 | if (is_inline_node(page)) { |
2049d4fc JK |
1587 | clear_inline_node(page); |
1588 | unlock_page(page); | |
1589 | flush_inline_data(sbi, ino_of_node(page)); | |
9a4cbc9e | 1590 | goto lock_node; |
2049d4fc JK |
1591 | } |
1592 | ||
fa3d2bdf JK |
1593 | f2fs_wait_on_page_writeback(page, NODE, true); |
1594 | ||
1595 | BUG_ON(PageWriteback(page)); | |
e05df3b1 JK |
1596 | if (!clear_page_dirty_for_io(page)) |
1597 | goto continue_unlock; | |
1598 | ||
52681375 JK |
1599 | set_fsync_mark(page, 0); |
1600 | set_dentry_mark(page, 0); | |
52746519 | 1601 | |
d68f735b JK |
1602 | ret = __write_node_page(page, false, &submitted, wbc); |
1603 | if (ret) | |
52746519 | 1604 | unlock_page(page); |
d68f735b | 1605 | else if (submitted) |
3f5f4959 | 1606 | nwritten++; |
e05df3b1 JK |
1607 | |
1608 | if (--wbc->nr_to_write == 0) | |
1609 | break; | |
1610 | } | |
1611 | pagevec_release(&pvec); | |
1612 | cond_resched(); | |
1613 | ||
1614 | if (wbc->nr_to_write == 0) { | |
1615 | step = 2; | |
1616 | break; | |
1617 | } | |
1618 | } | |
1619 | ||
1620 | if (step < 2) { | |
1621 | step++; | |
1622 | goto next_step; | |
1623 | } | |
3f5f4959 CY |
1624 | out: |
1625 | if (nwritten) | |
1626 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1627 | return ret; | |
e05df3b1 JK |
1628 | } |
1629 | ||
cfe58f9d JK |
1630 | int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino) |
1631 | { | |
80dd9c0e | 1632 | pgoff_t index = 0, end = ULONG_MAX; |
cfe58f9d | 1633 | struct pagevec pvec; |
280db3c8 | 1634 | int ret2, ret = 0; |
cfe58f9d JK |
1635 | |
1636 | pagevec_init(&pvec, 0); | |
4ef51a8f JK |
1637 | |
1638 | while (index <= end) { | |
1639 | int i, nr_pages; | |
1640 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1641 | PAGECACHE_TAG_WRITEBACK, | |
1642 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1643 | if (nr_pages == 0) | |
1644 | break; | |
cfe58f9d JK |
1645 | |
1646 | for (i = 0; i < nr_pages; i++) { | |
1647 | struct page *page = pvec.pages[i]; | |
1648 | ||
1649 | /* until radix tree lookup accepts end_index */ | |
cfb271d4 | 1650 | if (unlikely(page->index > end)) |
cfe58f9d JK |
1651 | continue; |
1652 | ||
4bf08ff6 | 1653 | if (ino && ino_of_node(page) == ino) { |
fec1d657 | 1654 | f2fs_wait_on_page_writeback(page, NODE, true); |
4bf08ff6 CY |
1655 | if (TestClearPageError(page)) |
1656 | ret = -EIO; | |
1657 | } | |
cfe58f9d JK |
1658 | } |
1659 | pagevec_release(&pvec); | |
1660 | cond_resched(); | |
1661 | } | |
1662 | ||
280db3c8 | 1663 | ret2 = filemap_check_errors(NODE_MAPPING(sbi)); |
cfe58f9d JK |
1664 | if (!ret) |
1665 | ret = ret2; | |
1666 | return ret; | |
1667 | } | |
1668 | ||
e05df3b1 JK |
1669 | static int f2fs_write_node_pages(struct address_space *mapping, |
1670 | struct writeback_control *wbc) | |
1671 | { | |
4081363f | 1672 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
9dfa1baf | 1673 | struct blk_plug plug; |
50c8cdb3 | 1674 | long diff; |
e05df3b1 | 1675 | |
4660f9c0 JK |
1676 | /* balancing f2fs's metadata in background */ |
1677 | f2fs_balance_fs_bg(sbi); | |
e05df3b1 | 1678 | |
a7fdffbd | 1679 | /* collect a number of dirty node pages and write together */ |
87d6f890 | 1680 | if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE)) |
d3baf95d | 1681 | goto skip_write; |
a7fdffbd | 1682 | |
d31c7c3f YH |
1683 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
1684 | ||
50c8cdb3 | 1685 | diff = nr_pages_to_write(sbi, NODE, wbc); |
fb5566da | 1686 | wbc->sync_mode = WB_SYNC_NONE; |
9dfa1baf | 1687 | blk_start_plug(&plug); |
52681375 | 1688 | sync_node_pages(sbi, wbc); |
9dfa1baf | 1689 | blk_finish_plug(&plug); |
50c8cdb3 | 1690 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
e05df3b1 | 1691 | return 0; |
d3baf95d JK |
1692 | |
1693 | skip_write: | |
1694 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES); | |
d31c7c3f | 1695 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
d3baf95d | 1696 | return 0; |
e05df3b1 JK |
1697 | } |
1698 | ||
1699 | static int f2fs_set_node_page_dirty(struct page *page) | |
1700 | { | |
26c6b887 JK |
1701 | trace_f2fs_set_page_dirty(page, NODE); |
1702 | ||
237c0790 JK |
1703 | if (!PageUptodate(page)) |
1704 | SetPageUptodate(page); | |
e05df3b1 | 1705 | if (!PageDirty(page)) { |
fe76b796 | 1706 | f2fs_set_page_dirty_nobuffers(page); |
4081363f | 1707 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 | 1708 | SetPagePrivate(page); |
9e4ded3f | 1709 | f2fs_trace_pid(page); |
e05df3b1 JK |
1710 | return 1; |
1711 | } | |
1712 | return 0; | |
1713 | } | |
1714 | ||
0a8165d7 | 1715 | /* |
e05df3b1 JK |
1716 | * Structure of the f2fs node operations |
