]>
Commit | Line | Data |
---|---|---|
0a8165d7 | 1 | /* |
e05df3b1 JK |
2 | * fs/f2fs/node.c |
3 | * | |
4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
5 | * http://www.samsung.com/ | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
11 | #include <linux/fs.h> | |
12 | #include <linux/f2fs_fs.h> | |
13 | #include <linux/mpage.h> | |
14 | #include <linux/backing-dev.h> | |
15 | #include <linux/blkdev.h> | |
16 | #include <linux/pagevec.h> | |
17 | #include <linux/swap.h> | |
18 | ||
19 | #include "f2fs.h" | |
20 | #include "node.h" | |
21 | #include "segment.h" | |
9e4ded3f | 22 | #include "trace.h" |
51dd6249 | 23 | #include <trace/events/f2fs.h> |
e05df3b1 | 24 | |
f978f5a0 GZ |
25 | #define on_build_free_nids(nmi) mutex_is_locked(&nm_i->build_lock) |
26 | ||
e05df3b1 JK |
27 | static struct kmem_cache *nat_entry_slab; |
28 | static struct kmem_cache *free_nid_slab; | |
aec71382 | 29 | static struct kmem_cache *nat_entry_set_slab; |
e05df3b1 | 30 | |
6fb03f3a | 31 | bool available_free_memory(struct f2fs_sb_info *sbi, int type) |
cdfc41c1 | 32 | { |
6fb03f3a | 33 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
cdfc41c1 | 34 | struct sysinfo val; |
e5e7ea3c | 35 | unsigned long avail_ram; |
cdfc41c1 | 36 | unsigned long mem_size = 0; |
6fb03f3a | 37 | bool res = false; |
cdfc41c1 JK |
38 | |
39 | si_meminfo(&val); | |
e5e7ea3c JK |
40 | |
41 | /* only uses low memory */ | |
42 | avail_ram = val.totalram - val.totalhigh; | |
43 | ||
429511cd CY |
44 | /* |
45 | * give 25%, 25%, 50%, 50%, 50% memory for each components respectively | |
46 | */ | |
6fb03f3a | 47 | if (type == FREE_NIDS) { |
e5e7ea3c JK |
48 | mem_size = (nm_i->fcnt * sizeof(struct free_nid)) >> |
49 | PAGE_CACHE_SHIFT; | |
50 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); | |
6fb03f3a | 51 | } else if (type == NAT_ENTRIES) { |
e5e7ea3c JK |
52 | mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >> |
53 | PAGE_CACHE_SHIFT; | |
54 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2); | |
6fb03f3a | 55 | } else if (type == DIRTY_DENTS) { |
2743f865 JK |
56 | if (sbi->sb->s_bdi->dirty_exceeded) |
57 | return false; | |
6fb03f3a | 58 | mem_size = get_pages(sbi, F2FS_DIRTY_DENTS); |
e5e7ea3c JK |
59 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
60 | } else if (type == INO_ENTRIES) { | |
61 | int i; | |
62 | ||
e5e7ea3c | 63 | for (i = 0; i <= UPDATE_INO; i++) |
67298804 CY |
64 | mem_size += (sbi->im[i].ino_num * |
65 | sizeof(struct ino_entry)) >> PAGE_CACHE_SHIFT; | |
e5e7ea3c | 66 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); |
429511cd CY |
67 | } else if (type == EXTENT_CACHE) { |
68 | mem_size = (sbi->total_ext_tree * sizeof(struct extent_tree) + | |
69 | atomic_read(&sbi->total_ext_node) * | |
70 | sizeof(struct extent_node)) >> PAGE_CACHE_SHIFT; | |
71 | res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1); | |
1e84371f JK |
72 | } else { |
73 | if (sbi->sb->s_bdi->dirty_exceeded) | |
74 | return false; | |
6fb03f3a JK |
75 | } |
76 | return res; | |
cdfc41c1 JK |
77 | } |
78 | ||
e05df3b1 JK |
79 | static void clear_node_page_dirty(struct page *page) |
80 | { | |
81 | struct address_space *mapping = page->mapping; | |
e05df3b1 JK |
82 | unsigned int long flags; |
83 | ||
84 | if (PageDirty(page)) { | |
85 | spin_lock_irqsave(&mapping->tree_lock, flags); | |
86 | radix_tree_tag_clear(&mapping->page_tree, | |
87 | page_index(page), | |
88 | PAGECACHE_TAG_DIRTY); | |
89 | spin_unlock_irqrestore(&mapping->tree_lock, flags); | |
90 | ||
91 | clear_page_dirty_for_io(page); | |
4081363f | 92 | dec_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_NODES); |
e05df3b1 JK |
93 | } |
94 | ClearPageUptodate(page); | |
95 | } | |
96 | ||
97 | static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
98 | { | |
99 | pgoff_t index = current_nat_addr(sbi, nid); | |
100 | return get_meta_page(sbi, index); | |
101 | } | |
102 | ||
103 | static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) | |
104 | { | |
105 | struct page *src_page; | |
106 | struct page *dst_page; | |
107 | pgoff_t src_off; | |
108 | pgoff_t dst_off; | |
109 | void *src_addr; | |
110 | void *dst_addr; | |
111 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
112 | ||
113 | src_off = current_nat_addr(sbi, nid); | |
114 | dst_off = next_nat_addr(sbi, src_off); | |
115 | ||
116 | /* get current nat block page with lock */ | |
117 | src_page = get_meta_page(sbi, src_off); | |
e05df3b1 | 118 | dst_page = grab_meta_page(sbi, dst_off); |
9850cf4a | 119 | f2fs_bug_on(sbi, PageDirty(src_page)); |
e05df3b1 JK |
120 | |
121 | src_addr = page_address(src_page); | |
122 | dst_addr = page_address(dst_page); | |
123 | memcpy(dst_addr, src_addr, PAGE_CACHE_SIZE); | |
124 | set_page_dirty(dst_page); | |
125 | f2fs_put_page(src_page, 1); | |
126 | ||
127 | set_to_next_nat(nm_i, nid); | |
128 | ||
129 | return dst_page; | |
130 | } | |
131 | ||
e05df3b1 JK |
132 | static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n) |
133 | { | |
134 | return radix_tree_lookup(&nm_i->nat_root, n); | |
135 | } | |
136 | ||
137 | static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i, | |
138 | nid_t start, unsigned int nr, struct nat_entry **ep) | |
139 | { | |
140 | return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr); | |
141 | } | |
142 | ||
143 | static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e) | |
144 | { | |
145 | list_del(&e->list); | |
146 | radix_tree_delete(&nm_i->nat_root, nat_get_nid(e)); | |
147 | nm_i->nat_cnt--; | |
148 | kmem_cache_free(nat_entry_slab, e); | |
149 | } | |
150 | ||
309cc2b6 JK |
151 | static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i, |
152 | struct nat_entry *ne) | |
153 | { | |
154 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); | |
155 | struct nat_entry_set *head; | |
156 | ||
157 | if (get_nat_flag(ne, IS_DIRTY)) | |
158 | return; | |
9be32d72 | 159 | |
309cc2b6 JK |
160 | head = radix_tree_lookup(&nm_i->nat_set_root, set); |
161 | if (!head) { | |
162 | head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_ATOMIC); | |
163 | ||
164 | INIT_LIST_HEAD(&head->entry_list); | |
165 | INIT_LIST_HEAD(&head->set_list); | |
166 | head->set = set; | |
167 | head->entry_cnt = 0; | |
9be32d72 | 168 | f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head); |
309cc2b6 JK |
169 | } |
170 | list_move_tail(&ne->list, &head->entry_list); | |
171 | nm_i->dirty_nat_cnt++; | |
172 | head->entry_cnt++; | |
173 | set_nat_flag(ne, IS_DIRTY, true); | |
174 | } | |
175 | ||
176 | static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i, | |
177 | struct nat_entry *ne) | |
178 | { | |
20d047c8 | 179 | nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid); |
309cc2b6 JK |
180 | struct nat_entry_set *head; |
181 | ||
182 | head = radix_tree_lookup(&nm_i->nat_set_root, set); | |
183 | if (head) { | |
184 | list_move_tail(&ne->list, &nm_i->nat_entries); | |
185 | set_nat_flag(ne, IS_DIRTY, false); | |
186 | head->entry_cnt--; | |
187 | nm_i->dirty_nat_cnt--; | |
188 | } | |
189 | } | |
190 | ||
191 | static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i, | |
192 | nid_t start, unsigned int nr, struct nat_entry_set **ep) | |
193 | { | |
194 | return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep, | |
195 | start, nr); | |
196 | } | |
197 | ||
2dcf51ab | 198 | int need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid) |
e05df3b1 JK |
199 | { |
200 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
201 | struct nat_entry *e; | |
2dcf51ab | 202 | bool need = false; |
e05df3b1 | 203 | |
8b26ef98 | 204 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 | 205 | e = __lookup_nat_cache(nm_i, nid); |
2dcf51ab JK |
206 | if (e) { |
207 | if (!get_nat_flag(e, IS_CHECKPOINTED) && | |
208 | !get_nat_flag(e, HAS_FSYNCED_INODE)) | |
209 | need = true; | |
210 | } | |
8b26ef98 | 211 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 212 | return need; |
e05df3b1 JK |
213 | } |
214 | ||
2dcf51ab | 215 | bool is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid) |
479f40c4 JK |
216 | { |
217 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
218 | struct nat_entry *e; | |
2dcf51ab | 219 | bool is_cp = true; |
479f40c4 | 220 | |
8b26ef98 | 221 | down_read(&nm_i->nat_tree_lock); |
2dcf51ab JK |
222 | e = __lookup_nat_cache(nm_i, nid); |
223 | if (e && !get_nat_flag(e, IS_CHECKPOINTED)) | |
224 | is_cp = false; | |
8b26ef98 | 225 | up_read(&nm_i->nat_tree_lock); |
2dcf51ab | 226 | return is_cp; |
479f40c4 JK |
227 | } |
228 | ||
88bd02c9 | 229 | bool need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino) |
b6fe5873 JK |
230 | { |
231 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
232 | struct nat_entry *e; | |
88bd02c9 | 233 | bool need_update = true; |
b6fe5873 | 234 | |
8b26ef98 | 235 | down_read(&nm_i->nat_tree_lock); |
88bd02c9 JK |
236 | e = __lookup_nat_cache(nm_i, ino); |
237 | if (e && get_nat_flag(e, HAS_LAST_FSYNC) && | |
238 | (get_nat_flag(e, IS_CHECKPOINTED) || | |
239 | get_nat_flag(e, HAS_FSYNCED_INODE))) | |
240 | need_update = false; | |
8b26ef98 | 241 | up_read(&nm_i->nat_tree_lock); |
88bd02c9 | 242 | return need_update; |
b6fe5873 JK |
243 | } |
244 | ||
e05df3b1 JK |
245 | static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid) |
246 | { | |
247 | struct nat_entry *new; | |
248 | ||
9be32d72 JK |
249 | new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_ATOMIC); |
250 | f2fs_radix_tree_insert(&nm_i->nat_root, nid, new); | |
e05df3b1 JK |
251 | memset(new, 0, sizeof(struct nat_entry)); |
252 | nat_set_nid(new, nid); | |
88bd02c9 | 253 | nat_reset_flag(new); |
e05df3b1 JK |
254 | list_add_tail(&new->list, &nm_i->nat_entries); |
255 | nm_i->nat_cnt++; | |
256 | return new; | |
257 | } | |
258 | ||
