]>
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" | |
51dd6249 | 22 | #include <trace/events/f2fs.h> |
e05df3b1 | 23 | |
f978f5a0 GZ |
24 | #define on_build_free_nids(nmi) mutex_is_locked(&nm_i->build_lock) |
25 | ||
e05df3b1 JK |
26 | static struct kmem_cache *nat_entry_slab; |
27 | static struct kmem_cache *free_nid_slab; | |
aec71382 | 28 | static struct kmem_cache *nat_entry_set_slab; |
e05df3b1 | 29 | |
6fb03f3a | 30 | bool available_free_memory(struct f2fs_sb_info *sbi, int type) |
cdfc41c1 | 31 | { |
6fb03f3a | 32 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
cdfc41c1 | 33 | struct sysinfo val; |
e5e7ea3c | 34 | unsigned long avail_ram; |
cdfc41c1 | 35 | unsigned long mem_size = 0; |
6fb03f3a | 36 | bool res = false; |
cdfc41c1 JK |
37 | |
38 | si_meminfo(&val); | |
e5e7ea3c JK |
39 | |
40 | /* only uses low memory */ | |
41 | avail_ram = val.totalram - val.totalhigh; | |
42 | ||
43 | /* give 25%, 25%, 50%, 50% memory for each components respectively */ | |
6fb03f3a | 44 | if (type == FREE_NIDS) { |
e5e7ea3c JK |
45 | mem_size = (nm_i->fcnt * sizeof(struct free_nid)) >> |
46 | PAGE_CACHE_SHIFT; | |
47 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); | |
6fb03f3a | 48 | } else if (type == NAT_ENTRIES) { |
e5e7ea3c JK |
49 | mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >> |
50 | PAGE_CACHE_SHIFT; | |
51 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); | |
6fb03f3a | 52 | } else if (type == DIRTY_DENTS) { |
2743f865 JK |
53 | if (sbi->sb->s_bdi->dirty_exceeded) |
54 | return false; | |
6fb03f3a | 55 | mem_size = get_pages(sbi, F2FS_DIRTY_DENTS); |
e5e7ea3c JK |
56 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
57 | } else if (type == INO_ENTRIES) { | |
58 | int i; | |
59 | ||
60 | if (sbi->sb->s_bdi->dirty_exceeded) | |
61 | return false; | |
62 | for (i = 0; i <= UPDATE_INO; i++) | |
67298804 CY |
63 | mem_size += (sbi->im[i].ino_num * |
64 | sizeof(struct ino_entry)) >> PAGE_CACHE_SHIFT; | |
e5e7ea3c | 65 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
6fb03f3a JK |
66 | } |
67 | return res; | |
cdfc41c1 JK |
68 | } |
69 | ||
e05df3b1 JK |
70 | static void clear_node_page_dirty(struct page *page) |
71 | { | |
72 | struct address_space *mapping = page->mapping; | |
e05df3b1 JK |
73 | unsigned int long flags; |
74 | ||
75 | if (PageDirty(page)) { | |
76 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
77 | radix_tree_tag_clear(&mapping->page_tree, | |
78 | page_index(page), | |
79 | PAGECACHE_TAG_DIRTY); | |
80 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
81 | ||
82 | clear_page_dirty_for_io(page); | |
4081363f | 83 | dec_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_NODES); |
e05df3b1 JK |
84 | } |
85 | ClearPageUptodate(page); | |
86 | } | |
87 | ||
88 | static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
89 | { | |
90 | pgoff_t index = current_nat_addr(sbi, nid); | |
91 | return get_meta_page(sbi, index); | |
92 | } | |
93 | ||
94 | static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
95 | { | |
96 | struct page *src_page; | |
97 | struct page *dst_page; | |
98 | pgoff_t src_off; | |
99 | pgoff_t dst_off; | |
100 | void *src_addr; | |
101 | void *dst_addr; | |
102 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
103 | ||
104 | src_off = current_nat_addr(sbi, nid); | |
105 | dst_off = next_nat_addr(sbi, src_off); | |
106 | ||
107 | /* get current nat block page with lock */ | |
108 | src_page = get_meta_page(sbi, src_off); | |
e05df3b1 | 109 | dst_page = grab_meta_page(sbi, dst_off); |
9850cf4a | 110 | f2fs_bug_on(sbi, PageDirty(src_page)); |
e05df3b1 JK |
111 | |
112 | src_addr = page_address(src_page); | |
113 | dst_addr = page_address(dst_page); | |
114 | memcpy(dst_addr, src_addr, PAGE_CACHE_SIZE); | |
115 | set_page_dirty(dst_page); | |
116 | f2fs_put_page(src_page, 1); | |
117 | ||
118 | set_to_next_nat(nm_i, nid); | |
119 | ||
120 | return dst_page; | |
121 | } | |
122 | ||
e05df3b1 JK |
123 | static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n) |
124 | { | |
125 | return radix_tree_lookup(&nm_i->nat_root, n); | |
126 | } | |
127 | ||
128 | static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i, | |
129 | nid_t start, unsigned int nr, struct nat_entry **ep) | |
130 | { | |
131 | return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr); | |
132 | } | |
133 | ||
134 | static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e) | |
135 | { | |
136 | list_del(&e->list); | |
137 | radix_tree_delete(&nm_i->nat_root, nat_get_nid(e)); | |
138 | nm_i->nat_cnt--; | |
139 | kmem_cache_free(nat_entry_slab, e); | |
140 | } | |
141 | ||
309cc2b6 JK |
142 | static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i, |
143 | struct nat_entry *ne) | |
144 | { | |
145 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); | |
146 | struct nat_entry_set *head; | |
147 | ||
148 | if (get_nat_flag(ne, IS_DIRTY)) | |
149 | return; | |
150 | retry: | |
151 | head = radix_tree_lookup(&nm_i->nat_set_root, set); | |
152 | if (!head) { | |
153 | head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_ATOMIC); | |
154 | ||
155 | INIT_LIST_HEAD(&head->entry_list); | |
156 | INIT_LIST_HEAD(&head->set_list); | |
157 | head->set = set; | |
158 | head->entry_cnt = 0; | |
159 | ||
160 | if (radix_tree_insert(&nm_i->nat_set_root, set, head)) { | |
4634d71e | 161 | kmem_cache_free(nat_entry_set_slab, head); |
309cc2b6 JK |
162 | goto retry; |
163 | } | |
164 | } | |
165 | list_move_tail(&ne->list, &head->entry_list); | |
166 | nm_i->dirty_nat_cnt++; | |
167 | head->entry_cnt++; | |
168 | set_nat_flag(ne, IS_DIRTY, true); | |
169 | } | |
170 | ||
171 | static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i, | |
172 | struct nat_entry *ne) | |
173 | { | |
20d047c8 | 174 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); |
309cc2b6 JK |
175 | struct nat_entry_set *head; |
176 | ||
177 | head = radix_tree_lookup(&nm_i->nat_set_root, set); | |
178 | if (head) { | |
179 | list_move_tail(&ne->list, &nm_i->nat_entries); | |
180 | set_nat_flag(ne, IS_DIRTY, false); | |
181 | head->entry_cnt--; | |
182 | nm_i->dirty_nat_cnt--; | |
183 | } | |
184 | } | |
185 | ||
186 | static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i, | |
187 | nid_t start, unsigned int nr, struct nat_entry_set **ep) | |
188 | { | |
189 | return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep, | |
190 | start, nr); | |
191 | } | |
192 | ||
88bd02c9 | 193 | bool is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 JK |
194 | { |
195 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
196 | struct nat_entry *e; | |
88bd02c9 | 197 | bool is_cp = true; |
e05df3b1 JK |
198 | |
199 | read_lock(&nm_i->nat_tree_lock); | |
200 | e = __lookup_nat_cache(nm_i, nid); | |
7ef35e3b | 201 | if (e && !get_nat_flag(e, IS_CHECKPOINTED)) |
88bd02c9 | 202 | is_cp = false; |
e05df3b1 JK |
203 | read_unlock(&nm_i->nat_tree_lock); |
204 | return is_cp; | |
205 | } | |
206 | ||
88bd02c9 | 207 | bool has_fsynced_inode(struct f2fs_sb_info *sbi, nid_t ino) |
479f40c4 JK |
208 | { |
209 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
210 | struct nat_entry *e; | |
88bd02c9 | 211 | bool fsynced = false; |
479f40c4 JK |
212 | |
213 | read_lock(&nm_i->nat_tree_lock); | |
88bd02c9 JK |
214 | e = __lookup_nat_cache(nm_i, ino); |
215 | if (e && get_nat_flag(e, HAS_FSYNCED_INODE)) | |
216 | fsynced = true; | |
479f40c4 | 217 | read_unlock(&nm_i->nat_tree_lock); |
88bd02c9 | 218 | return fsynced; |
479f40c4 JK |
219 | } |
220 | ||
88bd02c9 | 221 | bool need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino) |
b6fe5873 JK |
222 | { |
223 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
224 | struct nat_entry *e; | |
88bd02c9 | 225 | bool need_update = true; |
b6fe5873 | 226 | |
88bd02c9 JK |
227 | read_lock(&nm_i->nat_tree_lock); |
228 | e = __lookup_nat_cache(nm_i, ino); | |
229 | if (e && get_nat_flag(e, HAS_LAST_FSYNC) && | |
230 | (get_nat_flag(e, IS_CHECKPOINTED) || | |
231 | get_nat_flag(e, HAS_FSYNCED_INODE))) | |
232 | need_update = false; | |
233 | read_unlock(&nm_i->nat_tree_lock); | |
234 | return need_update; | |
b6fe5873 JK |
235 | } |
236 | ||
e05df3b1 JK |
237 | static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid) |
238 | { | |
239 | struct nat_entry *new; | |
240 | ||
241 | new = kmem_cache_alloc(nat_entry_slab, GFP_ATOMIC); | |
242 | if (!new) | |
243 | return NULL; | |
244 | if (radix_tree_insert(&nm_i->nat_root, nid, new)) { | |
245 | kmem_cache_free(nat_entry_slab, new); | |
246 | return NULL; | |
247 | } | |
248 | memset(new, 0, sizeof(struct nat_entry)); | |
249 | nat_set_nid(new, nid); | |
88bd02c9 | 250 | nat_reset_flag(new); |
e05df3b1 JK |
251 | list_add_tail(&new->list, &nm_i->nat_entries); |
252 | nm_i->nat_cnt++; | |
253 | return new; | |
254 | } | |
255 | ||
256 | static void cache_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid, | |
257 | struct f2fs_nat_entry *ne) | |
258 | { | |
259 | struct nat_entry *e; | |
260 | retry: | |
261 | write_lock(&nm_i->nat_tree_lock); | |
262 | e = __lookup_nat_cache(nm_i, nid); | |
263 | if (!e) { | |
264 | e = grab_nat_entry(nm_i, nid); | |
265 | if (!e) { | |
266 | write_unlock(&nm_i->nat_tree_lock); | |
267 | goto retry; | |
268 | } | |
94dac22e | 269 | node_info_from_raw_nat(&e->ni, ne); |
e05df3b1 JK |
270 | } |
271 | write_unlock(&nm_i->nat_tree_lock); | |
272 | } | |
273 | ||
274 | static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, | |
479f40c4 | 275 | block_t new_blkaddr, bool fsync_done) |
e05df3b1 JK |
276 | { |
277 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
278 | struct nat_entry *e; | |
279 | retry: | |
