]>
Commit | Line | Data |
---|---|---|
0a8165d7 | 1 | /* |
e05df3b1 JK |
2 | * fs/f2fs/node.c |
3 | * | |
4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
5 | * http://www.samsung.com/ | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
11 | #include <linux/fs.h> | |
12 | #include <linux/f2fs_fs.h> | |
13 | #include <linux/mpage.h> | |
14 | #include <linux/backing-dev.h> | |
15 | #include <linux/blkdev.h> | |
16 | #include <linux/pagevec.h> | |
17 | #include <linux/swap.h> | |
18 | ||
19 | #include "f2fs.h" | |
20 | #include "node.h" | |
21 | #include "segment.h" | |
9e4ded3f | 22 | #include "trace.h" |
51dd6249 | 23 | #include <trace/events/f2fs.h> |
e05df3b1 | 24 | |
f978f5a0 GZ |
25 | #define on_build_free_nids(nmi) mutex_is_locked(&nm_i->build_lock) |
26 | ||
e05df3b1 JK |
27 | static struct kmem_cache *nat_entry_slab; |
28 | static struct kmem_cache *free_nid_slab; | |
aec71382 | 29 | static struct kmem_cache *nat_entry_set_slab; |
e05df3b1 | 30 | |
6fb03f3a | 31 | bool available_free_memory(struct f2fs_sb_info *sbi, int type) |
cdfc41c1 | 32 | { |
6fb03f3a | 33 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
cdfc41c1 | 34 | struct sysinfo val; |
e5e7ea3c | 35 | unsigned long avail_ram; |
cdfc41c1 | 36 | unsigned long mem_size = 0; |
6fb03f3a | 37 | bool res = false; |
cdfc41c1 JK |
38 | |
39 | si_meminfo(&val); | |
e5e7ea3c JK |
40 | |
41 | /* only uses low memory */ | |
42 | avail_ram = val.totalram - val.totalhigh; | |
43 | ||
429511cd CY |
44 | /* |
45 | * give 25%, 25%, 50%, 50%, 50% memory for each components respectively | |
46 | */ | |
6fb03f3a | 47 | if (type == FREE_NIDS) { |
e5e7ea3c JK |
48 | mem_size = (nm_i->fcnt * sizeof(struct free_nid)) >> |
49 | PAGE_CACHE_SHIFT; | |
50 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); | |
6fb03f3a | 51 | } else if (type == NAT_ENTRIES) { |
e5e7ea3c JK |
52 | mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >> |
53 | PAGE_CACHE_SHIFT; | |
54 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); | |
a1257023 JK |
55 | } else if (type == DIRTY_DENTS) { |
56 | if (sbi->sb->s_bdi->wb.dirty_exceeded) | |
57 | return false; | |
58 | mem_size = get_pages(sbi, F2FS_DIRTY_DENTS); | |
59 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); | |
e5e7ea3c JK |
60 | } else if (type == INO_ENTRIES) { |
61 | int i; | |
62 | ||
e5e7ea3c | 63 | for (i = 0; i <= UPDATE_INO; i++) |
67298804 CY |
64 | mem_size += (sbi->im[i].ino_num * |
65 | sizeof(struct ino_entry)) >> PAGE_CACHE_SHIFT; | |
e5e7ea3c | 66 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
429511cd | 67 | } else if (type == EXTENT_CACHE) { |
7441ccef JK |
68 | mem_size = (atomic_read(&sbi->total_ext_tree) * |
69 | sizeof(struct extent_tree) + | |
429511cd CY |
70 | atomic_read(&sbi->total_ext_node) * |
71 | sizeof(struct extent_node)) >> PAGE_CACHE_SHIFT; | |
72 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); | |
1e84371f | 73 | } else { |
1663cae4 JK |
74 | if (!sbi->sb->s_bdi->wb.dirty_exceeded) |
75 | return true; | |
6fb03f3a JK |
76 | } |
77 | return res; | |
cdfc41c1 JK |
78 | } |
79 | ||
e05df3b1 JK |
80 | static void clear_node_page_dirty(struct page *page) |
81 | { | |
82 | struct address_space *mapping = page->mapping; | |
e05df3b1 JK |
83 | unsigned int long flags; |
84 | ||
85 | if (PageDirty(page)) { | |
86 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
87 | radix_tree_tag_clear(&mapping->page_tree, | |
88 | page_index(page), | |
89 | PAGECACHE_TAG_DIRTY); | |
90 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
91 | ||
92 | clear_page_dirty_for_io(page); | |
4081363f | 93 | dec_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_NODES); |
e05df3b1 JK |
94 | } |
95 | ClearPageUptodate(page); | |
96 | } | |
97 | ||
98 | static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
99 | { | |
100 | pgoff_t index = current_nat_addr(sbi, nid); | |
101 | return get_meta_page(sbi, index); | |
102 | } | |
103 | ||
104 | static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
105 | { | |
106 | struct page *src_page; | |
107 | struct page *dst_page; | |
108 | pgoff_t src_off; | |
109 | pgoff_t dst_off; | |
110 | void *src_addr; | |
111 | void *dst_addr; | |
112 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
113 | ||
114 | src_off = current_nat_addr(sbi, nid); | |
115 | dst_off = next_nat_addr(sbi, src_off); | |
116 | ||
117 | /* get current nat block page with lock */ | |
118 | src_page = get_meta_page(sbi, src_off); | |
e05df3b1 | 119 | dst_page = grab_meta_page(sbi, dst_off); |
9850cf4a | 120 | f2fs_bug_on(sbi, PageDirty(src_page)); |
e05df3b1 JK |
121 | |
122 | src_addr = page_address(src_page); | |
123 | dst_addr = page_address(dst_page); | |
124 | memcpy(dst_addr, src_addr, PAGE_CACHE_SIZE); | |
125 | set_page_dirty(dst_page); | |
126 | f2fs_put_page(src_page, 1); | |
127 | ||
128 | set_to_next_nat(nm_i, nid); | |
129 | ||
130 | return dst_page; | |
131 | } | |
132 | ||
e05df3b1 JK |
133 | static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n) |
134 | { | |
135 | return radix_tree_lookup(&nm_i->nat_root, n); | |
136 | } | |
137 | ||
138 | static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i, | |
139 | nid_t start, unsigned int nr, struct nat_entry **ep) | |
140 | { | |
141 | return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr); | |
142 | } | |
143 | ||
144 | static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e) | |
145 | { | |
146 | list_del(&e->list); | |
147 | radix_tree_delete(&nm_i->nat_root, nat_get_nid(e)); | |
148 | nm_i->nat_cnt--; | |
149 | kmem_cache_free(nat_entry_slab, e); | |
150 | } | |
151 | ||
309cc2b6 JK |
152 | static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i, |
153 | struct nat_entry *ne) | |
154 | { | |
155 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); | |
156 | struct nat_entry_set *head; | |
157 | ||
158 | if (get_nat_flag(ne, IS_DIRTY)) | |
159 | return; | |
9be32d72 | 160 | |
309cc2b6 JK |
161 | head = radix_tree_lookup(&nm_i->nat_set_root, set); |
162 | if (!head) { | |
80c54505 | 163 | head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS); |
309cc2b6 JK |
164 | |
165 | INIT_LIST_HEAD(&head->entry_list); | |
166 | INIT_LIST_HEAD(&head->set_list); | |
167 | head->set = set; | |
168 | head->entry_cnt = 0; | |
9be32d72 | 169 | f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head); |
309cc2b6 JK |
170 | } |
171 | list_move_tail(&ne->list, &head->entry_list); | |
172 | nm_i->dirty_nat_cnt++; | |
173 | head->entry_cnt++; | |
174 | set_nat_flag(ne, IS_DIRTY, true); | |
175 | } | |
176 | ||
177 | static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i, | |
178 | struct nat_entry *ne) | |
179 | { | |
20d047c8 | 180 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); |
309cc2b6 JK |
181 | struct nat_entry_set *head; |
182 | ||
183 | head = radix_tree_lookup(&nm_i->nat_set_root, set); | |
184 | if (head) { | |
185 | list_move_tail(&ne->list, &nm_i->nat_entries); | |
186 | set_nat_flag(ne, IS_DIRTY, false); | |
187 | head->entry_cnt--; | |
188 | nm_i->dirty_nat_cnt--; | |
189 | } | |
190 | } | |
191 | ||
192 | static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i, | |
193 | nid_t start, unsigned int nr, struct nat_entry_set **ep) | |
194 | { | |
195 | return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep, | |
196 | start, nr); | |
197 | } | |
198 | ||
2dcf51ab | 199 | int need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 JK |
200 | { |
201 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
202 | struct nat_entry *e; | |
2dcf51ab | 203 | bool need = false; |
e05df3b1 | 204 | |
8b26ef98 | 205 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 | 206 | e = __lookup_nat_cache(nm_i, nid); |
2dcf51ab JK |
207 | if (e) { |
208 | if (!get_nat_flag(e, IS_CHECKPOINTED) && | |
209 | !get_nat_flag(e, HAS_FSYNCED_INODE)) | |
210 | need = true; | |
211 | } | |
8b26ef98 | 212 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 213 | return need; |
e05df3b1 JK |
214 | } |
215 | ||
2dcf51ab | 216 | bool is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid) |
479f40c4 JK |
217 | { |
218 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
219 | struct nat_entry *e; | |
2dcf51ab | 220 | bool is_cp = true; |
479f40c4 | 221 | |
8b26ef98 | 222 | down_read(&nm_i->nat_tree_lock); |
2dcf51ab JK |
223 | e = __lookup_nat_cache(nm_i, nid); |
224 | if (e && !get_nat_flag(e, IS_CHECKPOINTED)) | |
225 | is_cp = false; | |
8b26ef98 | 226 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 227 | return is_cp; |
479f40c4 JK |
228 | } |
229 | ||
88bd02c9 | 230 | bool need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino) |
b6fe5873 JK |
231 | { |
232 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
233 | struct nat_entry *e; | |
88bd02c9 | 234 | bool need_update = true; |
b6fe5873 | 235 | |
8b26ef98 | 236 | down_read(&nm_i->nat_tree_lock); |
88bd02c9 JK |
237 | e = __lookup_nat_cache(nm_i, ino); |
238 | if (e && get_nat_flag(e, HAS_LAST_FSYNC) && | |
239 | (get_nat_flag(e, IS_CHECKPOINTED) || | |
240 | get_nat_flag(e, HAS_FSYNCED_INODE))) | |
241 | need_update = false; | |
8b26ef98 | 242 | up_read(&nm_i->nat_tree_lock); |
88bd02c9 | 243 | return need_update; |
b6fe5873 JK |
244 | } |
245 | ||
e05df3b1 JK |
246 | static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid) |
247 | { | |
248 | struct nat_entry *new; | |
249 | ||
80c54505 | 250 | new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_NOFS); |
9be32d72 | 251 | f2fs_radix_tree_insert(&nm_i->nat_root, nid, new); |
e05df3b1 JK |
252 | memset(new, 0, sizeof(struct nat_entry)); |
253 | nat_set_nid(new, nid); | |
88bd02c9 | 254 | nat_reset_flag(new); |
e05df3b1 JK |
255 | list_add_tail(&new->list, &nm_i->nat_entries); |
256 | nm_i->nat_cnt++; | |
257 | return new; | |
258 | } | |
259 | ||
1515aef0 | 260 | static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid, |
e05df3b1 JK |
261 | struct f2fs_nat_entry *ne) |
262 | { | |
1515aef0 | 263 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 264 | struct nat_entry *e; |
9be32d72 | 265 | |
e05df3b1 JK |
266 | e = __lookup_nat_cache(nm_i, nid); |
267 | if (!e) { | |
268 | e = grab_nat_entry(nm_i, nid); | |
94dac22e | 269 | node_info_from_raw_nat(&e->ni, ne); |
1515aef0 CY |
270 | } else { |
271 | f2fs_bug_on(sbi, nat_get_ino(e) != ne->ino || | |
272 | nat_get_blkaddr(e) != ne->block_addr || | |
273 | nat_get_version(e) != ne->version); | |
e05df3b1 | 274 | } |
e05df3b1 JK |
275 | } |
