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 | 48 | mem_size = (nm_i->fcnt * sizeof(struct free_nid)) >> |
09cbfeaf | 49 | 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); |
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 | 64 | mem_size += (sbi->im[i].ino_num * |
09cbfeaf | 65 | sizeof(struct ino_entry)) >> PAGE_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 | 70 | atomic_read(&sbi->total_ext_node) * |
09cbfeaf | 71 | sizeof(struct extent_node)) >> PAGE_SHIFT; |
429511cd | 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); | |
09cbfeaf | 124 | memcpy(dst_addr, src_addr, PAGE_SIZE); |
e05df3b1 JK |
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 | ||
999270de FL |
1087 | rcu_read_lock(); |
1088 | apage = radix_tree_lookup(&NODE_MAPPING(sbi)->page_tree, nid); | |
1089 | rcu_read_unlock(); | |
1090 | if (apage) | |
393ff91f | 1091 | return; |
e05df3b1 | 1092 | |
4ef51a8f | 1093 | apage = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1094 | if (!apage) |
1095 | return; | |
1096 | ||
56ae674c | 1097 | err = read_node_page(apage, READA); |
86531d6b | 1098 | f2fs_put_page(apage, err ? 1 : 0); |
e05df3b1 JK |
1099 | } |
1100 | ||
0e022ea8 CY |
1101 | /* |
1102 | * readahead MAX_RA_NODE number of node pages. | |
1103 | */ | |
17a0ee55 | 1104 | static void ra_node_pages(struct page *parent, int start) |
e05df3b1 | 1105 | { |
0e022ea8 CY |
1106 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); |
1107 | struct blk_plug plug; | |
1108 | int i, end; | |
1109 | nid_t nid; | |
4aa69d56 | 1110 | |
0e022ea8 | 1111 | blk_start_plug(&plug); |
e05df3b1 | 1112 | |
0e022ea8 CY |
1113 | /* Then, try readahead for siblings of the desired node */ |
1114 | end = start + MAX_RA_NODE; | |
1115 | end = min(end, NIDS_PER_BLOCK); | |
1116 | for (i = start; i < end; i++) { | |
1117 | nid = get_nid(parent, i, false); | |
1118 | ra_node_page(sbi, nid); | |
86531d6b | 1119 | } |
e05df3b1 | 1120 | |
0e022ea8 | 1121 | blk_finish_plug(&plug); |
e05df3b1 JK |
1122 | } |
1123 | ||
17a0ee55 | 1124 | static struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid, |
0e022ea8 | 1125 | struct page *parent, int start) |
e05df3b1 | 1126 | { |
e05df3b1 | 1127 | struct page *page; |
0e022ea8 | 1128 | int err; |
e05df3b1 | 1129 | |
e05df3b1 JK |
1130 | if (!nid) |
1131 | return ERR_PTR(-ENOENT); | |
e8458725 | 1132 | f2fs_bug_on(sbi, check_nid_range(sbi, nid)); |
afcb7ca0 | 1133 | repeat: |
4ef51a8f | 1134 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1135 | if (!page) |
1136 | return ERR_PTR(-ENOMEM); | |
1137 | ||
66d36a29 | 1138 | err = read_node_page(page, READ_SYNC); |
86531d6b JK |
1139 | if (err < 0) { |
1140 | f2fs_put_page(page, 1); | |
e05df3b1 | 1141 | return ERR_PTR(err); |
86531d6b | 1142 | } else if (err == LOCKED_PAGE) { |
56ae674c | 1143 | goto page_hit; |
86531d6b | 1144 | } |
e05df3b1 | 1145 | |
0e022ea8 CY |
1146 | if (parent) |
1147 | ra_node_pages(parent, start + 1); | |
c718379b | 1148 | |
e05df3b1 | 1149 | lock_page(page); |
0e022ea8 | 1150 | |
e1c51b9f CY |
1151 | if (unlikely(!PageUptodate(page))) { |
1152 | f2fs_put_page(page, 1); | |
1153 | return ERR_PTR(-EIO); | |
1154 | } | |
4ef51a8f | 1155 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1156 | f2fs_put_page(page, 1); |
1157 | goto repeat; | |
1158 | } | |
e0f56cb4 | 1159 | page_hit: |
e1c51b9f | 1160 | f2fs_bug_on(sbi, nid != nid_of_node(page)); |
e05df3b1 JK |
1161 | return page; |
1162 | } | |
1163 | ||
0e022ea8 CY |
1164 | struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) |
1165 | { | |
1166 | return __get_node_page(sbi, nid, NULL, 0); | |
1167 | } | |
1168 | ||
1169 | struct page *get_node_page_ra(struct page *parent, int start) | |
1170 | { | |
1171 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); | |
1172 | nid_t nid = get_nid(parent, start, false); | |
1173 | ||
1174 | return __get_node_page(sbi, nid, parent, start); | |
1175 | } | |
1176 | ||
e05df3b1 JK |
1177 | void sync_inode_page(struct dnode_of_data *dn) |
1178 | { | |
12719ae1 JK |
1179 | int ret = 0; |
1180 | ||
e05df3b1 | 1181 | if (IS_INODE(dn->node_page) || dn->inode_page == dn->node_page) { |
12719ae1 | 1182 | ret = update_inode(dn->inode, dn->node_page); |
e05df3b1 JK |
1183 | } else if (dn->inode_page) { |
1184 | if (!dn->inode_page_locked) | |
1185 | lock_page(dn->inode_page); | |
12719ae1 | 1186 | ret = update_inode(dn->inode, dn->inode_page); |
e05df3b1 JK |
1187 | if (!dn->inode_page_locked) |
1188 | unlock_page(dn->inode_page); | |
1189 | } else { | |
12719ae1 | 1190 | ret = update_inode_page(dn->inode); |
e05df3b1 | 1191 | } |
12719ae1 | 1192 | dn->node_changed = ret ? true: false; |
e05df3b1 JK |
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; | |
1199 | ||
1200 | /* should flush inline_data before evict_inode */ | |
1201 | inode = ilookup(sbi->sb, ino); | |
1202 | if (!inode) | |
1203 | return; | |
1204 | ||
19c7377b | 1205 | page = pagecache_get_page(inode->i_mapping, 0, FGP_NOWAIT, 0); |
2049d4fc JK |
1206 | if (!page) |
1207 | goto iput_out; | |
1208 | ||
19c7377b CY |
1209 | if (!trylock_page(page)) |
1210 | goto release_out; | |
1211 | ||
2049d4fc JK |
1212 | if (!PageUptodate(page)) |
1213 | goto page_out; | |
1214 | ||
1215 | if (!PageDirty(page)) | |
1216 | goto page_out; | |
1217 | ||
1218 | if (!clear_page_dirty_for_io(page)) | |
1219 | goto page_out; | |
1220 | ||
1221 | if (!f2fs_write_inline_data(inode, page)) | |
1222 | inode_dec_dirty_pages(inode); | |
1223 | else | |
1224 | set_page_dirty(page); | |
1225 | page_out: | |