280 | write_lock(&nm_i->nat_tree_lock); | |
281 | e = __lookup_nat_cache(nm_i, ni->nid); | |
282 | if (!e) { | |
283 | e = grab_nat_entry(nm_i, ni->nid); | |
284 | if (!e) { | |
285 | write_unlock(&nm_i->nat_tree_lock); | |
286 | goto retry; | |
287 | } | |
288 | e->ni = *ni; | |
9850cf4a | 289 | f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR); |
e05df3b1 JK |
290 | } else if (new_blkaddr == NEW_ADDR) { |
291 | /* | |
292 | * when nid is reallocated, | |
293 | * previous nat entry can be remained in nat cache. | |
294 | * So, reinitialize it with new information. | |
295 | */ | |
296 | e->ni = *ni; | |
9850cf4a | 297 | f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR); |
e05df3b1 JK |
298 | } |
299 | ||
e05df3b1 | 300 | /* sanity check */ |
9850cf4a JK |
301 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr); |
302 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR && | |
e05df3b1 | 303 | new_blkaddr == NULL_ADDR); |
9850cf4a | 304 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR && |
e05df3b1 | 305 | new_blkaddr == NEW_ADDR); |
9850cf4a | 306 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != NEW_ADDR && |
e05df3b1 JK |
307 | nat_get_blkaddr(e) != NULL_ADDR && |
308 | new_blkaddr == NEW_ADDR); | |
309 | ||
e1c42045 | 310 | /* increment version no as node is removed */ |
e05df3b1 JK |
311 | if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) { |
312 | unsigned char version = nat_get_version(e); | |
313 | nat_set_version(e, inc_node_version(version)); | |
314 | } | |
315 | ||
316 | /* change address */ | |
317 | nat_set_blkaddr(e, new_blkaddr); | |
88bd02c9 JK |
318 | if (new_blkaddr == NEW_ADDR || new_blkaddr == NULL_ADDR) |
319 | set_nat_flag(e, IS_CHECKPOINTED, false); | |
e05df3b1 | 320 | __set_nat_cache_dirty(nm_i, e); |
479f40c4 JK |
321 | |
322 | /* update fsync_mark if its inode nat entry is still alive */ | |
323 | e = __lookup_nat_cache(nm_i, ni->ino); | |
88bd02c9 JK |
324 | if (e) { |
325 | if (fsync_done && ni->nid == ni->ino) | |
326 | set_nat_flag(e, HAS_FSYNCED_INODE, true); | |
327 | set_nat_flag(e, HAS_LAST_FSYNC, fsync_done); | |
328 | } | |
e05df3b1 JK |
329 | write_unlock(&nm_i->nat_tree_lock); |
330 | } | |
331 | ||
4660f9c0 | 332 | int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) |
e05df3b1 JK |
333 | { |
334 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
335 | ||
6fb03f3a | 336 | if (available_free_memory(sbi, NAT_ENTRIES)) |
e05df3b1 JK |
337 | return 0; |
338 | ||
339 | write_lock(&nm_i->nat_tree_lock); | |
340 | while (nr_shrink && !list_empty(&nm_i->nat_entries)) { | |
341 | struct nat_entry *ne; | |
342 | ne = list_first_entry(&nm_i->nat_entries, | |
343 | struct nat_entry, list); | |
344 | __del_from_nat_cache(nm_i, ne); | |
345 | nr_shrink--; | |
346 | } | |
347 | write_unlock(&nm_i->nat_tree_lock); | |
348 | return nr_shrink; | |
349 | } | |
350 | ||
0a8165d7 | 351 | /* |
e1c42045 | 352 | * This function always returns success |
e05df3b1 JK |
353 | */ |
354 | void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni) | |
355 | { | |
356 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
357 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
358 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
359 | nid_t start_nid = START_NID(nid); | |
360 | struct f2fs_nat_block *nat_blk; | |
361 | struct page *page = NULL; | |
362 | struct f2fs_nat_entry ne; | |
363 | struct nat_entry *e; | |
364 | int i; | |
365 | ||
be4124f8 | 366 | memset(&ne, 0, sizeof(struct f2fs_nat_entry)); |
e05df3b1 JK |
367 | ni->nid = nid; |
368 | ||
369 | /* Check nat cache */ | |
370 | read_lock(&nm_i->nat_tree_lock); | |
371 | e = __lookup_nat_cache(nm_i, nid); | |
372 | if (e) { | |
373 | ni->ino = nat_get_ino(e); | |
374 | ni->blk_addr = nat_get_blkaddr(e); | |
375 | ni->version = nat_get_version(e); | |
376 | } | |
377 | read_unlock(&nm_i->nat_tree_lock); | |
378 | if (e) | |
379 | return; | |
380 | ||
381 | /* Check current segment summary */ | |
382 | mutex_lock(&curseg->curseg_mutex); | |
383 | i = lookup_journal_in_cursum(sum, NAT_JOURNAL, nid, 0); | |
384 | if (i >= 0) { | |
385 | ne = nat_in_journal(sum, i); | |
386 | node_info_from_raw_nat(ni, &ne); | |
387 | } | |
388 | mutex_unlock(&curseg->curseg_mutex); | |
389 | if (i >= 0) | |
390 | goto cache; | |
391 | ||
392 | /* Fill node_info from nat page */ | |
393 | page = get_current_nat_page(sbi, start_nid); | |
394 | nat_blk = (struct f2fs_nat_block *)page_address(page); | |
395 | ne = nat_blk->entries[nid - start_nid]; | |
396 | node_info_from_raw_nat(ni, &ne); | |
397 | f2fs_put_page(page, 1); | |
398 | cache: | |
399 | /* cache nat entry */ | |
400 | cache_nat_entry(NM_I(sbi), nid, &ne); | |
401 | } | |
402 | ||
0a8165d7 | 403 | /* |
e05df3b1 JK |
404 | * The maximum depth is four. |
405 | * Offset[0] will have raw inode offset. | |
406 | */ | |
de93653f JK |
407 | static int get_node_path(struct f2fs_inode_info *fi, long block, |
408 | int offset[4], unsigned int noffset[4]) | |
e05df3b1 | 409 | { |
de93653f | 410 | const long direct_index = ADDRS_PER_INODE(fi); |
e05df3b1 JK |
411 | const long direct_blks = ADDRS_PER_BLOCK; |
412 | const long dptrs_per_blk = NIDS_PER_BLOCK; | |
413 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
414 | const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK; | |
415 | int n = 0; | |
416 | int level = 0; | |
417 | ||
418 | noffset[0] = 0; | |
419 | ||
420 | if (block < direct_index) { | |
25c0a6e5 | 421 | offset[n] = block; |
e05df3b1 JK |
422 | goto got; |
423 | } | |
424 | block -= direct_index; | |
425 | if (block < direct_blks) { | |
426 | offset[n++] = NODE_DIR1_BLOCK; | |
427 | noffset[n] = 1; | |
25c0a6e5 | 428 | offset[n] = block; |
e05df3b1 JK |
429 | level = 1; |
430 | goto got; | |
431 | } | |
432 | block -= direct_blks; | |
433 | if (block < direct_blks) { | |
434 | offset[n++] = NODE_DIR2_BLOCK; | |
435 | noffset[n] = 2; | |
25c0a6e5 | 436 | offset[n] = block; |
e05df3b1 JK |
437 | level = 1; |
438 | goto got; | |
439 | } | |
440 | block -= direct_blks; | |
441 | if (block < indirect_blks) { | |
442 | offset[n++] = NODE_IND1_BLOCK; | |
443 | noffset[n] = 3; | |
444 | offset[n++] = block / direct_blks; | |
445 | noffset[n] = 4 + offset[n - 1]; | |
25c0a6e5 | 446 | offset[n] = block % direct_blks; |
e05df3b1 JK |
447 | level = 2; |
448 | goto got; | |
449 | } | |
450 | block -= indirect_blks; | |
451 | if (block < indirect_blks) { | |
452 | offset[n++] = NODE_IND2_BLOCK; | |
453 | noffset[n] = 4 + dptrs_per_blk; | |
454 | offset[n++] = block / direct_blks; | |
455 | noffset[n] = 5 + dptrs_per_blk + offset[n - 1]; | |
25c0a6e5 | 456 | offset[n] = block % direct_blks; |
e05df3b1 JK |
457 | level = 2; |
458 | goto got; | |
459 | } | |
460 | block -= indirect_blks; | |
461 | if (block < dindirect_blks) { | |
462 | offset[n++] = NODE_DIND_BLOCK; | |
463 | noffset[n] = 5 + (dptrs_per_blk * 2); | |
464 | offset[n++] = block / indirect_blks; | |
465 | noffset[n] = 6 + (dptrs_per_blk * 2) + | |
466 | offset[n - 1] * (dptrs_per_blk + 1); | |
467 | offset[n++] = (block / direct_blks) % dptrs_per_blk; | |
468 | noffset[n] = 7 + (dptrs_per_blk * 2) + | |
469 | offset[n - 2] * (dptrs_per_blk + 1) + | |
470 | offset[n - 1]; | |
25c0a6e5 | 471 | offset[n] = block % direct_blks; |
e05df3b1 JK |
472 | level = 3; |
473 | goto got; | |
474 | } else { | |
475 | BUG(); | |
476 | } | |
477 | got: | |
478 | return level; | |
479 | } | |
480 | ||
481 | /* | |
482 | * Caller should call f2fs_put_dnode(dn). | |
4f4124d0 CY |
483 | * Also, it should grab and release a rwsem by calling f2fs_lock_op() and |
484 | * f2fs_unlock_op() only if ro is not set RDONLY_NODE. | |
39936837 | 485 | * In the case of RDONLY_NODE, we don't need to care about mutex. |
e05df3b1 | 486 | */ |
266e97a8 | 487 | int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode) |
e05df3b1 | 488 | { |
4081363f | 489 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
490 | struct page *npage[4]; |
491 | struct page *parent; | |
492 | int offset[4]; | |
493 | unsigned int noffset[4]; | |
494 | nid_t nids[4]; | |
495 | int level, i; | |
496 | int err = 0; | |
497 | ||
de93653f | 498 | level = get_node_path(F2FS_I(dn->inode), index, offset, noffset); |
e05df3b1 JK |
499 | |
500 | nids[0] = dn->inode->i_ino; | |
1646cfac | 501 | npage[0] = dn->inode_page; |
e05df3b1 | 502 | |
1646cfac JK |
503 | if (!npage[0]) { |
504 | npage[0] = get_node_page(sbi, nids[0]); | |
505 | if (IS_ERR(npage[0])) | |
506 | return PTR_ERR(npage[0]); | |
507 | } | |
e05df3b1 | 508 | parent = npage[0]; |
52c2db3f CL |
509 | if (level != 0) |
510 | nids[1] = get_nid(parent, offset[0], true); | |
e05df3b1 JK |
511 | dn->inode_page = npage[0]; |
512 | dn->inode_page_locked = true; | |
513 | ||
514 | /* get indirect or direct nodes */ | |
515 | for (i = 1; i <= level; i++) { | |
516 | bool done = false; | |
517 | ||
266e97a8 | 518 | if (!nids[i] && mode == ALLOC_NODE) { |
e05df3b1 JK |
519 | /* alloc new node */ |
520 | if (!alloc_nid(sbi, &(nids[i]))) { | |
e05df3b1 JK |
521 | err = -ENOSPC; |
522 | goto release_pages; | |
523 | } | |
524 | ||
525 | dn->nid = nids[i]; | |
8ae8f162 | 526 | npage[i] = new_node_page(dn, noffset[i], NULL); |
e05df3b1 JK |
527 | if (IS_ERR(npage[i])) { |
528 | alloc_nid_failed(sbi, nids[i]); | |
e05df3b1 JK |
529 | err = PTR_ERR(npage[i]); |
530 | goto release_pages; | |
531 | } | |
532 | ||
533 | set_nid(parent, offset[i - 1], nids[i], i == 1); | |
534 | alloc_nid_done(sbi, nids[i]); | |
e05df3b1 | 535 | done = true; |
266e97a8 | 536 | } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) { |
e05df3b1 JK |
537 | npage[i] = get_node_page_ra(parent, offset[i - 1]); |