276 | ||
277 | static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, | |
479f40c4 | 278 | block_t new_blkaddr, bool fsync_done) |
e05df3b1 JK |
279 | { |
280 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
281 | struct nat_entry *e; | |
9be32d72 | 282 | |
8b26ef98 | 283 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
284 | e = __lookup_nat_cache(nm_i, ni->nid); |
285 | if (!e) { | |
286 | e = grab_nat_entry(nm_i, ni->nid); | |
5c27f4ee | 287 | copy_node_info(&e->ni, ni); |
9850cf4a | 288 | f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR); |
e05df3b1 JK |
289 | } else if (new_blkaddr == NEW_ADDR) { |
290 | /* | |
291 | * when nid is reallocated, | |
292 | * previous nat entry can be remained in nat cache. | |
293 | * So, reinitialize it with new information. | |
294 | */ | |
5c27f4ee | 295 | copy_node_info(&e->ni, ni); |
9850cf4a | 296 | f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR); |
e05df3b1 JK |
297 | } |
298 | ||
e05df3b1 | 299 | /* sanity check */ |
9850cf4a JK |
300 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr); |
301 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR && | |
e05df3b1 | 302 | new_blkaddr == NULL_ADDR); |
9850cf4a | 303 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR && |
e05df3b1 | 304 | new_blkaddr == NEW_ADDR); |
9850cf4a | 305 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != NEW_ADDR && |
e05df3b1 JK |
306 | nat_get_blkaddr(e) != NULL_ADDR && |
307 | new_blkaddr == NEW_ADDR); | |
308 | ||
e1c42045 | 309 | /* increment version no as node is removed */ |
e05df3b1 JK |
310 | if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) { |
311 | unsigned char version = nat_get_version(e); | |
312 | nat_set_version(e, inc_node_version(version)); | |
26834466 JK |
313 | |
314 | /* in order to reuse the nid */ | |
315 | if (nm_i->next_scan_nid > ni->nid) | |
316 | nm_i->next_scan_nid = ni->nid; | |
e05df3b1 JK |
317 | } |
318 | ||
319 | /* change address */ | |
320 | nat_set_blkaddr(e, new_blkaddr); | |
88bd02c9 JK |
321 | if (new_blkaddr == NEW_ADDR || new_blkaddr == NULL_ADDR) |
322 | set_nat_flag(e, IS_CHECKPOINTED, false); | |
e05df3b1 | 323 | __set_nat_cache_dirty(nm_i, e); |
479f40c4 JK |
324 | |
325 | /* update fsync_mark if its inode nat entry is still alive */ | |
d5b692b7 CY |
326 | if (ni->nid != ni->ino) |
327 | e = __lookup_nat_cache(nm_i, ni->ino); | |
88bd02c9 JK |
328 | if (e) { |
329 | if (fsync_done && ni->nid == ni->ino) | |
330 | set_nat_flag(e, HAS_FSYNCED_INODE, true); | |
331 | set_nat_flag(e, HAS_LAST_FSYNC, fsync_done); | |
332 | } | |
8b26ef98 | 333 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
334 | } |
335 | ||
4660f9c0 | 336 | int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) |
e05df3b1 JK |
337 | { |
338 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1b38dc8e | 339 | int nr = nr_shrink; |
e05df3b1 | 340 | |
1b38dc8e | 341 | if (!down_write_trylock(&nm_i->nat_tree_lock)) |
e05df3b1 JK |
342 | return 0; |
343 | ||
e05df3b1 JK |
344 | while (nr_shrink && !list_empty(&nm_i->nat_entries)) { |
345 | struct nat_entry *ne; | |
346 | ne = list_first_entry(&nm_i->nat_entries, | |
347 | struct nat_entry, list); | |
348 | __del_from_nat_cache(nm_i, ne); | |
349 | nr_shrink--; | |
350 | } | |
8b26ef98 | 351 | up_write(&nm_i->nat_tree_lock); |
1b38dc8e | 352 | return nr - nr_shrink; |
e05df3b1 JK |
353 | } |
354 | ||
0a8165d7 | 355 | /* |
e1c42045 | 356 | * This function always returns success |
e05df3b1 JK |
357 | */ |
358 | void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni) | |
359 | { | |
360 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
361 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 362 | struct f2fs_journal *journal = curseg->journal; |
e05df3b1 JK |
363 | nid_t start_nid = START_NID(nid); |
364 | struct f2fs_nat_block *nat_blk; | |
365 | struct page *page = NULL; | |
366 | struct f2fs_nat_entry ne; | |
367 | struct nat_entry *e; | |
368 | int i; | |
369 | ||
370 | ni->nid = nid; | |
371 | ||
372 | /* Check nat cache */ | |
8b26ef98 | 373 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
374 | e = __lookup_nat_cache(nm_i, nid); |
375 | if (e) { | |
376 | ni->ino = nat_get_ino(e); | |
377 | ni->blk_addr = nat_get_blkaddr(e); | |
378 | ni->version = nat_get_version(e); | |
1515aef0 | 379 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 | 380 | return; |
1515aef0 | 381 | } |
e05df3b1 | 382 | |
3547ea96 JK |
383 | memset(&ne, 0, sizeof(struct f2fs_nat_entry)); |
384 | ||
e05df3b1 | 385 | /* Check current segment summary */ |
b7ad7512 | 386 | down_read(&curseg->journal_rwsem); |
dfc08a12 | 387 | i = lookup_journal_in_cursum(journal, NAT_JOURNAL, nid, 0); |
e05df3b1 | 388 | if (i >= 0) { |
dfc08a12 | 389 | ne = nat_in_journal(journal, i); |
e05df3b1 JK |
390 | node_info_from_raw_nat(ni, &ne); |
391 | } | |
b7ad7512 | 392 | up_read(&curseg->journal_rwsem); |
e05df3b1 JK |
393 | if (i >= 0) |
394 | goto cache; | |
395 | ||
396 | /* Fill node_info from nat page */ | |
397 | page = get_current_nat_page(sbi, start_nid); | |
398 | nat_blk = (struct f2fs_nat_block *)page_address(page); | |
399 | ne = nat_blk->entries[nid - start_nid]; | |
400 | node_info_from_raw_nat(ni, &ne); | |
401 | f2fs_put_page(page, 1); | |
402 | cache: | |
1515aef0 | 403 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 | 404 | /* cache nat entry */ |
1515aef0 CY |
405 | down_write(&nm_i->nat_tree_lock); |
406 | cache_nat_entry(sbi, nid, &ne); | |
a5131193 | 407 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
408 | } |
409 | ||
3cf45747 CY |
410 | pgoff_t get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs) |
411 | { | |
412 | const long direct_index = ADDRS_PER_INODE(dn->inode); | |
413 | const long direct_blks = ADDRS_PER_BLOCK; | |
414 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
415 | unsigned int skipped_unit = ADDRS_PER_BLOCK; | |
416 | int cur_level = dn->cur_level; | |
417 | int max_level = dn->max_level; | |
418 | pgoff_t base = 0; | |
419 | ||
420 | if (!dn->max_level) | |
421 | return pgofs + 1; | |
422 | ||
423 | while (max_level-- > cur_level) | |
424 | skipped_unit *= NIDS_PER_BLOCK; | |
425 | ||
426 | switch (dn->max_level) { | |
427 | case 3: | |
428 | base += 2 * indirect_blks; | |
429 | case 2: | |
430 | base += 2 * direct_blks; | |
431 | case 1: | |
432 | base += direct_index; | |
433 | break; | |
434 | default: | |
435 | f2fs_bug_on(F2FS_I_SB(dn->inode), 1); | |
436 | } | |
437 | ||
438 | return ((pgofs - base) / skipped_unit + 1) * skipped_unit + base; | |
439 | } | |
440 | ||
0a8165d7 | 441 | /* |
e05df3b1 JK |
442 | * The maximum depth is four. |
443 | * Offset[0] will have raw inode offset. | |
444 | */ | |
81ca7350 | 445 | static int get_node_path(struct inode *inode, long block, |
de93653f | 446 | int offset[4], unsigned int noffset[4]) |
e05df3b1 | 447 | { |
81ca7350 | 448 | const long direct_index = ADDRS_PER_INODE(inode); |
e05df3b1 JK |
449 | const long direct_blks = ADDRS_PER_BLOCK; |
450 | const long dptrs_per_blk = NIDS_PER_BLOCK; | |
451 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
452 | const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK; | |
453 | int n = 0; | |
454 | int level = 0; | |
455 | ||
456 | noffset[0] = 0; | |
457 | ||
458 | if (block < direct_index) { | |
25c0a6e5 | 459 | offset[n] = block; |
e05df3b1 JK |
460 | goto got; |
461 | } | |
462 | block -= direct_index; | |
463 | if (block < direct_blks) { | |
464 | offset[n++] = NODE_DIR1_BLOCK; | |
465 | noffset[n] = 1; | |
25c0a6e5 | 466 | offset[n] = block; |
e05df3b1 JK |
467 | level = 1; |
468 | goto got; | |
469 | } | |
470 | block -= direct_blks; | |
471 | if (block < direct_blks) { | |
472 | offset[n++] = NODE_DIR2_BLOCK; | |
473 | noffset[n] = 2; | |
25c0a6e5 | 474 | offset[n] = block; |
e05df3b1 JK |
475 | level = 1; |
476 | goto got; | |
477 | } | |
478 | block -= direct_blks; | |
479 | if (block < indirect_blks) { | |
480 | offset[n++] = NODE_IND1_BLOCK; | |
481 | noffset[n] = 3; | |
482 | offset[n++] = block / direct_blks; | |
483 | noffset[n] = 4 + offset[n - 1]; | |
25c0a6e5 | 484 | offset[n] = block % direct_blks; |
e05df3b1 JK |
485 | level = 2; |
486 | goto got; | |
487 | } | |
488 | block -= indirect_blks; | |
489 | if (block < indirect_blks) { | |
490 | offset[n++] = NODE_IND2_BLOCK; | |
491 | noffset[n] = 4 + dptrs_per_blk; | |
492 | offset[n++] = block / direct_blks; | |
493 | noffset[n] = 5 + dptrs_per_blk + offset[n - 1]; | |
25c0a6e5 | 494 | offset[n] = block % direct_blks; |
e05df3b1 JK |
495 | level = 2; |
496 | goto got; | |
497 | } | |
498 | block -= indirect_blks; | |
499 | if (block < dindirect_blks) { | |
500 | offset[n++] = NODE_DIND_BLOCK; | |
501 | noffset[n] = 5 + (dptrs_per_blk * 2); | |
502 | offset[n++] = block / indirect_blks; | |
503 | noffset[n] = 6 + (dptrs_per_blk * 2) + | |
504 | offset[n - 1] * (dptrs_per_blk + 1); | |
505 | offset[n++] = (block / direct_blks) % dptrs_per_blk; | |
506 | noffset[n] = 7 + (dptrs_per_blk * 2) + | |
507 | offset[n - 2] * (dptrs_per_blk + 1) + | |
508 | offset[n - 1]; | |
25c0a6e5 | 509 | offset[n] = block % direct_blks; |
e05df3b1 JK |
510 | level = 3; |
511 | goto got; | |
512 | } else { | |
513 | BUG(); | |
514 | } | |
515 | got: | |
516 | return level; | |
517 | } | |
518 | ||
519 | /* | |
520 | * Caller should call f2fs_put_dnode(dn). | |
4f4124d0 CY |
521 | * Also, it should grab and release a rwsem by calling f2fs_lock_op() and |
522 | * f2fs_unlock_op() only if ro is not set RDONLY_NODE. | |
39936837 | 523 | * In the case of RDONLY_NODE, we don't need to care about mutex. |
e05df3b1 | 524 | */ |
266e97a8 | 525 | int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode) |
e05df3b1 | 526 | { |
4081363f | 527 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 | 528 | struct page *npage[4]; |
f1a3b98e | 529 | struct page *parent = NULL; |
e05df3b1 JK |
530 | int offset[4]; |
531 | unsigned int noffset[4]; | |
532 | nid_t nids[4]; | |
3cf45747 | 533 | int level, i = 0; |
e05df3b1 JK |
534 | int err = 0; |
535 | ||
81ca7350 | 536 | level = get_node_path(dn->inode, index, offset, noffset); |
e05df3b1 JK |
537 | |
538 | nids[0] = dn->inode->i_ino; | |
1646cfac | 539 | npage[0] = dn->inode_page; |
e05df3b1 | 540 | |
1646cfac JK |
541 | if (!npage[0]) { |
542 | npage[0] = get_node_page(sbi, nids[0]); | |
543 | if (IS_ERR(npage[0])) | |
544 | return PTR_ERR(npage[0]); | |
545 | } | |
f1a3b98e JK |
546 | |