1717 | */ | |
1718 | const struct address_space_operations f2fs_node_aops = { | |
1719 | .writepage = f2fs_write_node_page, | |
1720 | .writepages = f2fs_write_node_pages, | |
1721 | .set_page_dirty = f2fs_set_node_page_dirty, | |
487261f3 CY |
1722 | .invalidatepage = f2fs_invalidate_page, |
1723 | .releasepage = f2fs_release_page, | |
5b7a487c WG |
1724 | #ifdef CONFIG_MIGRATION |
1725 | .migratepage = f2fs_migrate_page, | |
1726 | #endif | |
e05df3b1 JK |
1727 | }; |
1728 | ||
8a7ed66a JK |
1729 | static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i, |
1730 | nid_t n) | |
e05df3b1 | 1731 | { |
8a7ed66a | 1732 | return radix_tree_lookup(&nm_i->free_nid_root, n); |
e05df3b1 JK |
1733 | } |
1734 | ||
eb0aa4b8 JK |
1735 | static int __insert_nid_to_list(struct f2fs_sb_info *sbi, |
1736 | struct free_nid *i, enum nid_list list, bool new) | |
e05df3b1 | 1737 | { |
b8559dc2 CY |
1738 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1739 | ||
eb0aa4b8 JK |
1740 | if (new) { |
1741 | int err = radix_tree_insert(&nm_i->free_nid_root, i->nid, i); | |
1742 | if (err) | |
1743 | return err; | |
1744 | } | |
1745 | ||
b8559dc2 CY |
1746 | f2fs_bug_on(sbi, list == FREE_NID_LIST ? i->state != NID_NEW : |
1747 | i->state != NID_ALLOC); | |
1748 | nm_i->nid_cnt[list]++; | |
1749 | list_add_tail(&i->list, &nm_i->nid_list[list]); | |
eb0aa4b8 | 1750 | return 0; |
b8559dc2 CY |
1751 | } |
1752 | ||
1753 | static void __remove_nid_from_list(struct f2fs_sb_info *sbi, | |
eb0aa4b8 | 1754 | struct free_nid *i, enum nid_list list, bool reuse) |
b8559dc2 CY |
1755 | { |
1756 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1757 | ||
1758 | f2fs_bug_on(sbi, list == FREE_NID_LIST ? i->state != NID_NEW : | |
1759 | i->state != NID_ALLOC); | |
1760 | nm_i->nid_cnt[list]--; | |
e05df3b1 | 1761 | list_del(&i->list); |
eb0aa4b8 JK |
1762 | if (!reuse) |
1763 | radix_tree_delete(&nm_i->free_nid_root, i->nid); | |
e05df3b1 JK |
1764 | } |
1765 | ||
6fb03f3a | 1766 | static int add_free_nid(struct f2fs_sb_info *sbi, nid_t nid, bool build) |
e05df3b1 | 1767 | { |
6fb03f3a | 1768 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1769 | struct free_nid *i; |
59bbd474 | 1770 | struct nat_entry *ne; |
eb0aa4b8 | 1771 | int err; |
9198aceb JK |
1772 | |
1773 | /* 0 nid should not be used */ | |
cfb271d4 | 1774 | if (unlikely(nid == 0)) |
9198aceb | 1775 | return 0; |
59bbd474 | 1776 | |
7bd59381 GZ |
1777 | if (build) { |
1778 | /* do not add allocated nids */ | |
7bd59381 | 1779 | ne = __lookup_nat_cache(nm_i, nid); |
a5131193 | 1780 | if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) || |
7ef35e3b | 1781 | nat_get_blkaddr(ne) != NULL_ADDR)) |
7bd59381 | 1782 | return 0; |
e05df3b1 | 1783 | } |
7bd59381 GZ |
1784 | |
1785 | i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); | |
e05df3b1 JK |
1786 | i->nid = nid; |
1787 | i->state = NID_NEW; | |
1788 | ||
769ec6e5 JK |
1789 | if (radix_tree_preload(GFP_NOFS)) { |
1790 | kmem_cache_free(free_nid_slab, i); | |
1791 | return 0; | |
1792 | } | |
1793 | ||
b8559dc2 | 1794 | spin_lock(&nm_i->nid_list_lock); |
eb0aa4b8 JK |
1795 | err = __insert_nid_to_list(sbi, i, FREE_NID_LIST, true); |
1796 | spin_unlock(&nm_i->nid_list_lock); | |
1797 | radix_tree_preload_end(); | |
1798 | if (err) { | |
e05df3b1 JK |
1799 | kmem_cache_free(free_nid_slab, i); |
1800 | return 0; | |
1801 | } | |
e05df3b1 JK |
1802 | return 1; |
1803 | } | |
1804 | ||
b8559dc2 | 1805 | static void remove_free_nid(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 | 1806 | { |
b8559dc2 | 1807 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1808 | struct free_nid *i; |
cf0ee0f0 CY |
1809 | bool need_free = false; |
1810 | ||
b8559dc2 | 1811 | spin_lock(&nm_i->nid_list_lock); |
8a7ed66a | 1812 | i = __lookup_free_nid_list(nm_i, nid); |
e05df3b1 | 1813 | if (i && i->state == NID_NEW) { |
eb0aa4b8 | 1814 | __remove_nid_from_list(sbi, i, FREE_NID_LIST, false); |
cf0ee0f0 | 1815 | need_free = true; |
e05df3b1 | 1816 | } |
b8559dc2 | 1817 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 CY |
1818 | |
1819 | if (need_free) | |
1820 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1821 | } |
1822 | ||
6fb03f3a | 1823 | static void scan_nat_page(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
1824 | struct page *nat_page, nid_t start_nid) |
1825 | { | |
6fb03f3a | 1826 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 JK |
1827 | struct f2fs_nat_block *nat_blk = page_address(nat_page); |
1828 | block_t blk_addr; | |
e05df3b1 JK |
1829 | int i; |
1830 | ||
e05df3b1 JK |
1831 | i = start_nid % NAT_ENTRY_PER_BLOCK; |
1832 | ||