538 | if (IS_ERR(npage[i])) { | |
539 | err = PTR_ERR(npage[i]); | |
540 | goto release_pages; | |
541 | } | |
542 | done = true; | |
543 | } | |
544 | if (i == 1) { | |
545 | dn->inode_page_locked = false; | |
546 | unlock_page(parent); | |
547 | } else { | |
548 | f2fs_put_page(parent, 1); | |
549 | } | |
550 | ||
551 | if (!done) { | |
552 | npage[i] = get_node_page(sbi, nids[i]); | |
553 | if (IS_ERR(npage[i])) { | |
554 | err = PTR_ERR(npage[i]); | |
555 | f2fs_put_page(npage[0], 0); | |
556 | goto release_out; | |
557 | } | |
558 | } | |
559 | if (i < level) { | |
560 | parent = npage[i]; | |
561 | nids[i + 1] = get_nid(parent, offset[i], false); | |
562 | } | |
563 | } | |
564 | dn->nid = nids[level]; | |
565 | dn->ofs_in_node = offset[level]; | |
566 | dn->node_page = npage[level]; | |
567 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
568 | return 0; | |
569 | ||
570 | release_pages: | |
571 | f2fs_put_page(parent, 1); | |
572 | if (i > 1) | |
573 | f2fs_put_page(npage[0], 0); | |
574 | release_out: | |
575 | dn->inode_page = NULL; | |
576 | dn->node_page = NULL; | |
577 | return err; | |
578 | } | |
579 | ||
580 | static void truncate_node(struct dnode_of_data *dn) | |
581 | { | |
4081363f | 582 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
583 | struct node_info ni; |
584 | ||
585 | get_node_info(sbi, dn->nid, &ni); | |
71e9fec5 | 586 | if (dn->inode->i_blocks == 0) { |
9850cf4a | 587 | f2fs_bug_on(sbi, ni.blk_addr != NULL_ADDR); |
71e9fec5 JK |
588 | goto invalidate; |
589 | } | |
9850cf4a | 590 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
e05df3b1 | 591 | |
e05df3b1 | 592 | /* Deallocate node address */ |
71e9fec5 | 593 | invalidate_blocks(sbi, ni.blk_addr); |
ef86d709 | 594 | dec_valid_node_count(sbi, dn->inode); |
479f40c4 | 595 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
e05df3b1 JK |
596 | |
597 | if (dn->nid == dn->inode->i_ino) { | |
598 | remove_orphan_inode(sbi, dn->nid); | |
599 | dec_valid_inode_count(sbi); | |
600 | } else { | |
601 | sync_inode_page(dn); | |
602 | } | |
71e9fec5 | 603 | invalidate: |
e05df3b1 JK |
604 | clear_node_page_dirty(dn->node_page); |
605 | F2FS_SET_SB_DIRT(sbi); | |
606 | ||
607 | f2fs_put_page(dn->node_page, 1); | |
bf39c00a JK |
608 | |
609 | invalidate_mapping_pages(NODE_MAPPING(sbi), | |
610 | dn->node_page->index, dn->node_page->index); | |
611 | ||
e05df3b1 | 612 | dn->node_page = NULL; |
51dd6249 | 613 | trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr); |
e05df3b1 JK |
614 | } |
615 | ||
616 | static int truncate_dnode(struct dnode_of_data *dn) | |
617 | { | |
e05df3b1 JK |
618 | struct page *page; |
619 | ||
620 | if (dn->nid == 0) | |
621 | return 1; | |
622 | ||
623 | /* get direct node */ | |
4081363f | 624 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
e05df3b1 JK |
625 | if (IS_ERR(page) && PTR_ERR(page) == -ENOENT) |
626 | return 1; | |
627 | else if (IS_ERR(page)) | |
628 | return PTR_ERR(page); | |
629 | ||
630 | /* Make dnode_of_data for parameter */ | |
631 | dn->node_page = page; | |
632 | dn->ofs_in_node = 0; | |
633 | truncate_data_blocks(dn); | |
634 | truncate_node(dn); | |
635 | return 1; | |
636 | } | |
637 | ||
638 | static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs, | |
639 | int ofs, int depth) | |
640 | { | |
e05df3b1 JK |
641 | struct dnode_of_data rdn = *dn; |
642 | struct page *page; | |
643 | struct f2fs_node *rn; | |
644 | nid_t child_nid; | |
645 | unsigned int child_nofs; | |
646 | int freed = 0; | |
647 | int i, ret; | |
648 | ||
649 | if (dn->nid == 0) | |
650 | return NIDS_PER_BLOCK + 1; | |
651 | ||
51dd6249 NJ |
652 | trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr); |
653 | ||
4081363f | 654 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
51dd6249 NJ |
655 | if (IS_ERR(page)) { |
656 | trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page)); | |
e05df3b1 | 657 | return PTR_ERR(page); |
51dd6249 | 658 | } |
e05df3b1 | 659 | |
45590710 | 660 | rn = F2FS_NODE(page); |
e05df3b1 JK |
661 | if (depth < 3) { |
662 | for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) { | |
663 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
664 | if (child_nid == 0) | |
665 | continue; | |
666 | rdn.nid = child_nid; | |
667 | ret = truncate_dnode(&rdn); | |
668 | if (ret < 0) | |
669 | goto out_err; | |
670 | set_nid(page, i, 0, false); | |
671 | } | |
672 | } else { | |
673 | child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1; | |
674 | for (i = ofs; i < NIDS_PER_BLOCK; i++) { | |
675 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
676 | if (child_nid == 0) { | |
677 | child_nofs += NIDS_PER_BLOCK + 1; | |
678 | continue; | |
679 | } | |
680 | rdn.nid = child_nid; | |
681 | ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1); | |
682 | if (ret == (NIDS_PER_BLOCK + 1)) { | |
683 | set_nid(page, i, 0, false); | |
684 | child_nofs += ret; | |
685 | } else if (ret < 0 && ret != -ENOENT) { | |
686 | goto out_err; | |
687 | } | |
688 | } | |
689 | freed = child_nofs; | |
690 | } | |
691 | ||
692 | if (!ofs) { | |
693 | /* remove current indirect node */ | |
694 | dn->node_page = page; | |
695 | truncate_node(dn); | |
696 | freed++; | |
697 | } else { | |
698 | f2fs_put_page(page, 1); | |
699 | } | |
51dd6249 | 700 | trace_f2fs_truncate_nodes_exit(dn->inode, freed); |
e05df3b1 JK |
701 | return freed; |
702 | ||
703 | out_err: | |
704 | f2fs_put_page(page, 1); | |
51dd6249 | 705 | trace_f2fs_truncate_nodes_exit(dn->inode, ret); |
e05df3b1 JK |
706 | return ret; |
707 | } | |
708 | ||
709 | static int truncate_partial_nodes(struct dnode_of_data *dn, | |
710 | struct f2fs_inode *ri, int *offset, int depth) | |
711 | { | |
e05df3b1 JK |
712 | struct page *pages[2]; |
713 | nid_t nid[3]; | |
714 | nid_t child_nid; | |
715 | int err = 0; | |
716 | int i; | |
717 | int idx = depth - 2; | |
718 | ||
719 | nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); | |
720 | if (!nid[0]) | |
721 | return 0; | |
722 | ||
723 | /* get indirect nodes in the path */ | |
a225dca3 | 724 | for (i = 0; i < idx + 1; i++) { |
e1c42045 | 725 | /* reference count'll be increased */ |
4081363f | 726 | pages[i] = get_node_page(F2FS_I_SB(dn->inode), nid[i]); |
e05df3b1 | 727 | if (IS_ERR(pages[i])) { |
e05df3b1 | 728 | err = PTR_ERR(pages[i]); |
a225dca3 | 729 | idx = i - 1; |
e05df3b1 JK |
730 | goto fail; |
731 | } | |
732 | nid[i + 1] = get_nid(pages[i], offset[i + 1], false); | |
733 | } | |
734 | ||
735 | /* free direct nodes linked to a partial indirect node */ | |
a225dca3 | 736 | for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) { |
e05df3b1 JK |
737 | child_nid = get_nid(pages[idx], i, false); |
738 | if (!child_nid) | |
739 | continue; | |
740 | dn->nid = child_nid; | |
741 | err = truncate_dnode(dn); | |
742 | if (err < 0) | |
743 | goto fail; | |
744 | set_nid(pages[idx], i, 0, false); | |
745 | } | |
746 | ||
a225dca3 | 747 | if (offset[idx + 1] == 0) { |
e05df3b1 JK |
748 | dn->node_page = pages[idx]; |
749 | dn->nid = nid[idx]; | |
750 | truncate_node(dn); | |
751 | } else { | |
752 | f2fs_put_page(pages[idx], 1); | |
753 | } | |
754 | offset[idx]++; | |
a225dca3 | 755 | offset[idx + 1] = 0; |
756 | idx--; | |
e05df3b1 | 757 | fail: |
a225dca3 | 758 | for (i = idx; i >= 0; i--) |
e05df3b1 | 759 | f2fs_put_page(pages[i], 1); |
51dd6249 NJ |
760 | |
761 | trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err); | |
762 | ||
e05df3b1 JK |
763 | return err; |
764 | } | |
765 | ||
0a8165d7 | 766 | /* |
e05df3b1 JK |
767 | * All the block addresses of data and nodes should be nullified. |
768 | */ | |
769 | int truncate_inode_blocks(struct inode *inode, pgoff_t from) | |
770 | { | |
4081363f | 771 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
e05df3b1 JK |
772 | int err = 0, cont = 1; |
773 | int level, offset[4], noffset[4]; | |
7dd690c8 | 774 | unsigned int nofs = 0; |
58bfaf44 | 775 | struct f2fs_inode *ri; |
e05df3b1 JK |
776 | struct dnode_of_data dn; |
777 | struct page *page; | |
778 | ||
51dd6249 NJ |
779 | trace_f2fs_truncate_inode_blocks_enter(inode, from); |
780 | ||
de93653f | 781 | level = get_node_path(F2FS_I(inode), from, offset, noffset); |
afcb7ca0 | 782 | restart: |
e05df3b1 | 783 | page = get_node_page(sbi, inode->i_ino); |
51dd6249 NJ |
784 | if (IS_ERR(page)) { |
785 | trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page)); | |
e05df3b1 | 786 | return PTR_ERR(page); |
51dd6249 | 787 | } |
e05df3b1 JK |
788 | |
789 | set_new_dnode(&dn, inode, page, NULL, 0); | |
790 | unlock_page(page); | |
791 | ||
58bfaf44 | 792 | ri = F2FS_INODE(page); |
e05df3b1 JK |
793 | switch (level) { |
794 | case 0: | |
795 | case 1: | |
796 | nofs = noffset[1]; | |
797 | break; | |
798 | case 2: | |
799 | nofs = noffset[1]; | |
800 | if (!offset[level - 1]) | |
801 | goto skip_partial; | |
58bfaf44 | 802 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
803 | if (err < 0 && err != -ENOENT) |
804 | goto fail; | |
805 | nofs += 1 + NIDS_PER_BLOCK; | |
806 | break; | |
807 | case 3: | |
808 | nofs = 5 + 2 * NIDS_PER_BLOCK; | |
809 | if (!offset[level - 1]) | |
810 | goto skip_partial; | |
58bfaf44 | 811 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
812 | if (err < 0 && err != -ENOENT) |
813 | goto fail; | |
814 | break; | |
815 | default: | |
816 | BUG(); | |
817 | } | |
818 | ||
819 | skip_partial: | |
820 | while (cont) { | |
58bfaf44 | 821 | dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); |
e05df3b1 JK |
822 | switch (offset[0]) { |
823 | case NODE_DIR1_BLOCK: | |
824 | case NODE_DIR2_BLOCK: | |
825 | err = truncate_dnode(&dn); | |
826 | break; | |
827 | ||
828 | case NODE_IND1_BLOCK: | |