259 | static void cache_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid, | |
260 | struct f2fs_nat_entry *ne) | |
261 | { | |
262 | struct nat_entry *e; | |
9be32d72 | 263 | |
8b26ef98 | 264 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
265 | e = __lookup_nat_cache(nm_i, nid); |
266 | if (!e) { | |
267 | e = grab_nat_entry(nm_i, nid); | |
94dac22e | 268 | node_info_from_raw_nat(&e->ni, ne); |
e05df3b1 | 269 | } |
8b26ef98 | 270 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
271 | } |
272 | ||
273 | static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, | |
479f40c4 | 274 | block_t new_blkaddr, bool fsync_done) |
e05df3b1 JK |
275 | { |
276 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
277 | struct nat_entry *e; | |
9be32d72 | 278 | |
8b26ef98 | 279 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
280 | e = __lookup_nat_cache(nm_i, ni->nid); |
281 | if (!e) { | |
282 | e = grab_nat_entry(nm_i, ni->nid); | |
5c27f4ee | 283 | copy_node_info(&e->ni, ni); |
9850cf4a | 284 | f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR); |
e05df3b1 JK |
285 | } else if (new_blkaddr == NEW_ADDR) { |
286 | /* | |
287 | * when nid is reallocated, | |
288 | * previous nat entry can be remained in nat cache. | |
289 | * So, reinitialize it with new information. | |
290 | */ | |
5c27f4ee | 291 | copy_node_info(&e->ni, ni); |
9850cf4a | 292 | f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR); |
e05df3b1 JK |
293 | } |
294 | ||
e05df3b1 | 295 | /* sanity check */ |
9850cf4a JK |
296 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr); |
297 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR && | |
e05df3b1 | 298 | new_blkaddr == NULL_ADDR); |
9850cf4a | 299 | f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR && |
e05df3b1 | 300 | new_blkaddr == NEW_ADDR); |
9850cf4a | 301 | f2fs_bug_on(sbi, nat_get_blkaddr(e) != NEW_ADDR && |
e05df3b1 JK |
302 | nat_get_blkaddr(e) != NULL_ADDR && |
303 | new_blkaddr == NEW_ADDR); | |
304 | ||
e1c42045 | 305 | /* increment version no as node is removed */ |
e05df3b1 JK |
306 | if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) { |
307 | unsigned char version = nat_get_version(e); | |
308 | nat_set_version(e, inc_node_version(version)); | |
309 | } | |
310 | ||
311 | /* change address */ | |
312 | nat_set_blkaddr(e, new_blkaddr); | |
88bd02c9 JK |
313 | if (new_blkaddr == NEW_ADDR || new_blkaddr == NULL_ADDR) |
314 | set_nat_flag(e, IS_CHECKPOINTED, false); | |
e05df3b1 | 315 | __set_nat_cache_dirty(nm_i, e); |
479f40c4 JK |
316 | |
317 | /* update fsync_mark if its inode nat entry is still alive */ | |
d5b692b7 CY |
318 | if (ni->nid != ni->ino) |
319 | e = __lookup_nat_cache(nm_i, ni->ino); | |
88bd02c9 JK |
320 | if (e) { |
321 | if (fsync_done && ni->nid == ni->ino) | |
322 | set_nat_flag(e, HAS_FSYNCED_INODE, true); | |
323 | set_nat_flag(e, HAS_LAST_FSYNC, fsync_done); | |
324 | } | |
8b26ef98 | 325 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
326 | } |
327 | ||
4660f9c0 | 328 | int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) |
e05df3b1 JK |
329 | { |
330 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
331 | ||
6fb03f3a | 332 | if (available_free_memory(sbi, NAT_ENTRIES)) |
e05df3b1 JK |
333 | return 0; |
334 | ||
8b26ef98 | 335 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
336 | while (nr_shrink && !list_empty(&nm_i->nat_entries)) { |
337 | struct nat_entry *ne; | |
338 | ne = list_first_entry(&nm_i->nat_entries, | |
339 | struct nat_entry, list); | |
340 | __del_from_nat_cache(nm_i, ne); | |
341 | nr_shrink--; | |
342 | } | |
8b26ef98 | 343 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
344 | return nr_shrink; |
345 | } | |
346 | ||
0a8165d7 | 347 | /* |
e1c42045 | 348 | * This function always returns success |
e05df3b1 JK |
349 | */ |
350 | void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni) | |
351 | { | |
352 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
353 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
354 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
355 | nid_t start_nid = START_NID(nid); | |
356 | struct f2fs_nat_block *nat_blk; | |
357 | struct page *page = NULL; | |
358 | struct f2fs_nat_entry ne; | |
359 | struct nat_entry *e; | |
360 | int i; | |
361 | ||
362 | ni->nid = nid; | |
363 | ||
364 | /* Check nat cache */ | |
8b26ef98 | 365 | down_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
366 | e = __lookup_nat_cache(nm_i, nid); |
367 | if (e) { | |
368 | ni->ino = nat_get_ino(e); | |
369 | ni->blk_addr = nat_get_blkaddr(e); | |
370 | ni->version = nat_get_version(e); | |
371 | } | |
8b26ef98 | 372 | up_read(&nm_i->nat_tree_lock); |
e05df3b1 JK |
373 | if (e) |
374 | return; | |
375 | ||
3547ea96 JK |
376 | memset(&ne, 0, sizeof(struct f2fs_nat_entry)); |
377 | ||
e05df3b1 JK |
378 | /* Check current segment summary */ |
379 | mutex_lock(&curseg->curseg_mutex); | |
380 | i = lookup_journal_in_cursum(sum, NAT_JOURNAL, nid, 0); | |
381 | if (i >= 0) { | |
382 | ne = nat_in_journal(sum, i); | |
383 | node_info_from_raw_nat(ni, &ne); | |
384 | } | |
385 | mutex_unlock(&curseg->curseg_mutex); | |
386 | if (i >= 0) | |
387 | goto cache; | |
388 | ||
389 | /* Fill node_info from nat page */ | |
390 | page = get_current_nat_page(sbi, start_nid); | |
391 | nat_blk = (struct f2fs_nat_block *)page_address(page); | |
392 | ne = nat_blk->entries[nid - start_nid]; | |
393 | node_info_from_raw_nat(ni, &ne); | |
394 | f2fs_put_page(page, 1); | |
395 | cache: | |
396 | /* cache nat entry */ | |
397 | cache_nat_entry(NM_I(sbi), nid, &ne); | |
398 | } | |
399 | ||
0a8165d7 | 400 | /* |
e05df3b1 JK |
401 | * The maximum depth is four. |
402 | * Offset[0] will have raw inode offset. | |
403 | */ | |
de93653f JK |
404 | static int get_node_path(struct f2fs_inode_info *fi, long block, |
405 | int offset[4], unsigned int noffset[4]) | |
e05df3b1 | 406 | { |
de93653f | 407 | const long direct_index = ADDRS_PER_INODE(fi); |
e05df3b1 JK |
408 | const long direct_blks = ADDRS_PER_BLOCK; |
409 | const long dptrs_per_blk = NIDS_PER_BLOCK; | |
410 | const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; | |
411 | const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK; | |
412 | int n = 0; | |
413 | int level = 0; | |
414 | ||
415 | noffset[0] = 0; | |
416 | ||
417 | if (block < direct_index) { | |
25c0a6e5 | 418 | offset[n] = block; |
e05df3b1 JK |
419 | goto got; |
420 | } | |
421 | block -= direct_index; | |
422 | if (block < direct_blks) { | |
423 | offset[n++] = NODE_DIR1_BLOCK; | |
424 | noffset[n] = 1; | |
25c0a6e5 | 425 | offset[n] = block; |
e05df3b1 JK |
426 | level = 1; |
427 | goto got; | |
428 | } | |
429 | block -= direct_blks; | |
430 | if (block < direct_blks) { | |
431 | offset[n++] = NODE_DIR2_BLOCK; | |
432 | noffset[n] = 2; | |
25c0a6e5 | 433 | offset[n] = block; |
e05df3b1 JK |
434 | level = 1; |
435 | goto got; | |
436 | } | |
437 | block -= direct_blks; | |
438 | if (block < indirect_blks) { | |
439 | offset[n++] = NODE_IND1_BLOCK; | |
440 | noffset[n] = 3; | |
441 | offset[n++] = block / direct_blks; | |
442 | noffset[n] = 4 + offset[n - 1]; | |
25c0a6e5 | 443 | offset[n] = block % direct_blks; |
e05df3b1 JK |
444 | level = 2; |
445 | goto got; | |
446 | } | |
447 | block -= indirect_blks; | |
448 | if (block < indirect_blks) { | |
449 | offset[n++] = NODE_IND2_BLOCK; | |
450 | noffset[n] = 4 + dptrs_per_blk; | |
451 | offset[n++] = block / direct_blks; | |
452 | noffset[n] = 5 + dptrs_per_blk + offset[n - 1]; | |
25c0a6e5 | 453 | offset[n] = block % direct_blks; |
e05df3b1 JK |
454 | level = 2; |
455 | goto got; | |
456 | } | |
457 | block -= indirect_blks; | |
458 | if (block < dindirect_blks) { | |
459 | offset[n++] = NODE_DIND_BLOCK; | |
460 | noffset[n] = 5 + (dptrs_per_blk * 2); | |
461 | offset[n++] = block / indirect_blks; | |
462 | noffset[n] = 6 + (dptrs_per_blk * 2) + | |
463 | offset[n - 1] * (dptrs_per_blk + 1); | |
464 | offset[n++] = (block / direct_blks) % dptrs_per_blk; | |
465 | noffset[n] = 7 + (dptrs_per_blk * 2) + | |
466 | offset[n - 2] * (dptrs_per_blk + 1) + | |
467 | offset[n - 1]; | |
25c0a6e5 | 468 | offset[n] = block % direct_blks; |
e05df3b1 JK |
469 | level = 3; |
470 | goto got; | |
471 | } else { | |
472 | BUG(); | |
473 | } | |
474 | got: | |
475 | return level; | |
476 | } | |
477 | ||
478 | /* | |
479 | * Caller should call f2fs_put_dnode(dn). | |
4f4124d0 CY |
480 | * Also, it should grab and release a rwsem by calling f2fs_lock_op() and |
481 | * f2fs_unlock_op() only if ro is not set RDONLY_NODE. | |
39936837 | 482 | * In the case of RDONLY_NODE, we don't need to care about mutex. |
e05df3b1 | 483 | */ |
266e97a8 | 484 | int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode) |
e05df3b1 | 485 | { |
4081363f | 486 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 | 487 | struct page *npage[4]; |
f1a3b98e | 488 | struct page *parent = NULL; |
e05df3b1 JK |
489 | int offset[4]; |
490 | unsigned int noffset[4]; | |
491 | nid_t nids[4]; | |
492 | int level, i; | |
493 | int err = 0; | |
494 | ||
de93653f | 495 | level = get_node_path(F2FS_I(dn->inode), index, offset, noffset); |
e05df3b1 JK |
496 | |
497 | nids[0] = dn->inode->i_ino; | |
1646cfac | 498 | npage[0] = dn->inode_page; |
e05df3b1 | 499 | |
1646cfac JK |
500 | if (!npage[0]) { |
501 | npage[0] = get_node_page(sbi, nids[0]); | |
502 | if (IS_ERR(npage[0])) | |
503 | return PTR_ERR(npage[0]); | |
504 | } | |
f1a3b98e JK |
505 | |
506 | /* if inline_data is set, should not report any block indices */ | |
507 | if (f2fs_has_inline_data(dn->inode) && index) { | |
76629165 | 508 | err = -ENOENT; |
f1a3b98e JK |
509 | f2fs_put_page(npage[0], 1); |
510 | goto release_out; | |
511 | } | |
512 | ||
e05df3b1 | 513 | parent = npage[0]; |
