]>
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 | ||
e5e7ea3c | 60 | for (i = 0; i <= UPDATE_INO; i++) |
67298804 CY |
61 | mem_size += (sbi->im[i].ino_num * |
62 | sizeof(struct ino_entry)) >> PAGE_CACHE_SHIFT; | |
e5e7ea3c | 63 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
1e84371f JK |
64 | } else { |
65 | if (sbi->sb->s_bdi->dirty_exceeded) | |
66 | return false; | |
6fb03f3a JK |
67 | } |
68 | return res; | |
cdfc41c1 JK |
69 | } |
70 | ||
e05df3b1 JK |
71 | static void clear_node_page_dirty(struct page *page) |
72 | { | |
73 | struct address_space *mapping = page->mapping; | |
e05df3b1 JK |
74 | unsigned int long flags; |
75 | ||
76 | if (PageDirty(page)) { | |
77 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
78 | radix_tree_tag_clear(&mapping->page_tree, | |
79 | page_index(page), | |
80 | PAGECACHE_TAG_DIRTY); | |
81 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
82 | ||
83 | clear_page_dirty_for_io(page); | |
4081363f | 84 | dec_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_NODES); |
e05df3b1 JK |
85 | } |
86 | ClearPageUptodate(page); | |
87 | } | |
88 | ||
89 | static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
90 | { | |
91 | pgoff_t index = current_nat_addr(sbi, nid); | |
92 | return get_meta_page(sbi, index); | |
93 | } | |
94 | ||
95 | static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
96 | { | |
97 | struct page *src_page; | |
98 | struct page *dst_page; | |
99 | pgoff_t src_off; | |
100 | pgoff_t dst_off; | |
101 | void *src_addr; | |
102 | void *dst_addr; | |
103 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
104 | ||
105 | src_off = current_nat_addr(sbi, nid); | |
106 | dst_off = next_nat_addr(sbi, src_off); | |
107 | ||
108 | /* get current nat block page with lock */ | |
109 | src_page = get_meta_page(sbi, src_off); | |
e05df3b1 | 110 | dst_page = grab_meta_page(sbi, dst_off); |
9850cf4a | 111 | f2fs_bug_on(sbi, PageDirty(src_page)); |
e05df3b1 JK |
112 | |
113 | src_addr = page_address(src_page); | |
114 | dst_addr = page_address(dst_page); | |
115 | memcpy(dst_addr, src_addr, PAGE_CACHE_SIZE); | |
116 | set_page_dirty(dst_page); | |
117 | f2fs_put_page(src_page, 1); | |
118 | ||
119 | set_to_next_nat(nm_i, nid); | |
120 | ||
121 | return dst_page; | |
122 | } | |
123 | ||
e05df3b1 JK |
124 | static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n) |
125 | { | |
126 | return radix_tree_lookup(&nm_i->nat_root, n); | |
127 | } | |
128 | ||
129 | static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i, | |
130 | nid_t start, unsigned int nr, struct nat_entry **ep) | |
131 | { | |
132 | return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr); | |
133 | } | |
134 | ||
135 | static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e) | |
136 | { | |
137 | list_del(&e->list); | |
138 | radix_tree_delete(&nm_i->nat_root, nat_get_nid(e)); | |
139 | nm_i->nat_cnt--; | |
140 | kmem_cache_free(nat_entry_slab, e); | |
141 | } | |
142 | ||
309cc2b6 JK |
143 | static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i, |
144 | struct nat_entry *ne) | |
145 | { | |
146 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); | |
147 | struct nat_entry_set *head; | |
148 | ||
149 | if (get_nat_flag(ne, IS_DIRTY)) | |
150 | return; | |
9be32d72 | 151 | |
309cc2b6 JK |
152 | head = radix_tree_lookup(&nm_i->nat_set_root, set); |
153 | if (!head) { | |
154 | head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_ATOMIC); | |
155 | ||
156 | INIT_LIST_HEAD(&head->entry_list); | |
157 | INIT_LIST_HEAD(&head->set_list); | |
158 | head->set = set; | |
159 | head->entry_cnt = 0; | |
9be32d72 | 160 | f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head); |
309cc2b6 JK |
161 | } |
162 | list_move_tail(&ne->list, &head->entry_list); | |
163 | nm_i->dirty_nat_cnt++; | |
164 | head->entry_cnt++; | |
165 | set_nat_flag(ne, IS_DIRTY, true); | |
166 | } | |
167 | ||
168 | static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i, | |
169 | struct nat_entry *ne) | |
170 | { | |
20d047c8 | 171 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); |
309cc2b6 JK |
172 | struct nat_entry_set *head; |
173 | ||
174 | head = radix_tree_lookup(&nm_i->nat_set_root, set); | |
175 | if (head) { | |
176 | list_move_tail(&ne->list, &nm_i->nat_entries); | |
177 | set_nat_flag(ne, IS_DIRTY, false); | |
178 | head->entry_cnt--; | |
179 | nm_i->dirty_nat_cnt--; | |
180 | } | |
181 | } | |
182 | ||
183 | static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i, | |
184 | nid_t start, unsigned int nr, struct nat_entry_set **ep) | |
185 | { | |
186 | return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep, | |
187 | start, nr); | |
188 | } | |
189 | ||
88bd02c9 | 190 | bool is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 JK |
191 | { |
192 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
193 | struct nat_entry *e; | |
88bd02c9 | 194 | bool is_cp = true; |
e05df3b1 | 195 | |
8b26ef98 | 196 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 | 197 | e = __lookup_nat_cache(nm_i, nid); |
7ef35e3b | 198 | if (e && !get_nat_flag(e, IS_CHECKPOINTED)) |
88bd02c9 | 199 | is_cp = false; |
8b26ef98 | 200 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
201 | return is_cp; |
202 | } | |
203 | ||
88bd02c9 | 204 | bool has_fsynced_inode(struct f2fs_sb_info *sbi, nid_t ino) |
479f40c4 JK |
205 | { |
206 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
207 | struct nat_entry *e; | |
88bd02c9 | 208 | bool fsynced = false; |
479f40c4 | 209 | |
8b26ef98 | 210 | down_read(&nm_i->nat_tree_lock); |
88bd02c9 JK |
211 | e = __lookup_nat_cache(nm_i, ino); |
212 | if (e && get_nat_flag(e, HAS_FSYNCED_INODE)) | |
213 | fsynced = true; | |
8b26ef98 | 214 | up_read(&nm_i->nat_tree_lock); |
88bd02c9 | 215 | return fsynced; |
479f40c4 JK |
216 | } |
217 | ||
88bd02c9 | 218 | bool need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino) |
b6fe5873 JK |
219 | { |
220 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
221 | struct nat_entry *e; | |
88bd02c9 | 222 | bool need_update = true; |
b6fe5873 | 223 | |
8b26ef98 | 224 | down_read(&nm_i->nat_tree_lock); |
88bd02c9 JK |
225 | e = __lookup_nat_cache(nm_i, ino); |
226 | if (e && get_nat_flag(e, HAS_LAST_FSYNC) && | |
227 | (get_nat_flag(e, IS_CHECKPOINTED) || | |
228 | get_nat_flag(e, HAS_FSYNCED_INODE))) | |
229 | need_update = false; | |
8b26ef98 | 230 | up_read(&nm_i->nat_tree_lock); |
88bd02c9 | 231 | return need_update; |
b6fe5873 JK |
232 | } |
233 | ||
e05df3b1 JK |
234 | static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid) |
235 | { | |
236 | struct nat_entry *new; | |
237 | ||
9be32d72 JK |
238 | new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_ATOMIC); |
239 | f2fs_radix_tree_insert(&nm_i->nat_root, nid, new); | |
e05df3b1 JK |
240 | memset(new, 0, sizeof(struct nat_entry)); |
241 | nat_set_nid(new, nid); | |
88bd02c9 | 242 | nat_reset_flag(new); |
e05df3b1 JK |
243 | list_add_tail(&new->list, &nm_i->nat_entries); |
244 | nm_i->nat_cnt++; | |
245 | return new; | |
246 | } | |
247 | ||
248 | static void cache_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid, | |
249 | struct f2fs_nat_entry *ne) | |
250 | { | |
251 | struct nat_entry *e; | |
9be32d72 | 252 | |
8b26ef98 | 253 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
254 | e = __lookup_nat_cache(nm_i, nid); |
255 | if (!e) { | |
256 | e = grab_nat_entry(nm_i, nid); | |
94dac22e | 257 | node_info_from_raw_nat(&e->ni, ne); |
e05df3b1 | 258 | } |
8b26ef98 | 259 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
260 | } |
261 | ||
262 | static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, | |
479f40c4 | 263 | block_t new_blkaddr, bool fsync_done) |
e05df3b1 JK |
264 | { |
265 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
266 | struct nat_entry *e; | |
9be32d72 | 267 | |
8b26ef98 | 268 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
269 | e = __lookup_nat_cache(nm_i, ni->nid); |
270 | if (!e) { | |
271 | e = grab_nat_entry(nm_i, ni->nid); | |
e05df3b1 | 272 | e->ni = *ni; |
9850cf4a | 273 | f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR); |
e05df3b1 JK |
274 | } else if (new_blkaddr == NEW_ADDR) { |
275 | /* | |
276 | * when nid is reallocated, | |
277 | * previous nat entry can be remained in nat cache. | |
278 | * So, reinitialize it with new information. | |
279 | */ | |
280 | e->ni = *ni; | |
9850cf4a | 281 | f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR); |
e05df3b1 JK |
282 | } |
283 | ||
e05df3b1 | 284 | /* sanity check */ |
9850cf4a JK |
285 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr); |
286 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR && | |
e05df3b1 | 287 | new_blkaddr == NULL_ADDR); |
9850cf4a | 288 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR && |
e05df3b1 | 289 | new_blkaddr == NEW_ADDR); |
9850cf4a | 290 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != NEW_ADDR && |
e05df3b1 JK |
291 | nat_get_blkaddr(e) != NULL_ADDR && |
292 | new_blkaddr == NEW_ADDR); | |
293 | ||
e1c42045 | 294 | /* increment version no as node is removed */ |
e05df3b1 JK |
295 | if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) { |
296 | unsigned char version = nat_get_version(e); | |
297 | nat_set_version(e, inc_node_version(version)); | |
298 | } | |
299 | ||
300 | /* change address */ | |
301 | nat_set_blkaddr(e, new_blkaddr); | |
88bd02c9 JK |
302 | if (new_blkaddr == NEW_ADDR || new_blkaddr == NULL_ADDR) |