829 | case NODE_IND2_BLOCK: | |
830 | err = truncate_nodes(&dn, nofs, offset[1], 2); | |
831 | break; | |
832 | ||
833 | case NODE_DIND_BLOCK: | |
834 | err = truncate_nodes(&dn, nofs, offset[1], 3); | |
835 | cont = 0; | |
836 | break; | |
837 | ||
838 | default: | |
839 | BUG(); | |
840 | } | |
841 | if (err < 0 && err != -ENOENT) | |
842 | goto fail; | |
843 | if (offset[1] == 0 && | |
58bfaf44 | 844 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) { |
e05df3b1 | 845 | lock_page(page); |
4ef51a8f | 846 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
847 | f2fs_put_page(page, 1); |
848 | goto restart; | |
849 | } | |
3cb5ad15 | 850 | f2fs_wait_on_page_writeback(page, NODE); |
58bfaf44 | 851 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0; |
e05df3b1 JK |
852 | set_page_dirty(page); |
853 | unlock_page(page); | |
854 | } | |
855 | offset[1] = 0; | |
856 | offset[0]++; | |
857 | nofs += err; | |
858 | } | |
859 | fail: | |
860 | f2fs_put_page(page, 0); | |
51dd6249 | 861 | trace_f2fs_truncate_inode_blocks_exit(inode, err); |
e05df3b1 JK |
862 | return err > 0 ? 0 : err; |
863 | } | |
864 | ||
4f16fb0f JK |
865 | int truncate_xattr_node(struct inode *inode, struct page *page) |
866 | { | |
4081363f | 867 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
4f16fb0f JK |
868 | nid_t nid = F2FS_I(inode)->i_xattr_nid; |
869 | struct dnode_of_data dn; | |
870 | struct page *npage; | |
871 | ||
872 | if (!nid) | |
873 | return 0; | |
874 | ||
875 | npage = get_node_page(sbi, nid); | |
876 | if (IS_ERR(npage)) | |
877 | return PTR_ERR(npage); | |
878 | ||
879 | F2FS_I(inode)->i_xattr_nid = 0; | |
65985d93 JK |
880 | |
881 | /* need to do checkpoint during fsync */ | |
882 | F2FS_I(inode)->xattr_ver = cur_cp_version(F2FS_CKPT(sbi)); | |
883 | ||
4f16fb0f JK |
884 | set_new_dnode(&dn, inode, page, npage, nid); |
885 | ||
886 | if (page) | |
01d2d1aa | 887 | dn.inode_page_locked = true; |
4f16fb0f JK |
888 | truncate_node(&dn); |
889 | return 0; | |
890 | } | |
891 | ||
39936837 | 892 | /* |
4f4124d0 CY |
893 | * Caller should grab and release a rwsem by calling f2fs_lock_op() and |
894 | * f2fs_unlock_op(). | |
39936837 | 895 | */ |
58e674d6 | 896 | void remove_inode_page(struct inode *inode) |
e05df3b1 | 897 | { |
e05df3b1 JK |
898 | struct dnode_of_data dn; |
899 | ||
c2e69583 JK |
900 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); |
901 | if (get_dnode_of_data(&dn, 0, LOOKUP_NODE)) | |
58e674d6 | 902 | return; |
e05df3b1 | 903 | |
c2e69583 JK |
904 | if (truncate_xattr_node(inode, dn.inode_page)) { |
905 | f2fs_put_dnode(&dn); | |
58e674d6 | 906 | return; |
e05df3b1 | 907 | } |
c2e69583 JK |
908 | |
909 | /* remove potential inline_data blocks */ | |
910 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
911 | S_ISLNK(inode->i_mode)) | |
912 | truncate_data_blocks_range(&dn, 1); | |
913 | ||
e1c42045 | 914 | /* 0 is possible, after f2fs_new_inode() has failed */ |
9850cf4a JK |
915 | f2fs_bug_on(F2FS_I_SB(inode), |
916 | inode->i_blocks != 0 && inode->i_blocks != 1); | |
c2e69583 JK |
917 | |
918 | /* will put inode & node pages */ | |
71e9fec5 | 919 | truncate_node(&dn); |
e05df3b1 JK |
920 | } |
921 | ||
a014e037 | 922 | struct page *new_inode_page(struct inode *inode) |
e05df3b1 | 923 | { |
e05df3b1 JK |
924 | struct dnode_of_data dn; |
925 | ||
926 | /* allocate inode page for new inode */ | |
927 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); | |
44a83ff6 JK |
928 | |
929 | /* caller should f2fs_put_page(page, 1); */ | |
8ae8f162 | 930 | return new_node_page(&dn, 0, NULL); |
e05df3b1 JK |
931 | } |
932 | ||
8ae8f162 JK |
933 | struct page *new_node_page(struct dnode_of_data *dn, |
934 | unsigned int ofs, struct page *ipage) | |
e05df3b1 | 935 | { |
4081363f | 936 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
937 | struct node_info old_ni, new_ni; |
938 | struct page *page; | |
939 | int err; | |
940 | ||
6bacf52f | 941 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
e05df3b1 JK |
942 | return ERR_PTR(-EPERM); |
943 | ||
54b591df | 944 | page = grab_cache_page(NODE_MAPPING(sbi), dn->nid); |
e05df3b1 JK |
945 | if (!page) |
946 | return ERR_PTR(-ENOMEM); | |
947 | ||
6bacf52f | 948 | if (unlikely(!inc_valid_node_count(sbi, dn->inode))) { |
9c02740c JK |
949 | err = -ENOSPC; |
950 | goto fail; | |
951 | } | |
e05df3b1 | 952 | |
9c02740c | 953 | get_node_info(sbi, dn->nid, &old_ni); |
e05df3b1 JK |
954 | |
955 | /* Reinitialize old_ni with new node page */ | |
9850cf4a | 956 | f2fs_bug_on(sbi, old_ni.blk_addr != NULL_ADDR); |
e05df3b1 JK |
957 | new_ni = old_ni; |
958 | new_ni.ino = dn->inode->i_ino; | |
479f40c4 | 959 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
9c02740c | 960 | |
54b591df | 961 | f2fs_wait_on_page_writeback(page, NODE); |
9c02740c | 962 | fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true); |
398b1ac5 | 963 | set_cold_node(dn->inode, page); |
9c02740c JK |
964 | SetPageUptodate(page); |
965 | set_page_dirty(page); | |
e05df3b1 | 966 | |
4bc8e9bc | 967 | if (f2fs_has_xattr_block(ofs)) |
479bd73a JK |
968 | F2FS_I(dn->inode)->i_xattr_nid = dn->nid; |
969 | ||
e05df3b1 | 970 | dn->node_page = page; |
8ae8f162 JK |
971 | if (ipage) |
972 | update_inode(dn->inode, ipage); | |
973 | else | |
974 | sync_inode_page(dn); | |
e05df3b1 JK |
975 | if (ofs == 0) |
976 | inc_valid_inode_count(sbi); | |
977 | ||
978 | return page; | |
979 | ||
980 | fail: | |
71e9fec5 | 981 | clear_node_page_dirty(page); |
e05df3b1 JK |
982 | f2fs_put_page(page, 1); |
983 | return ERR_PTR(err); | |
984 | } | |
985 | ||
56ae674c JK |
986 | /* |
987 | * Caller should do after getting the following values. | |
988 | * 0: f2fs_put_page(page, 0) | |
989 | * LOCKED_PAGE: f2fs_put_page(page, 1) | |
990 | * error: nothing | |
991 | */ | |
93dfe2ac | 992 | static int read_node_page(struct page *page, int rw) |
e05df3b1 | 993 | { |
4081363f | 994 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 JK |
995 | struct node_info ni; |
996 | ||
997 | get_node_info(sbi, page->index, &ni); | |
998 | ||
6bacf52f | 999 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
393ff91f | 1000 | f2fs_put_page(page, 1); |
e05df3b1 | 1001 | return -ENOENT; |
393ff91f JK |
1002 | } |
1003 | ||
56ae674c JK |
1004 | if (PageUptodate(page)) |
1005 | return LOCKED_PAGE; | |
393ff91f | 1006 | |
93dfe2ac | 1007 | return f2fs_submit_page_bio(sbi, page, ni.blk_addr, rw); |
e05df3b1 JK |
1008 | } |
1009 | ||
0a8165d7 | 1010 | /* |
e05df3b1 JK |
1011 | * Readahead a node page |
1012 | */ | |
1013 | void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid) | |
1014 | { | |
e05df3b1 | 1015 | struct page *apage; |
56ae674c | 1016 | int err; |
e05df3b1 | 1017 | |
4ef51a8f | 1018 | apage = find_get_page(NODE_MAPPING(sbi), nid); |
393ff91f JK |
1019 | if (apage && PageUptodate(apage)) { |
1020 | f2fs_put_page(apage, 0); | |
1021 | return; | |
1022 | } | |
e05df3b1 JK |
1023 | f2fs_put_page(apage, 0); |
1024 | ||
4ef51a8f | 1025 | apage = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1026 | if (!apage) |
1027 | return; | |
1028 | ||
56ae674c JK |
1029 | err = read_node_page(apage, READA); |
1030 | if (err == 0) | |
393ff91f | 1031 | f2fs_put_page(apage, 0); |
56ae674c JK |
1032 | else if (err == LOCKED_PAGE) |
1033 | f2fs_put_page(apage, 1); | |
e05df3b1 JK |
1034 | } |
1035 | ||
1036 | struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) | |
1037 | { | |
56ae674c JK |
1038 | struct page *page; |
1039 | int err; | |
afcb7ca0 | 1040 | repeat: |
54b591df | 1041 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1042 | if (!page) |
1043 | return ERR_PTR(-ENOMEM); | |
1044 | ||
1045 | err = read_node_page(page, READ_SYNC); | |
56ae674c | 1046 | if (err < 0) |
e05df3b1 | 1047 | return ERR_PTR(err); |
56ae674c JK |
1048 | else if (err == LOCKED_PAGE) |
1049 | goto got_it; | |
e05df3b1 | 1050 | |
393ff91f | 1051 | lock_page(page); |
3bb5e2c8 | 1052 | if (unlikely(!PageUptodate(page) || nid != nid_of_node(page))) { |
393ff91f JK |
1053 | f2fs_put_page(page, 1); |
1054 | return ERR_PTR(-EIO); | |
1055 | } | |
4ef51a8f | 1056 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1057 | f2fs_put_page(page, 1); |
1058 | goto repeat; | |
1059 | } | |
56ae674c | 1060 | got_it: |
e05df3b1 JK |
1061 | return page; |
1062 | } | |
1063 | ||
0a8165d7 | 1064 | /* |
e05df3b1 JK |
1065 | * Return a locked page for the desired node page. |
1066 | * And, readahead MAX_RA_NODE number of node pages. | |
1067 | */ | |
1068 | struct page *get_node_page_ra(struct page *parent, int start) | |
1069 | { | |
4081363f | 1070 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); |
c718379b | 1071 | struct blk_plug plug; |
e05df3b1 | 1072 | struct page *page; |
56ae674c JK |
1073 | int err, i, end; |
1074 | nid_t nid; | |
e05df3b1 JK |
1075 | |
1076 | /* First, try getting the desired direct node. */ | |
1077 | nid = get_nid(parent, start, false); | |
1078 | if (!nid) | |
1079 | return ERR_PTR(-ENOENT); | |
afcb7ca0 | 1080 | repeat: |
4ef51a8f | 1081 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1082 | if (!page) |
1083 | return ERR_PTR(-ENOMEM); | |
1084 | ||
66d36a29 | 1085 | err = read_node_page(page, READ_SYNC); |
56ae674c | 1086 | if (err < 0) |
e05df3b1 | 1087 | return ERR_PTR(err); |
56ae674c JK |
1088 | else if (err == LOCKED_PAGE) |
1089 | goto page_hit; | |
e05df3b1 | 1090 | |
c718379b JK |
1091 | blk_start_plug(&plug); |
1092 | ||
e05df3b1 JK |
1093 | /* Then, try readahead for siblings of the desired node */ |
1094 | end = start + MAX_RA_NODE; | |
1095 | end = min(end, NIDS_PER_BLOCK); | |
1096 | for (i = start + 1; i < end; i++) { | |
1097 | nid = get_nid(parent, i, false); | |