52c2db3f CL |
514 | if (level != 0) |
515 | nids[1] = get_nid(parent, offset[0], true); | |
e05df3b1 JK |
516 | dn->inode_page = npage[0]; |
517 | dn->inode_page_locked = true; | |
518 | ||
519 | /* get indirect or direct nodes */ | |
520 | for (i = 1; i <= level; i++) { | |
521 | bool done = false; | |
522 | ||
266e97a8 | 523 | if (!nids[i] && mode == ALLOC_NODE) { |
e05df3b1 JK |
524 | /* alloc new node */ |
525 | if (!alloc_nid(sbi, &(nids[i]))) { | |
e05df3b1 JK |
526 | err = -ENOSPC; |
527 | goto release_pages; | |
528 | } | |
529 | ||
530 | dn->nid = nids[i]; | |
8ae8f162 | 531 | npage[i] = new_node_page(dn, noffset[i], NULL); |
e05df3b1 JK |
532 | if (IS_ERR(npage[i])) { |
533 | alloc_nid_failed(sbi, nids[i]); | |
e05df3b1 JK |
534 | err = PTR_ERR(npage[i]); |
535 | goto release_pages; | |
536 | } | |
537 | ||
538 | set_nid(parent, offset[i - 1], nids[i], i == 1); | |
539 | alloc_nid_done(sbi, nids[i]); | |
e05df3b1 | 540 | done = true; |
266e97a8 | 541 | } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) { |
e05df3b1 JK |
542 | npage[i] = get_node_page_ra(parent, offset[i - 1]); |
543 | if (IS_ERR(npage[i])) { | |
544 | err = PTR_ERR(npage[i]); | |
545 | goto release_pages; | |
546 | } | |
547 | done = true; | |
548 | } | |
549 | if (i == 1) { | |
550 | dn->inode_page_locked = false; | |
551 | unlock_page(parent); | |
552 | } else { | |
553 | f2fs_put_page(parent, 1); | |
554 | } | |
555 | ||
556 | if (!done) { | |
557 | npage[i] = get_node_page(sbi, nids[i]); | |
558 | if (IS_ERR(npage[i])) { | |
559 | err = PTR_ERR(npage[i]); | |
560 | f2fs_put_page(npage[0], 0); | |
561 | goto release_out; | |
562 | } | |
563 | } | |
564 | if (i < level) { | |
565 | parent = npage[i]; | |
566 | nids[i + 1] = get_nid(parent, offset[i], false); | |
567 | } | |
568 | } | |
569 | dn->nid = nids[level]; | |
570 | dn->ofs_in_node = offset[level]; | |
571 | dn->node_page = npage[level]; | |
572 | dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); | |
573 | return 0; | |
574 | ||
575 | release_pages: | |
576 | f2fs_put_page(parent, 1); | |
577 | if (i > 1) | |
578 | f2fs_put_page(npage[0], 0); | |
579 | release_out: | |
580 | dn->inode_page = NULL; | |
581 | dn->node_page = NULL; | |
582 | return err; | |
583 | } | |
584 | ||
585 | static void truncate_node(struct dnode_of_data *dn) | |
586 | { | |
4081363f | 587 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
588 | struct node_info ni; |
589 | ||
590 | get_node_info(sbi, dn->nid, &ni); | |
71e9fec5 | 591 | if (dn->inode->i_blocks == 0) { |
9850cf4a | 592 | f2fs_bug_on(sbi, ni.blk_addr != NULL_ADDR); |
71e9fec5 JK |
593 | goto invalidate; |
594 | } | |
9850cf4a | 595 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
e05df3b1 | 596 | |
e05df3b1 | 597 | /* Deallocate node address */ |
71e9fec5 | 598 | invalidate_blocks(sbi, ni.blk_addr); |
ef86d709 | 599 | dec_valid_node_count(sbi, dn->inode); |
479f40c4 | 600 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
e05df3b1 JK |
601 | |
602 | if (dn->nid == dn->inode->i_ino) { | |
603 | remove_orphan_inode(sbi, dn->nid); | |
604 | dec_valid_inode_count(sbi); | |
605 | } else { | |
606 | sync_inode_page(dn); | |
607 | } | |
71e9fec5 | 608 | invalidate: |
e05df3b1 | 609 | clear_node_page_dirty(dn->node_page); |
caf0047e | 610 | set_sbi_flag(sbi, SBI_IS_DIRTY); |
e05df3b1 JK |
611 | |
612 | f2fs_put_page(dn->node_page, 1); | |
bf39c00a JK |
613 | |
614 | invalidate_mapping_pages(NODE_MAPPING(sbi), | |
615 | dn->node_page->index, dn->node_page->index); | |
616 | ||
e05df3b1 | 617 | dn->node_page = NULL; |
51dd6249 | 618 | trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr); |
e05df3b1 JK |
619 | } |
620 | ||
621 | static int truncate_dnode(struct dnode_of_data *dn) | |
622 | { | |
e05df3b1 JK |
623 | struct page *page; |
624 | ||
625 | if (dn->nid == 0) | |
626 | return 1; | |
627 | ||
628 | /* get direct node */ | |
4081363f | 629 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
e05df3b1 JK |
630 | if (IS_ERR(page) && PTR_ERR(page) == -ENOENT) |
631 | return 1; | |
632 | else if (IS_ERR(page)) | |
633 | return PTR_ERR(page); | |
634 | ||
635 | /* Make dnode_of_data for parameter */ | |
636 | dn->node_page = page; | |
637 | dn->ofs_in_node = 0; | |
638 | truncate_data_blocks(dn); | |
639 | truncate_node(dn); | |
640 | return 1; | |
641 | } | |
642 | ||
643 | static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs, | |
644 | int ofs, int depth) | |
645 | { | |
e05df3b1 JK |
646 | struct dnode_of_data rdn = *dn; |
647 | struct page *page; | |
648 | struct f2fs_node *rn; | |
649 | nid_t child_nid; | |
650 | unsigned int child_nofs; | |
651 | int freed = 0; | |
652 | int i, ret; | |
653 | ||
654 | if (dn->nid == 0) | |
655 | return NIDS_PER_BLOCK + 1; | |
656 | ||
51dd6249 NJ |
657 | trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr); |
658 | ||
4081363f | 659 | page = get_node_page(F2FS_I_SB(dn->inode), dn->nid); |
51dd6249 NJ |
660 | if (IS_ERR(page)) { |
661 | trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page)); | |
e05df3b1 | 662 | return PTR_ERR(page); |
51dd6249 | 663 | } |
e05df3b1 | 664 | |
45590710 | 665 | rn = F2FS_NODE(page); |
e05df3b1 JK |
666 | if (depth < 3) { |
667 | for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) { | |
668 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
669 | if (child_nid == 0) | |
670 | continue; | |
671 | rdn.nid = child_nid; | |
672 | ret = truncate_dnode(&rdn); | |
673 | if (ret < 0) | |
674 | goto out_err; | |
675 | set_nid(page, i, 0, false); | |
676 | } | |
677 | } else { | |
678 | child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1; | |
679 | for (i = ofs; i < NIDS_PER_BLOCK; i++) { | |
680 | child_nid = le32_to_cpu(rn->in.nid[i]); | |
681 | if (child_nid == 0) { | |
682 | child_nofs += NIDS_PER_BLOCK + 1; | |
683 | continue; | |
684 | } | |
685 | rdn.nid = child_nid; | |
686 | ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1); | |
687 | if (ret == (NIDS_PER_BLOCK + 1)) { | |
688 | set_nid(page, i, 0, false); | |
689 | child_nofs += ret; | |
690 | } else if (ret < 0 && ret != -ENOENT) { | |
691 | goto out_err; | |
692 | } | |
693 | } | |
694 | freed = child_nofs; | |
695 | } | |
696 | ||
697 | if (!ofs) { | |
698 | /* remove current indirect node */ | |
699 | dn->node_page = page; | |
700 | truncate_node(dn); | |
701 | freed++; | |
702 | } else { | |
703 | f2fs_put_page(page, 1); | |
704 | } | |
51dd6249 | 705 | trace_f2fs_truncate_nodes_exit(dn->inode, freed); |
e05df3b1 JK |
706 | return freed; |
707 | ||
708 | out_err: | |
709 | f2fs_put_page(page, 1); | |
51dd6249 | 710 | trace_f2fs_truncate_nodes_exit(dn->inode, ret); |
e05df3b1 JK |
711 | return ret; |
712 | } | |
713 | ||
714 | static int truncate_partial_nodes(struct dnode_of_data *dn, | |
715 | struct f2fs_inode *ri, int *offset, int depth) | |
716 | { | |
e05df3b1 JK |
717 | struct page *pages[2]; |
718 | nid_t nid[3]; | |
719 | nid_t child_nid; | |
720 | int err = 0; | |
721 | int i; | |
722 | int idx = depth - 2; | |
723 | ||
724 | nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); | |
725 | if (!nid[0]) | |
726 | return 0; | |
727 | ||
728 | /* get indirect nodes in the path */ | |
a225dca3 | 729 | for (i = 0; i < idx + 1; i++) { |
e1c42045 | 730 | /* reference count'll be increased */ |
4081363f | 731 | pages[i] = get_node_page(F2FS_I_SB(dn->inode), nid[i]); |
e05df3b1 | 732 | if (IS_ERR(pages[i])) { |
e05df3b1 | 733 | err = PTR_ERR(pages[i]); |
a225dca3 | 734 | idx = i - 1; |
e05df3b1 JK |
735 | goto fail; |
736 | } | |
737 | nid[i + 1] = get_nid(pages[i], offset[i + 1], false); | |
738 | } | |
739 | ||
740 | /* free direct nodes linked to a partial indirect node */ | |
a225dca3 | 741 | for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) { |
e05df3b1 JK |
742 | child_nid = get_nid(pages[idx], i, false); |
743 | if (!child_nid) | |
744 | continue; | |
745 | dn->nid = child_nid; | |
746 | err = truncate_dnode(dn); | |
747 | if (err < 0) | |
748 | goto fail; | |
749 | set_nid(pages[idx], i, 0, false); | |
750 | } | |
751 | ||
a225dca3 | 752 | if (offset[idx + 1] == 0) { |
e05df3b1 JK |
753 | dn->node_page = pages[idx]; |
754 | dn->nid = nid[idx]; | |
755 | truncate_node(dn); | |
756 | } else { | |
757 | f2fs_put_page(pages[idx], 1); | |
758 | } | |
759 | offset[idx]++; | |
a225dca3 | 760 | offset[idx + 1] = 0; |
761 | idx--; | |
e05df3b1 | 762 | fail: |
a225dca3 | 763 | for (i = idx; i >= 0; i--) |
e05df3b1 | 764 | f2fs_put_page(pages[i], 1); |
51dd6249 NJ |
765 | |
766 | trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err); | |
767 | ||
e05df3b1 JK |
768 | return err; |
769 | } | |
770 | ||
0a8165d7 | 771 | /* |
e05df3b1 JK |
772 | * All the block addresses of data and nodes should be nullified. |
773 | */ | |
774 | int truncate_inode_blocks(struct inode *inode, pgoff_t from) | |
775 | { | |
4081363f | 776 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
e05df3b1 JK |
777 | int err = 0, cont = 1; |
778 | int level, offset[4], noffset[4]; | |
7dd690c8 | 779 | unsigned int nofs = 0; |
58bfaf44 | 780 | struct f2fs_inode *ri; |
e05df3b1 JK |
781 | struct dnode_of_data dn; |
782 | struct page *page; | |
783 | ||
51dd6249 NJ |
784 | trace_f2fs_truncate_inode_blocks_enter(inode, from); |
785 | ||
de93653f | 786 | level = get_node_path(F2FS_I(inode), from, offset, noffset); |
afcb7ca0 | 787 | restart: |
e05df3b1 | 788 | page = get_node_page(sbi, inode->i_ino); |
51dd6249 NJ |
789 | if (IS_ERR(page)) { |
790 | trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page)); | |
e05df3b1 | 791 | return PTR_ERR(page); |
51dd6249 | 792 | } |
e05df3b1 JK |
793 | |
794 | set_new_dnode(&dn, inode, page, NULL, 0); | |
795 | unlock_page(page); | |
796 | ||
58bfaf44 | 797 | ri = F2FS_INODE(page); |
e05df3b1 JK |