547 | /* if inline_data is set, should not report any block indices */ | |
548 | if (f2fs_has_inline_data(dn->inode) && index) { | |
76629165 | 549 | err = -ENOENT; |
f1a3b98e JK |
550 | f2fs_put_page(npage[0], 1); |
551 | goto release_out; | |
552 | } | |
553 | ||
e05df3b1 | 554 | parent = npage[0]; |
52c2db3f CL |
555 | if (level != 0) |
556 | nids[1] = get_nid(parent, offset[0], true); | |
e05df3b1 JK |
557 | dn->inode_page = npage[0]; |
558 | dn->inode_page_locked = true; | |
559 | ||
560 | /* get indirect or direct nodes */ | |
561 | for (i = 1; i <= level; i++) { | |
562 | bool done = false; | |
563 | ||
266e97a8 | 564 | if (!nids[i] && mode == ALLOC_NODE) { |
e05df3b1 JK |
565 | /* alloc new node */ |
566 | if (!alloc_nid(sbi, &(nids[i]))) { | |
e05df3b1 JK |
567 | err = -ENOSPC; |
568 | goto release_pages; | |
569 | } | |
570 | ||
571 | dn->nid = nids[i]; | |
8ae8f162 | 572 | npage[i] = new_node_page(dn, noffset[i], NULL); |
e05df3b1 JK |
573 | if (IS_ERR(npage[i])) { |
574 | alloc_nid_failed(sbi, nids[i]); | |
e05df3b1 JK |
575 | err = PTR_ERR(npage[i]); |
576 | goto release_pages; | |
577 | } | |
578 | ||
579 | set_nid(parent, offset[i - 1], nids[i], i == 1); | |
580 | alloc_nid_done(sbi, nids[i]); | |
e05df3b1 | 581 | done = true; |
266e97a8 | 582 | } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) { |
e05df3b1 JK |
583 | npage[i] = get_node_page_ra(parent, offset[i - 1]); |
584 | if (IS_ERR(npage[i])) { | |
585 | err = PTR_ERR(npage[i]); | |
586 | goto release_pages; | |
587 | } | |
588 | done = true; | |
589 | } | |
590 | if (i == 1) { | |
591 | dn->inode_page_locked = false; | |
592 | unlock_page(parent); | |
593 | } else { | |
594 | f2fs_put_page(parent, 1); | |
595 | } | |
596 | ||
597 | if (!done) { | |
598 | npage[i] = get_node_page(sbi, nids[i]); | |
599 | if (IS_ERR(npage[i])) { | |
600 | err = PTR_ERR(npage[i]); | |
601 | f2fs_put_page(npage[0], 0); | |
602 | goto release_out; | |
603 | } | |
604 | } | |
605 | if (i < level) { | |
606 | parent = npage[i]; | |
607 | nids[i + 1] = get_nid(parent, offset[i], false); | |
608 | } | |
609 | } | |
610 | dn->nid = nids[level]; | |
611 | dn->ofs_in_node = offset[level]; | |
612 | dn->node_page = npage[level]; | |
613 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
614 | return 0; | |
615 | ||
616 | release_pages: | |
617 | f2fs_put_page(parent, 1); | |
618 | if (i > 1) | |
619 | f2fs_put_page(npage[0], 0); | |
620 | release_out: | |
621 | dn->inode_page = NULL; | |
622 | dn->node_page = NULL; | |
3cf45747 CY |
623 | if (err == -ENOENT) { |
624 | dn->cur_level = i; | |
625 | dn->max_level = level; | |
626 | } | |
e05df3b1 JK |
627 | return err; |
628 | } | |
629 | ||
630 | static void truncate_node(struct dnode_of_data *dn) | |
631 | { | |
4081363f | 632 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
633 | struct node_info ni; |
634 | ||
635 | get_node_info(sbi, dn->nid, &ni); | |
71e9fec5 | 636 | if (dn->inode->i_blocks == 0) { |
9850cf4a | 637 | f2fs_bug_on(sbi, ni.blk_addr != NULL_ADDR); |
71e9fec5 JK |
638 | goto invalidate; |
639 | } | |
9850cf4a | 640 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
e05df3b1 | 641 | |
e05df3b1 | 642 | /* Deallocate node address */ |
71e9fec5 | 643 | invalidate_blocks(sbi, ni.blk_addr); |
ef86d709 | 644 | dec_valid_node_count(sbi, dn->inode); |
479f40c4 | 645 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
e05df3b1 JK |
646 | |
647 | if (dn->nid == dn->inode->i_ino) { | |
648 | remove_orphan_inode(sbi, dn->nid); | |
649 | dec_valid_inode_count(sbi); | |
650 | } else { | |
651 | sync_inode_page(dn); | |
652 | } | |
71e9fec5 | 653 | invalidate: |
e05df3b1 | 654 | clear_node_page_dirty(dn->node_page); |
caf0047e | 655 | set_sbi_flag(sbi, SBI_IS_DIRTY); |
e05df3b1 JK |
656 | |
657 | f2fs_put_page(dn->node_page, 1); | |
bf39c00a JK |
658 | |
659 | invalidate_mapping_pages(NODE_MAPPING(sbi), | |
660 | dn->node_page->index, dn->node_page->index); | |
661 | ||
e05df3b1 | 662 | dn->node_page = NULL; |
51dd6249 | 663 | trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr); |
e05df3b1 JK |
664 | } |
665 | ||
666 | static int truncate_dnode(struct dnode_of_data *dn) | |
667 | { | |
e05df3b1 JK |
668 | struct page *page; |
669 | ||
670 | if (dn->nid == 0) | |
671 | return 1; | |
672 | ||
673 | /* get direct node */ | |
4081363f | 674 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
e05df3b1 JK |
675 | if (IS_ERR(page) && PTR_ERR(page) == -ENOENT) |
676 | return 1; | |
677 | else if (IS_ERR(page)) | |
678 | return PTR_ERR(page); | |
679 | ||
680 | /* Make dnode_of_data for parameter */ | |
681 | dn->node_page = page; | |
682 | dn->ofs_in_node = 0; | |
683 | truncate_data_blocks(dn); | |
684 | truncate_node(dn); | |
685 | return 1; | |
686 | } | |
687 | ||
688 | static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs, | |
689 | int ofs, int depth) | |
690 | { | |
e05df3b1 JK |
691 | struct dnode_of_data rdn = *dn; |
692 | struct page *page; | |
693 | struct f2fs_node *rn; | |
694 | nid_t child_nid; | |
695 | unsigned int child_nofs; | |
696 | int freed = 0; | |
697 | int i, ret; | |
698 | ||
699 | if (dn->nid == 0) | |
700 | return NIDS_PER_BLOCK + 1; | |
701 | ||
51dd6249 NJ |
702 | trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr); |
703 | ||
4081363f | 704 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
51dd6249 NJ |
705 | if (IS_ERR(page)) { |
706 | trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page)); | |
e05df3b1 | 707 | return PTR_ERR(page); |
51dd6249 | 708 | } |
e05df3b1 | 709 | |
45590710 | 710 | rn = F2FS_NODE(page); |
e05df3b1 JK |
711 | if (depth < 3) { |
712 | for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) { | |
713 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
714 | if (child_nid == 0) | |
715 | continue; | |
716 | rdn.nid = child_nid; | |
717 | ret = truncate_dnode(&rdn); | |
718 | if (ret < 0) | |
719 | goto out_err; | |
12719ae1 JK |
720 | if (set_nid(page, i, 0, false)) |
721 | dn->node_changed = true; | |
e05df3b1 JK |
722 | } |
723 | } else { | |
724 | child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1; | |
725 | for (i = ofs; i < NIDS_PER_BLOCK; i++) { | |
726 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
727 | if (child_nid == 0) { | |
728 | child_nofs += NIDS_PER_BLOCK + 1; | |
729 | continue; | |
730 | } | |
731 | rdn.nid = child_nid; | |
732 | ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1); | |
733 | if (ret == (NIDS_PER_BLOCK + 1)) { | |
12719ae1 JK |
734 | if (set_nid(page, i, 0, false)) |
735 | dn->node_changed = true; | |
e05df3b1 JK |
736 | child_nofs += ret; |
737 | } else if (ret < 0 && ret != -ENOENT) { | |
738 | goto out_err; | |
739 | } | |
740 | } | |
741 | freed = child_nofs; | |
742 | } | |
743 | ||
744 | if (!ofs) { | |
745 | /* remove current indirect node */ | |
746 | dn->node_page = page; | |
747 | truncate_node(dn); | |
748 | freed++; | |
749 | } else { | |
750 | f2fs_put_page(page, 1); | |
751 | } | |
51dd6249 | 752 | trace_f2fs_truncate_nodes_exit(dn->inode, freed); |
e05df3b1 JK |
753 | return freed; |
754 | ||
755 | out_err: | |
756 | f2fs_put_page(page, 1); | |
51dd6249 | 757 | trace_f2fs_truncate_nodes_exit(dn->inode, ret); |
e05df3b1 JK |
758 | return ret; |
759 | } | |
760 | ||
761 | static int truncate_partial_nodes(struct dnode_of_data *dn, | |
762 | struct f2fs_inode *ri, int *offset, int depth) | |
763 | { | |
e05df3b1 JK |
764 | struct page *pages[2]; |
765 | nid_t nid[3]; | |
766 | nid_t child_nid; | |
767 | int err = 0; | |
768 | int i; | |
769 | int idx = depth - 2; | |
770 | ||
771 | nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); | |
772 | if (!nid[0]) | |
773 | return 0; | |
774 | ||
775 | /* get indirect nodes in the path */ | |
a225dca3 | 776 | for (i = 0; i < idx + 1; i++) { |
e1c42045 | 777 | /* reference count'll be increased */ |
4081363f | 778 | pages[i] = get_node_page(F2FS_I_SB(dn->inode), nid[i]); |
e05df3b1 | 779 | if (IS_ERR(pages[i])) { |
e05df3b1 | 780 | err = PTR_ERR(pages[i]); |
a225dca3 | 781 | idx = i - 1; |
e05df3b1 JK |
782 | goto fail; |
783 | } | |
784 | nid[i + 1] = get_nid(pages[i], offset[i + 1], false); | |
785 | } | |
786 | ||
787 | /* free direct nodes linked to a partial indirect node */ | |
a225dca3 | 788 | for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) { |
e05df3b1 JK |
789 | child_nid = get_nid(pages[idx], i, false); |
790 | if (!child_nid) | |
791 | continue; | |
792 | dn->nid = child_nid; | |
793 | err = truncate_dnode(dn); | |
794 | if (err < 0) | |
795 | goto fail; | |
12719ae1 JK |
796 | if (set_nid(pages[idx], i, 0, false)) |
797 | dn->node_changed = true; | |
e05df3b1 JK |
798 | } |
799 | ||
a225dca3 | 800 | if (offset[idx + 1] == 0) { |
e05df3b1 JK |
801 | dn->node_page = pages[idx]; |
802 | dn->nid = nid[idx]; | |
803 | truncate_node(dn); | |
804 | } else { | |
805 | f2fs_put_page(pages[idx], 1); | |
806 | } | |
807 | offset[idx]++; | |
a225dca3 | 808 | offset[idx + 1] = 0; |
809 | idx--; | |
e05df3b1 | 810 | fail: |
a225dca3 | 811 | for (i = idx; i >= 0; i--) |
e05df3b1 | 812 | f2fs_put_page(pages[i], 1); |
51dd6249 NJ |
813 | |
814 | trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err); | |
815 | ||
e05df3b1 JK |
816 | return err; |
817 | } | |
818 | ||
0a8165d7 | 819 | /* |
e05df3b1 JK |
820 | * All the block addresses of data and nodes should be nullified. |
821 | */ | |
822 | int truncate_inode_blocks(struct inode *inode, pgoff_t from) | |
823 | { | |
4081363f | 824 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
e05df3b1 JK |
825 | int err = 0, cont = 1; |
826 | int level, offset[4], noffset[4]; | |
7dd690c8 | 827 | unsigned int nofs = 0; |
58bfaf44 | 828 | struct f2fs_inode *ri; |
e05df3b1 JK |
829 | struct dnode_of_data dn; |
830 | struct page *page; | |
831 | ||
51dd6249 NJ |
832 | trace_f2fs_truncate_inode_blocks_enter(inode, from); |
833 | ||
81ca7350 | 834 | level = get_node_path(inode, from, offset, noffset); |
afcb7ca0 | 835 | restart: |
e05df3b1 | 836 | page = get_node_page(sbi, inode->i_ino); |
51dd6249 NJ |
837 | if (IS_ERR(page)) { |
838 | trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page)); | |
e05df3b1 | 839 | return PTR_ERR(page); |
51dd6249 | 840 | } |
e05df3b1 JK |
841 | |
842 | set_new_dnode(&dn, inode, page, NULL, 0); | |
843 | unlock_page(page); | |
844 | ||
58bfaf44 | 845 | ri = F2FS_INODE(page); |
e05df3b1 JK |
846 | switch (level) { |
847 | case 0: | |
848 | case 1: | |
849 | nofs = noffset[1]; | |
850 | break; | |
851 | case 2: | |