19c7377b CY |
1226 | unlock_page(page); |
1227 | release_out: | |
1228 | f2fs_put_page(page, 0); | |
2049d4fc JK |
1229 | iput_out: |
1230 | iput(inode); | |
1231 | } | |
1232 | ||
e05df3b1 JK |
1233 | int sync_node_pages(struct f2fs_sb_info *sbi, nid_t ino, |
1234 | struct writeback_control *wbc) | |
1235 | { | |
e05df3b1 JK |
1236 | pgoff_t index, end; |
1237 | struct pagevec pvec; | |
1238 | int step = ino ? 2 : 0; | |
12bb0a8f | 1239 | int nwritten = 0; |
e05df3b1 JK |
1240 | |
1241 | pagevec_init(&pvec, 0); | |
1242 | ||
1243 | next_step: | |
1244 | index = 0; | |
80dd9c0e | 1245 | end = ULONG_MAX; |
e05df3b1 JK |
1246 | |
1247 | while (index <= end) { | |
1248 | int i, nr_pages; | |
4ef51a8f | 1249 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, |
e05df3b1 JK |
1250 | PAGECACHE_TAG_DIRTY, |
1251 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1252 | if (nr_pages == 0) | |
1253 | break; | |
1254 | ||
1255 | for (i = 0; i < nr_pages; i++) { | |
1256 | struct page *page = pvec.pages[i]; | |
1257 | ||
6d5a1495 CY |
1258 | if (unlikely(f2fs_cp_error(sbi))) { |
1259 | pagevec_release(&pvec); | |
1260 | return -EIO; | |
1261 | } | |
1262 | ||
e05df3b1 JK |
1263 | /* |
1264 | * flushing sequence with step: | |
1265 | * 0. indirect nodes | |
1266 | * 1. dentry dnodes | |
1267 | * 2. file dnodes | |
1268 | */ | |
1269 | if (step == 0 && IS_DNODE(page)) | |
1270 | continue; | |
1271 | if (step == 1 && (!IS_DNODE(page) || | |
1272 | is_cold_node(page))) | |
1273 | continue; | |
1274 | if (step == 2 && (!IS_DNODE(page) || | |
1275 | !is_cold_node(page))) | |
1276 | continue; | |
1277 | ||
1278 | /* | |
1279 | * If an fsync mode, | |
1280 | * we should not skip writing node pages. | |
1281 | */ | |
9a4cbc9e | 1282 | lock_node: |
e05df3b1 JK |
1283 | if (ino && ino_of_node(page) == ino) |
1284 | lock_page(page); | |
1285 | else if (!trylock_page(page)) | |
1286 | continue; | |
1287 | ||
4ef51a8f | 1288 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
e05df3b1 JK |
1289 | continue_unlock: |
1290 | unlock_page(page); | |
1291 | continue; | |
1292 | } | |
1293 | if (ino && ino_of_node(page) != ino) | |
1294 | goto continue_unlock; | |
1295 | ||
1296 | if (!PageDirty(page)) { | |
1297 | /* someone wrote it for us */ | |
1298 | goto continue_unlock; | |
1299 | } | |
1300 | ||
2049d4fc JK |
1301 | /* flush inline_data */ |
1302 | if (!ino && is_inline_node(page)) { | |
1303 | clear_inline_node(page); | |
1304 | unlock_page(page); | |
1305 | flush_inline_data(sbi, ino_of_node(page)); | |
9a4cbc9e | 1306 | goto lock_node; |
2049d4fc JK |
1307 | } |
1308 | ||
fa3d2bdf JK |
1309 | f2fs_wait_on_page_writeback(page, NODE, true); |
1310 | ||
1311 | BUG_ON(PageWriteback(page)); | |
e05df3b1 JK |
1312 | if (!clear_page_dirty_for_io(page)) |
1313 | goto continue_unlock; | |
1314 | ||
1315 | /* called by fsync() */ | |
1316 | if (ino && IS_DNODE(page)) { | |
e05df3b1 | 1317 | set_fsync_mark(page, 1); |
2dcf51ab JK |
1318 | if (IS_INODE(page)) |
1319 | set_dentry_mark(page, | |
1320 | need_dentry_mark(sbi, ino)); | |
e05df3b1 JK |
1321 | nwritten++; |
1322 | } else { | |
1323 | set_fsync_mark(page, 0); | |
1324 | set_dentry_mark(page, 0); | |
1325 | } | |
52746519 JK |
1326 | |
1327 | if (NODE_MAPPING(sbi)->a_ops->writepage(page, wbc)) | |
1328 | unlock_page(page); | |
e05df3b1 JK |
1329 | |
1330 | if (--wbc->nr_to_write == 0) | |
1331 | break; | |
1332 | } | |
1333 | pagevec_release(&pvec); | |
1334 | cond_resched(); | |
1335 | ||
1336 | if (wbc->nr_to_write == 0) { | |
1337 | step = 2; | |
1338 | break; | |
1339 | } | |
1340 | } | |
1341 | ||
1342 | if (step < 2) { | |
1343 | step++; | |
1344 | goto next_step; | |
1345 | } | |
e05df3b1 JK |
1346 | return nwritten; |
1347 | } | |
1348 | ||
cfe58f9d JK |
1349 | int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino) |
1350 | { | |
80dd9c0e | 1351 | pgoff_t index = 0, end = ULONG_MAX; |
cfe58f9d | 1352 | struct pagevec pvec; |
cfe58f9d JK |
1353 | int ret2 = 0, ret = 0; |
1354 | ||
1355 | pagevec_init(&pvec, 0); | |
4ef51a8f JK |
1356 | |
1357 | while (index <= end) { | |
1358 | int i, nr_pages; | |
1359 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1360 | PAGECACHE_TAG_WRITEBACK, | |
1361 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1362 | if (nr_pages == 0) | |
1363 | break; | |
cfe58f9d JK |
1364 | |
1365 | for (i = 0; i < nr_pages; i++) { | |
1366 | struct page *page = pvec.pages[i]; | |
1367 | ||
1368 | /* until radix tree lookup accepts end_index */ | |
cfb271d4 | 1369 | if (unlikely(page->index > end)) |
cfe58f9d JK |
1370 | continue; |
1371 | ||
4bf08ff6 | 1372 | if (ino && ino_of_node(page) == ino) { |
fec1d657 | 1373 | f2fs_wait_on_page_writeback(page, NODE, true); |
4bf08ff6 CY |
1374 | if (TestClearPageError(page)) |
1375 | ret = -EIO; | |
1376 | } | |
cfe58f9d JK |
1377 | } |
1378 | pagevec_release(&pvec); | |
1379 | cond_resched(); | |
1380 | } | |
1381 | ||
4ef51a8f | 1382 | if (unlikely(test_and_clear_bit(AS_ENOSPC, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d | 1383 | ret2 = -ENOSPC; |
4ef51a8f | 1384 | if (unlikely(test_and_clear_bit(AS_EIO, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d JK |
1385 | ret2 = -EIO; |
1386 | if (!ret) | |
1387 | ret = ret2; | |
1388 | return ret; | |
1389 | } | |
1390 | ||
e05df3b1 JK |
1391 | static int f2fs_write_node_page(struct page *page, |
1392 | struct writeback_control *wbc) | |
1393 | { | |
4081363f | 1394 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1395 | nid_t nid; |
e05df3b1 | 1396 | struct node_info ni; |