798 | switch (level) { |
799 | case 0: | |
800 | case 1: | |
801 | nofs = noffset[1]; | |
802 | break; | |
803 | case 2: | |
804 | nofs = noffset[1]; | |
805 | if (!offset[level - 1]) | |
806 | goto skip_partial; | |
58bfaf44 | 807 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
808 | if (err < 0 && err != -ENOENT) |
809 | goto fail; | |
810 | nofs += 1 + NIDS_PER_BLOCK; | |
811 | break; | |
812 | case 3: | |
813 | nofs = 5 + 2 * NIDS_PER_BLOCK; | |
814 | if (!offset[level - 1]) | |
815 | goto skip_partial; | |
58bfaf44 | 816 | err = truncate_partial_nodes(&dn, ri, offset, level); |
e05df3b1 JK |
817 | if (err < 0 && err != -ENOENT) |
818 | goto fail; | |
819 | break; | |
820 | default: | |
821 | BUG(); | |
822 | } | |
823 | ||
824 | skip_partial: | |
825 | while (cont) { | |
58bfaf44 | 826 | dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); |
e05df3b1 JK |
827 | switch (offset[0]) { |
828 | case NODE_DIR1_BLOCK: | |
829 | case NODE_DIR2_BLOCK: | |
830 | err = truncate_dnode(&dn); | |
831 | break; | |
832 | ||
833 | case NODE_IND1_BLOCK: | |
834 | case NODE_IND2_BLOCK: | |
835 | err = truncate_nodes(&dn, nofs, offset[1], 2); | |
836 | break; | |
837 | ||
838 | case NODE_DIND_BLOCK: | |
839 | err = truncate_nodes(&dn, nofs, offset[1], 3); | |
840 | cont = 0; | |
841 | break; | |
842 | ||
843 | default: | |
844 | BUG(); | |
845 | } | |
846 | if (err < 0 && err != -ENOENT) | |
847 | goto fail; | |
848 | if (offset[1] == 0 && | |
58bfaf44 | 849 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) { |
e05df3b1 | 850 | lock_page(page); |
4ef51a8f | 851 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
852 | f2fs_put_page(page, 1); |
853 | goto restart; | |
854 | } | |
3cb5ad15 | 855 | f2fs_wait_on_page_writeback(page, NODE); |
58bfaf44 | 856 | ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0; |
e05df3b1 JK |
857 | set_page_dirty(page); |
858 | unlock_page(page); | |
859 | } | |
860 | offset[1] = 0; | |
861 | offset[0]++; | |
862 | nofs += err; | |
863 | } | |
864 | fail: | |
865 | f2fs_put_page(page, 0); | |
51dd6249 | 866 | trace_f2fs_truncate_inode_blocks_exit(inode, err); |
e05df3b1 JK |
867 | return err > 0 ? 0 : err; |
868 | } | |
869 | ||
4f16fb0f JK |
870 | int truncate_xattr_node(struct inode *inode, struct page *page) |
871 | { | |
4081363f | 872 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
4f16fb0f JK |
873 | nid_t nid = F2FS_I(inode)->i_xattr_nid; |
874 | struct dnode_of_data dn; | |
875 | struct page *npage; | |
876 | ||
877 | if (!nid) | |
878 | return 0; | |
879 | ||
880 | npage = get_node_page(sbi, nid); | |
881 | if (IS_ERR(npage)) | |
882 | return PTR_ERR(npage); | |
883 | ||
884 | F2FS_I(inode)->i_xattr_nid = 0; | |
65985d93 JK |
885 | |
886 | /* need to do checkpoint during fsync */ | |
887 | F2FS_I(inode)->xattr_ver = cur_cp_version(F2FS_CKPT(sbi)); | |
888 | ||
4f16fb0f JK |
889 | set_new_dnode(&dn, inode, page, npage, nid); |
890 | ||
891 | if (page) | |
01d2d1aa | 892 | dn.inode_page_locked = true; |
4f16fb0f JK |
893 | truncate_node(&dn); |
894 | return 0; | |
895 | } | |
896 | ||
39936837 | 897 | /* |
4f4124d0 CY |
898 | * Caller should grab and release a rwsem by calling f2fs_lock_op() and |
899 | * f2fs_unlock_op(). | |
39936837 | 900 | */ |
58e674d6 | 901 | void remove_inode_page(struct inode *inode) |
e05df3b1 | 902 | { |
e05df3b1 JK |
903 | struct dnode_of_data dn; |
904 | ||
c2e69583 JK |
905 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); |
906 | if (get_dnode_of_data(&dn, 0, LOOKUP_NODE)) | |
58e674d6 | 907 | return; |
e05df3b1 | 908 | |
c2e69583 JK |
909 | if (truncate_xattr_node(inode, dn.inode_page)) { |
910 | f2fs_put_dnode(&dn); | |
58e674d6 | 911 | return; |
e05df3b1 | 912 | } |
c2e69583 JK |
913 | |
914 | /* remove potential inline_data blocks */ | |
915 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
916 | S_ISLNK(inode->i_mode)) | |
917 | truncate_data_blocks_range(&dn, 1); | |
918 | ||
e1c42045 | 919 | /* 0 is possible, after f2fs_new_inode() has failed */ |
9850cf4a JK |
920 | f2fs_bug_on(F2FS_I_SB(inode), |
921 | inode->i_blocks != 0 && inode->i_blocks != 1); | |
c2e69583 JK |
922 | |
923 | /* will put inode & node pages */ | |
71e9fec5 | 924 | truncate_node(&dn); |
e05df3b1 JK |
925 | } |
926 | ||
a014e037 | 927 | struct page *new_inode_page(struct inode *inode) |
e05df3b1 | 928 | { |
e05df3b1 JK |
929 | struct dnode_of_data dn; |
930 | ||
931 | /* allocate inode page for new inode */ | |
932 | set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); | |
44a83ff6 JK |
933 | |
934 | /* caller should f2fs_put_page(page, 1); */ | |
8ae8f162 | 935 | return new_node_page(&dn, 0, NULL); |
e05df3b1 JK |
936 | } |
937 | ||
8ae8f162 JK |
938 | struct page *new_node_page(struct dnode_of_data *dn, |
939 | unsigned int ofs, struct page *ipage) | |
e05df3b1 | 940 | { |
4081363f | 941 | struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode); |
e05df3b1 JK |
942 | struct node_info old_ni, new_ni; |
943 | struct page *page; | |
944 | int err; | |
945 | ||
6bacf52f | 946 | if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))) |
e05df3b1 JK |
947 | return ERR_PTR(-EPERM); |
948 | ||
54b591df | 949 | page = grab_cache_page(NODE_MAPPING(sbi), dn->nid); |
e05df3b1 JK |
950 | if (!page) |
951 | return ERR_PTR(-ENOMEM); | |
952 | ||
6bacf52f | 953 | if (unlikely(!inc_valid_node_count(sbi, dn->inode))) { |
9c02740c JK |
954 | err = -ENOSPC; |
955 | goto fail; | |
956 | } | |
e05df3b1 | 957 | |
9c02740c | 958 | get_node_info(sbi, dn->nid, &old_ni); |
e05df3b1 JK |
959 | |
960 | /* Reinitialize old_ni with new node page */ | |
9850cf4a | 961 | f2fs_bug_on(sbi, old_ni.blk_addr != NULL_ADDR); |
e05df3b1 JK |
962 | new_ni = old_ni; |
963 | new_ni.ino = dn->inode->i_ino; | |
479f40c4 | 964 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
9c02740c | 965 | |
54b591df | 966 | f2fs_wait_on_page_writeback(page, NODE); |
9c02740c | 967 | fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true); |
398b1ac5 | 968 | set_cold_node(dn->inode, page); |
9c02740c JK |
969 | SetPageUptodate(page); |
970 | set_page_dirty(page); | |
e05df3b1 | 971 | |
4bc8e9bc | 972 | if (f2fs_has_xattr_block(ofs)) |
479bd73a JK |
973 | F2FS_I(dn->inode)->i_xattr_nid = dn->nid; |
974 | ||
e05df3b1 | 975 | dn->node_page = page; |
8ae8f162 JK |
976 | if (ipage) |
977 | update_inode(dn->inode, ipage); | |
978 | else | |
979 | sync_inode_page(dn); | |
e05df3b1 JK |
980 | if (ofs == 0) |
981 | inc_valid_inode_count(sbi); | |
982 | ||
983 | return page; | |
984 | ||
985 | fail: | |
71e9fec5 | 986 | clear_node_page_dirty(page); |
e05df3b1 JK |
987 | f2fs_put_page(page, 1); |
988 | return ERR_PTR(err); | |
989 | } | |
990 | ||
56ae674c JK |
991 | /* |
992 | * Caller should do after getting the following values. | |
993 | * 0: f2fs_put_page(page, 0) | |
994 | * LOCKED_PAGE: f2fs_put_page(page, 1) | |
995 | * error: nothing | |
996 | */ | |
93dfe2ac | 997 | static int read_node_page(struct page *page, int rw) |
e05df3b1 | 998 | { |
4081363f | 999 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1000 | struct node_info ni; |
cf04e8eb | 1001 | struct f2fs_io_info fio = { |
05ca3632 | 1002 | .sbi = sbi, |
cf04e8eb JK |
1003 | .type = NODE, |
1004 | .rw = rw, | |
05ca3632 | 1005 | .page = page, |
cf04e8eb | 1006 | }; |
e05df3b1 JK |
1007 | |
1008 | get_node_info(sbi, page->index, &ni); | |
1009 | ||
6bacf52f | 1010 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
2bca1e23 | 1011 | ClearPageUptodate(page); |
393ff91f | 1012 | f2fs_put_page(page, 1); |
e05df3b1 | 1013 | return -ENOENT; |
393ff91f JK |
1014 | } |
1015 | ||
56ae674c JK |
1016 | if (PageUptodate(page)) |
1017 | return LOCKED_PAGE; | |
393ff91f | 1018 | |
cf04e8eb | 1019 | fio.blk_addr = ni.blk_addr; |
05ca3632 | 1020 | return f2fs_submit_page_bio(&fio); |
e05df3b1 JK |
1021 | } |
1022 | ||
0a8165d7 | 1023 | /* |
e05df3b1 JK |
1024 | * Readahead a node page |
1025 | */ | |
1026 | void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid) | |
1027 | { | |
e05df3b1 | 1028 | struct page *apage; |
56ae674c | 1029 | int err; |
e05df3b1 | 1030 | |
4ef51a8f | 1031 | apage = find_get_page(NODE_MAPPING(sbi), nid); |
393ff91f JK |
1032 | if (apage && PageUptodate(apage)) { |
1033 | f2fs_put_page(apage, 0); | |
1034 | return; | |
1035 | } | |
e05df3b1 JK |
1036 | f2fs_put_page(apage, 0); |
1037 | ||
4ef51a8f | 1038 | apage = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1039 | if (!apage) |
1040 | return; | |
1041 | ||
56ae674c JK |
1042 | err = read_node_page(apage, READA); |
1043 | if (err == 0) | |
393ff91f | 1044 | f2fs_put_page(apage, 0); |
56ae674c JK |
1045 | else if (err == LOCKED_PAGE) |
1046 | f2fs_put_page(apage, 1); | |
e05df3b1 JK |
1047 | } |
1048 | ||
1049 | struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) | |
1050 | { | |
56ae674c JK |
1051 | struct page *page; |
1052 | int err; | |
afcb7ca0 | 1053 | repeat: |
54b591df | 1054 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1055 | if (!page) |
1056 | return ERR_PTR(-ENOMEM); | |
1057 | ||
1058 | err = read_node_page(page, READ_SYNC); | |
56ae674c | 1059 | if (err < 0) |
e05df3b1 | 1060 | return ERR_PTR(err); |
aaf96075 JK |
1061 | else if (err != LOCKED_PAGE) |
1062 | lock_page(page); | |
e05df3b1 | 1063 | |
3bb5e2c8 | 1064 | if (unlikely(!PageUptodate(page) || nid != nid_of_node(page))) { |
aaf96075 | 1065 | ClearPageUptodate(page); |
393ff91f JK |
1066 | f2fs_put_page(page, 1); |
1067 | return ERR_PTR(-EIO); | |
1068 | } | |
4ef51a8f | 1069 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1070 | f2fs_put_page(page, 1); |
1071 | goto repeat; | |
1072 | } | |
e05df3b1 JK |
1073 | return page; |
1074 | } | |
1075 | ||
0a8165d7 | 1076 | /* |
e05df3b1 JK |
1077 | * Return a locked page for the desired node page. |