1833 | for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { | |
23d38844 | 1834 | |
cfb271d4 | 1835 | if (unlikely(start_nid >= nm_i->max_nid)) |
04431c44 | 1836 | break; |
23d38844 HL |
1837 | |
1838 | blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); | |
9850cf4a | 1839 | f2fs_bug_on(sbi, blk_addr == NEW_ADDR); |
3a2ad567 CY |
1840 | if (blk_addr == NULL_ADDR) |
1841 | add_free_nid(sbi, start_nid, true); | |
e05df3b1 | 1842 | } |
e05df3b1 JK |
1843 | } |
1844 | ||
3e7b5bbb | 1845 | static void __build_free_nids(struct f2fs_sb_info *sbi, bool sync) |
e05df3b1 | 1846 | { |
e05df3b1 JK |
1847 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1848 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 1849 | struct f2fs_journal *journal = curseg->journal; |
8760952d | 1850 | int i = 0; |
55008d84 | 1851 | nid_t nid = nm_i->next_scan_nid; |
e05df3b1 | 1852 | |
55008d84 | 1853 | /* Enough entries */ |
b8559dc2 | 1854 | if (nm_i->nid_cnt[FREE_NID_LIST] >= NAT_ENTRY_PER_BLOCK) |
55008d84 | 1855 | return; |
e05df3b1 | 1856 | |
3a2ad567 | 1857 | if (!sync && !available_free_memory(sbi, FREE_NIDS)) |
55008d84 | 1858 | return; |
e05df3b1 | 1859 | |
55008d84 | 1860 | /* readahead nat pages to be scanned */ |
26879fb1 CY |
1861 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES, |
1862 | META_NAT, true); | |
e05df3b1 | 1863 | |
b873b798 | 1864 | down_read(&nm_i->nat_tree_lock); |
a5131193 | 1865 | |
e05df3b1 JK |
1866 | while (1) { |
1867 | struct page *page = get_current_nat_page(sbi, nid); | |
1868 | ||
6fb03f3a | 1869 | scan_nat_page(sbi, page, nid); |
e05df3b1 JK |
1870 | f2fs_put_page(page, 1); |
1871 | ||
1872 | nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); | |
cfb271d4 | 1873 | if (unlikely(nid >= nm_i->max_nid)) |
e05df3b1 | 1874 | nid = 0; |
55008d84 | 1875 | |
a6d494b6 | 1876 | if (++i >= FREE_NID_PAGES) |
e05df3b1 JK |
1877 | break; |
1878 | } | |
1879 | ||
55008d84 JK |
1880 | /* go to the next free nat pages to find free nids abundantly */ |
1881 | nm_i->next_scan_nid = nid; | |
e05df3b1 JK |
1882 | |
1883 | /* find free nids from current sum_pages */ | |
b7ad7512 | 1884 | down_read(&curseg->journal_rwsem); |
dfc08a12 CY |
1885 | for (i = 0; i < nats_in_cursum(journal); i++) { |
1886 | block_t addr; | |
1887 | ||
1888 | addr = le32_to_cpu(nat_in_journal(journal, i).block_addr); | |
1889 | nid = le32_to_cpu(nid_in_journal(journal, i)); | |
e05df3b1 | 1890 | if (addr == NULL_ADDR) |
6fb03f3a | 1891 | add_free_nid(sbi, nid, true); |
e05df3b1 | 1892 | else |
b8559dc2 | 1893 | remove_free_nid(sbi, nid); |
e05df3b1 | 1894 | } |
b7ad7512 | 1895 | up_read(&curseg->journal_rwsem); |
b873b798 | 1896 | up_read(&nm_i->nat_tree_lock); |
2db2388f CY |
1897 | |
1898 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid), | |
ea1a29a0 | 1899 | nm_i->ra_nid_pages, META_NAT, false); |
e05df3b1 JK |
1900 | } |
1901 | ||
3a2ad567 | 1902 | void build_free_nids(struct f2fs_sb_info *sbi, bool sync) |
2411cf5b CY |
1903 | { |
1904 | mutex_lock(&NM_I(sbi)->build_lock); | |
3a2ad567 | 1905 | __build_free_nids(sbi, sync); |
2411cf5b CY |
1906 | mutex_unlock(&NM_I(sbi)->build_lock); |
1907 | } | |
1908 | ||
e05df3b1 JK |
1909 | /* |
1910 | * If this function returns success, caller can obtain a new nid | |
1911 | * from second parameter of this function. | |
1912 | * The returned nid could be used ino as well as nid when inode is created. | |
1913 | */ | |
1914 | bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) | |
1915 | { | |
1916 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1917 | struct free_nid *i = NULL; | |
e05df3b1 | 1918 | retry: |
cb78942b | 1919 | #ifdef CONFIG_F2FS_FAULT_INJECTION |
1ecc0c5c | 1920 | if (time_to_inject(sbi, FAULT_ALLOC_NID)) |
cb78942b JK |
1921 | return false; |
1922 | #endif | |
b8559dc2 | 1923 | spin_lock(&nm_i->nid_list_lock); |
e05df3b1 | 1924 | |
04d47e67 CY |
1925 | if (unlikely(nm_i->available_nids == 0)) { |
1926 | spin_unlock(&nm_i->nid_list_lock); | |
1927 | return false; | |
1928 | } | |
e05df3b1 | 1929 | |
55008d84 | 1930 | /* We should not use stale free nids created by build_free_nids */ |
b8559dc2 CY |
1931 | if (nm_i->nid_cnt[FREE_NID_LIST] && !on_build_free_nids(nm_i)) { |
1932 | f2fs_bug_on(sbi, list_empty(&nm_i->nid_list[FREE_NID_LIST])); | |
1933 | i = list_first_entry(&nm_i->nid_list[FREE_NID_LIST], | |
1934 | struct free_nid, list); | |
55008d84 | 1935 | *nid = i->nid; |
b8559dc2 | 1936 | |
eb0aa4b8 | 1937 | __remove_nid_from_list(sbi, i, FREE_NID_LIST, true); |
55008d84 | 1938 | i->state = NID_ALLOC; |
eb0aa4b8 | 1939 | __insert_nid_to_list(sbi, i, ALLOC_NID_LIST, false); |