852 | nofs = noffset[1]; | |
853 | if (!offset[level - 1]) | |
854 | goto skip_partial; | |
58bfaf44 | 855 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
856 | if (err < 0 && err != -ENOENT) |
857 | goto fail; | |
858 | nofs += 1 + NIDS_PER_BLOCK; | |
859 | break; | |
860 | case 3: | |
861 | nofs = 5 + 2 * NIDS_PER_BLOCK; | |
862 | if (!offset[level - 1]) | |
863 | goto skip_partial; | |
58bfaf44 | 864 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
865 | if (err < 0 && err != -ENOENT) |
866 | goto fail; | |
867 | break; | |
868 | default: | |
869 | BUG(); | |
870 | } | |
871 | ||
872 | skip_partial: | |
873 | while (cont) { | |
58bfaf44 | 874 | dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); |
e05df3b1 JK |
875 | switch (offset[0]) { |
876 | case NODE_DIR1_BLOCK: | |
877 | case NODE_DIR2_BLOCK: | |
878 | err = truncate_dnode(&dn); | |
879 | break; | |
880 | ||
881 | case NODE_IND1_BLOCK: | |
882 | case NODE_IND2_BLOCK: | |
883 | err = truncate_nodes(&dn, nofs, offset[1], 2); | |
884 | break; | |
885 | ||
886 | case NODE_DIND_BLOCK: | |
887 | err = truncate_nodes(&dn, nofs, offset[1], 3); | |
888 | cont = 0; | |
889 | break; | |
890 | ||
891 | default: | |
892 | BUG(); | |
893 | } | |
894 | if (err < 0 && err != -ENOENT) | |
895 | goto fail; | |
896 | if (offset[1] == 0 && | |
58bfaf44 | 897 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) { |
e05df3b1 | 898 | lock_page(page); |
4ef51a8f | 899 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
900 | f2fs_put_page(page, 1); |
901 | goto restart; | |
902 | } | |
fec1d657 | 903 | f2fs_wait_on_page_writeback(page, NODE, true); |
58bfaf44 | 904 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0; |
e05df3b1 JK |
905 | set_page_dirty(page); |
906 | unlock_page(page); | |
907 | } | |
908 | offset[1] = 0; | |
909 | offset[0]++; | |
910 | nofs += err; | |
911 | } | |
912 | fail: | |
913 | f2fs_put_page(page, 0); | |
51dd6249 | 914 | trace_f2fs_truncate_inode_blocks_exit(inode, err); |
e05df3b1 JK |
915 | return err > 0 ? 0 : err; |
916 | } | |
917 | ||
4f16fb0f JK |
918 | int truncate_xattr_node(struct inode *inode, struct page *page) |
919 | { | |
4081363f | 920 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
4f16fb0f JK |
921 | nid_t nid = F2FS_I(inode)->i_xattr_nid; |
922 | struct dnode_of_data dn; | |
923 | struct page *npage; | |
924 | ||
925 | if (!nid) | |
926 | return 0; | |
927 | ||
928 | npage = get_node_page(sbi, nid); | |
929 | if (IS_ERR(npage)) | |
930 | return PTR_ERR(npage); | |
931 | ||
932 | F2FS_I(inode)->i_xattr_nid = 0; | |
65985d93 JK |
933 | |
934 | /* need to do checkpoint during fsync */ | |
935 | F2FS_I(inode)->xattr_ver = cur_cp_version(F2FS_CKPT(sbi)); | |
936 | ||
4f16fb0f JK |
937 | set_new_dnode(&dn, inode, page, npage, nid); |
938 | ||
939 | if (page) | |
01d2d1aa | 940 | dn.inode_page_locked = true; |
4f16fb0f JK |
941 | truncate_node(&dn); |
942 | return 0; | |
943 | } | |
944 | ||
39936837 | 945 | /* |
4f4124d0 CY |
946 | * Caller should grab and release a rwsem by calling f2fs_lock_op() and |
947 | * f2fs_unlock_op(). | |
39936837 | 948 | */ |
13ec7297 | 949 | int remove_inode_page(struct inode *inode) |
e05df3b1 | 950 | { |
e05df3b1 | 951 | struct dnode_of_data dn; |
13ec7297 | 952 | int err; |
e05df3b1 | 953 | |
c2e69583 | 954 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); |
13ec7297 CY |
955 | err = get_dnode_of_data(&dn, 0, LOOKUP_NODE); |
956 | if (err) | |
957 | return err; | |
e05df3b1 | 958 | |
13ec7297 CY |
959 | err = truncate_xattr_node(inode, dn.inode_page); |
960 | if (err) { | |
c2e69583 | 961 | f2fs_put_dnode(&dn); |
13ec7297 | 962 | return err; |
e05df3b1 | 963 | } |
c2e69583 JK |
964 | |
965 | /* remove potential inline_data blocks */ | |
966 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
967 | S_ISLNK(inode->i_mode)) | |
968 | truncate_data_blocks_range(&dn, 1); | |
969 | ||
e1c42045 | 970 | /* 0 is possible, after f2fs_new_inode() has failed */ |
9850cf4a JK |
971 | f2fs_bug_on(F2FS_I_SB(inode), |
972 | inode->i_blocks != 0 && inode->i_blocks != 1); | |
c2e69583 JK |
973 | |
974 | /* will put inode & node pages */ | |
71e9fec5 | 975 | truncate_node(&dn); |
13ec7297 | 976 | return 0; |
e05df3b1 JK |
977 | } |
978 | ||
a014e037 | 979 | struct page *new_inode_page(struct inode *inode) |
e05df3b1 | 980 | { |
e05df3b1 JK |
981 | struct dnode_of_data dn; |
982 | ||
983 | /* allocate inode page for new inode */ | |
984 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); | |
44a83ff6 JK |
985 | |
986 | /* caller should f2fs_put_page(page, 1); */ | |
8ae8f162 | 987 | return new_node_page(&dn, 0, NULL); |
e05df3b1 JK |
988 | } |
989 | ||
8ae8f162 JK |
990 | struct page *new_node_page(struct dnode_of_data *dn, |
991 | unsigned int ofs, struct page *ipage) | |
e05df3b1 | 992 | { |
4081363f | 993 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
994 | struct node_info old_ni, new_ni; |
995 | struct page *page; | |
996 | int err; | |
997 | ||
6bacf52f | 998 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
e05df3b1 JK |
999 | return ERR_PTR(-EPERM); |
1000 | ||
54b591df | 1001 | page = grab_cache_page(NODE_MAPPING(sbi), dn->nid); |
e05df3b1 JK |
1002 | if (!page) |
1003 | return ERR_PTR(-ENOMEM); | |
1004 | ||
6bacf52f | 1005 | if (unlikely(!inc_valid_node_count(sbi, dn->inode))) { |
9c02740c JK |
1006 | err = -ENOSPC; |
1007 | goto fail; | |
1008 | } | |
e05df3b1 | 1009 | |
9c02740c | 1010 | get_node_info(sbi, dn->nid, &old_ni); |
e05df3b1 JK |
1011 | |
1012 | /* Reinitialize old_ni with new node page */ | |
9850cf4a | 1013 | f2fs_bug_on(sbi, old_ni.blk_addr != NULL_ADDR); |
e05df3b1 JK |
1014 | new_ni = old_ni; |
1015 | new_ni.ino = dn->inode->i_ino; | |
479f40c4 | 1016 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
9c02740c | 1017 | |
fec1d657 | 1018 | f2fs_wait_on_page_writeback(page, NODE, true); |
9c02740c | 1019 | fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true); |
398b1ac5 | 1020 | set_cold_node(dn->inode, page); |
9c02740c | 1021 | SetPageUptodate(page); |
12719ae1 JK |
1022 | if (set_page_dirty(page)) |
1023 | dn->node_changed = true; | |
e05df3b1 | 1024 | |
4bc8e9bc | 1025 | if (f2fs_has_xattr_block(ofs)) |
479bd73a JK |
1026 | F2FS_I(dn->inode)->i_xattr_nid = dn->nid; |
1027 | ||
e05df3b1 | 1028 | dn->node_page = page; |
8ae8f162 JK |
1029 | if (ipage) |
1030 | update_inode(dn->inode, ipage); | |
1031 | else | |
1032 | sync_inode_page(dn); | |
e05df3b1 JK |
1033 | if (ofs == 0) |
1034 | inc_valid_inode_count(sbi); | |
1035 | ||
1036 | return page; | |
1037 | ||
1038 | fail: | |
71e9fec5 | 1039 | clear_node_page_dirty(page); |
e05df3b1 JK |
1040 | f2fs_put_page(page, 1); |
1041 | return ERR_PTR(err); | |
1042 | } | |
1043 | ||
56ae674c JK |
1044 | /* |
1045 | * Caller should do after getting the following values. | |
1046 | * 0: f2fs_put_page(page, 0) | |
86531d6b | 1047 | * LOCKED_PAGE or error: f2fs_put_page(page, 1) |
56ae674c | 1048 | */ |
93dfe2ac | 1049 | static int read_node_page(struct page *page, int rw) |
e05df3b1 | 1050 | { |
4081363f | 1051 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1052 | struct node_info ni; |
cf04e8eb | 1053 | struct f2fs_io_info fio = { |
05ca3632 | 1054 | .sbi = sbi, |
cf04e8eb JK |
1055 | .type = NODE, |
1056 | .rw = rw, | |
05ca3632 | 1057 | .page = page, |
4375a336 | 1058 | .encrypted_page = NULL, |
cf04e8eb | 1059 | }; |
e05df3b1 JK |
1060 | |
1061 | get_node_info(sbi, page->index, &ni); | |
1062 | ||
6bacf52f | 1063 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
2bca1e23 | 1064 | ClearPageUptodate(page); |
e05df3b1 | 1065 | return -ENOENT; |
393ff91f JK |
1066 | } |
1067 | ||
56ae674c JK |
1068 | if (PageUptodate(page)) |
1069 | return LOCKED_PAGE; | |
393ff91f | 1070 | |
7a9d7548 | 1071 | fio.new_blkaddr = fio.old_blkaddr = ni.blk_addr; |
05ca3632 | 1072 | return f2fs_submit_page_bio(&fio); |
e05df3b1 JK |
1073 | } |
1074 | ||
0a8165d7 | 1075 | /* |
e05df3b1 JK |
1076 | * Readahead a node page |
1077 | */ | |
1078 | void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid) | |
1079 | { | |
e05df3b1 | 1080 | struct page *apage; |
56ae674c | 1081 | int err; |
e05df3b1 | 1082 | |
e8458725 CY |
1083 | if (!nid) |
1084 | return; | |
1085 | f2fs_bug_on(sbi, check_nid_range(sbi, nid)); | |
1086 | ||
4ef51a8f | 1087 | apage = find_get_page(NODE_MAPPING(sbi), nid); |
393ff91f JK |
1088 | if (apage && PageUptodate(apage)) { |
1089 | f2fs_put_page(apage, 0); | |
1090 | return; | |
1091 | } | |
e05df3b1 JK |
1092 | f2fs_put_page(apage, 0); |
1093 | ||
4ef51a8f | 1094 | apage = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1095 | if (!apage) |
1096 | return; | |
1097 | ||
56ae674c | 1098 | err = read_node_page(apage, READA); |
86531d6b | 1099 | f2fs_put_page(apage, err ? 1 : 0); |
e05df3b1 JK |
1100 | } |
1101 | ||
0e022ea8 CY |
1102 | /* |
1103 | * readahead MAX_RA_NODE number of node pages. | |
1104 | */ | |
1105 | void ra_node_pages(struct page *parent, int start) | |
e05df3b1 | 1106 | { |
0e022ea8 CY |
1107 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); |
1108 | struct blk_plug plug; | |
1109 | int i, end; | |
1110 | nid_t nid; | |
4aa69d56 | 1111 | |
0e022ea8 | 1112 | blk_start_plug(&plug); |
e05df3b1 | 1113 | |
0e022ea8 CY |
1114 | /* Then, try readahead for siblings of the desired node */ |
1115 | end = start + MAX_RA_NODE; | |
1116 | end = min(end, NIDS_PER_BLOCK); | |
1117 | for (i = start; i < end; i++) { | |
1118 | nid = get_nid(parent, i, false); | |
1119 | ra_node_page(sbi, nid); | |
86531d6b | 1120 | } |
e05df3b1 | 1121 | |
0e022ea8 | 1122 | blk_finish_plug(&plug); |
e05df3b1 JK |
1123 | } |
1124 | ||
0e022ea8 CY |
1125 | struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid, |
1126 | struct page *parent, int start) | |
e05df3b1 | 1127 | { |
e05df3b1 | 1128 | struct page *page; |
0e022ea8 | 1129 | int err; |
e05df3b1 | 1130 | |
e05df3b1 JK |
1131 | if (!nid) |
1132 | return ERR_PTR(-ENOENT); | |
e8458725 | 1133 | f2fs_bug_on(sbi, check_nid_range(sbi, nid)); |
afcb7ca0 | 1134 | repeat: |
4ef51a8f | 1135 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1136 | if (!page) |
1137 | return ERR_PTR(-ENOMEM); | |
1138 | ||
66d36a29 | 1139 | err = read_node_page(page, READ_SYNC); |
86531d6b JK |
1140 | if (err < 0) { |
1141 | f2fs_put_page(page, 1); | |
e05df3b1 | 1142 | return ERR_PTR(err); |
86531d6b | 1143 | } else if (err == LOCKED_PAGE) { |
56ae674c | 1144 | goto page_hit; |
86531d6b | 1145 | } |
e05df3b1 | 1146 | |