1078 | * And, readahead MAX_RA_NODE number of node pages. | |
1079 | */ | |
1080 | struct page *get_node_page_ra(struct page *parent, int start) | |
1081 | { | |
4081363f | 1082 | struct f2fs_sb_info *sbi = F2FS_P_SB(parent); |
c718379b | 1083 | struct blk_plug plug; |
e05df3b1 | 1084 | struct page *page; |
56ae674c JK |
1085 | int err, i, end; |
1086 | nid_t nid; | |
e05df3b1 JK |
1087 | |
1088 | /* First, try getting the desired direct node. */ | |
1089 | nid = get_nid(parent, start, false); | |
1090 | if (!nid) | |
1091 | return ERR_PTR(-ENOENT); | |
afcb7ca0 | 1092 | repeat: |
4ef51a8f | 1093 | page = grab_cache_page(NODE_MAPPING(sbi), nid); |
e05df3b1 JK |
1094 | if (!page) |
1095 | return ERR_PTR(-ENOMEM); | |
1096 | ||
66d36a29 | 1097 | err = read_node_page(page, READ_SYNC); |
56ae674c | 1098 | if (err < 0) |
e05df3b1 | 1099 | return ERR_PTR(err); |
56ae674c JK |
1100 | else if (err == LOCKED_PAGE) |
1101 | goto page_hit; | |
e05df3b1 | 1102 | |
c718379b JK |
1103 | blk_start_plug(&plug); |
1104 | ||
e05df3b1 JK |
1105 | /* Then, try readahead for siblings of the desired node */ |
1106 | end = start + MAX_RA_NODE; | |
1107 | end = min(end, NIDS_PER_BLOCK); | |
1108 | for (i = start + 1; i < end; i++) { | |
1109 | nid = get_nid(parent, i, false); | |
1110 | if (!nid) | |
1111 | continue; | |
1112 | ra_node_page(sbi, nid); | |
1113 | } | |
1114 | ||
c718379b JK |
1115 | blk_finish_plug(&plug); |
1116 | ||
e05df3b1 | 1117 | lock_page(page); |
4ef51a8f | 1118 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
afcb7ca0 JK |
1119 | f2fs_put_page(page, 1); |
1120 | goto repeat; | |
1121 | } | |
e0f56cb4 | 1122 | page_hit: |
6bacf52f | 1123 | if (unlikely(!PageUptodate(page))) { |
e05df3b1 JK |
1124 | f2fs_put_page(page, 1); |
1125 | return ERR_PTR(-EIO); | |
1126 | } | |
e05df3b1 JK |
1127 | return page; |
1128 | } | |
1129 | ||
1130 | void sync_inode_page(struct dnode_of_data *dn) | |
1131 | { | |
1132 | if (IS_INODE(dn->node_page) || dn->inode_page == dn->node_page) { | |
1133 | update_inode(dn->inode, dn->node_page); | |
1134 | } else if (dn->inode_page) { | |
1135 | if (!dn->inode_page_locked) | |
1136 | lock_page(dn->inode_page); | |
1137 | update_inode(dn->inode, dn->inode_page); | |
1138 | if (!dn->inode_page_locked) | |
1139 | unlock_page(dn->inode_page); | |
1140 | } else { | |
39936837 | 1141 | update_inode_page(dn->inode); |
e05df3b1 JK |
1142 | } |
1143 | } | |
1144 | ||
1145 | int sync_node_pages(struct f2fs_sb_info *sbi, nid_t ino, | |
1146 | struct writeback_control *wbc) | |
1147 | { | |
e05df3b1 JK |
1148 | pgoff_t index, end; |
1149 | struct pagevec pvec; | |
1150 | int step = ino ? 2 : 0; | |
1151 | int nwritten = 0, wrote = 0; | |
1152 | ||
1153 | pagevec_init(&pvec, 0); | |
1154 | ||
1155 | next_step: | |
1156 | index = 0; | |
1157 | end = LONG_MAX; | |
1158 | ||
1159 | while (index <= end) { | |
1160 | int i, nr_pages; | |
4ef51a8f | 1161 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, |
e05df3b1 JK |
1162 | PAGECACHE_TAG_DIRTY, |
1163 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1164 | if (nr_pages == 0) | |
1165 | break; | |
1166 | ||
1167 | for (i = 0; i < nr_pages; i++) { | |
1168 | struct page *page = pvec.pages[i]; | |
1169 | ||
1170 | /* | |
1171 | * flushing sequence with step: | |
1172 | * 0. indirect nodes | |
1173 | * 1. dentry dnodes | |
1174 | * 2. file dnodes | |
1175 | */ | |
1176 | if (step == 0 && IS_DNODE(page)) | |
1177 | continue; | |
1178 | if (step == 1 && (!IS_DNODE(page) || | |
1179 | is_cold_node(page))) | |
1180 | continue; | |
1181 | if (step == 2 && (!IS_DNODE(page) || | |
1182 | !is_cold_node(page))) | |
1183 | continue; | |
1184 | ||
1185 | /* | |
1186 | * If an fsync mode, | |
1187 | * we should not skip writing node pages. | |
1188 | */ | |
1189 | if (ino && ino_of_node(page) == ino) | |
1190 | lock_page(page); | |
1191 | else if (!trylock_page(page)) | |
1192 | continue; | |
1193 | ||
4ef51a8f | 1194 | if (unlikely(page->mapping != NODE_MAPPING(sbi))) { |
e05df3b1 JK |
1195 | continue_unlock: |
1196 | unlock_page(page); | |
1197 | continue; | |
1198 | } | |
1199 | if (ino && ino_of_node(page) != ino) | |
1200 | goto continue_unlock; | |
1201 | ||
1202 | if (!PageDirty(page)) { | |
1203 | /* someone wrote it for us */ | |
1204 | goto continue_unlock; | |
1205 | } | |
1206 | ||
1207 | if (!clear_page_dirty_for_io(page)) | |
1208 | goto continue_unlock; | |
1209 | ||
1210 | /* called by fsync() */ | |
1211 | if (ino && IS_DNODE(page)) { | |
e05df3b1 | 1212 | set_fsync_mark(page, 1); |
2dcf51ab JK |
1213 | if (IS_INODE(page)) |
1214 | set_dentry_mark(page, | |
1215 | need_dentry_mark(sbi, ino)); | |
e05df3b1 JK |
1216 | nwritten++; |
1217 | } else { | |
1218 | set_fsync_mark(page, 0); | |
1219 | set_dentry_mark(page, 0); | |
1220 | } | |
52746519 JK |
1221 | |
1222 | if (NODE_MAPPING(sbi)->a_ops->writepage(page, wbc)) | |
1223 | unlock_page(page); | |
1224 | else | |
1225 | wrote++; | |
e05df3b1 JK |
1226 | |
1227 | if (--wbc->nr_to_write == 0) | |
1228 | break; | |
1229 | } | |
1230 | pagevec_release(&pvec); | |
1231 | cond_resched(); | |
1232 | ||
1233 | if (wbc->nr_to_write == 0) { | |
1234 | step = 2; | |
1235 | break; | |
1236 | } | |
1237 | } | |
1238 | ||
1239 | if (step < 2) { | |
1240 | step++; | |
1241 | goto next_step; | |
1242 | } | |
1243 | ||
1244 | if (wrote) | |
458e6197 | 1245 | f2fs_submit_merged_bio(sbi, NODE, WRITE); |
e05df3b1 JK |
1246 | return nwritten; |
1247 | } | |
1248 | ||
cfe58f9d JK |
1249 | int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino) |
1250 | { | |
cfe58f9d JK |
1251 | pgoff_t index = 0, end = LONG_MAX; |
1252 | struct pagevec pvec; | |
cfe58f9d JK |
1253 | int ret2 = 0, ret = 0; |
1254 | ||
1255 | pagevec_init(&pvec, 0); | |
4ef51a8f JK |
1256 | |
1257 | while (index <= end) { | |
1258 | int i, nr_pages; | |
1259 | nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, | |
1260 | PAGECACHE_TAG_WRITEBACK, | |
1261 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); | |
1262 | if (nr_pages == 0) | |
1263 | break; | |
cfe58f9d JK |
1264 | |
1265 | for (i = 0; i < nr_pages; i++) { | |
1266 | struct page *page = pvec.pages[i]; | |
1267 | ||
1268 | /* until radix tree lookup accepts end_index */ | |
cfb271d4 | 1269 | if (unlikely(page->index > end)) |
cfe58f9d JK |
1270 | continue; |
1271 | ||
4bf08ff6 | 1272 | if (ino && ino_of_node(page) == ino) { |
3cb5ad15 | 1273 | f2fs_wait_on_page_writeback(page, NODE); |
4bf08ff6 CY |
1274 | if (TestClearPageError(page)) |
1275 | ret = -EIO; | |
1276 | } | |
cfe58f9d JK |
1277 | } |
1278 | pagevec_release(&pvec); | |
1279 | cond_resched(); | |
1280 | } | |
1281 | ||
4ef51a8f | 1282 | if (unlikely(test_and_clear_bit(AS_ENOSPC, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d | 1283 | ret2 = -ENOSPC; |
4ef51a8f | 1284 | if (unlikely(test_and_clear_bit(AS_EIO, &NODE_MAPPING(sbi)->flags))) |
cfe58f9d JK |
1285 | ret2 = -EIO; |
1286 | if (!ret) | |
1287 | ret = ret2; | |
1288 | return ret; | |
1289 | } | |
1290 | ||
e05df3b1 JK |
1291 | static int f2fs_write_node_page(struct page *page, |
1292 | struct writeback_control *wbc) | |
1293 | { | |
4081363f | 1294 | struct f2fs_sb_info *sbi = F2FS_P_SB(page); |
e05df3b1 | 1295 | nid_t nid; |
e05df3b1 | 1296 | struct node_info ni; |
fb5566da | 1297 | struct f2fs_io_info fio = { |
05ca3632 | 1298 | .sbi = sbi, |
fb5566da | 1299 | .type = NODE, |
6c311ec6 | 1300 | .rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE, |
05ca3632 | 1301 | .page = page, |
fb5566da | 1302 | }; |
e05df3b1 | 1303 | |
ecda0de3 CY |
1304 | trace_f2fs_writepage(page, NODE); |
1305 | ||
caf0047e | 1306 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
87a9bd26 | 1307 | goto redirty_out; |
cf779cab JK |
1308 | if (unlikely(f2fs_cp_error(sbi))) |
1309 | goto redirty_out; | |
87a9bd26 | 1310 | |
3cb5ad15 | 1311 | f2fs_wait_on_page_writeback(page, NODE); |
e05df3b1 | 1312 | |
e05df3b1 JK |
1313 | /* get old block addr of this node page */ |
1314 | nid = nid_of_node(page); | |
9850cf4a | 1315 | f2fs_bug_on(sbi, page->index != nid); |
e05df3b1 JK |
1316 | |
1317 | get_node_info(sbi, nid, &ni); | |
1318 | ||
1319 | /* This page is already truncated */ | |
6bacf52f | 1320 | if (unlikely(ni.blk_addr == NULL_ADDR)) { |
2bca1e23 | 1321 | ClearPageUptodate(page); |
39936837 JK |
1322 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
1323 | unlock_page(page); | |
1324 | return 0; | |
1325 | } | |
e05df3b1 | 1326 | |
2f97c326 JK |
1327 | if (wbc->for_reclaim) { |
1328 | if (!down_read_trylock(&sbi->node_write)) | |
1329 | goto redirty_out; | |
1330 | } else { | |
1331 | down_read(&sbi->node_write); | |
1332 | } | |
cf04e8eb | 1333 | |
e05df3b1 | 1334 | set_page_writeback(page); |
cf04e8eb | 1335 | fio.blk_addr = ni.blk_addr; |
05ca3632 | 1336 | write_node_page(nid, &fio); |
cf04e8eb | 1337 | set_node_addr(sbi, &ni, fio.blk_addr, is_fsync_dnode(page)); |
e05df3b1 | 1338 | dec_page_count(sbi, F2FS_DIRTY_NODES); |
b3582c68 | 1339 | up_read(&sbi->node_write); |
e05df3b1 | 1340 | unlock_page(page); |
27c6bd60 JK |
1341 | |
1342 | if (wbc->for_reclaim) | |
1343 | f2fs_submit_merged_bio(sbi, NODE, WRITE); | |
1344 | ||
e05df3b1 | 1345 | return 0; |
87a9bd26 JK |
1346 | |
1347 | redirty_out: | |
76f60268 | 1348 | redirty_page_for_writepage(wbc, page); |
87a9bd26 | 1349 | return AOP_WRITEPAGE_ACTIVATE; |
e05df3b1 JK |
1350 | } |
1351 | ||
1352 | static int f2fs_write_node_pages(struct address_space *mapping, | |
1353 | struct writeback_control *wbc) | |
1354 | { | |
4081363f | 1355 | struct f2fs_sb_info *sbi = F2FS_M_SB(mapping); |
50c8cdb3 | 1356 | long diff; |
e05df3b1 | 1357 | |
e5748434 CY |
1358 | trace_f2fs_writepages(mapping->host, wbc, NODE); |
1359 | ||
4660f9c0 JK |
1360 | /* balancing f2fs's metadata in background */ |