04d47e67 | 1940 | nm_i->available_nids--; |
b8559dc2 | 1941 | spin_unlock(&nm_i->nid_list_lock); |
55008d84 JK |
1942 | return true; |
1943 | } | |
b8559dc2 | 1944 | spin_unlock(&nm_i->nid_list_lock); |
55008d84 JK |
1945 | |
1946 | /* Let's scan nat pages and its caches to get free nids */ | |
3a2ad567 | 1947 | build_free_nids(sbi, true); |
55008d84 | 1948 | goto retry; |
e05df3b1 JK |
1949 | } |
1950 | ||
0a8165d7 | 1951 | /* |
e05df3b1 JK |
1952 | * alloc_nid() should be called prior to this function. |
1953 | */ | |
1954 | void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) | |
1955 | { | |
1956 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1957 | struct free_nid *i; | |
1958 | ||
b8559dc2 | 1959 | spin_lock(&nm_i->nid_list_lock); |
8a7ed66a | 1960 | i = __lookup_free_nid_list(nm_i, nid); |
b8559dc2 | 1961 | f2fs_bug_on(sbi, !i); |
eb0aa4b8 | 1962 | __remove_nid_from_list(sbi, i, ALLOC_NID_LIST, false); |
b8559dc2 | 1963 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 CY |
1964 | |
1965 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1966 | } |
1967 | ||
0a8165d7 | 1968 | /* |
e05df3b1 JK |
1969 | * alloc_nid() should be called prior to this function. |
1970 | */ | |
1971 | void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) | |
1972 | { | |
49952fa1 JK |
1973 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1974 | struct free_nid *i; | |
cf0ee0f0 | 1975 | bool need_free = false; |
49952fa1 | 1976 | |
65985d93 JK |
1977 | if (!nid) |
1978 | return; | |
1979 | ||
b8559dc2 | 1980 | spin_lock(&nm_i->nid_list_lock); |
8a7ed66a | 1981 | i = __lookup_free_nid_list(nm_i, nid); |
b8559dc2 CY |
1982 | f2fs_bug_on(sbi, !i); |
1983 | ||
6fb03f3a | 1984 | if (!available_free_memory(sbi, FREE_NIDS)) { |
eb0aa4b8 | 1985 | __remove_nid_from_list(sbi, i, ALLOC_NID_LIST, false); |
cf0ee0f0 | 1986 | need_free = true; |
95630cba | 1987 | } else { |
eb0aa4b8 | 1988 | __remove_nid_from_list(sbi, i, ALLOC_NID_LIST, true); |
95630cba | 1989 | i->state = NID_NEW; |
eb0aa4b8 | 1990 | __insert_nid_to_list(sbi, i, FREE_NID_LIST, false); |
95630cba | 1991 | } |
04d47e67 CY |
1992 | |
1993 | nm_i->available_nids++; | |
1994 | ||
b8559dc2 | 1995 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 CY |
1996 | |
1997 | if (need_free) | |
1998 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1999 | } |
2000 | ||
31696580 CY |
2001 | int try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink) |
2002 | { | |
2003 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2004 | struct free_nid *i, *next; | |
2005 | int nr = nr_shrink; | |
2006 | ||
b8559dc2 | 2007 | if (nm_i->nid_cnt[FREE_NID_LIST] <= MAX_FREE_NIDS) |
ad4edb83 JK |
2008 | return 0; |
2009 | ||
31696580 CY |
2010 | if (!mutex_trylock(&nm_i->build_lock)) |
2011 | return 0; | |
2012 | ||
b8559dc2 CY |
2013 | spin_lock(&nm_i->nid_list_lock); |
2014 | list_for_each_entry_safe(i, next, &nm_i->nid_list[FREE_NID_LIST], | |
2015 | list) { | |
2016 | if (nr_shrink <= 0 || | |
2017 | nm_i->nid_cnt[FREE_NID_LIST] <= MAX_FREE_NIDS) | |
31696580 | 2018 | break; |
b8559dc2 | 2019 | |
eb0aa4b8 | 2020 | __remove_nid_from_list(sbi, i, FREE_NID_LIST, false); |
31696580 CY |
2021 | kmem_cache_free(free_nid_slab, i); |
2022 | nr_shrink--; | |
31696580 | 2023 | } |
b8559dc2 | 2024 | spin_unlock(&nm_i->nid_list_lock); |
31696580 CY |
2025 | mutex_unlock(&nm_i->build_lock); |
2026 | ||
2027 | return nr - nr_shrink; | |
2028 | } | |
2029 | ||
70cfed88 | 2030 | void recover_inline_xattr(struct inode *inode, struct page *page) |
28cdce04 | 2031 | { |
28cdce04 CY |
2032 | void *src_addr, *dst_addr; |
2033 | size_t inline_size; | |
2034 | struct page *ipage; | |
2035 | struct f2fs_inode *ri; | |
2036 | ||
4081363f | 2037 | ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino); |
9850cf4a | 2038 | f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage)); |
28cdce04 | 2039 | |
e3b4d43f JK |
2040 | ri = F2FS_INODE(page); |
2041 | if (!(ri->i_inline & F2FS_INLINE_XATTR)) { | |
91942321 | 2042 | clear_inode_flag(inode, FI_INLINE_XATTR); |
e3b4d43f JK |
2043 | goto update_inode; |
2044 | } | |
2045 | ||
28cdce04 CY |
2046 | dst_addr = inline_xattr_addr(ipage); |
2047 | src_addr = inline_xattr_addr(page); | |
2048 | inline_size = inline_xattr_size(inode); | |
2049 | ||
fec1d657 | 2050 | f2fs_wait_on_page_writeback(ipage, NODE, true); |
28cdce04 | 2051 | memcpy(dst_addr, src_addr, inline_size); |
e3b4d43f | 2052 | update_inode: |
28cdce04 CY |
2053 | update_inode(inode, ipage); |
2054 | f2fs_put_page(ipage, 1); | |
2055 | } | |
2056 | ||
1c35a90e | 2057 | void recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr) |