303 | set_nat_flag(e, IS_CHECKPOINTED, false); | |
e05df3b1 | 304 | __set_nat_cache_dirty(nm_i, e); |
479f40c4 JK |
305 | |
306 | /* update fsync_mark if its inode nat entry is still alive */ | |
307 | e = __lookup_nat_cache(nm_i, ni->ino); | |
88bd02c9 JK |
308 | if (e) { |
309 | if (fsync_done && ni->nid == ni->ino) | |
310 | set_nat_flag(e, HAS_FSYNCED_INODE, true); | |
311 | set_nat_flag(e, HAS_LAST_FSYNC, fsync_done); | |
312 | } | |
8b26ef98 | 313 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
314 | } |
315 | ||
4660f9c0 | 316 | int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) |
e05df3b1 JK |
317 | { |
318 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
319 | ||
6fb03f3a | 320 | if (available_free_memory(sbi, NAT_ENTRIES)) |
e05df3b1 JK |
321 | return 0; |
322 | ||
8b26ef98 | 323 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
324 | while (nr_shrink && !list_empty(&nm_i->nat_entries)) { |
325 | struct nat_entry *ne; | |
326 | ne = list_first_entry(&nm_i->nat_entries, | |
327 | struct nat_entry, list); | |
328 | __del_from_nat_cache(nm_i, ne); | |
329 | nr_shrink--; | |
330 | } | |
8b26ef98 | 331 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
332 | return nr_shrink; |
333 | } | |
334 | ||
0a8165d7 | 335 | /* |
e1c42045 | 336 | * This function always returns success |
e05df3b1 JK |
337 | */ |
338 | void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni) | |
339 | { | |
340 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
341 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
342 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
343 | nid_t start_nid = START_NID(nid); | |
344 | struct f2fs_nat_block *nat_blk; | |
345 | struct page *page = NULL; | |
346 | struct f2fs_nat_entry ne; | |
347 | struct nat_entry *e; | |
348 | int i; | |
349 | ||
be4124f8 | 350 | memset(&ne, 0, sizeof(struct f2fs_nat_entry)); |
e05df3b1 JK |
351 | ni->nid = nid; |
352 | ||
353 | /* Check nat cache */ | |
8b26ef98 | 354 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
355 | e = __lookup_nat_cache(nm_i, nid); |
356 | if (e) { | |
357 | ni->ino = nat_get_ino(e); | |
358 | ni->blk_addr = nat_get_blkaddr(e); | |
359 | ni->version = nat_get_version(e); | |
360 | } | |
8b26ef98 | 361 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
362 | if (e) |
363 | return; | |
364 | ||
365 | /* Check current segment summary */ | |
366 | mutex_lock(&curseg->curseg_mutex); | |
367 | i = lookup_journal_in_cursum(sum, NAT_JOURNAL, nid, 0); | |
368 | if (i >= 0) { | |
369 | ne = nat_in_journal(sum, i); | |
370 | node_info_from_raw_nat(ni, &ne); | |
371 | } | |
372 | mutex_unlock(&curseg->curseg_mutex); | |
373 | if (i >= 0) | |
374 | goto cache; | |
375 | ||
376 | /* Fill node_info from nat page */ | |
377 | page = get_current_nat_page(sbi, start_nid); | |
378 | nat_blk = (struct f2fs_nat_block *)page_address(page); | |
379 | ne = nat_blk->entries[nid - start_nid]; | |
380 | node_info_from_raw_nat(ni, &ne); | |
381 | f2fs_put_page(page, 1); | |
382 | cache: | |
383 | /* cache nat entry */ | |
384 | cache_nat_entry(NM_I(sbi), nid, &ne); | |
385 | } | |
386 | ||
0a8165d7 | 387 | /* |
e05df3b1 JK |
388 | * The maximum depth is four. |
389 | * Offset[0] will have raw inode offset. | |
390 | */ | |
de93653f JK |
391 | static int get_node_path(struct f2fs_inode_info *fi, long block, |
392 | int offset[4], unsigned int noffset[4]) | |
e05df3b1 | 393 | { |
de93653f | 394 | const long direct_index = ADDRS_PER_INODE(fi); |
e05df3b1 JK |
395 | const long direct_blks = ADDRS_PER_BLOCK; |
396 | const long dptrs_per_blk = NIDS_PER_BLOCK; | |
397 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
398 | const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK; | |
399 | int n = 0; | |
400 | int level = 0; | |
401 | ||
402 | noffset[0] = 0; | |
403 | ||
404 | if (block < direct_index) { | |
25c0a6e5 | 405 | offset[n] = block; |
e05df3b1 JK |
406 | goto got; |
407 | } | |
408 | block -= direct_index; | |
409 | if (block < direct_blks) { | |
410 | offset[n++] = NODE_DIR1_BLOCK; | |
411 | noffset[n] = 1; | |
25c0a6e5 | 412 | offset[n] = block; |
e05df3b1 JK |
413 | level = 1; |
414 | goto got; | |
415 | } | |
416 | block -= direct_blks; | |
417 | if (block < direct_blks) { | |
418 | offset[n++] = NODE_DIR2_BLOCK; | |
419 | noffset[n] = 2; | |
25c0a6e5 | 420 | offset[n] = block; |
e05df3b1 JK |
421 | level = 1; |
422 | goto got; | |
423 | } | |
424 | block -= direct_blks; | |
425 | if (block < indirect_blks) { | |
426 | offset[n++] = NODE_IND1_BLOCK; | |
427 | noffset[n] = 3; | |
428 | offset[n++] = block / direct_blks; | |
429 | noffset[n] = 4 + offset[n - 1]; | |
25c0a6e5 | 430 | offset[n] = block % direct_blks; |
e05df3b1 JK |
431 | level = 2; |
432 | goto got; | |
433 | } | |
434 | block -= indirect_blks; | |
435 | if (block < indirect_blks) { | |
436 | offset[n++] = NODE_IND2_BLOCK; | |
437 | noffset[n] = 4 + dptrs_per_blk; | |
438 | offset[n++] = block / direct_blks; | |
439 | noffset[n] = 5 + dptrs_per_blk + offset[n - 1]; | |
25c0a6e5 | 440 | offset[n] = block % direct_blks; |
e05df3b1 JK |
441 | level = 2; |
442 | goto got; | |
443 | } | |
444 | block -= indirect_blks; | |
445 | if (block < dindirect_blks) { | |
446 | offset[n++] = NODE_DIND_BLOCK; | |
447 | noffset[n] = 5 + (dptrs_per_blk * 2); | |
448 | offset[n++] = block / indirect_blks; | |
449 | noffset[n] = 6 + (dptrs_per_blk * 2) + | |
450 | offset[n - 1] * (dptrs_per_blk + 1); | |
451 | offset[n++] = (block / direct_blks) % dptrs_per_blk; | |
452 | noffset[n] = 7 + (dptrs_per_blk * 2) + | |
453 | offset[n - 2] * (dptrs_per_blk + 1) + | |
454 | offset[n - 1]; | |
25c0a6e5 | 455 | offset[n] = block % direct_blks; |
e05df3b1 JK |
456 | level = 3; |
457 | goto got; | |
458 | } else { | |
459 | BUG(); | |
460 | } | |
461 | got: | |
462 | return level; | |
463 | } | |
464 | ||
465 | /* | |
466 | * Caller should call f2fs_put_dnode(dn). | |
4f4124d0 CY |
467 | * Also, it should grab and release a rwsem by calling f2fs_lock_op() and |
468 | * f2fs_unlock_op() only if ro is not set RDONLY_NODE. | |
39936837 | 469 | * In the case of RDONLY_NODE, we don't need to care about mutex. |
e05df3b1 | 470 | */ |
266e97a8 | 471 | int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode) |
e05df3b1 | 472 | { |
4081363f | 473 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
474 | struct page *npage[4]; |
475 | struct page *parent; | |
476 | int offset[4]; | |
477 | unsigned int noffset[4]; | |
478 | nid_t nids[4]; | |
479 | int level, i; | |
480 | int err = 0; | |
481 | ||
de93653f | 482 | level = get_node_path(F2FS_I(dn->inode), index, offset, noffset); |
e05df3b1 JK |
483 | |
484 | nids[0] = dn->inode->i_ino; | |
1646cfac | 485 | npage[0] = dn->inode_page; |
e05df3b1 | 486 | |
1646cfac JK |
487 | if (!npage[0]) { |
488 | npage[0] = get_node_page(sbi, nids[0]); | |
489 | if (IS_ERR(npage[0])) | |
490 | return PTR_ERR(npage[0]); | |
491 | } | |
e05df3b1 | 492 | parent = npage[0]; |
52c2db3f CL |
493 | if (level != 0) |
494 | nids[1] = get_nid(parent, offset[0], true); | |
e05df3b1 JK |
495 | dn->inode_page = npage[0]; |
496 | dn->inode_page_locked = true; | |
497 | ||
498 | /* get indirect or direct nodes */ | |
499 | for (i = 1; i <= level; i++) { | |
500 | bool done = false; | |
501 | ||
266e97a8 | 502 | if (!nids[i] && mode == ALLOC_NODE) { |
e05df3b1 JK |
503 | /* alloc new node */ |
504 | if (!alloc_nid(sbi, &(nids[i]))) { | |
e05df3b1 JK |
505 | err = -ENOSPC; |
506 | goto release_pages; | |
507 | } | |
508 | ||
509 | dn->nid = nids[i]; | |
8ae8f162 | 510 | npage[i] = new_node_page(dn, noffset[i], NULL); |
e05df3b1 JK |
511 | if (IS_ERR(npage[i])) { |
512 | alloc_nid_failed(sbi, nids[i]); | |
e05df3b1 JK |
513 | err = PTR_ERR(npage[i]); |
514 | goto release_pages; | |
515 | } | |
516 | ||
517 | set_nid(parent, offset[i - 1], nids[i], i == 1); | |
518 | alloc_nid_done(sbi, nids[i]); | |
e05df3b1 | 519 | done = true; |
266e97a8 | 520 | } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) { |
e05df3b1 JK |
521 | npage[i] = get_node_page_ra(parent, offset[i - 1]); |
522 | if (IS_ERR(npage[i])) { | |
523 | err = PTR_ERR(npage[i]); | |
524 | goto release_pages; | |
525 | } | |
526 | done = true; | |
527 | } | |
528 | if (i == 1) { | |
529 | dn->inode_page_locked = false; | |
530 | unlock_page(parent); | |
531 | } else { | |
532 | f2fs_put_page(parent, 1); | |
533 | } | |
534 | ||
535 | if (!done) { | |
536 | npage[i] = get_node_page(sbi, nids[i]); | |
537 | if (IS_ERR(npage[i])) { | |
538 | err = PTR_ERR(npage[i]); | |
539 | f2fs_put_page(npage[0], 0); | |
540 | goto release_out; | |
541 | } | |
542 | } | |
543 | if (i < level) { | |
544 | parent = npage[i]; | |
545 | nids[i + 1] = get_nid(parent, offset[i], false); | |
546 | } | |
547 | } | |
548 | dn->nid = nids[level]; | |
549 | dn->ofs_in_node = offset[level]; | |
550 | dn->node_page = npage[level]; | |
551 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
552 | return 0; | |
553 | ||
554 | release_pages: | |
555 | f2fs_put_page(parent, 1); | |
556 | if (i > 1) | |
557 | f2fs_put_page(npage[0], 0); | |
558 | release_out: | |
559 | dn->inode_page = NULL; | |
560 | dn->node_page = NULL; | |
561 | return err; | |
562 | } | |
563 | ||
564 | static void truncate_node(struct dnode_of_data *dn) | |
565 | { | |
4081363f | 566 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
567 | struct node_info ni; |
568 | ||
569 | get_node_info(sbi, dn->nid, &ni); | |
71e9fec5 | 570 | if (dn->inode->i_blocks == 0) { |