1098 | if (!nid) | |
1099 | continue; | |
1100 | ra_node_page(sbi, nid); | |
1101 | } | |
1102 | ||
c718379b JK |
1103 | blk_finish_plug(&plug); |
1104 | ||
e05df3b1 | 1105 | lock_page(page); |
4ef51a8f | 1106 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1107 | f2fs_put_page(page, 1); |
1108 | goto repeat; | |
1109 | } | |
e0f56cb4 | 1110 | page_hit: |
6bacf52f | 1111 | if (unlikely(!PageUptodate(page))) { |
e05df3b1 JK |
1112 | f2fs_put_page(page, 1); |
1113 | return ERR_PTR(-EIO); | |
1114 | } | |
e05df3b1 JK |
1115 | return page; |
1116 | } | |
1117 | ||
1118 | void sync_inode_page(struct dnode_of_data *dn) | |
1119 | { | |
1120 | if (IS_INODE(dn->node_page) || dn->inode_page == dn->node_page) { | |
1121 | update_inode(dn->inode, dn->node_page); | |
1122 | } else if (dn->inode_page) { | |
1123 | if (!dn->inode_page_locked) | |
1124 | lock_page(dn->inode_page); | |
1125 | update_inode(dn->inode, dn->inode_page); | |
1126 | if (!dn->inode_page_locked) | |
1127 | unlock_page(dn->inode_page); | |
1128 | } else { | |
39936837 | 1129 | update_inode_page(dn->inode); |
e05df3b1 JK |
1130 | } |
1131 | } | |
1132 | ||
1133 | int sync_node_pages(struct f2fs_sb_info *sbi, nid_t ino, | |
1134 | struct writeback_control *wbc) | |
1135 | { | |
e05df3b1 JK |
1136 | pgoff_t index, end; |
1137 | struct pagevec pvec; | |
1138 | int step = ino ? 2 : 0; | |
1139 | int nwritten = 0, wrote = 0; | |
1140 | ||
1141 | pagevec_init(&pvec, 0); | |
1142 | ||
1143 | next_step: | |
1144 | index = 0; | |
1145 | end = LONG_MAX; | |
1146 | ||
1147 | while (index <= end) { | |
1148 | int i, nr_pages; | |
4ef51a8f | 1149 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, |
e05df3b1 JK |
1150 | PAGECACHE_TAG_DIRTY, |
1151 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1152 | if (nr_pages == 0) | |
1153 | break; | |
1154 | ||
1155 | for (i = 0; i < nr_pages; i++) { | |
1156 | struct page *page = pvec.pages[i]; | |
1157 | ||
1158 | /* | |
1159 | * flushing sequence with step: | |
1160 | * 0. indirect nodes | |
1161 | * 1. dentry dnodes | |
1162 | * 2. file dnodes | |
1163 | */ | |
1164 | if (step == 0 && IS_DNODE(page)) | |
1165 | continue; | |
1166 | if (step == 1 && (!IS_DNODE(page) || | |
1167 | is_cold_node(page))) | |
1168 | continue; | |
1169 | if (step == 2 && (!IS_DNODE(page) || | |
1170 | !is_cold_node(page))) | |
1171 | continue; | |
1172 | ||
1173 | /* | |
1174 | * If an fsync mode, | |
1175 | * we should not skip writing node pages. | |
1176 | */ | |
1177 | if (ino && ino_of_node(page) == ino) | |
1178 | lock_page(page); | |
1179 | else if (!trylock_page(page)) | |
1180 | continue; | |
1181 | ||
4ef51a8f | 1182 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
e05df3b1 JK |
1183 | continue_unlock: |
1184 | unlock_page(page); | |
1185 | continue; | |
1186 | } | |
1187 | if (ino && ino_of_node(page) != ino) | |
1188 | goto continue_unlock; | |
1189 | ||
1190 | if (!PageDirty(page)) { | |
1191 | /* someone wrote it for us */ | |
1192 | goto continue_unlock; | |
1193 | } | |
1194 | ||
1195 | if (!clear_page_dirty_for_io(page)) | |
1196 | goto continue_unlock; | |
1197 | ||
1198 | /* called by fsync() */ | |
1199 | if (ino && IS_DNODE(page)) { | |
e05df3b1 | 1200 | set_fsync_mark(page, 1); |
88bd02c9 JK |
1201 | if (IS_INODE(page)) { |
1202 | if (!is_checkpointed_node(sbi, ino) && | |
1203 | !has_fsynced_inode(sbi, ino)) | |
1204 | set_dentry_mark(page, 1); | |
1205 | else | |
1206 | set_dentry_mark(page, 0); | |
1207 | } | |
e05df3b1 JK |
1208 | nwritten++; |
1209 | } else { | |
1210 | set_fsync_mark(page, 0); | |
1211 | set_dentry_mark(page, 0); | |
1212 | } | |
52746519 JK |
1213 | |
1214 | if (NODE_MAPPING(sbi)->a_ops->writepage(page, wbc)) | |
1215 | unlock_page(page); | |
1216 | else | |
1217 | wrote++; | |
e05df3b1 JK |
1218 | |
1219 | if (--wbc->nr_to_write == 0) | |
1220 | break; | |
1221 | } | |
1222 | pagevec_release(&pvec); | |
1223 | cond_resched(); | |
1224 | ||
1225 | if (wbc->nr_to_write == 0) { | |
1226 | step = 2; | |
1227 | break; | |
1228 | } | |
1229 | } | |
1230 | ||
1231 | if (step < 2) { | |
1232 | step++; | |
1233 | goto next_step; | |
1234 | } | |
1235 | ||
1236 | if (wrote) | |
458e6197 | 1237 | f2fs_submit_merged_bio(sbi, NODE, WRITE); |
e05df3b1 JK |
1238 | return nwritten; |
1239 | } | |
1240 | ||
cfe58f9d JK |
1241 | int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino) |
1242 | { | |
cfe58f9d JK |
1243 | pgoff_t index = 0, end = LONG_MAX; |
1244 | struct pagevec pvec; | |
cfe58f9d JK |
1245 | int ret2 = 0, ret = 0; |
1246 | ||
1247 | pagevec_init(&pvec, 0); | |
4ef51a8f JK |
1248 | |
1249 | while (index <= end) { | |
1250 | int i, nr_pages; | |
1251 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1252 | PAGECACHE_TAG_WRITEBACK, | |
1253 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1254 | if (nr_pages == 0) | |
1255 | break; | |
cfe58f9d JK |
1256 | |
1257 | for (i = 0; i < nr_pages; i++) { | |
1258 | struct page *page = pvec.pages[i]; | |
1259 | ||
1260 | /* until radix tree lookup accepts end_index */ | |
cfb271d4 | 1261 | if (unlikely(page->index > end)) |
cfe58f9d JK |
1262 | continue; |
1263 | ||
4bf08ff6 | 1264 | if (ino && ino_of_node(page) == ino) { |
3cb5ad15 | 1265 | f2fs_wait_on_page_writeback(page, NODE); |
4bf08ff6 CY |
1266 | if (TestClearPageError(page)) |
1267 | ret = -EIO; | |
1268 | } | |
cfe58f9d JK |
1269 | } |
1270 | pagevec_release(&pvec); | |
1271 | cond_resched(); | |
1272 | } | |
1273 | ||
4ef51a8f | 1274 | if (unlikely(test_and_clear_bit(AS_ENOSPC, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d | 1275 | ret2 = -ENOSPC; |
4ef51a8f | 1276 | if (unlikely(test_and_clear_bit(AS_EIO, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d JK |
1277 | ret2 = -EIO; |
1278 | if (!ret) | |
1279 | ret = ret2; | |
1280 | return ret; | |
1281 | } | |
1282 | ||
e05df3b1 JK |
1283 | static int f2fs_write_node_page(struct page *page, |
1284 | struct writeback_control *wbc) | |
1285 | { | |
4081363f | 1286 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1287 | nid_t nid; |
e05df3b1 JK |
1288 | block_t new_addr; |
1289 | struct node_info ni; | |
fb5566da JK |
1290 | struct f2fs_io_info fio = { |
1291 | .type = NODE, | |
6c311ec6 | 1292 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
fb5566da | 1293 | }; |
e05df3b1 | 1294 | |
ecda0de3 CY |
1295 | trace_f2fs_writepage(page, NODE); |
1296 | ||
cfb271d4 | 1297 | if (unlikely(sbi->por_doing)) |
87a9bd26 | 1298 | goto redirty_out; |
cf779cab JK |
1299 | if (unlikely(f2fs_cp_error(sbi))) |
1300 | goto redirty_out; | |
87a9bd26 | 1301 | |
3cb5ad15 | 1302 | f2fs_wait_on_page_writeback(page, NODE); |
e05df3b1 | 1303 | |
e05df3b1 JK |
1304 | /* get old block addr of this node page */ |
1305 | nid = nid_of_node(page); | |
9850cf4a | 1306 | f2fs_bug_on(sbi, page->index != nid); |
e05df3b1 JK |
1307 | |
1308 | get_node_info(sbi, nid, &ni); | |
1309 | ||
1310 | /* This page is already truncated */ | |
6bacf52f | 1311 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
39936837 JK |
1312 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
1313 | unlock_page(page); | |
1314 | return 0; | |
1315 | } | |
e05df3b1 | 1316 | |
2f97c326 JK |
1317 | if (wbc->for_reclaim) { |
1318 | if (!down_read_trylock(&sbi->node_write)) | |
1319 | goto redirty_out; | |
1320 | } else { | |
1321 | down_read(&sbi->node_write); | |
1322 | } | |
e05df3b1 | 1323 | set_page_writeback(page); |
fb5566da | 1324 | write_node_page(sbi, page, &fio, nid, ni.blk_addr, &new_addr); |
479f40c4 | 1325 | set_node_addr(sbi, &ni, new_addr, is_fsync_dnode(page)); |
e05df3b1 | 1326 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
b3582c68 | 1327 | up_read(&sbi->node_write); |
e05df3b1 | 1328 | unlock_page(page); |
27c6bd60 JK |
1329 | |
1330 | if (wbc->for_reclaim) | |
1331 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1332 | ||
e05df3b1 | 1333 | return 0; |
87a9bd26 JK |
1334 | |
1335 | redirty_out: | |
76f60268 | 1336 | redirty_page_for_writepage(wbc, page); |
87a9bd26 | 1337 | return AOP_WRITEPAGE_ACTIVATE; |
e05df3b1 JK |
1338 | } |
1339 | ||
1340 | static int f2fs_write_node_pages(struct address_space *mapping, | |
1341 | struct writeback_control *wbc) | |
1342 | { | |
4081363f | 1343 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 1344 | long diff; |
e05df3b1 | 1345 | |
e5748434 CY |
1346 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
1347 | ||
4660f9c0 JK |
1348 | /* balancing f2fs's metadata in background */ |
1349 | f2fs_balance_fs_bg(sbi); | |
e05df3b1 | 1350 | |
a7fdffbd | 1351 | /* collect a number of dirty node pages and write together */ |
87d6f890 | 1352 | if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE)) |
d3baf95d | 1353 | goto skip_write; |
a7fdffbd | 1354 | |
50c8cdb3 | 1355 | diff = nr_pages_to_write(sbi, NODE, wbc); |
fb5566da | 1356 | wbc->sync_mode = WB_SYNC_NONE; |
e05df3b1 | 1357 | sync_node_pages(sbi, 0, wbc); |
50c8cdb3 | 1358 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
e05df3b1 | 1359 | return 0; |
d3baf95d JK |
1360 | |
1361 | skip_write: | |
1362 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES); | |
1363 | return 0; | |
e05df3b1 JK |
1364 | } |
1365 | ||
1366 | static int f2fs_set_node_page_dirty(struct page *page) | |
1367 | { | |
26c6b887 JK |
1368 | trace_f2fs_set_page_dirty(page, NODE); |
1369 | ||
e05df3b1 JK |
1370 | SetPageUptodate(page); |
1371 | if (!PageDirty(page)) { | |
1372 | __set_page_dirty_nobuffers(page); | |
4081363f | 1373 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 JK |
1374 | SetPagePrivate(page); |
1375 | return 1; | |
1376 | } | |
1377 | return 0; | |
1378 | } | |
1379 | ||
d47992f8 LC |