abb2366c | 2058 | { |
4081363f | 2059 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
abb2366c JK |
2060 | nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid; |
2061 | nid_t new_xnid = nid_of_node(page); | |
2062 | struct node_info ni; | |
2063 | ||
abb2366c JK |
2064 | /* 1: invalidate the previous xattr nid */ |
2065 | if (!prev_xnid) | |
2066 | goto recover_xnid; | |
2067 | ||
2068 | /* Deallocate node address */ | |
2069 | get_node_info(sbi, prev_xnid, &ni); | |
9850cf4a | 2070 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
abb2366c JK |
2071 | invalidate_blocks(sbi, ni.blk_addr); |
2072 | dec_valid_node_count(sbi, inode); | |
479f40c4 | 2073 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
abb2366c JK |
2074 | |
2075 | recover_xnid: | |
2076 | /* 2: allocate new xattr nid */ | |
2077 | if (unlikely(!inc_valid_node_count(sbi, inode))) | |
9850cf4a | 2078 | f2fs_bug_on(sbi, 1); |
abb2366c | 2079 | |
b8559dc2 | 2080 | remove_free_nid(sbi, new_xnid); |
abb2366c JK |
2081 | get_node_info(sbi, new_xnid, &ni); |
2082 | ni.ino = inode->i_ino; | |
479f40c4 | 2083 | set_node_addr(sbi, &ni, NEW_ADDR, false); |
205b9822 | 2084 | f2fs_i_xnid_write(inode, new_xnid); |
abb2366c JK |
2085 | |
2086 | /* 3: update xattr blkaddr */ | |
2087 | refresh_sit_entry(sbi, NEW_ADDR, blkaddr); | |
479f40c4 | 2088 | set_node_addr(sbi, &ni, blkaddr, false); |
abb2366c JK |
2089 | } |
2090 | ||
e05df3b1 JK |
2091 | int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) |
2092 | { | |
58bfaf44 | 2093 | struct f2fs_inode *src, *dst; |
e05df3b1 JK |
2094 | nid_t ino = ino_of_node(page); |
2095 | struct node_info old_ni, new_ni; | |
2096 | struct page *ipage; | |
2097 | ||
e8271fa3 JK |
2098 | get_node_info(sbi, ino, &old_ni); |
2099 | ||
2100 | if (unlikely(old_ni.blk_addr != NULL_ADDR)) | |
2101 | return -EINVAL; | |
e8ea9b3d | 2102 | retry: |
300e129c | 2103 | ipage = f2fs_grab_cache_page(NODE_MAPPING(sbi), ino, false); |
e8ea9b3d JK |
2104 | if (!ipage) { |
2105 | congestion_wait(BLK_RW_ASYNC, HZ/50); | |
2106 | goto retry; | |
2107 | } | |
e05df3b1 | 2108 | |
e1c42045 | 2109 | /* Should not use this inode from free nid list */ |
b8559dc2 | 2110 | remove_free_nid(sbi, ino); |
e05df3b1 | 2111 | |
237c0790 JK |
2112 | if (!PageUptodate(ipage)) |
2113 | SetPageUptodate(ipage); | |
e05df3b1 JK |
2114 | fill_node_footer(ipage, ino, ino, 0, true); |
2115 | ||
58bfaf44 JK |
2116 | src = F2FS_INODE(page); |
2117 | dst = F2FS_INODE(ipage); | |
e05df3b1 | 2118 | |
58bfaf44 JK |
2119 | memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); |
2120 | dst->i_size = 0; | |
2121 | dst->i_blocks = cpu_to_le64(1); | |
2122 | dst->i_links = cpu_to_le32(1); | |
2123 | dst->i_xattr_nid = 0; | |
617deb8c | 2124 | dst->i_inline = src->i_inline & F2FS_INLINE_XATTR; |
e05df3b1 JK |
2125 | |
2126 | new_ni = old_ni; | |
2127 | new_ni.ino = ino; | |
2128 | ||
cfb271d4 | 2129 | if (unlikely(!inc_valid_node_count(sbi, NULL))) |
65e5cd0a | 2130 | WARN_ON(1); |
479f40c4 | 2131 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
e05df3b1 | 2132 | inc_valid_inode_count(sbi); |
617deb8c | 2133 | set_page_dirty(ipage); |
e05df3b1 JK |
2134 | f2fs_put_page(ipage, 1); |
2135 | return 0; | |
2136 | } | |
2137 | ||
2138 | int restore_node_summary(struct f2fs_sb_info *sbi, | |
2139 | unsigned int segno, struct f2fs_summary_block *sum) | |
2140 | { | |
2141 | struct f2fs_node *rn; | |
2142 | struct f2fs_summary *sum_entry; | |
e05df3b1 | 2143 | block_t addr; |
9ecf4b80 | 2144 | int i, idx, last_offset, nrpages; |
e05df3b1 JK |
2145 | |
2146 | /* scan the node segment */ | |
2147 | last_offset = sbi->blocks_per_seg; | |
2148 | addr = START_BLOCK(sbi, segno); | |
2149 | sum_entry = &sum->entries[0]; | |
2150 | ||
9ecf4b80 | 2151 | for (i = 0; i < last_offset; i += nrpages, addr += nrpages) { |
664ba972 | 2152 | nrpages = min(last_offset - i, BIO_MAX_PAGES); |
393ff91f | 2153 | |
e1c42045 | 2154 | /* readahead node pages */ |
26879fb1 | 2155 | ra_meta_pages(sbi, addr, nrpages, META_POR, true); |
e05df3b1 | 2156 | |
9ecf4b80 | 2157 | for (idx = addr; idx < addr + nrpages; idx++) { |
2b947003 | 2158 | struct page *page = get_tmp_page(sbi, idx); |
9af0ff1c | 2159 | |
9ecf4b80 CY |
2160 | rn = F2FS_NODE(page); |
2161 | sum_entry->nid = rn->footer.nid; | |
2162 | sum_entry->version = 0; | |
2163 | sum_entry->ofs_in_node = 0; | |
2164 | sum_entry++; | |
2165 | f2fs_put_page(page, 1); | |
9af0ff1c | 2166 | } |
bac4eef6 | 2167 | |
9ecf4b80 | 2168 | invalidate_mapping_pages(META_MAPPING(sbi), addr, |
bac4eef6 | 2169 | addr + nrpages); |
e05df3b1 | 2170 | } |
9ecf4b80 | 2171 | return 0; |