9850cf4a | 571 | f2fs_bug_on(sbi, ni.blk_addr != NULL_ADDR); |
71e9fec5 JK |
572 | goto invalidate; |
573 | } | |
9850cf4a | 574 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
e05df3b1 | 575 | |
e05df3b1 | 576 | /* Deallocate node address */ |
71e9fec5 | 577 | invalidate_blocks(sbi, ni.blk_addr); |
ef86d709 | 578 | dec_valid_node_count(sbi, dn->inode); |
479f40c4 | 579 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
e05df3b1 JK |
580 | |
581 | if (dn->nid == dn->inode->i_ino) { | |
582 | remove_orphan_inode(sbi, dn->nid); | |
583 | dec_valid_inode_count(sbi); | |
584 | } else { | |
585 | sync_inode_page(dn); | |
586 | } | |
71e9fec5 | 587 | invalidate: |
e05df3b1 JK |
588 | clear_node_page_dirty(dn->node_page); |
589 | F2FS_SET_SB_DIRT(sbi); | |
590 | ||
591 | f2fs_put_page(dn->node_page, 1); | |
bf39c00a JK |
592 | |
593 | invalidate_mapping_pages(NODE_MAPPING(sbi), | |
594 | dn->node_page->index, dn->node_page->index); | |
595 | ||
e05df3b1 | 596 | dn->node_page = NULL; |
51dd6249 | 597 | trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr); |
e05df3b1 JK |
598 | } |
599 | ||
600 | static int truncate_dnode(struct dnode_of_data *dn) | |
601 | { | |
e05df3b1 JK |
602 | struct page *page; |
603 | ||
604 | if (dn->nid == 0) | |
605 | return 1; | |
606 | ||
607 | /* get direct node */ | |
4081363f | 608 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
e05df3b1 JK |
609 | if (IS_ERR(page) && PTR_ERR(page) == -ENOENT) |
610 | return 1; | |
611 | else if (IS_ERR(page)) | |
612 | return PTR_ERR(page); | |
613 | ||
614 | /* Make dnode_of_data for parameter */ | |
615 | dn->node_page = page; | |
616 | dn->ofs_in_node = 0; | |
617 | truncate_data_blocks(dn); | |
618 | truncate_node(dn); | |
619 | return 1; | |
620 | } | |
621 | ||
622 | static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs, | |
623 | int ofs, int depth) | |
624 | { | |
e05df3b1 JK |
625 | struct dnode_of_data rdn = *dn; |
626 | struct page *page; | |
627 | struct f2fs_node *rn; | |
628 | nid_t child_nid; | |
629 | unsigned int child_nofs; | |
630 | int freed = 0; | |
631 | int i, ret; | |
632 | ||
633 | if (dn->nid == 0) | |
634 | return NIDS_PER_BLOCK + 1; | |
635 | ||
51dd6249 NJ |
636 | trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr); |
637 | ||
4081363f | 638 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
51dd6249 NJ |
639 | if (IS_ERR(page)) { |
640 | trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page)); | |
e05df3b1 | 641 | return PTR_ERR(page); |
51dd6249 | 642 | } |
e05df3b1 | 643 | |
45590710 | 644 | rn = F2FS_NODE(page); |
e05df3b1 JK |
645 | if (depth < 3) { |
646 | for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) { | |
647 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
648 | if (child_nid == 0) | |
649 | continue; | |
650 | rdn.nid = child_nid; | |
651 | ret = truncate_dnode(&rdn); | |
652 | if (ret < 0) | |
653 | goto out_err; | |
654 | set_nid(page, i, 0, false); | |
655 | } | |
656 | } else { | |
657 | child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1; | |
658 | for (i = ofs; i < NIDS_PER_BLOCK; i++) { | |
659 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
660 | if (child_nid == 0) { | |
661 | child_nofs += NIDS_PER_BLOCK + 1; | |
662 | continue; | |
663 | } | |
664 | rdn.nid = child_nid; | |
665 | ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1); | |
666 | if (ret == (NIDS_PER_BLOCK + 1)) { | |
667 | set_nid(page, i, 0, false); | |
668 | child_nofs += ret; | |
669 | } else if (ret < 0 && ret != -ENOENT) { | |
670 | goto out_err; | |
671 | } | |
672 | } | |
673 | freed = child_nofs; | |
674 | } | |
675 | ||
676 | if (!ofs) { | |
677 | /* remove current indirect node */ | |
678 | dn->node_page = page; | |
679 | truncate_node(dn); | |
680 | freed++; | |
681 | } else { | |
682 | f2fs_put_page(page, 1); | |
683 | } | |
51dd6249 | 684 | trace_f2fs_truncate_nodes_exit(dn->inode, freed); |
e05df3b1 JK |
685 | return freed; |
686 | ||
687 | out_err: | |
688 | f2fs_put_page(page, 1); | |
51dd6249 | 689 | trace_f2fs_truncate_nodes_exit(dn->inode, ret); |
e05df3b1 JK |
690 | return ret; |
691 | } | |
692 | ||
693 | static int truncate_partial_nodes(struct dnode_of_data *dn, | |
694 | struct f2fs_inode *ri, int *offset, int depth) | |
695 | { | |
e05df3b1 JK |
696 | struct page *pages[2]; |
697 | nid_t nid[3]; | |
698 | nid_t child_nid; | |
699 | int err = 0; | |
700 | int i; | |
701 | int idx = depth - 2; | |
702 | ||
703 | nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); | |
704 | if (!nid[0]) | |
705 | return 0; | |
706 | ||
707 | /* get indirect nodes in the path */ | |
a225dca3 | 708 | for (i = 0; i < idx + 1; i++) { |
e1c42045 | 709 | /* reference count'll be increased */ |
4081363f | 710 | pages[i] = get_node_page(F2FS_I_SB(dn->inode), nid[i]); |
e05df3b1 | 711 | if (IS_ERR(pages[i])) { |
e05df3b1 | 712 | err = PTR_ERR(pages[i]); |
a225dca3 | 713 | idx = i - 1; |
e05df3b1 JK |
714 | goto fail; |
715 | } | |
716 | nid[i + 1] = get_nid(pages[i], offset[i + 1], false); | |
717 | } | |
718 | ||
719 | /* free direct nodes linked to a partial indirect node */ | |
a225dca3 | 720 | for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) { |
e05df3b1 JK |
721 | child_nid = get_nid(pages[idx], i, false); |
722 | if (!child_nid) | |
723 | continue; | |
724 | dn->nid = child_nid; | |
725 | err = truncate_dnode(dn); | |
726 | if (err < 0) | |
727 | goto fail; | |
728 | set_nid(pages[idx], i, 0, false); | |
729 | } | |
730 | ||
a225dca3 | 731 | if (offset[idx + 1] == 0) { |
e05df3b1 JK |
732 | dn->node_page = pages[idx]; |
733 | dn->nid = nid[idx]; | |
734 | truncate_node(dn); | |
735 | } else { | |
736 | f2fs_put_page(pages[idx], 1); | |
737 | } | |
738 | offset[idx]++; | |
a225dca3 | 739 | offset[idx + 1] = 0; |
740 | idx--; | |
e05df3b1 | 741 | fail: |
a225dca3 | 742 | for (i = idx; i >= 0; i--) |
e05df3b1 | 743 | f2fs_put_page(pages[i], 1); |
51dd6249 NJ |
744 | |
745 | trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err); | |
746 | ||
e05df3b1 JK |
747 | return err; |
748 | } | |
749 | ||
0a8165d7 | 750 | /* |
e05df3b1 JK |
751 | * All the block addresses of data and nodes should be nullified. |
752 | */ | |
753 | int truncate_inode_blocks(struct inode *inode, pgoff_t from) | |
754 | { | |
4081363f | 755 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
e05df3b1 JK |
756 | int err = 0, cont = 1; |
757 | int level, offset[4], noffset[4]; | |
7dd690c8 | 758 | unsigned int nofs = 0; |
58bfaf44 | 759 | struct f2fs_inode *ri; |
e05df3b1 JK |
760 | struct dnode_of_data dn; |
761 | struct page *page; | |
762 | ||
51dd6249 NJ |
763 | trace_f2fs_truncate_inode_blocks_enter(inode, from); |
764 | ||
de93653f | 765 | level = get_node_path(F2FS_I(inode), from, offset, noffset); |
afcb7ca0 | 766 | restart: |
e05df3b1 | 767 | page = get_node_page(sbi, inode->i_ino); |
51dd6249 NJ |
768 | if (IS_ERR(page)) { |
769 | trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page)); | |
e05df3b1 | 770 | return PTR_ERR(page); |
51dd6249 | 771 | } |
e05df3b1 JK |
772 | |
773 | set_new_dnode(&dn, inode, page, NULL, 0); | |
774 | unlock_page(page); | |
775 | ||
58bfaf44 | 776 | ri = F2FS_INODE(page); |
e05df3b1 JK |
777 | switch (level) { |
778 | case 0: | |
779 | case 1: | |
780 | nofs = noffset[1]; | |
781 | break; | |
782 | case 2: | |
783 | nofs = noffset[1]; | |
784 | if (!offset[level - 1]) | |
785 | goto skip_partial; | |
58bfaf44 | 786 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
787 | if (err < 0 && err != -ENOENT) |
788 | goto fail; | |
789 | nofs += 1 + NIDS_PER_BLOCK; | |
790 | break; | |
791 | case 3: | |
792 | nofs = 5 + 2 * NIDS_PER_BLOCK; | |
793 | if (!offset[level - 1]) | |
794 | goto skip_partial; | |
58bfaf44 | 795 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
796 | if (err < 0 && err != -ENOENT) |
797 | goto fail; | |
798 | break; | |
799 | default: | |
800 | BUG(); | |
801 | } | |
802 | ||
803 | skip_partial: | |
804 | while (cont) { | |
58bfaf44 | 805 | dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); |
e05df3b1 JK |
806 | switch (offset[0]) { |
807 | case NODE_DIR1_BLOCK: | |
808 | case NODE_DIR2_BLOCK: | |
809 | err = truncate_dnode(&dn); | |
810 | break; | |
811 | ||
812 | case NODE_IND1_BLOCK: | |
813 | case NODE_IND2_BLOCK: | |
814 | err = truncate_nodes(&dn, nofs, offset[1], 2); | |
815 | break; | |
816 | ||
817 | case NODE_DIND_BLOCK: | |
818 | err = truncate_nodes(&dn, nofs, offset[1], 3); | |
819 | cont = 0; | |
820 | break; | |
821 | ||
822 | default: | |
823 | BUG(); | |
824 | } | |
825 | if (err < 0 && err != -ENOENT) | |
826 | goto fail; | |
827 | if (offset[1] == 0 && | |
58bfaf44 | 828 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) { |
e05df3b1 | 829 | lock_page(page); |
4ef51a8f | 830 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
831 | f2fs_put_page(page, 1); |
832 | goto restart; | |
833 | } | |
3cb5ad15 | 834 | f2fs_wait_on_page_writeback(page, NODE); |
58bfaf44 | 835 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0; |
e05df3b1 JK |
836 | set_page_dirty(page); |
837 | unlock_page(page); | |
838 | } | |
839 | offset[1] = 0; | |
840 | offset[0]++; | |
841 | nofs += err; | |
842 | } | |
843 | fail: | |
844 | f2fs_put_page(page, 0); | |
51dd6249 | 845 | trace_f2fs_truncate_inode_blocks_exit(inode, err); |
e05df3b1 JK |
846 | return err > 0 ? 0 : err; |
847 | } | |
848 | ||
4f16fb0f JK |
849 | int truncate_xattr_node(struct inode *inode, struct page *page) |