fb5566da | 1397 | struct f2fs_io_info fio = { |
05ca3632 | 1398 | .sbi = sbi, |
fb5566da | 1399 | .type = NODE, |
6c311ec6 | 1400 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
05ca3632 | 1401 | .page = page, |
4375a336 | 1402 | .encrypted_page = NULL, |
fb5566da | 1403 | }; |
e05df3b1 | 1404 | |
ecda0de3 CY |
1405 | trace_f2fs_writepage(page, NODE); |
1406 | ||
caf0047e | 1407 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
87a9bd26 | 1408 | goto redirty_out; |
cf779cab JK |
1409 | if (unlikely(f2fs_cp_error(sbi))) |
1410 | goto redirty_out; | |
87a9bd26 | 1411 | |
e05df3b1 JK |
1412 | /* get old block addr of this node page */ |
1413 | nid = nid_of_node(page); | |
9850cf4a | 1414 | f2fs_bug_on(sbi, page->index != nid); |
e05df3b1 | 1415 | |
25b93346 JK |
1416 | if (wbc->for_reclaim) { |
1417 | if (!down_read_trylock(&sbi->node_write)) | |
1418 | goto redirty_out; | |
1419 | } else { | |
1420 | down_read(&sbi->node_write); | |
1421 | } | |
1422 | ||
e05df3b1 JK |
1423 | get_node_info(sbi, nid, &ni); |
1424 | ||
1425 | /* This page is already truncated */ | |
6bacf52f | 1426 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
2bca1e23 | 1427 | ClearPageUptodate(page); |
39936837 | 1428 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
25b93346 | 1429 | up_read(&sbi->node_write); |
39936837 JK |
1430 | unlock_page(page); |
1431 | return 0; | |
1432 | } | |
e05df3b1 JK |
1433 | |
1434 | set_page_writeback(page); | |
7a9d7548 | 1435 | fio.old_blkaddr = ni.blk_addr; |
05ca3632 | 1436 | write_node_page(nid, &fio); |
7a9d7548 | 1437 | set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page)); |
e05df3b1 | 1438 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
b3582c68 | 1439 | up_read(&sbi->node_write); |
0c3a5797 CY |
1440 | |
1441 | if (wbc->for_reclaim) | |
1442 | f2fs_submit_merged_bio_cond(sbi, NULL, page, 0, NODE, WRITE); | |
1443 | ||
e05df3b1 | 1444 | unlock_page(page); |
27c6bd60 | 1445 | |
0c3a5797 | 1446 | if (unlikely(f2fs_cp_error(sbi))) |
27c6bd60 JK |
1447 | f2fs_submit_merged_bio(sbi, NODE, WRITE); |
1448 | ||
e05df3b1 | 1449 | return 0; |
87a9bd26 JK |
1450 | |
1451 | redirty_out: | |
76f60268 | 1452 | redirty_page_for_writepage(wbc, page); |
87a9bd26 | 1453 | return AOP_WRITEPAGE_ACTIVATE; |
e05df3b1 JK |
1454 | } |
1455 | ||
1456 | static int f2fs_write_node_pages(struct address_space *mapping, | |
1457 | struct writeback_control *wbc) | |
1458 | { | |
4081363f | 1459 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 1460 | long diff; |
e05df3b1 | 1461 | |
4660f9c0 JK |
1462 | /* balancing f2fs's metadata in background */ |
1463 | f2fs_balance_fs_bg(sbi); | |
e05df3b1 | 1464 | |
a7fdffbd | 1465 | /* collect a number of dirty node pages and write together */ |
87d6f890 | 1466 | if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE)) |
d3baf95d | 1467 | goto skip_write; |
a7fdffbd | 1468 | |
d31c7c3f YH |
1469 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
1470 | ||
50c8cdb3 | 1471 | diff = nr_pages_to_write(sbi, NODE, wbc); |
fb5566da | 1472 | wbc->sync_mode = WB_SYNC_NONE; |
e05df3b1 | 1473 | sync_node_pages(sbi, 0, wbc); |
50c8cdb3 | 1474 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
e05df3b1 | 1475 | return 0; |
d3baf95d JK |
1476 | |
1477 | skip_write: | |
1478 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES); | |
d31c7c3f | 1479 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
d3baf95d | 1480 | return 0; |
e05df3b1 JK |
1481 | } |
1482 | ||
1483 | static int f2fs_set_node_page_dirty(struct page *page) | |
1484 | { | |
26c6b887 JK |
1485 | trace_f2fs_set_page_dirty(page, NODE); |
1486 | ||
e05df3b1 JK |
1487 | SetPageUptodate(page); |
1488 | if (!PageDirty(page)) { | |
1489 | __set_page_dirty_nobuffers(page); | |
4081363f | 1490 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 | 1491 | SetPagePrivate(page); |
9e4ded3f | 1492 | f2fs_trace_pid(page); |
e05df3b1 JK |
1493 | return 1; |
1494 | } | |
1495 | return 0; | |
1496 | } | |
1497 | ||
0a8165d7 | 1498 | /* |
e05df3b1 JK |
1499 | * Structure of the f2fs node operations |
1500 | */ | |
1501 | const struct address_space_operations f2fs_node_aops = { | |
1502 | .writepage = f2fs_write_node_page, | |
1503 | .writepages = f2fs_write_node_pages, | |
1504 | .set_page_dirty = f2fs_set_node_page_dirty, | |
487261f3 CY |
1505 | .invalidatepage = f2fs_invalidate_page, |
1506 | .releasepage = f2fs_release_page, | |
e05df3b1 JK |
1507 | }; |
1508 | ||
8a7ed66a JK |
1509 | static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i, |
1510 | nid_t n) | |
e05df3b1 | 1511 | { |
8a7ed66a | 1512 | return radix_tree_lookup(&nm_i->free_nid_root, n); |
e05df3b1 JK |
1513 | } |
1514 | ||
8a7ed66a JK |
1515 | static void __del_from_free_nid_list(struct f2fs_nm_info *nm_i, |
1516 | struct free_nid *i) | |
e05df3b1 JK |
1517 | { |
1518 | list_del(&i->list); | |
8a7ed66a | 1519 | radix_tree_delete(&nm_i->free_nid_root, i->nid); |
e05df3b1 JK |
1520 | } |
1521 | ||
6fb03f3a | 1522 | static int add_free_nid(struct f2fs_sb_info *sbi, nid_t nid, bool build) |
e05df3b1 | 1523 | { |
6fb03f3a | 1524 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1525 | struct free_nid *i; |
59bbd474 JK |
1526 | struct nat_entry *ne; |
1527 | bool allocated = false; | |
e05df3b1 | 1528 | |
6fb03f3a | 1529 | if (!available_free_memory(sbi, FREE_NIDS)) |
23d38844 | 1530 | return -1; |
9198aceb JK |
1531 | |
1532 | /* 0 nid should not be used */ | |
cfb271d4 | 1533 | if (unlikely(nid == 0)) |
9198aceb | 1534 | return 0; |
59bbd474 | 1535 | |
7bd59381 GZ |