1380 | static void f2fs_invalidate_node_page(struct page *page, unsigned int offset, |
1381 | unsigned int length) | |
e05df3b1 JK |
1382 | { |
1383 | struct inode *inode = page->mapping->host; | |
e05df3b1 | 1384 | if (PageDirty(page)) |
4081363f | 1385 | dec_page_count(F2FS_I_SB(inode), F2FS_DIRTY_NODES); |
e05df3b1 JK |
1386 | ClearPagePrivate(page); |
1387 | } | |
1388 | ||
1389 | static int f2fs_release_node_page(struct page *page, gfp_t wait) | |
1390 | { | |
1391 | ClearPagePrivate(page); | |
c3850aa1 | 1392 | return 1; |
e05df3b1 JK |
1393 | } |
1394 | ||
0a8165d7 | 1395 | /* |
e05df3b1 JK |
1396 | * Structure of the f2fs node operations |
1397 | */ | |
1398 | const struct address_space_operations f2fs_node_aops = { | |
1399 | .writepage = f2fs_write_node_page, | |
1400 | .writepages = f2fs_write_node_pages, | |
1401 | .set_page_dirty = f2fs_set_node_page_dirty, | |
1402 | .invalidatepage = f2fs_invalidate_node_page, | |
1403 | .releasepage = f2fs_release_node_page, | |
1404 | }; | |
1405 | ||
8a7ed66a JK |
1406 | static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i, |
1407 | nid_t n) | |
e05df3b1 | 1408 | { |
8a7ed66a | 1409 | return radix_tree_lookup(&nm_i->free_nid_root, n); |
e05df3b1 JK |
1410 | } |
1411 | ||
8a7ed66a JK |
1412 | static void __del_from_free_nid_list(struct f2fs_nm_info *nm_i, |
1413 | struct free_nid *i) | |
e05df3b1 JK |
1414 | { |
1415 | list_del(&i->list); | |
8a7ed66a | 1416 | radix_tree_delete(&nm_i->free_nid_root, i->nid); |
e05df3b1 JK |
1417 | } |
1418 | ||
6fb03f3a | 1419 | static int add_free_nid(struct f2fs_sb_info *sbi, nid_t nid, bool build) |
e05df3b1 | 1420 | { |
6fb03f3a | 1421 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1422 | struct free_nid *i; |
59bbd474 JK |
1423 | struct nat_entry *ne; |
1424 | bool allocated = false; | |
e05df3b1 | 1425 | |
6fb03f3a | 1426 | if (!available_free_memory(sbi, FREE_NIDS)) |
23d38844 | 1427 | return -1; |
9198aceb JK |
1428 | |
1429 | /* 0 nid should not be used */ | |
cfb271d4 | 1430 | if (unlikely(nid == 0)) |
9198aceb | 1431 | return 0; |
59bbd474 | 1432 | |
7bd59381 GZ |
1433 | if (build) { |
1434 | /* do not add allocated nids */ | |
1435 | read_lock(&nm_i->nat_tree_lock); | |
1436 | ne = __lookup_nat_cache(nm_i, nid); | |
8a7ed66a | 1437 | if (ne && |
7ef35e3b JK |
1438 | (!get_nat_flag(ne, IS_CHECKPOINTED) || |
1439 | nat_get_blkaddr(ne) != NULL_ADDR)) | |
7bd59381 GZ |
1440 | allocated = true; |
1441 | read_unlock(&nm_i->nat_tree_lock); | |
1442 | if (allocated) | |
1443 | return 0; | |
e05df3b1 | 1444 | } |
7bd59381 GZ |
1445 | |
1446 | i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); | |
e05df3b1 JK |
1447 | i->nid = nid; |
1448 | i->state = NID_NEW; | |
1449 | ||
1450 | spin_lock(&nm_i->free_nid_list_lock); | |
8a7ed66a | 1451 | if (radix_tree_insert(&nm_i->free_nid_root, i->nid, i)) { |
e05df3b1 JK |
1452 | spin_unlock(&nm_i->free_nid_list_lock); |
1453 | kmem_cache_free(free_nid_slab, i); | |
1454 | return 0; | |
1455 | } | |
1456 | list_add_tail(&i->list, &nm_i->free_nid_list); | |
1457 | nm_i->fcnt++; | |
1458 | spin_unlock(&nm_i->free_nid_list_lock); | |
1459 | return 1; | |
1460 | } | |
1461 | ||
1462 | static void remove_free_nid(struct f2fs_nm_info *nm_i, nid_t nid) | |
1463 | { | |
1464 | struct free_nid *i; | |
cf0ee0f0 CY |
1465 | bool need_free = false; |
1466 | ||
e05df3b1 | 1467 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1468 | i = __lookup_free_nid_list(nm_i, nid); |
e05df3b1 | 1469 | if (i && i->state == NID_NEW) { |
8a7ed66a | 1470 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1471 | nm_i->fcnt--; |
cf0ee0f0 | 1472 | need_free = true; |
e05df3b1 JK |
1473 | } |
1474 | spin_unlock(&nm_i->free_nid_list_lock); | |
cf0ee0f0 CY |
1475 | |
1476 | if (need_free) | |
1477 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1478 | } |
1479 | ||
6fb03f3a | 1480 | static void scan_nat_page(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
1481 | struct page *nat_page, nid_t start_nid) |
1482 | { | |
6fb03f3a | 1483 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 JK |
1484 | struct f2fs_nat_block *nat_blk = page_address(nat_page); |
1485 | block_t blk_addr; | |
e05df3b1 JK |
1486 | int i; |
1487 | ||
e05df3b1 JK |
1488 | i = start_nid % NAT_ENTRY_PER_BLOCK; |
1489 | ||
1490 | for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { | |
23d38844 | 1491 | |
cfb271d4 | 1492 | if (unlikely(start_nid >= nm_i->max_nid)) |
04431c44 | 1493 | break; |
23d38844 HL |
1494 | |
1495 | blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); | |
9850cf4a | 1496 | f2fs_bug_on(sbi, blk_addr == NEW_ADDR); |
23d38844 | 1497 | if (blk_addr == NULL_ADDR) { |
6fb03f3a | 1498 | if (add_free_nid(sbi, start_nid, true) < 0) |
23d38844 HL |
1499 | break; |
1500 | } | |
e05df3b1 | 1501 | } |
e05df3b1 JK |
1502 | } |
1503 | ||
1504 | static void build_free_nids(struct f2fs_sb_info *sbi) | |
1505 | { | |
e05df3b1 JK |
1506 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1507 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1508 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
8760952d | 1509 | int i = 0; |
55008d84 | 1510 | nid_t nid = nm_i->next_scan_nid; |
e05df3b1 | 1511 | |
55008d84 JK |
1512 | /* Enough entries */ |
1513 | if (nm_i->fcnt > NAT_ENTRY_PER_BLOCK) | |
1514 | return; | |
e05df3b1 | 1515 | |
55008d84 | 1516 | /* readahead nat pages to be scanned */ |
662befda | 1517 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES, META_NAT); |
e05df3b1 JK |
1518 | |
1519 | while (1) { | |
1520 | struct page *page = get_current_nat_page(sbi, nid); | |
1521 | ||
6fb03f3a | 1522 | scan_nat_page(sbi, page, nid); |
e05df3b1 JK |
1523 | f2fs_put_page(page, 1); |
1524 | ||
1525 | nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); | |
cfb271d4 | 1526 | if (unlikely(nid >= nm_i->max_nid)) |
e05df3b1 | 1527 | nid = 0; |
55008d84 JK |
1528 | |
1529 | if (i++ == FREE_NID_PAGES) | |
e05df3b1 JK |
1530 | break; |
1531 | } | |
1532 | ||
55008d84 JK |
1533 | /* go to the next free nat pages to find free nids abundantly */ |
1534 | nm_i->next_scan_nid = nid; | |
e05df3b1 JK |
1535 | |
1536 | /* find free nids from current sum_pages */ | |
1537 | mutex_lock(&curseg->curseg_mutex); | |
1538 | for (i = 0; i < nats_in_cursum(sum); i++) { | |
1539 | block_t addr = le32_to_cpu(nat_in_journal(sum, i).block_addr); | |
1540 | nid = le32_to_cpu(nid_in_journal(sum, i)); | |
1541 | if (addr == NULL_ADDR) | |
6fb03f3a | 1542 | add_free_nid(sbi, nid, true); |
e05df3b1 JK |
1543 | else |
1544 | remove_free_nid(nm_i, nid); | |
1545 | } | |
1546 | mutex_unlock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1547 | } |
1548 | ||
1549 | /* | |
1550 | * If this function returns success, caller can obtain a new nid | |
1551 | * from second parameter of this function. | |
1552 | * The returned nid could be used ino as well as nid when inode is created. | |
1553 | */ | |
1554 | bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) | |
1555 | { | |
1556 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1557 | struct free_nid *i = NULL; | |
e05df3b1 | 1558 | retry: |
7ee0eeab | 1559 | if (unlikely(sbi->total_valid_node_count + 1 > nm_i->available_nids)) |
55008d84 | 1560 | return false; |
e05df3b1 | 1561 | |
e05df3b1 | 1562 | spin_lock(&nm_i->free_nid_list_lock); |
e05df3b1 | 1563 | |
55008d84 | 1564 | /* We should not use stale free nids created by build_free_nids */ |
f978f5a0 | 1565 | if (nm_i->fcnt && !on_build_free_nids(nm_i)) { |
9850cf4a | 1566 | f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list)); |
2d7b822a | 1567 | list_for_each_entry(i, &nm_i->free_nid_list, list) |
55008d84 JK |
1568 | if (i->state == NID_NEW) |
1569 | break; | |
e05df3b1 | 1570 | |
9850cf4a | 1571 | f2fs_bug_on(sbi, i->state != NID_NEW); |
55008d84 JK |
1572 | *nid = i->nid; |
1573 | i->state = NID_ALLOC; | |
1574 | nm_i->fcnt--; | |
1575 | spin_unlock(&nm_i->free_nid_list_lock); | |
1576 | return true; | |
1577 | } | |
e05df3b1 | 1578 | spin_unlock(&nm_i->free_nid_list_lock); |
55008d84 JK |
1579 | |
1580 | /* Let's scan nat pages and its caches to get free nids */ | |
1581 | mutex_lock(&nm_i->build_lock); | |
55008d84 | 1582 | build_free_nids(sbi); |
55008d84 JK |
1583 | mutex_unlock(&nm_i->build_lock); |
1584 | goto retry; | |
e05df3b1 JK |
1585 | } |
1586 | ||
0a8165d7 | 1587 | /* |
e05df3b1 JK |
1588 | * alloc_nid() should be called prior to this function. |
1589 | */ | |
1590 | void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) | |
1591 | { | |
1592 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1593 | struct free_nid *i; | |
1594 | ||
1595 | spin_lock(&nm_i->free_nid_list_lock); | |
8a7ed66a | 1596 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1597 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
8a7ed66a | 1598 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1599 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1600 | |
1601 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1602 | } |
1603 | ||
0a8165d7 | 1604 | /* |
e05df3b1 JK |
1605 | * alloc_nid() should be called prior to this function. |
1606 | */ | |
1607 | void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) | |
1608 | { | |
49952fa1 JK |
1609 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1610 | struct free_nid *i; | |
cf0ee0f0 | 1611 | bool need_free = false; |
49952fa1 | 1612 | |
65985d93 JK |
1613 | if (!nid) |
1614 | return; | |
1615 | ||
49952fa1 | 1616 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1617 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1618 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
6fb03f3a | 1619 | if (!available_free_memory(sbi, FREE_NIDS)) { |