850 | { | |
4081363f | 851 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
4f16fb0f JK |
852 | nid_t nid = F2FS_I(inode)->i_xattr_nid; |
853 | struct dnode_of_data dn; | |
854 | struct page *npage; | |
855 | ||
856 | if (!nid) | |
857 | return 0; | |
858 | ||
859 | npage = get_node_page(sbi, nid); | |
860 | if (IS_ERR(npage)) | |
861 | return PTR_ERR(npage); | |
862 | ||
863 | F2FS_I(inode)->i_xattr_nid = 0; | |
65985d93 JK |
864 | |
865 | /* need to do checkpoint during fsync */ | |
866 | F2FS_I(inode)->xattr_ver = cur_cp_version(F2FS_CKPT(sbi)); | |
867 | ||
4f16fb0f JK |
868 | set_new_dnode(&dn, inode, page, npage, nid); |
869 | ||
870 | if (page) | |
01d2d1aa | 871 | dn.inode_page_locked = true; |
4f16fb0f JK |
872 | truncate_node(&dn); |
873 | return 0; | |
874 | } | |
875 | ||
39936837 | 876 | /* |
4f4124d0 CY |
877 | * Caller should grab and release a rwsem by calling f2fs_lock_op() and |
878 | * f2fs_unlock_op(). | |
39936837 | 879 | */ |
58e674d6 | 880 | void remove_inode_page(struct inode *inode) |
e05df3b1 | 881 | { |
e05df3b1 JK |
882 | struct dnode_of_data dn; |
883 | ||
c2e69583 JK |
884 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); |
885 | if (get_dnode_of_data(&dn, 0, LOOKUP_NODE)) | |
58e674d6 | 886 | return; |
e05df3b1 | 887 | |
c2e69583 JK |
888 | if (truncate_xattr_node(inode, dn.inode_page)) { |
889 | f2fs_put_dnode(&dn); | |
58e674d6 | 890 | return; |
e05df3b1 | 891 | } |
c2e69583 JK |
892 | |
893 | /* remove potential inline_data blocks */ | |
894 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
895 | S_ISLNK(inode->i_mode)) | |
896 | truncate_data_blocks_range(&dn, 1); | |
897 | ||
e1c42045 | 898 | /* 0 is possible, after f2fs_new_inode() has failed */ |
9850cf4a JK |
899 | f2fs_bug_on(F2FS_I_SB(inode), |
900 | inode->i_blocks != 0 && inode->i_blocks != 1); | |
c2e69583 JK |
901 | |
902 | /* will put inode & node pages */ | |
71e9fec5 | 903 | truncate_node(&dn); |
e05df3b1 JK |
904 | } |
905 | ||
a014e037 | 906 | struct page *new_inode_page(struct inode *inode) |
e05df3b1 | 907 | { |
e05df3b1 JK |
908 | struct dnode_of_data dn; |
909 | ||
910 | /* allocate inode page for new inode */ | |
911 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); | |
44a83ff6 JK |
912 | |
913 | /* caller should f2fs_put_page(page, 1); */ | |
8ae8f162 | 914 | return new_node_page(&dn, 0, NULL); |
e05df3b1 JK |
915 | } |
916 | ||
8ae8f162 JK |
917 | struct page *new_node_page(struct dnode_of_data *dn, |
918 | unsigned int ofs, struct page *ipage) | |
e05df3b1 | 919 | { |
4081363f | 920 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
921 | struct node_info old_ni, new_ni; |
922 | struct page *page; | |
923 | int err; | |
924 | ||
6bacf52f | 925 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
e05df3b1 JK |
926 | return ERR_PTR(-EPERM); |
927 | ||
54b591df | 928 | page = grab_cache_page(NODE_MAPPING(sbi), dn->nid); |
e05df3b1 JK |
929 | if (!page) |
930 | return ERR_PTR(-ENOMEM); | |
931 | ||
6bacf52f | 932 | if (unlikely(!inc_valid_node_count(sbi, dn->inode))) { |
9c02740c JK |
933 | err = -ENOSPC; |
934 | goto fail; | |
935 | } | |
e05df3b1 | 936 | |
9c02740c | 937 | get_node_info(sbi, dn->nid, &old_ni); |
e05df3b1 JK |
938 | |
939 | /* Reinitialize old_ni with new node page */ | |
9850cf4a | 940 | f2fs_bug_on(sbi, old_ni.blk_addr != NULL_ADDR); |
e05df3b1 JK |
941 | new_ni = old_ni; |
942 | new_ni.ino = dn->inode->i_ino; | |
479f40c4 | 943 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
9c02740c | 944 | |
54b591df | 945 | f2fs_wait_on_page_writeback(page, NODE); |
9c02740c | 946 | fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true); |
398b1ac5 | 947 | set_cold_node(dn->inode, page); |
9c02740c JK |
948 | SetPageUptodate(page); |
949 | set_page_dirty(page); | |
e05df3b1 | 950 | |
4bc8e9bc | 951 | if (f2fs_has_xattr_block(ofs)) |
479bd73a JK |
952 | F2FS_I(dn->inode)->i_xattr_nid = dn->nid; |
953 | ||
e05df3b1 | 954 | dn->node_page = page; |
8ae8f162 JK |
955 | if (ipage) |
956 | update_inode(dn->inode, ipage); | |
957 | else | |
958 | sync_inode_page(dn); | |
e05df3b1 JK |
959 | if (ofs == 0) |
960 | inc_valid_inode_count(sbi); | |
961 | ||
962 | return page; | |
963 | ||
964 | fail: | |
71e9fec5 | 965 | clear_node_page_dirty(page); |
e05df3b1 JK |
966 | f2fs_put_page(page, 1); |
967 | return ERR_PTR(err); | |
968 | } | |
969 | ||
56ae674c JK |
970 | /* |
971 | * Caller should do after getting the following values. | |
972 | * 0: f2fs_put_page(page, 0) | |
973 | * LOCKED_PAGE: f2fs_put_page(page, 1) | |
974 | * error: nothing | |
975 | */ | |
93dfe2ac | 976 | static int read_node_page(struct page *page, int rw) |
e05df3b1 | 977 | { |
4081363f | 978 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 JK |
979 | struct node_info ni; |
980 | ||
981 | get_node_info(sbi, page->index, &ni); | |
982 | ||
6bacf52f | 983 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
393ff91f | 984 | f2fs_put_page(page, 1); |
e05df3b1 | 985 | return -ENOENT; |
393ff91f JK |
986 | } |
987 | ||
56ae674c JK |
988 | if (PageUptodate(page)) |
989 | return LOCKED_PAGE; | |
393ff91f | 990 | |
93dfe2ac | 991 | return f2fs_submit_page_bio(sbi, page, ni.blk_addr, rw); |
e05df3b1 JK |
992 | } |
993 | ||
0a8165d7 | 994 | /* |
e05df3b1 JK |
995 | * Readahead a node page |
996 | */ | |
997 | void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid) | |
998 | { | |
e05df3b1 | 999 | struct page *apage; |
56ae674c | 1000 | int err; |
e05df3b1 | 1001 | |
4ef51a8f | 1002 | apage = find_get_page(NODE_MAPPING(sbi), nid); |
393ff91f JK |
1003 | if (apage && PageUptodate(apage)) { |
1004 | f2fs_put_page(apage, 0); | |
1005 | return; | |
1006 | } | |
e05df3b1 JK |
1007 | f2fs_put_page(apage, 0); |
1008 | ||
4ef51a8f | 1009 | apage = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1010 | if (!apage) |
1011 | return; | |
1012 | ||
56ae674c JK |
1013 | err = read_node_page(apage, READA); |
1014 | if (err == 0) | |
393ff91f | 1015 | f2fs_put_page(apage, 0); |
56ae674c JK |
1016 | else if (err == LOCKED_PAGE) |
1017 | f2fs_put_page(apage, 1); | |
e05df3b1 JK |
1018 | } |
1019 | ||
1020 | struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) | |
1021 | { | |
56ae674c JK |
1022 | struct page *page; |
1023 | int err; | |
afcb7ca0 | 1024 | repeat: |
54b591df | 1025 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1026 | if (!page) |
1027 | return ERR_PTR(-ENOMEM); | |
1028 | ||
1029 | err = read_node_page(page, READ_SYNC); | |
56ae674c | 1030 | if (err < 0) |
e05df3b1 | 1031 | return ERR_PTR(err); |
56ae674c JK |
1032 | else if (err == LOCKED_PAGE) |
1033 | goto got_it; | |
e05df3b1 | 1034 | |
393ff91f | 1035 | lock_page(page); |
3bb5e2c8 | 1036 | if (unlikely(!PageUptodate(page) || nid != nid_of_node(page))) { |
393ff91f JK |
1037 | f2fs_put_page(page, 1); |
1038 | return ERR_PTR(-EIO); | |
1039 | } | |
4ef51a8f | 1040 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1041 | f2fs_put_page(page, 1); |
1042 | goto repeat; | |
1043 | } | |
56ae674c | 1044 | got_it: |
e05df3b1 JK |
1045 | return page; |
1046 | } | |
1047 | ||
0a8165d7 | 1048 | /* |
e05df3b1 JK |
1049 | * Return a locked page for the desired node page. |
1050 | * And, readahead MAX_RA_NODE number of node pages. | |
1051 | */ | |
1052 | struct page *get_node_page_ra(struct page *parent, int start) | |
1053 | { | |
4081363f | 1054 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); |
c718379b | 1055 | struct blk_plug plug; |
e05df3b1 | 1056 | struct page *page; |
56ae674c JK |
1057 | int err, i, end; |
1058 | nid_t nid; | |
e05df3b1 JK |
1059 | |
1060 | /* First, try getting the desired direct node. */ | |
1061 | nid = get_nid(parent, start, false); | |
1062 | if (!nid) | |
1063 | return ERR_PTR(-ENOENT); | |
afcb7ca0 | 1064 | repeat: |
4ef51a8f | 1065 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1066 | if (!page) |
1067 | return ERR_PTR(-ENOMEM); | |
1068 | ||
66d36a29 | 1069 | err = read_node_page(page, READ_SYNC); |
56ae674c | 1070 | if (err < 0) |
e05df3b1 | 1071 | return ERR_PTR(err); |
56ae674c JK |
1072 | else if (err == LOCKED_PAGE) |
1073 | goto page_hit; | |
e05df3b1 | 1074 | |
c718379b JK |
1075 | blk_start_plug(&plug); |
1076 | ||
e05df3b1 JK |
1077 | /* Then, try readahead for siblings of the desired node */ |
1078 | end = start + MAX_RA_NODE; | |
1079 | end = min(end, NIDS_PER_BLOCK); | |
1080 | for (i = start + 1; i < end; i++) { | |
1081 | nid = get_nid(parent, i, false); | |
1082 | if (!nid) | |
1083 | continue; | |
1084 | ra_node_page(sbi, nid); | |
1085 | } | |
1086 | ||
c718379b JK |
1087 | blk_finish_plug(&plug); |
1088 | ||
e05df3b1 | 1089 | lock_page(page); |
4ef51a8f | 1090 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1091 | f2fs_put_page(page, 1); |
1092 | goto repeat; | |
1093 | } | |
e0f56cb4 | 1094 | page_hit: |
6bacf52f | 1095 | if (unlikely(!PageUptodate(page))) { |
e05df3b1 JK |
1096 | f2fs_put_page(page, 1); |
1097 | return ERR_PTR(-EIO); | |
1098 | } | |
e05df3b1 JK |
1099 | return page; |
1100 | } | |
1101 | ||
1102 | void sync_inode_page(struct dnode_of_data *dn) | |
1103 | { | |
1104 | if (IS_INODE(dn->node_page) || dn->inode_page == dn->node_page) { | |
1105 | update_inode(dn->inode, dn->node_page); | |
1106 | } else if (dn->inode_page) { | |
1107 | if (!dn->inode_page_locked) | |
1108 | lock_page(dn->inode_page); | |
1109 | update_inode(dn->inode, dn->inode_page); | |
1110 | if (!dn->inode_page_locked) | |
1111 | unlock_page(dn->inode_page); | |
1112 | } else { | |