1536 | if (build) { |
1537 | /* do not add allocated nids */ | |
7bd59381 | 1538 | ne = __lookup_nat_cache(nm_i, nid); |
a5131193 | 1539 | if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) || |
7ef35e3b | 1540 | nat_get_blkaddr(ne) != NULL_ADDR)) |
7bd59381 | 1541 | allocated = true; |
7bd59381 GZ |
1542 | if (allocated) |
1543 | return 0; | |
e05df3b1 | 1544 | } |
7bd59381 GZ |
1545 | |
1546 | i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); | |
e05df3b1 JK |
1547 | i->nid = nid; |
1548 | i->state = NID_NEW; | |
1549 | ||
769ec6e5 JK |
1550 | if (radix_tree_preload(GFP_NOFS)) { |
1551 | kmem_cache_free(free_nid_slab, i); | |
1552 | return 0; | |
1553 | } | |
1554 | ||
e05df3b1 | 1555 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1556 | if (radix_tree_insert(&nm_i->free_nid_root, i->nid, i)) { |
e05df3b1 | 1557 | spin_unlock(&nm_i->free_nid_list_lock); |
769ec6e5 | 1558 | radix_tree_preload_end(); |
e05df3b1 JK |
1559 | kmem_cache_free(free_nid_slab, i); |
1560 | return 0; | |
1561 | } | |
1562 | list_add_tail(&i->list, &nm_i->free_nid_list); | |
1563 | nm_i->fcnt++; | |
1564 | spin_unlock(&nm_i->free_nid_list_lock); | |
769ec6e5 | 1565 | radix_tree_preload_end(); |
e05df3b1 JK |
1566 | return 1; |
1567 | } | |
1568 | ||
1569 | static void remove_free_nid(struct f2fs_nm_info *nm_i, nid_t nid) | |
1570 | { | |
1571 | struct free_nid *i; | |
cf0ee0f0 CY |
1572 | bool need_free = false; |
1573 | ||
e05df3b1 | 1574 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1575 | i = __lookup_free_nid_list(nm_i, nid); |
e05df3b1 | 1576 | if (i && i->state == NID_NEW) { |
8a7ed66a | 1577 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1578 | nm_i->fcnt--; |
cf0ee0f0 | 1579 | need_free = true; |
e05df3b1 JK |
1580 | } |
1581 | spin_unlock(&nm_i->free_nid_list_lock); | |
cf0ee0f0 CY |
1582 | |
1583 | if (need_free) | |
1584 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1585 | } |
1586 | ||
6fb03f3a | 1587 | static void scan_nat_page(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
1588 | struct page *nat_page, nid_t start_nid) |
1589 | { | |
6fb03f3a | 1590 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 JK |
1591 | struct f2fs_nat_block *nat_blk = page_address(nat_page); |
1592 | block_t blk_addr; | |
e05df3b1 JK |
1593 | int i; |
1594 | ||
e05df3b1 JK |
1595 | i = start_nid % NAT_ENTRY_PER_BLOCK; |
1596 | ||
1597 | for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { | |
23d38844 | 1598 | |
cfb271d4 | 1599 | if (unlikely(start_nid >= nm_i->max_nid)) |
04431c44 | 1600 | break; |
23d38844 HL |
1601 | |
1602 | blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); | |
9850cf4a | 1603 | f2fs_bug_on(sbi, blk_addr == NEW_ADDR); |
23d38844 | 1604 | if (blk_addr == NULL_ADDR) { |
6fb03f3a | 1605 | if (add_free_nid(sbi, start_nid, true) < 0) |
23d38844 HL |
1606 | break; |
1607 | } | |
e05df3b1 | 1608 | } |
e05df3b1 JK |
1609 | } |
1610 | ||
1611 | static void build_free_nids(struct f2fs_sb_info *sbi) | |
1612 | { | |
e05df3b1 JK |
1613 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1614 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 1615 | struct f2fs_journal *journal = curseg->journal; |
8760952d | 1616 | int i = 0; |
55008d84 | 1617 | nid_t nid = nm_i->next_scan_nid; |
e05df3b1 | 1618 | |
55008d84 JK |
1619 | /* Enough entries */ |
1620 | if (nm_i->fcnt > NAT_ENTRY_PER_BLOCK) | |
1621 | return; | |
e05df3b1 | 1622 | |
55008d84 | 1623 | /* readahead nat pages to be scanned */ |
26879fb1 CY |
1624 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES, |
1625 | META_NAT, true); | |
e05df3b1 | 1626 | |
a5131193 JK |
1627 | down_read(&nm_i->nat_tree_lock); |
1628 | ||
e05df3b1 JK |
1629 | while (1) { |
1630 | struct page *page = get_current_nat_page(sbi, nid); | |
1631 | ||
6fb03f3a | 1632 | scan_nat_page(sbi, page, nid); |
e05df3b1 JK |
1633 | f2fs_put_page(page, 1); |
1634 | ||
1635 | nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); | |
cfb271d4 | 1636 | if (unlikely(nid >= nm_i->max_nid)) |
e05df3b1 | 1637 | nid = 0; |
55008d84 | 1638 | |
a6d494b6 | 1639 | if (++i >= FREE_NID_PAGES) |
e05df3b1 JK |
1640 | break; |
1641 | } | |
1642 | ||
55008d84 JK |
1643 | /* go to the next free nat pages to find free nids abundantly */ |
1644 | nm_i->next_scan_nid = nid; | |
e05df3b1 JK |
1645 | |
1646 | /* find free nids from current sum_pages */ | |
b7ad7512 | 1647 | down_read(&curseg->journal_rwsem); |
dfc08a12 CY |
1648 | for (i = 0; i < nats_in_cursum(journal); i++) { |
1649 | block_t addr; | |
1650 | ||
1651 | addr = le32_to_cpu(nat_in_journal(journal, i).block_addr); | |
1652 | nid = le32_to_cpu(nid_in_journal(journal, i)); | |
e05df3b1 | 1653 | if (addr == NULL_ADDR) |
6fb03f3a | 1654 | add_free_nid(sbi, nid, true); |
e05df3b1 JK |
1655 | else |
1656 | remove_free_nid(nm_i, nid); | |
1657 | } | |
b7ad7512 | 1658 | up_read(&curseg->journal_rwsem); |
a5131193 | 1659 | up_read(&nm_i->nat_tree_lock); |
2db2388f CY |
1660 | |
1661 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid), | |
ea1a29a0 | 1662 | nm_i->ra_nid_pages, META_NAT, false); |
e05df3b1 JK |
1663 | } |
1664 | ||
1665 | /* | |
1666 | * If this function returns success, caller can obtain a new nid | |
1667 | * from second parameter of this function. | |
1668 | * The returned nid could be used ino as well as nid when inode is created. | |
1669 | */ | |
1670 | bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) | |
1671 | { | |
1672 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1673 | struct free_nid *i = NULL; | |