0e022ea8 CY |
1147 | if (parent) |
1148 | ra_node_pages(parent, start + 1); | |
c718379b | 1149 | |
e05df3b1 | 1150 | lock_page(page); |
0e022ea8 | 1151 | |
e1c51b9f CY |
1152 | if (unlikely(!PageUptodate(page))) { |
1153 | f2fs_put_page(page, 1); | |
1154 | return ERR_PTR(-EIO); | |
1155 | } | |
4ef51a8f | 1156 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1157 | f2fs_put_page(page, 1); |
1158 | goto repeat; | |
1159 | } | |
e0f56cb4 | 1160 | page_hit: |
e1c51b9f | 1161 | f2fs_bug_on(sbi, nid != nid_of_node(page)); |
e05df3b1 JK |
1162 | return page; |
1163 | } | |
1164 | ||
0e022ea8 CY |
1165 | struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) |
1166 | { | |
1167 | return __get_node_page(sbi, nid, NULL, 0); | |
1168 | } | |
1169 | ||
1170 | struct page *get_node_page_ra(struct page *parent, int start) | |
1171 | { | |
1172 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); | |
1173 | nid_t nid = get_nid(parent, start, false); | |
1174 | ||
1175 | return __get_node_page(sbi, nid, parent, start); | |
1176 | } | |
1177 | ||
e05df3b1 JK |
1178 | void sync_inode_page(struct dnode_of_data *dn) |
1179 | { | |
12719ae1 JK |
1180 | int ret = 0; |
1181 | ||
e05df3b1 | 1182 | if (IS_INODE(dn->node_page) || dn->inode_page == dn->node_page) { |
12719ae1 | 1183 | ret = update_inode(dn->inode, dn->node_page); |
e05df3b1 JK |
1184 | } else if (dn->inode_page) { |
1185 | if (!dn->inode_page_locked) | |
1186 | lock_page(dn->inode_page); | |
12719ae1 | 1187 | ret = update_inode(dn->inode, dn->inode_page); |
e05df3b1 JK |
1188 | if (!dn->inode_page_locked) |
1189 | unlock_page(dn->inode_page); | |
1190 | } else { | |
12719ae1 | 1191 | ret = update_inode_page(dn->inode); |
e05df3b1 | 1192 | } |
12719ae1 | 1193 | dn->node_changed = ret ? true: false; |
e05df3b1 JK |
1194 | } |
1195 | ||
2049d4fc JK |
1196 | static void flush_inline_data(struct f2fs_sb_info *sbi, nid_t ino) |
1197 | { | |
1198 | struct inode *inode; | |
1199 | struct page *page; | |
1200 | ||
1201 | /* should flush inline_data before evict_inode */ | |
1202 | inode = ilookup(sbi->sb, ino); | |
1203 | if (!inode) | |
1204 | return; | |
1205 | ||
19c7377b | 1206 | page = pagecache_get_page(inode->i_mapping, 0, FGP_NOWAIT, 0); |
2049d4fc JK |
1207 | if (!page) |
1208 | goto iput_out; | |
1209 | ||
19c7377b CY |
1210 | if (!trylock_page(page)) |
1211 | goto release_out; | |
1212 | ||
2049d4fc JK |
1213 | if (!PageUptodate(page)) |
1214 | goto page_out; | |
1215 | ||
1216 | if (!PageDirty(page)) | |
1217 | goto page_out; | |
1218 | ||
1219 | if (!clear_page_dirty_for_io(page)) | |
1220 | goto page_out; | |
1221 | ||
1222 | if (!f2fs_write_inline_data(inode, page)) | |
1223 | inode_dec_dirty_pages(inode); | |
1224 | else | |
1225 | set_page_dirty(page); | |
1226 | page_out: | |
19c7377b CY |
1227 | unlock_page(page); |
1228 | release_out: | |
1229 | f2fs_put_page(page, 0); | |
2049d4fc JK |
1230 | iput_out: |
1231 | iput(inode); | |
1232 | } | |
1233 | ||
e05df3b1 JK |
1234 | int sync_node_pages(struct f2fs_sb_info *sbi, nid_t ino, |
1235 | struct writeback_control *wbc) | |
1236 | { | |
e05df3b1 JK |
1237 | pgoff_t index, end; |
1238 | struct pagevec pvec; | |
1239 | int step = ino ? 2 : 0; | |
1240 | int nwritten = 0, wrote = 0; | |
1241 | ||
1242 | pagevec_init(&pvec, 0); | |
1243 | ||
1244 | next_step: | |
1245 | index = 0; | |
80dd9c0e | 1246 | end = ULONG_MAX; |
e05df3b1 JK |
1247 | |
1248 | while (index <= end) { | |
1249 | int i, nr_pages; | |
4ef51a8f | 1250 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, |
e05df3b1 JK |
1251 | PAGECACHE_TAG_DIRTY, |
1252 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1253 | if (nr_pages == 0) | |
1254 | break; | |
1255 | ||
1256 | for (i = 0; i < nr_pages; i++) { | |
1257 | struct page *page = pvec.pages[i]; | |
1258 | ||
6d5a1495 CY |
1259 | if (unlikely(f2fs_cp_error(sbi))) { |
1260 | pagevec_release(&pvec); | |
1261 | return -EIO; | |
1262 | } | |
1263 | ||
e05df3b1 JK |
1264 | /* |
1265 | * flushing sequence with step: | |
1266 | * 0. indirect nodes | |
1267 | * 1. dentry dnodes | |
1268 | * 2. file dnodes | |
1269 | */ | |
1270 | if (step == 0 && IS_DNODE(page)) | |
1271 | continue; | |
1272 | if (step == 1 && (!IS_DNODE(page) || | |
1273 | is_cold_node(page))) | |
1274 | continue; | |
1275 | if (step == 2 && (!IS_DNODE(page) || | |
1276 | !is_cold_node(page))) | |
1277 | continue; | |
1278 | ||
1279 | /* | |
1280 | * If an fsync mode, | |
1281 | * we should not skip writing node pages. | |
1282 | */ | |
9a4cbc9e | 1283 | lock_node: |
e05df3b1 JK |
1284 | if (ino && ino_of_node(page) == ino) |
1285 | lock_page(page); | |
1286 | else if (!trylock_page(page)) | |
1287 | continue; | |
1288 | ||
4ef51a8f | 1289 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
e05df3b1 JK |
1290 | continue_unlock: |
1291 | unlock_page(page); | |
1292 | continue; | |
1293 | } | |
1294 | if (ino && ino_of_node(page) != ino) | |
1295 | goto continue_unlock; | |
1296 | ||
1297 | if (!PageDirty(page)) { | |
1298 | /* someone wrote it for us */ | |
1299 | goto continue_unlock; | |
1300 | } | |
1301 | ||
2049d4fc JK |
1302 | /* flush inline_data */ |
1303 | if (!ino && is_inline_node(page)) { | |
1304 | clear_inline_node(page); | |
1305 | unlock_page(page); | |
1306 | flush_inline_data(sbi, ino_of_node(page)); | |
9a4cbc9e | 1307 | goto lock_node; |
2049d4fc JK |
1308 | } |
1309 | ||
fa3d2bdf JK |
1310 | f2fs_wait_on_page_writeback(page, NODE, true); |
1311 | ||
1312 | BUG_ON(PageWriteback(page)); | |
e05df3b1 JK |
1313 | if (!clear_page_dirty_for_io(page)) |
1314 | goto continue_unlock; | |
1315 | ||
1316 | /* called by fsync() */ | |
1317 | if (ino && IS_DNODE(page)) { | |
e05df3b1 | 1318 | set_fsync_mark(page, 1); |
2dcf51ab JK |
1319 | if (IS_INODE(page)) |
1320 | set_dentry_mark(page, | |
1321 | need_dentry_mark(sbi, ino)); | |
e05df3b1 JK |
1322 | nwritten++; |
1323 | } else { | |
1324 | set_fsync_mark(page, 0); | |
1325 | set_dentry_mark(page, 0); | |
1326 | } | |
52746519 JK |
1327 | |
1328 | if (NODE_MAPPING(sbi)->a_ops->writepage(page, wbc)) | |
1329 | unlock_page(page); | |
1330 | else | |
1331 | wrote++; | |
e05df3b1 JK |
1332 | |
1333 | if (--wbc->nr_to_write == 0) | |
1334 | break; | |
1335 | } | |
1336 | pagevec_release(&pvec); | |
1337 | cond_resched(); | |
1338 | ||
1339 | if (wbc->nr_to_write == 0) { | |
1340 | step = 2; | |
1341 | break; | |
1342 | } | |
1343 | } | |
1344 | ||
1345 | if (step < 2) { | |
1346 | step++; | |
1347 | goto next_step; | |
1348 | } | |
1349 | ||
0c3a5797 CY |
1350 | if (wrote) { |
1351 | if (ino) | |
1352 | f2fs_submit_merged_bio_cond(sbi, NULL, NULL, | |
1353 | ino, NODE, WRITE); | |
1354 | else | |
1355 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1356 | } | |
e05df3b1 JK |
1357 | return nwritten; |
1358 | } | |
1359 | ||
cfe58f9d JK |
1360 | int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino) |
1361 | { | |
80dd9c0e | 1362 | pgoff_t index = 0, end = ULONG_MAX; |
cfe58f9d | 1363 | struct pagevec pvec; |
cfe58f9d JK |
1364 | int ret2 = 0, ret = 0; |
1365 | ||
1366 | pagevec_init(&pvec, 0); | |
4ef51a8f JK |
1367 | |
1368 | while (index <= end) { | |
1369 | int i, nr_pages; | |
1370 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1371 | PAGECACHE_TAG_WRITEBACK, | |
1372 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1373 | if (nr_pages == 0) | |
1374 | break; | |
cfe58f9d JK |
1375 | |
1376 | for (i = 0; i < nr_pages; i++) { | |
1377 | struct page *page = pvec.pages[i]; | |
1378 | ||
1379 | /* until radix tree lookup accepts end_index */ | |
cfb271d4 | 1380 | if (unlikely(page->index > end)) |
cfe58f9d JK |
1381 | continue; |
1382 | ||
4bf08ff6 | 1383 | if (ino && ino_of_node(page) == ino) { |
fec1d657 | 1384 | f2fs_wait_on_page_writeback(page, NODE, true); |
4bf08ff6 CY |
1385 | if (TestClearPageError(page)) |
1386 | ret = -EIO; | |
1387 | } | |
cfe58f9d JK |
1388 | } |
1389 | pagevec_release(&pvec); | |
1390 | cond_resched(); | |
1391 | } | |
1392 | ||
4ef51a8f | 1393 | if (unlikely(test_and_clear_bit(AS_ENOSPC, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d | 1394 | ret2 = -ENOSPC; |
4ef51a8f | 1395 | if (unlikely(test_and_clear_bit(AS_EIO, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d JK |
1396 | ret2 = -EIO; |
1397 | if (!ret) | |
1398 | ret = ret2; | |
1399 | return ret; | |
1400 | } | |
1401 | ||
e05df3b1 JK |
1402 | static int f2fs_write_node_page(struct page *page, |
1403 | struct writeback_control *wbc) | |
1404 | { | |
4081363f | 1405 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1406 | nid_t nid; |
e05df3b1 | 1407 | struct node_info ni; |
fb5566da | 1408 | struct f2fs_io_info fio = { |
05ca3632 | 1409 | .sbi = sbi, |
fb5566da | 1410 | .type = NODE, |
6c311ec6 | 1411 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
05ca3632 | 1412 | .page = page, |
4375a336 | 1413 | .encrypted_page = NULL, |
fb5566da | 1414 | }; |
e05df3b1 | 1415 | |
ecda0de3 CY |
1416 | trace_f2fs_writepage(page, NODE); |
1417 | ||
caf0047e | 1418 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
87a9bd26 | 1419 | goto redirty_out; |
cf779cab JK |
1420 | if (unlikely(f2fs_cp_error(sbi))) |
1421 | goto redirty_out; | |
87a9bd26 | 1422 | |
e05df3b1 JK |
1423 | /* get old block addr of this node page */ |
1424 | nid = nid_of_node(page); | |
9850cf4a | 1425 | f2fs_bug_on(sbi, page->index != nid); |
e05df3b1 | 1426 | |
25b93346 JK |
1427 | if (wbc->for_reclaim) { |
1428 | if (!down_read_trylock(&sbi->node_write)) | |
1429 | goto redirty_out; | |
1430 | } else { | |
1431 | down_read(&sbi->node_write); | |
1432 | } | |
1433 | ||
e05df3b1 JK |
1434 | get_node_info(sbi, nid, &ni); |
1435 | ||
1436 | /* This page is already truncated */ | |
6bacf52f | 1437 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
2bca1e23 | 1438 | ClearPageUptodate(page); |
39936837 | 1439 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
25b93346 | 1440 | up_read(&sbi->node_write); |
39936837 JK |
1441 | unlock_page(page); |
1442 | return 0; | |
1443 | } | |
e05df3b1 JK |
1444 | |
1445 | set_page_writeback(page); | |
7a9d7548 | 1446 | fio.old_blkaddr = ni.blk_addr; |
05ca3632 | 1447 | write_node_page(nid, &fio); |
7a9d7548 | 1448 | set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page)); |
e05df3b1 | 1449 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
b3582c68 | 1450 | up_read(&sbi->node_write); |
0c3a5797 CY |
1451 | |
1452 | if (wbc->for_reclaim) | |
1453 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, NODE, WRITE); | |