39936837 | 1113 | update_inode_page(dn->inode); |
e05df3b1 JK |
1114 | } |
1115 | } | |
1116 | ||
1117 | int sync_node_pages(struct f2fs_sb_info *sbi, nid_t ino, | |
1118 | struct writeback_control *wbc) | |
1119 | { | |
e05df3b1 JK |
1120 | pgoff_t index, end; |
1121 | struct pagevec pvec; | |
1122 | int step = ino ? 2 : 0; | |
1123 | int nwritten = 0, wrote = 0; | |
1124 | ||
1125 | pagevec_init(&pvec, 0); | |
1126 | ||
1127 | next_step: | |
1128 | index = 0; | |
1129 | end = LONG_MAX; | |
1130 | ||
1131 | while (index <= end) { | |
1132 | int i, nr_pages; | |
4ef51a8f | 1133 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, |
e05df3b1 JK |
1134 | PAGECACHE_TAG_DIRTY, |
1135 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1136 | if (nr_pages == 0) | |
1137 | break; | |
1138 | ||
1139 | for (i = 0; i < nr_pages; i++) { | |
1140 | struct page *page = pvec.pages[i]; | |
1141 | ||
1142 | /* | |
1143 | * flushing sequence with step: | |
1144 | * 0. indirect nodes | |
1145 | * 1. dentry dnodes | |
1146 | * 2. file dnodes | |
1147 | */ | |
1148 | if (step == 0 && IS_DNODE(page)) | |
1149 | continue; | |
1150 | if (step == 1 && (!IS_DNODE(page) || | |
1151 | is_cold_node(page))) | |
1152 | continue; | |
1153 | if (step == 2 && (!IS_DNODE(page) || | |
1154 | !is_cold_node(page))) | |
1155 | continue; | |
1156 | ||
1157 | /* | |
1158 | * If an fsync mode, | |
1159 | * we should not skip writing node pages. | |
1160 | */ | |
1161 | if (ino && ino_of_node(page) == ino) | |
1162 | lock_page(page); | |
1163 | else if (!trylock_page(page)) | |
1164 | continue; | |
1165 | ||
4ef51a8f | 1166 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
e05df3b1 JK |
1167 | continue_unlock: |
1168 | unlock_page(page); | |
1169 | continue; | |
1170 | } | |
1171 | if (ino && ino_of_node(page) != ino) | |
1172 | goto continue_unlock; | |
1173 | ||
1174 | if (!PageDirty(page)) { | |
1175 | /* someone wrote it for us */ | |
1176 | goto continue_unlock; | |
1177 | } | |
1178 | ||
1179 | if (!clear_page_dirty_for_io(page)) | |
1180 | goto continue_unlock; | |
1181 | ||
1182 | /* called by fsync() */ | |
1183 | if (ino && IS_DNODE(page)) { | |
e05df3b1 | 1184 | set_fsync_mark(page, 1); |
88bd02c9 JK |
1185 | if (IS_INODE(page)) { |
1186 | if (!is_checkpointed_node(sbi, ino) && | |
1187 | !has_fsynced_inode(sbi, ino)) | |
1188 | set_dentry_mark(page, 1); | |
1189 | else | |
1190 | set_dentry_mark(page, 0); | |
1191 | } | |
e05df3b1 JK |
1192 | nwritten++; |
1193 | } else { | |
1194 | set_fsync_mark(page, 0); | |
1195 | set_dentry_mark(page, 0); | |
1196 | } | |
52746519 JK |
1197 | |
1198 | if (NODE_MAPPING(sbi)->a_ops->writepage(page, wbc)) | |
1199 | unlock_page(page); | |
1200 | else | |
1201 | wrote++; | |
e05df3b1 JK |
1202 | |
1203 | if (--wbc->nr_to_write == 0) | |
1204 | break; | |
1205 | } | |
1206 | pagevec_release(&pvec); | |
1207 | cond_resched(); | |
1208 | ||
1209 | if (wbc->nr_to_write == 0) { | |
1210 | step = 2; | |
1211 | break; | |
1212 | } | |
1213 | } | |
1214 | ||
1215 | if (step < 2) { | |
1216 | step++; | |
1217 | goto next_step; | |
1218 | } | |
1219 | ||
1220 | if (wrote) | |
458e6197 | 1221 | f2fs_submit_merged_bio(sbi, NODE, WRITE); |
e05df3b1 JK |
1222 | return nwritten; |
1223 | } | |
1224 | ||
cfe58f9d JK |
1225 | int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino) |
1226 | { | |
cfe58f9d JK |
1227 | pgoff_t index = 0, end = LONG_MAX; |
1228 | struct pagevec pvec; | |
cfe58f9d JK |
1229 | int ret2 = 0, ret = 0; |
1230 | ||
1231 | pagevec_init(&pvec, 0); | |
4ef51a8f JK |
1232 | |
1233 | while (index <= end) { | |
1234 | int i, nr_pages; | |
1235 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1236 | PAGECACHE_TAG_WRITEBACK, | |
1237 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1238 | if (nr_pages == 0) | |
1239 | break; | |
cfe58f9d JK |
1240 | |
1241 | for (i = 0; i < nr_pages; i++) { | |
1242 | struct page *page = pvec.pages[i]; | |
1243 | ||
1244 | /* until radix tree lookup accepts end_index */ | |
cfb271d4 | 1245 | if (unlikely(page->index > end)) |
cfe58f9d JK |
1246 | continue; |
1247 | ||
4bf08ff6 | 1248 | if (ino && ino_of_node(page) == ino) { |
3cb5ad15 | 1249 | f2fs_wait_on_page_writeback(page, NODE); |
4bf08ff6 CY |
1250 | if (TestClearPageError(page)) |
1251 | ret = -EIO; | |
1252 | } | |
cfe58f9d JK |
1253 | } |
1254 | pagevec_release(&pvec); | |
1255 | cond_resched(); | |
1256 | } | |
1257 | ||
4ef51a8f | 1258 | if (unlikely(test_and_clear_bit(AS_ENOSPC, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d | 1259 | ret2 = -ENOSPC; |
4ef51a8f | 1260 | if (unlikely(test_and_clear_bit(AS_EIO, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d JK |
1261 | ret2 = -EIO; |
1262 | if (!ret) | |
1263 | ret = ret2; | |
1264 | return ret; | |
1265 | } | |
1266 | ||
e05df3b1 JK |
1267 | static int f2fs_write_node_page(struct page *page, |
1268 | struct writeback_control *wbc) | |
1269 | { | |
4081363f | 1270 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1271 | nid_t nid; |
e05df3b1 JK |
1272 | block_t new_addr; |
1273 | struct node_info ni; | |
fb5566da JK |
1274 | struct f2fs_io_info fio = { |
1275 | .type = NODE, | |
6c311ec6 | 1276 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
fb5566da | 1277 | }; |
e05df3b1 | 1278 | |
ecda0de3 CY |
1279 | trace_f2fs_writepage(page, NODE); |
1280 | ||
cfb271d4 | 1281 | if (unlikely(sbi->por_doing)) |
87a9bd26 | 1282 | goto redirty_out; |
cf779cab JK |
1283 | if (unlikely(f2fs_cp_error(sbi))) |
1284 | goto redirty_out; | |
87a9bd26 | 1285 | |
3cb5ad15 | 1286 | f2fs_wait_on_page_writeback(page, NODE); |
e05df3b1 | 1287 | |
e05df3b1 JK |
1288 | /* get old block addr of this node page */ |
1289 | nid = nid_of_node(page); | |
9850cf4a | 1290 | f2fs_bug_on(sbi, page->index != nid); |
e05df3b1 JK |
1291 | |
1292 | get_node_info(sbi, nid, &ni); | |
1293 | ||
1294 | /* This page is already truncated */ | |
6bacf52f | 1295 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
39936837 JK |
1296 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
1297 | unlock_page(page); | |
1298 | return 0; | |
1299 | } | |
e05df3b1 | 1300 | |
2f97c326 JK |
1301 | if (wbc->for_reclaim) { |
1302 | if (!down_read_trylock(&sbi->node_write)) | |
1303 | goto redirty_out; | |
1304 | } else { | |
1305 | down_read(&sbi->node_write); | |
1306 | } | |
e05df3b1 | 1307 | set_page_writeback(page); |
fb5566da | 1308 | write_node_page(sbi, page, &fio, nid, ni.blk_addr, &new_addr); |
479f40c4 | 1309 | set_node_addr(sbi, &ni, new_addr, is_fsync_dnode(page)); |
e05df3b1 | 1310 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
b3582c68 | 1311 | up_read(&sbi->node_write); |
e05df3b1 | 1312 | unlock_page(page); |
27c6bd60 JK |
1313 | |
1314 | if (wbc->for_reclaim) | |
1315 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1316 | ||
e05df3b1 | 1317 | return 0; |
87a9bd26 JK |
1318 | |
1319 | redirty_out: | |
76f60268 | 1320 | redirty_page_for_writepage(wbc, page); |
87a9bd26 | 1321 | return AOP_WRITEPAGE_ACTIVATE; |
e05df3b1 JK |
1322 | } |
1323 | ||
1324 | static int f2fs_write_node_pages(struct address_space *mapping, | |
1325 | struct writeback_control *wbc) | |
1326 | { | |
4081363f | 1327 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 1328 | long diff; |
e05df3b1 | 1329 | |
e5748434 CY |
1330 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
1331 | ||
4660f9c0 JK |
1332 | /* balancing f2fs's metadata in background */ |
1333 | f2fs_balance_fs_bg(sbi); | |
e05df3b1 | 1334 | |
a7fdffbd | 1335 | /* collect a number of dirty node pages and write together */ |
87d6f890 | 1336 | if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE)) |
d3baf95d | 1337 | goto skip_write; |
a7fdffbd | 1338 | |
50c8cdb3 | 1339 | diff = nr_pages_to_write(sbi, NODE, wbc); |
fb5566da | 1340 | wbc->sync_mode = WB_SYNC_NONE; |
e05df3b1 | 1341 | sync_node_pages(sbi, 0, wbc); |
50c8cdb3 | 1342 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
e05df3b1 | 1343 | return 0; |
d3baf95d JK |
1344 | |
1345 | skip_write: | |
1346 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES); | |
1347 | return 0; | |
e05df3b1 JK |
1348 | } |
1349 | ||
1350 | static int f2fs_set_node_page_dirty(struct page *page) | |
1351 | { | |
26c6b887 JK |
1352 | trace_f2fs_set_page_dirty(page, NODE); |
1353 | ||
e05df3b1 JK |
1354 | SetPageUptodate(page); |
1355 | if (!PageDirty(page)) { | |
1356 | __set_page_dirty_nobuffers(page); | |
4081363f | 1357 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 JK |
1358 | SetPagePrivate(page); |
1359 | return 1; | |
1360 | } | |
1361 | return 0; | |
1362 | } | |
1363 | ||
d47992f8 LC |
1364 | static void f2fs_invalidate_node_page(struct page *page, unsigned int offset, |
1365 | unsigned int length) | |
e05df3b1 JK |
1366 | { |
1367 | struct inode *inode = page->mapping->host; | |
e05df3b1 | 1368 | if (PageDirty(page)) |
4081363f | 1369 | dec_page_count(F2FS_I_SB(inode), F2FS_DIRTY_NODES); |
e05df3b1 JK |
1370 | ClearPagePrivate(page); |
1371 | } | |
1372 | ||
1373 | static int f2fs_release_node_page(struct page *page, gfp_t wait) | |
1374 | { | |
1375 | ClearPagePrivate(page); | |
c3850aa1 | 1376 | return 1; |
e05df3b1 JK |
1377 | } |
1378 | ||
0a8165d7 | 1379 | /* |
e05df3b1 JK |
1380 | * Structure of the f2fs node operations |
1381 | */ | |
1382 | const struct address_space_operations f2fs_node_aops = { | |
1383 | .writepage = f2fs_write_node_page, | |
1384 | .writepages = f2fs_write_node_pages, | |
1385 | .set_page_dirty = f2fs_set_node_page_dirty, | |