e05df3b1 | 1674 | retry: |
7ee0eeab | 1675 | if (unlikely(sbi->total_valid_node_count + 1 > nm_i->available_nids)) |
55008d84 | 1676 | return false; |
e05df3b1 | 1677 | |
e05df3b1 | 1678 | spin_lock(&nm_i->free_nid_list_lock); |
e05df3b1 | 1679 | |
55008d84 | 1680 | /* We should not use stale free nids created by build_free_nids */ |
f978f5a0 | 1681 | if (nm_i->fcnt && !on_build_free_nids(nm_i)) { |
9850cf4a | 1682 | f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list)); |
2d7b822a | 1683 | list_for_each_entry(i, &nm_i->free_nid_list, list) |
55008d84 JK |
1684 | if (i->state == NID_NEW) |
1685 | break; | |
e05df3b1 | 1686 | |
9850cf4a | 1687 | f2fs_bug_on(sbi, i->state != NID_NEW); |
55008d84 JK |
1688 | *nid = i->nid; |
1689 | i->state = NID_ALLOC; | |
1690 | nm_i->fcnt--; | |
1691 | spin_unlock(&nm_i->free_nid_list_lock); | |
1692 | return true; | |
1693 | } | |
e05df3b1 | 1694 | spin_unlock(&nm_i->free_nid_list_lock); |
55008d84 JK |
1695 | |
1696 | /* Let's scan nat pages and its caches to get free nids */ | |
1697 | mutex_lock(&nm_i->build_lock); | |
55008d84 | 1698 | build_free_nids(sbi); |
55008d84 JK |
1699 | mutex_unlock(&nm_i->build_lock); |
1700 | goto retry; | |
e05df3b1 JK |
1701 | } |
1702 | ||
0a8165d7 | 1703 | /* |
e05df3b1 JK |
1704 | * alloc_nid() should be called prior to this function. |
1705 | */ | |
1706 | void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) | |
1707 | { | |
1708 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1709 | struct free_nid *i; | |
1710 | ||
1711 | spin_lock(&nm_i->free_nid_list_lock); | |
8a7ed66a | 1712 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1713 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
8a7ed66a | 1714 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1715 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1716 | |
1717 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1718 | } |
1719 | ||
0a8165d7 | 1720 | /* |
e05df3b1 JK |
1721 | * alloc_nid() should be called prior to this function. |
1722 | */ | |
1723 | void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) | |
1724 | { | |
49952fa1 JK |
1725 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1726 | struct free_nid *i; | |
cf0ee0f0 | 1727 | bool need_free = false; |
49952fa1 | 1728 | |
65985d93 JK |
1729 | if (!nid) |
1730 | return; | |
1731 | ||
49952fa1 | 1732 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1733 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1734 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
6fb03f3a | 1735 | if (!available_free_memory(sbi, FREE_NIDS)) { |
8a7ed66a | 1736 | __del_from_free_nid_list(nm_i, i); |
cf0ee0f0 | 1737 | need_free = true; |
95630cba HL |
1738 | } else { |
1739 | i->state = NID_NEW; | |
1740 | nm_i->fcnt++; | |
1741 | } | |
49952fa1 | 1742 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1743 | |
1744 | if (need_free) | |
1745 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1746 | } |
1747 | ||
31696580 CY |
1748 | int try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink) |
1749 | { | |
1750 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1751 | struct free_nid *i, *next; | |
1752 | int nr = nr_shrink; | |
1753 | ||
1754 | if (!mutex_trylock(&nm_i->build_lock)) | |
1755 | return 0; | |
1756 | ||
1757 | spin_lock(&nm_i->free_nid_list_lock); | |
1758 | list_for_each_entry_safe(i, next, &nm_i->free_nid_list, list) { | |
1759 | if (nr_shrink <= 0 || nm_i->fcnt <= NAT_ENTRY_PER_BLOCK) | |
1760 | break; | |
1761 | if (i->state == NID_ALLOC) | |
1762 | continue; | |
1763 | __del_from_free_nid_list(nm_i, i); | |
31696580 | 1764 | kmem_cache_free(free_nid_slab, i); |
f7409d0f | 1765 | nm_i->fcnt--; |
31696580 | 1766 | nr_shrink--; |
31696580 CY |
1767 | } |
1768 | spin_unlock(&nm_i->free_nid_list_lock); | |
1769 | mutex_unlock(&nm_i->build_lock); | |
1770 | ||
1771 | return nr - nr_shrink; | |
1772 | } | |
1773 | ||
70cfed88 | 1774 | void recover_inline_xattr(struct inode *inode, struct page *page) |
28cdce04 | 1775 | { |
28cdce04 CY |
1776 | void *src_addr, *dst_addr; |
1777 | size_t inline_size; | |
1778 | struct page *ipage; | |
1779 | struct f2fs_inode *ri; | |
1780 | ||
4081363f | 1781 | ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino); |
9850cf4a | 1782 | f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage)); |
28cdce04 | 1783 | |
e3b4d43f JK |
1784 | ri = F2FS_INODE(page); |
1785 | if (!(ri->i_inline & F2FS_INLINE_XATTR)) { | |
1786 | clear_inode_flag(F2FS_I(inode), FI_INLINE_XATTR); | |
1787 | goto update_inode; | |
1788 | } | |
1789 | ||
28cdce04 CY |
1790 | dst_addr = inline_xattr_addr(ipage); |
1791 | src_addr = inline_xattr_addr(page); | |
1792 | inline_size = inline_xattr_size(inode); | |
1793 | ||
fec1d657 | 1794 | f2fs_wait_on_page_writeback(ipage, NODE, true); |
28cdce04 | 1795 | memcpy(dst_addr, src_addr, inline_size); |
e3b4d43f | 1796 | update_inode: |
28cdce04 CY |
1797 | update_inode(inode, ipage); |
1798 | f2fs_put_page(ipage, 1); | |
1799 | } | |
1800 | ||
1c35a90e | 1801 | void recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr) |
abb2366c | 1802 | { |
4081363f | 1803 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
abb2366c JK |
1804 | nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid; |
1805 | nid_t new_xnid = nid_of_node(page); | |
1806 | struct node_info ni; | |
1807 | ||
abb2366c JK |
1808 | /* 1: invalidate the previous xattr nid */ |
1809 | if (!prev_xnid) | |
1810 | goto recover_xnid; | |
1811 | ||
1812 | /* Deallocate node address */ | |