1361 | f2fs_balance_fs_bg(sbi); | |
e05df3b1 | 1362 | |
a7fdffbd | 1363 | /* collect a number of dirty node pages and write together */ |
87d6f890 | 1364 | if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE)) |
d3baf95d | 1365 | goto skip_write; |
a7fdffbd | 1366 | |
50c8cdb3 | 1367 | diff = nr_pages_to_write(sbi, NODE, wbc); |
fb5566da | 1368 | wbc->sync_mode = WB_SYNC_NONE; |
e05df3b1 | 1369 | sync_node_pages(sbi, 0, wbc); |
50c8cdb3 | 1370 | wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff); |
e05df3b1 | 1371 | return 0; |
d3baf95d JK |
1372 | |
1373 | skip_write: | |
1374 | wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES); | |
1375 | return 0; | |
e05df3b1 JK |
1376 | } |
1377 | ||
1378 | static int f2fs_set_node_page_dirty(struct page *page) | |
1379 | { | |
26c6b887 JK |
1380 | trace_f2fs_set_page_dirty(page, NODE); |
1381 | ||
e05df3b1 JK |
1382 | SetPageUptodate(page); |
1383 | if (!PageDirty(page)) { | |
1384 | __set_page_dirty_nobuffers(page); | |
4081363f | 1385 | inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES); |
e05df3b1 | 1386 | SetPagePrivate(page); |
9e4ded3f | 1387 | f2fs_trace_pid(page); |
e05df3b1 JK |
1388 | return 1; |
1389 | } | |
1390 | return 0; | |
1391 | } | |
1392 | ||
0a8165d7 | 1393 | /* |
e05df3b1 JK |
1394 | * Structure of the f2fs node operations |
1395 | */ | |
1396 | const struct address_space_operations f2fs_node_aops = { | |
1397 | .writepage = f2fs_write_node_page, | |
1398 | .writepages = f2fs_write_node_pages, | |
1399 | .set_page_dirty = f2fs_set_node_page_dirty, | |
487261f3 CY |
1400 | .invalidatepage = f2fs_invalidate_page, |
1401 | .releasepage = f2fs_release_page, | |
e05df3b1 JK |
1402 | }; |
1403 | ||
8a7ed66a JK |
1404 | static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i, |
1405 | nid_t n) | |
e05df3b1 | 1406 | { |
8a7ed66a | 1407 | return radix_tree_lookup(&nm_i->free_nid_root, n); |
e05df3b1 JK |
1408 | } |
1409 | ||
8a7ed66a JK |
1410 | static void __del_from_free_nid_list(struct f2fs_nm_info *nm_i, |
1411 | struct free_nid *i) | |
e05df3b1 JK |
1412 | { |
1413 | list_del(&i->list); | |
8a7ed66a | 1414 | radix_tree_delete(&nm_i->free_nid_root, i->nid); |
e05df3b1 JK |
1415 | } |
1416 | ||
6fb03f3a | 1417 | static int add_free_nid(struct f2fs_sb_info *sbi, nid_t nid, bool build) |
e05df3b1 | 1418 | { |
6fb03f3a | 1419 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1420 | struct free_nid *i; |
59bbd474 JK |
1421 | struct nat_entry *ne; |
1422 | bool allocated = false; | |
e05df3b1 | 1423 | |
6fb03f3a | 1424 | if (!available_free_memory(sbi, FREE_NIDS)) |
23d38844 | 1425 | return -1; |
9198aceb JK |
1426 | |
1427 | /* 0 nid should not be used */ | |
cfb271d4 | 1428 | if (unlikely(nid == 0)) |
9198aceb | 1429 | return 0; |
59bbd474 | 1430 | |
7bd59381 GZ |
1431 | if (build) { |
1432 | /* do not add allocated nids */ | |
8b26ef98 | 1433 | down_read(&nm_i->nat_tree_lock); |
7bd59381 | 1434 | ne = __lookup_nat_cache(nm_i, nid); |
8a7ed66a | 1435 | if (ne && |
7ef35e3b JK |
1436 | (!get_nat_flag(ne, IS_CHECKPOINTED) || |
1437 | nat_get_blkaddr(ne) != NULL_ADDR)) | |
7bd59381 | 1438 | allocated = true; |
8b26ef98 | 1439 | up_read(&nm_i->nat_tree_lock); |
7bd59381 GZ |
1440 | if (allocated) |
1441 | return 0; | |
e05df3b1 | 1442 | } |
7bd59381 GZ |
1443 | |
1444 | i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS); | |
e05df3b1 JK |
1445 | i->nid = nid; |
1446 | i->state = NID_NEW; | |
1447 | ||
769ec6e5 JK |
1448 | if (radix_tree_preload(GFP_NOFS)) { |
1449 | kmem_cache_free(free_nid_slab, i); | |
1450 | return 0; | |
1451 | } | |
1452 | ||
e05df3b1 | 1453 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1454 | if (radix_tree_insert(&nm_i->free_nid_root, i->nid, i)) { |
e05df3b1 | 1455 | spin_unlock(&nm_i->free_nid_list_lock); |
769ec6e5 | 1456 | radix_tree_preload_end(); |
e05df3b1 JK |
1457 | kmem_cache_free(free_nid_slab, i); |
1458 | return 0; | |
1459 | } | |
1460 | list_add_tail(&i->list, &nm_i->free_nid_list); | |
1461 | nm_i->fcnt++; | |
1462 | spin_unlock(&nm_i->free_nid_list_lock); | |
769ec6e5 | 1463 | radix_tree_preload_end(); |
e05df3b1 JK |
1464 | return 1; |
1465 | } | |
1466 | ||
1467 | static void remove_free_nid(struct f2fs_nm_info *nm_i, nid_t nid) | |
1468 | { | |
1469 | struct free_nid *i; | |
cf0ee0f0 CY |
1470 | bool need_free = false; |
1471 | ||
e05df3b1 | 1472 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1473 | i = __lookup_free_nid_list(nm_i, nid); |
e05df3b1 | 1474 | if (i && i->state == NID_NEW) { |
8a7ed66a | 1475 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1476 | nm_i->fcnt--; |
cf0ee0f0 | 1477 | need_free = true; |
e05df3b1 JK |
1478 | } |
1479 | spin_unlock(&nm_i->free_nid_list_lock); | |
cf0ee0f0 CY |
1480 | |
1481 | if (need_free) | |
1482 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1483 | } |
1484 | ||
6fb03f3a | 1485 | static void scan_nat_page(struct f2fs_sb_info *sbi, |
e05df3b1 JK |
1486 | struct page *nat_page, nid_t start_nid) |
1487 | { | |
6fb03f3a | 1488 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 JK |
1489 | struct f2fs_nat_block *nat_blk = page_address(nat_page); |
1490 | block_t blk_addr; | |
e05df3b1 JK |
1491 | int i; |
1492 | ||
e05df3b1 JK |
1493 | i = start_nid % NAT_ENTRY_PER_BLOCK; |
1494 | ||
1495 | for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { | |
23d38844 | 1496 | |
cfb271d4 | 1497 | if (unlikely(start_nid >= nm_i->max_nid)) |
04431c44 | 1498 | break; |
23d38844 HL |
1499 | |
1500 | blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); | |
9850cf4a | 1501 | f2fs_bug_on(sbi, blk_addr == NEW_ADDR); |
23d38844 | 1502 | if (blk_addr == NULL_ADDR) { |
6fb03f3a | 1503 | if (add_free_nid(sbi, start_nid, true) < 0) |
23d38844 HL |
1504 | break; |
1505 | } | |
e05df3b1 | 1506 | } |
e05df3b1 JK |
1507 | } |
1508 | ||
1509 | static void build_free_nids(struct f2fs_sb_info *sbi) | |
1510 | { | |
e05df3b1 JK |
1511 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1512 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1513 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
8760952d | 1514 | int i = 0; |
55008d84 | 1515 | nid_t nid = nm_i->next_scan_nid; |
e05df3b1 | 1516 | |
55008d84 JK |
1517 | /* Enough entries */ |
1518 | if (nm_i->fcnt > NAT_ENTRY_PER_BLOCK) | |
1519 | return; | |
e05df3b1 | 1520 | |
55008d84 | 1521 | /* readahead nat pages to be scanned */ |
662befda | 1522 | ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES, META_NAT); |
e05df3b1 JK |
1523 | |
1524 | while (1) { | |
1525 | struct page *page = get_current_nat_page(sbi, nid); | |
1526 | ||
6fb03f3a | 1527 | scan_nat_page(sbi, page, nid); |
e05df3b1 JK |
1528 | f2fs_put_page(page, 1); |
1529 | ||
1530 | nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); | |
cfb271d4 | 1531 | if (unlikely(nid >= nm_i->max_nid)) |
e05df3b1 | 1532 | nid = 0; |
55008d84 JK |
1533 | |
1534 | if (i++ == FREE_NID_PAGES) | |
e05df3b1 JK |
1535 | break; |
1536 | } | |
1537 | ||
55008d84 JK |
1538 | /* go to the next free nat pages to find free nids abundantly */ |
1539 | nm_i->next_scan_nid = nid; | |
e05df3b1 JK |
1540 | |
1541 | /* find free nids from current sum_pages */ | |
1542 | mutex_lock(&curseg->curseg_mutex); | |
1543 | for (i = 0; i < nats_in_cursum(sum); i++) { | |
1544 | block_t addr = le32_to_cpu(nat_in_journal(sum, i).block_addr); | |
1545 | nid = le32_to_cpu(nid_in_journal(sum, i)); | |
1546 | if (addr == NULL_ADDR) | |
6fb03f3a | 1547 | add_free_nid(sbi, nid, true); |
e05df3b1 JK |
1548 | else |
1549 | remove_free_nid(nm_i, nid); | |
1550 | } | |
1551 | mutex_unlock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1552 | } |
1553 | ||
1554 | /* | |
1555 | * If this function returns success, caller can obtain a new nid | |
1556 | * from second parameter of this function. | |
1557 | * The returned nid could be used ino as well as nid when inode is created. | |
1558 | */ | |
1559 | bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) | |
1560 | { | |
1561 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1562 | struct free_nid *i = NULL; | |
e05df3b1 | 1563 | retry: |
7ee0eeab | 1564 | if (unlikely(sbi->total_valid_node_count + 1 > nm_i->available_nids)) |
55008d84 | 1565 | return false; |
e05df3b1 | 1566 | |
e05df3b1 | 1567 | spin_lock(&nm_i->free_nid_list_lock); |
e05df3b1 | 1568 | |
55008d84 | 1569 | /* We should not use stale free nids created by build_free_nids */ |
f978f5a0 | 1570 | if (nm_i->fcnt && !on_build_free_nids(nm_i)) { |
9850cf4a | 1571 | f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list)); |
2d7b822a | 1572 | list_for_each_entry(i, &nm_i->free_nid_list, list) |
55008d84 JK |
1573 | if (i->state == NID_NEW) |
1574 | break; | |
e05df3b1 | 1575 | |
9850cf4a | 1576 | f2fs_bug_on(sbi, i->state != NID_NEW); |
55008d84 JK |
1577 | *nid = i->nid; |
1578 | i->state = NID_ALLOC; | |
1579 | nm_i->fcnt--; | |
1580 | spin_unlock(&nm_i->free_nid_list_lock); | |
1581 | return true; | |
1582 | } | |
e05df3b1 | 1583 | spin_unlock(&nm_i->free_nid_list_lock); |
55008d84 JK |
1584 | |
1585 | /* Let's scan nat pages and its caches to get free nids */ | |
1586 | mutex_lock(&nm_i->build_lock); | |
55008d84 | 1587 | build_free_nids(sbi); |
55008d84 JK |
1588 | mutex_unlock(&nm_i->build_lock); |
1589 | goto retry; | |
e05df3b1 JK |
1590 | } |
1591 | ||
0a8165d7 | 1592 | /* |
e05df3b1 JK |
1593 | * alloc_nid() should be called prior to this function. |
1594 | */ | |
1595 | void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) | |
1596 | { | |
1597 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1598 | struct free_nid *i; | |
1599 | ||
1600 | spin_lock(&nm_i->free_nid_list_lock); | |
8a7ed66a | 1601 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1602 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