1386 | .invalidatepage = f2fs_invalidate_node_page, | |
1387 | .releasepage = f2fs_release_node_page, | |
1388 | }; | |
1389 | ||
8a7ed66a JK |
1390 | static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i, |
1391 | nid_t n) | |
e05df3b1 | 1392 | { |
8a7ed66a | 1393 | return radix_tree_lookup(&nm_i->free_nid_root, n); |
e05df3b1 JK |
1394 | } |
1395 | ||
8a7ed66a JK |
1396 | static void __del_from_free_nid_list(struct f2fs_nm_info *nm_i, |
1397 | struct free_nid *i) | |
e05df3b1 JK |
1398 | { |
1399 | list_del(&i->list); | |
8a7ed66a | 1400 | radix_tree_delete(&nm_i->free_nid_root, i->nid); |
e05df3b1 JK |
1401 | } |
1402 | ||
6fb03f3a | 1403 | static int add_free_nid(struct f2fs_sb_info *sbi, nid_t nid, bool build) |
e05df3b1 | 1404 | { |
6fb03f3a | 1405 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1406 | struct free_nid *i; |
59bbd474 JK |
1407 | struct nat_entry *ne; |
1408 | bool allocated = false; | |
e05df3b1 | 1409 | |
6fb03f3a | 1410 | if (!available_free_memory(sbi, FREE_NIDS)) |
23d38844 | 1411 | return -1; |
9198aceb JK |
1412 | |
1413 | /* 0 nid should not be used */ | |
cfb271d4 | 1414 | if (unlikely(nid == 0)) |
9198aceb | 1415 | return 0; |
59bbd474 | 1416 | |
7bd59381 GZ |
1417 | if (build) { |
1418 | /* do not add allocated nids */ | |
8b26ef98 | 1419 | down_read(&nm_i->nat_tree_lock); |
7bd59381 | 1420 | ne = __lookup_nat_cache(nm_i, nid); |
8a7ed66a | 1421 | if (ne && |
7ef35e3b JK |
1422 | (!get_nat_flag(ne, IS_CHECKPOINTED) || |
1423 | nat_get_blkaddr(ne) != NULL_ADDR)) | |
7bd59381 | 1424 | allocated = true; |
8b26ef98 | 1425 | up_read(&nm_i->nat_tree_lock); |
7bd59381 GZ |
1426 | if (allocated) |
1427 | return 0; | |
e05df3b1 | 1428 | } |
7bd59381 GZ |
1429 | |
1430 | i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); | |
e05df3b1 JK |
1431 | i->nid = nid; |
1432 | i->state = NID_NEW; | |
1433 | ||
769ec6e5 JK |
1434 | if (radix_tree_preload(GFP_NOFS)) { |
1435 | kmem_cache_free(free_nid_slab, i); | |
1436 | return 0; | |
1437 | } | |
1438 | ||
e05df3b1 | 1439 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1440 | if (radix_tree_insert(&nm_i->free_nid_root, i->nid, i)) { |
e05df3b1 | 1441 | spin_unlock(&nm_i->free_nid_list_lock); |
769ec6e5 | 1442 | radix_tree_preload_end(); |
e05df3b1 JK |
1443 | kmem_cache_free(free_nid_slab, i); |
1444 | return 0; | |
1445 | } | |
1446 | list_add_tail(&i->list, &nm_i->free_nid_list); | |
1447 | nm_i->fcnt++; | |
1448 | spin_unlock(&nm_i->free_nid_list_lock); | |
769ec6e5 | 1449 | radix_tree_preload_end(); |
e05df3b1 JK |
1450 | return 1; |
1451 | } | |
1452 | ||
1453 | static void remove_free_nid(struct f2fs_nm_info *nm_i, nid_t nid) | |
1454 | { | |
1455 | struct free_nid *i; | |
cf0ee0f0 CY |
1456 | bool need_free = false; |
1457 | ||
e05df3b1 | 1458 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1459 | i = __lookup_free_nid_list(nm_i, nid); |
e05df3b1 | 1460 | if (i && i->state == NID_NEW) { |
8a7ed66a | 1461 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1462 | nm_i->fcnt--; |
cf0ee0f0 | 1463 | need_free = true; |
e05df3b1 JK |
1464 | } |
1465 | spin_unlock(&nm_i->free_nid_list_lock); | |
cf0ee0f0 CY |
1466 | |
1467 | if (need_free) | |
1468 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1469 | } |
1470 | ||
6fb03f3a | 1471 | static void scan_nat_page(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
1472 | struct page *nat_page, nid_t start_nid) |
1473 | { | |
6fb03f3a | 1474 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 JK |
1475 | struct f2fs_nat_block *nat_blk = page_address(nat_page); |
1476 | block_t blk_addr; | |
e05df3b1 JK |
1477 | int i; |
1478 | ||
e05df3b1 JK |
1479 | i = start_nid % NAT_ENTRY_PER_BLOCK; |
1480 | ||
1481 | for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { | |
23d38844 | 1482 | |
cfb271d4 | 1483 | if (unlikely(start_nid >= nm_i->max_nid)) |
04431c44 | 1484 | break; |
23d38844 HL |
1485 | |
1486 | blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); | |
9850cf4a | 1487 | f2fs_bug_on(sbi, blk_addr == NEW_ADDR); |
23d38844 | 1488 | if (blk_addr == NULL_ADDR) { |
6fb03f3a | 1489 | if (add_free_nid(sbi, start_nid, true) < 0) |
23d38844 HL |
1490 | break; |
1491 | } | |
e05df3b1 | 1492 | } |
e05df3b1 JK |
1493 | } |
1494 | ||
1495 | static void build_free_nids(struct f2fs_sb_info *sbi) | |
1496 | { | |
e05df3b1 JK |
1497 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1498 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1499 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
8760952d | 1500 | int i = 0; |
55008d84 | 1501 | nid_t nid = nm_i->next_scan_nid; |
e05df3b1 | 1502 | |
55008d84 JK |
1503 | /* Enough entries */ |
1504 | if (nm_i->fcnt > NAT_ENTRY_PER_BLOCK) | |
1505 | return; | |
e05df3b1 | 1506 | |
55008d84 | 1507 | /* readahead nat pages to be scanned */ |
662befda | 1508 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES, META_NAT); |
e05df3b1 JK |
1509 | |
1510 | while (1) { | |
1511 | struct page *page = get_current_nat_page(sbi, nid); | |
1512 | ||
6fb03f3a | 1513 | scan_nat_page(sbi, page, nid); |
e05df3b1 JK |
1514 | f2fs_put_page(page, 1); |
1515 | ||
1516 | nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); | |
cfb271d4 | 1517 | if (unlikely(nid >= nm_i->max_nid)) |
e05df3b1 | 1518 | nid = 0; |
55008d84 JK |
1519 | |
1520 | if (i++ == FREE_NID_PAGES) | |
e05df3b1 JK |
1521 | break; |
1522 | } | |
1523 | ||
55008d84 JK |
1524 | /* go to the next free nat pages to find free nids abundantly */ |
1525 | nm_i->next_scan_nid = nid; | |
e05df3b1 JK |
1526 | |
1527 | /* find free nids from current sum_pages */ | |
1528 | mutex_lock(&curseg->curseg_mutex); | |
1529 | for (i = 0; i < nats_in_cursum(sum); i++) { | |
1530 | block_t addr = le32_to_cpu(nat_in_journal(sum, i).block_addr); | |
1531 | nid = le32_to_cpu(nid_in_journal(sum, i)); | |
1532 | if (addr == NULL_ADDR) | |
6fb03f3a | 1533 | add_free_nid(sbi, nid, true); |
e05df3b1 JK |
1534 | else |
1535 | remove_free_nid(nm_i, nid); | |
1536 | } | |
1537 | mutex_unlock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1538 | } |
1539 | ||
1540 | /* | |
1541 | * If this function returns success, caller can obtain a new nid | |
1542 | * from second parameter of this function. | |
1543 | * The returned nid could be used ino as well as nid when inode is created. | |
1544 | */ | |
1545 | bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) | |
1546 | { | |
1547 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1548 | struct free_nid *i = NULL; | |
e05df3b1 | 1549 | retry: |
7ee0eeab | 1550 | if (unlikely(sbi->total_valid_node_count + 1 > nm_i->available_nids)) |
55008d84 | 1551 | return false; |
e05df3b1 | 1552 | |
e05df3b1 | 1553 | spin_lock(&nm_i->free_nid_list_lock); |
e05df3b1 | 1554 | |
55008d84 | 1555 | /* We should not use stale free nids created by build_free_nids */ |
f978f5a0 | 1556 | if (nm_i->fcnt && !on_build_free_nids(nm_i)) { |
9850cf4a | 1557 | f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list)); |
2d7b822a | 1558 | list_for_each_entry(i, &nm_i->free_nid_list, list) |
55008d84 JK |
1559 | if (i->state == NID_NEW) |
1560 | break; | |
e05df3b1 | 1561 | |
9850cf4a | 1562 | f2fs_bug_on(sbi, i->state != NID_NEW); |
55008d84 JK |
1563 | *nid = i->nid; |
1564 | i->state = NID_ALLOC; | |
1565 | nm_i->fcnt--; | |
1566 | spin_unlock(&nm_i->free_nid_list_lock); | |
1567 | return true; | |
1568 | } | |
e05df3b1 | 1569 | spin_unlock(&nm_i->free_nid_list_lock); |
55008d84 JK |
1570 | |
1571 | /* Let's scan nat pages and its caches to get free nids */ | |
1572 | mutex_lock(&nm_i->build_lock); | |
55008d84 | 1573 | build_free_nids(sbi); |
55008d84 JK |
1574 | mutex_unlock(&nm_i->build_lock); |
1575 | goto retry; | |
e05df3b1 JK |
1576 | } |
1577 | ||
0a8165d7 | 1578 | /* |
e05df3b1 JK |
1579 | * alloc_nid() should be called prior to this function. |
1580 | */ | |
1581 | void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) | |
1582 | { | |
1583 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1584 | struct free_nid *i; | |
1585 | ||
1586 | spin_lock(&nm_i->free_nid_list_lock); | |
8a7ed66a | 1587 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1588 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
8a7ed66a | 1589 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1590 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1591 | |
1592 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1593 | } |
1594 | ||
0a8165d7 | 1595 | /* |
e05df3b1 JK |
1596 | * alloc_nid() should be called prior to this function. |
1597 | */ | |
1598 | void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) | |
1599 | { | |
49952fa1 JK |
1600 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1601 | struct free_nid *i; | |
cf0ee0f0 | 1602 | bool need_free = false; |
49952fa1 | 1603 | |
65985d93 JK |
1604 | if (!nid) |
1605 | return; | |
1606 | ||
49952fa1 | 1607 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1608 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1609 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
6fb03f3a | 1610 | if (!available_free_memory(sbi, FREE_NIDS)) { |
8a7ed66a | 1611 | __del_from_free_nid_list(nm_i, i); |
cf0ee0f0 | 1612 | need_free = true; |
95630cba HL |
1613 | } else { |
1614 | i->state = NID_NEW; | |
1615 | nm_i->fcnt++; | |
1616 | } | |
49952fa1 | 1617 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1618 | |
1619 | if (need_free) | |
1620 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1621 | } |