1454 | ||
e05df3b1 | 1455 | unlock_page(page); |
27c6bd60 | 1456 | |
0c3a5797 | 1457 | if (unlikely(f2fs_cp_error(sbi))) |
27c6bd60 JK |
1458 | f2fs_submit_merged_bio(sbi, NODE, WRITE); |
1459 | ||
e05df3b1 | 1460 | return 0; |
87a9bd26 JK |
1461 | |
1462 | redirty_out: | |
76f60268 | 1463 | redirty_page_for_writepage(wbc, page); |
87a9bd26 | 1464 | return AOP_WRITEPAGE_ACTIVATE; |
e05df3b1 JK |
1465 | } |
1466 | ||
1467 | static int f2fs_write_node_pages(struct address_space *mapping, | |
1468 | struct writeback_control *wbc) | |
1469 | { | |
4081363f | 1470 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 1471 | long diff; |
e05df3b1 | 1472 | |
4660f9c0 JK |
1473 | /* balancing f2fs's metadata in background */ |
1474 | f2fs_balance_fs_bg(sbi); | |
e05df3b1 | 1475 | |
a7fdffbd | 1476 | /* collect a number of dirty node pages and write together */ |
87d6f890 | 1477 | if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE)) |
d3baf95d | 1478 | goto skip_write; |
a7fdffbd | 1479 | |
d31c7c3f YH |
1480 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
1481 | ||
50c8cdb3 | 1482 | diff = nr_pages_to_write(sbi, NODE, wbc); |
fb5566da | 1483 | wbc->sync_mode = WB_SYNC_NONE; |
e05df3b1 | 1484 | sync_node_pages(sbi, 0, wbc); |
50c8cdb3 | 1485 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
e05df3b1 | 1486 | return 0; |
d3baf95d JK |
1487 | |
1488 | skip_write: | |
1489 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES); | |
d31c7c3f | 1490 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
d3baf95d | 1491 | return 0; |
e05df3b1 JK |
1492 | } |
1493 | ||
1494 | static int f2fs_set_node_page_dirty(struct page *page) | |
1495 | { | |
26c6b887 JK |
1496 | trace_f2fs_set_page_dirty(page, NODE); |
1497 | ||
e05df3b1 JK |
1498 | SetPageUptodate(page); |
1499 | if (!PageDirty(page)) { | |
1500 | __set_page_dirty_nobuffers(page); | |
4081363f | 1501 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 | 1502 | SetPagePrivate(page); |
9e4ded3f | 1503 | f2fs_trace_pid(page); |
e05df3b1 JK |
1504 | return 1; |
1505 | } | |
1506 | return 0; | |
1507 | } | |
1508 | ||
0a8165d7 | 1509 | /* |
e05df3b1 JK |
1510 | * Structure of the f2fs node operations |
1511 | */ | |
1512 | const struct address_space_operations f2fs_node_aops = { | |
1513 | .writepage = f2fs_write_node_page, | |
1514 | .writepages = f2fs_write_node_pages, | |
1515 | .set_page_dirty = f2fs_set_node_page_dirty, | |
487261f3 CY |
1516 | .invalidatepage = f2fs_invalidate_page, |
1517 | .releasepage = f2fs_release_page, | |
e05df3b1 JK |
1518 | }; |
1519 | ||
8a7ed66a JK |
1520 | static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i, |
1521 | nid_t n) | |
e05df3b1 | 1522 | { |
8a7ed66a | 1523 | return radix_tree_lookup(&nm_i->free_nid_root, n); |
e05df3b1 JK |
1524 | } |
1525 | ||
8a7ed66a JK |
1526 | static void __del_from_free_nid_list(struct f2fs_nm_info *nm_i, |
1527 | struct free_nid *i) | |
e05df3b1 JK |
1528 | { |
1529 | list_del(&i->list); | |
8a7ed66a | 1530 | radix_tree_delete(&nm_i->free_nid_root, i->nid); |
e05df3b1 JK |
1531 | } |
1532 | ||
6fb03f3a | 1533 | static int add_free_nid(struct f2fs_sb_info *sbi, nid_t nid, bool build) |
e05df3b1 | 1534 | { |
6fb03f3a | 1535 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1536 | struct free_nid *i; |
59bbd474 JK |
1537 | struct nat_entry *ne; |
1538 | bool allocated = false; | |
e05df3b1 | 1539 | |
6fb03f3a | 1540 | if (!available_free_memory(sbi, FREE_NIDS)) |
23d38844 | 1541 | return -1; |
9198aceb JK |
1542 | |
1543 | /* 0 nid should not be used */ | |
cfb271d4 | 1544 | if (unlikely(nid == 0)) |
9198aceb | 1545 | return 0; |
59bbd474 | 1546 | |
7bd59381 GZ |
1547 | if (build) { |
1548 | /* do not add allocated nids */ | |
7bd59381 | 1549 | ne = __lookup_nat_cache(nm_i, nid); |
a5131193 | 1550 | if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) || |
7ef35e3b | 1551 | nat_get_blkaddr(ne) != NULL_ADDR)) |
7bd59381 | 1552 | allocated = true; |
7bd59381 GZ |
1553 | if (allocated) |
1554 | return 0; | |
e05df3b1 | 1555 | } |
7bd59381 GZ |
1556 | |
1557 | i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); | |
e05df3b1 JK |
1558 | i->nid = nid; |
1559 | i->state = NID_NEW; | |
1560 | ||
769ec6e5 JK |
1561 | if (radix_tree_preload(GFP_NOFS)) { |
1562 | kmem_cache_free(free_nid_slab, i); | |
1563 | return 0; | |
1564 | } | |
1565 | ||
e05df3b1 | 1566 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1567 | if (radix_tree_insert(&nm_i->free_nid_root, i->nid, i)) { |
e05df3b1 | 1568 | spin_unlock(&nm_i->free_nid_list_lock); |
769ec6e5 | 1569 | radix_tree_preload_end(); |
e05df3b1 JK |
1570 | kmem_cache_free(free_nid_slab, i); |
1571 | return 0; | |
1572 | } | |
1573 | list_add_tail(&i->list, &nm_i->free_nid_list); | |
1574 | nm_i->fcnt++; | |
1575 | spin_unlock(&nm_i->free_nid_list_lock); | |
769ec6e5 | 1576 | radix_tree_preload_end(); |
e05df3b1 JK |
1577 | return 1; |
1578 | } | |
1579 | ||
1580 | static void remove_free_nid(struct f2fs_nm_info *nm_i, nid_t nid) | |
1581 | { | |
1582 | struct free_nid *i; | |
cf0ee0f0 CY |
1583 | bool need_free = false; |
1584 | ||
e05df3b1 | 1585 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1586 | i = __lookup_free_nid_list(nm_i, nid); |
e05df3b1 | 1587 | if (i && i->state == NID_NEW) { |
8a7ed66a | 1588 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1589 | nm_i->fcnt--; |
cf0ee0f0 | 1590 | need_free = true; |
e05df3b1 JK |
1591 | } |
1592 | spin_unlock(&nm_i->free_nid_list_lock); | |
cf0ee0f0 CY |
1593 | |
1594 | if (need_free) | |
1595 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1596 | } |
1597 | ||
6fb03f3a | 1598 | static void scan_nat_page(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
1599 | struct page *nat_page, nid_t start_nid) |
1600 | { | |
6fb03f3a | 1601 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 JK |
1602 | struct f2fs_nat_block *nat_blk = page_address(nat_page); |
1603 | block_t blk_addr; | |
e05df3b1 JK |
1604 | int i; |
1605 | ||
e05df3b1 JK |
1606 | i = start_nid % NAT_ENTRY_PER_BLOCK; |
1607 | ||
1608 | for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { | |
23d38844 | 1609 | |
cfb271d4 | 1610 | if (unlikely(start_nid >= nm_i->max_nid)) |
04431c44 | 1611 | break; |
23d38844 HL |
1612 | |
1613 | blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); | |
9850cf4a | 1614 | f2fs_bug_on(sbi, blk_addr == NEW_ADDR); |
23d38844 | 1615 | if (blk_addr == NULL_ADDR) { |
6fb03f3a | 1616 | if (add_free_nid(sbi, start_nid, true) < 0) |
23d38844 HL |
1617 | break; |
1618 | } | |
e05df3b1 | 1619 | } |
e05df3b1 JK |
1620 | } |
1621 | ||
1622 | static void build_free_nids(struct f2fs_sb_info *sbi) | |
1623 | { | |
e05df3b1 JK |
1624 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1625 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 1626 | struct f2fs_journal *journal = curseg->journal; |
8760952d | 1627 | int i = 0; |
55008d84 | 1628 | nid_t nid = nm_i->next_scan_nid; |
e05df3b1 | 1629 | |
55008d84 JK |
1630 | /* Enough entries */ |
1631 | if (nm_i->fcnt > NAT_ENTRY_PER_BLOCK) | |
1632 | return; | |
e05df3b1 | 1633 | |
55008d84 | 1634 | /* readahead nat pages to be scanned */ |
26879fb1 CY |
1635 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES, |
1636 | META_NAT, true); | |
e05df3b1 | 1637 | |
a5131193 JK |
1638 | down_read(&nm_i->nat_tree_lock); |
1639 | ||
e05df3b1 JK |
1640 | while (1) { |
1641 | struct page *page = get_current_nat_page(sbi, nid); | |
1642 | ||
6fb03f3a | 1643 | scan_nat_page(sbi, page, nid); |
e05df3b1 JK |
1644 | f2fs_put_page(page, 1); |
1645 | ||
1646 | nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); | |
cfb271d4 | 1647 | if (unlikely(nid >= nm_i->max_nid)) |
e05df3b1 | 1648 | nid = 0; |
55008d84 | 1649 | |
a6d494b6 | 1650 | if (++i >= FREE_NID_PAGES) |
e05df3b1 JK |
1651 | break; |
1652 | } | |
1653 | ||
55008d84 JK |
1654 | /* go to the next free nat pages to find free nids abundantly */ |
1655 | nm_i->next_scan_nid = nid; | |
e05df3b1 JK |
1656 | |
1657 | /* find free nids from current sum_pages */ | |
b7ad7512 | 1658 | down_read(&curseg->journal_rwsem); |
dfc08a12 CY |
1659 | for (i = 0; i < nats_in_cursum(journal); i++) { |
1660 | block_t addr; | |
1661 | ||
1662 | addr = le32_to_cpu(nat_in_journal(journal, i).block_addr); | |
1663 | nid = le32_to_cpu(nid_in_journal(journal, i)); | |
e05df3b1 | 1664 | if (addr == NULL_ADDR) |
6fb03f3a | 1665 | add_free_nid(sbi, nid, true); |
e05df3b1 JK |
1666 | else |
1667 | remove_free_nid(nm_i, nid); | |
1668 | } | |
b7ad7512 | 1669 | up_read(&curseg->journal_rwsem); |
a5131193 | 1670 | up_read(&nm_i->nat_tree_lock); |
2db2388f CY |
1671 | |
1672 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid), | |
ea1a29a0 | 1673 | nm_i->ra_nid_pages, META_NAT, false); |
e05df3b1 JK |
1674 | } |
1675 | ||
1676 | /* | |
1677 | * If this function returns success, caller can obtain a new nid | |
1678 | * from second parameter of this function. | |
1679 | * The returned nid could be used ino as well as nid when inode is created. | |
1680 | */ | |
1681 | bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) | |
1682 | { | |
1683 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1684 | struct free_nid *i = NULL; | |
e05df3b1 | 1685 | retry: |
7ee0eeab | 1686 | if (unlikely(sbi->total_valid_node_count + 1 > nm_i->available_nids)) |
55008d84 | 1687 | return false; |
e05df3b1 | 1688 | |
e05df3b1 | 1689 | spin_lock(&nm_i->free_nid_list_lock); |
e05df3b1 | 1690 | |
55008d84 | 1691 | /* We should not use stale free nids created by build_free_nids */ |
f978f5a0 | 1692 | if (nm_i->fcnt && !on_build_free_nids(nm_i)) { |
9850cf4a | 1693 | f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list)); |
2d7b822a | 1694 | list_for_each_entry(i, &nm_i->free_nid_list, list) |
55008d84 JK |
1695 | if (i->state == NID_NEW) |
1696 | break; | |
e05df3b1 | 1697 | |
9850cf4a | 1698 | f2fs_bug_on(sbi, i->state != NID_NEW); |
55008d84 JK |
1699 | *nid = i->nid; |
1700 | i->state = NID_ALLOC; | |
1701 | nm_i->fcnt--; | |
1702 | spin_unlock(&nm_i->free_nid_list_lock); | |
1703 | return true; | |
1704 | } | |
e05df3b1 | 1705 | spin_unlock(&nm_i->free_nid_list_lock); |
55008d84 JK |
1706 | |
1707 | /* Let's scan nat pages and its caches to get free nids */ | |
1708 | mutex_lock(&nm_i->build_lock); | |
55008d84 | 1709 | build_free_nids(sbi); |
55008d84 JK |
1710 | mutex_unlock(&nm_i->build_lock); |
1711 | goto retry; | |
e05df3b1 JK |
1712 | } |
1713 | ||