1813 | get_node_info(sbi, prev_xnid, &ni); | |
9850cf4a | 1814 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
abb2366c JK |
1815 | invalidate_blocks(sbi, ni.blk_addr); |
1816 | dec_valid_node_count(sbi, inode); | |
479f40c4 | 1817 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
abb2366c JK |
1818 | |
1819 | recover_xnid: | |
1820 | /* 2: allocate new xattr nid */ | |
1821 | if (unlikely(!inc_valid_node_count(sbi, inode))) | |
9850cf4a | 1822 | f2fs_bug_on(sbi, 1); |
abb2366c JK |
1823 | |
1824 | remove_free_nid(NM_I(sbi), new_xnid); | |
1825 | get_node_info(sbi, new_xnid, &ni); | |
1826 | ni.ino = inode->i_ino; | |
479f40c4 | 1827 | set_node_addr(sbi, &ni, NEW_ADDR, false); |
abb2366c JK |
1828 | F2FS_I(inode)->i_xattr_nid = new_xnid; |
1829 | ||
1830 | /* 3: update xattr blkaddr */ | |
1831 | refresh_sit_entry(sbi, NEW_ADDR, blkaddr); | |
479f40c4 | 1832 | set_node_addr(sbi, &ni, blkaddr, false); |
abb2366c JK |
1833 | |
1834 | update_inode_page(inode); | |
abb2366c JK |
1835 | } |
1836 | ||
e05df3b1 JK |
1837 | int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) |
1838 | { | |
58bfaf44 | 1839 | struct f2fs_inode *src, *dst; |
e05df3b1 JK |
1840 | nid_t ino = ino_of_node(page); |
1841 | struct node_info old_ni, new_ni; | |
1842 | struct page *ipage; | |
1843 | ||
e8271fa3 JK |
1844 | get_node_info(sbi, ino, &old_ni); |
1845 | ||
1846 | if (unlikely(old_ni.blk_addr != NULL_ADDR)) | |
1847 | return -EINVAL; | |
1848 | ||
4ef51a8f | 1849 | ipage = grab_cache_page(NODE_MAPPING(sbi), ino); |
e05df3b1 JK |
1850 | if (!ipage) |
1851 | return -ENOMEM; | |
1852 | ||
e1c42045 | 1853 | /* Should not use this inode from free nid list */ |
e05df3b1 JK |
1854 | remove_free_nid(NM_I(sbi), ino); |
1855 | ||
e05df3b1 JK |
1856 | SetPageUptodate(ipage); |
1857 | fill_node_footer(ipage, ino, ino, 0, true); | |
1858 | ||
58bfaf44 JK |
1859 | src = F2FS_INODE(page); |
1860 | dst = F2FS_INODE(ipage); | |
e05df3b1 | 1861 | |
58bfaf44 JK |
1862 | memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); |
1863 | dst->i_size = 0; | |
1864 | dst->i_blocks = cpu_to_le64(1); | |
1865 | dst->i_links = cpu_to_le32(1); | |
1866 | dst->i_xattr_nid = 0; | |
617deb8c | 1867 | dst->i_inline = src->i_inline & F2FS_INLINE_XATTR; |
e05df3b1 JK |
1868 | |
1869 | new_ni = old_ni; | |
1870 | new_ni.ino = ino; | |
1871 | ||
cfb271d4 | 1872 | if (unlikely(!inc_valid_node_count(sbi, NULL))) |
65e5cd0a | 1873 | WARN_ON(1); |
479f40c4 | 1874 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
e05df3b1 | 1875 | inc_valid_inode_count(sbi); |
617deb8c | 1876 | set_page_dirty(ipage); |
e05df3b1 JK |
1877 | f2fs_put_page(ipage, 1); |
1878 | return 0; | |
1879 | } | |
1880 | ||
1881 | int restore_node_summary(struct f2fs_sb_info *sbi, | |
1882 | unsigned int segno, struct f2fs_summary_block *sum) | |
1883 | { | |
1884 | struct f2fs_node *rn; | |
1885 | struct f2fs_summary *sum_entry; | |
e05df3b1 | 1886 | block_t addr; |
90a893c7 | 1887 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
9ecf4b80 | 1888 | int i, idx, last_offset, nrpages; |
e05df3b1 JK |
1889 | |
1890 | /* scan the node segment */ | |
1891 | last_offset = sbi->blocks_per_seg; | |
1892 | addr = START_BLOCK(sbi, segno); | |
1893 | sum_entry = &sum->entries[0]; | |
1894 | ||
9ecf4b80 | 1895 | for (i = 0; i < last_offset; i += nrpages, addr += nrpages) { |
9af0ff1c | 1896 | nrpages = min(last_offset - i, bio_blocks); |
393ff91f | 1897 | |
e1c42045 | 1898 | /* readahead node pages */ |
26879fb1 | 1899 | ra_meta_pages(sbi, addr, nrpages, META_POR, true); |
e05df3b1 | 1900 | |
9ecf4b80 | 1901 | for (idx = addr; idx < addr + nrpages; idx++) { |
2b947003 | 1902 | struct page *page = get_tmp_page(sbi, idx); |
9af0ff1c | 1903 | |
9ecf4b80 CY |
1904 | rn = F2FS_NODE(page); |
1905 | sum_entry->nid = rn->footer.nid; | |
1906 | sum_entry->version = 0; | |
1907 | sum_entry->ofs_in_node = 0; | |
1908 | sum_entry++; | |
1909 | f2fs_put_page(page, 1); | |
9af0ff1c | 1910 | } |
bac4eef6 | 1911 | |
9ecf4b80 | 1912 | invalidate_mapping_pages(META_MAPPING(sbi), addr, |
bac4eef6 | 1913 | addr + nrpages); |
e05df3b1 | 1914 | } |
9ecf4b80 | 1915 | return 0; |
e05df3b1 JK |
1916 | } |
1917 | ||
aec71382 | 1918 | static void remove_nats_in_journal(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
1919 | { |
1920 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1921 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 1922 | struct f2fs_journal *journal = curseg->journal; |
e05df3b1 JK |
1923 | int i; |
1924 | ||
b7ad7512 | 1925 | down_write(&curseg->journal_rwsem); |
dfc08a12 | 1926 | for (i = 0; i < nats_in_cursum(journal); i++) { |
e05df3b1 JK |
1927 | struct nat_entry *ne; |
1928 | struct f2fs_nat_entry raw_ne; | |
dfc08a12 | 1929 | nid_t nid = le32_to_cpu(nid_in_journal(journal, i)); |
e05df3b1 | 1930 | |
dfc08a12 | 1931 | raw_ne = nat_in_journal(journal, i); |
9be32d72 | 1932 | |
e05df3b1 | 1933 | ne = __lookup_nat_cache(nm_i, nid); |
e05df3b1 | 1934 | if (!ne) { |
9be32d72 JK |
1935 | ne = grab_nat_entry(nm_i, nid); |
1936 | node_info_from_raw_nat(&ne->ni, &raw_ne); | |
e05df3b1 | 1937 | } |
e05df3b1 | 1938 | __set_nat_cache_dirty(nm_i, ne); |
e05df3b1 | 1939 | } |
dfc08a12 | 1940 | update_nats_in_cursum(journal, -i); |
b7ad7512 | 1941 | up_write(&curseg->journal_rwsem); |
e05df3b1 JK |
1942 | } |
1943 | ||
309cc2b6 JK |
1944 | static void __adjust_nat_entry_set(struct nat_entry_set *nes, |
1945 | struct list_head *head, int max) | |
e05df3b1 | 1946 | { |
309cc2b6 | 1947 | struct nat_entry_set *cur; |
e05df3b1 | 1948 | |