1622 | ||
70cfed88 | 1623 | void recover_inline_xattr(struct inode *inode, struct page *page) |
28cdce04 | 1624 | { |
28cdce04 CY |
1625 | void *src_addr, *dst_addr; |
1626 | size_t inline_size; | |
1627 | struct page *ipage; | |
1628 | struct f2fs_inode *ri; | |
1629 | ||
4081363f | 1630 | ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino); |
9850cf4a | 1631 | f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage)); |
28cdce04 | 1632 | |
e3b4d43f JK |
1633 | ri = F2FS_INODE(page); |
1634 | if (!(ri->i_inline & F2FS_INLINE_XATTR)) { | |
1635 | clear_inode_flag(F2FS_I(inode), FI_INLINE_XATTR); | |
1636 | goto update_inode; | |
1637 | } | |
1638 | ||
28cdce04 CY |
1639 | dst_addr = inline_xattr_addr(ipage); |
1640 | src_addr = inline_xattr_addr(page); | |
1641 | inline_size = inline_xattr_size(inode); | |
1642 | ||
54b591df | 1643 | f2fs_wait_on_page_writeback(ipage, NODE); |
28cdce04 | 1644 | memcpy(dst_addr, src_addr, inline_size); |
e3b4d43f | 1645 | update_inode: |
28cdce04 CY |
1646 | update_inode(inode, ipage); |
1647 | f2fs_put_page(ipage, 1); | |
1648 | } | |
1649 | ||
1c35a90e | 1650 | void recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr) |
abb2366c | 1651 | { |
4081363f | 1652 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
abb2366c JK |
1653 | nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid; |
1654 | nid_t new_xnid = nid_of_node(page); | |
1655 | struct node_info ni; | |
1656 | ||
abb2366c JK |
1657 | /* 1: invalidate the previous xattr nid */ |
1658 | if (!prev_xnid) | |
1659 | goto recover_xnid; | |
1660 | ||
1661 | /* Deallocate node address */ | |
1662 | get_node_info(sbi, prev_xnid, &ni); | |
9850cf4a | 1663 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
abb2366c JK |
1664 | invalidate_blocks(sbi, ni.blk_addr); |
1665 | dec_valid_node_count(sbi, inode); | |
479f40c4 | 1666 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
abb2366c JK |
1667 | |
1668 | recover_xnid: | |
1669 | /* 2: allocate new xattr nid */ | |
1670 | if (unlikely(!inc_valid_node_count(sbi, inode))) | |
9850cf4a | 1671 | f2fs_bug_on(sbi, 1); |
abb2366c JK |
1672 | |
1673 | remove_free_nid(NM_I(sbi), new_xnid); | |
1674 | get_node_info(sbi, new_xnid, &ni); | |
1675 | ni.ino = inode->i_ino; | |
479f40c4 | 1676 | set_node_addr(sbi, &ni, NEW_ADDR, false); |
abb2366c JK |
1677 | F2FS_I(inode)->i_xattr_nid = new_xnid; |
1678 | ||
1679 | /* 3: update xattr blkaddr */ | |
1680 | refresh_sit_entry(sbi, NEW_ADDR, blkaddr); | |
479f40c4 | 1681 | set_node_addr(sbi, &ni, blkaddr, false); |
abb2366c JK |
1682 | |
1683 | update_inode_page(inode); | |
abb2366c JK |
1684 | } |
1685 | ||
e05df3b1 JK |
1686 | int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) |
1687 | { | |
58bfaf44 | 1688 | struct f2fs_inode *src, *dst; |
e05df3b1 JK |
1689 | nid_t ino = ino_of_node(page); |
1690 | struct node_info old_ni, new_ni; | |
1691 | struct page *ipage; | |
1692 | ||
e8271fa3 JK |
1693 | get_node_info(sbi, ino, &old_ni); |
1694 | ||
1695 | if (unlikely(old_ni.blk_addr != NULL_ADDR)) | |
1696 | return -EINVAL; | |
1697 | ||
4ef51a8f | 1698 | ipage = grab_cache_page(NODE_MAPPING(sbi), ino); |
e05df3b1 JK |
1699 | if (!ipage) |
1700 | return -ENOMEM; | |
1701 | ||
e1c42045 | 1702 | /* Should not use this inode from free nid list */ |
e05df3b1 JK |
1703 | remove_free_nid(NM_I(sbi), ino); |
1704 | ||
e05df3b1 JK |
1705 | SetPageUptodate(ipage); |
1706 | fill_node_footer(ipage, ino, ino, 0, true); | |
1707 | ||
58bfaf44 JK |
1708 | src = F2FS_INODE(page); |
1709 | dst = F2FS_INODE(ipage); | |
e05df3b1 | 1710 | |
58bfaf44 JK |
1711 | memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); |
1712 | dst->i_size = 0; | |
1713 | dst->i_blocks = cpu_to_le64(1); | |
1714 | dst->i_links = cpu_to_le32(1); | |
1715 | dst->i_xattr_nid = 0; | |
617deb8c | 1716 | dst->i_inline = src->i_inline & F2FS_INLINE_XATTR; |
e05df3b1 JK |
1717 | |
1718 | new_ni = old_ni; | |
1719 | new_ni.ino = ino; | |
1720 | ||
cfb271d4 | 1721 | if (unlikely(!inc_valid_node_count(sbi, NULL))) |
65e5cd0a | 1722 | WARN_ON(1); |
479f40c4 | 1723 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
e05df3b1 | 1724 | inc_valid_inode_count(sbi); |
617deb8c | 1725 | set_page_dirty(ipage); |
e05df3b1 JK |
1726 | f2fs_put_page(ipage, 1); |
1727 | return 0; | |
1728 | } | |
1729 | ||
9af0ff1c CY |
1730 | /* |
1731 | * ra_sum_pages() merge contiguous pages into one bio and submit. | |
e1c42045 | 1732 | * these pre-read pages are allocated in bd_inode's mapping tree. |
9af0ff1c | 1733 | */ |
bac4eef6 | 1734 | static int ra_sum_pages(struct f2fs_sb_info *sbi, struct page **pages, |
9af0ff1c CY |
1735 | int start, int nrpages) |
1736 | { | |
bac4eef6 CY |
1737 | struct inode *inode = sbi->sb->s_bdev->bd_inode; |
1738 | struct address_space *mapping = inode->i_mapping; | |
1739 | int i, page_idx = start; | |
458e6197 JK |
1740 | struct f2fs_io_info fio = { |
1741 | .type = META, | |
7e8f2308 | 1742 | .rw = READ_SYNC | REQ_META | REQ_PRIO |
458e6197 | 1743 | }; |
9af0ff1c | 1744 | |
bac4eef6 CY |
1745 | for (i = 0; page_idx < start + nrpages; page_idx++, i++) { |
1746 | /* alloc page in bd_inode for reading node summary info */ | |
1747 | pages[i] = grab_cache_page(mapping, page_idx); | |
1748 | if (!pages[i]) | |
d653788a | 1749 | break; |
bac4eef6 | 1750 | f2fs_submit_page_mbio(sbi, pages[i], page_idx, &fio); |
9af0ff1c CY |
1751 | } |
1752 | ||
458e6197 | 1753 | f2fs_submit_merged_bio(sbi, META, READ); |
bac4eef6 | 1754 | return i; |
9af0ff1c CY |
1755 | } |
1756 | ||
e05df3b1 JK |
1757 | int restore_node_summary(struct f2fs_sb_info *sbi, |
1758 | unsigned int segno, struct f2fs_summary_block *sum) | |
1759 | { | |
1760 | struct f2fs_node *rn; | |
1761 | struct f2fs_summary *sum_entry; | |
bac4eef6 | 1762 | struct inode *inode = sbi->sb->s_bdev->bd_inode; |
e05df3b1 | 1763 | block_t addr; |
90a893c7 | 1764 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
bac4eef6 CY |
1765 | struct page *pages[bio_blocks]; |
1766 | int i, idx, last_offset, nrpages, err = 0; | |
e05df3b1 JK |
1767 | |
1768 | /* scan the node segment */ | |
1769 | last_offset = sbi->blocks_per_seg; | |
1770 | addr = START_BLOCK(sbi, segno); | |
1771 | sum_entry = &sum->entries[0]; | |
1772 | ||
d653788a | 1773 | for (i = 0; !err && i < last_offset; i += nrpages, addr += nrpages) { |
9af0ff1c | 1774 | nrpages = min(last_offset - i, bio_blocks); |
393ff91f | 1775 | |
e1c42045 | 1776 | /* readahead node pages */ |
bac4eef6 | 1777 | nrpages = ra_sum_pages(sbi, pages, addr, nrpages); |
d653788a GZ |
1778 | if (!nrpages) |
1779 | return -ENOMEM; | |
e05df3b1 | 1780 | |
bac4eef6 | 1781 | for (idx = 0; idx < nrpages; idx++) { |
d653788a GZ |
1782 | if (err) |
1783 | goto skip; | |
9af0ff1c | 1784 | |
bac4eef6 CY |
1785 | lock_page(pages[idx]); |
1786 | if (unlikely(!PageUptodate(pages[idx]))) { | |
6bacf52f JK |
1787 | err = -EIO; |
1788 | } else { | |
bac4eef6 | 1789 | rn = F2FS_NODE(pages[idx]); |
9af0ff1c CY |
1790 | sum_entry->nid = rn->footer.nid; |
1791 | sum_entry->version = 0; | |
1792 | sum_entry->ofs_in_node = 0; | |
1793 | sum_entry++; | |
9af0ff1c | 1794 | } |
bac4eef6 | 1795 | unlock_page(pages[idx]); |
d653788a | 1796 | skip: |
bac4eef6 | 1797 | page_cache_release(pages[idx]); |
9af0ff1c | 1798 | } |
bac4eef6 CY |
1799 | |
1800 | invalidate_mapping_pages(inode->i_mapping, addr, | |
1801 | addr + nrpages); | |
e05df3b1 | 1802 | } |
9af0ff1c | 1803 | return err; |
e05df3b1 JK |
1804 | } |
1805 | ||
aec71382 | 1806 | static void remove_nats_in_journal(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
1807 | { |
1808 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1809 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1810 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1811 | int i; | |
1812 | ||
1813 | mutex_lock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1814 | for (i = 0; i < nats_in_cursum(sum); i++) { |
1815 | struct nat_entry *ne; | |
1816 | struct f2fs_nat_entry raw_ne; | |
1817 | nid_t nid = le32_to_cpu(nid_in_journal(sum, i)); | |
1818 | ||
1819 | raw_ne = nat_in_journal(sum, i); | |
9be32d72 | 1820 | |
8b26ef98 | 1821 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 | 1822 | ne = __lookup_nat_cache(nm_i, nid); |
e05df3b1 | 1823 | if (!ne) { |
9be32d72 JK |
1824 | ne = grab_nat_entry(nm_i, nid); |
1825 | node_info_from_raw_nat(&ne->ni, &raw_ne); | |
e05df3b1 | 1826 | } |
e05df3b1 | 1827 | __set_nat_cache_dirty(nm_i, ne); |
8b26ef98 | 1828 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
1829 | } |
1830 | update_nats_in_cursum(sum, -i); | |
1831 | mutex_unlock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1832 | } |
1833 | ||
309cc2b6 JK |
1834 | static void __adjust_nat_entry_set(struct nat_entry_set *nes, |
1835 | struct list_head *head, int max) | |
e05df3b1 | 1836 | { |
309cc2b6 | 1837 | struct nat_entry_set *cur; |
e05df3b1 | 1838 | |
309cc2b6 JK |
1839 | if (nes->entry_cnt >= max) |
1840 | goto add_out; | |
e05df3b1 | 1841 | |
309cc2b6 JK |
1842 | list_for_each_entry(cur, head, set_list) { |
1843 | if (cur->entry_cnt >= nes->entry_cnt) { | |
1844 | list_add(&nes->set_list, cur->set_list.prev); | |
1845 | return; | |
1846 | } | |
aec71382 | 1847 | } |
309cc2b6 JK |
1848 | add_out: |
1849 | list_add_tail(&nes->set_list, head); | |
1850 | } | |
e05df3b1 | 1851 | |
309cc2b6 JK |
1852 | static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, |
1853 | struct nat_entry_set *set) | |
1854 | { | |