e05df3b1 JK |
2172 | } |
2173 | ||
aec71382 | 2174 | static void remove_nats_in_journal(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
2175 | { |
2176 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2177 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2178 | struct f2fs_journal *journal = curseg->journal; |
e05df3b1 JK |
2179 | int i; |
2180 | ||
b7ad7512 | 2181 | down_write(&curseg->journal_rwsem); |
dfc08a12 | 2182 | for (i = 0; i < nats_in_cursum(journal); i++) { |
e05df3b1 JK |
2183 | struct nat_entry *ne; |
2184 | struct f2fs_nat_entry raw_ne; | |
dfc08a12 | 2185 | nid_t nid = le32_to_cpu(nid_in_journal(journal, i)); |
e05df3b1 | 2186 | |
dfc08a12 | 2187 | raw_ne = nat_in_journal(journal, i); |
9be32d72 | 2188 | |
e05df3b1 | 2189 | ne = __lookup_nat_cache(nm_i, nid); |
e05df3b1 | 2190 | if (!ne) { |
5c9e4184 | 2191 | ne = grab_nat_entry(nm_i, nid, true); |
9be32d72 | 2192 | node_info_from_raw_nat(&ne->ni, &raw_ne); |
e05df3b1 | 2193 | } |
04d47e67 CY |
2194 | |
2195 | /* | |
2196 | * if a free nat in journal has not been used after last | |
2197 | * checkpoint, we should remove it from available nids, | |
2198 | * since later we will add it again. | |
2199 | */ | |
2200 | if (!get_nat_flag(ne, IS_DIRTY) && | |
2201 | le32_to_cpu(raw_ne.block_addr) == NULL_ADDR) { | |
2202 | spin_lock(&nm_i->nid_list_lock); | |
2203 | nm_i->available_nids--; | |
2204 | spin_unlock(&nm_i->nid_list_lock); | |
2205 | } | |
2206 | ||
e05df3b1 | 2207 | __set_nat_cache_dirty(nm_i, ne); |
e05df3b1 | 2208 | } |
dfc08a12 | 2209 | update_nats_in_cursum(journal, -i); |
b7ad7512 | 2210 | up_write(&curseg->journal_rwsem); |
e05df3b1 JK |
2211 | } |
2212 | ||
309cc2b6 JK |
2213 | static void __adjust_nat_entry_set(struct nat_entry_set *nes, |
2214 | struct list_head *head, int max) | |
e05df3b1 | 2215 | { |
309cc2b6 | 2216 | struct nat_entry_set *cur; |
e05df3b1 | 2217 | |
309cc2b6 JK |
2218 | if (nes->entry_cnt >= max) |
2219 | goto add_out; | |
e05df3b1 | 2220 | |
309cc2b6 JK |
2221 | list_for_each_entry(cur, head, set_list) { |
2222 | if (cur->entry_cnt >= nes->entry_cnt) { | |
2223 | list_add(&nes->set_list, cur->set_list.prev); | |
2224 | return; | |
2225 | } | |
aec71382 | 2226 | } |
309cc2b6 JK |
2227 | add_out: |
2228 | list_add_tail(&nes->set_list, head); | |
2229 | } | |
e05df3b1 | 2230 | |
309cc2b6 JK |
2231 | static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, |
2232 | struct nat_entry_set *set) | |
2233 | { | |
2234 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2235 | struct f2fs_journal *journal = curseg->journal; |
309cc2b6 JK |
2236 | nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK; |
2237 | bool to_journal = true; | |
2238 | struct f2fs_nat_block *nat_blk; | |
2239 | struct nat_entry *ne, *cur; | |
2240 | struct page *page = NULL; | |
e05df3b1 | 2241 | |
aec71382 CY |
2242 | /* |
2243 | * there are two steps to flush nat entries: | |
2244 | * #1, flush nat entries to journal in current hot data summary block. | |
2245 | * #2, flush nat entries to nat page. | |
2246 | */ | |
dfc08a12 | 2247 | if (!__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL)) |
309cc2b6 JK |
2248 | to_journal = false; |
2249 | ||
2250 | if (to_journal) { | |
b7ad7512 | 2251 | down_write(&curseg->journal_rwsem); |
309cc2b6 JK |
2252 | } else { |
2253 | page = get_next_nat_page(sbi, start_nid); | |
2254 | nat_blk = page_address(page); | |
2255 | f2fs_bug_on(sbi, !nat_blk); | |
2256 | } | |
aec71382 | 2257 | |
309cc2b6 JK |
2258 | /* flush dirty nats in nat entry set */ |
2259 | list_for_each_entry_safe(ne, cur, &set->entry_list, list) { | |
2260 | struct f2fs_nat_entry *raw_ne; | |
2261 | nid_t nid = nat_get_nid(ne); | |
2262 | int offset; | |
2263 | ||
2264 | if (nat_get_blkaddr(ne) == NEW_ADDR) | |
2265 | continue; | |
aec71382 CY |
2266 | |
2267 | if (to_journal) { | |
dfc08a12 | 2268 | offset = lookup_journal_in_cursum(journal, |
309cc2b6 JK |
2269 | NAT_JOURNAL, nid, 1); |
2270 | f2fs_bug_on(sbi, offset < 0); | |
dfc08a12 CY |
2271 | raw_ne = &nat_in_journal(journal, offset); |
2272 | nid_in_journal(journal, offset) = cpu_to_le32(nid); | |
aec71382 | 2273 | } else { |
309cc2b6 | 2274 | raw_ne = &nat_blk->entries[nid - start_nid]; |
e05df3b1 | 2275 | } |
309cc2b6 | 2276 | raw_nat_from_node_info(raw_ne, &ne->ni); |
309cc2b6 JK |
2277 | nat_reset_flag(ne); |
2278 | __clear_nat_cache_dirty(NM_I(sbi), ne); | |
04d47e67 | 2279 | if (nat_get_blkaddr(ne) == NULL_ADDR) { |
309cc2b6 | 2280 | add_free_nid(sbi, nid, false); |
04d47e67 CY |
2281 | spin_lock(&NM_I(sbi)->nid_list_lock); |
2282 | NM_I(sbi)->available_nids++; | |