8a7ed66a | 1603 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 1604 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1605 | |
1606 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1607 | } |
1608 | ||
0a8165d7 | 1609 | /* |
e05df3b1 JK |
1610 | * alloc_nid() should be called prior to this function. |
1611 | */ | |
1612 | void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) | |
1613 | { | |
49952fa1 JK |
1614 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
1615 | struct free_nid *i; | |
cf0ee0f0 | 1616 | bool need_free = false; |
49952fa1 | 1617 | |
65985d93 JK |
1618 | if (!nid) |
1619 | return; | |
1620 | ||
49952fa1 | 1621 | spin_lock(&nm_i->free_nid_list_lock); |
8a7ed66a | 1622 | i = __lookup_free_nid_list(nm_i, nid); |
9850cf4a | 1623 | f2fs_bug_on(sbi, !i || i->state != NID_ALLOC); |
6fb03f3a | 1624 | if (!available_free_memory(sbi, FREE_NIDS)) { |
8a7ed66a | 1625 | __del_from_free_nid_list(nm_i, i); |
cf0ee0f0 | 1626 | need_free = true; |
95630cba HL |
1627 | } else { |
1628 | i->state = NID_NEW; | |
1629 | nm_i->fcnt++; | |
1630 | } | |
49952fa1 | 1631 | spin_unlock(&nm_i->free_nid_list_lock); |
cf0ee0f0 CY |
1632 | |
1633 | if (need_free) | |
1634 | kmem_cache_free(free_nid_slab, i); | |
e05df3b1 JK |
1635 | } |
1636 | ||
70cfed88 | 1637 | void recover_inline_xattr(struct inode *inode, struct page *page) |
28cdce04 | 1638 | { |
28cdce04 CY |
1639 | void *src_addr, *dst_addr; |
1640 | size_t inline_size; | |
1641 | struct page *ipage; | |
1642 | struct f2fs_inode *ri; | |
1643 | ||
4081363f | 1644 | ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino); |
9850cf4a | 1645 | f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage)); |
28cdce04 | 1646 | |
e3b4d43f JK |
1647 | ri = F2FS_INODE(page); |
1648 | if (!(ri->i_inline & F2FS_INLINE_XATTR)) { | |
1649 | clear_inode_flag(F2FS_I(inode), FI_INLINE_XATTR); | |
1650 | goto update_inode; | |
1651 | } | |
1652 | ||
28cdce04 CY |
1653 | dst_addr = inline_xattr_addr(ipage); |
1654 | src_addr = inline_xattr_addr(page); | |
1655 | inline_size = inline_xattr_size(inode); | |
1656 | ||
54b591df | 1657 | f2fs_wait_on_page_writeback(ipage, NODE); |
28cdce04 | 1658 | memcpy(dst_addr, src_addr, inline_size); |
e3b4d43f | 1659 | update_inode: |
28cdce04 CY |
1660 | update_inode(inode, ipage); |
1661 | f2fs_put_page(ipage, 1); | |
1662 | } | |
1663 | ||
1c35a90e | 1664 | void recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr) |
abb2366c | 1665 | { |
4081363f | 1666 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
abb2366c JK |
1667 | nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid; |
1668 | nid_t new_xnid = nid_of_node(page); | |
1669 | struct node_info ni; | |
1670 | ||
abb2366c JK |
1671 | /* 1: invalidate the previous xattr nid */ |
1672 | if (!prev_xnid) | |
1673 | goto recover_xnid; | |
1674 | ||
1675 | /* Deallocate node address */ | |
1676 | get_node_info(sbi, prev_xnid, &ni); | |
9850cf4a | 1677 | f2fs_bug_on(sbi, ni.blk_addr == NULL_ADDR); |
abb2366c JK |
1678 | invalidate_blocks(sbi, ni.blk_addr); |
1679 | dec_valid_node_count(sbi, inode); | |
479f40c4 | 1680 | set_node_addr(sbi, &ni, NULL_ADDR, false); |
abb2366c JK |
1681 | |
1682 | recover_xnid: | |
1683 | /* 2: allocate new xattr nid */ | |
1684 | if (unlikely(!inc_valid_node_count(sbi, inode))) | |
9850cf4a | 1685 | f2fs_bug_on(sbi, 1); |
abb2366c JK |
1686 | |
1687 | remove_free_nid(NM_I(sbi), new_xnid); | |
1688 | get_node_info(sbi, new_xnid, &ni); | |
1689 | ni.ino = inode->i_ino; | |
479f40c4 | 1690 | set_node_addr(sbi, &ni, NEW_ADDR, false); |
abb2366c JK |
1691 | F2FS_I(inode)->i_xattr_nid = new_xnid; |
1692 | ||
1693 | /* 3: update xattr blkaddr */ | |
1694 | refresh_sit_entry(sbi, NEW_ADDR, blkaddr); | |
479f40c4 | 1695 | set_node_addr(sbi, &ni, blkaddr, false); |
abb2366c JK |
1696 | |
1697 | update_inode_page(inode); | |
abb2366c JK |
1698 | } |
1699 | ||
e05df3b1 JK |
1700 | int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) |
1701 | { | |
58bfaf44 | 1702 | struct f2fs_inode *src, *dst; |
e05df3b1 JK |
1703 | nid_t ino = ino_of_node(page); |
1704 | struct node_info old_ni, new_ni; | |
1705 | struct page *ipage; | |
1706 | ||
e8271fa3 JK |
1707 | get_node_info(sbi, ino, &old_ni); |
1708 | ||
1709 | if (unlikely(old_ni.blk_addr != NULL_ADDR)) | |
1710 | return -EINVAL; | |
1711 | ||
4ef51a8f | 1712 | ipage = grab_cache_page(NODE_MAPPING(sbi), ino); |
e05df3b1 JK |
1713 | if (!ipage) |
1714 | return -ENOMEM; | |
1715 | ||
e1c42045 | 1716 | /* Should not use this inode from free nid list */ |
e05df3b1 JK |
1717 | remove_free_nid(NM_I(sbi), ino); |
1718 | ||
e05df3b1 JK |
1719 | SetPageUptodate(ipage); |
1720 | fill_node_footer(ipage, ino, ino, 0, true); | |
1721 | ||
58bfaf44 JK |
1722 | src = F2FS_INODE(page); |
1723 | dst = F2FS_INODE(ipage); | |
e05df3b1 | 1724 | |
58bfaf44 JK |
1725 | memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src); |
1726 | dst->i_size = 0; | |
1727 | dst->i_blocks = cpu_to_le64(1); | |
1728 | dst->i_links = cpu_to_le32(1); | |
1729 | dst->i_xattr_nid = 0; | |
617deb8c | 1730 | dst->i_inline = src->i_inline & F2FS_INLINE_XATTR; |
e05df3b1 JK |
1731 | |
1732 | new_ni = old_ni; | |
1733 | new_ni.ino = ino; | |
1734 | ||
cfb271d4 | 1735 | if (unlikely(!inc_valid_node_count(sbi, NULL))) |
65e5cd0a | 1736 | WARN_ON(1); |
479f40c4 | 1737 | set_node_addr(sbi, &new_ni, NEW_ADDR, false); |
e05df3b1 | 1738 | inc_valid_inode_count(sbi); |
617deb8c | 1739 | set_page_dirty(ipage); |
e05df3b1 JK |
1740 | f2fs_put_page(ipage, 1); |
1741 | return 0; | |
1742 | } | |
1743 | ||
1744 | int restore_node_summary(struct f2fs_sb_info *sbi, | |
1745 | unsigned int segno, struct f2fs_summary_block *sum) | |
1746 | { | |
1747 | struct f2fs_node *rn; | |
1748 | struct f2fs_summary *sum_entry; | |
e05df3b1 | 1749 | block_t addr; |
90a893c7 | 1750 | int bio_blocks = MAX_BIO_BLOCKS(sbi); |
9ecf4b80 | 1751 | int i, idx, last_offset, nrpages; |
e05df3b1 JK |
1752 | |
1753 | /* scan the node segment */ | |
1754 | last_offset = sbi->blocks_per_seg; | |
1755 | addr = START_BLOCK(sbi, segno); | |
1756 | sum_entry = &sum->entries[0]; | |
1757 | ||
9ecf4b80 | 1758 | for (i = 0; i < last_offset; i += nrpages, addr += nrpages) { |
9af0ff1c | 1759 | nrpages = min(last_offset - i, bio_blocks); |
393ff91f | 1760 | |
e1c42045 | 1761 | /* readahead node pages */ |
9ecf4b80 | 1762 | ra_meta_pages(sbi, addr, nrpages, META_POR); |
e05df3b1 | 1763 | |
9ecf4b80 CY |
1764 | for (idx = addr; idx < addr + nrpages; idx++) { |
1765 | struct page *page = get_meta_page(sbi, idx); | |
9af0ff1c | 1766 | |
9ecf4b80 CY |
1767 | rn = F2FS_NODE(page); |
1768 | sum_entry->nid = rn->footer.nid; | |
1769 | sum_entry->version = 0; | |
1770 | sum_entry->ofs_in_node = 0; | |
1771 | sum_entry++; | |
1772 | f2fs_put_page(page, 1); | |
9af0ff1c | 1773 | } |
bac4eef6 | 1774 | |
9ecf4b80 | 1775 | invalidate_mapping_pages(META_MAPPING(sbi), addr, |
bac4eef6 | 1776 | addr + nrpages); |
e05df3b1 | 1777 | } |
9ecf4b80 | 1778 | return 0; |
e05df3b1 JK |
1779 | } |
1780 | ||
aec71382 | 1781 | static void remove_nats_in_journal(struct f2fs_sb_info *sbi) |
e05df3b1 JK |
1782 | { |
1783 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1784 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1785 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1786 | int i; | |
1787 | ||
1788 | mutex_lock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1789 | for (i = 0; i < nats_in_cursum(sum); i++) { |
1790 | struct nat_entry *ne; | |
1791 | struct f2fs_nat_entry raw_ne; | |
1792 | nid_t nid = le32_to_cpu(nid_in_journal(sum, i)); | |
1793 | ||
1794 | raw_ne = nat_in_journal(sum, i); | |
9be32d72 | 1795 | |
8b26ef98 | 1796 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 | 1797 | ne = __lookup_nat_cache(nm_i, nid); |
e05df3b1 | 1798 | if (!ne) { |
9be32d72 JK |
1799 | ne = grab_nat_entry(nm_i, nid); |
1800 | node_info_from_raw_nat(&ne->ni, &raw_ne); | |
e05df3b1 | 1801 | } |
e05df3b1 | 1802 | __set_nat_cache_dirty(nm_i, ne); |
8b26ef98 | 1803 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
1804 | } |
1805 | update_nats_in_cursum(sum, -i); | |
1806 | mutex_unlock(&curseg->curseg_mutex); | |
e05df3b1 JK |
1807 | } |
1808 | ||
309cc2b6 JK |
1809 | static void __adjust_nat_entry_set(struct nat_entry_set *nes, |
1810 | struct list_head *head, int max) | |
e05df3b1 | 1811 | { |
309cc2b6 | 1812 | struct nat_entry_set *cur; |
e05df3b1 | 1813 | |
309cc2b6 JK |
1814 | if (nes->entry_cnt >= max) |
1815 | goto add_out; | |
e05df3b1 | 1816 | |
309cc2b6 JK |
1817 | list_for_each_entry(cur, head, set_list) { |
1818 | if (cur->entry_cnt >= nes->entry_cnt) { | |
1819 | list_add(&nes->set_list, cur->set_list.prev); | |
1820 | return; | |
1821 | } | |
aec71382 | 1822 | } |
309cc2b6 JK |
1823 | add_out: |
1824 | list_add_tail(&nes->set_list, head); | |
1825 | } | |
e05df3b1 | 1826 | |
309cc2b6 JK |
1827 | static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, |
1828 | struct nat_entry_set *set) | |
1829 | { | |
1830 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1831 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
1832 | nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK; | |
1833 | bool to_journal = true; | |
1834 | struct f2fs_nat_block *nat_blk; | |
1835 | struct nat_entry *ne, *cur; | |
1836 | struct page *page = NULL; | |
57ed1e95 | 1837 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
e05df3b1 | 1838 | |
aec71382 CY |
1839 | /* |
1840 | * there are two steps to flush nat entries: | |
1841 | * #1, flush nat entries to journal in current hot data summary block. | |
1842 | * #2, flush nat entries to nat page. | |
1843 | */ | |
309cc2b6 JK |
1844 | if (!__has_cursum_space(sum, set->entry_cnt, NAT_JOURNAL)) |
1845 | to_journal = false; | |
1846 | ||