0a8165d7 | 1714 | /* |
e05df3b1 JK |
1715 | * alloc_nid() should be called prior to this function. |
1716 | */ | |
1717 | void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) | |
1718 | { | |
1719 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1720 | struct free_nid *i; | |
1721 | ||
1722 | spin_lock(&nm_i->free_nid_list_lock); | |
8a7ed66a | 1723 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1724 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
8a7ed66a | 1725 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1726 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1727 | |
1728 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1729 | } |
1730 | ||
0a8165d7 | 1731 | /* |
e05df3b1 JK |
1732 | * alloc_nid() should be called prior to this function. |
1733 | */ | |
1734 | void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) | |
1735 | { | |
49952fa1 JK |
1736 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1737 | struct free_nid *i; | |
cf0ee0f0 | 1738 | bool need_free = false; |
49952fa1 | 1739 | |
65985d93 JK |
1740 | if (!nid) |
1741 | return; | |
1742 | ||
49952fa1 | 1743 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1744 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1745 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
6fb03f3a | 1746 | if (!available_free_memory(sbi, FREE_NIDS)) { |
8a7ed66a | 1747 | __del_from_free_nid_list(nm_i, i); |
cf0ee0f0 | 1748 | need_free = true; |
95630cba HL |
1749 | } else { |
1750 | i->state = NID_NEW; | |
1751 | nm_i->fcnt++; | |
1752 | } | |
49952fa1 | 1753 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1754 | |
1755 | if (need_free) | |
1756 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1757 | } |
1758 | ||
31696580 CY |
1759 | int try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink) |
1760 | { | |
1761 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1762 | struct free_nid *i, *next; | |
1763 | int nr = nr_shrink; | |
1764 | ||
1765 | if (!mutex_trylock(&nm_i->build_lock)) | |
1766 | return 0; | |
1767 | ||
1768 | spin_lock(&nm_i->free_nid_list_lock); | |
1769 | list_for_each_entry_safe(i, next, &nm_i->free_nid_list, list) { | |
1770 | if (nr_shrink <= 0 || nm_i->fcnt <= NAT_ENTRY_PER_BLOCK) | |
1771 | break; | |
1772 | if (i->state == NID_ALLOC) | |
1773 | continue; | |
1774 | __del_from_free_nid_list(nm_i, i); | |
31696580 | 1775 | kmem_cache_free(free_nid_slab, i); |
f7409d0f | 1776 | nm_i->fcnt--; |
31696580 | 1777 | nr_shrink--; |
31696580 CY |
1778 | } |
1779 | spin_unlock(&nm_i->free_nid_list_lock); | |
1780 | mutex_unlock(&nm_i->build_lock); | |
1781 | ||
1782 | return nr - nr_shrink; | |
1783 | } | |
1784 | ||
70cfed88 | 1785 | void recover_inline_xattr(struct inode *inode, struct page *page) |
28cdce04 | 1786 | { |
28cdce04 CY |
1787 | void *src_addr, *dst_addr; |
1788 | size_t inline_size; | |
1789 | struct page *ipage; | |
1790 | struct f2fs_inode *ri; | |
1791 | ||
4081363f | 1792 | ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino); |
9850cf4a | 1793 | f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage)); |
28cdce04 | 1794 | |
e3b4d43f JK |
1795 | ri = F2FS_INODE(page); |
1796 | if (!(ri->i_inline & F2FS_INLINE_XATTR)) { | |
1797 | clear_inode_flag(F2FS_I(inode), FI_INLINE_XATTR); | |
1798 | goto update_inode; | |
1799 | } | |
1800 | ||
28cdce04 CY |
1801 | dst_addr = inline_xattr_addr(ipage); |
1802 | src_addr = inline_xattr_addr(page); | |
1803 | inline_size = inline_xattr_size(inode); | |
1804 | ||
fec1d657 | 1805 | f2fs_wait_on_page_writeback(ipage, NODE, true); |
28cdce04 | 1806 | memcpy(dst_addr, src_addr, inline_size); |
e3b4d43f | 1807 | update_inode: |
28cdce04 CY |
1808 | update_inode(inode, ipage); |
1809 | f2fs_put_page(ipage, 1); | |
1810 | } | |
1811 | ||
1c35a90e | 1812 | void recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr) |
abb2366c | 1813 | { |
4081363f | 1814 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
abb2366c JK |
1815 | nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid; |
1816 | nid_t new_xnid = nid_of_node(page); | |
1817 | struct node_info ni; | |
1818 | ||
abb2366c JK |
1819 | /* 1: invalidate the previous xattr nid */ |
1820 | if (!prev_xnid) | |
1821 | goto recover_xnid; | |
1822 | ||
1823 | /* Deallocate node address */ | |
1824 | get_node_info(sbi, prev_xnid, &ni); | |
9850cf4a | 1825 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
abb2366c JK |
1826 | invalidate_blocks(sbi, ni.blk_addr); |
1827 | dec_valid_node_count(sbi, inode); | |
479f40c4 | 1828 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
abb2366c JK |
1829 | |
1830 | recover_xnid: | |
1831 | /* 2: allocate new xattr nid */ | |
1832 | if (unlikely(!inc_valid_node_count(sbi, inode))) | |
9850cf4a | 1833 | f2fs_bug_on(sbi, 1); |
abb2366c JK |
1834 | |
1835 | remove_free_nid(NM_I(sbi), new_xnid); | |
1836 | get_node_info(sbi, new_xnid, &ni); | |
1837 | ni.ino = inode->i_ino; | |
479f40c4 | 1838 | set_node_addr(sbi, &ni, NEW_ADDR, false); |
abb2366c JK |
1839 | F2FS_I(inode)->i_xattr_nid = new_xnid; |
1840 | ||
1841 | /* 3: update xattr blkaddr */ | |
1842 | refresh_sit_entry(sbi, NEW_ADDR, blkaddr); | |
479f40c4 | 1843 | set_node_addr(sbi, &ni, blkaddr, false); |
abb2366c JK |
1844 | |
1845 | update_inode_page(inode); | |
abb2366c JK |
1846 | } |
1847 | ||
e05df3b1 JK |
1848 | int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) |
1849 | { | |
58bfaf44 | 1850 | struct f2fs_inode *src, *dst; |
e05df3b1 JK |
1851 | nid_t ino = ino_of_node(page); |
1852 | struct node_info old_ni, new_ni; | |
1853 | struct page *ipage; | |
1854 | ||
e8271fa3 JK |
1855 | get_node_info(sbi, ino, &old_ni); |
1856 | ||
1857 | if (unlikely(old_ni.blk_addr != NULL_ADDR)) | |
1858 | return -EINVAL; | |
1859 | ||
4ef51a8f | 1860 | ipage = grab_cache_page(NODE_MAPPING(sbi), ino); |
e05df3b1 JK |
1861 | if (!ipage) |
1862 | return -ENOMEM; | |
1863 | ||
e1c42045 | 1864 | /* Should not use this inode from free nid list */ |
e05df3b1 JK |
1865 | remove_free_nid(NM_I(sbi), ino); |
1866 | ||
e05df3b1 JK |
1867 | SetPageUptodate(ipage); |
1868 | fill_node_footer(ipage, ino, ino, 0, true); | |
1869 | ||
58bfaf44 JK |
1870 | src = F2FS_INODE(page); |
1871 | dst = F2FS_INODE(ipage); | |
e05df3b1 | 1872 | |
58bfaf44 JK |
1873 | memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); |
1874 | dst->i_size = 0; | |
1875 | dst->i_blocks = cpu_to_le64(1); | |
1876 | dst->i_links = cpu_to_le32(1); | |
1877 | dst->i_xattr_nid = 0; | |
617deb8c | 1878 | dst->i_inline = src->i_inline & F2FS_INLINE_XATTR; |
e05df3b1 JK |
1879 | |
1880 | new_ni = old_ni; | |
1881 | new_ni.ino = ino; | |
1882 | ||
cfb271d4 | 1883 | if (unlikely(!inc_valid_node_count(sbi, NULL))) |
65e5cd0a | 1884 | WARN_ON(1); |
479f40c4 | 1885 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
e05df3b1 | 1886 | inc_valid_inode_count(sbi); |
617deb8c | 1887 | set_page_dirty(ipage); |
e05df3b1 JK |
1888 | f2fs_put_page(ipage, 1); |
1889 | return 0; | |
1890 | } | |
1891 | ||
1892 | int restore_node_summary(struct f2fs_sb_info *sbi, | |
1893 | unsigned int segno, struct f2fs_summary_block *sum) | |
1894 | { | |
1895 | struct f2fs_node *rn; | |
1896 | struct f2fs_summary *sum_entry; | |
e05df3b1 | 1897 | block_t addr; |
90a893c7 | 1898 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
9ecf4b80 | 1899 | int i, idx, last_offset, nrpages; |
e05df3b1 JK |
1900 | |
1901 | /* scan the node segment */ | |
1902 | last_offset = sbi->blocks_per_seg; | |
1903 | addr = START_BLOCK(sbi, segno); | |
1904 | sum_entry = &sum->entries[0]; | |
1905 | ||
9ecf4b80 | 1906 | for (i = 0; i < last_offset; i += nrpages, addr += nrpages) { |
9af0ff1c | 1907 | nrpages = min(last_offset - i, bio_blocks); |
393ff91f | 1908 | |
e1c42045 | 1909 | /* readahead node pages */ |
26879fb1 | 1910 | ra_meta_pages(sbi, addr, nrpages, META_POR, true); |
e05df3b1 | 1911 | |
9ecf4b80 | 1912 | for (idx = addr; idx < addr + nrpages; idx++) { |
2b947003 | 1913 | struct page *page = get_tmp_page(sbi, idx); |
9af0ff1c | 1914 | |
9ecf4b80 CY |
1915 | rn = F2FS_NODE(page); |
1916 | sum_entry->nid = rn->footer.nid; | |
1917 | sum_entry->version = 0; | |
1918 | sum_entry->ofs_in_node = 0; | |
1919 | sum_entry++; | |
1920 | f2fs_put_page(page, 1); | |
9af0ff1c | 1921 | } |
bac4eef6 | 1922 | |
9ecf4b80 | 1923 | invalidate_mapping_pages(META_MAPPING(sbi), addr, |
bac4eef6 | 1924 | addr + nrpages); |
e05df3b1 | 1925 | } |
9ecf4b80 | 1926 | return 0; |
e05df3b1 JK |
1927 | } |
1928 | ||
aec71382 | 1929 | static void remove_nats_in_journal(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
1930 | { |
1931 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1932 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 1933 | struct f2fs_journal *journal = curseg->journal; |
e05df3b1 JK |
1934 | int i; |
1935 | ||
b7ad7512 | 1936 | down_write(&curseg->journal_rwsem); |
dfc08a12 | 1937 | for (i = 0; i < nats_in_cursum(journal); i++) { |
e05df3b1 JK |
1938 | struct nat_entry *ne; |
1939 | struct f2fs_nat_entry raw_ne; | |
dfc08a12 | 1940 | nid_t nid = le32_to_cpu(nid_in_journal(journal, i)); |
e05df3b1 | 1941 | |
dfc08a12 | 1942 | raw_ne = nat_in_journal(journal, i); |
9be32d72 | 1943 | |
e05df3b1 | 1944 | ne = __lookup_nat_cache(nm_i, nid); |
e05df3b1 | 1945 | if (!ne) { |
9be32d72 JK |
1946 | ne = grab_nat_entry(nm_i, nid); |
1947 | node_info_from_raw_nat(&ne->ni, &raw_ne); | |
e05df3b1 | 1948 | } |
e05df3b1 | 1949 | __set_nat_cache_dirty(nm_i, ne); |
e05df3b1 | 1950 | } |
dfc08a12 | 1951 | update_nats_in_cursum(journal, -i); |
b7ad7512 | 1952 | up_write(&curseg->journal_rwsem); |
e05df3b1 JK |
1953 | } |
1954 | ||
309cc2b6 JK |
1955 | static void __adjust_nat_entry_set(struct nat_entry_set *nes, |
1956 | struct list_head *head, int max) | |
e05df3b1 | 1957 | { |
309cc2b6 | 1958 | struct nat_entry_set *cur; |
e05df3b1 | 1959 | |
309cc2b6 JK |
1960 | if (nes->entry_cnt >= max) |
1961 | goto add_out; | |
e05df3b1 | 1962 | |
309cc2b6 JK |
1963 | list_for_each_entry(cur, head, set_list) { |
1964 | if (cur->entry_cnt >= nes->entry_cnt) { | |
1965 | list_add(&nes->set_list, cur->set_list.prev); | |
1966 | return; | |
1967 | } | |
aec71382 | 1968 | } |
309cc2b6 JK |
1969 | add_out: |
1970 | list_add_tail(&nes->set_list, head); | |
1971 | } | |
e05df3b1 | 1972 | |
309cc2b6 JK |
1973 | static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, |
1974 | struct nat_entry_set *set) | |
1975 | { | |