2283 | spin_unlock(&NM_I(sbi)->nid_list_lock); | |
2284 | } | |
309cc2b6 | 2285 | } |
e05df3b1 | 2286 | |
309cc2b6 | 2287 | if (to_journal) |
b7ad7512 | 2288 | up_write(&curseg->journal_rwsem); |
309cc2b6 JK |
2289 | else |
2290 | f2fs_put_page(page, 1); | |
aec71382 | 2291 | |
80ec2e91 CL |
2292 | f2fs_bug_on(sbi, set->entry_cnt); |
2293 | ||
2294 | radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); | |
2295 | kmem_cache_free(nat_entry_set_slab, set); | |
309cc2b6 | 2296 | } |
aec71382 | 2297 | |
309cc2b6 JK |
2298 | /* |
2299 | * This function is called during the checkpointing process. | |
2300 | */ | |
2301 | void flush_nat_entries(struct f2fs_sb_info *sbi) | |
2302 | { | |
2303 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2304 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2305 | struct f2fs_journal *journal = curseg->journal; |
7aed0d45 | 2306 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
309cc2b6 JK |
2307 | struct nat_entry_set *set, *tmp; |
2308 | unsigned int found; | |
2309 | nid_t set_idx = 0; | |
2310 | LIST_HEAD(sets); | |
2311 | ||
20d047c8 CL |
2312 | if (!nm_i->dirty_nat_cnt) |
2313 | return; | |
a5131193 | 2314 | |
b873b798 | 2315 | down_write(&nm_i->nat_tree_lock); |
a5131193 | 2316 | |
309cc2b6 JK |
2317 | /* |
2318 | * if there are no enough space in journal to store dirty nat | |
2319 | * entries, remove all entries from journal and merge them | |
2320 | * into nat entry set. | |
2321 | */ | |
dfc08a12 | 2322 | if (!__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL)) |
309cc2b6 JK |
2323 | remove_nats_in_journal(sbi); |
2324 | ||
309cc2b6 | 2325 | while ((found = __gang_lookup_nat_set(nm_i, |
7aed0d45 | 2326 | set_idx, SETVEC_SIZE, setvec))) { |
309cc2b6 JK |
2327 | unsigned idx; |
2328 | set_idx = setvec[found - 1]->set + 1; | |
2329 | for (idx = 0; idx < found; idx++) | |
2330 | __adjust_nat_entry_set(setvec[idx], &sets, | |
dfc08a12 | 2331 | MAX_NAT_JENTRIES(journal)); |
e05df3b1 | 2332 | } |
aec71382 | 2333 | |
309cc2b6 JK |
2334 | /* flush dirty nats in nat entry set */ |
2335 | list_for_each_entry_safe(set, tmp, &sets, set_list) | |
2336 | __flush_nat_entry_set(sbi, set); | |
2337 | ||
b873b798 | 2338 | up_write(&nm_i->nat_tree_lock); |
a5131193 | 2339 | |
9850cf4a | 2340 | f2fs_bug_on(sbi, nm_i->dirty_nat_cnt); |
e05df3b1 JK |
2341 | } |
2342 | ||
2343 | static int init_node_manager(struct f2fs_sb_info *sbi) | |
2344 | { | |
2345 | struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); | |
2346 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2347 | unsigned char *version_bitmap; | |
2348 | unsigned int nat_segs, nat_blocks; | |
2349 | ||
2350 | nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); | |
2351 | ||
2352 | /* segment_count_nat includes pair segment so divide to 2. */ | |
2353 | nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; | |
2354 | nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); | |
b63da15e | 2355 | |
7ee0eeab JK |
2356 | nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks; |
2357 | ||
b63da15e | 2358 | /* not used nids: 0, node, meta, (and root counted as valid node) */ |
04d47e67 CY |
2359 | nm_i->available_nids = nm_i->max_nid - sbi->total_valid_node_count - |
2360 | F2FS_RESERVED_NODE_NUM; | |
b8559dc2 CY |
2361 | nm_i->nid_cnt[FREE_NID_LIST] = 0; |
2362 | nm_i->nid_cnt[ALLOC_NID_LIST] = 0; | |
e05df3b1 | 2363 | nm_i->nat_cnt = 0; |
cdfc41c1 | 2364 | nm_i->ram_thresh = DEF_RAM_THRESHOLD; |
ea1a29a0 | 2365 | nm_i->ra_nid_pages = DEF_RA_NID_PAGES; |
2304cb0c | 2366 | nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD; |
e05df3b1 | 2367 | |
8a7ed66a | 2368 | INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC); |
b8559dc2 CY |
2369 | INIT_LIST_HEAD(&nm_i->nid_list[FREE_NID_LIST]); |
2370 | INIT_LIST_HEAD(&nm_i->nid_list[ALLOC_NID_LIST]); | |
769ec6e5 JK |
2371 | INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO); |
2372 | INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO); | |
e05df3b1 | 2373 | INIT_LIST_HEAD(&nm_i->nat_entries); |
e05df3b1 JK |
2374 | |
2375 | mutex_init(&nm_i->build_lock); | |
b8559dc2 | 2376 | spin_lock_init(&nm_i->nid_list_lock); |
b873b798 | 2377 | init_rwsem(&nm_i->nat_tree_lock); |
e05df3b1 | 2378 | |
e05df3b1 | 2379 | nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); |
79b5793b | 2380 | nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); |
e05df3b1 JK |
2381 | version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); |
2382 | if (!version_bitmap) | |
2383 | return -EFAULT; | |
2384 | ||
79b5793b AG |
2385 | nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, |