8a7ed66a | 1620 | __del_from_free_nid_list(nm_i, i); |
cf0ee0f0 | 1621 | need_free = true; |
95630cba HL |
1622 | } else { |
1623 | i->state = NID_NEW; | |
1624 | nm_i->fcnt++; | |
1625 | } | |
49952fa1 | 1626 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1627 | |
1628 | if (need_free) | |
1629 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1630 | } |
1631 | ||
70cfed88 | 1632 | void recover_inline_xattr(struct inode *inode, struct page *page) |
28cdce04 | 1633 | { |
28cdce04 CY |
1634 | void *src_addr, *dst_addr; |
1635 | size_t inline_size; | |
1636 | struct page *ipage; | |
1637 | struct f2fs_inode *ri; | |
1638 | ||
4081363f | 1639 | ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino); |
9850cf4a | 1640 | f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage)); |
28cdce04 | 1641 | |
e3b4d43f JK |
1642 | ri = F2FS_INODE(page); |
1643 | if (!(ri->i_inline & F2FS_INLINE_XATTR)) { | |
1644 | clear_inode_flag(F2FS_I(inode), FI_INLINE_XATTR); | |
1645 | goto update_inode; | |
1646 | } | |
1647 | ||
28cdce04 CY |
1648 | dst_addr = inline_xattr_addr(ipage); |
1649 | src_addr = inline_xattr_addr(page); | |
1650 | inline_size = inline_xattr_size(inode); | |
1651 | ||
54b591df | 1652 | f2fs_wait_on_page_writeback(ipage, NODE); |
28cdce04 | 1653 | memcpy(dst_addr, src_addr, inline_size); |
e3b4d43f | 1654 | update_inode: |
28cdce04 CY |
1655 | update_inode(inode, ipage); |
1656 | f2fs_put_page(ipage, 1); | |
1657 | } | |
1658 | ||
1c35a90e | 1659 | void recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr) |
abb2366c | 1660 | { |
4081363f | 1661 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
abb2366c JK |
1662 | nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid; |
1663 | nid_t new_xnid = nid_of_node(page); | |
1664 | struct node_info ni; | |
1665 | ||
abb2366c JK |
1666 | /* 1: invalidate the previous xattr nid */ |
1667 | if (!prev_xnid) | |
1668 | goto recover_xnid; | |
1669 | ||
1670 | /* Deallocate node address */ | |
1671 | get_node_info(sbi, prev_xnid, &ni); | |
9850cf4a | 1672 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
abb2366c JK |
1673 | invalidate_blocks(sbi, ni.blk_addr); |
1674 | dec_valid_node_count(sbi, inode); | |
479f40c4 | 1675 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
abb2366c JK |
1676 | |
1677 | recover_xnid: | |
1678 | /* 2: allocate new xattr nid */ | |
1679 | if (unlikely(!inc_valid_node_count(sbi, inode))) | |
9850cf4a | 1680 | f2fs_bug_on(sbi, 1); |
abb2366c JK |
1681 | |
1682 | remove_free_nid(NM_I(sbi), new_xnid); | |
1683 | get_node_info(sbi, new_xnid, &ni); | |
1684 | ni.ino = inode->i_ino; | |
479f40c4 | 1685 | set_node_addr(sbi, &ni, NEW_ADDR, false); |
abb2366c JK |
1686 | F2FS_I(inode)->i_xattr_nid = new_xnid; |
1687 | ||
1688 | /* 3: update xattr blkaddr */ | |
1689 | refresh_sit_entry(sbi, NEW_ADDR, blkaddr); | |
479f40c4 | 1690 | set_node_addr(sbi, &ni, blkaddr, false); |
abb2366c JK |
1691 | |
1692 | update_inode_page(inode); | |
abb2366c JK |
1693 | } |
1694 | ||
e05df3b1 JK |
1695 | int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) |
1696 | { | |
58bfaf44 | 1697 | struct f2fs_inode *src, *dst; |
e05df3b1 JK |
1698 | nid_t ino = ino_of_node(page); |
1699 | struct node_info old_ni, new_ni; | |
1700 | struct page *ipage; | |
1701 | ||
e8271fa3 JK |
1702 | get_node_info(sbi, ino, &old_ni); |
1703 | ||
1704 | if (unlikely(old_ni.blk_addr != NULL_ADDR)) | |
1705 | return -EINVAL; | |
1706 | ||
4ef51a8f | 1707 | ipage = grab_cache_page(NODE_MAPPING(sbi), ino); |
e05df3b1 JK |
1708 | if (!ipage) |
1709 | return -ENOMEM; | |
1710 | ||
e1c42045 | 1711 | /* Should not use this inode from free nid list */ |
e05df3b1 JK |
1712 | remove_free_nid(NM_I(sbi), ino); |
1713 | ||
e05df3b1 JK |
1714 | SetPageUptodate(ipage); |
1715 | fill_node_footer(ipage, ino, ino, 0, true); | |
1716 | ||
58bfaf44 JK |
1717 | src = F2FS_INODE(page); |
1718 | dst = F2FS_INODE(ipage); | |
e05df3b1 | 1719 | |
58bfaf44 JK |
1720 | memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); |
1721 | dst->i_size = 0; | |
1722 | dst->i_blocks = cpu_to_le64(1); | |
1723 | dst->i_links = cpu_to_le32(1); | |
1724 | dst->i_xattr_nid = 0; | |
617deb8c | 1725 | dst->i_inline = src->i_inline & F2FS_INLINE_XATTR; |
e05df3b1 JK |
1726 | |
1727 | new_ni = old_ni; | |
1728 | new_ni.ino = ino; | |
1729 | ||
cfb271d4 | 1730 | if (unlikely(!inc_valid_node_count(sbi, NULL))) |
65e5cd0a | 1731 | WARN_ON(1); |
479f40c4 | 1732 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
e05df3b1 | 1733 | inc_valid_inode_count(sbi); |
617deb8c | 1734 | set_page_dirty(ipage); |
e05df3b1 JK |
1735 | f2fs_put_page(ipage, 1); |
1736 | return 0; | |
1737 | } | |
1738 | ||
9af0ff1c CY |
1739 | /* |
1740 | * ra_sum_pages() merge contiguous pages into one bio and submit. | |
e1c42045 | 1741 | * these pre-read pages are allocated in bd_inode's mapping tree. |
9af0ff1c | 1742 | */ |
bac4eef6 | 1743 | static int ra_sum_pages(struct f2fs_sb_info *sbi, struct page **pages, |
9af0ff1c CY |
1744 | int start, int nrpages) |
1745 | { | |
bac4eef6 CY |
1746 | struct inode *inode = sbi->sb->s_bdev->bd_inode; |
1747 | struct address_space *mapping = inode->i_mapping; | |
1748 | int i, page_idx = start; | |
458e6197 JK |
1749 | struct f2fs_io_info fio = { |
1750 | .type = META, | |
7e8f2308 | 1751 | .rw = READ_SYNC | REQ_META | REQ_PRIO |
458e6197 | 1752 | }; |
9af0ff1c | 1753 | |
bac4eef6 CY |
1754 | for (i = 0; page_idx < start + nrpages; page_idx++, i++) { |
1755 | /* alloc page in bd_inode for reading node summary info */ | |
1756 | pages[i] = grab_cache_page(mapping, page_idx); | |
1757 | if (!pages[i]) | |
d653788a | 1758 | break; |
bac4eef6 | 1759 | f2fs_submit_page_mbio(sbi, pages[i], page_idx, &fio); |
9af0ff1c CY |
1760 | } |
1761 | ||
458e6197 | 1762 | f2fs_submit_merged_bio(sbi, META, READ); |
bac4eef6 | 1763 | return i; |
9af0ff1c CY |
1764 | } |
1765 | ||
e05df3b1 JK |
1766 | int restore_node_summary(struct f2fs_sb_info *sbi, |
1767 | unsigned int segno, struct f2fs_summary_block *sum) | |
1768 | { | |
1769 | struct f2fs_node *rn; | |
1770 | struct f2fs_summary *sum_entry; | |
bac4eef6 | 1771 | struct inode *inode = sbi->sb->s_bdev->bd_inode; |
e05df3b1 | 1772 | block_t addr; |
90a893c7 | 1773 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
bac4eef6 CY |
1774 | struct page *pages[bio_blocks]; |
1775 | int i, idx, last_offset, nrpages, err = 0; | |
e05df3b1 JK |
1776 | |
1777 | /* scan the node segment */ | |
1778 | last_offset = sbi->blocks_per_seg; | |
1779 | addr = START_BLOCK(sbi, segno); | |
1780 | sum_entry = &sum->entries[0]; | |
1781 | ||
d653788a | 1782 | for (i = 0; !err && i < last_offset; i += nrpages, addr += nrpages) { |
9af0ff1c | 1783 | nrpages = min(last_offset - i, bio_blocks); |
393ff91f | 1784 | |
e1c42045 | 1785 | /* readahead node pages */ |
bac4eef6 | 1786 | nrpages = ra_sum_pages(sbi, pages, addr, nrpages); |
d653788a GZ |
1787 | if (!nrpages) |
1788 | return -ENOMEM; | |
e05df3b1 | 1789 | |
bac4eef6 | 1790 | for (idx = 0; idx < nrpages; idx++) { |
d653788a GZ |
1791 | if (err) |
1792 | goto skip; | |
9af0ff1c | 1793 | |
bac4eef6 CY |
1794 | lock_page(pages[idx]); |
1795 | if (unlikely(!PageUptodate(pages[idx]))) { | |
6bacf52f JK |
1796 | err = -EIO; |
1797 | } else { | |
bac4eef6 | 1798 | rn = F2FS_NODE(pages[idx]); |
9af0ff1c CY |
1799 | sum_entry->nid = rn->footer.nid; |
1800 | sum_entry->version = 0; | |
1801 | sum_entry->ofs_in_node = 0; | |
1802 | sum_entry++; | |
9af0ff1c | 1803 | } |
bac4eef6 | 1804 | unlock_page(pages[idx]); |
d653788a | 1805 | skip: |
bac4eef6 | 1806 | page_cache_release(pages[idx]); |
9af0ff1c | 1807 | } |
bac4eef6 CY |
1808 | |
1809 | invalidate_mapping_pages(inode->i_mapping, addr, | |
1810 | addr + nrpages); | |
e05df3b1 | 1811 | } |
9af0ff1c | 1812 | return err; |
e05df3b1 JK |
1813 | } |
1814 | ||
aec71382 | 1815 | static void remove_nats_in_journal(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
1816 | { |
1817 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1818 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1819 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1820 | int i; | |
1821 | ||
1822 | mutex_lock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1823 | for (i = 0; i < nats_in_cursum(sum); i++) { |
1824 | struct nat_entry *ne; | |
1825 | struct f2fs_nat_entry raw_ne; | |
1826 | nid_t nid = le32_to_cpu(nid_in_journal(sum, i)); | |
1827 | ||
1828 | raw_ne = nat_in_journal(sum, i); | |
1829 | retry: | |
1830 | write_lock(&nm_i->nat_tree_lock); | |
1831 | ne = __lookup_nat_cache(nm_i, nid); | |
aec71382 CY |
1832 | if (ne) |
1833 | goto found; | |
1834 | ||
e05df3b1 JK |
1835 | ne = grab_nat_entry(nm_i, nid); |
1836 | if (!ne) { | |
1837 | write_unlock(&nm_i->nat_tree_lock); | |
1838 | goto retry; | |
1839 | } | |
94dac22e | 1840 | node_info_from_raw_nat(&ne->ni, &raw_ne); |
aec71382 | 1841 | found: |
e05df3b1 JK |
1842 | __set_nat_cache_dirty(nm_i, ne); |
1843 | write_unlock(&nm_i->nat_tree_lock); | |
1844 | } | |
1845 | update_nats_in_cursum(sum, -i); | |
1846 | mutex_unlock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1847 | } |
1848 | ||
309cc2b6 JK |
1849 | static void __adjust_nat_entry_set(struct nat_entry_set *nes, |
1850 | struct list_head *head, int max) | |
e05df3b1 | 1851 | { |
309cc2b6 | 1852 | struct nat_entry_set *cur; |
e05df3b1 | 1853 | |
309cc2b6 JK |
1854 | if (nes->entry_cnt >= max) |
1855 | goto add_out; | |
e05df3b1 | 1856 | |
309cc2b6 JK |
1857 | list_for_each_entry(cur, head, set_list) { |
1858 | if (cur->entry_cnt >= nes->entry_cnt) { | |
1859 | list_add(&nes->set_list, cur->set_list.prev); | |
1860 | return; | |
1861 | } | |
aec71382 | 1862 | } |
309cc2b6 JK |