1976 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 1977 | struct f2fs_journal *journal = curseg->journal; |
309cc2b6 JK |
1978 | nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK; |
1979 | bool to_journal = true; | |
1980 | struct f2fs_nat_block *nat_blk; | |
1981 | struct nat_entry *ne, *cur; | |
1982 | struct page *page = NULL; | |
e05df3b1 | 1983 | |
aec71382 CY |
1984 | /* |
1985 | * there are two steps to flush nat entries: | |
1986 | * #1, flush nat entries to journal in current hot data summary block. | |
1987 | * #2, flush nat entries to nat page. | |
1988 | */ | |
dfc08a12 | 1989 | if (!__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL)) |
309cc2b6 JK |
1990 | to_journal = false; |
1991 | ||
1992 | if (to_journal) { | |
b7ad7512 | 1993 | down_write(&curseg->journal_rwsem); |
309cc2b6 JK |
1994 | } else { |
1995 | page = get_next_nat_page(sbi, start_nid); | |
1996 | nat_blk = page_address(page); | |
1997 | f2fs_bug_on(sbi, !nat_blk); | |
1998 | } | |
aec71382 | 1999 | |
309cc2b6 JK |
2000 | /* flush dirty nats in nat entry set */ |
2001 | list_for_each_entry_safe(ne, cur, &set->entry_list, list) { | |
2002 | struct f2fs_nat_entry *raw_ne; | |
2003 | nid_t nid = nat_get_nid(ne); | |
2004 | int offset; | |
2005 | ||
2006 | if (nat_get_blkaddr(ne) == NEW_ADDR) | |
2007 | continue; | |
aec71382 CY |
2008 | |
2009 | if (to_journal) { | |
dfc08a12 | 2010 | offset = lookup_journal_in_cursum(journal, |
309cc2b6 JK |
2011 | NAT_JOURNAL, nid, 1); |
2012 | f2fs_bug_on(sbi, offset < 0); | |
dfc08a12 CY |
2013 | raw_ne = &nat_in_journal(journal, offset); |
2014 | nid_in_journal(journal, offset) = cpu_to_le32(nid); | |
aec71382 | 2015 | } else { |
309cc2b6 | 2016 | raw_ne = &nat_blk->entries[nid - start_nid]; |
e05df3b1 | 2017 | } |
309cc2b6 | 2018 | raw_nat_from_node_info(raw_ne, &ne->ni); |
309cc2b6 JK |
2019 | nat_reset_flag(ne); |
2020 | __clear_nat_cache_dirty(NM_I(sbi), ne); | |
309cc2b6 JK |
2021 | if (nat_get_blkaddr(ne) == NULL_ADDR) |
2022 | add_free_nid(sbi, nid, false); | |
2023 | } | |
e05df3b1 | 2024 | |
309cc2b6 | 2025 | if (to_journal) |
b7ad7512 | 2026 | up_write(&curseg->journal_rwsem); |
309cc2b6 JK |
2027 | else |
2028 | f2fs_put_page(page, 1); | |
aec71382 | 2029 | |
80ec2e91 CL |
2030 | f2fs_bug_on(sbi, set->entry_cnt); |
2031 | ||
2032 | radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); | |
2033 | kmem_cache_free(nat_entry_set_slab, set); | |
309cc2b6 | 2034 | } |
aec71382 | 2035 | |
309cc2b6 JK |
2036 | /* |
2037 | * This function is called during the checkpointing process. | |
2038 | */ | |
2039 | void flush_nat_entries(struct f2fs_sb_info *sbi) | |
2040 | { | |
2041 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2042 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2043 | struct f2fs_journal *journal = curseg->journal; |
7aed0d45 | 2044 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
309cc2b6 JK |
2045 | struct nat_entry_set *set, *tmp; |
2046 | unsigned int found; | |
2047 | nid_t set_idx = 0; | |
2048 | LIST_HEAD(sets); | |
2049 | ||
20d047c8 CL |
2050 | if (!nm_i->dirty_nat_cnt) |
2051 | return; | |
a5131193 JK |
2052 | |
2053 | down_write(&nm_i->nat_tree_lock); | |
2054 | ||
309cc2b6 JK |
2055 | /* |
2056 | * if there are no enough space in journal to store dirty nat | |
2057 | * entries, remove all entries from journal and merge them | |
2058 | * into nat entry set. | |
2059 | */ | |
dfc08a12 | 2060 | if (!__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL)) |
309cc2b6 JK |
2061 | remove_nats_in_journal(sbi); |
2062 | ||
309cc2b6 | 2063 | while ((found = __gang_lookup_nat_set(nm_i, |
7aed0d45 | 2064 | set_idx, SETVEC_SIZE, setvec))) { |
309cc2b6 JK |
2065 | unsigned idx; |
2066 | set_idx = setvec[found - 1]->set + 1; | |
2067 | for (idx = 0; idx < found; idx++) | |
2068 | __adjust_nat_entry_set(setvec[idx], &sets, | |
dfc08a12 | 2069 | MAX_NAT_JENTRIES(journal)); |
e05df3b1 | 2070 | } |
aec71382 | 2071 | |
309cc2b6 JK |
2072 | /* flush dirty nats in nat entry set */ |
2073 | list_for_each_entry_safe(set, tmp, &sets, set_list) | |
2074 | __flush_nat_entry_set(sbi, set); | |
2075 | ||
a5131193 JK |
2076 | up_write(&nm_i->nat_tree_lock); |
2077 | ||
9850cf4a | 2078 | f2fs_bug_on(sbi, nm_i->dirty_nat_cnt); |
e05df3b1 JK |
2079 | } |
2080 | ||
2081 | static int init_node_manager(struct f2fs_sb_info *sbi) | |
2082 | { | |
2083 | struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); | |
2084 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2085 | unsigned char *version_bitmap; | |
2086 | unsigned int nat_segs, nat_blocks; | |
2087 | ||
2088 | nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); | |
2089 | ||
2090 | /* segment_count_nat includes pair segment so divide to 2. */ | |
2091 | nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; | |
2092 | nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); | |
b63da15e | 2093 | |
7ee0eeab JK |
2094 | nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks; |
2095 | ||
b63da15e | 2096 | /* not used nids: 0, node, meta, (and root counted as valid node) */ |
c200b1aa | 2097 | nm_i->available_nids = nm_i->max_nid - F2FS_RESERVED_NODE_NUM; |
e05df3b1 JK |
2098 | nm_i->fcnt = 0; |
2099 | nm_i->nat_cnt = 0; | |
cdfc41c1 | 2100 | nm_i->ram_thresh = DEF_RAM_THRESHOLD; |
ea1a29a0 | 2101 | nm_i->ra_nid_pages = DEF_RA_NID_PAGES; |
2304cb0c | 2102 | nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD; |
e05df3b1 | 2103 | |
8a7ed66a | 2104 | INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC); |
e05df3b1 | 2105 | INIT_LIST_HEAD(&nm_i->free_nid_list); |
769ec6e5 JK |
2106 | INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO); |
2107 | INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO); | |
e05df3b1 | 2108 | INIT_LIST_HEAD(&nm_i->nat_entries); |
e05df3b1 JK |
2109 | |
2110 | mutex_init(&nm_i->build_lock); | |
2111 | spin_lock_init(&nm_i->free_nid_list_lock); | |
8b26ef98 | 2112 | init_rwsem(&nm_i->nat_tree_lock); |
e05df3b1 | 2113 | |
e05df3b1 | 2114 | nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); |
79b5793b | 2115 | nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); |
e05df3b1 JK |
2116 | version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); |
2117 | if (!version_bitmap) | |
2118 | return -EFAULT; | |
2119 | ||
79b5793b AG |
2120 | nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, |
2121 | GFP_KERNEL); | |
2122 | if (!nm_i->nat_bitmap) | |
2123 | return -ENOMEM; | |
e05df3b1 JK |
2124 | return 0; |
2125 | } | |
2126 | ||
2127 | int build_node_manager(struct f2fs_sb_info *sbi) | |
2128 | { | |
2129 | int err; | |
2130 | ||
2131 | sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL); | |
2132 | if (!sbi->nm_info) | |
2133 | return -ENOMEM; | |
2134 | ||
2135 | err = init_node_manager(sbi); | |
2136 | if (err) | |
2137 | return err; | |
2138 | ||
2139 | build_free_nids(sbi); | |
2140 | return 0; | |
2141 | } | |
2142 | ||
2143 | void destroy_node_manager(struct f2fs_sb_info *sbi) | |
2144 | { | |
2145 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2146 | struct free_nid *i, *next_i; | |
2147 | struct nat_entry *natvec[NATVEC_SIZE]; | |
7aed0d45 | 2148 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
e05df3b1 JK |
2149 | nid_t nid = 0; |
2150 | unsigned int found; | |
2151 | ||
2152 | if (!nm_i) | |
2153 | return; | |
2154 | ||
2155 | /* destroy free nid list */ | |
2156 | spin_lock(&nm_i->free_nid_list_lock); | |
2157 | list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) { | |
9850cf4a | 2158 | f2fs_bug_on(sbi, i->state == NID_ALLOC); |
8a7ed66a | 2159 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 2160 | nm_i->fcnt--; |
cf0ee0f0 CY |
2161 | spin_unlock(&nm_i->free_nid_list_lock); |
2162 | kmem_cache_free(free_nid_slab, i); | |
2163 | spin_lock(&nm_i->free_nid_list_lock); | |
e05df3b1 | 2164 | } |
9850cf4a | 2165 | f2fs_bug_on(sbi, nm_i->fcnt); |
e05df3b1 JK |
2166 | spin_unlock(&nm_i->free_nid_list_lock); |
2167 | ||
2168 | /* destroy nat cache */ | |
8b26ef98 | 2169 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2170 | while ((found = __gang_lookup_nat_cache(nm_i, |
2171 | nid, NATVEC_SIZE, natvec))) { | |
2172 | unsigned idx; | |
7aed0d45 | 2173 | |
b6ce391e GZ |
2174 | nid = nat_get_nid(natvec[found - 1]) + 1; |
2175 | for (idx = 0; idx < found; idx++) | |
2176 | __del_from_nat_cache(nm_i, natvec[idx]); | |
e05df3b1 | 2177 | } |
9850cf4a | 2178 | f2fs_bug_on(sbi, nm_i->nat_cnt); |
7aed0d45 JK |
2179 | |
2180 | /* destroy nat set cache */ | |
2181 | nid = 0; | |
2182 | while ((found = __gang_lookup_nat_set(nm_i, | |
2183 | nid, SETVEC_SIZE, setvec))) { | |
2184 | unsigned idx; | |
2185 | ||
2186 | nid = setvec[found - 1]->set + 1; | |
2187 | for (idx = 0; idx < found; idx++) { | |
2188 | /* entry_cnt is not zero, when cp_error was occurred */ | |
2189 | f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list)); | |
2190 | radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set); | |
2191 | kmem_cache_free(nat_entry_set_slab, setvec[idx]); | |
2192 | } | |
2193 | } | |
8b26ef98 | 2194 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2195 | |
2196 | kfree(nm_i->nat_bitmap); | |
2197 | sbi->nm_info = NULL; | |
2198 | kfree(nm_i); | |
2199 | } | |
2200 | ||
6e6093a8 | 2201 | int __init create_node_manager_caches(void) |
e05df3b1 JK |
2202 | { |
2203 | nat_entry_slab = f2fs_kmem_cache_create("nat_entry", | |
e8512d2e | 2204 | sizeof(struct nat_entry)); |
e05df3b1 | 2205 | if (!nat_entry_slab) |
aec71382 | 2206 | goto fail; |
e05df3b1 JK |
2207 | |
2208 | free_nid_slab = f2fs_kmem_cache_create("free_nid", | |
e8512d2e | 2209 | sizeof(struct free_nid)); |
aec71382 | 2210 | if (!free_nid_slab) |
ce3e6d25 | 2211 | goto destroy_nat_entry; |
aec71382 CY |
2212 | |
2213 | nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set", | |
2214 | sizeof(struct nat_entry_set)); | |
2215 | if (!nat_entry_set_slab) | |
ce3e6d25 | 2216 | goto destroy_free_nid; |
e05df3b1 | 2217 | return 0; |
aec71382 | 2218 | |
ce3e6d25 | 2219 | destroy_free_nid: |
aec71382 | 2220 | kmem_cache_destroy(free_nid_slab); |
ce3e6d25 | 2221 | destroy_nat_entry: |
aec71382 CY |
2222 | kmem_cache_destroy(nat_entry_slab); |
2223 | fail: | |
2224 | return -ENOMEM; | |
e05df3b1 JK |
2225 | } |
2226 | ||
2227 | void destroy_node_manager_caches(void) | |
2228 | { | |
aec71382 | 2229 | kmem_cache_destroy(nat_entry_set_slab); |
e05df3b1 JK |
2230 | kmem_cache_destroy(free_nid_slab); |
2231 | kmem_cache_destroy(nat_entry_slab); | |
2232 | } |