1847 | if (to_journal) { | |
1848 | mutex_lock(&curseg->curseg_mutex); | |
1849 | } else { | |
1850 | page = get_next_nat_page(sbi, start_nid); | |
1851 | nat_blk = page_address(page); | |
1852 | f2fs_bug_on(sbi, !nat_blk); | |
1853 | } | |
aec71382 | 1854 | |
309cc2b6 JK |
1855 | /* flush dirty nats in nat entry set */ |
1856 | list_for_each_entry_safe(ne, cur, &set->entry_list, list) { | |
1857 | struct f2fs_nat_entry *raw_ne; | |
1858 | nid_t nid = nat_get_nid(ne); | |
1859 | int offset; | |
1860 | ||
1861 | if (nat_get_blkaddr(ne) == NEW_ADDR) | |
1862 | continue; | |
aec71382 CY |
1863 | |
1864 | if (to_journal) { | |
309cc2b6 JK |
1865 | offset = lookup_journal_in_cursum(sum, |
1866 | NAT_JOURNAL, nid, 1); | |
1867 | f2fs_bug_on(sbi, offset < 0); | |
1868 | raw_ne = &nat_in_journal(sum, offset); | |
1869 | nid_in_journal(sum, offset) = cpu_to_le32(nid); | |
aec71382 | 1870 | } else { |
309cc2b6 | 1871 | raw_ne = &nat_blk->entries[nid - start_nid]; |
e05df3b1 | 1872 | } |
309cc2b6 | 1873 | raw_nat_from_node_info(raw_ne, &ne->ni); |
e05df3b1 | 1874 | |
8b26ef98 | 1875 | down_write(&NM_I(sbi)->nat_tree_lock); |
309cc2b6 JK |
1876 | nat_reset_flag(ne); |
1877 | __clear_nat_cache_dirty(NM_I(sbi), ne); | |
8b26ef98 | 1878 | up_write(&NM_I(sbi)->nat_tree_lock); |
aec71382 | 1879 | |
309cc2b6 JK |
1880 | if (nat_get_blkaddr(ne) == NULL_ADDR) |
1881 | add_free_nid(sbi, nid, false); | |
1882 | } | |
e05df3b1 | 1883 | |
309cc2b6 JK |
1884 | if (to_journal) |
1885 | mutex_unlock(&curseg->curseg_mutex); | |
1886 | else | |
1887 | f2fs_put_page(page, 1); | |
aec71382 | 1888 | |
80ec2e91 CL |
1889 | f2fs_bug_on(sbi, set->entry_cnt); |
1890 | ||
57ed1e95 | 1891 | down_write(&nm_i->nat_tree_lock); |
80ec2e91 | 1892 | radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); |
57ed1e95 | 1893 | up_write(&nm_i->nat_tree_lock); |
80ec2e91 | 1894 | kmem_cache_free(nat_entry_set_slab, set); |
309cc2b6 | 1895 | } |
aec71382 | 1896 | |
309cc2b6 JK |
1897 | /* |
1898 | * This function is called during the checkpointing process. | |
1899 | */ | |
1900 | void flush_nat_entries(struct f2fs_sb_info *sbi) | |
1901 | { | |
1902 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1903 | struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); | |
1904 | struct f2fs_summary_block *sum = curseg->sum_blk; | |
7aed0d45 | 1905 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
309cc2b6 JK |
1906 | struct nat_entry_set *set, *tmp; |
1907 | unsigned int found; | |
1908 | nid_t set_idx = 0; | |
1909 | LIST_HEAD(sets); | |
1910 | ||
20d047c8 CL |
1911 | if (!nm_i->dirty_nat_cnt) |
1912 | return; | |
309cc2b6 JK |
1913 | /* |
1914 | * if there are no enough space in journal to store dirty nat | |
1915 | * entries, remove all entries from journal and merge them | |
1916 | * into nat entry set. | |
1917 | */ | |
1918 | if (!__has_cursum_space(sum, nm_i->dirty_nat_cnt, NAT_JOURNAL)) | |
1919 | remove_nats_in_journal(sbi); | |
1920 | ||
57ed1e95 | 1921 | down_write(&nm_i->nat_tree_lock); |
309cc2b6 | 1922 | while ((found = __gang_lookup_nat_set(nm_i, |
7aed0d45 | 1923 | set_idx, SETVEC_SIZE, setvec))) { |
309cc2b6 JK |
1924 | unsigned idx; |
1925 | set_idx = setvec[found - 1]->set + 1; | |
1926 | for (idx = 0; idx < found; idx++) | |
1927 | __adjust_nat_entry_set(setvec[idx], &sets, | |
1928 | MAX_NAT_JENTRIES(sum)); | |
e05df3b1 | 1929 | } |
57ed1e95 | 1930 | up_write(&nm_i->nat_tree_lock); |
aec71382 | 1931 | |
309cc2b6 JK |
1932 | /* flush dirty nats in nat entry set */ |
1933 | list_for_each_entry_safe(set, tmp, &sets, set_list) | |
1934 | __flush_nat_entry_set(sbi, set); | |
1935 | ||
9850cf4a | 1936 | f2fs_bug_on(sbi, nm_i->dirty_nat_cnt); |
e05df3b1 JK |
1937 | } |
1938 | ||
1939 | static int init_node_manager(struct f2fs_sb_info *sbi) | |
1940 | { | |
1941 | struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); | |
1942 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
1943 | unsigned char *version_bitmap; | |
1944 | unsigned int nat_segs, nat_blocks; | |
1945 | ||
1946 | nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); | |
1947 | ||
1948 | /* segment_count_nat includes pair segment so divide to 2. */ | |
1949 | nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; | |
1950 | nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); | |
b63da15e | 1951 | |
7ee0eeab JK |
1952 | nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks; |
1953 | ||
b63da15e | 1954 | /* not used nids: 0, node, meta, (and root counted as valid node) */ |
c200b1aa | 1955 | nm_i->available_nids = nm_i->max_nid - F2FS_RESERVED_NODE_NUM; |
e05df3b1 JK |
1956 | nm_i->fcnt = 0; |
1957 | nm_i->nat_cnt = 0; | |
cdfc41c1 | 1958 | nm_i->ram_thresh = DEF_RAM_THRESHOLD; |
e05df3b1 | 1959 | |
8a7ed66a | 1960 | INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC); |
e05df3b1 | 1961 | INIT_LIST_HEAD(&nm_i->free_nid_list); |
769ec6e5 JK |
1962 | INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO); |
1963 | INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO); | |
e05df3b1 | 1964 | INIT_LIST_HEAD(&nm_i->nat_entries); |
e05df3b1 JK |
1965 | |
1966 | mutex_init(&nm_i->build_lock); | |
1967 | spin_lock_init(&nm_i->free_nid_list_lock); | |
8b26ef98 | 1968 | init_rwsem(&nm_i->nat_tree_lock); |
e05df3b1 | 1969 | |
e05df3b1 | 1970 | nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); |
79b5793b | 1971 | nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); |
e05df3b1 JK |
1972 | version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); |
1973 | if (!version_bitmap) | |
1974 | return -EFAULT; | |
1975 | ||
79b5793b AG |
1976 | nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, |
1977 | GFP_KERNEL); | |
1978 | if (!nm_i->nat_bitmap) | |
1979 | return -ENOMEM; | |
e05df3b1 JK |
1980 | return 0; |
1981 | } | |
1982 | ||
1983 | int build_node_manager(struct f2fs_sb_info *sbi) | |
1984 | { | |
1985 | int err; | |
1986 | ||
1987 | sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL); | |
1988 | if (!sbi->nm_info) | |
1989 | return -ENOMEM; | |
1990 | ||
1991 | err = init_node_manager(sbi); | |
1992 | if (err) | |
1993 | return err; | |
1994 | ||
1995 | build_free_nids(sbi); | |
1996 | return 0; | |
1997 | } | |
1998 | ||
1999 | void destroy_node_manager(struct f2fs_sb_info *sbi) | |
2000 | { | |
2001 | struct f2fs_nm_info *nm_i = NM_I(sbi); | |
2002 | struct free_nid *i, *next_i; | |
2003 | struct nat_entry *natvec[NATVEC_SIZE]; | |
7aed0d45 | 2004 | struct nat_entry_set *setvec[SETVEC_SIZE]; |
e05df3b1 JK |
2005 | nid_t nid = 0; |
2006 | unsigned int found; | |
2007 | ||
2008 | if (!nm_i) | |
2009 | return; | |
2010 | ||
2011 | /* destroy free nid list */ | |
2012 | spin_lock(&nm_i->free_nid_list_lock); | |
2013 | list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) { | |
9850cf4a | 2014 | f2fs_bug_on(sbi, i->state == NID_ALLOC); |
8a7ed66a | 2015 | __del_from_free_nid_list(nm_i, i); |
e05df3b1 | 2016 | nm_i->fcnt--; |
cf0ee0f0 CY |
2017 | spin_unlock(&nm_i->free_nid_list_lock); |
2018 | kmem_cache_free(free_nid_slab, i); | |
2019 | spin_lock(&nm_i->free_nid_list_lock); | |
e05df3b1 | 2020 | } |
9850cf4a | 2021 | f2fs_bug_on(sbi, nm_i->fcnt); |
e05df3b1 JK |
2022 | spin_unlock(&nm_i->free_nid_list_lock); |
2023 | ||
2024 | /* destroy nat cache */ | |
8b26ef98 | 2025 | down_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2026 | while ((found = __gang_lookup_nat_cache(nm_i, |
2027 | nid, NATVEC_SIZE, natvec))) { | |
2028 | unsigned idx; | |
7aed0d45 | 2029 | |
b6ce391e GZ |
2030 | nid = nat_get_nid(natvec[found - 1]) + 1; |
2031 | for (idx = 0; idx < found; idx++) | |
2032 | __del_from_nat_cache(nm_i, natvec[idx]); | |
e05df3b1 | 2033 | } |
9850cf4a | 2034 | f2fs_bug_on(sbi, nm_i->nat_cnt); |
7aed0d45 JK |
2035 | |
2036 | /* destroy nat set cache */ | |
2037 | nid = 0; | |
2038 | while ((found = __gang_lookup_nat_set(nm_i, | |
2039 | nid, SETVEC_SIZE, setvec))) { | |
2040 | unsigned idx; | |
2041 | ||
2042 | nid = setvec[found - 1]->set + 1; | |
2043 | for (idx = 0; idx < found; idx++) { | |
2044 | /* entry_cnt is not zero, when cp_error was occurred */ | |
2045 | f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list)); | |
2046 | radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set); | |
2047 | kmem_cache_free(nat_entry_set_slab, setvec[idx]); | |
2048 | } | |
2049 | } | |
8b26ef98 | 2050 | up_write(&nm_i->nat_tree_lock); |
e05df3b1 JK |
2051 | |
2052 | kfree(nm_i->nat_bitmap); | |
2053 | sbi->nm_info = NULL; | |
2054 | kfree(nm_i); | |
2055 | } | |
2056 | ||
6e6093a8 | 2057 | int __init create_node_manager_caches(void) |
e05df3b1 JK |
2058 | { |
2059 | nat_entry_slab = f2fs_kmem_cache_create("nat_entry", | |
e8512d2e | 2060 | sizeof(struct nat_entry)); |
e05df3b1 | 2061 | if (!nat_entry_slab) |
aec71382 | 2062 | goto fail; |
e05df3b1 JK |
2063 | |
2064 | free_nid_slab = f2fs_kmem_cache_create("free_nid", | |
e8512d2e | 2065 | sizeof(struct free_nid)); |
aec71382 | 2066 | if (!free_nid_slab) |
ce3e6d25 | 2067 | goto destroy_nat_entry; |
aec71382 CY |
2068 | |
2069 | nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set", | |
2070 | sizeof(struct nat_entry_set)); | |
2071 | if (!nat_entry_set_slab) | |
ce3e6d25 | 2072 | goto destroy_free_nid; |
e05df3b1 | 2073 | return 0; |
aec71382 | 2074 | |
ce3e6d25 | 2075 | destroy_free_nid: |
aec71382 | 2076 | kmem_cache_destroy(free_nid_slab); |
ce3e6d25 | 2077 | destroy_nat_entry: |
aec71382 CY |
2078 | kmem_cache_destroy(nat_entry_slab); |
2079 | fail: | |
2080 | return -ENOMEM; | |
e05df3b1 JK |
2081 | } |
2082 | ||
2083 | void destroy_node_manager_caches(void) | |
2084 | { | |
aec71382 | 2085 | kmem_cache_destroy(nat_entry_set_slab); |
e05df3b1 JK |
2086 | kmem_cache_destroy(free_nid_slab); |
2087 | kmem_cache_destroy(nat_entry_slab); | |
2088 | } |