2386 | GFP_KERNEL); | |
2387 | if (!nm_i->nat_bitmap) | |
2388 | return -ENOMEM; | |
599a09b2 CY |
2389 | |
2390 | #ifdef CONFIG_F2FS_CHECK_FS | |
2391 | nm_i->nat_bitmap_mir = kmemdup(version_bitmap, nm_i->bitmap_size, | |
2392 | GFP_KERNEL); | |
2393 | if (!nm_i->nat_bitmap_mir) | |
2394 | return -ENOMEM; | |
2395 | #endif | |
2396 | ||
e05df3b1 JK |
2397 | return 0; |
2398 | } | |
2399 | ||
2400 | int build_node_manager(struct f2fs_sb_info *sbi) | |
2401 | { | |
2402 | int err; | |
2403 | ||
2404 | sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL); | |
2405 | if (!sbi->nm_info) | |
2406 | return -ENOMEM; | |
2407 | ||
2408 | err = init_node_manager(sbi); | |
2409 | if (err) | |
2410 | return err; | |
2411 | ||
3a2ad567 | 2412 | build_free_nids(sbi, true); |
e05df3b1 JK |
2413 | return 0; |
2414 | } | |
2415 | ||
2416 | void destroy_node_manager(struct f2fs_sb_info *sbi) | |
2417 | { | |
2418 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2419 | struct free_nid *i, *next_i; | |
2420 | struct nat_entry *natvec[NATVEC_SIZE]; | |
7aed0d45 | 2421 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
e05df3b1 JK |
2422 | nid_t nid = 0; |
2423 | unsigned int found; | |
2424 | ||
2425 | if (!nm_i) | |
2426 | return; | |
2427 | ||
2428 | /* destroy free nid list */ | |
b8559dc2 CY |
2429 | spin_lock(&nm_i->nid_list_lock); |
2430 | list_for_each_entry_safe(i, next_i, &nm_i->nid_list[FREE_NID_LIST], | |
2431 | list) { | |
eb0aa4b8 | 2432 | __remove_nid_from_list(sbi, i, FREE_NID_LIST, false); |
b8559dc2 | 2433 | spin_unlock(&nm_i->nid_list_lock); |
cf0ee0f0 | 2434 | kmem_cache_free(free_nid_slab, i); |
b8559dc2 | 2435 | spin_lock(&nm_i->nid_list_lock); |
e05df3b1 | 2436 | } |
b8559dc2 CY |
2437 | f2fs_bug_on(sbi, nm_i->nid_cnt[FREE_NID_LIST]); |
2438 | f2fs_bug_on(sbi, nm_i->nid_cnt[ALLOC_NID_LIST]); | |
2439 | f2fs_bug_on(sbi, !list_empty(&nm_i->nid_list[ALLOC_NID_LIST])); | |
2440 | spin_unlock(&nm_i->nid_list_lock); | |
e05df3b1 JK |
2441 | |
2442 | /* destroy nat cache */ | |
b873b798 | 2443 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2444 | while ((found = __gang_lookup_nat_cache(nm_i, |
2445 | nid, NATVEC_SIZE, natvec))) { | |
2446 | unsigned idx; | |
7aed0d45 | 2447 | |
b6ce391e GZ |
2448 | nid = nat_get_nid(natvec[found - 1]) + 1; |
2449 | for (idx = 0; idx < found; idx++) | |
2450 | __del_from_nat_cache(nm_i, natvec[idx]); | |
e05df3b1 | 2451 | } |
9850cf4a | 2452 | f2fs_bug_on(sbi, nm_i->nat_cnt); |
7aed0d45 JK |
2453 | |
2454 | /* destroy nat set cache */ | |
2455 | nid = 0; | |
2456 | while ((found = __gang_lookup_nat_set(nm_i, | |
2457 | nid, SETVEC_SIZE, setvec))) { | |
2458 | unsigned idx; | |
2459 | ||
2460 | nid = setvec[found - 1]->set + 1; | |
2461 | for (idx = 0; idx < found; idx++) { | |
2462 | /* entry_cnt is not zero, when cp_error was occurred */ | |
2463 | f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list)); | |
2464 | radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set); | |
2465 | kmem_cache_free(nat_entry_set_slab, setvec[idx]); | |
2466 | } | |
2467 | } | |
b873b798 | 2468 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2469 | |
2470 | kfree(nm_i->nat_bitmap); | |
599a09b2 CY |
2471 | #ifdef CONFIG_F2FS_CHECK_FS |
2472 | kfree(nm_i->nat_bitmap_mir); | |
2473 | #endif | |
e05df3b1 JK |
2474 | sbi->nm_info = NULL; |
2475 | kfree(nm_i); | |
2476 | } | |
2477 | ||
6e6093a8 | 2478 | int __init create_node_manager_caches(void) |
e05df3b1 JK |
2479 | { |
2480 | nat_entry_slab = f2fs_kmem_cache_create("nat_entry", | |
e8512d2e | 2481 | sizeof(struct nat_entry)); |
e05df3b1 | 2482 | if (!nat_entry_slab) |
aec71382 | 2483 | goto fail; |
e05df3b1 JK |
2484 | |
2485 | free_nid_slab = f2fs_kmem_cache_create("free_nid", | |
e8512d2e | 2486 | sizeof(struct free_nid)); |
aec71382 | 2487 | if (!free_nid_slab) |
ce3e6d25 | 2488 | goto destroy_nat_entry; |
aec71382 CY |
2489 | |
2490 | nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set", | |
2491 | sizeof(struct nat_entry_set)); | |
2492 | if (!nat_entry_set_slab) | |
ce3e6d25 | 2493 | goto destroy_free_nid; |
e05df3b1 | 2494 | return 0; |
aec71382 | 2495 | |
ce3e6d25 | 2496 | destroy_free_nid: |
aec71382 | 2497 | kmem_cache_destroy(free_nid_slab); |
ce3e6d25 | 2498 | destroy_nat_entry: |
aec71382 CY |
2499 | kmem_cache_destroy(nat_entry_slab); |
2500 | fail: | |
2501 | return -ENOMEM; | |
e05df3b1 JK |
2502 | } |
2503 | ||
2504 | void destroy_node_manager_caches(void) | |
2505 | { | |
aec71382 | 2506 | kmem_cache_destroy(nat_entry_set_slab); |
e05df3b1 JK |
2507 | kmem_cache_destroy(free_nid_slab); |
2508 | kmem_cache_destroy(nat_entry_slab); | |
2509 | } |