1855 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1856 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1857 | nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK; | |
1858 | bool to_journal = true; | |
1859 | struct f2fs_nat_block *nat_blk; | |
1860 | struct nat_entry *ne, *cur; | |
1861 | struct page *page = NULL; | |
e05df3b1 | 1862 | |
aec71382 CY |
1863 | /* |
1864 | * there are two steps to flush nat entries: | |
1865 | * #1, flush nat entries to journal in current hot data summary block. | |
1866 | * #2, flush nat entries to nat page. | |
1867 | */ | |
309cc2b6 JK |
1868 | if (!__has_cursum_space(sum, set->entry_cnt, NAT_JOURNAL)) |
1869 | to_journal = false; | |
1870 | ||
1871 | if (to_journal) { | |
1872 | mutex_lock(&curseg->curseg_mutex); | |
1873 | } else { | |
1874 | page = get_next_nat_page(sbi, start_nid); | |
1875 | nat_blk = page_address(page); | |
1876 | f2fs_bug_on(sbi, !nat_blk); | |
1877 | } | |
aec71382 | 1878 | |
309cc2b6 JK |
1879 | /* flush dirty nats in nat entry set */ |
1880 | list_for_each_entry_safe(ne, cur, &set->entry_list, list) { | |
1881 | struct f2fs_nat_entry *raw_ne; | |
1882 | nid_t nid = nat_get_nid(ne); | |
1883 | int offset; | |
1884 | ||
1885 | if (nat_get_blkaddr(ne) == NEW_ADDR) | |
1886 | continue; | |
aec71382 CY |
1887 | |
1888 | if (to_journal) { | |
309cc2b6 JK |
1889 | offset = lookup_journal_in_cursum(sum, |
1890 | NAT_JOURNAL, nid, 1); | |
1891 | f2fs_bug_on(sbi, offset < 0); | |
1892 | raw_ne = &nat_in_journal(sum, offset); | |
1893 | nid_in_journal(sum, offset) = cpu_to_le32(nid); | |
aec71382 | 1894 | } else { |
309cc2b6 | 1895 | raw_ne = &nat_blk->entries[nid - start_nid]; |
e05df3b1 | 1896 | } |
309cc2b6 | 1897 | raw_nat_from_node_info(raw_ne, &ne->ni); |
e05df3b1 | 1898 | |
8b26ef98 | 1899 | down_write(&NM_I(sbi)->nat_tree_lock); |
309cc2b6 JK |
1900 | nat_reset_flag(ne); |
1901 | __clear_nat_cache_dirty(NM_I(sbi), ne); | |
8b26ef98 | 1902 | up_write(&NM_I(sbi)->nat_tree_lock); |
aec71382 | 1903 | |
309cc2b6 JK |
1904 | if (nat_get_blkaddr(ne) == NULL_ADDR) |
1905 | add_free_nid(sbi, nid, false); | |
1906 | } | |
e05df3b1 | 1907 | |
309cc2b6 JK |
1908 | if (to_journal) |
1909 | mutex_unlock(&curseg->curseg_mutex); | |
1910 | else | |
1911 | f2fs_put_page(page, 1); | |
aec71382 | 1912 | |
80ec2e91 CL |
1913 | f2fs_bug_on(sbi, set->entry_cnt); |
1914 | ||
1915 | radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); | |
1916 | kmem_cache_free(nat_entry_set_slab, set); | |
309cc2b6 | 1917 | } |
aec71382 | 1918 | |
309cc2b6 JK |
1919 | /* |
1920 | * This function is called during the checkpointing process. | |
1921 | */ | |
1922 | void flush_nat_entries(struct f2fs_sb_info *sbi) | |
1923 | { | |
1924 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1925 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1926 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1927 | struct nat_entry_set *setvec[NATVEC_SIZE]; | |
1928 | struct nat_entry_set *set, *tmp; | |
1929 | unsigned int found; | |
1930 | nid_t set_idx = 0; | |
1931 | LIST_HEAD(sets); | |
1932 | ||
20d047c8 CL |
1933 | if (!nm_i->dirty_nat_cnt) |
1934 | return; | |
309cc2b6 JK |
1935 | /* |
1936 | * if there are no enough space in journal to store dirty nat | |
1937 | * entries, remove all entries from journal and merge them | |
1938 | * into nat entry set. | |
1939 | */ | |
1940 | if (!__has_cursum_space(sum, nm_i->dirty_nat_cnt, NAT_JOURNAL)) | |
1941 | remove_nats_in_journal(sbi); | |
1942 | ||
309cc2b6 JK |
1943 | while ((found = __gang_lookup_nat_set(nm_i, |
1944 | set_idx, NATVEC_SIZE, setvec))) { | |
1945 | unsigned idx; | |
1946 | set_idx = setvec[found - 1]->set + 1; | |
1947 | for (idx = 0; idx < found; idx++) | |
1948 | __adjust_nat_entry_set(setvec[idx], &sets, | |
1949 | MAX_NAT_JENTRIES(sum)); | |
e05df3b1 | 1950 | } |
aec71382 | 1951 | |
309cc2b6 JK |
1952 | /* flush dirty nats in nat entry set */ |
1953 | list_for_each_entry_safe(set, tmp, &sets, set_list) | |
1954 | __flush_nat_entry_set(sbi, set); | |
1955 | ||
9850cf4a | 1956 | f2fs_bug_on(sbi, nm_i->dirty_nat_cnt); |
e05df3b1 JK |
1957 | } |
1958 | ||
1959 | static int init_node_manager(struct f2fs_sb_info *sbi) | |
1960 | { | |
1961 | struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); | |
1962 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1963 | unsigned char *version_bitmap; | |
1964 | unsigned int nat_segs, nat_blocks; | |
1965 | ||
1966 | nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); | |
1967 | ||
1968 | /* segment_count_nat includes pair segment so divide to 2. */ | |
1969 | nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; | |
1970 | nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); | |
b63da15e | 1971 | |
7ee0eeab JK |
1972 | nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks; |
1973 | ||
b63da15e | 1974 | /* not used nids: 0, node, meta, (and root counted as valid node) */ |
c200b1aa | 1975 | nm_i->available_nids = nm_i->max_nid - F2FS_RESERVED_NODE_NUM; |
e05df3b1 JK |
1976 | nm_i->fcnt = 0; |
1977 | nm_i->nat_cnt = 0; | |
cdfc41c1 | 1978 | nm_i->ram_thresh = DEF_RAM_THRESHOLD; |
e05df3b1 | 1979 | |
8a7ed66a | 1980 | INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC); |
e05df3b1 | 1981 | INIT_LIST_HEAD(&nm_i->free_nid_list); |
769ec6e5 JK |
1982 | INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO); |
1983 | INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO); | |
e05df3b1 | 1984 | INIT_LIST_HEAD(&nm_i->nat_entries); |
e05df3b1 JK |
1985 | |
1986 | mutex_init(&nm_i->build_lock); | |
1987 | spin_lock_init(&nm_i->free_nid_list_lock); | |
8b26ef98 | 1988 | init_rwsem(&nm_i->nat_tree_lock); |
e05df3b1 | 1989 | |
e05df3b1 | 1990 | nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); |
79b5793b | 1991 | nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); |
e05df3b1 JK |
1992 | version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); |
1993 | if (!version_bitmap) | |
1994 | return -EFAULT; | |
1995 | ||
79b5793b AG |
1996 | nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, |
1997 | GFP_KERNEL); | |
1998 | if (!nm_i->nat_bitmap) | |
1999 | return -ENOMEM; | |
e05df3b1 JK |
2000 | return 0; |
2001 | } | |
2002 | ||
2003 | int build_node_manager(struct f2fs_sb_info *sbi) | |
2004 | { | |
2005 | int err; | |
2006 | ||
2007 | sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL); | |
2008 | if (!sbi->nm_info) | |
2009 | return -ENOMEM; | |
2010 | ||
2011 | err = init_node_manager(sbi); | |
2012 | if (err) | |
2013 | return err; | |
2014 | ||
2015 | build_free_nids(sbi); | |
2016 | return 0; | |
2017 | } | |
2018 | ||
2019 | void destroy_node_manager(struct f2fs_sb_info *sbi) | |
2020 | { | |
2021 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2022 | struct free_nid *i, *next_i; | |
2023 | struct nat_entry *natvec[NATVEC_SIZE]; | |
2024 | nid_t nid = 0; | |
2025 | unsigned int found; | |
2026 | ||
2027 | if (!nm_i) | |
2028 | return; | |
2029 | ||
2030 | /* destroy free nid list */ | |
2031 | spin_lock(&nm_i->free_nid_list_lock); | |
2032 | list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) { | |
9850cf4a | 2033 | f2fs_bug_on(sbi, i->state == NID_ALLOC); |
8a7ed66a | 2034 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 2035 | nm_i->fcnt--; |
cf0ee0f0 CY |
2036 | spin_unlock(&nm_i->free_nid_list_lock); |
2037 | kmem_cache_free(free_nid_slab, i); | |
2038 | spin_lock(&nm_i->free_nid_list_lock); | |
e05df3b1 | 2039 | } |
9850cf4a | 2040 | f2fs_bug_on(sbi, nm_i->fcnt); |
e05df3b1 JK |
2041 | spin_unlock(&nm_i->free_nid_list_lock); |
2042 | ||
2043 | /* destroy nat cache */ | |
8b26ef98 | 2044 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2045 | while ((found = __gang_lookup_nat_cache(nm_i, |
2046 | nid, NATVEC_SIZE, natvec))) { | |
2047 | unsigned idx; | |
b6ce391e GZ |
2048 | nid = nat_get_nid(natvec[found - 1]) + 1; |
2049 | for (idx = 0; idx < found; idx++) | |
2050 | __del_from_nat_cache(nm_i, natvec[idx]); | |
e05df3b1 | 2051 | } |
9850cf4a | 2052 | f2fs_bug_on(sbi, nm_i->nat_cnt); |
8b26ef98 | 2053 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2054 | |
2055 | kfree(nm_i->nat_bitmap); | |
2056 | sbi->nm_info = NULL; | |
2057 | kfree(nm_i); | |
2058 | } | |
2059 | ||
6e6093a8 | 2060 | int __init create_node_manager_caches(void) |
e05df3b1 JK |
2061 | { |
2062 | nat_entry_slab = f2fs_kmem_cache_create("nat_entry", | |
e8512d2e | 2063 | sizeof(struct nat_entry)); |
e05df3b1 | 2064 | if (!nat_entry_slab) |
aec71382 | 2065 | goto fail; |
e05df3b1 JK |
2066 | |
2067 | free_nid_slab = f2fs_kmem_cache_create("free_nid", | |
e8512d2e | 2068 | sizeof(struct free_nid)); |
aec71382 | 2069 | if (!free_nid_slab) |
ce3e6d25 | 2070 | goto destroy_nat_entry; |
aec71382 CY |
2071 | |
2072 | nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set", | |
2073 | sizeof(struct nat_entry_set)); | |
2074 | if (!nat_entry_set_slab) | |
ce3e6d25 | 2075 | goto destroy_free_nid; |
e05df3b1 | 2076 | return 0; |
aec71382 | 2077 | |
ce3e6d25 | 2078 | destroy_free_nid: |
aec71382 | 2079 | kmem_cache_destroy(free_nid_slab); |
ce3e6d25 | 2080 | destroy_nat_entry: |
aec71382 CY |
2081 | kmem_cache_destroy(nat_entry_slab); |
2082 | fail: | |
2083 | return -ENOMEM; | |
e05df3b1 JK |
2084 | } |
2085 | ||
2086 | void destroy_node_manager_caches(void) | |
2087 | { | |
aec71382 | 2088 | kmem_cache_destroy(nat_entry_set_slab); |
e05df3b1 JK |
2089 | kmem_cache_destroy(free_nid_slab); |
2090 | kmem_cache_destroy(nat_entry_slab); | |
2091 | } |