1863 | add_out: |
1864 | list_add_tail(&nes->set_list, head); | |
1865 | } | |
e05df3b1 | 1866 | |
309cc2b6 JK |
1867 | static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, |
1868 | struct nat_entry_set *set) | |
1869 | { | |
1870 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1871 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1872 | nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK; | |
1873 | bool to_journal = true; | |
1874 | struct f2fs_nat_block *nat_blk; | |
1875 | struct nat_entry *ne, *cur; | |
1876 | struct page *page = NULL; | |
e05df3b1 | 1877 | |
aec71382 CY |
1878 | /* |
1879 | * there are two steps to flush nat entries: | |
1880 | * #1, flush nat entries to journal in current hot data summary block. | |
1881 | * #2, flush nat entries to nat page. | |
1882 | */ | |
309cc2b6 JK |
1883 | if (!__has_cursum_space(sum, set->entry_cnt, NAT_JOURNAL)) |
1884 | to_journal = false; | |
1885 | ||
1886 | if (to_journal) { | |
1887 | mutex_lock(&curseg->curseg_mutex); | |
1888 | } else { | |
1889 | page = get_next_nat_page(sbi, start_nid); | |
1890 | nat_blk = page_address(page); | |
1891 | f2fs_bug_on(sbi, !nat_blk); | |
1892 | } | |
aec71382 | 1893 | |
309cc2b6 JK |
1894 | /* flush dirty nats in nat entry set */ |
1895 | list_for_each_entry_safe(ne, cur, &set->entry_list, list) { | |
1896 | struct f2fs_nat_entry *raw_ne; | |
1897 | nid_t nid = nat_get_nid(ne); | |
1898 | int offset; | |
1899 | ||
1900 | if (nat_get_blkaddr(ne) == NEW_ADDR) | |
1901 | continue; | |
aec71382 CY |
1902 | |
1903 | if (to_journal) { | |
309cc2b6 JK |
1904 | offset = lookup_journal_in_cursum(sum, |
1905 | NAT_JOURNAL, nid, 1); | |
1906 | f2fs_bug_on(sbi, offset < 0); | |
1907 | raw_ne = &nat_in_journal(sum, offset); | |
1908 | nid_in_journal(sum, offset) = cpu_to_le32(nid); | |
aec71382 | 1909 | } else { |
309cc2b6 | 1910 | raw_ne = &nat_blk->entries[nid - start_nid]; |
e05df3b1 | 1911 | } |
309cc2b6 | 1912 | raw_nat_from_node_info(raw_ne, &ne->ni); |
e05df3b1 | 1913 | |
309cc2b6 JK |
1914 | write_lock(&NM_I(sbi)->nat_tree_lock); |
1915 | nat_reset_flag(ne); | |
1916 | __clear_nat_cache_dirty(NM_I(sbi), ne); | |
1917 | write_unlock(&NM_I(sbi)->nat_tree_lock); | |
aec71382 | 1918 | |
309cc2b6 JK |
1919 | if (nat_get_blkaddr(ne) == NULL_ADDR) |
1920 | add_free_nid(sbi, nid, false); | |
1921 | } | |
e05df3b1 | 1922 | |
309cc2b6 JK |
1923 | if (to_journal) |
1924 | mutex_unlock(&curseg->curseg_mutex); | |
1925 | else | |
1926 | f2fs_put_page(page, 1); | |
aec71382 | 1927 | |
80ec2e91 CL |
1928 | f2fs_bug_on(sbi, set->entry_cnt); |
1929 | ||
1930 | radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); | |
1931 | kmem_cache_free(nat_entry_set_slab, set); | |
309cc2b6 | 1932 | } |
aec71382 | 1933 | |
309cc2b6 JK |
1934 | /* |
1935 | * This function is called during the checkpointing process. | |
1936 | */ | |
1937 | void flush_nat_entries(struct f2fs_sb_info *sbi) | |
1938 | { | |
1939 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1940 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1941 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1942 | struct nat_entry_set *setvec[NATVEC_SIZE]; | |
1943 | struct nat_entry_set *set, *tmp; | |
1944 | unsigned int found; | |
1945 | nid_t set_idx = 0; | |
1946 | LIST_HEAD(sets); | |
1947 | ||
20d047c8 CL |
1948 | if (!nm_i->dirty_nat_cnt) |
1949 | return; | |
309cc2b6 JK |
1950 | /* |
1951 | * if there are no enough space in journal to store dirty nat | |
1952 | * entries, remove all entries from journal and merge them | |
1953 | * into nat entry set. | |
1954 | */ | |
1955 | if (!__has_cursum_space(sum, nm_i->dirty_nat_cnt, NAT_JOURNAL)) | |
1956 | remove_nats_in_journal(sbi); | |
1957 | ||
309cc2b6 JK |
1958 | while ((found = __gang_lookup_nat_set(nm_i, |
1959 | set_idx, NATVEC_SIZE, setvec))) { | |
1960 | unsigned idx; | |
1961 | set_idx = setvec[found - 1]->set + 1; | |
1962 | for (idx = 0; idx < found; idx++) | |
1963 | __adjust_nat_entry_set(setvec[idx], &sets, | |
1964 | MAX_NAT_JENTRIES(sum)); | |
e05df3b1 | 1965 | } |
aec71382 | 1966 | |
309cc2b6 JK |
1967 | /* flush dirty nats in nat entry set */ |
1968 | list_for_each_entry_safe(set, tmp, &sets, set_list) | |
1969 | __flush_nat_entry_set(sbi, set); | |
1970 | ||
9850cf4a | 1971 | f2fs_bug_on(sbi, nm_i->dirty_nat_cnt); |
e05df3b1 JK |
1972 | } |
1973 | ||
1974 | static int init_node_manager(struct f2fs_sb_info *sbi) | |
1975 | { | |
1976 | struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); | |
1977 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1978 | unsigned char *version_bitmap; | |
1979 | unsigned int nat_segs, nat_blocks; | |
1980 | ||
1981 | nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); | |
1982 | ||
1983 | /* segment_count_nat includes pair segment so divide to 2. */ | |
1984 | nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; | |
1985 | nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); | |
b63da15e | 1986 | |
7ee0eeab JK |
1987 | nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks; |
1988 | ||
b63da15e | 1989 | /* not used nids: 0, node, meta, (and root counted as valid node) */ |
c200b1aa | 1990 | nm_i->available_nids = nm_i->max_nid - F2FS_RESERVED_NODE_NUM; |
e05df3b1 JK |
1991 | nm_i->fcnt = 0; |
1992 | nm_i->nat_cnt = 0; | |
cdfc41c1 | 1993 | nm_i->ram_thresh = DEF_RAM_THRESHOLD; |
e05df3b1 | 1994 | |
8a7ed66a | 1995 | INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC); |
e05df3b1 JK |
1996 | INIT_LIST_HEAD(&nm_i->free_nid_list); |
1997 | INIT_RADIX_TREE(&nm_i->nat_root, GFP_ATOMIC); | |
309cc2b6 | 1998 | INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_ATOMIC); |
e05df3b1 | 1999 | INIT_LIST_HEAD(&nm_i->nat_entries); |
e05df3b1 JK |
2000 | |
2001 | mutex_init(&nm_i->build_lock); | |
2002 | spin_lock_init(&nm_i->free_nid_list_lock); | |
2003 | rwlock_init(&nm_i->nat_tree_lock); | |
2004 | ||
e05df3b1 | 2005 | nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); |
79b5793b | 2006 | nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); |
e05df3b1 JK |
2007 | version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); |
2008 | if (!version_bitmap) | |
2009 | return -EFAULT; | |
2010 | ||
79b5793b AG |
2011 | nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, |
2012 | GFP_KERNEL); | |
2013 | if (!nm_i->nat_bitmap) | |
2014 | return -ENOMEM; | |
e05df3b1 JK |
2015 | return 0; |
2016 | } | |
2017 | ||
2018 | int build_node_manager(struct f2fs_sb_info *sbi) | |
2019 | { | |
2020 | int err; | |
2021 | ||
2022 | sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL); | |
2023 | if (!sbi->nm_info) | |
2024 | return -ENOMEM; | |
2025 | ||
2026 | err = init_node_manager(sbi); | |
2027 | if (err) | |
2028 | return err; | |
2029 | ||
2030 | build_free_nids(sbi); | |
2031 | return 0; | |
2032 | } | |
2033 | ||
2034 | void destroy_node_manager(struct f2fs_sb_info *sbi) | |
2035 | { | |
2036 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2037 | struct free_nid *i, *next_i; | |
2038 | struct nat_entry *natvec[NATVEC_SIZE]; | |
2039 | nid_t nid = 0; | |
2040 | unsigned int found; | |
2041 | ||
2042 | if (!nm_i) | |
2043 | return; | |
2044 | ||
2045 | /* destroy free nid list */ | |
2046 | spin_lock(&nm_i->free_nid_list_lock); | |
2047 | list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) { | |
9850cf4a | 2048 | f2fs_bug_on(sbi, i->state == NID_ALLOC); |
8a7ed66a | 2049 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 2050 | nm_i->fcnt--; |
cf0ee0f0 CY |
2051 | spin_unlock(&nm_i->free_nid_list_lock); |
2052 | kmem_cache_free(free_nid_slab, i); | |
2053 | spin_lock(&nm_i->free_nid_list_lock); | |
e05df3b1 | 2054 | } |
9850cf4a | 2055 | f2fs_bug_on(sbi, nm_i->fcnt); |
e05df3b1 JK |
2056 | spin_unlock(&nm_i->free_nid_list_lock); |
2057 | ||
2058 | /* destroy nat cache */ | |
2059 | write_lock(&nm_i->nat_tree_lock); | |
2060 | while ((found = __gang_lookup_nat_cache(nm_i, | |
2061 | nid, NATVEC_SIZE, natvec))) { | |
2062 | unsigned idx; | |
b6ce391e GZ |
2063 | nid = nat_get_nid(natvec[found - 1]) + 1; |
2064 | for (idx = 0; idx < found; idx++) | |
2065 | __del_from_nat_cache(nm_i, natvec[idx]); | |
e05df3b1 | 2066 | } |
9850cf4a | 2067 | f2fs_bug_on(sbi, nm_i->nat_cnt); |
e05df3b1 JK |
2068 | write_unlock(&nm_i->nat_tree_lock); |
2069 | ||
2070 | kfree(nm_i->nat_bitmap); | |
2071 | sbi->nm_info = NULL; | |
2072 | kfree(nm_i); | |
2073 | } | |
2074 | ||
6e6093a8 | 2075 | int __init create_node_manager_caches(void) |
e05df3b1 JK |
2076 | { |
2077 | nat_entry_slab = f2fs_kmem_cache_create("nat_entry", | |
e8512d2e | 2078 | sizeof(struct nat_entry)); |
e05df3b1 | 2079 | if (!nat_entry_slab) |
aec71382 | 2080 | goto fail; |
e05df3b1 JK |
2081 | |
2082 | free_nid_slab = f2fs_kmem_cache_create("free_nid", | |
e8512d2e | 2083 | sizeof(struct free_nid)); |
aec71382 | 2084 | if (!free_nid_slab) |
ce3e6d25 | 2085 | goto destroy_nat_entry; |
aec71382 CY |
2086 | |
2087 | nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set", | |
2088 | sizeof(struct nat_entry_set)); | |
2089 | if (!nat_entry_set_slab) | |
ce3e6d25 | 2090 | goto destroy_free_nid; |
e05df3b1 | 2091 | return 0; |
aec71382 | 2092 | |
ce3e6d25 | 2093 | destroy_free_nid: |
aec71382 | 2094 | kmem_cache_destroy(free_nid_slab); |
ce3e6d25 | 2095 | destroy_nat_entry: |
aec71382 CY |
2096 | kmem_cache_destroy(nat_entry_slab); |
2097 | fail: | |
2098 | return -ENOMEM; | |
e05df3b1 JK |
2099 | } |
2100 | ||
2101 | void destroy_node_manager_caches(void) | |
2102 | { | |
aec71382 | 2103 | kmem_cache_destroy(nat_entry_set_slab); |
e05df3b1 JK |
2104 | kmem_cache_destroy(free_nid_slab); |
2105 | kmem_cache_destroy(nat_entry_slab); | |
2106 | } |