309cc2b6 JK |
1949 | if (nes->entry_cnt >= max) |
1950 | goto add_out; | |
e05df3b1 | 1951 | |
309cc2b6 JK |
1952 | list_for_each_entry(cur, head, set_list) { |
1953 | if (cur->entry_cnt >= nes->entry_cnt) { | |
1954 | list_add(&nes->set_list, cur->set_list.prev); | |
1955 | return; | |
1956 | } | |
aec71382 | 1957 | } |
309cc2b6 JK |
1958 | add_out: |
1959 | list_add_tail(&nes->set_list, head); | |
1960 | } | |
e05df3b1 | 1961 | |
309cc2b6 JK |
1962 | static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, |
1963 | struct nat_entry_set *set) | |
1964 | { | |
1965 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 1966 | struct f2fs_journal *journal = curseg->journal; |
309cc2b6 JK |
1967 | nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK; |
1968 | bool to_journal = true; | |
1969 | struct f2fs_nat_block *nat_blk; | |
1970 | struct nat_entry *ne, *cur; | |
1971 | struct page *page = NULL; | |
e05df3b1 | 1972 | |
aec71382 CY |
1973 | /* |
1974 | * there are two steps to flush nat entries: | |
1975 | * #1, flush nat entries to journal in current hot data summary block. | |
1976 | * #2, flush nat entries to nat page. | |
1977 | */ | |
dfc08a12 | 1978 | if (!__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL)) |
309cc2b6 JK |
1979 | to_journal = false; |
1980 | ||
1981 | if (to_journal) { | |
b7ad7512 | 1982 | down_write(&curseg->journal_rwsem); |
309cc2b6 JK |
1983 | } else { |
1984 | page = get_next_nat_page(sbi, start_nid); | |
1985 | nat_blk = page_address(page); | |
1986 | f2fs_bug_on(sbi, !nat_blk); | |
1987 | } | |
aec71382 | 1988 | |
309cc2b6 JK |
1989 | /* flush dirty nats in nat entry set */ |
1990 | list_for_each_entry_safe(ne, cur, &set->entry_list, list) { | |
1991 | struct f2fs_nat_entry *raw_ne; | |
1992 | nid_t nid = nat_get_nid(ne); | |
1993 | int offset; | |
1994 | ||
1995 | if (nat_get_blkaddr(ne) == NEW_ADDR) | |
1996 | continue; | |
aec71382 CY |
1997 | |
1998 | if (to_journal) { | |
dfc08a12 | 1999 | offset = lookup_journal_in_cursum(journal, |
309cc2b6 JK |
2000 | NAT_JOURNAL, nid, 1); |
2001 | f2fs_bug_on(sbi, offset < 0); | |
dfc08a12 CY |
2002 | raw_ne = &nat_in_journal(journal, offset); |
2003 | nid_in_journal(journal, offset) = cpu_to_le32(nid); | |
aec71382 | 2004 | } else { |
309cc2b6 | 2005 | raw_ne = &nat_blk->entries[nid - start_nid]; |
e05df3b1 | 2006 | } |
309cc2b6 | 2007 | raw_nat_from_node_info(raw_ne, &ne->ni); |
309cc2b6 JK |
2008 | nat_reset_flag(ne); |
2009 | __clear_nat_cache_dirty(NM_I(sbi), ne); | |
309cc2b6 JK |
2010 | if (nat_get_blkaddr(ne) == NULL_ADDR) |
2011 | add_free_nid(sbi, nid, false); | |
2012 | } | |
e05df3b1 | 2013 | |
309cc2b6 | 2014 | if (to_journal) |
b7ad7512 | 2015 | up_write(&curseg->journal_rwsem); |
309cc2b6 JK |
2016 | else |
2017 | f2fs_put_page(page, 1); | |
aec71382 | 2018 | |
80ec2e91 CL |
2019 | f2fs_bug_on(sbi, set->entry_cnt); |
2020 | ||
2021 | radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); | |
2022 | kmem_cache_free(nat_entry_set_slab, set); | |
309cc2b6 | 2023 | } |
aec71382 | 2024 | |
309cc2b6 JK |
2025 | /* |
2026 | * This function is called during the checkpointing process. | |
2027 | */ | |
2028 | void flush_nat_entries(struct f2fs_sb_info *sbi) | |
2029 | { | |
2030 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2031 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
b7ad7512 | 2032 | struct f2fs_journal *journal = curseg->journal; |
7aed0d45 | 2033 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
309cc2b6 JK |
2034 | struct nat_entry_set *set, *tmp; |
2035 | unsigned int found; | |
2036 | nid_t set_idx = 0; | |
2037 | LIST_HEAD(sets); | |
2038 | ||
20d047c8 CL |
2039 | if (!nm_i->dirty_nat_cnt) |
2040 | return; | |
a5131193 JK |
2041 | |
2042 | down_write(&nm_i->nat_tree_lock); | |
2043 | ||
309cc2b6 JK |
2044 | /* |
2045 | * if there are no enough space in journal to store dirty nat | |
2046 | * entries, remove all entries from journal and merge them | |
2047 | * into nat entry set. | |
2048 | */ | |
dfc08a12 | 2049 | if (!__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL)) |
309cc2b6 JK |
2050 | remove_nats_in_journal(sbi); |
2051 | ||
309cc2b6 | 2052 | while ((found = __gang_lookup_nat_set(nm_i, |
7aed0d45 | 2053 | set_idx, SETVEC_SIZE, setvec))) { |
309cc2b6 JK |
2054 | unsigned idx; |
2055 | set_idx = setvec[found - 1]->set + 1; | |
2056 | for (idx = 0; idx < found; idx++) | |
2057 | __adjust_nat_entry_set(setvec[idx], &sets, | |
dfc08a12 | 2058 | MAX_NAT_JENTRIES(journal)); |
e05df3b1 | 2059 | } |
aec71382 | 2060 | |
309cc2b6 JK |
2061 | /* flush dirty nats in nat entry set */ |
2062 | list_for_each_entry_safe(set, tmp, &sets, set_list) | |
2063 | __flush_nat_entry_set(sbi, set); | |
2064 | ||
a5131193 JK |
2065 | up_write(&nm_i->nat_tree_lock); |
2066 | ||
9850cf4a | 2067 | f2fs_bug_on(sbi, nm_i->dirty_nat_cnt); |
e05df3b1 JK |
2068 | } |
2069 | ||
2070 | static int init_node_manager(struct f2fs_sb_info *sbi) | |
2071 | { | |
2072 | struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); | |
2073 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2074 | unsigned char *version_bitmap; | |
2075 | unsigned int nat_segs, nat_blocks; | |
2076 | ||
2077 | nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); | |
2078 | ||
2079 | /* segment_count_nat includes pair segment so divide to 2. */ | |
2080 | nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; | |
2081 | nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); | |
b63da15e | 2082 | |
7ee0eeab JK |
2083 | nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks; |
2084 | ||
b63da15e | 2085 | /* not used nids: 0, node, meta, (and root counted as valid node) */ |
c200b1aa | 2086 | nm_i->available_nids = nm_i->max_nid - F2FS_RESERVED_NODE_NUM; |
e05df3b1 JK |
2087 | nm_i->fcnt = 0; |
2088 | nm_i->nat_cnt = 0; | |
cdfc41c1 | 2089 | nm_i->ram_thresh = DEF_RAM_THRESHOLD; |
ea1a29a0 | 2090 | nm_i->ra_nid_pages = DEF_RA_NID_PAGES; |
2304cb0c | 2091 | nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD; |
e05df3b1 | 2092 | |
8a7ed66a | 2093 | INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC); |
e05df3b1 | 2094 | INIT_LIST_HEAD(&nm_i->free_nid_list); |
769ec6e5 JK |
2095 | INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO); |
2096 | INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO); | |
e05df3b1 | 2097 | INIT_LIST_HEAD(&nm_i->nat_entries); |
e05df3b1 JK |
2098 | |
2099 | mutex_init(&nm_i->build_lock); | |
2100 | spin_lock_init(&nm_i->free_nid_list_lock); | |
8b26ef98 | 2101 | init_rwsem(&nm_i->nat_tree_lock); |
e05df3b1 | 2102 | |
e05df3b1 | 2103 | nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); |
79b5793b | 2104 | nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); |
e05df3b1 JK |
2105 | version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); |
2106 | if (!version_bitmap) | |
2107 | return -EFAULT; | |
2108 | ||
79b5793b AG |
2109 | nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, |
2110 | GFP_KERNEL); | |
2111 | if (!nm_i->nat_bitmap) | |
2112 | return -ENOMEM; | |
e05df3b1 JK |
2113 | return 0; |
2114 | } | |
2115 | ||
2116 | int build_node_manager(struct f2fs_sb_info *sbi) | |
2117 | { | |
2118 | int err; | |
2119 | ||
2120 | sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL); | |
2121 | if (!sbi->nm_info) | |
2122 | return -ENOMEM; | |
2123 | ||
2124 | err = init_node_manager(sbi); | |
2125 | if (err) | |
2126 | return err; | |
2127 | ||
2128 | build_free_nids(sbi); | |
2129 | return 0; | |
2130 | } | |
2131 | ||
2132 | void destroy_node_manager(struct f2fs_sb_info *sbi) | |
2133 | { | |
2134 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2135 | struct free_nid *i, *next_i; | |
2136 | struct nat_entry *natvec[NATVEC_SIZE]; | |
7aed0d45 | 2137 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
e05df3b1 JK |
2138 | nid_t nid = 0; |
2139 | unsigned int found; | |
2140 | ||
2141 | if (!nm_i) | |
2142 | return; | |
2143 | ||
2144 | /* destroy free nid list */ | |
2145 | spin_lock(&nm_i->free_nid_list_lock); | |
2146 | list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) { | |
9850cf4a | 2147 | f2fs_bug_on(sbi, i->state == NID_ALLOC); |
8a7ed66a | 2148 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 2149 | nm_i->fcnt--; |
cf0ee0f0 CY |
2150 | spin_unlock(&nm_i->free_nid_list_lock); |
2151 | kmem_cache_free(free_nid_slab, i); | |
2152 | spin_lock(&nm_i->free_nid_list_lock); | |
e05df3b1 | 2153 | } |
9850cf4a | 2154 | f2fs_bug_on(sbi, nm_i->fcnt); |
e05df3b1 JK |
2155 | spin_unlock(&nm_i->free_nid_list_lock); |
2156 | ||
2157 | /* destroy nat cache */ | |
8b26ef98 | 2158 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2159 | while ((found = __gang_lookup_nat_cache(nm_i, |
2160 | nid, NATVEC_SIZE, natvec))) { | |
2161 | unsigned idx; | |
7aed0d45 | 2162 | |
b6ce391e GZ |
2163 | nid = nat_get_nid(natvec[found - 1]) + 1; |
2164 | for (idx = 0; idx < found; idx++) | |
2165 | __del_from_nat_cache(nm_i, natvec[idx]); | |
e05df3b1 | 2166 | } |
9850cf4a | 2167 | f2fs_bug_on(sbi, nm_i->nat_cnt); |
7aed0d45 JK |
2168 | |
2169 | /* destroy nat set cache */ | |
2170 | nid = 0; | |
2171 | while ((found = __gang_lookup_nat_set(nm_i, | |
2172 | nid, SETVEC_SIZE, setvec))) { | |
2173 | unsigned idx; | |
2174 | ||
2175 | nid = setvec[found - 1]->set + 1; | |
2176 | for (idx = 0; idx < found; idx++) { | |
2177 | /* entry_cnt is not zero, when cp_error was occurred */ | |
2178 | f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list)); | |
2179 | radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set); | |
2180 | kmem_cache_free(nat_entry_set_slab, setvec[idx]); | |
2181 | } | |
2182 | } | |
8b26ef98 | 2183 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2184 | |
2185 | kfree(nm_i->nat_bitmap); | |
2186 | sbi->nm_info = NULL; | |
2187 | kfree(nm_i); | |
2188 | } | |
2189 | ||
6e6093a8 | 2190 | int __init create_node_manager_caches(void) |
e05df3b1 JK |
2191 | { |
2192 | nat_entry_slab = f2fs_kmem_cache_create("nat_entry", | |
e8512d2e | 2193 | sizeof(struct nat_entry)); |
e05df3b1 | 2194 | if (!nat_entry_slab) |
aec71382 | 2195 | goto fail; |
e05df3b1 JK |
2196 | |
2197 | free_nid_slab = f2fs_kmem_cache_create("free_nid", | |
e8512d2e | 2198 | sizeof(struct free_nid)); |
aec71382 | 2199 | if (!free_nid_slab) |
ce3e6d25 | 2200 | goto destroy_nat_entry; |
aec71382 CY |
2201 | |
2202 | nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set", | |
2203 | sizeof(struct nat_entry_set)); | |
2204 | if (!nat_entry_set_slab) | |
ce3e6d25 | 2205 | goto destroy_free_nid; |
e05df3b1 | 2206 | return 0; |
aec71382 | 2207 | |
ce3e6d25 | 2208 | destroy_free_nid: |
aec71382 | 2209 | kmem_cache_destroy(free_nid_slab); |
ce3e6d25 | 2210 | destroy_nat_entry: |
aec71382 CY |
2211 | kmem_cache_destroy(nat_entry_slab); |
2212 | fail: | |
2213 | return -ENOMEM; | |
e05df3b1 JK |
2214 | } |
2215 | ||
2216 | void destroy_node_manager_caches(void) | |
2217 | { | |
aec71382 | 2218 | kmem_cache_destroy(nat_entry_set_slab); |
e05df3b1 JK |
2219 | kmem_cache_destroy(free_nid_slab); |
2220 | kmem_cache_destroy(nat_entry_slab); | |
2221 | } |